WSL2-Linux-Kernel/drivers/cpufreq/cpufreq.c

2683 строки
69 KiB
C
Исходник Обычный вид История

/*
* linux/drivers/cpufreq/cpufreq.c
*
* Copyright (C) 2001 Russell King
* (C) 2002 - 2003 Dominik Brodowski <linux@brodo.de>
* (C) 2013 Viresh Kumar <viresh.kumar@linaro.org>
*
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 01:59:54 +03:00
* Oct 2005 - Ashok Raj <ashok.raj@intel.com>
* Added handling for CPU hotplug
* Feb 2006 - Jacob Shin <jacob.shin@amd.com>
* Fix handling for CPU hotplug -- affected CPUs
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 01:59:54 +03:00
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#include <linux/cpu.h>
#include <linux/cpufreq.h>
#include <linux/cpu_cooling.h>
#include <linux/delay.h>
#include <linux/device.h>
#include <linux/init.h>
#include <linux/kernel_stat.h>
#include <linux/module.h>
#include <linux/mutex.h>
#include <linux/slab.h>
#include <linux/suspend.h>
#include <linux/syscore_ops.h>
#include <linux/tick.h>
#include <trace/events/power.h>
static LIST_HEAD(cpufreq_policy_list);
/* Macros to iterate over CPU policies */
#define for_each_suitable_policy(__policy, __active) \
list_for_each_entry(__policy, &cpufreq_policy_list, policy_list) \
if ((__active) == !policy_is_inactive(__policy))
#define for_each_active_policy(__policy) \
for_each_suitable_policy(__policy, true)
#define for_each_inactive_policy(__policy) \
for_each_suitable_policy(__policy, false)
#define for_each_policy(__policy) \
list_for_each_entry(__policy, &cpufreq_policy_list, policy_list)
/* Iterate over governors */
static LIST_HEAD(cpufreq_governor_list);
#define for_each_governor(__governor) \
list_for_each_entry(__governor, &cpufreq_governor_list, governor_list)
/**
* The "cpufreq driver" - the arch- or hardware-dependent low
* level driver of CPUFreq support, and its spinlock. This lock
* also protects the cpufreq_cpu_data array.
*/
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
static struct cpufreq_driver *cpufreq_driver;
static DEFINE_PER_CPU(struct cpufreq_policy *, cpufreq_cpu_data);
static DEFINE_RWLOCK(cpufreq_driver_lock);
/* Flag to suspend/resume CPUFreq governors */
static bool cpufreq_suspended;
cpufreq: Implement light weight ->target_index() routine Currently, the prototype of cpufreq_drivers target routines is: int target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation); And most of the drivers call cpufreq_frequency_table_target() to get a valid index of their frequency table which is closest to the target_freq. And they don't use target_freq and relation after that. So, it makes sense to just do this work in cpufreq core before calling cpufreq_frequency_table_target() and simply pass index instead. But this can be done only with drivers which expose their frequency table with cpufreq core. For others we need to stick with the old prototype of target() until those drivers are converted to expose frequency tables. This patch implements the new light weight prototype for target_index() routine. It looks like this: int target_index(struct cpufreq_policy *policy, unsigned int index); CPUFreq core will call cpufreq_frequency_table_target() before calling this routine and pass index to it. Because CPUFreq core now requires to call routines present in freq_table.c CONFIG_CPU_FREQ_TABLE must be enabled all the time. This also marks target() interface as deprecated. So, that new drivers avoid using it. And Documentation is updated accordingly. It also converts existing .target() to newly defined light weight .target_index() routine for many driver. Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no> Acked-by: Jesper Nilsson <jesper.nilsson@axis.com> Acked-by: Linus Walleij <linus.walleij@linaro.org> Acked-by: Russell King <linux@arm.linux.org.uk> Acked-by: David S. Miller <davem@davemloft.net> Tested-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rjw@rjwysocki.net>
2013-10-25 18:15:48 +04:00
static inline bool has_target(void)
{
return cpufreq_driver->target_index || cpufreq_driver->target;
}
/* internal prototypes */
static unsigned int __cpufreq_get(struct cpufreq_policy *policy);
static int cpufreq_init_governor(struct cpufreq_policy *policy);
static void cpufreq_exit_governor(struct cpufreq_policy *policy);
static int cpufreq_start_governor(struct cpufreq_policy *policy);
static void cpufreq_stop_governor(struct cpufreq_policy *policy);
static void cpufreq_governor_limits(struct cpufreq_policy *policy);
/**
* Two notifier lists: the "policy" list is involved in the
* validation process for a new CPU frequency policy; the
* "transition" list for kernel code that needs to handle
* changes to devices when the CPU clock speed changes.
* The mutex locks both lists.
*/
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 13:16:30 +04:00
static BLOCKING_NOTIFIER_HEAD(cpufreq_policy_notifier_list);
SRCU_NOTIFIER_HEAD_STATIC(cpufreq_transition_notifier_list);
static int off __read_mostly;
static int cpufreq_disabled(void)
{
return off;
}
void disable_cpufreq(void)
{
off = 1;
}
static DEFINE_MUTEX(cpufreq_governor_mutex);
bool have_governor_per_policy(void)
{
return !!(cpufreq_driver->flags & CPUFREQ_HAVE_GOVERNOR_PER_POLICY);
}
EXPORT_SYMBOL_GPL(have_governor_per_policy);
struct kobject *get_governor_parent_kobj(struct cpufreq_policy *policy)
{
if (have_governor_per_policy())
return &policy->kobj;
else
return cpufreq_global_kobject;
}
EXPORT_SYMBOL_GPL(get_governor_parent_kobj);
static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
{
u64 idle_time;
u64 cur_wall_time;
u64 busy_time;
cur_wall_time = jiffies64_to_nsecs(get_jiffies_64());
busy_time = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];
idle_time = cur_wall_time - busy_time;
if (wall)
*wall = div_u64(cur_wall_time, NSEC_PER_USEC);
return div_u64(idle_time, NSEC_PER_USEC);
}
u64 get_cpu_idle_time(unsigned int cpu, u64 *wall, int io_busy)
{
u64 idle_time = get_cpu_idle_time_us(cpu, io_busy ? wall : NULL);
if (idle_time == -1ULL)
return get_cpu_idle_time_jiffy(cpu, wall);
else if (!io_busy)
idle_time += get_cpu_iowait_time_us(cpu, wall);
return idle_time;
}
EXPORT_SYMBOL_GPL(get_cpu_idle_time);
__weak void arch_set_freq_scale(struct cpumask *cpus, unsigned long cur_freq,
unsigned long max_freq)
{
}
EXPORT_SYMBOL_GPL(arch_set_freq_scale);
/*
* This is a generic cpufreq init() routine which can be used by cpufreq
* drivers of SMP systems. It will do following:
* - validate & show freq table passed
* - set policies transition latency
* - policy->cpus with all possible CPUs
*/
int cpufreq_generic_init(struct cpufreq_policy *policy,
struct cpufreq_frequency_table *table,
unsigned int transition_latency)
{
policy->freq_table = table;
policy->cpuinfo.transition_latency = transition_latency;
/*
* The driver only supports the SMP configuration where all processors
* share the clock and voltage and clock.
