WSL2-Linux-Kernel/kernel/time/timer.c

2090 строки
60 KiB
C
Исходник Обычный вид История

time: Add SPDX license identifiers Update the time(r) core files files with the correct SPDX license identifier based on the license text in the file itself. The SPDX identifier is a legally binding shorthand, which can be used instead of the full boiler plate text. This work is based on a script and data from Philippe Ombredanne, Kate Stewart and myself. The data has been created with two independent license scanners and manual inspection. The following files do not contain any direct license information and have been omitted from the big initial SPDX changes: timeconst.bc: The .bc files were not touched time.c, timer.c, timekeeping.c: Licence was deduced from EXPORT_SYMBOL_GPL As those files do not contain direct license references they fall under the project license, i.e. GPL V2 only. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Kees Cook <keescook@chromium.org> Acked-by: Ingo Molnar <mingo@kernel.org> Acked-by: John Stultz <john.stultz@linaro.org> Acked-by: Corey Minyard <cminyard@mvista.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Kate Stewart <kstewart@linuxfoundation.org> Cc: Philippe Ombredanne <pombredanne@nexb.com> Cc: Russell King <rmk+kernel@armlinux.org.uk> Cc: Richard Cochran <richardcochran@gmail.com> Cc: Nicolas Pitre <nicolas.pitre@linaro.org> Cc: David Riley <davidriley@chromium.org> Cc: Colin Cross <ccross@android.com> Cc: Mark Brown <broonie@kernel.org> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Link: https://lkml.kernel.org/r/20181031182252.879109557@linutronix.de
2018-10-31 21:21:09 +03:00
// SPDX-License-Identifier: GPL-2.0
/*
* Kernel internal timers
*
* Copyright (C) 1991, 1992 Linus Torvalds
*
* 1997-01-28 Modified by Finn Arne Gangstad to make timers scale better.
*
* 1997-09-10 Updated NTP code according to technical memorandum Jan '96
* "A Kernel Model for Precision Timekeeping" by Dave Mills
* 1998-12-24 Fixed a xtime SMP race (we need the xtime_lock rw spinlock to
* serialize accesses to xtime/lost_ticks).
* Copyright (C) 1998 Andrea Arcangeli
* 1999-03-10 Improved NTP compatibility by Ulrich Windl
* 2002-05-31 Move sys_sysinfo here and make its locking sane, Robert Love
* 2000-10-05 Implemented scalable SMP per-CPU timer handling.
* Copyright (C) 2000, 2001, 2002 Ingo Molnar
* Designed by David S. Miller, Alexey Kuznetsov and Ingo Molnar
*/
#include <linux/kernel_stat.h>
#include <linux/export.h>
#include <linux/interrupt.h>
#include <linux/percpu.h>
#include <linux/init.h>
#include <linux/mm.h>
#include <linux/swap.h>
#include <linux/pid_namespace.h>
#include <linux/notifier.h>
#include <linux/thread_info.h>
#include <linux/time.h>
#include <linux/jiffies.h>
#include <linux/posix-timers.h>
#include <linux/cpu.h>
#include <linux/syscalls.h>
#include <linux/delay.h>
#include <linux/tick.h>
[PATCH] Add debugging feature /proc/timer_stat Add /proc/timer_stats support: debugging feature to profile timer expiration. Both the starting site, process/PID and the expiration function is captured. This allows the quick identification of timer event sources in a system. Sample output: # echo 1 > /proc/timer_stats # cat /proc/timer_stats Timer Stats Version: v0.1 Sample period: 4.010 s 24, 0 swapper hrtimer_stop_sched_tick (hrtimer_sched_tick) 11, 0 swapper sk_reset_timer (tcp_delack_timer) 6, 0 swapper hrtimer_stop_sched_tick (hrtimer_sched_tick) 2, 1 swapper queue_delayed_work_on (delayed_work_timer_fn) 17, 0 swapper hrtimer_restart_sched_tick (hrtimer_sched_tick) 2, 1 swapper queue_delayed_work_on (delayed_work_timer_fn) 4, 2050 pcscd do_nanosleep (hrtimer_wakeup) 5, 4179 sshd sk_reset_timer (tcp_write_timer) 4, 2248 yum-updatesd schedule_timeout (process_timeout) 18, 0 swapper hrtimer_restart_sched_tick (hrtimer_sched_tick) 3, 0 swapper sk_reset_timer (tcp_delack_timer) 1, 1 swapper neigh_table_init_no_netlink (neigh_periodic_timer) 2, 1 swapper e1000_up (e1000_watchdog) 1, 1 init schedule_timeout (process_timeout) 100 total events, 25.24 events/sec [ cleanups and hrtimers support from Thomas Gleixner <tglx@linutronix.de> ] [bunk@stusta.de: nr_entries can become static] Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: john stultz <johnstul@us.ibm.com> Cc: Roman Zippel <zippel@linux-m68k.org> Cc: Andi Kleen <ak@suse.de> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-02-16 12:28:13 +03:00
#include <linux/kallsyms.h>
#include <linux/irq_work.h>
#include <linux/sched/signal.h>
#include <linux/sched/sysctl.h>
#include <linux/sched/nohz.h>
#include <linux/sched/debug.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 11:04:11 +03:00
#include <linux/slab.h>
#include <linux/compat.h>
#include <linux/uaccess.h>
#include <asm/unistd.h>
#include <asm/div64.h>
#include <asm/timex.h>
#include <asm/io.h>
#include "tick-internal.h"
#define CREATE_TRACE_POINTS
#include <trace/events/timer.h>
__visible u64 jiffies_64 __cacheline_aligned_in_smp = INITIAL_JIFFIES;
EXPORT_SYMBOL(jiffies_64);
/*
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
* The timer wheel has LVL_DEPTH array levels. Each level provides an array of
* LVL_SIZE buckets. Each level is driven by its own clock and therefor each
* level has a different granularity.
*
* The level granularity is: LVL_CLK_DIV ^ lvl
* The level clock frequency is: HZ / (LVL_CLK_DIV ^ level)
*
* The array level of a newly armed timer depends on the relative expiry
* time. The farther the expiry time is away the higher the array level and
* therefor the granularity becomes.
*
* Contrary to the original timer wheel implementation, which aims for 'exact'
* expiry of the timers, this implementation removes the need for recascading
* the timers into the lower array levels. The previous 'classic' timer wheel
* implementation of the kernel already violated the 'exact' expiry by adding
* slack to the expiry time to provide batched expiration. The granularity
* levels provide implicit batching.
*
* This is an optimization of the original timer wheel implementation for the
* majority of the timer wheel use cases: timeouts. The vast majority of
* timeout timers (networking, disk I/O ...) are canceled before expiry. If
* the timeout expires it indicates that normal operation is disturbed, so it
* does not matter much whether the timeout comes with a slight delay.
*
* The only exception to this are networking timers with a small expiry
* time. They rely on the granularity. Those fit into the first wheel level,
* which has HZ granularity.
*
* We don't have cascading anymore. timers with a expiry time above the
* capacity of the last wheel level are force expired at the maximum timeout
* value of the last wheel level. From data sampling we know that the maximum
* value observed is 5 days (network connection tracking), so this should not
* be an issue.
*
* The currently chosen array constants values are a good compromise between
* array size and granularity.
*
* This results in the following granularity and range levels:
*
* HZ 1000 steps
* Level Offset Granularity Range
* 0 0 1 ms 0 ms - 63 ms
* 1 64 8 ms 64 ms - 511 ms
* 2 128 64 ms 512 ms - 4095 ms (512ms - ~4s)
* 3 192 512 ms 4096 ms - 32767 ms (~4s - ~32s)
* 4 256 4096 ms (~4s) 32768 ms - 262143 ms (~32s - ~4m)
* 5 320 32768 ms (~32s) 262144 ms - 2097151 ms (~4m - ~34m)
* 6 384 262144 ms (~4m) 2097152 ms - 16777215 ms (~34m - ~4h)
* 7 448 2097152 ms (~34m) 16777216 ms - 134217727 ms (~4h - ~1d)
* 8 512 16777216 ms (~4h) 134217728 ms - 1073741822 ms (~1d - ~12d)
*
* HZ 300
* Level Offset Granularity Range
* 0 0 3 ms 0 ms - 210 ms
* 1 64 26 ms 213 ms - 1703 ms (213ms - ~1s)
* 2 128 213 ms 1706 ms - 13650 ms (~1s - ~13s)
* 3 192 1706 ms (~1s) 13653 ms - 109223 ms (~13s - ~1m)
* 4 256 13653 ms (~13s) 109226 ms - 873810 ms (~1m - ~14m)
* 5 320 109226 ms (~1m) 873813 ms - 6990503 ms (~14m - ~1h)
* 6 384 873813 ms (~14m) 6990506 ms - 55924050 ms (~1h - ~15h)
* 7 448 6990506 ms (~1h) 55924053 ms - 447392423 ms (~15h - ~5d)
* 8 512 55924053 ms (~15h) 447392426 ms - 3579139406 ms (~5d - ~41d)
*
* HZ 250
* Level Offset Granularity Range
* 0 0 4 ms 0 ms - 255 ms
* 1 64 32 ms 256 ms - 2047 ms (256ms - ~2s)
* 2 128 256 ms 2048 ms - 16383 ms (~2s - ~16s)
* 3 192 2048 ms (~2s) 16384 ms - 131071 ms (~16s - ~2m)
* 4 256 16384 ms (~16s) 131072 ms - 1048575 ms (~2m - ~17m)
* 5 320 131072 ms (~2m) 1048576 ms - 8388607 ms (~17m - ~2h)
* 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h)
* 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d)
* 8 512 67108864 ms (~18h) 536870912 ms - 4294967288 ms (~6d - ~49d)
*
* HZ 100
* Level Offset Granularity Range
* 0 0 10 ms 0 ms - 630 ms
* 1 64 80 ms 640 ms - 5110 ms (640ms - ~5s)
* 2 128 640 ms 5120 ms - 40950 ms (~5s - ~40s)
* 3 192 5120 ms (~5s) 40960 ms - 327670 ms (~40s - ~5m)
* 4 256 40960 ms (~40s) 327680 ms - 2621430 ms (~5m - ~43m)
* 5 320 327680 ms (~5m) 2621440 ms - 20971510 ms (~43m - ~5h)
* 6 384 2621440 ms (~43m) 20971520 ms - 167772150 ms (~5h - ~1d)
* 7 448 20971520 ms (~5h) 167772160 ms - 1342177270 ms (~1d - ~15d)
*/
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
/* Clock divisor for the next level */
#define LVL_CLK_SHIFT 3
#define LVL_CLK_DIV (1UL << LVL_CLK_SHIFT)
#define LVL_CLK_MASK (LVL_CLK_DIV - 1)
#define LVL_SHIFT(n) ((n) * LVL_CLK_SHIFT)
#define LVL_GRAN(n) (1UL << LVL_SHIFT(n))
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
/*
* The time start value for each level to select the bucket at enqueue
* time.
*/
#define LVL_START(n) ((LVL_SIZE - 1) << (((n) - 1) * LVL_CLK_SHIFT))
/* Size of each clock level */
#define LVL_BITS 6
#define LVL_SIZE (1UL << LVL_BITS)
#define LVL_MASK (LVL_SIZE - 1)
#define LVL_OFFS(n) ((n) * LVL_SIZE)
/* Level depth */
#if HZ > 100
# define LVL_DEPTH 9
# else
# define LVL_DEPTH 8
#endif
/* The cutoff (max. capacity of the wheel) */
#define WHEEL_TIMEOUT_CUTOFF (LVL_START(LVL_DEPTH))
#define WHEEL_TIMEOUT_MAX (WHEEL_TIMEOUT_CUTOFF - LVL_GRAN(LVL_DEPTH - 1))
/*
* The resulting wheel size. If NOHZ is configured we allocate two
* wheels so we have a separate storage for the deferrable timers.
*/
#define WHEEL_SIZE (LVL_SIZE * LVL_DEPTH)
#ifdef CONFIG_NO_HZ_COMMON
# define NR_BASES 2
# define BASE_STD 0
# define BASE_DEF 1
#else
# define NR_BASES 1
# define BASE_STD 0
# define BASE_DEF 0
#endif
struct timer_base {
raw_spinlock_t lock;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
struct timer_list *running_timer;
timers: Prepare support for PREEMPT_RT When PREEMPT_RT is enabled, the soft interrupt thread can be preempted. If the soft interrupt thread is preempted in the middle of a timer callback, then calling del_timer_sync() can lead to two issues: - If the caller is on a remote CPU then it has to spin wait for the timer handler to complete. This can result in unbound priority inversion. - If the caller originates from the task which preempted the timer handler on the same CPU, then spin waiting for the timer handler to complete is never going to end. To avoid these issues, add a new lock to the timer base which is held around the execution of the timer callbacks. If del_timer_sync() detects that the timer callback is currently running, it blocks on the expiry lock. When the callback is finished, the expiry lock is dropped by the softirq thread which wakes up the waiter and the system makes progress. This addresses both the priority inversion and the life lock issues. This mechanism is not used for timers which are marked IRQSAFE as for those preemption is disabled accross the callback and therefore this situation cannot happen. The callbacks for such timers need to be individually audited for RT compliance. The same issue can happen in virtual machines when the vCPU which runs a timer callback is scheduled out. If a second vCPU of the same guest calls del_timer_sync() it will spin wait for the other vCPU to be scheduled back in. The expiry lock mechanism would avoid that. It'd be trivial to enable this when paravirt spinlocks are enabled in a guest, but it's not clear whether this is an actual problem in the wild, so for now it's an RT only mechanism. As the softirq thread can be preempted with PREEMPT_RT=y, the SMP variant of del_timer_sync() needs to be used on UP as well. [ tglx: Refactored it for mainline ] Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20190726185753.832418500@linutronix.de
2019-07-26 21:31:00 +03:00
#ifdef CONFIG_PREEMPT_RT
spinlock_t expiry_lock;
atomic_t timer_waiters;
#endif
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
unsigned long clk;
unsigned long next_expiry;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
unsigned int cpu;
bool is_idle;
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
bool must_forward_clk;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
DECLARE_BITMAP(pending_map, WHEEL_SIZE);
struct hlist_head vectors[WHEEL_SIZE];
Add support for deferrable timers Introduce a new flag for timers - deferrable: Timers that work normally when system is busy. But, will not cause CPU to come out of idle (just to service this timer), when CPU is idle. Instead, this timer will be serviced when CPU eventually wakes up with a subsequent non-deferrable timer. The main advantage of this is to avoid unnecessary timer interrupts when CPU is idle. If the routine currently called by a timer can wait until next event without any issues, this new timer can be used to setup timer event for that routine. This, with dynticks, allows CPUs to be lazy, allowing them to stay in idle for extended period of time by reducing unnecesary wakeup and thereby reducing the power consumption. This patch: Builds this new timer on top of existing timer infrastructure. It uses last bit in 'base' pointer of timer_list structure to store this deferrable timer flag. __next_timer_interrupt() function skips over these deferrable timers when CPU looks for next timer event for which it has to wake up. This is exported by a new interface init_timer_deferrable() that can be called in place of regular init_timer(). [akpm@linux-foundation.org: Privatise a #define] Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Oleg Nesterov <oleg@tv-sign.ru> Cc: Dave Jones <davej@codemonkey.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-08 11:27:44 +04:00
} ____cacheline_aligned;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
static DEFINE_PER_CPU(struct timer_base, timer_bases[NR_BASES]);
Add support for deferrable timers Introduce a new flag for timers - deferrable: Timers that work normally when system is busy. But, will not cause CPU to come out of idle (just to service this timer), when CPU is idle. Instead, this timer will be serviced when CPU eventually wakes up with a subsequent non-deferrable timer. The main advantage of this is to avoid unnecessary timer interrupts when CPU is idle. If the routine currently called by a timer can wait until next event without any issues, this new timer can be used to setup timer event for that routine. This, with dynticks, allows CPUs to be lazy, allowing them to stay in idle for extended period of time by reducing unnecesary wakeup and thereby reducing the power consumption. This patch: Builds this new timer on top of existing timer infrastructure. It uses last bit in 'base' pointer of timer_list structure to store this deferrable timer flag. __next_timer_interrupt() function skips over these deferrable timers when CPU looks for next timer event for which it has to wake up. This is exported by a new interface init_timer_deferrable() that can be called in place of regular init_timer(). [akpm@linux-foundation.org: Privatise a #define] Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Oleg Nesterov <oleg@tv-sign.ru> Cc: Dave Jones <davej@codemonkey.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-08 11:27:44 +04:00
#ifdef CONFIG_NO_HZ_COMMON
hrtimer: Unify remote enqueue handling hrtimer_reprogram() is conditionally invoked from hrtimer_start_range_ns() when hrtimer_cpu_base.hres_active is true. In the !hres_active case there is a special condition for the nohz_active case: If the newly enqueued timer expires before the first expiring timer on a remote CPU then the remote CPU needs to be notified and woken up from a NOHZ idle sleep to take the new first expiring timer into account. Previous changes have already established the prerequisites to make the remote enqueue behaviour the same whether high resolution mode is active or not: If the to be enqueued timer expires before the first expiring timer on a remote CPU, then it cannot be enqueued there. This was done for the high resolution mode because there is no way to access the remote CPU timer hardware. The same is true for NOHZ, but was handled differently by unconditionally enqueuing the timer and waking up the remote CPU so it can reprogram its timer. Again there is no compelling reason for this difference. hrtimer_check_target(), which makes the 'can remote enqueue' decision is already unconditional, but not yet functional because nothing updates hrtimer_cpu_base.expires_next in the !hres_active case. To unify this the following changes are required: 1) Make the store of the new first expiry time unconditonal in hrtimer_reprogram() and check __hrtimer_hres_active() before proceeding to the actual hardware access. This check also lets the compiler eliminate the rest of the function in case of CONFIG_HIGH_RES_TIMERS=n. 2) Invoke hrtimer_reprogram() unconditionally from hrtimer_start_range_ns() 3) Remove the remote wakeup special case for the !high_res && nohz_active case. Confine the timers_nohz_active static key to timer.c which is the only user now. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Cc: Christoph Hellwig <hch@lst.de> Cc: John Stultz <john.stultz@linaro.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: keescook@chromium.org Link: http://lkml.kernel.org/r/20171221104205.7269-21-anna-maria@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-12-21 13:41:49 +03:00
static DEFINE_STATIC_KEY_FALSE(timers_nohz_active);
static DEFINE_MUTEX(timer_keys_mutex);
static void timer_update_keys(struct work_struct *work);
static DECLARE_WORK(timer_update_work, timer_update_keys);
#ifdef CONFIG_SMP
timer: Reduce timer migration overhead if disabled Eric reported that the timer_migration sysctl is not really nice performance wise as it needs to check at every timer insertion whether the feature is enabled or not. Further the check does not live in the timer code, so we have an extra function call which checks an extra cache line to figure out that it is disabled. We can do better and store that information in the per cpu (hr)timer bases. I pondered to use a static key, but that's a nightmare to update from the nohz code and the timer base cache line is hot anyway when we select a timer base. The old logic enabled the timer migration unconditionally if CONFIG_NO_HZ was set even if nohz was disabled on the kernel command line. With this modification, we start off with migration disabled. The user visible sysctl is still set to enabled. If the kernel switches to NOHZ migration is enabled, if the user did not disable it via the sysctl prior to the switch. If nohz=off is on the kernel command line, migration stays disabled no matter what. Before: 47.76% hog [.] main 14.84% [kernel] [k] _raw_spin_lock_irqsave 9.55% [kernel] [k] _raw_spin_unlock_irqrestore 6.71% [kernel] [k] mod_timer 6.24% [kernel] [k] lock_timer_base.isra.38 3.76% [kernel] [k] detach_if_pending 3.71% [kernel] [k] del_timer 2.50% [kernel] [k] internal_add_timer 1.51% [kernel] [k] get_nohz_timer_target 1.28% [kernel] [k] __internal_add_timer 0.78% [kernel] [k] timerfn 0.48% [kernel] [k] wake_up_nohz_cpu After: 48.10% hog [.] main 15.25% [kernel] [k] _raw_spin_lock_irqsave 9.76% [kernel] [k] _raw_spin_unlock_irqrestore 6.50% [kernel] [k] mod_timer 6.44% [kernel] [k] lock_timer_base.isra.38 3.87% [kernel] [k] detach_if_pending 3.80% [kernel] [k] del_timer 2.67% [kernel] [k] internal_add_timer 1.33% [kernel] [k] __internal_add_timer 0.73% [kernel] [k] timerfn 0.54% [kernel] [k] wake_up_nohz_cpu Reported-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Link: http://lkml.kernel.org/r/20150526224512.127050787@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:33 +03:00
unsigned int sysctl_timer_migration = 1;
DEFINE_STATIC_KEY_FALSE(timers_migration_enabled);
static void timers_update_migration(void)
timer: Reduce timer migration overhead if disabled Eric reported that the timer_migration sysctl is not really nice performance wise as it needs to check at every timer insertion whether the feature is enabled or not. Further the check does not live in the timer code, so we have an extra function call which checks an extra cache line to figure out that it is disabled. We can do better and store that information in the per cpu (hr)timer bases. I pondered to use a static key, but that's a nightmare to update from the nohz code and the timer base cache line is hot anyway when we select a timer base. The old logic enabled the timer migration unconditionally if CONFIG_NO_HZ was set even if nohz was disabled on the kernel command line. With this modification, we start off with migration disabled. The user visible sysctl is still set to enabled. If the kernel switches to NOHZ migration is enabled, if the user did not disable it via the sysctl prior to the switch. If nohz=off is on the kernel command line, migration stays disabled no matter what. Before: 47.76% hog [.] main 14.84% [kernel] [k] _raw_spin_lock_irqsave 9.55% [kernel] [k] _raw_spin_unlock_irqrestore 6.71% [kernel] [k] mod_timer 6.24% [kernel] [k] lock_timer_base.isra.38 3.76% [kernel] [k] detach_if_pending 3.71% [kernel] [k] del_timer 2.50% [kernel] [k] internal_add_timer 1.51% [kernel] [k] get_nohz_timer_target 1.28% [kernel] [k] __internal_add_timer 0.78% [kernel] [k] timerfn 0.48% [kernel] [k] wake_up_nohz_cpu After: 48.10% hog [.] main 15.25% [kernel] [k] _raw_spin_lock_irqsave 9.76% [kernel] [k] _raw_spin_unlock_irqrestore 6.50% [kernel] [k] mod_timer 6.44% [kernel] [k] lock_timer_base.isra.38 3.87% [kernel] [k] detach_if_pending 3.80% [kernel] [k] del_timer 2.67% [kernel] [k] internal_add_timer 1.33% [kernel] [k] __internal_add_timer 0.73% [kernel] [k] timerfn 0.54% [kernel] [k] wake_up_nohz_cpu Reported-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Link: http://lkml.kernel.org/r/20150526224512.127050787@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:33 +03:00
{
if (sysctl_timer_migration && tick_nohz_active)
static_branch_enable(&timers_migration_enabled);
else
static_branch_disable(&timers_migration_enabled);
}
#else
static inline void timers_update_migration(void) { }
#endif /* !CONFIG_SMP */
timer: Reduce timer migration overhead if disabled Eric reported that the timer_migration sysctl is not really nice performance wise as it needs to check at every timer insertion whether the feature is enabled or not. Further the check does not live in the timer code, so we have an extra function call which checks an extra cache line to figure out that it is disabled. We can do better and store that information in the per cpu (hr)timer bases. I pondered to use a static key, but that's a nightmare to update from the nohz code and the timer base cache line is hot anyway when we select a timer base. The old logic enabled the timer migration unconditionally if CONFIG_NO_HZ was set even if nohz was disabled on the kernel command line. With this modification, we start off with migration disabled. The user visible sysctl is still set to enabled. If the kernel switches to NOHZ migration is enabled, if the user did not disable it via the sysctl prior to the switch. If nohz=off is on the kernel command line, migration stays disabled no matter what. Before: 47.76% hog [.] main 14.84% [kernel] [k] _raw_spin_lock_irqsave 9.55% [kernel] [k] _raw_spin_unlock_irqrestore 6.71% [kernel] [k] mod_timer 6.24% [kernel] [k] lock_timer_base.isra.38 3.76% [kernel] [k] detach_if_pending 3.71% [kernel] [k] del_timer 2.50% [kernel] [k] internal_add_timer 1.51% [kernel] [k] get_nohz_timer_target 1.28% [kernel] [k] __internal_add_timer 0.78% [kernel] [k] timerfn 0.48% [kernel] [k] wake_up_nohz_cpu After: 48.10% hog [.] main 15.25% [kernel] [k] _raw_spin_lock_irqsave 9.76% [kernel] [k] _raw_spin_unlock_irqrestore 6.50% [kernel] [k] mod_timer 6.44% [kernel] [k] lock_timer_base.isra.38 3.87% [kernel] [k] detach_if_pending 3.80% [kernel] [k] del_timer 2.67% [kernel] [k] internal_add_timer 1.33% [kernel] [k] __internal_add_timer 0.73% [kernel] [k] timerfn 0.54% [kernel] [k] wake_up_nohz_cpu Reported-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Link: http://lkml.kernel.org/r/20150526224512.127050787@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:33 +03:00
static void timer_update_keys(struct work_struct *work)
{
mutex_lock(&timer_keys_mutex);
timers_update_migration();
static_branch_enable(&timers_nohz_active);
mutex_unlock(&timer_keys_mutex);
}
timer: Reduce timer migration overhead if disabled Eric reported that the timer_migration sysctl is not really nice performance wise as it needs to check at every timer insertion whether the feature is enabled or not. Further the check does not live in the timer code, so we have an extra function call which checks an extra cache line to figure out that it is disabled. We can do better and store that information in the per cpu (hr)timer bases. I pondered to use a static key, but that's a nightmare to update from the nohz code and the timer base cache line is hot anyway when we select a timer base. The old logic enabled the timer migration unconditionally if CONFIG_NO_HZ was set even if nohz was disabled on the kernel command line. With this modification, we start off with migration disabled. The user visible sysctl is still set to enabled. If the kernel switches to NOHZ migration is enabled, if the user did not disable it via the sysctl prior to the switch. If nohz=off is on the kernel command line, migration stays disabled no matter what. Before: 47.76% hog [.] main 14.84% [kernel] [k] _raw_spin_lock_irqsave 9.55% [kernel] [k] _raw_spin_unlock_irqrestore 6.71% [kernel] [k] mod_timer 6.24% [kernel] [k] lock_timer_base.isra.38 3.76% [kernel] [k] detach_if_pending 3.71% [kernel] [k] del_timer 2.50% [kernel] [k] internal_add_timer 1.51% [kernel] [k] get_nohz_timer_target 1.28% [kernel] [k] __internal_add_timer 0.78% [kernel] [k] timerfn 0.48% [kernel] [k] wake_up_nohz_cpu After: 48.10% hog [.] main 15.25% [kernel] [k] _raw_spin_lock_irqsave 9.76% [kernel] [k] _raw_spin_unlock_irqrestore 6.50% [kernel] [k] mod_timer 6.44% [kernel] [k] lock_timer_base.isra.38 3.87% [kernel] [k] detach_if_pending 3.80% [kernel] [k] del_timer 2.67% [kernel] [k] internal_add_timer 1.33% [kernel] [k] __internal_add_timer 0.73% [kernel] [k] timerfn 0.54% [kernel] [k] wake_up_nohz_cpu Reported-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Link: http://lkml.kernel.org/r/20150526224512.127050787@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:33 +03:00
void timers_update_nohz(void)
{
schedule_work(&timer_update_work);
timer: Reduce timer migration overhead if disabled Eric reported that the timer_migration sysctl is not really nice performance wise as it needs to check at every timer insertion whether the feature is enabled or not. Further the check does not live in the timer code, so we have an extra function call which checks an extra cache line to figure out that it is disabled. We can do better and store that information in the per cpu (hr)timer bases. I pondered to use a static key, but that's a nightmare to update from the nohz code and the timer base cache line is hot anyway when we select a timer base. The old logic enabled the timer migration unconditionally if CONFIG_NO_HZ was set even if nohz was disabled on the kernel command line. With this modification, we start off with migration disabled. The user visible sysctl is still set to enabled. If the kernel switches to NOHZ migration is enabled, if the user did not disable it via the sysctl prior to the switch. If nohz=off is on the kernel command line, migration stays disabled no matter what. Before: 47.76% hog [.] main 14.84% [kernel] [k] _raw_spin_lock_irqsave 9.55% [kernel] [k] _raw_spin_unlock_irqrestore 6.71% [kernel] [k] mod_timer 6.24% [kernel] [k] lock_timer_base.isra.38 3.76% [kernel] [k] detach_if_pending 3.71% [kernel] [k] del_timer 2.50% [kernel] [k] internal_add_timer 1.51% [kernel] [k] get_nohz_timer_target 1.28% [kernel] [k] __internal_add_timer 0.78% [kernel] [k] timerfn 0.48% [kernel] [k] wake_up_nohz_cpu After: 48.10% hog [.] main 15.25% [kernel] [k] _raw_spin_lock_irqsave 9.76% [kernel] [k] _raw_spin_unlock_irqrestore 6.50% [kernel] [k] mod_timer 6.44% [kernel] [k] lock_timer_base.isra.38 3.87% [kernel] [k] detach_if_pending 3.80% [kernel] [k] del_timer 2.67% [kernel] [k] internal_add_timer 1.33% [kernel] [k] __internal_add_timer 0.73% [kernel] [k] timerfn 0.54% [kernel] [k] wake_up_nohz_cpu Reported-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Link: http://lkml.kernel.org/r/20150526224512.127050787@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:33 +03:00
}
int timer_migration_handler(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp,
loff_t *ppos)
{
int ret;
mutex_lock(&timer_keys_mutex);
ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
timer: Reduce timer migration overhead if disabled Eric reported that the timer_migration sysctl is not really nice performance wise as it needs to check at every timer insertion whether the feature is enabled or not. Further the check does not live in the timer code, so we have an extra function call which checks an extra cache line to figure out that it is disabled. We can do better and store that information in the per cpu (hr)timer bases. I pondered to use a static key, but that's a nightmare to update from the nohz code and the timer base cache line is hot anyway when we select a timer base. The old logic enabled the timer migration unconditionally if CONFIG_NO_HZ was set even if nohz was disabled on the kernel command line. With this modification, we start off with migration disabled. The user visible sysctl is still set to enabled. If the kernel switches to NOHZ migration is enabled, if the user did not disable it via the sysctl prior to the switch. If nohz=off is on the kernel command line, migration stays disabled no matter what. Before: 47.76% hog [.] main 14.84% [kernel] [k] _raw_spin_lock_irqsave 9.55% [kernel] [k] _raw_spin_unlock_irqrestore 6.71% [kernel] [k] mod_timer 6.24% [kernel] [k] lock_timer_base.isra.38 3.76% [kernel] [k] detach_if_pending 3.71% [kernel] [k] del_timer 2.50% [kernel] [k] internal_add_timer 1.51% [kernel] [k] get_nohz_timer_target 1.28% [kernel] [k] __internal_add_timer 0.78% [kernel] [k] timerfn 0.48% [kernel] [k] wake_up_nohz_cpu After: 48.10% hog [.] main 15.25% [kernel] [k] _raw_spin_lock_irqsave 9.76% [kernel] [k] _raw_spin_unlock_irqrestore 6.50% [kernel] [k] mod_timer 6.44% [kernel] [k] lock_timer_base.isra.38 3.87% [kernel] [k] detach_if_pending 3.80% [kernel] [k] del_timer 2.67% [kernel] [k] internal_add_timer 1.33% [kernel] [k] __internal_add_timer 0.73% [kernel] [k] timerfn 0.54% [kernel] [k] wake_up_nohz_cpu Reported-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Link: http://lkml.kernel.org/r/20150526224512.127050787@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:33 +03:00
if (!ret && write)
timers_update_migration();
mutex_unlock(&timer_keys_mutex);
timer: Reduce timer migration overhead if disabled Eric reported that the timer_migration sysctl is not really nice performance wise as it needs to check at every timer insertion whether the feature is enabled or not. Further the check does not live in the timer code, so we have an extra function call which checks an extra cache line to figure out that it is disabled. We can do better and store that information in the per cpu (hr)timer bases. I pondered to use a static key, but that's a nightmare to update from the nohz code and the timer base cache line is hot anyway when we select a timer base. The old logic enabled the timer migration unconditionally if CONFIG_NO_HZ was set even if nohz was disabled on the kernel command line. With this modification, we start off with migration disabled. The user visible sysctl is still set to enabled. If the kernel switches to NOHZ migration is enabled, if the user did not disable it via the sysctl prior to the switch. If nohz=off is on the kernel command line, migration stays disabled no matter what. Before: 47.76% hog [.] main 14.84% [kernel] [k] _raw_spin_lock_irqsave 9.55% [kernel] [k] _raw_spin_unlock_irqrestore 6.71% [kernel] [k] mod_timer 6.24% [kernel] [k] lock_timer_base.isra.38 3.76% [kernel] [k] detach_if_pending 3.71% [kernel] [k] del_timer 2.50% [kernel] [k] internal_add_timer 1.51% [kernel] [k] get_nohz_timer_target 1.28% [kernel] [k] __internal_add_timer 0.78% [kernel] [k] timerfn 0.48% [kernel] [k] wake_up_nohz_cpu After: 48.10% hog [.] main 15.25% [kernel] [k] _raw_spin_lock_irqsave 9.76% [kernel] [k] _raw_spin_unlock_irqrestore 6.50% [kernel] [k] mod_timer 6.44% [kernel] [k] lock_timer_base.isra.38 3.87% [kernel] [k] detach_if_pending 3.80% [kernel] [k] del_timer 2.67% [kernel] [k] internal_add_timer 1.33% [kernel] [k] __internal_add_timer 0.73% [kernel] [k] timerfn 0.54% [kernel] [k] wake_up_nohz_cpu Reported-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Link: http://lkml.kernel.org/r/20150526224512.127050787@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:33 +03:00
return ret;
}
hrtimer: Unify remote enqueue handling hrtimer_reprogram() is conditionally invoked from hrtimer_start_range_ns() when hrtimer_cpu_base.hres_active is true. In the !hres_active case there is a special condition for the nohz_active case: If the newly enqueued timer expires before the first expiring timer on a remote CPU then the remote CPU needs to be notified and woken up from a NOHZ idle sleep to take the new first expiring timer into account. Previous changes have already established the prerequisites to make the remote enqueue behaviour the same whether high resolution mode is active or not: If the to be enqueued timer expires before the first expiring timer on a remote CPU, then it cannot be enqueued there. This was done for the high resolution mode because there is no way to access the remote CPU timer hardware. The same is true for NOHZ, but was handled differently by unconditionally enqueuing the timer and waking up the remote CPU so it can reprogram its timer. Again there is no compelling reason for this difference. hrtimer_check_target(), which makes the 'can remote enqueue' decision is already unconditional, but not yet functional because nothing updates hrtimer_cpu_base.expires_next in the !hres_active case. To unify this the following changes are required: 1) Make the store of the new first expiry time unconditonal in hrtimer_reprogram() and check __hrtimer_hres_active() before proceeding to the actual hardware access. This check also lets the compiler eliminate the rest of the function in case of CONFIG_HIGH_RES_TIMERS=n. 2) Invoke hrtimer_reprogram() unconditionally from hrtimer_start_range_ns() 3) Remove the remote wakeup special case for the !high_res && nohz_active case. Confine the timers_nohz_active static key to timer.c which is the only user now. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Cc: Christoph Hellwig <hch@lst.de> Cc: John Stultz <john.stultz@linaro.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: keescook@chromium.org Link: http://lkml.kernel.org/r/20171221104205.7269-21-anna-maria@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-12-21 13:41:49 +03:00
static inline bool is_timers_nohz_active(void)
{
return static_branch_unlikely(&timers_nohz_active);
}
#else
static inline bool is_timers_nohz_active(void) { return false; }
#endif /* NO_HZ_COMMON */
timer: Reduce timer migration overhead if disabled Eric reported that the timer_migration sysctl is not really nice performance wise as it needs to check at every timer insertion whether the feature is enabled or not. Further the check does not live in the timer code, so we have an extra function call which checks an extra cache line to figure out that it is disabled. We can do better and store that information in the per cpu (hr)timer bases. I pondered to use a static key, but that's a nightmare to update from the nohz code and the timer base cache line is hot anyway when we select a timer base. The old logic enabled the timer migration unconditionally if CONFIG_NO_HZ was set even if nohz was disabled on the kernel command line. With this modification, we start off with migration disabled. The user visible sysctl is still set to enabled. If the kernel switches to NOHZ migration is enabled, if the user did not disable it via the sysctl prior to the switch. If nohz=off is on the kernel command line, migration stays disabled no matter what. Before: 47.76% hog [.] main 14.84% [kernel] [k] _raw_spin_lock_irqsave 9.55% [kernel] [k] _raw_spin_unlock_irqrestore 6.71% [kernel] [k] mod_timer 6.24% [kernel] [k] lock_timer_base.isra.38 3.76% [kernel] [k] detach_if_pending 3.71% [kernel] [k] del_timer 2.50% [kernel] [k] internal_add_timer 1.51% [kernel] [k] get_nohz_timer_target 1.28% [kernel] [k] __internal_add_timer 0.78% [kernel] [k] timerfn 0.48% [kernel] [k] wake_up_nohz_cpu After: 48.10% hog [.] main 15.25% [kernel] [k] _raw_spin_lock_irqsave 9.76% [kernel] [k] _raw_spin_unlock_irqrestore 6.50% [kernel] [k] mod_timer 6.44% [kernel] [k] lock_timer_base.isra.38 3.87% [kernel] [k] detach_if_pending 3.80% [kernel] [k] del_timer 2.67% [kernel] [k] internal_add_timer 1.33% [kernel] [k] __internal_add_timer 0.73% [kernel] [k] timerfn 0.54% [kernel] [k] wake_up_nohz_cpu Reported-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Link: http://lkml.kernel.org/r/20150526224512.127050787@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:33 +03:00
static unsigned long round_jiffies_common(unsigned long j, int cpu,
bool force_up)
{
int rem;
unsigned long original = j;
/*
* We don't want all cpus firing their timers at once hitting the
* same lock or cachelines, so we skew each extra cpu with an extra
* 3 jiffies. This 3 jiffies came originally from the mm/ code which
* already did this.