*/
cpumask_setall(policy->cpus);
return 0;
}
EXPORT_SYMBOL_GPL(cpufreq_generic_init);
struct cpufreq_policy *cpufreq_cpu_get_raw(unsigned int cpu)
{
struct cpufreq_policy *policy = per_cpu(cpufreq_cpu_data, cpu);
return policy && cpumask_test_cpu(cpu, policy->cpus) ? policy : NULL;
}
EXPORT_SYMBOL_GPL(cpufreq_cpu_get_raw);
unsigned int cpufreq_generic_get(unsigned int cpu)
{
struct cpufreq_policy *policy = cpufreq_cpu_get_raw(cpu);
if (!policy || IS_ERR(policy->clk)) {
pr_err("%s: No %s associated to cpu: %d\n",
__func__, policy ? "clk" : "policy", cpu);
return 0;
}
return clk_get_rate(policy->clk) / 1000;
}
EXPORT_SYMBOL_GPL(cpufreq_generic_get);
/**
* cpufreq_cpu_get - Return policy for a CPU and mark it as busy.
* @cpu: CPU to find the policy for.
*
* Call cpufreq_cpu_get_raw() to obtain a cpufreq policy for @cpu and increment
* the kobject reference counter of that policy. Return a valid policy on
* success or NULL on failure.
*
* The policy returned by this function has to be released with the help of
* cpufreq_cpu_put() to balance its kobject reference counter properly.
*/
struct cpufreq_policy *cpufreq_cpu_get(unsigned int cpu)
{
struct cpufreq_policy *policy = NULL;
unsigned long flags;
if (WARN_ON(cpu >= nr_cpu_ids))
return NULL;
/* get the cpufreq driver */
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
read_lock_irqsave(&cpufreq_driver_lock, flags);
if (cpufreq_driver) {
/* get the CPU */
policy = cpufreq_cpu_get_raw(cpu);
if (policy)
kobject_get(&policy->kobj);
}
read_unlock_irqrestore(&cpufreq_driver_lock, flags);
return policy;
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-20 22:14:38 +04:00
}
EXPORT_SYMBOL_GPL(cpufreq_cpu_get);
/**
* cpufreq_cpu_put - Decrement kobject usage counter for cpufreq policy.
* @policy: cpufreq policy returned by cpufreq_cpu_get().
*/
void cpufreq_cpu_put(struct cpufreq_policy *policy)
{
kobject_put(&policy->kobj);
}
EXPORT_SYMBOL_GPL(cpufreq_cpu_put);
/**
* cpufreq_cpu_release - Unlock a policy and decrement its usage counter.
* @policy: cpufreq policy returned by cpufreq_cpu_acquire().
*/
void cpufreq_cpu_release(struct cpufreq_policy *policy)
{
if (WARN_ON(!policy))
return;
lockdep_assert_held(&policy->rwsem);
up_write(&policy->rwsem);
cpufreq_cpu_put(policy);
}
/**
* cpufreq_cpu_acquire - Find policy for a CPU, mark it as busy and lock it.
* @cpu: CPU to find the policy for.
*
* Call cpufreq_cpu_get() to get a reference on the cpufreq policy for @cpu and
* if the policy returned by it is not NULL, acquire its rwsem for writing.
* Return the policy if it is active or release it and return NULL otherwise.
*
* The policy returned by this function has to be released with the help of
* cpufreq_cpu_release() in order to release its rwsem and balance its usage
* counter properly.
*/
struct cpufreq_policy *cpufreq_cpu_acquire(unsigned int cpu)
{
struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
if (!policy)
return NULL;
down_write(&policy->rwsem);
if (policy_is_inactive(policy)) {
cpufreq_cpu_release(policy);
return NULL;
}
return policy;
}
/*********************************************************************
* EXTERNALLY AFFECTING FREQUENCY CHANGES *
*********************************************************************/
/**
* adjust_jiffies - adjust the system "loops_per_jiffy"
*
* This function alters the system "loops_per_jiffy" for the clock
* speed change. Note that loops_per_jiffy cannot be updated on SMP
* systems as each CPU might be scaled differently. So, use the arch
* per-CPU loops_per_jiffy value wherever possible.
*/
static void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci)
{
#ifndef CONFIG_SMP
static unsigned long l_p_j_ref;
static unsigned int l_p_j_ref_freq;
if (ci->flags & CPUFREQ_CONST_LOOPS)
return;
if (!l_p_j_ref_freq) {
l_p_j_ref = loops_per_jiffy;
l_p_j_ref_freq = ci->old;
pr_debug("saving %lu as reference value for loops_per_jiffy; freq is %u kHz\n",
l_p_j_ref, l_p_j_ref_freq);
}
if (val == CPUFREQ_POSTCHANGE && ci->old != ci->new) {
loops_per_jiffy = cpufreq_scale(l_p_j_ref, l_p_j_ref_freq,
ci->new);
pr_debug("scaling loops_per_jiffy to %lu for frequency %u kHz\n",
loops_per_jiffy, ci->new);
}
#endif
}
/**
* cpufreq_notify_transition - Notify frequency transition and adjust_jiffies.