* The skew is done by adding 3*cpunr, then round, then subtract this
* extra offset again.
*/
j += cpu * 3;
rem = j % HZ;
/*
* If the target jiffie is just after a whole second (which can happen
* due to delays of the timer irq, long irq off times etc etc) then
* we should round down to the whole second, not up. Use 1/4th second
* as cutoff for this rounding as an extreme upper bound for this.
* But never round down if @force_up is set.
*/
if (rem < HZ/4 && !force_up) /* round down */
j = j - rem;
else /* round up */
j = j - rem + HZ;
/* now that we have rounded, subtract the extra skew again */
j -= cpu * 3;
/*
* Make sure j is still in the future. Otherwise return the
* unmodified value.
*/
return time_is_after_jiffies(j) ? j : original;
}
/**
* __round_jiffies - function to round jiffies to a full second
* @j: the time in (absolute) jiffies that should be rounded
* @cpu: the processor number on which the timeout will happen
*
* __round_jiffies() rounds an absolute time in the future (in jiffies)
* up or down to (approximately) full seconds. This is useful for timers
* for which the exact time they fire does not matter too much, as long as
* they fire approximately every X seconds.
*
* By rounding these timers to whole seconds, all such timers will fire
* at the same time, rather than at various times spread out. The goal
* of this is to have the CPU wake up less, which saves power.
*
* The exact rounding is skewed for each processor to avoid all
* processors firing at the exact same time, which could lead
* to lock contention or spurious cache line bouncing.
*
* The return value is the rounded version of the @j parameter.
*/
unsigned long __round_jiffies(unsigned long j, int cpu)
{
return round_jiffies_common(j, cpu, false);
}
EXPORT_SYMBOL_GPL(__round_jiffies);
/**
* __round_jiffies_relative - function to round jiffies to a full second
* @j: the time in (relative) jiffies that should be rounded
* @cpu: the processor number on which the timeout will happen
*
* __round_jiffies_relative() rounds a time delta in the future (in jiffies)
* up or down to (approximately) full seconds. This is useful for timers
* for which the exact time they fire does not matter too much, as long as
* they fire approximately every X seconds.
*
* By rounding these timers to whole seconds, all such timers will fire
* at the same time, rather than at various times spread out. The goal
* of this is to have the CPU wake up less, which saves power.
*
* The exact rounding is skewed for each processor to avoid all
* processors firing at the exact same time, which could lead
* to lock contention or spurious cache line bouncing.
*
* The return value is the rounded version of the @j parameter.
*/
unsigned long __round_jiffies_relative(unsigned long j, int cpu)
{
unsigned long j0 = jiffies;
/* Use j0 because jiffies might change while we run */
return round_jiffies_common(j + j0, cpu, false) - j0;
}
EXPORT_SYMBOL_GPL(__round_jiffies_relative);
/**
* round_jiffies - function to round jiffies to a full second
* @j: the time in (absolute) jiffies that should be rounded
*
* round_jiffies() rounds an absolute time in the future (in jiffies)
* up or down to (approximately) full seconds. This is useful for timers
* for which the exact time they fire does not matter too much, as long as
* they fire approximately every X seconds.
*
* By rounding these timers to whole seconds, all such timers will fire
* at the same time, rather than at various times spread out. The goal
* of this is to have the CPU wake up less, which saves power.
*
* The return value is the rounded version of the @j parameter.
*/
unsigned long round_jiffies(unsigned long j)
{
return round_jiffies_common(j, raw_smp_processor_id(), false);
}
EXPORT_SYMBOL_GPL(round_jiffies);
/**
* round_jiffies_relative - function to round jiffies to a full second
* @j: the time in (relative) jiffies that should be rounded
*
* round_jiffies_relative() rounds a time delta in the future (in jiffies)
* up or down to (approximately) full seconds. This is useful for timers
* for which the exact time they fire does not matter too much, as long as
* they fire approximately every X seconds.
*
* By rounding these timers to whole seconds, all such timers will fire
* at the same time, rather than at various times spread out. The goal
* of this is to have the CPU wake up less, which saves power.
*
* The return value is the rounded version of the @j parameter.
*/
unsigned long round_jiffies_relative(unsigned long j)
{
return __round_jiffies_relative(j, raw_smp_processor_id());
}
EXPORT_SYMBOL_GPL(round_jiffies_relative);
/**
* __round_jiffies_up - function to round jiffies up to a full second
* @j: the time in (absolute) jiffies that should be rounded
* @cpu: the processor number on which the timeout will happen
*
* This is the same as __round_jiffies() except that it will never
* round down. This is useful for timeouts for which the exact time
* of firing does not matter too much, as long as they don't fire too
* early.
*/
unsigned long __round_jiffies_up(unsigned long j, int cpu)
{
return round_jiffies_common(j, cpu, true);
}
EXPORT_SYMBOL_GPL(__round_jiffies_up);
/**
* __round_jiffies_up_relative - function to round jiffies up to a full second
* @j: the time in (relative) jiffies that should be rounded
* @cpu: the processor number on which the timeout will happen
*
* This is the same as __round_jiffies_relative() except that it will never
* round down. This is useful for timeouts for which the exact time
* of firing does not matter too much, as long as they don't fire too
* early.
*/
unsigned long __round_jiffies_up_relative(unsigned long j, int cpu)
{
unsigned long j0 = jiffies;
/* Use j0 because jiffies might change while we run */
return round_jiffies_common(j + j0, cpu, true) - j0;
}
EXPORT_SYMBOL_GPL(__round_jiffies_up_relative);
/**
* round_jiffies_up - function to round jiffies up to a full second
* @j: the time in (absolute) jiffies that should be rounded
*
* This is the same as round_jiffies() except that it will never
* round down. This is useful for timeouts for which the exact time
* of firing does not matter too much, as long as they don't fire too
* early.
*/
unsigned long round_jiffies_up(unsigned long j)
{
return round_jiffies_common(j, raw_smp_processor_id(), true);
}
EXPORT_SYMBOL_GPL(round_jiffies_up);
/**
* round_jiffies_up_relative - function to round jiffies up to a full second
* @j: the time in (relative) jiffies that should be rounded
*
* This is the same as round_jiffies_relative() except that it will never
* round down. This is useful for timeouts for which the exact time
* of firing does not matter too much, as long as they don't fire too
* early.
*/
unsigned long round_jiffies_up_relative(unsigned long j)
{
return __round_jiffies_up_relative(j, raw_smp_processor_id());
}
EXPORT_SYMBOL_GPL(round_jiffies_up_relative);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
static inline unsigned int timer_get_idx(struct timer_list *timer)
{
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
return (timer->flags & TIMER_ARRAYMASK) >> TIMER_ARRAYSHIFT;
}
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
static inline void timer_set_idx(struct timer_list *timer, unsigned int idx)
{
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
timer->flags = (timer->flags & ~TIMER_ARRAYMASK) |
idx << TIMER_ARRAYSHIFT;
}
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
/*
* Helper function to calculate the array index for a given expiry
* time.
*/
static inline unsigned calc_index(unsigned expires, unsigned lvl)
{
expires = (expires + LVL_GRAN(lvl)) >> LVL_SHIFT(lvl);
return LVL_OFFS(lvl) + (expires & LVL_MASK);
}
static int calc_wheel_index(unsigned long expires, unsigned long clk)
{
unsigned long delta = expires - clk;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
unsigned int idx;
if (delta < LVL_START(1)) {
idx = calc_index(expires, 0);
} else if (delta < LVL_START(2)) {
idx = calc_index(expires, 1);
} else if (delta < LVL_START(3)) {
idx = calc_index(expires, 2);
} else if (delta < LVL_START(4)) {
idx = calc_index(expires, 3);
} else if (delta < LVL_START(5)) {
idx = calc_index(expires, 4);
} else if (delta < LVL_START(6)) {
idx = calc_index(expires, 5);
} else if (delta < LVL_START(7)) {
idx = calc_index(expires, 6);
} else if (LVL_DEPTH > 8 && delta < LVL_START(8)) {
idx = calc_index(expires, 7);
} else if ((long) delta < 0) {
idx = clk & LVL_MASK;
} else {
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
/*
* Force expire obscene large timeouts to expire at the
* capacity limit of the wheel.
*/
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
if (expires >= WHEEL_TIMEOUT_CUTOFF)
expires = WHEEL_TIMEOUT_MAX;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
idx = calc_index(expires, LVL_DEPTH - 1);
}
return idx;
}
/*
* Enqueue the timer into the hash bucket, mark it pending in
* the bitmap and store the index in the timer flags.
*/
static void enqueue_timer(struct timer_base *base, struct timer_list *timer,
unsigned int idx)
{
hlist_add_head(&timer->entry, base->vectors + idx);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
__set_bit(idx, base->pending_map);
timer_set_idx(timer, idx);
trace_timer_start(timer, timer->expires, timer->flags);
}
static void
__internal_add_timer(struct timer_base *base, struct timer_list *timer)
{
unsigned int idx;
idx = calc_wheel_index(timer->expires, base->clk);
enqueue_timer(base, timer, idx);
}
timer: Kick dynticks targets on mod_timer*() calls When a timer is enqueued or modified on a dynticks target, that CPU must re-evaluate the next tick to service that timer. The tick re-evaluation is performed by an IPI kick on the target. Now while we correctly call wake_up_nohz_cpu() from add_timer_on(), the mod_timer*() API family doesn't support so well dynticks targets. The reason for this is likely that __mod_timer() isn't supposed to select an idle target for a timer, unless that target is the current CPU, in which case a dynticks idle kick isn't actually needed. But there is a small race window lurking behind that assumption: the elected target has all the time to turn dynticks idle between the call to get_nohz_timer_target() and the locking of its base. Hence a risk that we enqueue a timer on a dynticks idle destination without kicking it. As a result, the timer might be serviced too late in the future. Also a target elected by __mod_timer() can be in full dynticks mode and thus require to be kicked as well. And unlike idle dynticks, this concern both local and remote targets. To fix this whole issue, lets centralize the dynticks kick to internal_add_timer() so that it is well handled for all sort of timer enqueue. Even timer migration is concerned so that a full dynticks target is correctly kicked as needed when timers are migrating to it. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Link: http://lkml.kernel.org/r/1403393357-2070-3-git-send-email-fweisbec@gmail.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2014-06-22 03:29:14 +04:00
static void
trigger_dyntick_cpu(struct timer_base *base, struct timer_list *timer)
{
if (!is_timers_nohz_active())
return;
/*
* TODO: This wants some optimizing similar to the code below, but we
* will do that when we switch from push to pull for deferrable timers.
*/
if (timer->flags & TIMER_DEFERRABLE) {
if (tick_nohz_full_cpu(base->cpu))
timer: Minimize nohz off overhead If nohz is disabled on the kernel command line the [hr]timer code still calls wake_up_nohz_cpu() and tick_nohz_full_cpu(), a pretty pointless exercise. Cache nohz_active in [hr]timer per cpu bases and avoid the overhead. Before: 48.10% hog [.] main 15.25% [kernel] [k] _raw_spin_lock_irqsave 9.76% [kernel] [k] _raw_spin_unlock_irqrestore 6.50% [kernel] [k] mod_timer 6.44% [kernel] [k] lock_timer_base.isra.38 3.87% [kernel] [k] detach_if_pending 3.80% [kernel] [k] del_timer 2.67% [kernel] [k] internal_add_timer 1.33% [kernel] [k] __internal_add_timer 0.73% [kernel] [k] timerfn 0.54% [kernel] [k] wake_up_nohz_cpu After: 48.73% hog [.] main 15.36% [kernel] [k] _raw_spin_lock_irqsave 9.77% [kernel] [k] _raw_spin_unlock_irqrestore 6.61% [kernel] [k] lock_timer_base.isra.38 6.42% [kernel] [k] mod_timer 3.90% [kernel] [k] detach_if_pending 3.76% [kernel] [k] del_timer 2.41% [kernel] [k] internal_add_timer 1.39% [kernel] [k] __internal_add_timer 0.76% [kernel] [k] timerfn We probably should have a cached value for nohz full in the per cpu bases as well to avoid the cpumask check. The base cache line is hot already, the cpumask not necessarily. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Link: http://lkml.kernel.org/r/20150526224512.207378134@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:35 +03:00
wake_up_nohz_cpu(base->cpu);
return;
}
timer: Kick dynticks targets on mod_timer*() calls When a timer is enqueued or modified on a dynticks target, that CPU must re-evaluate the next tick to service that timer. The tick re-evaluation is performed by an IPI kick on the target. Now while we correctly call wake_up_nohz_cpu() from add_timer_on(), the mod_timer*() API family doesn't support so well dynticks targets. The reason for this is likely that __mod_timer() isn't supposed to select an idle target for a timer, unless that target is the current CPU, in which case a dynticks idle kick isn't actually needed. But there is a small race window lurking behind that assumption: the elected target has all the time to turn dynticks idle between the call to get_nohz_timer_target() and the locking of its base. Hence a risk that we enqueue a timer on a dynticks idle destination without kicking it. As a result, the timer might be serviced too late in the future. Also a target elected by __mod_timer() can be in full dynticks mode and thus require to be kicked as well. And unlike idle dynticks, this concern both local and remote targets. To fix this whole issue, lets centralize the dynticks kick to internal_add_timer() so that it is well handled for all sort of timer enqueue. Even timer migration is concerned so that a full dynticks target is correctly kicked as needed when timers are migrating to it. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Link: http://lkml.kernel.org/r/1403393357-2070-3-git-send-email-fweisbec@gmail.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2014-06-22 03:29:14 +04:00
/*
* We might have to IPI the remote CPU if the base is idle and the
* timer is not deferrable. If the other CPU is on the way to idle
* then it can't set base->is_idle as we hold the base lock:
timer: Kick dynticks targets on mod_timer*() calls When a timer is enqueued or modified on a dynticks target, that CPU must re-evaluate the next tick to service that timer. The tick re-evaluation is performed by an IPI kick on the target. Now while we correctly call wake_up_nohz_cpu() from add_timer_on(), the mod_timer*() API family doesn't support so well dynticks targets. The reason for this is likely that __mod_timer() isn't supposed to select an idle target for a timer, unless that target is the current CPU, in which case a dynticks idle kick isn't actually needed. But there is a small race window lurking behind that assumption: the elected target has all the time to turn dynticks idle between the call to get_nohz_timer_target() and the locking of its base. Hence a risk that we enqueue a timer on a dynticks idle destination without kicking it. As a result, the timer might be serviced too late in the future. Also a target elected by __mod_timer() can be in full dynticks mode and thus require to be kicked as well. And unlike idle dynticks, this concern both local and remote targets. To fix this whole issue, lets centralize the dynticks kick to internal_add_timer() so that it is well handled for all sort of timer enqueue. Even timer migration is concerned so that a full dynticks target is correctly kicked as needed when timers are migrating to it. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Link: http://lkml.kernel.org/r/1403393357-2070-3-git-send-email-fweisbec@gmail.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2014-06-22 03:29:14 +04:00
*/
if (!base->is_idle)
return;
/* Check whether this is the new first expiring timer: */
if (time_after_eq(timer->expires, base->next_expiry))
return;
/*
* Set the next expiry time and kick the CPU so it can reevaluate the
* wheel:
*/
base->next_expiry = timer->expires;
wake_up_nohz_cpu(base->cpu);
}
static void
internal_add_timer(struct timer_base *base, struct timer_list *timer)
{
__internal_add_timer(base, timer);
trigger_dyntick_cpu(base, timer);
}
#ifdef CONFIG_DEBUG_OBJECTS_TIMERS
static struct debug_obj_descr timer_debug_descr;
static void *timer_debug_hint(void *addr)
{
return ((struct timer_list *) addr)->function;
}
debugobjects: insulate non-fixup logic related to static obj from fixup callbacks When activating a static object we need make sure that the object is tracked in the object tracker. If it is a non-static object then the activation is illegal. In previous implementation, each subsystem need take care of this in their fixup callbacks. Actually we can put it into debugobjects core. Thus we can save duplicated code, and have *pure* fixup callbacks. To achieve this, a new callback "is_static_object" is introduced to let the type specific code decide whether a object is static or not. If yes, we take it into object tracker, otherwise give warning and invoke fixup callback. This change has paassed debugobjects selftest, and I also do some test with all debugobjects supports enabled. At last, I have a concern about the fixups that can it change the object which is in incorrect state on fixup? Because the 'addr' may not point to any valid object if a non-static object is not tracked. Then Change such object can overwrite someone's memory and cause unexpected behaviour. For example, the timer_fixup_activate bind timer to function stub_timer. Link: http://lkml.kernel.org/r/1462576157-14539-1-git-send-email-changbin.du@intel.com [changbin.du@intel.com: improve code comments where invoke the new is_static_object callback] Link: http://lkml.kernel.org/r/1462777431-8171-1-git-send-email-changbin.du@intel.com Signed-off-by: Du, Changbin <changbin.du@intel.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Josh Triplett <josh@kernel.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tejun Heo <tj@kernel.org> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-05-20 03:09:41 +03:00
static bool timer_is_static_object(void *addr)
{
struct timer_list *timer = addr;
return (timer->entry.pprev == NULL &&
timer->entry.next == TIMER_ENTRY_STATIC);
}
/*
* fixup_init is called when:
* - an active object is initialized
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
*/
static bool timer_fixup_init(void *addr, enum debug_obj_state state)
{
struct timer_list *timer = addr;
switch (state) {
case ODEBUG_STATE_ACTIVE:
del_timer_sync(timer);
debug_object_init(timer, &timer_debug_descr);
return true;
default:
return false;
}
}
/* Stub timer callback for improperly used timers. */
static void stub_timer(struct timer_list *unused)
{
WARN_ON(1);
}
/*
* fixup_activate is called when:
* - an active object is activated
debugobjects: insulate non-fixup logic related to static obj from fixup callbacks When activating a static object we need make sure that the object is tracked in the object tracker. If it is a non-static object then the activation is illegal. In previous implementation, each subsystem need take care of this in their fixup callbacks. Actually we can put it into debugobjects core. Thus we can save duplicated code, and have *pure* fixup callbacks. To achieve this, a new callback "is_static_object" is introduced to let the type specific code decide whether a object is static or not. If yes, we take it into object tracker, otherwise give warning and invoke fixup callback. This change has paassed debugobjects selftest, and I also do some test with all debugobjects supports enabled. At last, I have a concern about the fixups that can it change the object which is in incorrect state on fixup? Because the 'addr' may not point to any valid object if a non-static object is not tracked. Then Change such object can overwrite someone's memory and cause unexpected behaviour. For example, the timer_fixup_activate bind timer to function stub_timer. Link: http://lkml.kernel.org/r/1462576157-14539-1-git-send-email-changbin.du@intel.com [changbin.du@intel.com: improve code comments where invoke the new is_static_object callback] Link: http://lkml.kernel.org/r/1462777431-8171-1-git-send-email-changbin.du@intel.com Signed-off-by: Du, Changbin <changbin.du@intel.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Josh Triplett <josh@kernel.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tejun Heo <tj@kernel.org> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-05-20 03:09:41 +03:00
* - an unknown non-static object is activated
*/
static bool timer_fixup_activate(void *addr, enum debug_obj_state state)
{
struct timer_list *timer = addr;
switch (state) {
case ODEBUG_STATE_NOTAVAILABLE:
timer_setup(timer, stub_timer, 0);
debugobjects: insulate non-fixup logic related to static obj from fixup callbacks When activating a static object we need make sure that the object is tracked in the object tracker. If it is a non-static object then the activation is illegal. In previous implementation, each subsystem need take care of this in their fixup callbacks. Actually we can put it into debugobjects core. Thus we can save duplicated code, and have *pure* fixup callbacks. To achieve this, a new callback "is_static_object" is introduced to let the type specific code decide whether a object is static or not. If yes, we take it into object tracker, otherwise give warning and invoke fixup callback. This change has paassed debugobjects selftest, and I also do some test with all debugobjects supports enabled. At last, I have a concern about the fixups that can it change the object which is in incorrect state on fixup? Because the 'addr' may not point to any valid object if a non-static object is not tracked. Then Change such object can overwrite someone's memory and cause unexpected behaviour. For example, the timer_fixup_activate bind timer to function stub_timer. Link: http://lkml.kernel.org/r/1462576157-14539-1-git-send-email-changbin.du@intel.com [changbin.du@intel.com: improve code comments where invoke the new is_static_object callback] Link: http://lkml.kernel.org/r/1462777431-8171-1-git-send-email-changbin.du@intel.com Signed-off-by: Du, Changbin <changbin.du@intel.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Josh Triplett <josh@kernel.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tejun Heo <tj@kernel.org> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-05-20 03:09:41 +03:00
return true;
case ODEBUG_STATE_ACTIVE:
WARN_ON(1);
/* fall through */
default:
return false;
}
}
/*
* fixup_free is called when:
* - an active object is freed
*/
static bool timer_fixup_free(void *addr, enum debug_obj_state state)
{
struct timer_list *timer = addr;
switch (state) {
case ODEBUG_STATE_ACTIVE:
del_timer_sync(timer);
debug_object_free(timer, &timer_debug_descr);
return true;
default:
return false;
}
}
/*
* fixup_assert_init is called when:
* - an untracked/uninit-ed object is found
*/
static bool timer_fixup_assert_init(void *addr, enum debug_obj_state state)
{
struct timer_list *timer = addr;
switch (state) {
case ODEBUG_STATE_NOTAVAILABLE:
timer_setup(timer, stub_timer, 0);
debugobjects: insulate non-fixup logic related to static obj from fixup callbacks When activating a static object we need make sure that the object is tracked in the object tracker. If it is a non-static object then the activation is illegal. In previous implementation, each subsystem need take care of this in their fixup callbacks. Actually we can put it into debugobjects core. Thus we can save duplicated code, and have *pure* fixup callbacks. To achieve this, a new callback "is_static_object" is introduced to let the type specific code decide whether a object is static or not. If yes, we take it into object tracker, otherwise give warning and invoke fixup callback. This change has paassed debugobjects selftest, and I also do some test with all debugobjects supports enabled. At last, I have a concern about the fixups that can it change the object which is in incorrect state on fixup? Because the 'addr' may not point to any valid object if a non-static object is not tracked. Then Change such object can overwrite someone's memory and cause unexpected behaviour. For example, the timer_fixup_activate bind timer to function stub_timer. Link: http://lkml.kernel.org/r/1462576157-14539-1-git-send-email-changbin.du@intel.com [changbin.du@intel.com: improve code comments where invoke the new is_static_object callback] Link: http://lkml.kernel.org/r/1462777431-8171-1-git-send-email-changbin.du@intel.com Signed-off-by: Du, Changbin <changbin.du@intel.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Josh Triplett <josh@kernel.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tejun Heo <tj@kernel.org> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-05-20 03:09:41 +03:00
return true;
default:
return false;
}
}
static struct debug_obj_descr timer_debug_descr = {
.name = "timer_list",
.debug_hint = timer_debug_hint,
debugobjects: insulate non-fixup logic related to static obj from fixup callbacks When activating a static object we need make sure that the object is tracked in the object tracker. If it is a non-static object then the activation is illegal. In previous implementation, each subsystem need take care of this in their fixup callbacks. Actually we can put it into debugobjects core. Thus we can save duplicated code, and have *pure* fixup callbacks. To achieve this, a new callback "is_static_object" is introduced to let the type specific code decide whether a object is static or not. If yes, we take it into object tracker, otherwise give warning and invoke fixup callback. This change has paassed debugobjects selftest, and I also do some test with all debugobjects supports enabled. At last, I have a concern about the fixups that can it change the object which is in incorrect state on fixup? Because the 'addr' may not point to any valid object if a non-static object is not tracked. Then Change such object can overwrite someone's memory and cause unexpected behaviour. For example, the timer_fixup_activate bind timer to function stub_timer. Link: http://lkml.kernel.org/r/1462576157-14539-1-git-send-email-changbin.du@intel.com [changbin.du@intel.com: improve code comments where invoke the new is_static_object callback] Link: http://lkml.kernel.org/r/1462777431-8171-1-git-send-email-changbin.du@intel.com Signed-off-by: Du, Changbin <changbin.du@intel.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Josh Triplett <josh@kernel.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tejun Heo <tj@kernel.org> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-05-20 03:09:41 +03:00
.is_static_object = timer_is_static_object,
.fixup_init = timer_fixup_init,
.fixup_activate = timer_fixup_activate,
.fixup_free = timer_fixup_free,
.fixup_assert_init = timer_fixup_assert_init,
};
static inline void debug_timer_init(struct timer_list *timer)
{
debug_object_init(timer, &timer_debug_descr);
}
static inline void debug_timer_activate(struct timer_list *timer)
{
debug_object_activate(timer, &timer_debug_descr);
}
static inline void debug_timer_deactivate(struct timer_list *timer)
{
debug_object_deactivate(timer, &timer_debug_descr);
}
static inline void debug_timer_free(struct timer_list *timer)
{
debug_object_free(timer, &timer_debug_descr);
}
static inline void debug_timer_assert_init(struct timer_list *timer)
{
debug_object_assert_init(timer, &timer_debug_descr);
}
static void do_init_timer(struct timer_list *timer,
void (*func)(struct timer_list *),
unsigned int flags,
const char *name, struct lock_class_key *key);
void init_timer_on_stack_key(struct timer_list *timer,
void (*func)(struct timer_list *),
unsigned int flags,
const char *name, struct lock_class_key *key)
{
debug_object_init_on_stack(timer, &timer_debug_descr);
do_init_timer(timer, func, flags, name, key);
}
EXPORT_SYMBOL_GPL(init_timer_on_stack_key);
void destroy_timer_on_stack(struct timer_list *timer)
{
debug_object_free(timer, &timer_debug_descr);
}
EXPORT_SYMBOL_GPL(destroy_timer_on_stack);
#else
static inline void debug_timer_init(struct timer_list *timer) { }
static inline void debug_timer_activate(struct timer_list *timer) { }
static inline void debug_timer_deactivate(struct timer_list *timer) { }
static inline void debug_timer_assert_init(struct timer_list *timer) { }
#endif
static inline void debug_init(struct timer_list *timer)
{
debug_timer_init(timer);
trace_timer_init(timer);
}
static inline void debug_deactivate(struct timer_list *timer)
{
debug_timer_deactivate(timer);
trace_timer_cancel(timer);
}
static inline void debug_assert_init(struct timer_list *timer)
{
debug_timer_assert_init(timer);
}
static void do_init_timer(struct timer_list *timer,
void (*func)(struct timer_list *),
unsigned int flags,
const char *name, struct lock_class_key *key)
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
{
timer->entry.pprev = NULL;
timer->function = func;
timer: Replace timer base by a cpu index Instead of storing a pointer to the per cpu tvec_base we can simply cache a CPU index in the timer_list and use that to get hold of the correct per cpu tvec_base. This is only used in lock_timer_base() and the slightly larger code is peanuts versus the spinlock operation and the d-cache foot print of the timer wheel. Aside of that this allows to get rid of following nuisances: - boot_tvec_base That statically allocated 4k bss data is just kept around so the timer has a home when it gets statically initialized. It serves no other purpose. With the CPU index we assign the timer to CPU0 at static initialization time and therefor can avoid the whole boot_tvec_base dance. That also simplifies the init code, which just can use the per cpu base. Before: text data bss dec hex filename 17491 9201 4160 30852 7884 ../build/kernel/time/timer.o After: text data bss dec hex filename 17440 9193 0 26633 6809 ../build/kernel/time/timer.o - Overloading the base pointer with various flags The CPU index has enough space to hold the flags (deferrable, irqsafe) so we can get rid of the extra masking and bit fiddling with the base pointer. As a benefit we reduce the size of struct timer_list on 64 bit machines. 4 - 8 bytes, a size reduction up to 15% per struct timer_list, which is a real win as we have tons of them embedded in other structs. This changes also the newly added deferrable printout of the timer start trace point to capture and print all timer->flags, which allows us to decode the target cpu of the timer as well. We might have used bitfields for this, but that would change the static initializers and the init function for no value to accomodate big endian bitfields. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Badhri Jagan Sridharan <Badhri@google.com> Link: http://lkml.kernel.org/r/20150526224511.950084301@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:29 +03:00
timer->flags = flags | raw_smp_processor_id();
lockdep_init_map(&timer->lockdep_map, name, key, 0);
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
}
/**
* init_timer_key - initialize a timer
* @timer: the timer to be initialized
* @func: timer callback function
* @flags: timer flags
* @name: name of the timer
* @key: lockdep class key of the fake lock used for tracking timer
* sync lock dependencies
*
* init_timer_key() must be done to a timer prior calling *any* of the
* other timer functions.
*/
void init_timer_key(struct timer_list *timer,
void (*func)(struct timer_list *), unsigned int flags,
const char *name, struct lock_class_key *key)
{
debug_init(timer);
do_init_timer(timer, func, flags, name, key);
}
EXPORT_SYMBOL(init_timer_key);
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
static inline void detach_timer(struct timer_list *timer, bool clear_pending)
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
{
struct hlist_node *entry = &timer->entry;
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
debug_deactivate(timer);
__hlist_del(entry);
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
if (clear_pending)
entry->pprev = NULL;
entry->next = LIST_POISON2;
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
}
static int detach_if_pending(struct timer_list *timer, struct timer_base *base,
bool clear_pending)
{
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
unsigned idx = timer_get_idx(timer);
if (!timer_pending(timer))
return 0;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
if (hlist_is_singular_node(&timer->entry, base->vectors + idx))
__clear_bit(idx, base->pending_map);
detach_timer(timer, clear_pending);
return 1;
}
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
static inline struct timer_base *get_timer_cpu_base(u32 tflags, u32 cpu)
{
struct timer_base *base = per_cpu_ptr(&timer_bases[BASE_STD], cpu);
/*
* If the timer is deferrable and NO_HZ_COMMON is set then we need
* to use the deferrable base.
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
*/
if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE))
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
base = per_cpu_ptr(&timer_bases[BASE_DEF], cpu);
return base;
}
static inline struct timer_base *get_timer_this_cpu_base(u32 tflags)
{
struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
/*
* If the timer is deferrable and NO_HZ_COMMON is set then we need
* to use the deferrable base.
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
*/
if (IS_ENABLED(CONFIG_NO_HZ_COMMON) && (tflags & TIMER_DEFERRABLE))
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
base = this_cpu_ptr(&timer_bases[BASE_DEF]);
return base;
}
static inline struct timer_base *get_timer_base(u32 tflags)
{
return get_timer_cpu_base(tflags, tflags & TIMER_CPUMASK);
}
static inline struct timer_base *
get_target_base(struct timer_base *base, unsigned tflags)
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
{
#if defined(CONFIG_SMP) && defined(CONFIG_NO_HZ_COMMON)
if (static_branch_likely(&timers_migration_enabled) &&
!(tflags & TIMER_PINNED))
return get_timer_cpu_base(tflags, get_nohz_timer_target());
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
#endif
return get_timer_this_cpu_base(tflags);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
}
static inline void forward_timer_base(struct timer_base *base)
{
#ifdef CONFIG_NO_HZ_COMMON
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
unsigned long jnow;
/*
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
* We only forward the base when we are idle or have just come out of
* idle (must_forward_clk logic), and have a delta between base clock
* and jiffies. In the common case, run_timers will take care of it.
*/
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
if (likely(!base->must_forward_clk))
return;
jnow = READ_ONCE(jiffies);
base->must_forward_clk = base->is_idle;
if ((long)(jnow - base->clk) < 2)
return;
/*
* If the next expiry value is > jiffies, then we fast forward to
* jiffies otherwise we forward to the next expiry value.
*/
if (time_after(base->next_expiry, jnow))
base->clk = jnow;
else
base->clk = base->next_expiry;
#endif
}
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
/*
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
* We are using hashed locking: Holding per_cpu(timer_bases[x]).lock means
* that all timers which are tied to this base are locked, and the base itself
* is locked too.
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
*
* So __run_timers/migrate_timers can safely modify all timers which could
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
* be found in the base->vectors array.
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
*
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
* When a timer is migrating then the TIMER_MIGRATING flag is set and we need
* to wait until the migration is done.
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
*/
static struct timer_base *lock_timer_base(struct timer_list *timer,
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
unsigned long *flags)
__acquires(timer->base->lock)
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
{
for (;;) {
struct timer_base *base;
u32 tf;
/*
* We need to use READ_ONCE() here, otherwise the compiler
* might re-read @tf between the check for TIMER_MIGRATING
* and spin_lock().
*/
tf = READ_ONCE(timer->flags);
timer: Replace timer base by a cpu index Instead of storing a pointer to the per cpu tvec_base we can simply cache a CPU index in the timer_list and use that to get hold of the correct per cpu tvec_base. This is only used in lock_timer_base() and the slightly larger code is peanuts versus the spinlock operation and the d-cache foot print of the timer wheel. Aside of that this allows to get rid of following nuisances: - boot_tvec_base That statically allocated 4k bss data is just kept around so the timer has a home when it gets statically initialized. It serves no other purpose. With the CPU index we assign the timer to CPU0 at static initialization time and therefor can avoid the whole boot_tvec_base dance. That also simplifies the init code, which just can use the per cpu base. Before: text data bss dec hex filename 17491 9201 4160 30852 7884 ../build/kernel/time/timer.o After: text data bss dec hex filename 17440 9193 0 26633 6809 ../build/kernel/time/timer.o - Overloading the base pointer with various flags The CPU index has enough space to hold the flags (deferrable, irqsafe) so we can get rid of the extra masking and bit fiddling with the base pointer. As a benefit we reduce the size of struct timer_list on 64 bit machines. 4 - 8 bytes, a size reduction up to 15% per struct timer_list, which is a real win as we have tons of them embedded in other structs. This changes also the newly added deferrable printout of the timer start trace point to capture and print all timer->flags, which allows us to decode the target cpu of the timer as well. We might have used bitfields for this, but that would change the static initializers and the init function for no value to accomodate big endian bitfields. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Badhri Jagan Sridharan <Badhri@google.com> Link: http://lkml.kernel.org/r/20150526224511.950084301@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:29 +03:00
if (!(tf & TIMER_MIGRATING)) {
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
base = get_timer_base(tf);
raw_spin_lock_irqsave(&base->lock, *flags);
timer: Replace timer base by a cpu index Instead of storing a pointer to the per cpu tvec_base we can simply cache a CPU index in the timer_list and use that to get hold of the correct per cpu tvec_base. This is only used in lock_timer_base() and the slightly larger code is peanuts versus the spinlock operation and the d-cache foot print of the timer wheel. Aside of that this allows to get rid of following nuisances: - boot_tvec_base That statically allocated 4k bss data is just kept around so the timer has a home when it gets statically initialized. It serves no other purpose. With the CPU index we assign the timer to CPU0 at static initialization time and therefor can avoid the whole boot_tvec_base dance. That also simplifies the init code, which just can use the per cpu base. Before: text data bss dec hex filename 17491 9201 4160 30852 7884 ../build/kernel/time/timer.o After: text data bss dec hex filename 17440 9193 0 26633 6809 ../build/kernel/time/timer.o - Overloading the base pointer with various flags The CPU index has enough space to hold the flags (deferrable, irqsafe) so we can get rid of the extra masking and bit fiddling with the base pointer. As a benefit we reduce the size of struct timer_list on 64 bit machines. 4 - 8 bytes, a size reduction up to 15% per struct timer_list, which is a real win as we have tons of them embedded in other structs. This changes also the newly added deferrable printout of the timer start trace point to capture and print all timer->flags, which allows us to decode the target cpu of the timer as well. We might have used bitfields for this, but that would change the static initializers and the init function for no value to accomodate big endian bitfields. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Badhri Jagan Sridharan <Badhri@google.com> Link: http://lkml.kernel.org/r/20150526224511.950084301@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:29 +03:00
if (timer->flags == tf)
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
return base;
raw_spin_unlock_irqrestore(&base->lock, *flags);
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
}
cpu_relax();
}
}
timers: Add a function to start/reduce a timer Add a function, similar to mod_timer(), that will start a timer if it isn't running and will modify it if it is running and has an expiry time longer than the new time. If the timer is running with an expiry time that's the same or sooner, no change is made. The function looks like: int timer_reduce(struct timer_list *timer, unsigned long expires); This can be used by code such as networking code to make it easier to share a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code, the rxrpc_call struct will maintain a number of timeouts: unsigned long ack_at; unsigned long resend_at; unsigned long ping_at; unsigned long expect_rx_by; unsigned long expect_req_by; unsigned long expect_term_by; each of which is set independently of the others. With timer reduction available, when the code needs to set one of the timeouts, it only needs to look at that timeout and then call timer_reduce() to modify the timer, starting it or bringing it forward if necessary. There is no need to refer to the other timeouts to see which is earliest and no need to take any lock other than, potentially, the timer lock inside timer_reduce(). Note, that this does not protect against concurrent invocations of any of the timer functions. As an example, the expect_rx_by timeout above, which terminates a call if we don't get a packet from the server within a certain time window, would be set something like this: unsigned long now = jiffies; unsigned long expect_rx_by = now + packet_receive_timeout; WRITE_ONCE(call->expect_rx_by, expect_rx_by); timer_reduce(&call->timer, expect_rx_by); The timer service code (which might, say, be in a work function) would then check all the timeouts to see which, if any, had triggered, deal with those: t = READ_ONCE(call->ack_at); if (time_after_eq(now, t)) { cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET); set_bit(RXRPC_CALL_EV_ACK, &call->events); } and then restart the timer if necessary by finding the soonest timeout that hasn't yet passed and then calling timer_reduce(). The disadvantage of doing things this way rather than comparing the timers each time and calling mod_timer() is that you *will* take timer events unless you can finish what you're doing and delete the timer in time. The advantage of doing things this way is that you don't need to use a lock to work out when the next timer should be set, other than the timer's own lock - which you might not have to take. [ tglx: Fixed weird formatting and adopted it to pending changes ] Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: keyrings@vger.kernel.org Cc: linux-afs@lists.infradead.org Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
2017-11-09 15:35:07 +03:00
#define MOD_TIMER_PENDING_ONLY 0x01
#define MOD_TIMER_REDUCE 0x02
static inline int
timers: Add a function to start/reduce a timer Add a function, similar to mod_timer(), that will start a timer if it isn't running and will modify it if it is running and has an expiry time longer than the new time. If the timer is running with an expiry time that's the same or sooner, no change is made. The function looks like: int timer_reduce(struct timer_list *timer, unsigned long expires); This can be used by code such as networking code to make it easier to share a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code, the rxrpc_call struct will maintain a number of timeouts: unsigned long ack_at; unsigned long resend_at; unsigned long ping_at; unsigned long expect_rx_by; unsigned long expect_req_by; unsigned long expect_term_by; each of which is set independently of the others. With timer reduction available, when the code needs to set one of the timeouts, it only needs to look at that timeout and then call timer_reduce() to modify the timer, starting it or bringing it forward if necessary. There is no need to refer to the other timeouts to see which is earliest and no need to take any lock other than, potentially, the timer lock inside timer_reduce(). Note, that this does not protect against concurrent invocations of any of the timer functions. As an example, the expect_rx_by timeout above, which terminates a call if we don't get a packet from the server within a certain time window, would be set something like this: unsigned long now = jiffies; unsigned long expect_rx_by = now + packet_receive_timeout; WRITE_ONCE(call->expect_rx_by, expect_rx_by); timer_reduce(&call->timer, expect_rx_by); The timer service code (which might, say, be in a work function) would then check all the timeouts to see which, if any, had triggered, deal with those: t = READ_ONCE(call->ack_at); if (time_after_eq(now, t)) { cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET); set_bit(RXRPC_CALL_EV_ACK, &call->events); } and then restart the timer if necessary by finding the soonest timeout that hasn't yet passed and then calling timer_reduce(). The disadvantage of doing things this way rather than comparing the timers each time and calling mod_timer() is that you *will* take timer events unless you can finish what you're doing and delete the timer in time. The advantage of doing things this way is that you don't need to use a lock to work out when the next timer should be set, other than the timer's own lock - which you might not have to take. [ tglx: Fixed weird formatting and adopted it to pending changes ] Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: keyrings@vger.kernel.org Cc: linux-afs@lists.infradead.org Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
2017-11-09 15:35:07 +03:00
__mod_timer(struct timer_list *timer, unsigned long expires, unsigned int options)
{
struct timer_base *base, *new_base;
timers: Implement optimization for same expiry time in mod_timer() The existing optimization for same expiry time in mod_timer() checks whether the timer expiry time is the same as the new requested expiry time. In the old timer wheel implementation this does not take the slack batching into account, neither does the new implementation evaluate whether the new expiry time will requeue the timer to the same bucket. To optimize that, we can calculate the resulting bucket and check if the new expiry time is different from the current expiry time. This calculation happens outside the base lock held region. If the resulting bucket is the same we can avoid taking the base lock and requeueing the timer. If the timer needs to be requeued then we have to check under the base lock whether the base time has changed between the lockless calculation and taking the lock. If it has changed we need to recalculate under the lock. This optimization takes effect for timers which are enqueued into the less granular wheel levels (1 and above). With a simple test case the functionality has been verified: Before After Match: 5.5% 86.6% Requeue: 94.5% 13.4% Recalc: <0.01% In the non optimized case the timer is requeued in 94.5% of the cases. With the index optimization in place the requeue rate drops to 13.4%. The case where the lockless index calculation has to be redone is less than 0.01%. With a real world test case (networking) we observed the following changes: Before After Match: 97.8% 99.7% Requeue: 2.2% 0.3% Recalc: <0.001% That means two percent fewer lock/requeue/unlock operations done in one of the hot path use cases of timers. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.778527749@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:40 +03:00
unsigned int idx = UINT_MAX;
unsigned long clk = 0, flags;
timer: Reduce timer migration overhead if disabled Eric reported that the timer_migration sysctl is not really nice performance wise as it needs to check at every timer insertion whether the feature is enabled or not. Further the check does not live in the timer code, so we have an extra function call which checks an extra cache line to figure out that it is disabled. We can do better and store that information in the per cpu (hr)timer bases. I pondered to use a static key, but that's a nightmare to update from the nohz code and the timer base cache line is hot anyway when we select a timer base. The old logic enabled the timer migration unconditionally if CONFIG_NO_HZ was set even if nohz was disabled on the kernel command line. With this modification, we start off with migration disabled. The user visible sysctl is still set to enabled. If the kernel switches to NOHZ migration is enabled, if the user did not disable it via the sysctl prior to the switch. If nohz=off is on the kernel command line, migration stays disabled no matter what. Before: 47.76% hog [.] main 14.84% [kernel] [k] _raw_spin_lock_irqsave 9.55% [kernel] [k] _raw_spin_unlock_irqrestore 6.71% [kernel] [k] mod_timer 6.24% [kernel] [k] lock_timer_base.isra.38 3.76% [kernel] [k] detach_if_pending 3.71% [kernel] [k] del_timer 2.50% [kernel] [k] internal_add_timer 1.51% [kernel] [k] get_nohz_timer_target 1.28% [kernel] [k] __internal_add_timer 0.78% [kernel] [k] timerfn 0.48% [kernel] [k] wake_up_nohz_cpu After: 48.10% hog [.] main 15.25% [kernel] [k] _raw_spin_lock_irqsave 9.76% [kernel] [k] _raw_spin_unlock_irqrestore 6.50% [kernel] [k] mod_timer 6.44% [kernel] [k] lock_timer_base.isra.38 3.87% [kernel] [k] detach_if_pending 3.80% [kernel] [k] del_timer 2.67% [kernel] [k] internal_add_timer 1.33% [kernel] [k] __internal_add_timer 0.73% [kernel] [k] timerfn 0.54% [kernel] [k] wake_up_nohz_cpu Reported-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Link: http://lkml.kernel.org/r/20150526224512.127050787@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:33 +03:00
int ret = 0;
BUG_ON(!timer->function);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
/*
timers: Implement optimization for same expiry time in mod_timer() The existing optimization for same expiry time in mod_timer() checks whether the timer expiry time is the same as the new requested expiry time. In the old timer wheel implementation this does not take the slack batching into account, neither does the new implementation evaluate whether the new expiry time will requeue the timer to the same bucket. To optimize that, we can calculate the resulting bucket and check if the new expiry time is different from the current expiry time. This calculation happens outside the base lock held region. If the resulting bucket is the same we can avoid taking the base lock and requeueing the timer. If the timer needs to be requeued then we have to check under the base lock whether the base time has changed between the lockless calculation and taking the lock. If it has changed we need to recalculate under the lock. This optimization takes effect for timers which are enqueued into the less granular wheel levels (1 and above). With a simple test case the functionality has been verified: Before After Match: 5.5% 86.6% Requeue: 94.5% 13.4% Recalc: <0.01% In the non optimized case the timer is requeued in 94.5% of the cases. With the index optimization in place the requeue rate drops to 13.4%. The case where the lockless index calculation has to be redone is less than 0.01%. With a real world test case (networking) we observed the following changes: Before After Match: 97.8% 99.7% Requeue: 2.2% 0.3% Recalc: <0.001% That means two percent fewer lock/requeue/unlock operations done in one of the hot path use cases of timers. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.778527749@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:40 +03:00
* This is a common optimization triggered by the networking code - if
* the timer is re-modified to have the same timeout or ends up in the
* same array bucket then just return:
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
*/
if (timer_pending(timer)) {
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
/*
* The downside of this optimization is that it can result in
* larger granularity than you would get from adding a new
* timer with this expiry.
*/
timers: Add a function to start/reduce a timer Add a function, similar to mod_timer(), that will start a timer if it isn't running and will modify it if it is running and has an expiry time longer than the new time. If the timer is running with an expiry time that's the same or sooner, no change is made. The function looks like: int timer_reduce(struct timer_list *timer, unsigned long expires); This can be used by code such as networking code to make it easier to share a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code, the rxrpc_call struct will maintain a number of timeouts: unsigned long ack_at; unsigned long resend_at; unsigned long ping_at; unsigned long expect_rx_by; unsigned long expect_req_by; unsigned long expect_term_by; each of which is set independently of the others. With timer reduction available, when the code needs to set one of the timeouts, it only needs to look at that timeout and then call timer_reduce() to modify the timer, starting it or bringing it forward if necessary. There is no need to refer to the other timeouts to see which is earliest and no need to take any lock other than, potentially, the timer lock inside timer_reduce(). Note, that this does not protect against concurrent invocations of any of the timer functions. As an example, the expect_rx_by timeout above, which terminates a call if we don't get a packet from the server within a certain time window, would be set something like this: unsigned long now = jiffies; unsigned long expect_rx_by = now + packet_receive_timeout; WRITE_ONCE(call->expect_rx_by, expect_rx_by); timer_reduce(&call->timer, expect_rx_by); The timer service code (which might, say, be in a work function) would then check all the timeouts to see which, if any, had triggered, deal with those: t = READ_ONCE(call->ack_at); if (time_after_eq(now, t)) { cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET); set_bit(RXRPC_CALL_EV_ACK, &call->events); } and then restart the timer if necessary by finding the soonest timeout that hasn't yet passed and then calling timer_reduce(). The disadvantage of doing things this way rather than comparing the timers each time and calling mod_timer() is that you *will* take timer events unless you can finish what you're doing and delete the timer in time. The advantage of doing things this way is that you don't need to use a lock to work out when the next timer should be set, other than the timer's own lock - which you might not have to take. [ tglx: Fixed weird formatting and adopted it to pending changes ] Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: keyrings@vger.kernel.org Cc: linux-afs@lists.infradead.org Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
2017-11-09 15:35:07 +03:00
long diff = timer->expires - expires;
if (!diff)
return 1;
if (options & MOD_TIMER_REDUCE && diff <= 0)
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
return 1;
timers: Implement optimization for same expiry time in mod_timer() The existing optimization for same expiry time in mod_timer() checks whether the timer expiry time is the same as the new requested expiry time. In the old timer wheel implementation this does not take the slack batching into account, neither does the new implementation evaluate whether the new expiry time will requeue the timer to the same bucket. To optimize that, we can calculate the resulting bucket and check if the new expiry time is different from the current expiry time. This calculation happens outside the base lock held region. If the resulting bucket is the same we can avoid taking the base lock and requeueing the timer. If the timer needs to be requeued then we have to check under the base lock whether the base time has changed between the lockless calculation and taking the lock. If it has changed we need to recalculate under the lock. This optimization takes effect for timers which are enqueued into the less granular wheel levels (1 and above). With a simple test case the functionality has been verified: Before After Match: 5.5% 86.6% Requeue: 94.5% 13.4% Recalc: <0.01% In the non optimized case the timer is requeued in 94.5% of the cases. With the index optimization in place the requeue rate drops to 13.4%. The case where the lockless index calculation has to be redone is less than 0.01%. With a real world test case (networking) we observed the following changes: Before After Match: 97.8% 99.7% Requeue: 2.2% 0.3% Recalc: <0.001% That means two percent fewer lock/requeue/unlock operations done in one of the hot path use cases of timers. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.778527749@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:40 +03:00
/*
* We lock timer base and calculate the bucket index right
* here. If the timer ends up in the same bucket, then we
* just update the expiry time and avoid the whole
* dequeue/enqueue dance.
timers: Implement optimization for same expiry time in mod_timer() The existing optimization for same expiry time in mod_timer() checks whether the timer expiry time is the same as the new requested expiry time. In the old timer wheel implementation this does not take the slack batching into account, neither does the new implementation evaluate whether the new expiry time will requeue the timer to the same bucket. To optimize that, we can calculate the resulting bucket and check if the new expiry time is different from the current expiry time. This calculation happens outside the base lock held region. If the resulting bucket is the same we can avoid taking the base lock and requeueing the timer. If the timer needs to be requeued then we have to check under the base lock whether the base time has changed between the lockless calculation and taking the lock. If it has changed we need to recalculate under the lock. This optimization takes effect for timers which are enqueued into the less granular wheel levels (1 and above). With a simple test case the functionality has been verified: Before After Match: 5.5% 86.6% Requeue: 94.5% 13.4% Recalc: <0.01% In the non optimized case the timer is requeued in 94.5% of the cases. With the index optimization in place the requeue rate drops to 13.4%. The case where the lockless index calculation has to be redone is less than 0.01%. With a real world test case (networking) we observed the following changes: Before After Match: 97.8% 99.7% Requeue: 2.2% 0.3% Recalc: <0.001% That means two percent fewer lock/requeue/unlock operations done in one of the hot path use cases of timers. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.778527749@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:40 +03:00
*/
base = lock_timer_base(timer, &flags);
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
forward_timer_base(base);
timers: Implement optimization for same expiry time in mod_timer() The existing optimization for same expiry time in mod_timer() checks whether the timer expiry time is the same as the new requested expiry time. In the old timer wheel implementation this does not take the slack batching into account, neither does the new implementation evaluate whether the new expiry time will requeue the timer to the same bucket. To optimize that, we can calculate the resulting bucket and check if the new expiry time is different from the current expiry time. This calculation happens outside the base lock held region. If the resulting bucket is the same we can avoid taking the base lock and requeueing the timer. If the timer needs to be requeued then we have to check under the base lock whether the base time has changed between the lockless calculation and taking the lock. If it has changed we need to recalculate under the lock. This optimization takes effect for timers which are enqueued into the less granular wheel levels (1 and above). With a simple test case the functionality has been verified: Before After Match: 5.5% 86.6% Requeue: 94.5% 13.4% Recalc: <0.01% In the non optimized case the timer is requeued in 94.5% of the cases. With the index optimization in place the requeue rate drops to 13.4%. The case where the lockless index calculation has to be redone is less than 0.01%. With a real world test case (networking) we observed the following changes: Before After Match: 97.8% 99.7% Requeue: 2.2% 0.3% Recalc: <0.001% That means two percent fewer lock/requeue/unlock operations done in one of the hot path use cases of timers. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.778527749@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:40 +03:00
timers: Add a function to start/reduce a timer Add a function, similar to mod_timer(), that will start a timer if it isn't running and will modify it if it is running and has an expiry time longer than the new time. If the timer is running with an expiry time that's the same or sooner, no change is made. The function looks like: int timer_reduce(struct timer_list *timer, unsigned long expires); This can be used by code such as networking code to make it easier to share a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code, the rxrpc_call struct will maintain a number of timeouts: unsigned long ack_at; unsigned long resend_at; unsigned long ping_at; unsigned long expect_rx_by; unsigned long expect_req_by; unsigned long expect_term_by; each of which is set independently of the others. With timer reduction available, when the code needs to set one of the timeouts, it only needs to look at that timeout and then call timer_reduce() to modify the timer, starting it or bringing it forward if necessary. There is no need to refer to the other timeouts to see which is earliest and no need to take any lock other than, potentially, the timer lock inside timer_reduce(). Note, that this does not protect against concurrent invocations of any of the timer functions. As an example, the expect_rx_by timeout above, which terminates a call if we don't get a packet from the server within a certain time window, would be set something like this: unsigned long now = jiffies; unsigned long expect_rx_by = now + packet_receive_timeout; WRITE_ONCE(call->expect_rx_by, expect_rx_by); timer_reduce(&call->timer, expect_rx_by); The timer service code (which might, say, be in a work function) would then check all the timeouts to see which, if any, had triggered, deal with those: t = READ_ONCE(call->ack_at); if (time_after_eq(now, t)) { cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET); set_bit(RXRPC_CALL_EV_ACK, &call->events); } and then restart the timer if necessary by finding the soonest timeout that hasn't yet passed and then calling timer_reduce(). The disadvantage of doing things this way rather than comparing the timers each time and calling mod_timer() is that you *will* take timer events unless you can finish what you're doing and delete the timer in time. The advantage of doing things this way is that you don't need to use a lock to work out when the next timer should be set, other than the timer's own lock - which you might not have to take. [ tglx: Fixed weird formatting and adopted it to pending changes ] Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: keyrings@vger.kernel.org Cc: linux-afs@lists.infradead.org Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
2017-11-09 15:35:07 +03:00
if (timer_pending(timer) && (options & MOD_TIMER_REDUCE) &&
time_before_eq(timer->expires, expires)) {
ret = 1;
goto out_unlock;
}
clk = base->clk;
timers: Implement optimization for same expiry time in mod_timer() The existing optimization for same expiry time in mod_timer() checks whether the timer expiry time is the same as the new requested expiry time. In the old timer wheel implementation this does not take the slack batching into account, neither does the new implementation evaluate whether the new expiry time will requeue the timer to the same bucket. To optimize that, we can calculate the resulting bucket and check if the new expiry time is different from the current expiry time. This calculation happens outside the base lock held region. If the resulting bucket is the same we can avoid taking the base lock and requeueing the timer. If the timer needs to be requeued then we have to check under the base lock whether the base time has changed between the lockless calculation and taking the lock. If it has changed we need to recalculate under the lock. This optimization takes effect for timers which are enqueued into the less granular wheel levels (1 and above). With a simple test case the functionality has been verified: Before After Match: 5.5% 86.6% Requeue: 94.5% 13.4% Recalc: <0.01% In the non optimized case the timer is requeued in 94.5% of the cases. With the index optimization in place the requeue rate drops to 13.4%. The case where the lockless index calculation has to be redone is less than 0.01%. With a real world test case (networking) we observed the following changes: Before After Match: 97.8% 99.7% Requeue: 2.2% 0.3% Recalc: <0.001% That means two percent fewer lock/requeue/unlock operations done in one of the hot path use cases of timers. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.778527749@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:40 +03:00
idx = calc_wheel_index(expires, clk);
/*
* Retrieve and compare the array index of the pending
* timer. If it matches set the expiry to the new value so a
* subsequent call will exit in the expires check above.