* @policy: cpufreq policy to enable fast frequency switching for.
* @freqs: contain details of the frequency update.
* @state: set to CPUFREQ_PRECHANGE or CPUFREQ_POSTCHANGE.
*
* This function calls the transition notifiers and the "adjust_jiffies"
* function. It is called twice on all CPU frequency changes that have
* external effects.
*/
static void cpufreq_notify_transition(struct cpufreq_policy *policy,
struct cpufreq_freqs *freqs,
unsigned int state)
{
BUG_ON(irqs_disabled());
if (cpufreq_disabled())
return;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
freqs->flags = cpufreq_driver->flags;
pr_debug("notification %u of frequency transition to %u kHz\n",
state, freqs->new);
switch (state) {
case CPUFREQ_PRECHANGE:
/*
* Detect if the driver reported a value as "old frequency"
* which is not equal to what the cpufreq core thinks is
* "old frequency".
*/
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
if (!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) {
if (policy->cur && (policy->cur != freqs->old)) {
pr_debug("Warning: CPU frequency is %u, cpufreq assumed %u kHz\n",
freqs->old, policy->cur);
freqs->old = policy->cur;
}
}
for_each_cpu(freqs->cpu, policy->cpus) {
srcu_notifier_call_chain(&cpufreq_transition_notifier_list,
CPUFREQ_PRECHANGE, freqs);
}
adjust_jiffies(CPUFREQ_PRECHANGE, freqs);
break;
case CPUFREQ_POSTCHANGE:
adjust_jiffies(CPUFREQ_POSTCHANGE, freqs);
pr_debug("FREQ: %u - CPUs: %*pbl\n", freqs->new,
cpumask_pr_args(policy->cpus));
for_each_cpu(freqs->cpu, policy->cpus) {
trace_cpu_frequency(freqs->new, freqs->cpu);
srcu_notifier_call_chain(&cpufreq_transition_notifier_list,
CPUFREQ_POSTCHANGE, freqs);
}
cpufreq_stats_record_transition(policy, freqs->new);
policy->cur = freqs->new;
}
}
/* Do post notifications when there are chances that transition has failed */
static void cpufreq_notify_post_transition(struct cpufreq_policy *policy,
struct cpufreq_freqs *freqs, int transition_failed)
{
cpufreq_notify_transition(policy, freqs, CPUFREQ_POSTCHANGE);
if (!transition_failed)
return;
swap(freqs->old, freqs->new);
cpufreq_notify_transition(policy, freqs, CPUFREQ_PRECHANGE);
cpufreq_notify_transition(policy, freqs, CPUFREQ_POSTCHANGE);
}
cpufreq: Make sure frequency transitions are serialized Whenever we change the frequency of a CPU, we call the PRECHANGE and POSTCHANGE notifiers. They must be serialized, i.e. PRECHANGE and POSTCHANGE notifiers should strictly alternate, thereby preventing two different sets of PRECHANGE or POSTCHANGE notifiers from interleaving arbitrarily. The following examples illustrate why this is important: Scenario 1: ----------- A thread reading the value of cpuinfo_cur_freq, will call __cpufreq_cpu_get()->cpufreq_out_of_sync()->cpufreq_notify_transition() The ondemand governor can decide to change the frequency of the CPU at the same time and hence it can end up sending the notifications via ->target(). If the notifiers are not serialized, the following sequence can occur: - PRECHANGE Notification for freq A (from cpuinfo_cur_freq) - PRECHANGE Notification for freq B (from target()) - Freq changed by target() to B - POSTCHANGE Notification for freq B - POSTCHANGE Notification for freq A We can see from the above that the last POSTCHANGE Notification happens for freq A but the hardware is set to run at freq B. Where would we break then?: adjust_jiffies() in cpufreq.c & cpufreq_callback() in arch/arm/kernel/smp.c (which also adjusts the jiffies). All the loops_per_jiffy calculations will get messed up. Scenario 2: ----------- The governor calls __cpufreq_driver_target() to change the frequency. At the same time, if we change scaling_{min|max}_freq from sysfs, it will end up calling the governor's CPUFREQ_GOV_LIMITS notification, which will also call __cpufreq_driver_target(). And hence we end up issuing concurrent calls to ->target(). Typically, platforms have the following logic in their ->target() routines: (Eg: cpufreq-cpu0, omap, exynos, etc) A. If new freq is more than old: Increase voltage B. Change freq C. If new freq is less than old: decrease voltage Now, if the two concurrent calls to ->target() are X and Y, where X is trying to increase the freq and Y is trying to decrease it, we get the following race condition: X.A: voltage gets increased for larger freq Y.A: nothing happens Y.B: freq gets decreased Y.C: voltage gets decreased X.B: freq gets increased X.C: nothing happens Thus we can end up setting a freq which is not supported by the voltage we have set. That will probably make the clock to the CPU unstable and the system might not work properly anymore. This patch introduces a set of synchronization primitives to serialize frequency transitions, which are to be used as shown below: cpufreq_freq_transition_begin(); //Perform the frequency change cpufreq_freq_transition_end(); The _begin() call sends the PRECHANGE notification whereas the _end() call sends the POSTCHANGE notification. Also, all the necessary synchronization is handled within these calls. In particular, even drivers which set the ASYNC_NOTIFICATION flag can also use these APIs for performing frequency transitions (ie., you can call _begin() from one task, and call the corresponding _end() from a different task). The actual synchronization underneath is not that complicated: The key challenge is to allow drivers to begin the transition from one thread and end it in a completely different thread (this is to enable drivers that do asynchronous POSTCHANGE notification from bottom-halves, to also use the same interface). To achieve this, a 'transition_ongoing' flag, a 'transition_lock' spinlock and a wait-queue are added per-policy. The flag and the wait-queue are used in conjunction to create an "uninterrupted flow" from _begin() to _end(). The spinlock is used to ensure that only one such "flow" is in flight at any given time. Put together, this provides us all the necessary synchronization. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-03-24 12:05:44 +04:00
void cpufreq_freq_transition_begin(struct cpufreq_policy *policy,
struct cpufreq_freqs *freqs)
{