*/
if (idx == timer_get_idx(timer)) {
timers: Add a function to start/reduce a timer Add a function, similar to mod_timer(), that will start a timer if it isn't running and will modify it if it is running and has an expiry time longer than the new time. If the timer is running with an expiry time that's the same or sooner, no change is made. The function looks like: int timer_reduce(struct timer_list *timer, unsigned long expires); This can be used by code such as networking code to make it easier to share a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code, the rxrpc_call struct will maintain a number of timeouts: unsigned long ack_at; unsigned long resend_at; unsigned long ping_at; unsigned long expect_rx_by; unsigned long expect_req_by; unsigned long expect_term_by; each of which is set independently of the others. With timer reduction available, when the code needs to set one of the timeouts, it only needs to look at that timeout and then call timer_reduce() to modify the timer, starting it or bringing it forward if necessary. There is no need to refer to the other timeouts to see which is earliest and no need to take any lock other than, potentially, the timer lock inside timer_reduce(). Note, that this does not protect against concurrent invocations of any of the timer functions. As an example, the expect_rx_by timeout above, which terminates a call if we don't get a packet from the server within a certain time window, would be set something like this: unsigned long now = jiffies; unsigned long expect_rx_by = now + packet_receive_timeout; WRITE_ONCE(call->expect_rx_by, expect_rx_by); timer_reduce(&call->timer, expect_rx_by); The timer service code (which might, say, be in a work function) would then check all the timeouts to see which, if any, had triggered, deal with those: t = READ_ONCE(call->ack_at); if (time_after_eq(now, t)) { cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET); set_bit(RXRPC_CALL_EV_ACK, &call->events); } and then restart the timer if necessary by finding the soonest timeout that hasn't yet passed and then calling timer_reduce(). The disadvantage of doing things this way rather than comparing the timers each time and calling mod_timer() is that you *will* take timer events unless you can finish what you're doing and delete the timer in time. The advantage of doing things this way is that you don't need to use a lock to work out when the next timer should be set, other than the timer's own lock - which you might not have to take. [ tglx: Fixed weird formatting and adopted it to pending changes ] Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: keyrings@vger.kernel.org Cc: linux-afs@lists.infradead.org Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
2017-11-09 15:35:07 +03:00
if (!(options & MOD_TIMER_REDUCE))
timer->expires = expires;
else if (time_after(timer->expires, expires))
timer->expires = expires;
ret = 1;
goto out_unlock;
timers: Implement optimization for same expiry time in mod_timer() The existing optimization for same expiry time in mod_timer() checks whether the timer expiry time is the same as the new requested expiry time. In the old timer wheel implementation this does not take the slack batching into account, neither does the new implementation evaluate whether the new expiry time will requeue the timer to the same bucket. To optimize that, we can calculate the resulting bucket and check if the new expiry time is different from the current expiry time. This calculation happens outside the base lock held region. If the resulting bucket is the same we can avoid taking the base lock and requeueing the timer. If the timer needs to be requeued then we have to check under the base lock whether the base time has changed between the lockless calculation and taking the lock. If it has changed we need to recalculate under the lock. This optimization takes effect for timers which are enqueued into the less granular wheel levels (1 and above). With a simple test case the functionality has been verified: Before After Match: 5.5% 86.6% Requeue: 94.5% 13.4% Recalc: <0.01% In the non optimized case the timer is requeued in 94.5% of the cases. With the index optimization in place the requeue rate drops to 13.4%. The case where the lockless index calculation has to be redone is less than 0.01%. With a real world test case (networking) we observed the following changes: Before After Match: 97.8% 99.7% Requeue: 2.2% 0.3% Recalc: <0.001% That means two percent fewer lock/requeue/unlock operations done in one of the hot path use cases of timers. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.778527749@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:40 +03:00
}
} else {
base = lock_timer_base(timer, &flags);
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
forward_timer_base(base);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
}
ret = detach_if_pending(timer, base, false);
timers: Add a function to start/reduce a timer Add a function, similar to mod_timer(), that will start a timer if it isn't running and will modify it if it is running and has an expiry time longer than the new time. If the timer is running with an expiry time that's the same or sooner, no change is made. The function looks like: int timer_reduce(struct timer_list *timer, unsigned long expires); This can be used by code such as networking code to make it easier to share a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code, the rxrpc_call struct will maintain a number of timeouts: unsigned long ack_at; unsigned long resend_at; unsigned long ping_at; unsigned long expect_rx_by; unsigned long expect_req_by; unsigned long expect_term_by; each of which is set independently of the others. With timer reduction available, when the code needs to set one of the timeouts, it only needs to look at that timeout and then call timer_reduce() to modify the timer, starting it or bringing it forward if necessary. There is no need to refer to the other timeouts to see which is earliest and no need to take any lock other than, potentially, the timer lock inside timer_reduce(). Note, that this does not protect against concurrent invocations of any of the timer functions. As an example, the expect_rx_by timeout above, which terminates a call if we don't get a packet from the server within a certain time window, would be set something like this: unsigned long now = jiffies; unsigned long expect_rx_by = now + packet_receive_timeout; WRITE_ONCE(call->expect_rx_by, expect_rx_by); timer_reduce(&call->timer, expect_rx_by); The timer service code (which might, say, be in a work function) would then check all the timeouts to see which, if any, had triggered, deal with those: t = READ_ONCE(call->ack_at); if (time_after_eq(now, t)) { cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET); set_bit(RXRPC_CALL_EV_ACK, &call->events); } and then restart the timer if necessary by finding the soonest timeout that hasn't yet passed and then calling timer_reduce(). The disadvantage of doing things this way rather than comparing the timers each time and calling mod_timer() is that you *will* take timer events unless you can finish what you're doing and delete the timer in time. The advantage of doing things this way is that you don't need to use a lock to work out when the next timer should be set, other than the timer's own lock - which you might not have to take. [ tglx: Fixed weird formatting and adopted it to pending changes ] Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: keyrings@vger.kernel.org Cc: linux-afs@lists.infradead.org Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
2017-11-09 15:35:07 +03:00
if (!ret && (options & MOD_TIMER_PENDING_ONLY))
goto out_unlock;
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
new_base = get_target_base(base, timer->flags);
if (base != new_base) {
/*
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
* We are trying to schedule the timer on the new base.
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
* However we can't change timer's base while it is running,
* otherwise del_timer_sync() can't detect that the timer's
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
* handler yet has not finished. This also guarantees that the
* timer is serialized wrt itself.
*/
if (likely(base->running_timer != timer)) {
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
/* See the comment in lock_timer_base() */
timer: Replace timer base by a cpu index Instead of storing a pointer to the per cpu tvec_base we can simply cache a CPU index in the timer_list and use that to get hold of the correct per cpu tvec_base. This is only used in lock_timer_base() and the slightly larger code is peanuts versus the spinlock operation and the d-cache foot print of the timer wheel. Aside of that this allows to get rid of following nuisances: - boot_tvec_base That statically allocated 4k bss data is just kept around so the timer has a home when it gets statically initialized. It serves no other purpose. With the CPU index we assign the timer to CPU0 at static initialization time and therefor can avoid the whole boot_tvec_base dance. That also simplifies the init code, which just can use the per cpu base. Before: text data bss dec hex filename 17491 9201 4160 30852 7884 ../build/kernel/time/timer.o After: text data bss dec hex filename 17440 9193 0 26633 6809 ../build/kernel/time/timer.o - Overloading the base pointer with various flags The CPU index has enough space to hold the flags (deferrable, irqsafe) so we can get rid of the extra masking and bit fiddling with the base pointer. As a benefit we reduce the size of struct timer_list on 64 bit machines. 4 - 8 bytes, a size reduction up to 15% per struct timer_list, which is a real win as we have tons of them embedded in other structs. This changes also the newly added deferrable printout of the timer start trace point to capture and print all timer->flags, which allows us to decode the target cpu of the timer as well. We might have used bitfields for this, but that would change the static initializers and the init function for no value to accomodate big endian bitfields. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Badhri Jagan Sridharan <Badhri@google.com> Link: http://lkml.kernel.org/r/20150526224511.950084301@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:29 +03:00
timer->flags |= TIMER_MIGRATING;
raw_spin_unlock(&base->lock);
base = new_base;
raw_spin_lock(&base->lock);
WRITE_ONCE(timer->flags,
(timer->flags & ~TIMER_BASEMASK) | base->cpu);
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
forward_timer_base(base);
}
}
debug_timer_activate(timer);
timer->expires = expires;
timers: Implement optimization for same expiry time in mod_timer() The existing optimization for same expiry time in mod_timer() checks whether the timer expiry time is the same as the new requested expiry time. In the old timer wheel implementation this does not take the slack batching into account, neither does the new implementation evaluate whether the new expiry time will requeue the timer to the same bucket. To optimize that, we can calculate the resulting bucket and check if the new expiry time is different from the current expiry time. This calculation happens outside the base lock held region. If the resulting bucket is the same we can avoid taking the base lock and requeueing the timer. If the timer needs to be requeued then we have to check under the base lock whether the base time has changed between the lockless calculation and taking the lock. If it has changed we need to recalculate under the lock. This optimization takes effect for timers which are enqueued into the less granular wheel levels (1 and above). With a simple test case the functionality has been verified: Before After Match: 5.5% 86.6% Requeue: 94.5% 13.4% Recalc: <0.01% In the non optimized case the timer is requeued in 94.5% of the cases. With the index optimization in place the requeue rate drops to 13.4%. The case where the lockless index calculation has to be redone is less than 0.01%. With a real world test case (networking) we observed the following changes: Before After Match: 97.8% 99.7% Requeue: 2.2% 0.3% Recalc: <0.001% That means two percent fewer lock/requeue/unlock operations done in one of the hot path use cases of timers. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.778527749@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:40 +03:00
/*
* If 'idx' was calculated above and the base time did not advance
* between calculating 'idx' and possibly switching the base, only
* enqueue_timer() and trigger_dyntick_cpu() is required. Otherwise
* we need to (re)calculate the wheel index via
* internal_add_timer().
timers: Implement optimization for same expiry time in mod_timer() The existing optimization for same expiry time in mod_timer() checks whether the timer expiry time is the same as the new requested expiry time. In the old timer wheel implementation this does not take the slack batching into account, neither does the new implementation evaluate whether the new expiry time will requeue the timer to the same bucket. To optimize that, we can calculate the resulting bucket and check if the new expiry time is different from the current expiry time. This calculation happens outside the base lock held region. If the resulting bucket is the same we can avoid taking the base lock and requeueing the timer. If the timer needs to be requeued then we have to check under the base lock whether the base time has changed between the lockless calculation and taking the lock. If it has changed we need to recalculate under the lock. This optimization takes effect for timers which are enqueued into the less granular wheel levels (1 and above). With a simple test case the functionality has been verified: Before After Match: 5.5% 86.6% Requeue: 94.5% 13.4% Recalc: <0.01% In the non optimized case the timer is requeued in 94.5% of the cases. With the index optimization in place the requeue rate drops to 13.4%. The case where the lockless index calculation has to be redone is less than 0.01%. With a real world test case (networking) we observed the following changes: Before After Match: 97.8% 99.7% Requeue: 2.2% 0.3% Recalc: <0.001% That means two percent fewer lock/requeue/unlock operations done in one of the hot path use cases of timers. Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.778527749@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:40 +03:00
*/
if (idx != UINT_MAX && clk == base->clk) {
enqueue_timer(base, timer, idx);
trigger_dyntick_cpu(base, timer);
} else {
internal_add_timer(base, timer);
}
out_unlock:
raw_spin_unlock_irqrestore(&base->lock, flags);
return ret;
}
/**
* mod_timer_pending - modify a pending timer's timeout
* @timer: the pending timer to be modified
* @expires: new timeout in jiffies
*
* mod_timer_pending() is the same for pending timers as mod_timer(),
* but will not re-activate and modify already deleted timers.
*
* It is useful for unserialized use of timers.
*/
int mod_timer_pending(struct timer_list *timer, unsigned long expires)
{
timers: Add a function to start/reduce a timer Add a function, similar to mod_timer(), that will start a timer if it isn't running and will modify it if it is running and has an expiry time longer than the new time. If the timer is running with an expiry time that's the same or sooner, no change is made. The function looks like: int timer_reduce(struct timer_list *timer, unsigned long expires); This can be used by code such as networking code to make it easier to share a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code, the rxrpc_call struct will maintain a number of timeouts: unsigned long ack_at; unsigned long resend_at; unsigned long ping_at; unsigned long expect_rx_by; unsigned long expect_req_by; unsigned long expect_term_by; each of which is set independently of the others. With timer reduction available, when the code needs to set one of the timeouts, it only needs to look at that timeout and then call timer_reduce() to modify the timer, starting it or bringing it forward if necessary. There is no need to refer to the other timeouts to see which is earliest and no need to take any lock other than, potentially, the timer lock inside timer_reduce(). Note, that this does not protect against concurrent invocations of any of the timer functions. As an example, the expect_rx_by timeout above, which terminates a call if we don't get a packet from the server within a certain time window, would be set something like this: unsigned long now = jiffies; unsigned long expect_rx_by = now + packet_receive_timeout; WRITE_ONCE(call->expect_rx_by, expect_rx_by); timer_reduce(&call->timer, expect_rx_by); The timer service code (which might, say, be in a work function) would then check all the timeouts to see which, if any, had triggered, deal with those: t = READ_ONCE(call->ack_at); if (time_after_eq(now, t)) { cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET); set_bit(RXRPC_CALL_EV_ACK, &call->events); } and then restart the timer if necessary by finding the soonest timeout that hasn't yet passed and then calling timer_reduce(). The disadvantage of doing things this way rather than comparing the timers each time and calling mod_timer() is that you *will* take timer events unless you can finish what you're doing and delete the timer in time. The advantage of doing things this way is that you don't need to use a lock to work out when the next timer should be set, other than the timer's own lock - which you might not have to take. [ tglx: Fixed weird formatting and adopted it to pending changes ] Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: keyrings@vger.kernel.org Cc: linux-afs@lists.infradead.org Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
2017-11-09 15:35:07 +03:00
return __mod_timer(timer, expires, MOD_TIMER_PENDING_ONLY);
}
EXPORT_SYMBOL(mod_timer_pending);
/**
* mod_timer - modify a timer's timeout
* @timer: the timer to be modified
* @expires: new timeout in jiffies
*
* mod_timer() is a more efficient way to update the expire field of an
* active timer (if the timer is inactive it will be activated)
*
* mod_timer(timer, expires) is equivalent to:
*
* del_timer(timer); timer->expires = expires; add_timer(timer);
*
* Note that if there are multiple unserialized concurrent users of the
* same timer, then mod_timer() is the only safe way to modify the timeout,
* since add_timer() cannot modify an already running timer.
*
* The function returns whether it has modified a pending timer or not.
* (ie. mod_timer() of an inactive timer returns 0, mod_timer() of an
* active timer returns 1.)
*/
int mod_timer(struct timer_list *timer, unsigned long expires)
{
timers: Add a function to start/reduce a timer Add a function, similar to mod_timer(), that will start a timer if it isn't running and will modify it if it is running and has an expiry time longer than the new time. If the timer is running with an expiry time that's the same or sooner, no change is made. The function looks like: int timer_reduce(struct timer_list *timer, unsigned long expires); This can be used by code such as networking code to make it easier to share a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code, the rxrpc_call struct will maintain a number of timeouts: unsigned long ack_at; unsigned long resend_at; unsigned long ping_at; unsigned long expect_rx_by; unsigned long expect_req_by; unsigned long expect_term_by; each of which is set independently of the others. With timer reduction available, when the code needs to set one of the timeouts, it only needs to look at that timeout and then call timer_reduce() to modify the timer, starting it or bringing it forward if necessary. There is no need to refer to the other timeouts to see which is earliest and no need to take any lock other than, potentially, the timer lock inside timer_reduce(). Note, that this does not protect against concurrent invocations of any of the timer functions. As an example, the expect_rx_by timeout above, which terminates a call if we don't get a packet from the server within a certain time window, would be set something like this: unsigned long now = jiffies; unsigned long expect_rx_by = now + packet_receive_timeout; WRITE_ONCE(call->expect_rx_by, expect_rx_by); timer_reduce(&call->timer, expect_rx_by); The timer service code (which might, say, be in a work function) would then check all the timeouts to see which, if any, had triggered, deal with those: t = READ_ONCE(call->ack_at); if (time_after_eq(now, t)) { cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET); set_bit(RXRPC_CALL_EV_ACK, &call->events); } and then restart the timer if necessary by finding the soonest timeout that hasn't yet passed and then calling timer_reduce(). The disadvantage of doing things this way rather than comparing the timers each time and calling mod_timer() is that you *will* take timer events unless you can finish what you're doing and delete the timer in time. The advantage of doing things this way is that you don't need to use a lock to work out when the next timer should be set, other than the timer's own lock - which you might not have to take. [ tglx: Fixed weird formatting and adopted it to pending changes ] Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: keyrings@vger.kernel.org Cc: linux-afs@lists.infradead.org Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
2017-11-09 15:35:07 +03:00
return __mod_timer(timer, expires, 0);
}
EXPORT_SYMBOL(mod_timer);
timers: Add a function to start/reduce a timer Add a function, similar to mod_timer(), that will start a timer if it isn't running and will modify it if it is running and has an expiry time longer than the new time. If the timer is running with an expiry time that's the same or sooner, no change is made. The function looks like: int timer_reduce(struct timer_list *timer, unsigned long expires); This can be used by code such as networking code to make it easier to share a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code, the rxrpc_call struct will maintain a number of timeouts: unsigned long ack_at; unsigned long resend_at; unsigned long ping_at; unsigned long expect_rx_by; unsigned long expect_req_by; unsigned long expect_term_by; each of which is set independently of the others. With timer reduction available, when the code needs to set one of the timeouts, it only needs to look at that timeout and then call timer_reduce() to modify the timer, starting it or bringing it forward if necessary. There is no need to refer to the other timeouts to see which is earliest and no need to take any lock other than, potentially, the timer lock inside timer_reduce(). Note, that this does not protect against concurrent invocations of any of the timer functions. As an example, the expect_rx_by timeout above, which terminates a call if we don't get a packet from the server within a certain time window, would be set something like this: unsigned long now = jiffies; unsigned long expect_rx_by = now + packet_receive_timeout; WRITE_ONCE(call->expect_rx_by, expect_rx_by); timer_reduce(&call->timer, expect_rx_by); The timer service code (which might, say, be in a work function) would then check all the timeouts to see which, if any, had triggered, deal with those: t = READ_ONCE(call->ack_at); if (time_after_eq(now, t)) { cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET); set_bit(RXRPC_CALL_EV_ACK, &call->events); } and then restart the timer if necessary by finding the soonest timeout that hasn't yet passed and then calling timer_reduce(). The disadvantage of doing things this way rather than comparing the timers each time and calling mod_timer() is that you *will* take timer events unless you can finish what you're doing and delete the timer in time. The advantage of doing things this way is that you don't need to use a lock to work out when the next timer should be set, other than the timer's own lock - which you might not have to take. [ tglx: Fixed weird formatting and adopted it to pending changes ] Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: keyrings@vger.kernel.org Cc: linux-afs@lists.infradead.org Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
2017-11-09 15:35:07 +03:00
/**
* timer_reduce - Modify a timer's timeout if it would reduce the timeout
* @timer: The timer to be modified
* @expires: New timeout in jiffies
*
* timer_reduce() is very similar to mod_timer(), except that it will only
* modify a running timer if that would reduce the expiration time (it will
* start a timer that isn't running).
*/
int timer_reduce(struct timer_list *timer, unsigned long expires)
{
return __mod_timer(timer, expires, MOD_TIMER_REDUCE);
}
EXPORT_SYMBOL(timer_reduce);
/**
* add_timer - start a timer
* @timer: the timer to be added
*
* The kernel will do a ->function(@timer) callback from the
* timer interrupt at the ->expires point in the future. The
* current time is 'jiffies'.
*
* The timer's ->expires, ->function fields must be set prior calling this
* function.
*
* Timers with an ->expires field in the past will be executed in the next
* timer tick.
*/
void add_timer(struct timer_list *timer)
{
BUG_ON(timer_pending(timer));
mod_timer(timer, timer->expires);
}
EXPORT_SYMBOL(add_timer);
/**
* add_timer_on - start a timer on a particular CPU
* @timer: the timer to be added
* @cpu: the CPU to start it on
*
* This is not very scalable on SMP. Double adds are not possible.
*/
void add_timer_on(struct timer_list *timer, int cpu)
{
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
struct timer_base *new_base, *base;
unsigned long flags;
BUG_ON(timer_pending(timer) || !timer->function);
timers: Use proper base migration in add_timer_on() Regardless of the previous CPU a timer was on, add_timer_on() currently simply sets timer->flags to the new CPU. As the caller must be seeing the timer as idle, this is locally fine, but the timer leaving the old base while unlocked can lead to race conditions as follows. Let's say timer was on cpu 0. cpu 0 cpu 1 ----------------------------------------------------------------------------- del_timer(timer) succeeds del_timer(timer) lock_timer_base(timer) locks cpu_0_base add_timer_on(timer, 1) spin_lock(&cpu_1_base->lock) timer->flags set to cpu_1_base operates on @timer operates on @timer This triggered with mod_delayed_work_on() which contains "if (del_timer()) add_timer_on()" sequence eventually leading to the following oops. BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<ffffffff810ca6e9>] detach_if_pending+0x69/0x1a0 ... Workqueue: wqthrash wqthrash_workfunc [wqthrash] task: ffff8800172ca680 ti: ffff8800172d0000 task.ti: ffff8800172d0000 RIP: 0010:[<ffffffff810ca6e9>] [<ffffffff810ca6e9>] detach_if_pending+0x69/0x1a0 ... Call Trace: [<ffffffff810cb0b4>] del_timer+0x44/0x60 [<ffffffff8106e836>] try_to_grab_pending+0xb6/0x160 [<ffffffff8106e913>] mod_delayed_work_on+0x33/0x80 [<ffffffffa0000081>] wqthrash_workfunc+0x61/0x90 [wqthrash] [<ffffffff8106dba8>] process_one_work+0x1e8/0x650 [<ffffffff8106e05e>] worker_thread+0x4e/0x450 [<ffffffff810746af>] kthread+0xef/0x110 [<ffffffff8185980f>] ret_from_fork+0x3f/0x70 Fix it by updating add_timer_on() to perform proper migration as __mod_timer() does. Reported-and-tested-by: Jeff Layton <jlayton@poochiereds.net> Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Chris Worley <chris.worley@primarydata.com> Cc: bfields@fieldses.org Cc: Michael Skralivetsky <michael.skralivetsky@primarydata.com> Cc: Trond Myklebust <trond.myklebust@primarydata.com> Cc: Shaohua Li <shli@fb.com> Cc: Jeff Layton <jlayton@poochiereds.net> Cc: kernel-team@fb.com Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20151029103113.2f893924@tlielax.poochiereds.net Link: http://lkml.kernel.org/r/20151104171533.GI5749@mtj.duckdns.org Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-11-04 20:15:33 +03:00
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
new_base = get_timer_cpu_base(timer->flags, cpu);
timers: Use proper base migration in add_timer_on() Regardless of the previous CPU a timer was on, add_timer_on() currently simply sets timer->flags to the new CPU. As the caller must be seeing the timer as idle, this is locally fine, but the timer leaving the old base while unlocked can lead to race conditions as follows. Let's say timer was on cpu 0. cpu 0 cpu 1 ----------------------------------------------------------------------------- del_timer(timer) succeeds del_timer(timer) lock_timer_base(timer) locks cpu_0_base add_timer_on(timer, 1) spin_lock(&cpu_1_base->lock) timer->flags set to cpu_1_base operates on @timer operates on @timer This triggered with mod_delayed_work_on() which contains "if (del_timer()) add_timer_on()" sequence eventually leading to the following oops. BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<ffffffff810ca6e9>] detach_if_pending+0x69/0x1a0 ... Workqueue: wqthrash wqthrash_workfunc [wqthrash] task: ffff8800172ca680 ti: ffff8800172d0000 task.ti: ffff8800172d0000 RIP: 0010:[<ffffffff810ca6e9>] [<ffffffff810ca6e9>] detach_if_pending+0x69/0x1a0 ... Call Trace: [<ffffffff810cb0b4>] del_timer+0x44/0x60 [<ffffffff8106e836>] try_to_grab_pending+0xb6/0x160 [<ffffffff8106e913>] mod_delayed_work_on+0x33/0x80 [<ffffffffa0000081>] wqthrash_workfunc+0x61/0x90 [wqthrash] [<ffffffff8106dba8>] process_one_work+0x1e8/0x650 [<ffffffff8106e05e>] worker_thread+0x4e/0x450 [<ffffffff810746af>] kthread+0xef/0x110 [<ffffffff8185980f>] ret_from_fork+0x3f/0x70 Fix it by updating add_timer_on() to perform proper migration as __mod_timer() does. Reported-and-tested-by: Jeff Layton <jlayton@poochiereds.net> Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Chris Worley <chris.worley@primarydata.com> Cc: bfields@fieldses.org Cc: Michael Skralivetsky <michael.skralivetsky@primarydata.com> Cc: Trond Myklebust <trond.myklebust@primarydata.com> Cc: Shaohua Li <shli@fb.com> Cc: Jeff Layton <jlayton@poochiereds.net> Cc: kernel-team@fb.com Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20151029103113.2f893924@tlielax.poochiereds.net Link: http://lkml.kernel.org/r/20151104171533.GI5749@mtj.duckdns.org Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-11-04 20:15:33 +03:00
/*
* If @timer was on a different CPU, it should be migrated with the
* old base locked to prevent other operations proceeding with the
* wrong base locked. See lock_timer_base().
*/
base = lock_timer_base(timer, &flags);
if (base != new_base) {
timer->flags |= TIMER_MIGRATING;
raw_spin_unlock(&base->lock);
timers: Use proper base migration in add_timer_on() Regardless of the previous CPU a timer was on, add_timer_on() currently simply sets timer->flags to the new CPU. As the caller must be seeing the timer as idle, this is locally fine, but the timer leaving the old base while unlocked can lead to race conditions as follows. Let's say timer was on cpu 0. cpu 0 cpu 1 ----------------------------------------------------------------------------- del_timer(timer) succeeds del_timer(timer) lock_timer_base(timer) locks cpu_0_base add_timer_on(timer, 1) spin_lock(&cpu_1_base->lock) timer->flags set to cpu_1_base operates on @timer operates on @timer This triggered with mod_delayed_work_on() which contains "if (del_timer()) add_timer_on()" sequence eventually leading to the following oops. BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<ffffffff810ca6e9>] detach_if_pending+0x69/0x1a0 ... Workqueue: wqthrash wqthrash_workfunc [wqthrash] task: ffff8800172ca680 ti: ffff8800172d0000 task.ti: ffff8800172d0000 RIP: 0010:[<ffffffff810ca6e9>] [<ffffffff810ca6e9>] detach_if_pending+0x69/0x1a0 ... Call Trace: [<ffffffff810cb0b4>] del_timer+0x44/0x60 [<ffffffff8106e836>] try_to_grab_pending+0xb6/0x160 [<ffffffff8106e913>] mod_delayed_work_on+0x33/0x80 [<ffffffffa0000081>] wqthrash_workfunc+0x61/0x90 [wqthrash] [<ffffffff8106dba8>] process_one_work+0x1e8/0x650 [<ffffffff8106e05e>] worker_thread+0x4e/0x450 [<ffffffff810746af>] kthread+0xef/0x110 [<ffffffff8185980f>] ret_from_fork+0x3f/0x70 Fix it by updating add_timer_on() to perform proper migration as __mod_timer() does. Reported-and-tested-by: Jeff Layton <jlayton@poochiereds.net> Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Chris Worley <chris.worley@primarydata.com> Cc: bfields@fieldses.org Cc: Michael Skralivetsky <michael.skralivetsky@primarydata.com> Cc: Trond Myklebust <trond.myklebust@primarydata.com> Cc: Shaohua Li <shli@fb.com> Cc: Jeff Layton <jlayton@poochiereds.net> Cc: kernel-team@fb.com Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20151029103113.2f893924@tlielax.poochiereds.net Link: http://lkml.kernel.org/r/20151104171533.GI5749@mtj.duckdns.org Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-11-04 20:15:33 +03:00
base = new_base;
raw_spin_lock(&base->lock);
timers: Use proper base migration in add_timer_on() Regardless of the previous CPU a timer was on, add_timer_on() currently simply sets timer->flags to the new CPU. As the caller must be seeing the timer as idle, this is locally fine, but the timer leaving the old base while unlocked can lead to race conditions as follows. Let's say timer was on cpu 0. cpu 0 cpu 1 ----------------------------------------------------------------------------- del_timer(timer) succeeds del_timer(timer) lock_timer_base(timer) locks cpu_0_base add_timer_on(timer, 1) spin_lock(&cpu_1_base->lock) timer->flags set to cpu_1_base operates on @timer operates on @timer This triggered with mod_delayed_work_on() which contains "if (del_timer()) add_timer_on()" sequence eventually leading to the following oops. BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<ffffffff810ca6e9>] detach_if_pending+0x69/0x1a0 ... Workqueue: wqthrash wqthrash_workfunc [wqthrash] task: ffff8800172ca680 ti: ffff8800172d0000 task.ti: ffff8800172d0000 RIP: 0010:[<ffffffff810ca6e9>] [<ffffffff810ca6e9>] detach_if_pending+0x69/0x1a0 ... Call Trace: [<ffffffff810cb0b4>] del_timer+0x44/0x60 [<ffffffff8106e836>] try_to_grab_pending+0xb6/0x160 [<ffffffff8106e913>] mod_delayed_work_on+0x33/0x80 [<ffffffffa0000081>] wqthrash_workfunc+0x61/0x90 [wqthrash] [<ffffffff8106dba8>] process_one_work+0x1e8/0x650 [<ffffffff8106e05e>] worker_thread+0x4e/0x450 [<ffffffff810746af>] kthread+0xef/0x110 [<ffffffff8185980f>] ret_from_fork+0x3f/0x70 Fix it by updating add_timer_on() to perform proper migration as __mod_timer() does. Reported-and-tested-by: Jeff Layton <jlayton@poochiereds.net> Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Chris Worley <chris.worley@primarydata.com> Cc: bfields@fieldses.org Cc: Michael Skralivetsky <michael.skralivetsky@primarydata.com> Cc: Trond Myklebust <trond.myklebust@primarydata.com> Cc: Shaohua Li <shli@fb.com> Cc: Jeff Layton <jlayton@poochiereds.net> Cc: kernel-team@fb.com Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20151029103113.2f893924@tlielax.poochiereds.net Link: http://lkml.kernel.org/r/20151104171533.GI5749@mtj.duckdns.org Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-11-04 20:15:33 +03:00
WRITE_ONCE(timer->flags,
(timer->flags & ~TIMER_BASEMASK) | cpu);
}
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
forward_timer_base(base);
timers: Use proper base migration in add_timer_on() Regardless of the previous CPU a timer was on, add_timer_on() currently simply sets timer->flags to the new CPU. As the caller must be seeing the timer as idle, this is locally fine, but the timer leaving the old base while unlocked can lead to race conditions as follows. Let's say timer was on cpu 0. cpu 0 cpu 1 ----------------------------------------------------------------------------- del_timer(timer) succeeds del_timer(timer) lock_timer_base(timer) locks cpu_0_base add_timer_on(timer, 1) spin_lock(&cpu_1_base->lock) timer->flags set to cpu_1_base operates on @timer operates on @timer This triggered with mod_delayed_work_on() which contains "if (del_timer()) add_timer_on()" sequence eventually leading to the following oops. BUG: unable to handle kernel NULL pointer dereference at (null) IP: [<ffffffff810ca6e9>] detach_if_pending+0x69/0x1a0 ... Workqueue: wqthrash wqthrash_workfunc [wqthrash] task: ffff8800172ca680 ti: ffff8800172d0000 task.ti: ffff8800172d0000 RIP: 0010:[<ffffffff810ca6e9>] [<ffffffff810ca6e9>] detach_if_pending+0x69/0x1a0 ... Call Trace: [<ffffffff810cb0b4>] del_timer+0x44/0x60 [<ffffffff8106e836>] try_to_grab_pending+0xb6/0x160 [<ffffffff8106e913>] mod_delayed_work_on+0x33/0x80 [<ffffffffa0000081>] wqthrash_workfunc+0x61/0x90 [wqthrash] [<ffffffff8106dba8>] process_one_work+0x1e8/0x650 [<ffffffff8106e05e>] worker_thread+0x4e/0x450 [<ffffffff810746af>] kthread+0xef/0x110 [<ffffffff8185980f>] ret_from_fork+0x3f/0x70 Fix it by updating add_timer_on() to perform proper migration as __mod_timer() does. Reported-and-tested-by: Jeff Layton <jlayton@poochiereds.net> Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Chris Worley <chris.worley@primarydata.com> Cc: bfields@fieldses.org Cc: Michael Skralivetsky <michael.skralivetsky@primarydata.com> Cc: Trond Myklebust <trond.myklebust@primarydata.com> Cc: Shaohua Li <shli@fb.com> Cc: Jeff Layton <jlayton@poochiereds.net> Cc: kernel-team@fb.com Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20151029103113.2f893924@tlielax.poochiereds.net Link: http://lkml.kernel.org/r/20151104171533.GI5749@mtj.duckdns.org Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-11-04 20:15:33 +03:00
debug_timer_activate(timer);
internal_add_timer(base, timer);
raw_spin_unlock_irqrestore(&base->lock, flags);
}
EXPORT_SYMBOL_GPL(add_timer_on);
/**
* del_timer - deactivate a timer.
* @timer: the timer to be deactivated
*
* del_timer() deactivates a timer - this works on both active and inactive
* timers.
*
* The function returns whether it has deactivated a pending timer or not.
* (ie. del_timer() of an inactive timer returns 0, del_timer() of an
* active timer returns 1.)
*/
int del_timer(struct timer_list *timer)
{
struct timer_base *base;
unsigned long flags;
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
int ret = 0;
debug_assert_init(timer);
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
if (timer_pending(timer)) {
base = lock_timer_base(timer, &flags);
ret = detach_if_pending(timer, base, true);
raw_spin_unlock_irqrestore(&base->lock, flags);
}
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
return ret;
}
EXPORT_SYMBOL(del_timer);
/**
* try_to_del_timer_sync - Try to deactivate a timer
* @timer: timer to delete
*
* This function tries to deactivate a timer. Upon successful (ret >= 0)
* exit the timer is not queued and the handler is not running on any CPU.
*/
int try_to_del_timer_sync(struct timer_list *timer)
{
struct timer_base *base;
unsigned long flags;
int ret = -1;
debug_assert_init(timer);
base = lock_timer_base(timer, &flags);
time: Remove CONFIG_TIMER_STATS Currently CONFIG_TIMER_STATS exposes process information across namespaces: kernel/time/timer_list.c print_timer(): SEQ_printf(m, ", %s/%d", tmp, timer->start_pid); /proc/timer_list: #11: <0000000000000000>, hrtimer_wakeup, S:01, do_nanosleep, cron/2570 Given that the tracer can give the same information, this patch entirely removes CONFIG_TIMER_STATS. Suggested-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Kees Cook <keescook@chromium.org> Acked-by: John Stultz <john.stultz@linaro.org> Cc: Nicolas Pitre <nicolas.pitre@linaro.org> Cc: linux-doc@vger.kernel.org Cc: Lai Jiangshan <jiangshanlai@gmail.com> Cc: Shuah Khan <shuah@kernel.org> Cc: Xing Gao <xgao01@email.wm.edu> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Jessica Frazelle <me@jessfraz.com> Cc: kernel-hardening@lists.openwall.com Cc: Nicolas Iooss <nicolas.iooss_linux@m4x.org> Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> Cc: Petr Mladek <pmladek@suse.com> Cc: Richard Cochran <richardcochran@gmail.com> Cc: Tejun Heo <tj@kernel.org> Cc: Michal Marek <mmarek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Olof Johansson <olof@lixom.net> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: linux-api@vger.kernel.org Cc: Arjan van de Ven <arjan@linux.intel.com> Link: http://lkml.kernel.org/r/20170208192659.GA32582@beast Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2017-02-08 22:26:59 +03:00
if (base->running_timer != timer)
ret = detach_if_pending(timer, base, true);
time: Remove CONFIG_TIMER_STATS Currently CONFIG_TIMER_STATS exposes process information across namespaces: kernel/time/timer_list.c print_timer(): SEQ_printf(m, ", %s/%d", tmp, timer->start_pid); /proc/timer_list: #11: <0000000000000000>, hrtimer_wakeup, S:01, do_nanosleep, cron/2570 Given that the tracer can give the same information, this patch entirely removes CONFIG_TIMER_STATS. Suggested-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Kees Cook <keescook@chromium.org> Acked-by: John Stultz <john.stultz@linaro.org> Cc: Nicolas Pitre <nicolas.pitre@linaro.org> Cc: linux-doc@vger.kernel.org Cc: Lai Jiangshan <jiangshanlai@gmail.com> Cc: Shuah Khan <shuah@kernel.org> Cc: Xing Gao <xgao01@email.wm.edu> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Jessica Frazelle <me@jessfraz.com> Cc: kernel-hardening@lists.openwall.com Cc: Nicolas Iooss <nicolas.iooss_linux@m4x.org> Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> Cc: Petr Mladek <pmladek@suse.com> Cc: Richard Cochran <richardcochran@gmail.com> Cc: Tejun Heo <tj@kernel.org> Cc: Michal Marek <mmarek@suse.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Dmitry Vyukov <dvyukov@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Olof Johansson <olof@lixom.net> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: linux-api@vger.kernel.org Cc: Arjan van de Ven <arjan@linux.intel.com> Link: http://lkml.kernel.org/r/20170208192659.GA32582@beast Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2017-02-08 22:26:59 +03:00
raw_spin_unlock_irqrestore(&base->lock, flags);
return ret;
}
EXPORT_SYMBOL(try_to_del_timer_sync);
timers: Prepare support for PREEMPT_RT When PREEMPT_RT is enabled, the soft interrupt thread can be preempted. If the soft interrupt thread is preempted in the middle of a timer callback, then calling del_timer_sync() can lead to two issues: - If the caller is on a remote CPU then it has to spin wait for the timer handler to complete. This can result in unbound priority inversion. - If the caller originates from the task which preempted the timer handler on the same CPU, then spin waiting for the timer handler to complete is never going to end. To avoid these issues, add a new lock to the timer base which is held around the execution of the timer callbacks. If del_timer_sync() detects that the timer callback is currently running, it blocks on the expiry lock. When the callback is finished, the expiry lock is dropped by the softirq thread which wakes up the waiter and the system makes progress. This addresses both the priority inversion and the life lock issues. This mechanism is not used for timers which are marked IRQSAFE as for those preemption is disabled accross the callback and therefore this situation cannot happen. The callbacks for such timers need to be individually audited for RT compliance. The same issue can happen in virtual machines when the vCPU which runs a timer callback is scheduled out. If a second vCPU of the same guest calls del_timer_sync() it will spin wait for the other vCPU to be scheduled back in. The expiry lock mechanism would avoid that. It'd be trivial to enable this when paravirt spinlocks are enabled in a guest, but it's not clear whether this is an actual problem in the wild, so for now it's an RT only mechanism. As the softirq thread can be preempted with PREEMPT_RT=y, the SMP variant of del_timer_sync() needs to be used on UP as well. [ tglx: Refactored it for mainline ] Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20190726185753.832418500@linutronix.de
2019-07-26 21:31:00 +03:00
#ifdef CONFIG_PREEMPT_RT
static __init void timer_base_init_expiry_lock(struct timer_base *base)
{
spin_lock_init(&base->expiry_lock);
}
static inline void timer_base_lock_expiry(struct timer_base *base)
{
spin_lock(&base->expiry_lock);
}
static inline void timer_base_unlock_expiry(struct timer_base *base)
{
spin_unlock(&base->expiry_lock);
}
/*
* The counterpart to del_timer_wait_running().
*
* If there is a waiter for base->expiry_lock, then it was waiting for the
* timer callback to finish. Drop expiry_lock and reaquire it. That allows
* the waiter to acquire the lock and make progress.
*/
static void timer_sync_wait_running(struct timer_base *base)
{
if (atomic_read(&base->timer_waiters)) {
spin_unlock(&base->expiry_lock);
spin_lock(&base->expiry_lock);
}
}
/*
* This function is called on PREEMPT_RT kernels when the fast path
* deletion of a timer failed because the timer callback function was
* running.
*
* This prevents priority inversion, if the softirq thread on a remote CPU
* got preempted, and it prevents a life lock when the task which tries to
* delete a timer preempted the softirq thread running the timer callback
* function.
*/
static void del_timer_wait_running(struct timer_list *timer)
{
u32 tf;
tf = READ_ONCE(timer->flags);
if (!(tf & TIMER_MIGRATING)) {
struct timer_base *base = get_timer_base(tf);
/*
* Mark the base as contended and grab the expiry lock,
* which is held by the softirq across the timer
* callback. Drop the lock immediately so the softirq can
* expire the next timer. In theory the timer could already
* be running again, but that's more than unlikely and just
* causes another wait loop.
*/
atomic_inc(&base->timer_waiters);
spin_lock_bh(&base->expiry_lock);
atomic_dec(&base->timer_waiters);
spin_unlock_bh(&base->expiry_lock);
}
}
#else
static inline void timer_base_init_expiry_lock(struct timer_base *base) { }
static inline void timer_base_lock_expiry(struct timer_base *base) { }
static inline void timer_base_unlock_expiry(struct timer_base *base) { }
static inline void timer_sync_wait_running(struct timer_base *base) { }
static inline void del_timer_wait_running(struct timer_list *timer) { }
#endif
#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT_RT)
/**
* del_timer_sync - deactivate a timer and wait for the handler to finish.
* @timer: the timer to be deactivated
*
* This function only differs from del_timer() on SMP: besides deactivating
* the timer it also makes sure the handler has finished executing on other
* CPUs.
*
* Synchronization rules: Callers must prevent restarting of the timer,
* otherwise this function is meaningless. It must not be called from
* interrupt contexts unless the timer is an irqsafe one. The caller must
* not hold locks which would prevent completion of the timer's
* handler. The timer's handler must not call add_timer_on(). Upon exit the
* timer is not queued and the handler is not running on any CPU.
*
* Note: For !irqsafe timers, you must not hold locks that are held in
* interrupt context while calling this function. Even if the lock has
* nothing to do with the timer in question. Here's why::
*
* CPU0 CPU1
* ---- ----
* <SOFTIRQ>
* call_timer_fn();
* base->running_timer = mytimer;
* spin_lock_irq(somelock);
* <IRQ>
* spin_lock(somelock);
* del_timer_sync(mytimer);
* while (base->running_timer == mytimer);
*
* Now del_timer_sync() will never return and never release somelock.
* The interrupt on the other CPU is waiting to grab somelock but
* it has interrupted the softirq that CPU0 is waiting to finish.
*
* The function returns whether it has deactivated a pending timer or not.
*/
int del_timer_sync(struct timer_list *timer)
{
timers: Prepare support for PREEMPT_RT When PREEMPT_RT is enabled, the soft interrupt thread can be preempted. If the soft interrupt thread is preempted in the middle of a timer callback, then calling del_timer_sync() can lead to two issues: - If the caller is on a remote CPU then it has to spin wait for the timer handler to complete. This can result in unbound priority inversion. - If the caller originates from the task which preempted the timer handler on the same CPU, then spin waiting for the timer handler to complete is never going to end. To avoid these issues, add a new lock to the timer base which is held around the execution of the timer callbacks. If del_timer_sync() detects that the timer callback is currently running, it blocks on the expiry lock. When the callback is finished, the expiry lock is dropped by the softirq thread which wakes up the waiter and the system makes progress. This addresses both the priority inversion and the life lock issues. This mechanism is not used for timers which are marked IRQSAFE as for those preemption is disabled accross the callback and therefore this situation cannot happen. The callbacks for such timers need to be individually audited for RT compliance. The same issue can happen in virtual machines when the vCPU which runs a timer callback is scheduled out. If a second vCPU of the same guest calls del_timer_sync() it will spin wait for the other vCPU to be scheduled back in. The expiry lock mechanism would avoid that. It'd be trivial to enable this when paravirt spinlocks are enabled in a guest, but it's not clear whether this is an actual problem in the wild, so for now it's an RT only mechanism. As the softirq thread can be preempted with PREEMPT_RT=y, the SMP variant of del_timer_sync() needs to be used on UP as well. [ tglx: Refactored it for mainline ] Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20190726185753.832418500@linutronix.de
2019-07-26 21:31:00 +03:00
int ret;
#ifdef CONFIG_LOCKDEP
unsigned long flags;
/*
* If lockdep gives a backtrace here, please reference
* the synchronization rules above.
*/
local_irq_save(flags);
lock_map_acquire(&timer->lockdep_map);
lock_map_release(&timer->lockdep_map);
local_irq_restore(flags);
#endif
/*
* don't use it in hardirq context, because it
* could lead to deadlock.
*/
timer: Replace timer base by a cpu index Instead of storing a pointer to the per cpu tvec_base we can simply cache a CPU index in the timer_list and use that to get hold of the correct per cpu tvec_base. This is only used in lock_timer_base() and the slightly larger code is peanuts versus the spinlock operation and the d-cache foot print of the timer wheel. Aside of that this allows to get rid of following nuisances: - boot_tvec_base That statically allocated 4k bss data is just kept around so the timer has a home when it gets statically initialized. It serves no other purpose. With the CPU index we assign the timer to CPU0 at static initialization time and therefor can avoid the whole boot_tvec_base dance. That also simplifies the init code, which just can use the per cpu base. Before: text data bss dec hex filename 17491 9201 4160 30852 7884 ../build/kernel/time/timer.o After: text data bss dec hex filename 17440 9193 0 26633 6809 ../build/kernel/time/timer.o - Overloading the base pointer with various flags The CPU index has enough space to hold the flags (deferrable, irqsafe) so we can get rid of the extra masking and bit fiddling with the base pointer. As a benefit we reduce the size of struct timer_list on 64 bit machines. 4 - 8 bytes, a size reduction up to 15% per struct timer_list, which is a real win as we have tons of them embedded in other structs. This changes also the newly added deferrable printout of the timer start trace point to capture and print all timer->flags, which allows us to decode the target cpu of the timer as well. We might have used bitfields for this, but that would change the static initializers and the init function for no value to accomodate big endian bitfields. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Badhri Jagan Sridharan <Badhri@google.com> Link: http://lkml.kernel.org/r/20150526224511.950084301@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:29 +03:00
WARN_ON(in_irq() && !(timer->flags & TIMER_IRQSAFE));
timers: Prepare support for PREEMPT_RT When PREEMPT_RT is enabled, the soft interrupt thread can be preempted. If the soft interrupt thread is preempted in the middle of a timer callback, then calling del_timer_sync() can lead to two issues: - If the caller is on a remote CPU then it has to spin wait for the timer handler to complete. This can result in unbound priority inversion. - If the caller originates from the task which preempted the timer handler on the same CPU, then spin waiting for the timer handler to complete is never going to end. To avoid these issues, add a new lock to the timer base which is held around the execution of the timer callbacks. If del_timer_sync() detects that the timer callback is currently running, it blocks on the expiry lock. When the callback is finished, the expiry lock is dropped by the softirq thread which wakes up the waiter and the system makes progress. This addresses both the priority inversion and the life lock issues. This mechanism is not used for timers which are marked IRQSAFE as for those preemption is disabled accross the callback and therefore this situation cannot happen. The callbacks for such timers need to be individually audited for RT compliance. The same issue can happen in virtual machines when the vCPU which runs a timer callback is scheduled out. If a second vCPU of the same guest calls del_timer_sync() it will spin wait for the other vCPU to be scheduled back in. The expiry lock mechanism would avoid that. It'd be trivial to enable this when paravirt spinlocks are enabled in a guest, but it's not clear whether this is an actual problem in the wild, so for now it's an RT only mechanism. As the softirq thread can be preempted with PREEMPT_RT=y, the SMP variant of del_timer_sync() needs to be used on UP as well. [ tglx: Refactored it for mainline ] Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20190726185753.832418500@linutronix.de
2019-07-26 21:31:00 +03:00
do {
ret = try_to_del_timer_sync(timer);
if (unlikely(ret < 0)) {
del_timer_wait_running(timer);
cpu_relax();
}
} while (ret < 0);
return ret;
}
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
EXPORT_SYMBOL(del_timer_sync);
#endif
static void call_timer_fn(struct timer_list *timer,
void (*fn)(struct timer_list *),
unsigned long baseclk)
{
int count = preempt_count();
#ifdef CONFIG_LOCKDEP
/*
* It is permissible to free the timer from inside the
* function that is called from it, this we need to take into
* account for lockdep too. To avoid bogus "held lock freed"
* warnings as well as problems when looking into
* timer->lockdep_map, make a copy and use that here.
*/
lockdep: fix oops in processing workqueue Under memory load, on x86_64, with lockdep enabled, the workqueue's process_one_work() has been seen to oops in __lock_acquire(), barfing on a 0xffffffff00000000 pointer in the lockdep_map's class_cache[]. Because it's permissible to free a work_struct from its callout function, the map used is an onstack copy of the map given in the work_struct: and that copy is made without any locking. Surprisingly, gcc (4.5.1 in Hugh's case) uses "rep movsl" rather than "rep movsq" for that structure copy: which might race with a workqueue user's wait_on_work() doing lock_map_acquire() on the source of the copy, putting a pointer into the class_cache[], but only in time for the top half of that pointer to be copied to the destination map. Boom when process_one_work() subsequently does lock_map_acquire() on its onstack copy of the lockdep_map. Fix this, and a similar instance in call_timer_fn(), with a lockdep_copy_map() function which additionally NULLs the class_cache[]. Note: this oops was actually seen on 3.4-next, where flush_work() newly does the racing lock_map_acquire(); but Tejun points out that 3.4 and earlier are already vulnerable to the same through wait_on_work(). * Patch orginally from Peter. Hugh modified it a bit and wrote the description. Signed-off-by: Peter Zijlstra <peterz@infradead.org> Reported-by: Hugh Dickins <hughd@google.com> LKML-Reference: <alpine.LSU.2.00.1205070951170.1544@eggly.anvils> Signed-off-by: Tejun Heo <tj@kernel.org>
2012-05-15 19:06:19 +04:00
struct lockdep_map lockdep_map;
lockdep_copy_map(&lockdep_map, &timer->lockdep_map);
#endif
/*
* Couple the lock chain with the lock chain at
* del_timer_sync() by acquiring the lock_map around the fn()
* call here and in del_timer_sync().
*/
lock_map_acquire(&lockdep_map);
trace_timer_expire_entry(timer, baseclk);
fn(timer);
trace_timer_expire_exit(timer);
lock_map_release(&lockdep_map);
if (count != preempt_count()) {
2019-03-25 22:32:28 +03:00
WARN_ONCE(1, "timer: %pS preempt leak: %08x -> %08x\n",
fn, count, preempt_count());
/*
* Restore the preempt count. That gives us a decent
* chance to survive and extract information. If the
* callback kept a lock held, bad luck, but not worse
* than the BUG() we had.
*/
preempt_count_set(count);
}
}
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
static void expire_timers(struct timer_base *base, struct hlist_head *head)
{
/*
* This value is required only for tracing. base->clk was
* incremented directly before expire_timers was called. But expiry
* is related to the old base->clk value.
*/
unsigned long baseclk = base->clk - 1;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
while (!hlist_empty(head)) {
struct timer_list *timer;
void (*fn)(struct timer_list *);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
timer = hlist_entry(head->first, struct timer_list, entry);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
base->running_timer = timer;
detach_timer(timer, true);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
fn = timer->function;
if (timer->flags & TIMER_IRQSAFE) {
raw_spin_unlock(&base->lock);
call_timer_fn(timer, fn, baseclk);
timers: Prepare support for PREEMPT_RT When PREEMPT_RT is enabled, the soft interrupt thread can be preempted. If the soft interrupt thread is preempted in the middle of a timer callback, then calling del_timer_sync() can lead to two issues: - If the caller is on a remote CPU then it has to spin wait for the timer handler to complete. This can result in unbound priority inversion. - If the caller originates from the task which preempted the timer handler on the same CPU, then spin waiting for the timer handler to complete is never going to end. To avoid these issues, add a new lock to the timer base which is held around the execution of the timer callbacks. If del_timer_sync() detects that the timer callback is currently running, it blocks on the expiry lock. When the callback is finished, the expiry lock is dropped by the softirq thread which wakes up the waiter and the system makes progress. This addresses both the priority inversion and the life lock issues. This mechanism is not used for timers which are marked IRQSAFE as for those preemption is disabled accross the callback and therefore this situation cannot happen. The callbacks for such timers need to be individually audited for RT compliance. The same issue can happen in virtual machines when the vCPU which runs a timer callback is scheduled out. If a second vCPU of the same guest calls del_timer_sync() it will spin wait for the other vCPU to be scheduled back in. The expiry lock mechanism would avoid that. It'd be trivial to enable this when paravirt spinlocks are enabled in a guest, but it's not clear whether this is an actual problem in the wild, so for now it's an RT only mechanism. As the softirq thread can be preempted with PREEMPT_RT=y, the SMP variant of del_timer_sync() needs to be used on UP as well. [ tglx: Refactored it for mainline ] Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20190726185753.832418500@linutronix.de
2019-07-26 21:31:00 +03:00
base->running_timer = NULL;
raw_spin_lock(&base->lock);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
} else {
raw_spin_unlock_irq(&base->lock);
call_timer_fn(timer, fn, baseclk);
timers: Prepare support for PREEMPT_RT When PREEMPT_RT is enabled, the soft interrupt thread can be preempted. If the soft interrupt thread is preempted in the middle of a timer callback, then calling del_timer_sync() can lead to two issues: - If the caller is on a remote CPU then it has to spin wait for the timer handler to complete. This can result in unbound priority inversion. - If the caller originates from the task which preempted the timer handler on the same CPU, then spin waiting for the timer handler to complete is never going to end. To avoid these issues, add a new lock to the timer base which is held around the execution of the timer callbacks. If del_timer_sync() detects that the timer callback is currently running, it blocks on the expiry lock. When the callback is finished, the expiry lock is dropped by the softirq thread which wakes up the waiter and the system makes progress. This addresses both the priority inversion and the life lock issues. This mechanism is not used for timers which are marked IRQSAFE as for those preemption is disabled accross the callback and therefore this situation cannot happen. The callbacks for such timers need to be individually audited for RT compliance. The same issue can happen in virtual machines when the vCPU which runs a timer callback is scheduled out. If a second vCPU of the same guest calls del_timer_sync() it will spin wait for the other vCPU to be scheduled back in. The expiry lock mechanism would avoid that. It'd be trivial to enable this when paravirt spinlocks are enabled in a guest, but it's not clear whether this is an actual problem in the wild, so for now it's an RT only mechanism. As the softirq thread can be preempted with PREEMPT_RT=y, the SMP variant of del_timer_sync() needs to be used on UP as well. [ tglx: Refactored it for mainline ] Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20190726185753.832418500@linutronix.de
2019-07-26 21:31:00 +03:00
base->running_timer = NULL;
timer_sync_wait_running(base);
raw_spin_lock_irq(&base->lock);
}
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
}
}
static int __collect_expired_timers(struct timer_base *base,
struct hlist_head *heads)
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
{
unsigned long clk = base->clk;
struct hlist_head *vec;
int i, levels = 0;
unsigned int idx;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
for (i = 0; i < LVL_DEPTH; i++) {
idx = (clk & LVL_MASK) + i * LVL_SIZE;
if (__test_and_clear_bit(idx, base->pending_map)) {
vec = base->vectors + idx;
hlist_move_list(vec, heads++);
levels++;
}
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
/* Is it time to look at the next level? */
if (clk & LVL_CLK_MASK)
break;
/* Shift clock for the next level granularity */
clk >>= LVL_CLK_SHIFT;
}
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
return levels;
}
nohz: Rename CONFIG_NO_HZ to CONFIG_NO_HZ_COMMON We are planning to convert the dynticks Kconfig options layout into a choice menu. The user must be able to easily pick any of the following implementations: constant periodic tick, idle dynticks, full dynticks. As this implies a mutual exclusion, the two dynticks implementions need to converge on the selection of a common Kconfig option in order to ease the sharing of a common infrastructure. It would thus seem pretty natural to reuse CONFIG_NO_HZ to that end. It already implements all the idle dynticks code and the full dynticks depends on all that code for now. So ideally the choice menu would propose CONFIG_NO_HZ_IDLE and CONFIG_NO_HZ_EXTENDED then both would select CONFIG_NO_HZ. On the other hand we want to stay backward compatible: if CONFIG_NO_HZ is set in an older config file, we want to enable CONFIG_NO_HZ_IDLE by default. But we can't afford both at the same time or we run into a circular dependency: 1) CONFIG_NO_HZ_IDLE and CONFIG_NO_HZ_EXTENDED both select CONFIG_NO_HZ 2) If CONFIG_NO_HZ is set, we default to CONFIG_NO_HZ_IDLE We might be able to support that from Kconfig/Kbuild but it may not be wise to introduce such a confusing behaviour. So to solve this, create a new CONFIG_NO_HZ_COMMON option which gathers the common code between idle and full dynticks (that common code for now is simply the idle dynticks code) and select it from their referring Kconfig. Then we'll later create CONFIG_NO_HZ_IDLE and map CONFIG_NO_HZ to it for backward compatibility. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Christoph Lameter <cl@linux.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Gilad Ben Yossef <gilad@benyossef.com> Cc: Hakan Akkan <hakanakkan@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Kevin Hilman <khilman@linaro.org> Cc: Li Zhong <zhong@linux.vnet.ibm.com> Cc: Namhyung Kim <namhyung.kim@lge.com> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Thomas Gleixner <tglx@linutronix.de>
2011-08-11 01:21:01 +04:00
#ifdef CONFIG_NO_HZ_COMMON
/*
* Find the next pending bucket of a level. Search from level start (@offset)
* + @clk upwards and if nothing there, search from start of the level
* (@offset) up to @offset + clk.
*/
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
static int next_pending_bucket(struct timer_base *base, unsigned offset,
unsigned clk)
{
unsigned pos, start = offset + clk;
unsigned end = offset + LVL_SIZE;
pos = find_next_bit(base->pending_map, end, start);
if (pos < end)
return pos - start;
pos = find_next_bit(base->pending_map, start, offset);
return pos < start ? pos + LVL_SIZE - start : -1;
}
/*
* Search the first expiring timer in the various clock levels. Caller must
* hold base->lock.
*/
static unsigned long __next_timer_interrupt(struct timer_base *base)
{
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
unsigned long clk, next, adj;
unsigned lvl, offset = 0;
next = base->clk + NEXT_TIMER_MAX_DELTA;
clk = base->clk;
for (lvl = 0; lvl < LVL_DEPTH; lvl++, offset += LVL_SIZE) {
int pos = next_pending_bucket(base, offset, clk & LVL_MASK);
if (pos >= 0) {
unsigned long tmp = clk + (unsigned long) pos;
tmp <<= LVL_SHIFT(lvl);
if (time_before(tmp, next))
next = tmp;
}
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
/*
* Clock for the next level. If the current level clock lower
* bits are zero, we look at the next level as is. If not we
* need to advance it by one because that's going to be the
* next expiring bucket in that level. base->clk is the next
* expiring jiffie. So in case of:
*
* LVL5 LVL4 LVL3 LVL2 LVL1 LVL0
* 0 0 0 0 0 0
*
* we have to look at all levels @index 0. With
*
* LVL5 LVL4 LVL3 LVL2 LVL1 LVL0
* 0 0 0 0 0 2
*
* LVL0 has the next expiring bucket @index 2. The upper
* levels have the next expiring bucket @index 1.
*
* In case that the propagation wraps the next level the same
* rules apply:
*
* LVL5 LVL4 LVL3 LVL2 LVL1 LVL0
* 0 0 0 0 F 2
*
* So after looking at LVL0 we get:
*
* LVL5 LVL4 LVL3 LVL2 LVL1
* 0 0 0 1 0
*
* So no propagation from LVL1 to LVL2 because that happened
* with the add already, but then we need to propagate further
* from LVL2 to LVL3.