cpufreq: Catch double invocations of cpufreq_freq_transition_begin/end Some cpufreq drivers were redundantly invoking the _begin() and _end() APIs around frequency transitions, and this double invocation (one from the cpufreq core and the other from the cpufreq driver) used to result in a self-deadlock, leading to system hangs during boot. (The _begin() API makes contending callers wait until the previous invocation is complete. Hence, the cpufreq driver would end up waiting on itself!). Now all such drivers have been fixed, but debugging this issue was not very straight-forward (even lockdep didn't catch this). So let us add a debug infrastructure to the cpufreq core to catch such issues more easily in the future. We add a new field called 'transition_task' to the policy structure, to keep track of the task which is performing the frequency transition. Using this field, we make note of this task during _begin() and print a warning if we find a case where the same task is calling _begin() again, before completing the previous frequency transition using the corresponding _end(). We have left out ASYNC_NOTIFICATION drivers from this debug infrastructure for 2 reasons: 1. At the moment, we have no way to avoid a particular scenario where this debug infrastructure can emit false-positive warnings for such drivers. The scenario is depicted below: Task A Task B /* 1st freq transition */ Invoke _begin() { ... ... } Change the frequency /* 2nd freq transition */ Invoke _begin() { ... //waiting for B to ... //finish _end() for ... //the 1st transition ... | Got interrupt for successful ... | change of frequency (1st one). ... | ... | /* 1st freq transition */ ... | Invoke _end() { ... | ... ... V } ... ... } This scenario is actually deadlock-free because, once Task A changes the frequency, it is Task B's responsibility to invoke the corresponding _end() for the 1st frequency transition. Hence it is perfectly legal for Task A to go ahead and attempt another frequency transition in the meantime. (Of course it won't be able to proceed until Task B finishes the 1st _end(), but this doesn't cause a deadlock or a hang). The debug infrastructure cannot handle this scenario and will treat it as a deadlock and print a warning. To avoid this, we exclude such drivers from the purview of this code. 2. Luckily, we don't _need_ this infrastructure for ASYNC_NOTIFICATION drivers at all! The cpufreq core does not automatically invoke the _begin() and _end() APIs during frequency transitions in such drivers. Thus, the driver alone is responsible for invoking _begin()/_end() and hence there shouldn't be any conflicts which lead to double invocations. So, we can skip these drivers, since the probability that such drivers will hit this problem is extremely low, as outlined above. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-05-05 11:22:39 +04:00
/*
* Catch double invocations of _begin() which lead to self-deadlock.
* ASYNC_NOTIFICATION drivers are left out because the cpufreq core
* doesn't invoke _begin() on their behalf, and hence the chances of
* double invocations are very low. Moreover, there are scenarios
* where these checks can emit false-positive warnings in these
* drivers; so we avoid that by skipping them altogether.
*/
WARN_ON(!(cpufreq_driver->flags & CPUFREQ_ASYNC_NOTIFICATION)
&& current == policy->transition_task);
cpufreq: Make sure frequency transitions are serialized Whenever we change the frequency of a CPU, we call the PRECHANGE and POSTCHANGE notifiers. They must be serialized, i.e. PRECHANGE and POSTCHANGE notifiers should strictly alternate, thereby preventing two different sets of PRECHANGE or POSTCHANGE notifiers from interleaving arbitrarily. The following examples illustrate why this is important: Scenario 1: ----------- A thread reading the value of cpuinfo_cur_freq, will call __cpufreq_cpu_get()->cpufreq_out_of_sync()->cpufreq_notify_transition() The ondemand governor can decide to change the frequency of the CPU at the same time and hence it can end up sending the notifications via ->target(). If the notifiers are not serialized, the following sequence can occur: - PRECHANGE Notification for freq A (from cpuinfo_cur_freq) - PRECHANGE Notification for freq B (from target()) - Freq changed by target() to B - POSTCHANGE Notification for freq B - POSTCHANGE Notification for freq A We can see from the above that the last POSTCHANGE Notification happens for freq A but the hardware is set to run at freq B. Where would we break then?: adjust_jiffies() in cpufreq.c & cpufreq_callback() in arch/arm/kernel/smp.c (which also adjusts the jiffies). All the loops_per_jiffy calculations will get messed up. Scenario 2: ----------- The governor calls __cpufreq_driver_target() to change the frequency. At the same time, if we change scaling_{min|max}_freq from sysfs, it will end up calling the governor's CPUFREQ_GOV_LIMITS notification, which will also call __cpufreq_driver_target(). And hence we end up issuing concurrent calls to ->target(). Typically, platforms have the following logic in their ->target() routines: (Eg: cpufreq-cpu0, omap, exynos, etc) A. If new freq is more than old: Increase voltage B. Change freq C. If new freq is less than old: decrease voltage Now, if the two concurrent calls to ->target() are X and Y, where X is trying to increase the freq and Y is trying to decrease it, we get the following race condition: X.A: voltage gets increased for larger freq Y.A: nothing happens Y.B: freq gets decreased Y.C: voltage gets decreased X.B: freq gets increased X.C: nothing happens Thus we can end up setting a freq which is not supported by the voltage we have set. That will probably make the clock to the CPU unstable and the system might not work properly anymore. This patch introduces a set of synchronization primitives to serialize frequency transitions, which are to be used as shown below: cpufreq_freq_transition_begin(); //Perform the frequency change cpufreq_freq_transition_end(); The _begin() call sends the PRECHANGE notification whereas the _end() call sends the POSTCHANGE notification. Also, all the necessary synchronization is handled within these calls. In particular, even drivers which set the ASYNC_NOTIFICATION flag can also use these APIs for performing frequency transitions (ie., you can call _begin() from one task, and call the corresponding _end() from a different task). The actual synchronization underneath is not that complicated: The key challenge is to allow drivers to begin the transition from one thread and end it in a completely different thread (this is to enable drivers that do asynchronous POSTCHANGE notification from bottom-halves, to also use the same interface). To achieve this, a 'transition_ongoing' flag, a 'transition_lock' spinlock and a wait-queue are added per-policy. The flag and the wait-queue are used in conjunction to create an "uninterrupted flow" from _begin() to _end(). The spinlock is used to ensure that only one such "flow" is in flight at any given time. Put together, this provides us all the necessary synchronization. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-03-24 12:05:44 +04:00