*
* So the simple check whether the lower bits of the current
* level are 0 or not is sufficient for all cases.
*/
adj = clk & LVL_CLK_MASK ? 1 : 0;
clk >>= LVL_CLK_SHIFT;
clk += adj;
}
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
return next;
}
/*
* Check, if the next hrtimer event is before the next timer wheel
* event:
*/
static u64 cmp_next_hrtimer_event(u64 basem, u64 expires)
{
u64 nextevt = hrtimer_get_next_event();
/*
* If high resolution timers are enabled
* hrtimer_get_next_event() returns KTIME_MAX.
*/
if (expires <= nextevt)
return expires;
/*
* If the next timer is already expired, return the tick base
* time so the tick is fired immediately.
*/
if (nextevt <= basem)
return basem;
/*
* Round up to the next jiffie. High resolution timers are
* off, so the hrtimers are expired in the tick and we need to
* make sure that this tick really expires the timer to avoid
* a ping pong of the nohz stop code.
*
* Use DIV_ROUND_UP_ULL to prevent gcc calling __divdi3
*/
return DIV_ROUND_UP_ULL(nextevt, TICK_NSEC) * TICK_NSEC;
}
/**
* get_next_timer_interrupt - return the time (clock mono) of the next timer
* @basej: base time jiffies
* @basem: base time clock monotonic
*
* Returns the tick aligned clock monotonic time of the next pending
* timer or KTIME_MAX if no timer is pending.
*/
u64 get_next_timer_interrupt(unsigned long basej, u64 basem)
{
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
u64 expires = KTIME_MAX;
unsigned long nextevt;
timers: Fix get_next_timer_interrupt() computation The tick_nohz_stop_sched_tick() routine is not properly canceling the sched timer when nothing is pending, because get_next_timer_interrupt() is no longer returning KTIME_MAX in that case. This causes periodic interrupts when none are needed. When determining the next interrupt time, we first use __next_timer_interrupt() to get the first expiring timer in the timer wheel. If no timer is found, we return the base clock value plus NEXT_TIMER_MAX_DELTA to indicate there is no timer in the timer wheel. Back in get_next_timer_interrupt(), we set the "expires" value by converting the timer wheel expiry (in ticks) to a nsec value. But we don't want to do this if the timer wheel expiry value indicates no timer; we want to return KTIME_MAX. Prior to commit 500462a9de65 ("timers: Switch to a non-cascading wheel") we checked base->active_timers to see if any timers were active, and if not, we didn't touch the expiry value and so properly returned KTIME_MAX. Now we don't have active_timers. To fix this, we now just check the timer wheel expiry value to see if it is "now + NEXT_TIMER_MAX_DELTA", and if it is, we don't try to compute a new value based on it, but instead simply let the KTIME_MAX value in expires remain. Fixes: 500462a9de65 "timers: Switch to a non-cascading wheel" Signed-off-by: Chris Metcalf <cmetcalf@mellanox.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Christoph Lameter <cl@linux.com> Cc: John Stultz <john.stultz@linaro.org> Link: http://lkml.kernel.org/r/1470688147-22287-1-git-send-email-cmetcalf@mellanox.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2016-08-08 23:29:07 +03:00
bool is_max_delta;
nohz: Fix get_next_timer_interrupt() vs cpu hotplug This fixes a bug as seen on 2.6.32 based kernels where timers got enqueued on offline cpus. If a cpu goes offline it might still have pending timers. These will be migrated during CPU_DEAD handling after the cpu is offline. However while the cpu is going offline it will schedule the idle task which will then call tick_nohz_stop_sched_tick(). That function in turn will call get_next_timer_intterupt() to figure out if the tick of the cpu can be stopped or not. If it turns out that the next tick is just one jiffy off (delta_jiffies == 1) tick_nohz_stop_sched_tick() incorrectly assumes that the tick should not stop and takes an early exit and thus it won't update the load balancer cpu. Just afterwards the cpu will be killed and the load balancer cpu could be the offline cpu. On 2.6.32 based kernel get_nohz_load_balancer() gets called to decide on which cpu a timer should be enqueued (see __mod_timer()). Which leads to the possibility that timers get enqueued on an offline cpu. These will never expire and can cause a system hang. This has been observed 2.6.32 kernels. On current kernels __mod_timer() uses get_nohz_timer_target() which doesn't have that problem. However there might be other problems because of the too early exit tick_nohz_stop_sched_tick() in case a cpu goes offline. The easiest and probably safest fix seems to be to let get_next_timer_interrupt() just lie and let it say there isn't any pending timer if the current cpu is offline. I also thought of moving migrate_[hr]timers() from CPU_DEAD to CPU_DYING, but seeing that there already have been fixes at least in the hrtimer code in this area I'm afraid that this could add new subtle bugs. Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <20101201091109.GA8984@osiris.boeblingen.de.ibm.com> Cc: stable@kernel.org Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-12-01 12:11:09 +03:00
/*
* Pretend that there is no timer pending if the cpu is offline.
* Possible pending timers will be migrated later to an active cpu.
*/
if (cpu_is_offline(smp_processor_id()))
timers: Improve get_next_timer_interrupt() Gilad reported at http://lkml.kernel.org/r/1336056962-10465-2-git-send-email-gilad@benyossef.com "Current timer code fails to correctly return a value meaning that there is no future timer event, with the result that the timer keeps getting re-armed in HZ one shot mode even when we could turn it off, generating unneeded interrupts. What is happening is that when __next_timer_interrupt() wishes to return a value that signifies "there is no future timer event", it returns (base->timer_jiffies + NEXT_TIMER_MAX_DELTA). However, the code in tick_nohz_stop_sched_tick(), which called __next_timer_interrupt() via get_next_timer_interrupt(), compares the return value to (last_jiffies + NEXT_TIMER_MAX_DELTA) to see if the timer needs to be re-armed. base->timer_jiffies != last_jiffies and so tick_nohz_stop_sched_tick() interperts the return value as indication that there is a distant future event 12 days from now and programs the timer to fire next after KTIME_MAX nsecs instead of avoiding to arm it. This ends up causing a needless interrupt once every KTIME_MAX nsecs." Fix this by using the new active timer accounting. This avoids scans when no active timer is enqueued completely, so we don't have to rely on base->timer_next and base->timer_jiffies anymore. Reported-by: Gilad Ben-Yossef <gilad@benyossef.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Link: http://lkml.kernel.org/r/20120525214819.317535385@linutronix.de
2012-05-26 02:08:59 +04:00
return expires;
raw_spin_lock(&base->lock);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
nextevt = __next_timer_interrupt(base);
timers: Fix get_next_timer_interrupt() computation The tick_nohz_stop_sched_tick() routine is not properly canceling the sched timer when nothing is pending, because get_next_timer_interrupt() is no longer returning KTIME_MAX in that case. This causes periodic interrupts when none are needed. When determining the next interrupt time, we first use __next_timer_interrupt() to get the first expiring timer in the timer wheel. If no timer is found, we return the base clock value plus NEXT_TIMER_MAX_DELTA to indicate there is no timer in the timer wheel. Back in get_next_timer_interrupt(), we set the "expires" value by converting the timer wheel expiry (in ticks) to a nsec value. But we don't want to do this if the timer wheel expiry value indicates no timer; we want to return KTIME_MAX. Prior to commit 500462a9de65 ("timers: Switch to a non-cascading wheel") we checked base->active_timers to see if any timers were active, and if not, we didn't touch the expiry value and so properly returned KTIME_MAX. Now we don't have active_timers. To fix this, we now just check the timer wheel expiry value to see if it is "now + NEXT_TIMER_MAX_DELTA", and if it is, we don't try to compute a new value based on it, but instead simply let the KTIME_MAX value in expires remain. Fixes: 500462a9de65 "timers: Switch to a non-cascading wheel" Signed-off-by: Chris Metcalf <cmetcalf@mellanox.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Christoph Lameter <cl@linux.com> Cc: John Stultz <john.stultz@linaro.org> Link: http://lkml.kernel.org/r/1470688147-22287-1-git-send-email-cmetcalf@mellanox.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2016-08-08 23:29:07 +03:00
is_max_delta = (nextevt == base->clk + NEXT_TIMER_MAX_DELTA);
base->next_expiry = nextevt;
/*
timers: Prevent base clock rewind when forwarding clock Ashton and Michael reported, that kernel versions 4.8 and later suffer from USB timeouts which are caused by the timer wheel rework. This is caused by a bug in the base clock forwarding mechanism, which leads to timers expiring early. The scenario which leads to this is: run_timers() while (jiffies >= base->clk) { collect_expired_timers(); base->clk++; expire_timers(); } So base->clk = jiffies + 1. Now the cpu goes idle: idle() get_next_timer_interrupt() nextevt = __next_time_interrupt(); if (time_after(nextevt, base->clk)) base->clk = jiffies; jiffies has not advanced since run_timers(), so this assignment effectively decrements base->clk by one. base->clk is the index into the timer wheel arrays. So let's assume the following state after the base->clk increment in run_timers(): jiffies = 0 base->clk = 1 A timer gets enqueued with an expiry delta of 63 ticks (which is the case with the USB timeout and HZ=250) so the resulting bucket index is: base->clk + delta = 1 + 63 = 64 The timer goes into the first wheel level. The array size is 64 so it ends up in bucket 0, which is correct as it takes 63 ticks to advance base->clk to index into bucket 0 again. If the cpu goes idle before jiffies advance, then the bug in the forwarding mechanism sets base->clk back to 0, so the next invocation of run_timers() at the next tick will index into bucket 0 and therefore expire the timer 62 ticks too early. Instead of blindly setting base->clk to jiffies we must make the forwarding conditional on jiffies > base->clk, but we cannot use jiffies for this as we might run into the following issue: if (time_after(jiffies, base->clk) { if (time_after(nextevt, base->clk)) base->clk = jiffies; jiffies can increment between the check and the assigment far enough to advance beyond nextevt. So we need to use a stable value for checking. get_next_timer_interrupt() has the basej argument which is the jiffies value snapshot taken in the calling code. So we can just that. Thanks to Ashton for bisecting and providing trace data! Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Reported-by: Ashton Holmes <scoopta@gmail.com> Reported-by: Michael Thayer <michael.thayer@oracle.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Michal Necasek <michal.necasek@oracle.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: knut.osmundsen@oracle.com Cc: stable@vger.kernel.org Cc: stern@rowland.harvard.edu Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20161022110552.175308322@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2016-10-22 14:07:35 +03:00
* We have a fresh next event. Check whether we can forward the
* base. We can only do that when @basej is past base->clk
* otherwise we might rewind base->clk.
*/
timers: Prevent base clock rewind when forwarding clock Ashton and Michael reported, that kernel versions 4.8 and later suffer from USB timeouts which are caused by the timer wheel rework. This is caused by a bug in the base clock forwarding mechanism, which leads to timers expiring early. The scenario which leads to this is: run_timers() while (jiffies >= base->clk) { collect_expired_timers(); base->clk++; expire_timers(); } So base->clk = jiffies + 1. Now the cpu goes idle: idle() get_next_timer_interrupt() nextevt = __next_time_interrupt(); if (time_after(nextevt, base->clk)) base->clk = jiffies; jiffies has not advanced since run_timers(), so this assignment effectively decrements base->clk by one. base->clk is the index into the timer wheel arrays. So let's assume the following state after the base->clk increment in run_timers(): jiffies = 0 base->clk = 1 A timer gets enqueued with an expiry delta of 63 ticks (which is the case with the USB timeout and HZ=250) so the resulting bucket index is: base->clk + delta = 1 + 63 = 64 The timer goes into the first wheel level. The array size is 64 so it ends up in bucket 0, which is correct as it takes 63 ticks to advance base->clk to index into bucket 0 again. If the cpu goes idle before jiffies advance, then the bug in the forwarding mechanism sets base->clk back to 0, so the next invocation of run_timers() at the next tick will index into bucket 0 and therefore expire the timer 62 ticks too early. Instead of blindly setting base->clk to jiffies we must make the forwarding conditional on jiffies > base->clk, but we cannot use jiffies for this as we might run into the following issue: if (time_after(jiffies, base->clk) { if (time_after(nextevt, base->clk)) base->clk = jiffies; jiffies can increment between the check and the assigment far enough to advance beyond nextevt. So we need to use a stable value for checking. get_next_timer_interrupt() has the basej argument which is the jiffies value snapshot taken in the calling code. So we can just that. Thanks to Ashton for bisecting and providing trace data! Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Reported-by: Ashton Holmes <scoopta@gmail.com> Reported-by: Michael Thayer <michael.thayer@oracle.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Michal Necasek <michal.necasek@oracle.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: knut.osmundsen@oracle.com Cc: stable@vger.kernel.org Cc: stern@rowland.harvard.edu Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20161022110552.175308322@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2016-10-22 14:07:35 +03:00
if (time_after(basej, base->clk)) {
if (time_after(nextevt, basej))
base->clk = basej;
else if (time_after(nextevt, base->clk))
base->clk = nextevt;
}
if (time_before_eq(nextevt, basej)) {
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
expires = basem;
base->is_idle = false;
} else {
timers: Fix get_next_timer_interrupt() computation The tick_nohz_stop_sched_tick() routine is not properly canceling the sched timer when nothing is pending, because get_next_timer_interrupt() is no longer returning KTIME_MAX in that case. This causes periodic interrupts when none are needed. When determining the next interrupt time, we first use __next_timer_interrupt() to get the first expiring timer in the timer wheel. If no timer is found, we return the base clock value plus NEXT_TIMER_MAX_DELTA to indicate there is no timer in the timer wheel. Back in get_next_timer_interrupt(), we set the "expires" value by converting the timer wheel expiry (in ticks) to a nsec value. But we don't want to do this if the timer wheel expiry value indicates no timer; we want to return KTIME_MAX. Prior to commit 500462a9de65 ("timers: Switch to a non-cascading wheel") we checked base->active_timers to see if any timers were active, and if not, we didn't touch the expiry value and so properly returned KTIME_MAX. Now we don't have active_timers. To fix this, we now just check the timer wheel expiry value to see if it is "now + NEXT_TIMER_MAX_DELTA", and if it is, we don't try to compute a new value based on it, but instead simply let the KTIME_MAX value in expires remain. Fixes: 500462a9de65 "timers: Switch to a non-cascading wheel" Signed-off-by: Chris Metcalf <cmetcalf@mellanox.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Christoph Lameter <cl@linux.com> Cc: John Stultz <john.stultz@linaro.org> Link: http://lkml.kernel.org/r/1470688147-22287-1-git-send-email-cmetcalf@mellanox.com Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2016-08-08 23:29:07 +03:00
if (!is_max_delta)
expires = basem + (u64)(nextevt - basej) * TICK_NSEC;
/*
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
* If we expect to sleep more than a tick, mark the base idle.
* Also the tick is stopped so any added timer must forward
* the base clk itself to keep granularity small. This idle
* logic is only maintained for the BASE_STD base, deferrable
* timers may still see large granularity skew (by design).
*/
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
if ((expires - basem) > TICK_NSEC) {
base->must_forward_clk = true;
base->is_idle = true;
timers: Fix excessive granularity of new timers after a nohz idle When a timer base is idle, it is forwarded when a new timer is added to ensure that granularity does not become excessive. When not idle, the timer tick is expected to increment the base. However there are several problems: - If an existing timer is modified, the base is forwarded only after the index is calculated. - The base is not forwarded by add_timer_on. - There is a window after a timer is restarted from a nohz idle, after it is marked not-idle and before the timer tick on this CPU, where a timer may be added but the ancient base does not get forwarded. These result in excessive granularity (a 1 jiffy timeout can blow out to 100s of jiffies), which cause the rcu lockup detector to trigger, among other things. Fix this by keeping track of whether the timer base has been idle since it was last run or forwarded, and if so then forward it before adding a new timer. There is still a case where mod_timer optimises the case of a pending timer mod with the same expiry time, where the timer can see excessive granularity relative to the new, shorter interval. A comment is added, but it's not changed because it is an important fastpath for networking. This has been tested and found to fix the RCU softlockup messages. Testing was also done with tracing to measure requested versus achieved wakeup latencies for all non-deferrable timers in an idle system (with no lockup watchdogs running). Wakeup latency relative to absolute latency is calculated (note this suffers from round-up skew at low absolute times) and analysed: max avg std upstream 506.0 1.20 4.68 patched 2.0 1.08 0.15 The bug was noticed due to the lockup detector Kconfig changes dropping it out of people's .configs and resulting in larger base clk skew When the lockup detectors are enabled, no CPU can go idle for longer than 4 seconds, which limits the granularity errors. Sub-optimal timer behaviour is observable on a smaller scale in that case: max avg std upstream 9.0 1.05 0.19 patched 2.0 1.04 0.11 Fixes: Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Signed-off-by: Nicholas Piggin <npiggin@gmail.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Tested-by: David Miller <davem@davemloft.net> Cc: dzickus@redhat.com Cc: sfr@canb.auug.org.au Cc: mpe@ellerman.id.au Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: linuxarm@huawei.com Cc: abdhalee@linux.vnet.ibm.com Cc: John Stultz <john.stultz@linaro.org> Cc: akpm@linux-foundation.org Cc: paulmck@linux.vnet.ibm.com Cc: torvalds@linux-foundation.org Cc: stable@vger.kernel.org Link: http://lkml.kernel.org/r/20170822084348.21436-1-npiggin@gmail.com
2017-08-22 11:43:48 +03:00
}
timers: Improve get_next_timer_interrupt() Gilad reported at http://lkml.kernel.org/r/1336056962-10465-2-git-send-email-gilad@benyossef.com "Current timer code fails to correctly return a value meaning that there is no future timer event, with the result that the timer keeps getting re-armed in HZ one shot mode even when we could turn it off, generating unneeded interrupts. What is happening is that when __next_timer_interrupt() wishes to return a value that signifies "there is no future timer event", it returns (base->timer_jiffies + NEXT_TIMER_MAX_DELTA). However, the code in tick_nohz_stop_sched_tick(), which called __next_timer_interrupt() via get_next_timer_interrupt(), compares the return value to (last_jiffies + NEXT_TIMER_MAX_DELTA) to see if the timer needs to be re-armed. base->timer_jiffies != last_jiffies and so tick_nohz_stop_sched_tick() interperts the return value as indication that there is a distant future event 12 days from now and programs the timer to fire next after KTIME_MAX nsecs instead of avoiding to arm it. This ends up causing a needless interrupt once every KTIME_MAX nsecs." Fix this by using the new active timer accounting. This avoids scans when no active timer is enqueued completely, so we don't have to rely on base->timer_next and base->timer_jiffies anymore. Reported-by: Gilad Ben-Yossef <gilad@benyossef.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Link: http://lkml.kernel.org/r/20120525214819.317535385@linutronix.de
2012-05-26 02:08:59 +04:00
}
raw_spin_unlock(&base->lock);
return cmp_next_hrtimer_event(basem, expires);
}
/**
* timer_clear_idle - Clear the idle state of the timer base
*
* Called with interrupts disabled
*/
void timer_clear_idle(void)
{
struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
/*
* We do this unlocked. The worst outcome is a remote enqueue sending
* a pointless IPI, but taking the lock would just make the window for
* sending the IPI a few instructions smaller for the cost of taking
* the lock in the exit from idle path.
*/
base->is_idle = false;
}
static int collect_expired_timers(struct timer_base *base,
struct hlist_head *heads)
{
/*
* NOHZ optimization. After a long idle sleep we need to forward the
* base to current jiffies. Avoid a loop by searching the bitfield for
* the next expiring timer.
*/
if ((long)(jiffies - base->clk) > 2) {
unsigned long next = __next_timer_interrupt(base);
/*
* If the next timer is ahead of time forward to current
* jiffies, otherwise forward to the next expiry time:
*/
if (time_after(next, jiffies)) {
/*
* The call site will increment base->clk and then
* terminate the expiry loop immediately.
*/
base->clk = jiffies;
return 0;
}
base->clk = next;
}
return __collect_expired_timers(base, heads);
}
#else
static inline int collect_expired_timers(struct timer_base *base,
struct hlist_head *heads)
{
return __collect_expired_timers(base, heads);
}
#endif
/*
* Called from the timer interrupt handler to charge one tick to the current
* process. user_tick is 1 if the tick is user time, 0 for system.
*/
void update_process_times(int user_tick)
{
struct task_struct *p = current;
/* Note: this timer irq context must be accounted for as well. */
account_process_tick(p, user_tick);
run_local_timers();
rcu_sched_clock_irq(user_tick);
#ifdef CONFIG_IRQ_WORK
if (in_irq())
irq_work: Force raised irq work to run on irq work interrupt The nohz full kick, which restarts the tick when any resource depend on it, can't be executed anywhere given the operation it does on timers. If it is called from the scheduler or timers code, chances are that we run into a deadlock. This is why we run the nohz full kick from an irq work. That way we make sure that the kick runs on a virgin context. However if that's the case when irq work runs in its own dedicated self-ipi, things are different for the big bunch of archs that don't support the self triggered way. In order to support them, irq works are also handled by the timer interrupt as fallback. Now when irq works run on the timer interrupt, the context isn't blank. More precisely, they can run in the context of the hrtimer that runs the tick. But the nohz kick cancels and restarts this hrtimer and cancelling an hrtimer from itself isn't allowed. This is why we run in an endless loop: Kernel panic - not syncing: Watchdog detected hard LOCKUP on cpu 2 CPU: 2 PID: 7538 Comm: kworker/u8:8 Not tainted 3.16.0+ #34 Workqueue: btrfs-endio-write normal_work_helper [btrfs] ffff880244c06c88 000000001b486fe1 ffff880244c06bf0 ffffffff8a7f1e37 ffffffff8ac52a18 ffff880244c06c78 ffffffff8a7ef928 0000000000000010 ffff880244c06c88 ffff880244c06c20 000000001b486fe1 0000000000000000 Call Trace: <NMI[<ffffffff8a7f1e37>] dump_stack+0x4e/0x7a [<ffffffff8a7ef928>] panic+0xd4/0x207 [<ffffffff8a1450e8>] watchdog_overflow_callback+0x118/0x120 [<ffffffff8a186b0e>] __perf_event_overflow+0xae/0x350 [<ffffffff8a184f80>] ? perf_event_task_disable+0xa0/0xa0 [<ffffffff8a01a4cf>] ? x86_perf_event_set_period+0xbf/0x150 [<ffffffff8a187934>] perf_event_overflow+0x14/0x20 [<ffffffff8a020386>] intel_pmu_handle_irq+0x206/0x410 [<ffffffff8a01937b>] perf_event_nmi_handler+0x2b/0x50 [<ffffffff8a007b72>] nmi_handle+0xd2/0x390 [<ffffffff8a007aa5>] ? nmi_handle+0x5/0x390 [<ffffffff8a0cb7f8>] ? match_held_lock+0x8/0x1b0 [<ffffffff8a008062>] default_do_nmi+0x72/0x1c0 [<ffffffff8a008268>] do_nmi+0xb8/0x100 [<ffffffff8a7ff66a>] end_repeat_nmi+0x1e/0x2e [<ffffffff8a0cb7f8>] ? match_held_lock+0x8/0x1b0 [<ffffffff8a0cb7f8>] ? match_held_lock+0x8/0x1b0 [<ffffffff8a0cb7f8>] ? match_held_lock+0x8/0x1b0 <<EOE><IRQ[<ffffffff8a0ccd2f>] lock_acquired+0xaf/0x450 [<ffffffff8a0f74c5>] ? lock_hrtimer_base.isra.20+0x25/0x50 [<ffffffff8a7fc678>] _raw_spin_lock_irqsave+0x78/0x90 [<ffffffff8a0f74c5>] ? lock_hrtimer_base.isra.20+0x25/0x50 [<ffffffff8a0f74c5>] lock_hrtimer_base.isra.20+0x25/0x50 [<ffffffff8a0f7723>] hrtimer_try_to_cancel+0x33/0x1e0 [<ffffffff8a0f78ea>] hrtimer_cancel+0x1a/0x30 [<ffffffff8a109237>] tick_nohz_restart+0x17/0x90 [<ffffffff8a10a213>] __tick_nohz_full_check+0xc3/0x100 [<ffffffff8a10a25e>] nohz_full_kick_work_func+0xe/0x10 [<ffffffff8a17c884>] irq_work_run_list+0x44/0x70 [<ffffffff8a17c8da>] irq_work_run+0x2a/0x50 [<ffffffff8a0f700b>] update_process_times+0x5b/0x70 [<ffffffff8a109005>] tick_sched_handle.isra.21+0x25/0x60 [<ffffffff8a109b81>] tick_sched_timer+0x41/0x60 [<ffffffff8a0f7aa2>] __run_hrtimer+0x72/0x470 [<ffffffff8a109b40>] ? tick_sched_do_timer+0xb0/0xb0 [<ffffffff8a0f8707>] hrtimer_interrupt+0x117/0x270 [<ffffffff8a034357>] local_apic_timer_interrupt+0x37/0x60 [<ffffffff8a80010f>] smp_apic_timer_interrupt+0x3f/0x50 [<ffffffff8a7fe52f>] apic_timer_interrupt+0x6f/0x80 To fix this we force non-lazy irq works to run on irq work self-IPIs when available. That ability of the arch to trigger irq work self IPIs is available with arch_irq_work_has_interrupt(). Reported-by: Catalin Iacob <iacobcatalin@gmail.com> Reported-by: Dave Jones <davej@redhat.com> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Ingo Molnar <mingo@kernel.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
2014-08-16 20:37:19 +04:00
irq_work_tick();
#endif
scheduler_tick();
if (IS_ENABLED(CONFIG_POSIX_TIMERS))
run_posix_cpu_timers();
}
/**
* __run_timers - run all expired timers (if any) on this CPU.
* @base: the timer vector to be processed.
*/
static inline void __run_timers(struct timer_base *base)
{
struct hlist_head heads[LVL_DEPTH];
int levels;
if (!time_after_eq(jiffies, base->clk))
return;
timers: Prepare support for PREEMPT_RT When PREEMPT_RT is enabled, the soft interrupt thread can be preempted. If the soft interrupt thread is preempted in the middle of a timer callback, then calling del_timer_sync() can lead to two issues: - If the caller is on a remote CPU then it has to spin wait for the timer handler to complete. This can result in unbound priority inversion. - If the caller originates from the task which preempted the timer handler on the same CPU, then spin waiting for the timer handler to complete is never going to end. To avoid these issues, add a new lock to the timer base which is held around the execution of the timer callbacks. If del_timer_sync() detects that the timer callback is currently running, it blocks on the expiry lock. When the callback is finished, the expiry lock is dropped by the softirq thread which wakes up the waiter and the system makes progress. This addresses both the priority inversion and the life lock issues. This mechanism is not used for timers which are marked IRQSAFE as for those preemption is disabled accross the callback and therefore this situation cannot happen. The callbacks for such timers need to be individually audited for RT compliance. The same issue can happen in virtual machines when the vCPU which runs a timer callback is scheduled out. If a second vCPU of the same guest calls del_timer_sync() it will spin wait for the other vCPU to be scheduled back in. The expiry lock mechanism would avoid that. It'd be trivial to enable this when paravirt spinlocks are enabled in a guest, but it's not clear whether this is an actual problem in the wild, so for now it's an RT only mechanism. As the softirq thread can be preempted with PREEMPT_RT=y, the SMP variant of del_timer_sync() needs to be used on UP as well. [ tglx: Refactored it for mainline ] Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20190726185753.832418500@linutronix.de
2019-07-26 21:31:00 +03:00
timer_base_lock_expiry(base);
raw_spin_lock_irq(&base->lock);
/*
* timer_base::must_forward_clk must be cleared before running
* timers so that any timer functions that call mod_timer() will
* not try to forward the base. Idle tracking / clock forwarding
* logic is only used with BASE_STD timers.
*
* The must_forward_clk flag is cleared unconditionally also for
* the deferrable base. The deferrable base is not affected by idle
* tracking and never forwarded, so clearing the flag is a NOOP.
*
* The fact that the deferrable base is never forwarded can cause
* large variations in granularity for deferrable timers, but they
* can be deferred for long periods due to idle anyway.
*/
base->must_forward_clk = false;
while (time_after_eq(jiffies, base->clk)) {
levels = collect_expired_timers(base, heads);
base->clk++;
while (levels--)
expire_timers(base, heads + levels);
}
raw_spin_unlock_irq(&base->lock);
timers: Prepare support for PREEMPT_RT When PREEMPT_RT is enabled, the soft interrupt thread can be preempted. If the soft interrupt thread is preempted in the middle of a timer callback, then calling del_timer_sync() can lead to two issues: - If the caller is on a remote CPU then it has to spin wait for the timer handler to complete. This can result in unbound priority inversion. - If the caller originates from the task which preempted the timer handler on the same CPU, then spin waiting for the timer handler to complete is never going to end. To avoid these issues, add a new lock to the timer base which is held around the execution of the timer callbacks. If del_timer_sync() detects that the timer callback is currently running, it blocks on the expiry lock. When the callback is finished, the expiry lock is dropped by the softirq thread which wakes up the waiter and the system makes progress. This addresses both the priority inversion and the life lock issues. This mechanism is not used for timers which are marked IRQSAFE as for those preemption is disabled accross the callback and therefore this situation cannot happen. The callbacks for such timers need to be individually audited for RT compliance. The same issue can happen in virtual machines when the vCPU which runs a timer callback is scheduled out. If a second vCPU of the same guest calls del_timer_sync() it will spin wait for the other vCPU to be scheduled back in. The expiry lock mechanism would avoid that. It'd be trivial to enable this when paravirt spinlocks are enabled in a guest, but it's not clear whether this is an actual problem in the wild, so for now it's an RT only mechanism. As the softirq thread can be preempted with PREEMPT_RT=y, the SMP variant of del_timer_sync() needs to be used on UP as well. [ tglx: Refactored it for mainline ] Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20190726185753.832418500@linutronix.de
2019-07-26 21:31:00 +03:00
timer_base_unlock_expiry(base);
}
/*
* This function runs timers and the timer-tq in bottom half context.
*/
static __latent_entropy void run_timer_softirq(struct softirq_action *h)
{
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
__run_timers(base);
if (IS_ENABLED(CONFIG_NO_HZ_COMMON))
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
__run_timers(this_cpu_ptr(&timer_bases[BASE_DEF]));
}
/*
* Called by the local, per-CPU timer interrupt on SMP.
*/
void run_local_timers(void)
{
struct timer_base *base = this_cpu_ptr(&timer_bases[BASE_STD]);
hrtimer_run_queues();
/* Raise the softirq only if required. */
if (time_before(jiffies, base->clk)) {
if (!IS_ENABLED(CONFIG_NO_HZ_COMMON))
return;
/* CPU is awake, so check the deferrable base. */
base++;
if (time_before(jiffies, base->clk))
return;
}
raise_softirq(TIMER_SOFTIRQ);
}
timer: Convert schedule_timeout() to use from_timer() In preparation for unconditionally passing the struct timer_list pointer to all timer callbacks, switch to using the new from_timer() helper and passing the timer pointer explicitly. Since this special timer is on the stack, it needs to have a wrapper structure to carry state once .data is eliminated. Signed-off-by: Kees Cook <keescook@chromium.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mips@linux-mips.org Cc: Petr Mladek <pmladek@suse.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Lai Jiangshan <jiangshanlai@gmail.com> Cc: Sebastian Reichel <sre@kernel.org> Cc: Kalle Valo <kvalo@qca.qualcomm.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Pavel Machek <pavel@ucw.cz> Cc: linux1394-devel@lists.sourceforge.net Cc: Chris Metcalf <cmetcalf@mellanox.com> Cc: linux-s390@vger.kernel.org Cc: linux-wireless@vger.kernel.org Cc: "James E.J. Bottomley" <jejb@linux.vnet.ibm.com> Cc: Wim Van Sebroeck <wim@iguana.be> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Ursula Braun <ubraun@linux.vnet.ibm.com> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Harish Patil <harish.patil@cavium.com> Cc: Guenter Roeck <linux@roeck-us.net> Cc: Manish Chopra <manish.chopra@cavium.com> Cc: Len Brown <len.brown@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: linux-pm@vger.kernel.org Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Julian Wiedmann <jwi@linux.vnet.ibm.com> Cc: John Stultz <john.stultz@linaro.org> Cc: Mark Gross <mark.gross@intel.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: linux-watchdog@vger.kernel.org Cc: linux-scsi@vger.kernel.org Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Stefan Richter <stefanr@s5r6.in-berlin.de> Cc: Michael Reed <mdr@sgi.com> Cc: netdev@vger.kernel.org Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: linuxppc-dev@lists.ozlabs.org Cc: Sudip Mukherjee <sudipm.mukherjee@gmail.com> Link: https://lkml.kernel.org/r/1507159627-127660-2-git-send-email-keescook@chromium.org
2017-10-05 02:26:55 +03:00
/*
* Since schedule_timeout()'s timer is defined on the stack, it must store
* the target task on the stack as well.
*/
struct process_timer {
struct timer_list timer;
struct task_struct *task;
};
static void process_timeout(struct timer_list *t)
{
timer: Convert schedule_timeout() to use from_timer() In preparation for unconditionally passing the struct timer_list pointer to all timer callbacks, switch to using the new from_timer() helper and passing the timer pointer explicitly. Since this special timer is on the stack, it needs to have a wrapper structure to carry state once .data is eliminated. Signed-off-by: Kees Cook <keescook@chromium.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mips@linux-mips.org Cc: Petr Mladek <pmladek@suse.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Lai Jiangshan <jiangshanlai@gmail.com> Cc: Sebastian Reichel <sre@kernel.org> Cc: Kalle Valo <kvalo@qca.qualcomm.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Pavel Machek <pavel@ucw.cz> Cc: linux1394-devel@lists.sourceforge.net Cc: Chris Metcalf <cmetcalf@mellanox.com> Cc: linux-s390@vger.kernel.org Cc: linux-wireless@vger.kernel.org Cc: "James E.J. Bottomley" <jejb@linux.vnet.ibm.com> Cc: Wim Van Sebroeck <wim@iguana.be> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Ursula Braun <ubraun@linux.vnet.ibm.com> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Harish Patil <harish.patil@cavium.com> Cc: Guenter Roeck <linux@roeck-us.net> Cc: Manish Chopra <manish.chopra@cavium.com> Cc: Len Brown <len.brown@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: linux-pm@vger.kernel.org Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Julian Wiedmann <jwi@linux.vnet.ibm.com> Cc: John Stultz <john.stultz@linaro.org> Cc: Mark Gross <mark.gross@intel.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: linux-watchdog@vger.kernel.org Cc: linux-scsi@vger.kernel.org Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Stefan Richter <stefanr@s5r6.in-berlin.de> Cc: Michael Reed <mdr@sgi.com> Cc: netdev@vger.kernel.org Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: linuxppc-dev@lists.ozlabs.org Cc: Sudip Mukherjee <sudipm.mukherjee@gmail.com> Link: https://lkml.kernel.org/r/1507159627-127660-2-git-send-email-keescook@chromium.org
2017-10-05 02:26:55 +03:00
struct process_timer *timeout = from_timer(timeout, t, timer);
wake_up_process(timeout->task);
}
/**
* schedule_timeout - sleep until timeout
* @timeout: timeout value in jiffies
*
* Make the current task sleep until @timeout jiffies have
* elapsed. The routine will return immediately unless
* the current task state has been set (see set_current_state()).
*
* You can set the task state as follows -
*
* %TASK_UNINTERRUPTIBLE - at least @timeout jiffies are guaranteed to
* pass before the routine returns unless the current task is explicitly
* woken up, (e.g. by wake_up_process())".
*
* %TASK_INTERRUPTIBLE - the routine may return early if a signal is
* delivered to the current task or the current task is explicitly woken
* up.
*
* The current task state is guaranteed to be TASK_RUNNING when this
* routine returns.
*
* Specifying a @timeout value of %MAX_SCHEDULE_TIMEOUT will schedule
* the CPU away without a bound on the timeout. In this case the return
* value will be %MAX_SCHEDULE_TIMEOUT.
*
* Returns 0 when the timer has expired otherwise the remaining time in
* jiffies will be returned. In all cases the return value is guaranteed
* to be non-negative.
*/
signed long __sched schedule_timeout(signed long timeout)
{
timer: Convert schedule_timeout() to use from_timer() In preparation for unconditionally passing the struct timer_list pointer to all timer callbacks, switch to using the new from_timer() helper and passing the timer pointer explicitly. Since this special timer is on the stack, it needs to have a wrapper structure to carry state once .data is eliminated. Signed-off-by: Kees Cook <keescook@chromium.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mips@linux-mips.org Cc: Petr Mladek <pmladek@suse.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Lai Jiangshan <jiangshanlai@gmail.com> Cc: Sebastian Reichel <sre@kernel.org> Cc: Kalle Valo <kvalo@qca.qualcomm.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Pavel Machek <pavel@ucw.cz> Cc: linux1394-devel@lists.sourceforge.net Cc: Chris Metcalf <cmetcalf@mellanox.com> Cc: linux-s390@vger.kernel.org Cc: linux-wireless@vger.kernel.org Cc: "James E.J. Bottomley" <jejb@linux.vnet.ibm.com> Cc: Wim Van Sebroeck <wim@iguana.be> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Ursula Braun <ubraun@linux.vnet.ibm.com> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Harish Patil <harish.patil@cavium.com> Cc: Guenter Roeck <linux@roeck-us.net> Cc: Manish Chopra <manish.chopra@cavium.com> Cc: Len Brown <len.brown@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: linux-pm@vger.kernel.org Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Julian Wiedmann <jwi@linux.vnet.ibm.com> Cc: John Stultz <john.stultz@linaro.org> Cc: Mark Gross <mark.gross@intel.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: linux-watchdog@vger.kernel.org Cc: linux-scsi@vger.kernel.org Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Stefan Richter <stefanr@s5r6.in-berlin.de> Cc: Michael Reed <mdr@sgi.com> Cc: netdev@vger.kernel.org Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: linuxppc-dev@lists.ozlabs.org Cc: Sudip Mukherjee <sudipm.mukherjee@gmail.com> Link: https://lkml.kernel.org/r/1507159627-127660-2-git-send-email-keescook@chromium.org
2017-10-05 02:26:55 +03:00
struct process_timer timer;
unsigned long expire;
switch (timeout)
{
case MAX_SCHEDULE_TIMEOUT:
/*
* These two special cases are useful to be comfortable
* in the caller. Nothing more. We could take
* MAX_SCHEDULE_TIMEOUT from one of the negative value
* but I' d like to return a valid offset (>=0) to allow
* the caller to do everything it want with the retval.
*/
schedule();
goto out;
default:
/*
* Another bit of PARANOID. Note that the retval will be
* 0 since no piece of kernel is supposed to do a check
* for a negative retval of schedule_timeout() (since it
* should never happens anyway). You just have the printk()
* that will tell you if something is gone wrong and where.
*/
if (timeout < 0) {
printk(KERN_ERR "schedule_timeout: wrong timeout "
"value %lx\n", timeout);
dump_stack();
current->state = TASK_RUNNING;
goto out;
}
}
expire = timeout + jiffies;
timer: Convert schedule_timeout() to use from_timer() In preparation for unconditionally passing the struct timer_list pointer to all timer callbacks, switch to using the new from_timer() helper and passing the timer pointer explicitly. Since this special timer is on the stack, it needs to have a wrapper structure to carry state once .data is eliminated. Signed-off-by: Kees Cook <keescook@chromium.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mips@linux-mips.org Cc: Petr Mladek <pmladek@suse.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Lai Jiangshan <jiangshanlai@gmail.com> Cc: Sebastian Reichel <sre@kernel.org> Cc: Kalle Valo <kvalo@qca.qualcomm.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Pavel Machek <pavel@ucw.cz> Cc: linux1394-devel@lists.sourceforge.net Cc: Chris Metcalf <cmetcalf@mellanox.com> Cc: linux-s390@vger.kernel.org Cc: linux-wireless@vger.kernel.org Cc: "James E.J. Bottomley" <jejb@linux.vnet.ibm.com> Cc: Wim Van Sebroeck <wim@iguana.be> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Ursula Braun <ubraun@linux.vnet.ibm.com> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Harish Patil <harish.patil@cavium.com> Cc: Guenter Roeck <linux@roeck-us.net> Cc: Manish Chopra <manish.chopra@cavium.com> Cc: Len Brown <len.brown@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: linux-pm@vger.kernel.org Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Julian Wiedmann <jwi@linux.vnet.ibm.com> Cc: John Stultz <john.stultz@linaro.org> Cc: Mark Gross <mark.gross@intel.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: linux-watchdog@vger.kernel.org Cc: linux-scsi@vger.kernel.org Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Stefan Richter <stefanr@s5r6.in-berlin.de> Cc: Michael Reed <mdr@sgi.com> Cc: netdev@vger.kernel.org Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: linuxppc-dev@lists.ozlabs.org Cc: Sudip Mukherjee <sudipm.mukherjee@gmail.com> Link: https://lkml.kernel.org/r/1507159627-127660-2-git-send-email-keescook@chromium.org
2017-10-05 02:26:55 +03:00
timer.task = current;
timer_setup_on_stack(&timer.timer, process_timeout, 0);
timers: Add a function to start/reduce a timer Add a function, similar to mod_timer(), that will start a timer if it isn't running and will modify it if it is running and has an expiry time longer than the new time. If the timer is running with an expiry time that's the same or sooner, no change is made. The function looks like: int timer_reduce(struct timer_list *timer, unsigned long expires); This can be used by code such as networking code to make it easier to share a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code, the rxrpc_call struct will maintain a number of timeouts: unsigned long ack_at; unsigned long resend_at; unsigned long ping_at; unsigned long expect_rx_by; unsigned long expect_req_by; unsigned long expect_term_by; each of which is set independently of the others. With timer reduction available, when the code needs to set one of the timeouts, it only needs to look at that timeout and then call timer_reduce() to modify the timer, starting it or bringing it forward if necessary. There is no need to refer to the other timeouts to see which is earliest and no need to take any lock other than, potentially, the timer lock inside timer_reduce(). Note, that this does not protect against concurrent invocations of any of the timer functions. As an example, the expect_rx_by timeout above, which terminates a call if we don't get a packet from the server within a certain time window, would be set something like this: unsigned long now = jiffies; unsigned long expect_rx_by = now + packet_receive_timeout; WRITE_ONCE(call->expect_rx_by, expect_rx_by); timer_reduce(&call->timer, expect_rx_by); The timer service code (which might, say, be in a work function) would then check all the timeouts to see which, if any, had triggered, deal with those: t = READ_ONCE(call->ack_at); if (time_after_eq(now, t)) { cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET); set_bit(RXRPC_CALL_EV_ACK, &call->events); } and then restart the timer if necessary by finding the soonest timeout that hasn't yet passed and then calling timer_reduce(). The disadvantage of doing things this way rather than comparing the timers each time and calling mod_timer() is that you *will* take timer events unless you can finish what you're doing and delete the timer in time. The advantage of doing things this way is that you don't need to use a lock to work out when the next timer should be set, other than the timer's own lock - which you might not have to take. [ tglx: Fixed weird formatting and adopted it to pending changes ] Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: keyrings@vger.kernel.org Cc: linux-afs@lists.infradead.org Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
2017-11-09 15:35:07 +03:00
__mod_timer(&timer.timer, expire, 0);
schedule();
timer: Convert schedule_timeout() to use from_timer() In preparation for unconditionally passing the struct timer_list pointer to all timer callbacks, switch to using the new from_timer() helper and passing the timer pointer explicitly. Since this special timer is on the stack, it needs to have a wrapper structure to carry state once .data is eliminated. Signed-off-by: Kees Cook <keescook@chromium.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mips@linux-mips.org Cc: Petr Mladek <pmladek@suse.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Lai Jiangshan <jiangshanlai@gmail.com> Cc: Sebastian Reichel <sre@kernel.org> Cc: Kalle Valo <kvalo@qca.qualcomm.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Pavel Machek <pavel@ucw.cz> Cc: linux1394-devel@lists.sourceforge.net Cc: Chris Metcalf <cmetcalf@mellanox.com> Cc: linux-s390@vger.kernel.org Cc: linux-wireless@vger.kernel.org Cc: "James E.J. Bottomley" <jejb@linux.vnet.ibm.com> Cc: Wim Van Sebroeck <wim@iguana.be> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Ursula Braun <ubraun@linux.vnet.ibm.com> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Harish Patil <harish.patil@cavium.com> Cc: Guenter Roeck <linux@roeck-us.net> Cc: Manish Chopra <manish.chopra@cavium.com> Cc: Len Brown <len.brown@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: linux-pm@vger.kernel.org Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Julian Wiedmann <jwi@linux.vnet.ibm.com> Cc: John Stultz <john.stultz@linaro.org> Cc: Mark Gross <mark.gross@intel.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: linux-watchdog@vger.kernel.org Cc: linux-scsi@vger.kernel.org Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Stefan Richter <stefanr@s5r6.in-berlin.de> Cc: Michael Reed <mdr@sgi.com> Cc: netdev@vger.kernel.org Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: linuxppc-dev@lists.ozlabs.org Cc: Sudip Mukherjee <sudipm.mukherjee@gmail.com> Link: https://lkml.kernel.org/r/1507159627-127660-2-git-send-email-keescook@chromium.org
2017-10-05 02:26:55 +03:00
del_singleshot_timer_sync(&timer.timer);
/* Remove the timer from the object tracker */
timer: Convert schedule_timeout() to use from_timer() In preparation for unconditionally passing the struct timer_list pointer to all timer callbacks, switch to using the new from_timer() helper and passing the timer pointer explicitly. Since this special timer is on the stack, it needs to have a wrapper structure to carry state once .data is eliminated. Signed-off-by: Kees Cook <keescook@chromium.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: linux-mips@linux-mips.org Cc: Petr Mladek <pmladek@suse.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Lai Jiangshan <jiangshanlai@gmail.com> Cc: Sebastian Reichel <sre@kernel.org> Cc: Kalle Valo <kvalo@qca.qualcomm.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Pavel Machek <pavel@ucw.cz> Cc: linux1394-devel@lists.sourceforge.net Cc: Chris Metcalf <cmetcalf@mellanox.com> Cc: linux-s390@vger.kernel.org Cc: linux-wireless@vger.kernel.org Cc: "James E.J. Bottomley" <jejb@linux.vnet.ibm.com> Cc: Wim Van Sebroeck <wim@iguana.be> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Ursula Braun <ubraun@linux.vnet.ibm.com> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: Harish Patil <harish.patil@cavium.com> Cc: Guenter Roeck <linux@roeck-us.net> Cc: Manish Chopra <manish.chopra@cavium.com> Cc: Len Brown <len.brown@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: linux-pm@vger.kernel.org Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Julian Wiedmann <jwi@linux.vnet.ibm.com> Cc: John Stultz <john.stultz@linaro.org> Cc: Mark Gross <mark.gross@intel.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: linux-watchdog@vger.kernel.org Cc: linux-scsi@vger.kernel.org Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Stephen Boyd <sboyd@codeaurora.org> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Stefan Richter <stefanr@s5r6.in-berlin.de> Cc: Michael Reed <mdr@sgi.com> Cc: netdev@vger.kernel.org Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: linuxppc-dev@lists.ozlabs.org Cc: Sudip Mukherjee <sudipm.mukherjee@gmail.com> Link: https://lkml.kernel.org/r/1507159627-127660-2-git-send-email-keescook@chromium.org
2017-10-05 02:26:55 +03:00
destroy_timer_on_stack(&timer.timer);
timeout = expire - jiffies;
out:
return timeout < 0 ? 0 : timeout;
}
EXPORT_SYMBOL(schedule_timeout);
/*
* We can use __set_current_state() here because schedule_timeout() calls
* schedule() unconditionally.
*/
signed long __sched schedule_timeout_interruptible(signed long timeout)
{
__set_current_state(TASK_INTERRUPTIBLE);
return schedule_timeout(timeout);
}
EXPORT_SYMBOL(schedule_timeout_interruptible);
signed long __sched schedule_timeout_killable(signed long timeout)
{
__set_current_state(TASK_KILLABLE);
return schedule_timeout(timeout);
}
EXPORT_SYMBOL(schedule_timeout_killable);
signed long __sched schedule_timeout_uninterruptible(signed long timeout)
{
__set_current_state(TASK_UNINTERRUPTIBLE);
return schedule_timeout(timeout);
}
EXPORT_SYMBOL(schedule_timeout_uninterruptible);
/*
* Like schedule_timeout_uninterruptible(), except this task will not contribute
* to load average.
*/
signed long __sched schedule_timeout_idle(signed long timeout)
{
__set_current_state(TASK_IDLE);
return schedule_timeout(timeout);
}
EXPORT_SYMBOL(schedule_timeout_idle);
#ifdef CONFIG_HOTPLUG_CPU
static void migrate_timer_list(struct timer_base *new_base, struct hlist_head *head)
{
struct timer_list *timer;
timer: Replace timer base by a cpu index Instead of storing a pointer to the per cpu tvec_base we can simply cache a CPU index in the timer_list and use that to get hold of the correct per cpu tvec_base. This is only used in lock_timer_base() and the slightly larger code is peanuts versus the spinlock operation and the d-cache foot print of the timer wheel. Aside of that this allows to get rid of following nuisances: - boot_tvec_base That statically allocated 4k bss data is just kept around so the timer has a home when it gets statically initialized. It serves no other purpose. With the CPU index we assign the timer to CPU0 at static initialization time and therefor can avoid the whole boot_tvec_base dance. That also simplifies the init code, which just can use the per cpu base. Before: text data bss dec hex filename 17491 9201 4160 30852 7884 ../build/kernel/time/timer.o After: text data bss dec hex filename 17440 9193 0 26633 6809 ../build/kernel/time/timer.o - Overloading the base pointer with various flags The CPU index has enough space to hold the flags (deferrable, irqsafe) so we can get rid of the extra masking and bit fiddling with the base pointer. As a benefit we reduce the size of struct timer_list on 64 bit machines. 4 - 8 bytes, a size reduction up to 15% per struct timer_list, which is a real win as we have tons of them embedded in other structs. This changes also the newly added deferrable printout of the timer start trace point to capture and print all timer->flags, which allows us to decode the target cpu of the timer as well. We might have used bitfields for this, but that would change the static initializers and the init function for no value to accomodate big endian bitfields. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Badhri Jagan Sridharan <Badhri@google.com> Link: http://lkml.kernel.org/r/20150526224511.950084301@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:29 +03:00
int cpu = new_base->cpu;
while (!hlist_empty(head)) {
timer = hlist_entry(head->first, struct timer_list, entry);
detach_timer(timer, false);
timer: Replace timer base by a cpu index Instead of storing a pointer to the per cpu tvec_base we can simply cache a CPU index in the timer_list and use that to get hold of the correct per cpu tvec_base. This is only used in lock_timer_base() and the slightly larger code is peanuts versus the spinlock operation and the d-cache foot print of the timer wheel. Aside of that this allows to get rid of following nuisances: - boot_tvec_base That statically allocated 4k bss data is just kept around so the timer has a home when it gets statically initialized. It serves no other purpose. With the CPU index we assign the timer to CPU0 at static initialization time and therefor can avoid the whole boot_tvec_base dance. That also simplifies the init code, which just can use the per cpu base. Before: text data bss dec hex filename 17491 9201 4160 30852 7884 ../build/kernel/time/timer.o After: text data bss dec hex filename 17440 9193 0 26633 6809 ../build/kernel/time/timer.o - Overloading the base pointer with various flags The CPU index has enough space to hold the flags (deferrable, irqsafe) so we can get rid of the extra masking and bit fiddling with the base pointer. As a benefit we reduce the size of struct timer_list on 64 bit machines. 4 - 8 bytes, a size reduction up to 15% per struct timer_list, which is a real win as we have tons of them embedded in other structs. This changes also the newly added deferrable printout of the timer start trace point to capture and print all timer->flags, which allows us to decode the target cpu of the timer as well. We might have used bitfields for this, but that would change the static initializers and the init function for no value to accomodate big endian bitfields. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Badhri Jagan Sridharan <Badhri@google.com> Link: http://lkml.kernel.org/r/20150526224511.950084301@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:29 +03:00
timer->flags = (timer->flags & ~TIMER_BASEMASK) | cpu;
internal_add_timer(new_base, timer);
}
}
int timers_prepare_cpu(unsigned int cpu)
{
struct timer_base *base;
int b;
for (b = 0; b < NR_BASES; b++) {
base = per_cpu_ptr(&timer_bases[b], cpu);
base->clk = jiffies;
base->next_expiry = base->clk + NEXT_TIMER_MAX_DELTA;
base->is_idle = false;
base->must_forward_clk = true;
}
return 0;
}
int timers_dead_cpu(unsigned int cpu)
{
struct timer_base *old_base;
struct timer_base *new_base;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
int b, i;
BUG_ON(cpu_online(cpu));
[PATCH] timers fixes/improvements This patch tries to solve following problems: 1. del_timer_sync() is racy. The timer can be fired again after del_timer_sync have checked all cpus and before it will recheck timer_pending(). 2. It has scalability problems. All cpus are scanned to determine if the timer is running on that cpu. With this patch del_timer_sync is O(1) and no slower than plain del_timer(pending_timer), unless it has to actually wait for completion of the currently running timer. The only restriction is that the recurring timer should not use add_timer_on(). 3. The timers are not serialized wrt to itself. If CPU_0 does mod_timer(jiffies+1) while the timer is currently running on CPU 1, it is quite possible that local interrupt on CPU_0 will start that timer before it finished on CPU_1. 4. The timers locking is suboptimal. __mod_timer() takes 3 locks at once and still requires wmb() in del_timer/run_timers. The new implementation takes 2 locks sequentially and does not need memory barriers. Currently ->base != NULL means that the timer is pending. In that case ->base.lock is used to lock the timer. __mod_timer also takes timer->lock because ->base can be == NULL. This patch uses timer->entry.next != NULL as indication that the timer is pending. So it does __list_del(), entry->next = NULL instead of list_del() when the timer is deleted. The ->base field is used for hashed locking only, it is initialized in init_timer() which sets ->base = per_cpu(tvec_bases). When the tvec_bases.lock is locked, it means that all timers which are tied to this base via timer->base are locked, and the base itself is locked too. So __run_timers/migrate_timers can safely modify all timers which could be found on ->tvX lists (pending timers). When the timer's base is locked, and the timer removed from ->entry list (which means that _run_timers/migrate_timers can't see this timer), it is possible to set timer->base = NULL and drop the lock: the timer remains locked. This patch adds lock_timer_base() helper, which waits for ->base != NULL, locks the ->base, and checks it is still the same. __mod_timer() schedules the timer on the local CPU and changes it's base. However, it does not lock both old and new bases at once. It locks the timer via lock_timer_base(), deletes the timer, sets ->base = NULL, and unlocks old base. Then __mod_timer() locks new_base, sets ->base = new_base, and adds this timer. This simplifies the code, because AB-BA deadlock is not possible. __mod_timer() also ensures that the timer's base is not changed while the timer's handler is running on the old base. __run_timers(), del_timer() do not change ->base anymore, they only clear pending flag. So del_timer_sync() can test timer->base->running_timer == timer to detect whether it is running or not. We don't need timer_list->lock anymore, this patch kills it. We also don't need barriers. del_timer() and __run_timers() used smp_wmb() before clearing timer's pending flag. It was needed because __mod_timer() did not lock old_base if the timer is not pending, so __mod_timer()->list_add() could race with del_timer()->list_del(). With this patch these functions are serialized through base->lock. One problem. TIMER_INITIALIZER can't use per_cpu(tvec_bases). So this patch adds global struct timer_base_s { spinlock_t lock; struct timer_list *running_timer; } __init_timer_base; which is used by TIMER_INITIALIZER. The corresponding fields in tvec_t_base_s struct are replaced by struct timer_base_s t_base. It is indeed ugly. But this can't have scalability problems. The global __init_timer_base.lock is used only when __mod_timer() is called for the first time AND the timer was compile time initialized. After that the timer migrates to the local CPU. Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru> Acked-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Renaud Lienhart <renaud.lienhart@free.fr> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 11:08:56 +04:00
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
for (b = 0; b < NR_BASES; b++) {
old_base = per_cpu_ptr(&timer_bases[b], cpu);
new_base = get_cpu_ptr(&timer_bases[b]);
/*
* The caller is globally serialized and nobody else
* takes two locks at once, deadlock is not possible.
*/
raw_spin_lock_irq(&new_base->lock);
raw_spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
timers: Forward timer base before migrating timers On CPU hotunplug the enqueued timers of the unplugged CPU are migrated to a live CPU. This happens from the control thread which initiated the unplug. If the CPU on which the control thread runs came out from a longer idle period then the base clock of that CPU might be stale because the control thread runs prior to any event which forwards the clock. In such a case the timers from the unplugged CPU are queued on the live CPU based on the stale clock which can cause large delays due to increased granularity of the outer timer wheels which are far away from base:;clock. But there is a worse problem than that. The following sequence of events illustrates it: - CPU0 timer1 is queued expires = 59969 and base->clk = 59131. The timer is queued at wheel level 2, with resulting expiry time = 60032 (due to level granularity). - CPU1 enters idle @60007, with next timer expiry @60020. - CPU0 is hotplugged at @60009 - CPU1 exits idle and runs the control thread which migrates the timers from CPU0 timer1 is now queued in level 0 for immediate handling in the next softirq because the requested expiry time 59969 is before CPU1 base->clk 60007 - CPU1 runs code which forwards the base clock which succeeds because the next expiring timer. which was collected at idle entry time is still set to 60020. So it forwards beyond 60007 and therefore misses to expire the migrated timer1. That timer gets expired when the wheel wraps around again, which takes between 63 and 630ms depending on the HZ setting. Address both problems by invoking forward_timer_base() for the control CPUs timer base. All other places, which might run into a similar problem (mod_timer()/add_timer_on()) already invoke forward_timer_base() to avoid that. [ tglx: Massaged comment and changelog ] Fixes: a683f390b93f ("timers: Forward the wheel clock whenever possible") Co-developed-by: Neeraj Upadhyay <neeraju@codeaurora.org> Signed-off-by: Neeraj Upadhyay <neeraju@codeaurora.org> Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Anna-Maria Gleixner <anna-maria@linutronix.de> Cc: linux-arm-msm@vger.kernel.org Cc: stable@vger.kernel.org Link: https://lkml.kernel.org/r/20180118115022.6368-1-clingutla@codeaurora.org
2018-01-18 14:50:22 +03:00
/*
* The current CPUs base clock might be stale. Update it
* before moving the timers over.