wait:
wait_event(policy->transition_wait, !policy->transition_ongoing);
spin_lock(&policy->transition_lock);
if (unlikely(policy->transition_ongoing)) {
spin_unlock(&policy->transition_lock);
goto wait;
}
policy->transition_ongoing = true;
cpufreq: Catch double invocations of cpufreq_freq_transition_begin/end Some cpufreq drivers were redundantly invoking the _begin() and _end() APIs around frequency transitions, and this double invocation (one from the cpufreq core and the other from the cpufreq driver) used to result in a self-deadlock, leading to system hangs during boot. (The _begin() API makes contending callers wait until the previous invocation is complete. Hence, the cpufreq driver would end up waiting on itself!). Now all such drivers have been fixed, but debugging this issue was not very straight-forward (even lockdep didn't catch this). So let us add a debug infrastructure to the cpufreq core to catch such issues more easily in the future. We add a new field called 'transition_task' to the policy structure, to keep track of the task which is performing the frequency transition. Using this field, we make note of this task during _begin() and print a warning if we find a case where the same task is calling _begin() again, before completing the previous frequency transition using the corresponding _end(). We have left out ASYNC_NOTIFICATION drivers from this debug infrastructure for 2 reasons: 1. At the moment, we have no way to avoid a particular scenario where this debug infrastructure can emit false-positive warnings for such drivers. The scenario is depicted below: Task A Task B /* 1st freq transition */ Invoke _begin() { ... ... } Change the frequency /* 2nd freq transition */ Invoke _begin() { ... //waiting for B to ... //finish _end() for ... //the 1st transition ... | Got interrupt for successful ... | change of frequency (1st one). ... | ... | /* 1st freq transition */ ... | Invoke _end() { ... | ... ... V } ... ... } This scenario is actually deadlock-free because, once Task A changes the frequency, it is Task B's responsibility to invoke the corresponding _end() for the 1st frequency transition. Hence it is perfectly legal for Task A to go ahead and attempt another frequency transition in the meantime. (Of course it won't be able to proceed until Task B finishes the 1st _end(), but this doesn't cause a deadlock or a hang). The debug infrastructure cannot handle this scenario and will treat it as a deadlock and print a warning. To avoid this, we exclude such drivers from the purview of this code. 2. Luckily, we don't _need_ this infrastructure for ASYNC_NOTIFICATION drivers at all! The cpufreq core does not automatically invoke the _begin() and _end() APIs during frequency transitions in such drivers. Thus, the driver alone is responsible for invoking _begin()/_end() and hence there shouldn't be any conflicts which lead to double invocations. So, we can skip these drivers, since the probability that such drivers will hit this problem is extremely low, as outlined above. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-05-05 11:22:39 +04:00
policy->transition_task = current;
cpufreq: Make sure frequency transitions are serialized Whenever we change the frequency of a CPU, we call the PRECHANGE and POSTCHANGE notifiers. They must be serialized, i.e. PRECHANGE and POSTCHANGE notifiers should strictly alternate, thereby preventing two different sets of PRECHANGE or POSTCHANGE notifiers from interleaving arbitrarily. The following examples illustrate why this is important: Scenario 1: ----------- A thread reading the value of cpuinfo_cur_freq, will call __cpufreq_cpu_get()->cpufreq_out_of_sync()->cpufreq_notify_transition() The ondemand governor can decide to change the frequency of the CPU at the same time and hence it can end up sending the notifications via ->target(). If the notifiers are not serialized, the following sequence can occur: - PRECHANGE Notification for freq A (from cpuinfo_cur_freq) - PRECHANGE Notification for freq B (from target()) - Freq changed by target() to B - POSTCHANGE Notification for freq B - POSTCHANGE Notification for freq A We can see from the above that the last POSTCHANGE Notification happens for freq A but the hardware is set to run at freq B. Where would we break then?: adjust_jiffies() in cpufreq.c & cpufreq_callback() in arch/arm/kernel/smp.c (which also adjusts the jiffies). All the loops_per_jiffy calculations will get messed up. Scenario 2: ----------- The governor calls __cpufreq_driver_target() to change the frequency. At the same time, if we change scaling_{min|max}_freq from sysfs, it will end up calling the governor's CPUFREQ_GOV_LIMITS notification, which will also call __cpufreq_driver_target(). And hence we end up issuing concurrent calls to ->target(). Typically, platforms have the following logic in their ->target() routines: (Eg: cpufreq-cpu0, omap, exynos, etc) A. If new freq is more than old: Increase voltage B. Change freq C. If new freq is less than old: decrease voltage Now, if the two concurrent calls to ->target() are X and Y, where X is trying to increase the freq and Y is trying to decrease it, we get the following race condition: X.A: voltage gets increased for larger freq Y.A: nothing happens Y.B: freq gets decreased Y.C: voltage gets decreased X.B: freq gets increased X.C: nothing happens Thus we can end up setting a freq which is not supported by the voltage we have set. That will probably make the clock to the CPU unstable and the system might not work properly anymore. This patch introduces a set of synchronization primitives to serialize frequency transitions, which are to be used as shown below: cpufreq_freq_transition_begin(); //Perform the frequency change cpufreq_freq_transition_end(); The _begin() call sends the PRECHANGE notification whereas the _end() call sends the POSTCHANGE notification. Also, all the necessary synchronization is handled within these calls. In particular, even drivers which set the ASYNC_NOTIFICATION flag can also use these APIs for performing frequency transitions (ie., you can call _begin() from one task, and call the corresponding _end() from a different task). The actual synchronization underneath is not that complicated: The key challenge is to allow drivers to begin the transition from one thread and end it in a completely different thread (this is to enable drivers that do asynchronous POSTCHANGE notification from bottom-halves, to also use the same interface). To achieve this, a 'transition_ongoing' flag, a 'transition_lock' spinlock and a wait-queue are added per-policy. The flag and the wait-queue are used in conjunction to create an "uninterrupted flow" from _begin() to _end(). The spinlock is used to ensure that only one such "flow" is in flight at any given time. Put together, this provides us all the necessary synchronization. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-03-24 12:05:44 +04:00