*/
forward_timer_base(new_base);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
BUG_ON(old_base->running_timer);
for (i = 0; i < WHEEL_SIZE; i++)
migrate_timer_list(new_base, old_base->vectors + i);
raw_spin_unlock(&old_base->lock);
raw_spin_unlock_irq(&new_base->lock);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
put_cpu_ptr(&timer_bases);
}
return 0;
}
#endif /* CONFIG_HOTPLUG_CPU */
timer: Replace timer base by a cpu index Instead of storing a pointer to the per cpu tvec_base we can simply cache a CPU index in the timer_list and use that to get hold of the correct per cpu tvec_base. This is only used in lock_timer_base() and the slightly larger code is peanuts versus the spinlock operation and the d-cache foot print of the timer wheel. Aside of that this allows to get rid of following nuisances: - boot_tvec_base That statically allocated 4k bss data is just kept around so the timer has a home when it gets statically initialized. It serves no other purpose. With the CPU index we assign the timer to CPU0 at static initialization time and therefor can avoid the whole boot_tvec_base dance. That also simplifies the init code, which just can use the per cpu base. Before: text data bss dec hex filename 17491 9201 4160 30852 7884 ../build/kernel/time/timer.o After: text data bss dec hex filename 17440 9193 0 26633 6809 ../build/kernel/time/timer.o - Overloading the base pointer with various flags The CPU index has enough space to hold the flags (deferrable, irqsafe) so we can get rid of the extra masking and bit fiddling with the base pointer. As a benefit we reduce the size of struct timer_list on 64 bit machines. 4 - 8 bytes, a size reduction up to 15% per struct timer_list, which is a real win as we have tons of them embedded in other structs. This changes also the newly added deferrable printout of the timer start trace point to capture and print all timer->flags, which allows us to decode the target cpu of the timer as well. We might have used bitfields for this, but that would change the static initializers and the init function for no value to accomodate big endian bitfields. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Badhri Jagan Sridharan <Badhri@google.com> Link: http://lkml.kernel.org/r/20150526224511.950084301@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:29 +03:00
static void __init init_timer_cpu(int cpu)
{
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
struct timer_base *base;
int i;
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
for (i = 0; i < NR_BASES; i++) {
base = per_cpu_ptr(&timer_bases[i], cpu);
base->cpu = cpu;
raw_spin_lock_init(&base->lock);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
base->clk = jiffies;
timers: Prepare support for PREEMPT_RT When PREEMPT_RT is enabled, the soft interrupt thread can be preempted. If the soft interrupt thread is preempted in the middle of a timer callback, then calling del_timer_sync() can lead to two issues: - If the caller is on a remote CPU then it has to spin wait for the timer handler to complete. This can result in unbound priority inversion. - If the caller originates from the task which preempted the timer handler on the same CPU, then spin waiting for the timer handler to complete is never going to end. To avoid these issues, add a new lock to the timer base which is held around the execution of the timer callbacks. If del_timer_sync() detects that the timer callback is currently running, it blocks on the expiry lock. When the callback is finished, the expiry lock is dropped by the softirq thread which wakes up the waiter and the system makes progress. This addresses both the priority inversion and the life lock issues. This mechanism is not used for timers which are marked IRQSAFE as for those preemption is disabled accross the callback and therefore this situation cannot happen. The callbacks for such timers need to be individually audited for RT compliance. The same issue can happen in virtual machines when the vCPU which runs a timer callback is scheduled out. If a second vCPU of the same guest calls del_timer_sync() it will spin wait for the other vCPU to be scheduled back in. The expiry lock mechanism would avoid that. It'd be trivial to enable this when paravirt spinlocks are enabled in a guest, but it's not clear whether this is an actual problem in the wild, so for now it's an RT only mechanism. As the softirq thread can be preempted with PREEMPT_RT=y, the SMP variant of del_timer_sync() needs to be used on UP as well. [ tglx: Refactored it for mainline ] Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20190726185753.832418500@linutronix.de
2019-07-26 21:31:00 +03:00
timer_base_init_expiry_lock(base);
timers: Switch to a non-cascading wheel The current timer wheel has some drawbacks: 1) Cascading: Cascading can be an unbound operation and is completely pointless in most cases because the vast majority of the timer wheel timers are canceled or rearmed before expiration. (They are used as timeout safeguards, not as real timers to measure time.) 2) No fast lookup of the next expiring timer: In NOHZ scenarios the first timer soft interrupt after a long NOHZ period must fast forward the base time to the current value of jiffies. As we have no way to find the next expiring timer fast, the code loops linearly and increments the base time one by one and checks for expired timers in each step. This causes unbound overhead spikes exactly in the moment when we should wake up as fast as possible. After a thorough analysis of real world data gathered on laptops, workstations, webservers and other machines (thanks Chris!) I came to the conclusion that the current 'classic' timer wheel implementation can be modified to address the above issues. The vast majority of timer wheel timers is canceled or rearmed before expiry. Most of them are timeouts for networking and other I/O tasks. The nature of timeouts is to catch the exception from normal operation (TCP ack timed out, disk does not respond, etc.). For these kinds of timeouts the accuracy of the timeout is not really a concern. Timeouts are very often approximate worst-case values and in case the timeout fires, we already waited for a long time and performance is down the drain already. The few timers which actually expire can be split into two categories: 1) Short expiry times which expect halfways accurate expiry 2) Long term expiry times are inaccurate today already due to the batching which is done for NOHZ automatically and also via the set_timer_slack() API. So for long term expiry timers we can avoid the cascading property and just leave them in the less granular outer wheels until expiry or cancelation. Timers which are armed with a timeout larger than the wheel capacity are no longer cascaded. We expire them with the longest possible timeout (6+ days). We have not observed such timeouts in our data collection, but at least we handle them, applying the rule of the least surprise. To avoid extending the wheel levels for HZ=1000 so we can accomodate the longest observed timeouts (5 days in the network conntrack code) we reduce the first level granularity on HZ=1000 to 4ms, which effectively is the same as the HZ=250 behaviour. From our data analysis there is nothing which relies on that 1ms granularity and as a side effect we get better batching and timer locality for the networking code as well. Contrary to the classic wheel the granularity of the next wheel is not the capacity of the first wheel. The granularities of the wheels are in the currently chosen setting 8 times the granularity of the previous wheel. So for HZ=250 we end up with the following granularity levels: Level Offset Granularity Range 0 0 4 ms 0 ms - 252 ms 1 64 32 ms 256 ms - 2044 ms (256ms - ~2s) 2 128 256 ms 2048 ms - 16380 ms (~2s - ~16s) 3 192 2048 ms (~2s) 16384 ms - 131068 ms (~16s - ~2m) 4 256 16384 ms (~16s) 131072 ms - 1048572 ms (~2m - ~17m) 5 320 131072 ms (~2m) 1048576 ms - 8388604 ms (~17m - ~2h) 6 384 1048576 ms (~17m) 8388608 ms - 67108863 ms (~2h - ~18h) 7 448 8388608 ms (~2h) 67108864 ms - 536870911 ms (~18h - ~6d) That's a worst case inaccuracy of 12.5% for the timers which are queued at the beginning of a level. So the new wheel concept addresses the old issues: 1) Cascading is avoided completely 2) By keeping the timers in the bucket until expiry/cancelation we can track the buckets which have timers enqueued in a bucket bitmap and therefore can look up the next expiring timer very fast and O(1). A further benefit of the concept is that the slack calculation which is done on every timer start is no longer necessary because the granularity levels provide natural batching already. Our extensive testing with various loads did not show any performance degradation vs. the current wheel implementation. This patch does not address the 'fast lookup' issue as we wanted to make sure that there is no regression introduced by the wheel redesign. The optimizations are in follow up patches. This patch contains fixes from Anna-Maria Gleixner and Richard Cochran. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Chris Mason <clm@fb.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: George Spelvin <linux@sciencehorizons.net> Cc: Josh Triplett <josh@joshtriplett.org> Cc: Len Brown <lenb@kernel.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rik van Riel <riel@redhat.com> Cc: rt@linutronix.de Link: http://lkml.kernel.org/r/20160704094342.108621834@linutronix.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-07-04 12:50:30 +03:00
}
}
static void __init init_timer_cpus(void)
{
int cpu;
timer: Replace timer base by a cpu index Instead of storing a pointer to the per cpu tvec_base we can simply cache a CPU index in the timer_list and use that to get hold of the correct per cpu tvec_base. This is only used in lock_timer_base() and the slightly larger code is peanuts versus the spinlock operation and the d-cache foot print of the timer wheel. Aside of that this allows to get rid of following nuisances: - boot_tvec_base That statically allocated 4k bss data is just kept around so the timer has a home when it gets statically initialized. It serves no other purpose. With the CPU index we assign the timer to CPU0 at static initialization time and therefor can avoid the whole boot_tvec_base dance. That also simplifies the init code, which just can use the per cpu base. Before: text data bss dec hex filename 17491 9201 4160 30852 7884 ../build/kernel/time/timer.o After: text data bss dec hex filename 17440 9193 0 26633 6809 ../build/kernel/time/timer.o - Overloading the base pointer with various flags The CPU index has enough space to hold the flags (deferrable, irqsafe) so we can get rid of the extra masking and bit fiddling with the base pointer. As a benefit we reduce the size of struct timer_list on 64 bit machines. 4 - 8 bytes, a size reduction up to 15% per struct timer_list, which is a real win as we have tons of them embedded in other structs. This changes also the newly added deferrable printout of the timer start trace point to capture and print all timer->flags, which allows us to decode the target cpu of the timer as well. We might have used bitfields for this, but that would change the static initializers and the init function for no value to accomodate big endian bitfields. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul McKenney <paulmck@linux.vnet.ibm.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Eric Dumazet <edumazet@google.com> Cc: Viresh Kumar <viresh.kumar@linaro.org> Cc: John Stultz <john.stultz@linaro.org> Cc: Joonwoo Park <joonwoop@codeaurora.org> Cc: Wenbo Wang <wenbo.wang@memblaze.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Badhri Jagan Sridharan <Badhri@google.com> Link: http://lkml.kernel.org/r/20150526224511.950084301@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2015-05-27 01:50:29 +03:00
for_each_possible_cpu(cpu)
init_timer_cpu(cpu);
}
void __init init_timers(void)
{
init_timer_cpus();
open_softirq(TIMER_SOFTIRQ, run_timer_softirq);
}
/**
* msleep - sleep safely even with waitqueue interruptions
* @msecs: Time in milliseconds to sleep for
*/
void msleep(unsigned int msecs)
{
unsigned long timeout = msecs_to_jiffies(msecs) + 1;
while (timeout)
timeout = schedule_timeout_uninterruptible(timeout);
}
EXPORT_SYMBOL(msleep);
/**
* msleep_interruptible - sleep waiting for signals
* @msecs: Time in milliseconds to sleep for
*/
unsigned long msleep_interruptible(unsigned int msecs)
{
unsigned long timeout = msecs_to_jiffies(msecs) + 1;
while (timeout && !signal_pending(current))
timeout = schedule_timeout_interruptible(timeout);
return jiffies_to_msecs(timeout);
}
EXPORT_SYMBOL(msleep_interruptible);
timer: Added usleep_range timer usleep_range is a finer precision implementations of msleep and is designed to be a drop-in replacement for udelay where a precise sleep / busy-wait is unnecessary. Since an easy interface to hrtimers could lead to an undesired proliferation of interrupts, we provide only a "range" API, forcing the caller to think about an acceptable tolerance on both ends and hopefully avoiding introducing another interrupt. INTRO As discussed here ( http://lkml.org/lkml/2007/8/3/250 ), msleep(1) is not precise enough for many drivers (yes, sleep precision is an unfair notion, but consistently sleeping for ~an order of magnitude greater than requested is worth fixing). This patch adds a usleep API so that udelay does not have to be used. Obviously not every udelay can be replaced (those in atomic contexts or being used for simple bitbanging come to mind), but there are many, many examples of mydriver_write(...) /* Wait for hardware to latch */ udelay(100) in various drivers where a busy-wait loop is neither beneficial nor necessary, but msleep simply does not provide enough precision and people are using a busy-wait loop instead. CONCERNS FROM THE RFC Why is udelay a problem / necessary? Most callers of udelay are in device/ driver initialization code, which is serial... As I see it, there is only benefit to sleeping over a delay; the notion of "refactoring" areas that use udelay was presented, but I see usleep as the refactoring. Consider i2c, if the bus is busy, you need to wait a bit (say 100us) before trying again, your current options are: * udelay(100) * msleep(1) <-- As noted above, actually as high as ~20ms on some platforms, so not really an option * Manually set up an hrtimer to try again in 100us (which is what usleep does anyway...) People choose the udelay route because it is EASY; we need to provide a better easy route. Device / driver / boot code is *currently* serial, but every few months someone makes noise about parallelizing boot, and IMHO, a little forward-thinking now is one less thing to worry about if/when that ever happens udelay's could be preempted Sure, but if udelay plans on looping 1000 times, and it gets preempted on loop 200, whenever it's scheduled again, it is going to do the next 800 loops. Is the interruptible case needed? Probably not, but I see usleep as a very logical parallel to msleep, so it made sense to include the "full" API. Processors are getting faster (albeit not as quickly as they are becoming more parallel), so if someone wanted to be interruptible for a few usecs, why not let them? If this is a contentious point, I'm happy to remove it. OTHER THOUGHTS I believe there is also value in exposing the usleep_range option; it gives the scheduler a lot more flexibility and allows the programmer to express his intent much more clearly; it's something I would hope future driver writers will take advantage of. To get the results in the NUMBERS section below, I literally s/udelay/usleep the kernel tree; I had to go in and undo the changes to the USB drivers, but everything else booted successfully; I find that extremely telling in and of itself -- many people are using a delay API where a sleep will suit them just fine. SOME ATTEMPTS AT NUMBERS It turns out that calculating quantifiable benefit on this is challenging, so instead I will simply present the current state of things, and I hope this to be sufficient: How many udelay calls are there in 2.6.35-rc5? udealy(ARG) >= | COUNT 1000 | 319 500 | 414 100 | 1146 20 | 1832 I am working on Android, so that is my focus for this. The following table is a modified usleep that simply printk's the amount of time requested to sleep; these tests were run on a kernel with udelay >= 20 --> usleep "boot" is power-on to lock screen "power collapse" is when the power button is pushed and the device suspends "resume" is when the power button is pushed and the lock screen is displayed (no touchscreen events or anything, just turning on the display) "use device" is from the unlock swipe to clicking around a bit; there is no sd card in this phone, so fail loading music, video, camera ACTION | TOTAL NUMBER OF USLEEP CALLS | NET TIME (us) boot | 22 | 1250 power-collapse | 9 | 1200 resume | 5 | 500 use device | 59 | 7700 The most interesting category to me is the "use device" field; 7700us of busy-wait time that could be put towards better responsiveness, or at the least less power usage. Signed-off-by: Patrick Pannuto <ppannuto@codeaurora.org> Cc: apw@canonical.com Cc: corbet@lwn.net Cc: arjan@linux.intel.com Cc: Randy Dunlap <rdunlap@xenotime.net> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2010-08-03 02:01:04 +04:00
/**
* usleep_range - Sleep for an approximate time
timer: Added usleep_range timer usleep_range is a finer precision implementations of msleep and is designed to be a drop-in replacement for udelay where a precise sleep / busy-wait is unnecessary. Since an easy interface to hrtimers could lead to an undesired proliferation of interrupts, we provide only a "range" API, forcing the caller to think about an acceptable tolerance on both ends and hopefully avoiding introducing another interrupt. INTRO As discussed here ( http://lkml.org/lkml/2007/8/3/250 ), msleep(1) is not precise enough for many drivers (yes, sleep precision is an unfair notion, but consistently sleeping for ~an order of magnitude greater than requested is worth fixing). This patch adds a usleep API so that udelay does not have to be used. Obviously not every udelay can be replaced (those in atomic contexts or being used for simple bitbanging come to mind), but there are many, many examples of mydriver_write(...) /* Wait for hardware to latch */ udelay(100) in various drivers where a busy-wait loop is neither beneficial nor necessary, but msleep simply does not provide enough precision and people are using a busy-wait loop instead. CONCERNS FROM THE RFC Why is udelay a problem / necessary? Most callers of udelay are in device/ driver initialization code, which is serial... As I see it, there is only benefit to sleeping over a delay; the notion of "refactoring" areas that use udelay was presented, but I see usleep as the refactoring. Consider i2c, if the bus is busy, you need to wait a bit (say 100us) before trying again, your current options are: * udelay(100) * msleep(1) <-- As noted above, actually as high as ~20ms on some platforms, so not really an option * Manually set up an hrtimer to try again in 100us (which is what usleep does anyway...) People choose the udelay route because it is EASY; we need to provide a better easy route. Device / driver / boot code is *currently* serial, but every few months someone makes noise about parallelizing boot, and IMHO, a little forward-thinking now is one less thing to worry about if/when that ever happens udelay's could be preempted Sure, but if udelay plans on looping 1000 times, and it gets preempted on loop 200, whenever it's scheduled again, it is going to do the next 800 loops. Is the interruptible case needed? Probably not, but I see usleep as a very logical parallel to msleep, so it made sense to include the "full" API. Processors are getting faster (albeit not as quickly as they are becoming more parallel), so if someone wanted to be interruptible for a few usecs, why not let them? If this is a contentious point, I'm happy to remove it. OTHER THOUGHTS I believe there is also value in exposing the usleep_range option; it gives the scheduler a lot more flexibility and allows the programmer to express his intent much more clearly; it's something I would hope future driver writers will take advantage of. To get the results in the NUMBERS section below, I literally s/udelay/usleep the kernel tree; I had to go in and undo the changes to the USB drivers, but everything else booted successfully; I find that extremely telling in and of itself -- many people are using a delay API where a sleep will suit them just fine. SOME ATTEMPTS AT NUMBERS It turns out that calculating quantifiable benefit on this is challenging, so instead I will simply present the current state of things, and I hope this to be sufficient: How many udelay calls are there in 2.6.35-rc5? udealy(ARG) >= | COUNT 1000 | 319 500 | 414 100 | 1146 20 | 1832 I am working on Android, so that is my focus for this. The following table is a modified usleep that simply printk's the amount of time requested to sleep; these tests were run on a kernel with udelay >= 20 --> usleep "boot" is power-on to lock screen "power collapse" is when the power button is pushed and the device suspends "resume" is when the power button is pushed and the lock screen is displayed (no touchscreen events or anything, just turning on the display) "use device" is from the unlock swipe to clicking around a bit; there is no sd card in this phone, so fail loading music, video, camera ACTION | TOTAL NUMBER OF USLEEP CALLS | NET TIME (us) boot | 22 | 1250 power-collapse | 9 | 1200 resume | 5 | 500 use device | 59 | 7700 The most interesting category to me is the "use device" field; 7700us of busy-wait time that could be put towards better responsiveness, or at the least less power usage. Signed-off-by: Patrick Pannuto <ppannuto@codeaurora.org> Cc: apw@canonical.com Cc: corbet@lwn.net Cc: arjan@linux.intel.com Cc: Randy Dunlap <rdunlap@xenotime.net> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2010-08-03 02:01:04 +04:00
* @min: Minimum time in usecs to sleep
* @max: Maximum time in usecs to sleep
*
* In non-atomic context where the exact wakeup time is flexible, use
* usleep_range() instead of udelay(). The sleep improves responsiveness
* by avoiding the CPU-hogging busy-wait of udelay(), and the range reduces
* power usage by allowing hrtimers to take advantage of an already-
* scheduled interrupt instead of scheduling a new one just for this sleep.
timer: Added usleep_range timer usleep_range is a finer precision implementations of msleep and is designed to be a drop-in replacement for udelay where a precise sleep / busy-wait is unnecessary. Since an easy interface to hrtimers could lead to an undesired proliferation of interrupts, we provide only a "range" API, forcing the caller to think about an acceptable tolerance on both ends and hopefully avoiding introducing another interrupt. INTRO As discussed here ( http://lkml.org/lkml/2007/8/3/250 ), msleep(1) is not precise enough for many drivers (yes, sleep precision is an unfair notion, but consistently sleeping for ~an order of magnitude greater than requested is worth fixing). This patch adds a usleep API so that udelay does not have to be used. Obviously not every udelay can be replaced (those in atomic contexts or being used for simple bitbanging come to mind), but there are many, many examples of mydriver_write(...) /* Wait for hardware to latch */ udelay(100) in various drivers where a busy-wait loop is neither beneficial nor necessary, but msleep simply does not provide enough precision and people are using a busy-wait loop instead. CONCERNS FROM THE RFC Why is udelay a problem / necessary? Most callers of udelay are in device/ driver initialization code, which is serial... As I see it, there is only benefit to sleeping over a delay; the notion of "refactoring" areas that use udelay was presented, but I see usleep as the refactoring. Consider i2c, if the bus is busy, you need to wait a bit (say 100us) before trying again, your current options are: * udelay(100) * msleep(1) <-- As noted above, actually as high as ~20ms on some platforms, so not really an option * Manually set up an hrtimer to try again in 100us (which is what usleep does anyway...) People choose the udelay route because it is EASY; we need to provide a better easy route. Device / driver / boot code is *currently* serial, but every few months someone makes noise about parallelizing boot, and IMHO, a little forward-thinking now is one less thing to worry about if/when that ever happens udelay's could be preempted Sure, but if udelay plans on looping 1000 times, and it gets preempted on loop 200, whenever it's scheduled again, it is going to do the next 800 loops. Is the interruptible case needed? Probably not, but I see usleep as a very logical parallel to msleep, so it made sense to include the "full" API. Processors are getting faster (albeit not as quickly as they are becoming more parallel), so if someone wanted to be interruptible for a few usecs, why not let them? If this is a contentious point, I'm happy to remove it. OTHER THOUGHTS I believe there is also value in exposing the usleep_range option; it gives the scheduler a lot more flexibility and allows the programmer to express his intent much more clearly; it's something I would hope future driver writers will take advantage of. To get the results in the NUMBERS section below, I literally s/udelay/usleep the kernel tree; I had to go in and undo the changes to the USB drivers, but everything else booted successfully; I find that extremely telling in and of itself -- many people are using a delay API where a sleep will suit them just fine. SOME ATTEMPTS AT NUMBERS It turns out that calculating quantifiable benefit on this is challenging, so instead I will simply present the current state of things, and I hope this to be sufficient: How many udelay calls are there in 2.6.35-rc5? udealy(ARG) >= | COUNT 1000 | 319 500 | 414 100 | 1146 20 | 1832 I am working on Android, so that is my focus for this. The following table is a modified usleep that simply printk's the amount of time requested to sleep; these tests were run on a kernel with udelay >= 20 --> usleep "boot" is power-on to lock screen "power collapse" is when the power button is pushed and the device suspends "resume" is when the power button is pushed and the lock screen is displayed (no touchscreen events or anything, just turning on the display) "use device" is from the unlock swipe to clicking around a bit; there is no sd card in this phone, so fail loading music, video, camera ACTION | TOTAL NUMBER OF USLEEP CALLS | NET TIME (us) boot | 22 | 1250 power-collapse | 9 | 1200 resume | 5 | 500 use device | 59 | 7700 The most interesting category to me is the "use device" field; 7700us of busy-wait time that could be put towards better responsiveness, or at the least less power usage. Signed-off-by: Patrick Pannuto <ppannuto@codeaurora.org> Cc: apw@canonical.com Cc: corbet@lwn.net Cc: arjan@linux.intel.com Cc: Randy Dunlap <rdunlap@xenotime.net> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2010-08-03 02:01:04 +04:00
*/
void __sched usleep_range(unsigned long min, unsigned long max)
timer: Added usleep_range timer usleep_range is a finer precision implementations of msleep and is designed to be a drop-in replacement for udelay where a precise sleep / busy-wait is unnecessary. Since an easy interface to hrtimers could lead to an undesired proliferation of interrupts, we provide only a "range" API, forcing the caller to think about an acceptable tolerance on both ends and hopefully avoiding introducing another interrupt. INTRO As discussed here ( http://lkml.org/lkml/2007/8/3/250 ), msleep(1) is not precise enough for many drivers (yes, sleep precision is an unfair notion, but consistently sleeping for ~an order of magnitude greater than requested is worth fixing). This patch adds a usleep API so that udelay does not have to be used. Obviously not every udelay can be replaced (those in atomic contexts or being used for simple bitbanging come to mind), but there are many, many examples of mydriver_write(...) /* Wait for hardware to latch */ udelay(100) in various drivers where a busy-wait loop is neither beneficial nor necessary, but msleep simply does not provide enough precision and people are using a busy-wait loop instead. CONCERNS FROM THE RFC Why is udelay a problem / necessary? Most callers of udelay are in device/ driver initialization code, which is serial... As I see it, there is only benefit to sleeping over a delay; the notion of "refactoring" areas that use udelay was presented, but I see usleep as the refactoring. Consider i2c, if the bus is busy, you need to wait a bit (say 100us) before trying again, your current options are: * udelay(100) * msleep(1) <-- As noted above, actually as high as ~20ms on some platforms, so not really an option * Manually set up an hrtimer to try again in 100us (which is what usleep does anyway...) People choose the udelay route because it is EASY; we need to provide a better easy route. Device / driver / boot code is *currently* serial, but every few months someone makes noise about parallelizing boot, and IMHO, a little forward-thinking now is one less thing to worry about if/when that ever happens udelay's could be preempted Sure, but if udelay plans on looping 1000 times, and it gets preempted on loop 200, whenever it's scheduled again, it is going to do the next 800 loops. Is the interruptible case needed? Probably not, but I see usleep as a very logical parallel to msleep, so it made sense to include the "full" API. Processors are getting faster (albeit not as quickly as they are becoming more parallel), so if someone wanted to be interruptible for a few usecs, why not let them? If this is a contentious point, I'm happy to remove it. OTHER THOUGHTS I believe there is also value in exposing the usleep_range option; it gives the scheduler a lot more flexibility and allows the programmer to express his intent much more clearly; it's something I would hope future driver writers will take advantage of. To get the results in the NUMBERS section below, I literally s/udelay/usleep the kernel tree; I had to go in and undo the changes to the USB drivers, but everything else booted successfully; I find that extremely telling in and of itself -- many people are using a delay API where a sleep will suit them just fine. SOME ATTEMPTS AT NUMBERS It turns out that calculating quantifiable benefit on this is challenging, so instead I will simply present the current state of things, and I hope this to be sufficient: How many udelay calls are there in 2.6.35-rc5? udealy(ARG) >= | COUNT 1000 | 319 500 | 414 100 | 1146 20 | 1832 I am working on Android, so that is my focus for this. The following table is a modified usleep that simply printk's the amount of time requested to sleep; these tests were run on a kernel with udelay >= 20 --> usleep "boot" is power-on to lock screen "power collapse" is when the power button is pushed and the device suspends "resume" is when the power button is pushed and the lock screen is displayed (no touchscreen events or anything, just turning on the display) "use device" is from the unlock swipe to clicking around a bit; there is no sd card in this phone, so fail loading music, video, camera ACTION | TOTAL NUMBER OF USLEEP CALLS | NET TIME (us) boot | 22 | 1250 power-collapse | 9 | 1200 resume | 5 | 500 use device | 59 | 7700 The most interesting category to me is the "use device" field; 7700us of busy-wait time that could be put towards better responsiveness, or at the least less power usage. Signed-off-by: Patrick Pannuto <ppannuto@codeaurora.org> Cc: apw@canonical.com Cc: corbet@lwn.net Cc: arjan@linux.intel.com Cc: Randy Dunlap <rdunlap@xenotime.net> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2010-08-03 02:01:04 +04:00
{
timers: Fix usleep_range() in the context of wake_up_process() Users of usleep_range() expect that it will _never_ return in less time than the minimum passed parameter. However, nothing in the code ensures this, when the sleeping task is woken by wake_up_process() or any other mechanism which can wake a task from uninterruptible state. Neither usleep_range() nor schedule_hrtimeout_range*() have any protection against wakeups. schedule_hrtimeout_range*() is designed this way despite the fact that the API documentation does not mention it. msleep() already has code to handle this case since it will loop as long as there was still time left. usleep_range() has no such loop, add it. Presumably this problem was not detected before because usleep_range() is only used in a few places and the function is mostly used in contexts which are not exposed to wakeups of any form. An effort was made to look for users relying on the old behavior by looking for usleep_range() in the same file as wake_up_process(). No problems were found by this search, though it is conceivable that someone could have put the sleep and wakeup in two different files. An effort was made to ask several upstream maintainers if they were aware of people relying on wake_up_process() to wake up usleep_range(). No maintainers were aware of that but they were aware of many people relying on usleep_range() never returning before the minimum. Reported-by: Tao Huang <huangtao@rock-chips.com> Signed-off-by: Douglas Anderson <dianders@chromium.org> Cc: heiko@sntech.de Cc: broonie@kernel.org Cc: briannorris@chromium.org Cc: Andreas Mohr <andi@lisas.de> Cc: linux-rockchip@lists.infradead.org Cc: tony.xie@rock-chips.com Cc: John Stultz <john.stultz@linaro.org> Cc: djkurtz@chromium.org Cc: linux@roeck-us.net Cc: tskd08@gmail.com Link: http://lkml.kernel.org/r/1477065531-30342-1-git-send-email-dianders@chromium.org Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2016-10-21 18:58:50 +03:00
ktime_t exp = ktime_add_us(ktime_get(), min);
u64 delta = (u64)(max - min) * NSEC_PER_USEC;
for (;;) {
__set_current_state(TASK_UNINTERRUPTIBLE);
/* Do not return before the requested sleep time has elapsed */
if (!schedule_hrtimeout_range(&exp, delta, HRTIMER_MODE_ABS))
break;
}
timer: Added usleep_range timer usleep_range is a finer precision implementations of msleep and is designed to be a drop-in replacement for udelay where a precise sleep / busy-wait is unnecessary. Since an easy interface to hrtimers could lead to an undesired proliferation of interrupts, we provide only a "range" API, forcing the caller to think about an acceptable tolerance on both ends and hopefully avoiding introducing another interrupt. INTRO As discussed here ( http://lkml.org/lkml/2007/8/3/250 ), msleep(1) is not precise enough for many drivers (yes, sleep precision is an unfair notion, but consistently sleeping for ~an order of magnitude greater than requested is worth fixing). This patch adds a usleep API so that udelay does not have to be used. Obviously not every udelay can be replaced (those in atomic contexts or being used for simple bitbanging come to mind), but there are many, many examples of mydriver_write(...) /* Wait for hardware to latch */ udelay(100) in various drivers where a busy-wait loop is neither beneficial nor necessary, but msleep simply does not provide enough precision and people are using a busy-wait loop instead. CONCERNS FROM THE RFC Why is udelay a problem / necessary? Most callers of udelay are in device/ driver initialization code, which is serial... As I see it, there is only benefit to sleeping over a delay; the notion of "refactoring" areas that use udelay was presented, but I see usleep as the refactoring. Consider i2c, if the bus is busy, you need to wait a bit (say 100us) before trying again, your current options are: * udelay(100) * msleep(1) <-- As noted above, actually as high as ~20ms on some platforms, so not really an option * Manually set up an hrtimer to try again in 100us (which is what usleep does anyway...) People choose the udelay route because it is EASY; we need to provide a better easy route. Device / driver / boot code is *currently* serial, but every few months someone makes noise about parallelizing boot, and IMHO, a little forward-thinking now is one less thing to worry about if/when that ever happens udelay's could be preempted Sure, but if udelay plans on looping 1000 times, and it gets preempted on loop 200, whenever it's scheduled again, it is going to do the next 800 loops. Is the interruptible case needed? Probably not, but I see usleep as a very logical parallel to msleep, so it made sense to include the "full" API. Processors are getting faster (albeit not as quickly as they are becoming more parallel), so if someone wanted to be interruptible for a few usecs, why not let them? If this is a contentious point, I'm happy to remove it. OTHER THOUGHTS I believe there is also value in exposing the usleep_range option; it gives the scheduler a lot more flexibility and allows the programmer to express his intent much more clearly; it's something I would hope future driver writers will take advantage of. To get the results in the NUMBERS section below, I literally s/udelay/usleep the kernel tree; I had to go in and undo the changes to the USB drivers, but everything else booted successfully; I find that extremely telling in and of itself -- many people are using a delay API where a sleep will suit them just fine. SOME ATTEMPTS AT NUMBERS It turns out that calculating quantifiable benefit on this is challenging, so instead I will simply present the current state of things, and I hope this to be sufficient: How many udelay calls are there in 2.6.35-rc5? udealy(ARG) >= | COUNT 1000 | 319 500 | 414 100 | 1146 20 | 1832 I am working on Android, so that is my focus for this. The following table is a modified usleep that simply printk's the amount of time requested to sleep; these tests were run on a kernel with udelay >= 20 --> usleep "boot" is power-on to lock screen "power collapse" is when the power button is pushed and the device suspends "resume" is when the power button is pushed and the lock screen is displayed (no touchscreen events or anything, just turning on the display) "use device" is from the unlock swipe to clicking around a bit; there is no sd card in this phone, so fail loading music, video, camera ACTION | TOTAL NUMBER OF USLEEP CALLS | NET TIME (us) boot | 22 | 1250 power-collapse | 9 | 1200 resume | 5 | 500 use device | 59 | 7700 The most interesting category to me is the "use device" field; 7700us of busy-wait time that could be put towards better responsiveness, or at the least less power usage. Signed-off-by: Patrick Pannuto <ppannuto@codeaurora.org> Cc: apw@canonical.com Cc: corbet@lwn.net Cc: arjan@linux.intel.com Cc: Randy Dunlap <rdunlap@xenotime.net> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2010-08-03 02:01:04 +04:00
}
EXPORT_SYMBOL(usleep_range);