spin_unlock(&policy->transition_lock);
cpufreq_notify_transition(policy, freqs, CPUFREQ_PRECHANGE);
}
EXPORT_SYMBOL_GPL(cpufreq_freq_transition_begin);
void cpufreq_freq_transition_end(struct cpufreq_policy *policy,
struct cpufreq_freqs *freqs, int transition_failed)
{
if (WARN_ON(!policy->transition_ongoing))
cpufreq: Make sure frequency transitions are serialized Whenever we change the frequency of a CPU, we call the PRECHANGE and POSTCHANGE notifiers. They must be serialized, i.e. PRECHANGE and POSTCHANGE notifiers should strictly alternate, thereby preventing two different sets of PRECHANGE or POSTCHANGE notifiers from interleaving arbitrarily. The following examples illustrate why this is important: Scenario 1: ----------- A thread reading the value of cpuinfo_cur_freq, will call __cpufreq_cpu_get()->cpufreq_out_of_sync()->cpufreq_notify_transition() The ondemand governor can decide to change the frequency of the CPU at the same time and hence it can end up sending the notifications via ->target(). If the notifiers are not serialized, the following sequence can occur: - PRECHANGE Notification for freq A (from cpuinfo_cur_freq) - PRECHANGE Notification for freq B (from target()) - Freq changed by target() to B - POSTCHANGE Notification for freq B - POSTCHANGE Notification for freq A We can see from the above that the last POSTCHANGE Notification happens for freq A but the hardware is set to run at freq B. Where would we break then?: adjust_jiffies() in cpufreq.c & cpufreq_callback() in arch/arm/kernel/smp.c (which also adjusts the jiffies). All the loops_per_jiffy calculations will get messed up. Scenario 2: ----------- The governor calls __cpufreq_driver_target() to change the frequency. At the same time, if we change scaling_{min|max}_freq from sysfs, it will end up calling the governor's CPUFREQ_GOV_LIMITS notification, which will also call __cpufreq_driver_target(). And hence we end up issuing concurrent calls to ->target(). Typically, platforms have the following logic in their ->target() routines: (Eg: cpufreq-cpu0, omap, exynos, etc) A. If new freq is more than old: Increase voltage B. Change freq C. If new freq is less than old: decrease voltage Now, if the two concurrent calls to ->target() are X and Y, where X is trying to increase the freq and Y is trying to decrease it, we get the following race condition: X.A: voltage gets increased for larger freq Y.A: nothing happens Y.B: freq gets decreased Y.C: voltage gets decreased X.B: freq gets increased X.C: nothing happens Thus we can end up setting a freq which is not supported by the voltage we have set. That will probably make the clock to the CPU unstable and the system might not work properly anymore. This patch introduces a set of synchronization primitives to serialize frequency transitions, which are to be used as shown below: cpufreq_freq_transition_begin(); //Perform the frequency change cpufreq_freq_transition_end(); The _begin() call sends the PRECHANGE notification whereas the _end() call sends the POSTCHANGE notification. Also, all the necessary synchronization is handled within these calls. In particular, even drivers which set the ASYNC_NOTIFICATION flag can also use these APIs for performing frequency transitions (ie., you can call _begin() from one task, and call the corresponding _end() from a different task). The actual synchronization underneath is not that complicated: The key challenge is to allow drivers to begin the transition from one thread and end it in a completely different thread (this is to enable drivers that do asynchronous POSTCHANGE notification from bottom-halves, to also use the same interface). To achieve this, a 'transition_ongoing' flag, a 'transition_lock' spinlock and a wait-queue are added per-policy. The flag and the wait-queue are used in conjunction to create an "uninterrupted flow" from _begin() to _end(). The spinlock is used to ensure that only one such "flow" is in flight at any given time. Put together, this provides us all the necessary synchronization. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-03-24 12:05:44 +04:00
return;
cpufreq_notify_post_transition(policy, freqs, transition_failed);
policy->transition_ongoing = false;
cpufreq: Catch double invocations of cpufreq_freq_transition_begin/end Some cpufreq drivers were redundantly invoking the _begin() and _end() APIs around frequency transitions, and this double invocation (one from the cpufreq core and the other from the cpufreq driver) used to result in a self-deadlock, leading to system hangs during boot. (The _begin() API makes contending callers wait until the previous invocation is complete. Hence, the cpufreq driver would end up waiting on itself!). Now all such drivers have been fixed, but debugging this issue was not very straight-forward (even lockdep didn't catch this). So let us add a debug infrastructure to the cpufreq core to catch such issues more easily in the future. We add a new field called 'transition_task' to the policy structure, to keep track of the task which is performing the frequency transition. Using this field, we make note of this task during _begin() and print a warning if we find a case where the same task is calling _begin() again, before completing the previous frequency transition using the corresponding _end(). We have left out ASYNC_NOTIFICATION drivers from this debug infrastructure for 2 reasons: 1. At the moment, we have no way to avoid a particular scenario where this debug infrastructure can emit false-positive warnings for such drivers. The scenario is depicted below: Task A Task B /* 1st freq transition */ Invoke _begin() { ... ... } Change the frequency /* 2nd freq transition */ Invoke _begin() { ... //waiting for B to ... //finish _end() for ... //the 1st transition ... | Got interrupt for successful ... | change of frequency (1st one). ... | ... | /* 1st freq transition */ ... | Invoke _end() { ... | ... ... V } ... ... } This scenario is actually deadlock-free because, once Task A changes the frequency, it is Task B's responsibility to invoke the corresponding _end() for the 1st frequency transition. Hence it is perfectly legal for Task A to go ahead and attempt another frequency transition in the meantime. (Of course it won't be able to proceed until Task B finishes the 1st _end(), but this doesn't cause a deadlock or a hang). The debug infrastructure cannot handle this scenario and will treat it as a deadlock and print a warning. To avoid this, we exclude such drivers from the purview of this code. 2. Luckily, we don't _need_ this infrastructure for ASYNC_NOTIFICATION drivers at all! The cpufreq core does not automatically invoke the _begin() and _end() APIs during frequency transitions in such drivers. Thus, the driver alone is responsible for invoking _begin()/_end() and hence there shouldn't be any conflicts which lead to double invocations. So, we can skip these drivers, since the probability that such drivers will hit this problem is extremely low, as outlined above. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-05-05 11:22:39 +04:00
policy->transition_task = NULL;
cpufreq: Make sure frequency transitions are serialized Whenever we change the frequency of a CPU, we call the PRECHANGE and POSTCHANGE notifiers. They must be serialized, i.e. PRECHANGE and POSTCHANGE notifiers should strictly alternate, thereby preventing two different sets of PRECHANGE or POSTCHANGE notifiers from interleaving arbitrarily. The following examples illustrate why this is important: Scenario 1: ----------- A thread reading the value of cpuinfo_cur_freq, will call __cpufreq_cpu_get()->cpufreq_out_of_sync()->cpufreq_notify_transition() The ondemand governor can decide to change the frequency of the CPU at the same time and hence it can end up sending the notifications via ->target(). If the notifiers are not serialized, the following sequence can occur: - PRECHANGE Notification for freq A (from cpuinfo_cur_freq) - PRECHANGE Notification for freq B (from target()) - Freq changed by target() to B - POSTCHANGE Notification for freq B - POSTCHANGE Notification for freq A We can see from the above that the last POSTCHANGE Notification happens for freq A but the hardware is set to run at freq B. Where would we break then?: adjust_jiffies() in cpufreq.c & cpufreq_callback() in arch/arm/kernel/smp.c (which also adjusts the jiffies). All the loops_per_jiffy calculations will get messed up. Scenario 2: ----------- The governor calls __cpufreq_driver_target() to change the frequency. At the same time, if we change scaling_{min|max}_freq from sysfs, it will end up calling the governor's CPUFREQ_GOV_LIMITS notification, which will also call __cpufreq_driver_target(). And hence we end up issuing concurrent calls to ->target(). Typically, platforms have the following logic in their ->target() routines: (Eg: cpufreq-cpu0, omap, exynos, etc) A. If new freq is more than old: Increase voltage B. Change freq C. If new freq is less than old: decrease voltage Now, if the two concurrent calls to ->target() are X and Y, where X is trying to increase the freq and Y is trying to decrease it, we get the following race condition: X.A: voltage gets increased for larger freq Y.A: nothing happens Y.B: freq gets decreased Y.C: voltage gets decreased X.B: freq gets increased X.C: nothing happens Thus we can end up setting a freq which is not supported by the voltage we have set. That will probably make the clock to the CPU unstable and the system might not work properly anymore. This patch introduces a set of synchronization primitives to serialize frequency transitions, which are to be used as shown below: cpufreq_freq_transition_begin(); //Perform the frequency change cpufreq_freq_transition_end(); The _begin() call sends the PRECHANGE notification whereas the _end() call sends the POSTCHANGE notification. Also, all the necessary synchronization is handled within these calls. In particular, even drivers which set the ASYNC_NOTIFICATION flag can also use these APIs for performing frequency transitions (ie., you can call _begin() from one task, and call the corresponding _end() from a different task). The actual synchronization underneath is not that complicated: The key challenge is to allow drivers to begin the transition from one thread and end it in a completely different thread (this is to enable drivers that do asynchronous POSTCHANGE notification from bottom-halves, to also use the same interface). To achieve this, a 'transition_ongoing' flag, a 'transition_lock' spinlock and a wait-queue are added per-policy. The flag and the wait-queue are used in conjunction to create an "uninterrupted flow" from _begin() to _end(). The spinlock is used to ensure that only one such "flow" is in flight at any given time. Put together, this provides us all the necessary synchronization. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-03-24 12:05:44 +04:00
wake_up(&policy->transition_wait);
}
EXPORT_SYMBOL_GPL(cpufreq_freq_transition_end);
/*
* Fast frequency switching status count. Positive means "enabled", negative
* means "disabled" and 0 means "not decided yet".
*/
static int cpufreq_fast_switch_count;
static DEFINE_MUTEX(cpufreq_fast_switch_lock);
static void cpufreq_list_transition_notifiers(void)
{
struct notifier_block *nb;
pr_info("Registered transition notifiers:\n");
mutex_lock(&cpufreq_transition_notifier_list.mutex);
for (nb = cpufreq_transition_notifier_list.head; nb; nb = nb->next)
pr_info("%pF\n", nb->notifier_call);
mutex_unlock(&cpufreq_transition_notifier_list.mutex);
}
/**
* cpufreq_enable_fast_switch - Enable fast frequency switching for policy.
* @policy: cpufreq policy to enable fast frequency switching for.
*
* Try to enable fast frequency switching for @policy.
*
* The attempt will fail if there is at least one transition notifier registered
* at this point, as fast frequency switching is quite fundamentally at odds
* with transition notifiers. Thus if successful, it will make registration of
* transition notifiers fail going forward.
*/
void cpufreq_enable_fast_switch(struct cpufreq_policy *policy)
{
lockdep_assert_held(&policy->rwsem);
if (!policy->fast_switch_possible)
return;
mutex_lock(&cpufreq_fast_switch_lock);
if (cpufreq_fast_switch_count >= 0) {
cpufreq_fast_switch_count++;
policy->fast_switch_enabled = true;
} else {
pr_warn("CPU%u: Fast frequency switching not enabled\n",
policy->cpu);
cpufreq_list_transition_notifiers();
}
mutex_unlock(&cpufreq_fast_switch_lock);
}
EXPORT_SYMBOL_GPL(cpufreq_enable_fast_switch);
/**
* cpufreq_disable_fast_switch - Disable fast frequency switching for policy.
* @policy: cpufreq policy to disable fast frequency switching for.
*/
void cpufreq_disable_fast_switch(struct cpufreq_policy *policy)
{
mutex_lock(&cpufreq_fast_switch_lock);
if (policy->fast_switch_enabled) {
policy->fast_switch_enabled = false;
if (!WARN_ON(cpufreq_fast_switch_count <= 0))
cpufreq_fast_switch_count--;
}
mutex_unlock(&cpufreq_fast_switch_lock);
}
EXPORT_SYMBOL_GPL(cpufreq_disable_fast_switch);
/**
* cpufreq_driver_resolve_freq - Map a target frequency to a driver-supported
* one.
* @target_freq: target frequency to resolve.
*
* The target to driver frequency mapping is cached in the policy.
*
* Return: Lowest driver-supported frequency greater than or equal to the
* given target_freq, subject to policy (min/max) and driver limitations.
*/
unsigned int cpufreq_driver_resolve_freq(struct cpufreq_policy *policy,
unsigned int target_freq)
{
target_freq = clamp_val(target_freq, policy->min, policy->max);
policy->cached_target_freq = target_freq;
if (cpufreq_driver->target_index) {
int idx;
idx = cpufreq_frequency_table_target(policy, target_freq,
CPUFREQ_RELATION_L);
policy->cached_resolved_idx = idx;
return policy->freq_table[idx].frequency;
}
if (cpufreq_driver->resolve_freq)
return cpufreq_driver->resolve_freq(policy, target_freq);
return target_freq;
}
EXPORT_SYMBOL_GPL(cpufreq_driver_resolve_freq);
unsigned int cpufreq_policy_transition_delay_us(struct cpufreq_policy *policy)
{
unsigned int latency;
if (policy->transition_delay_us)
return policy->transition_delay_us;
latency = policy->cpuinfo.transition_latency / NSEC_PER_USEC;
if (latency) {
/*
* For platforms that can change the frequency very fast (< 10
* us), the above formula gives a decent transition delay. But
* for platforms where transition_latency is in milliseconds, it
* ends up giving unrealistic values.
*
* Cap the default transition delay to 10 ms, which seems to be
* a reasonable amount of time after which we should reevaluate
* the frequency.
*/
return min(latency * LATENCY_MULTIPLIER, (unsigned int)10000);
}
return LATENCY_MULTIPLIER;
}
EXPORT_SYMBOL_GPL(cpufreq_policy_transition_delay_us);
/*********************************************************************
* SYSFS INTERFACE *
*********************************************************************/
static ssize_t show_boost(struct kobject *kobj,
struct kobj_attribute *attr, char *buf)
{
return sprintf(buf, "%d\n", cpufreq_driver->boost_enabled);
}
static ssize_t store_boost(struct kobject *kobj, struct kobj_attribute *attr,
const char *buf, size_t count)
{
int ret, enable;
ret = sscanf(buf, "%d", &enable);
if (ret != 1 || enable < 0 || enable > 1)
return -EINVAL;
if (cpufreq_boost_trigger_state(enable)) {
pr_err("%s: Cannot %s BOOST!\n",
__func__, enable ? "enable" : "disable");
return -EINVAL;
}
pr_debug("%s: cpufreq BOOST %s\n",
__func__, enable ? "enabled" : "disabled");
return count;
}
define_one_global_rw(boost);
static struct cpufreq_governor *find_governor(const char *str_governor)
{
struct cpufreq_governor *t;
for_each_governor(t)
if (!strncasecmp(str_governor, t->name, CPUFREQ_NAME_LEN))
return t;
return NULL;
}
/**
* cpufreq_parse_governor - parse a governor string
*/
static int cpufreq_parse_governor(char *str_governor,
struct cpufreq_policy *policy)
{
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
if (cpufreq_driver->setpolicy) {
if (!strncasecmp(str_governor, "performance", CPUFREQ_NAME_LEN)) {
policy->policy = CPUFREQ_POLICY_PERFORMANCE;
return 0;
}
if (!strncasecmp(str_governor, "powersave", CPUFREQ_NAME_LEN)) {
policy->policy = CPUFREQ_POLICY_POWERSAVE;
return 0;
}
} else {
struct cpufreq_governor *t;
mutex_lock(&cpufreq_governor_mutex);
t = find_governor(str_governor);
if (!t) {
int ret;
mutex_unlock(&cpufreq_governor_mutex);
ret = request_module("cpufreq_%s", str_governor);
if (ret)
return -EINVAL;
mutex_lock(&cpufreq_governor_mutex);
t = find_governor(str_governor);
}
if (t && !try_module_get(t->owner))
t = NULL;
mutex_unlock(&cpufreq_governor_mutex);
if (t) {
policy->governor = t;
return 0;
}
}
return -EINVAL;
}
/**
* cpufreq_per_cpu_attr_read() / show_##file_name() -
* print out cpufreq information
*
* Write out information from cpufreq_driver->policy[cpu]; object must be
* "unsigned int".
*/
#define show_one(file_name, object) \
static ssize_t show_##file_name \
(struct cpufreq_policy *policy, char *buf) \
{ \
return sprintf(buf, "%u\n", policy->object); \
}
show_one(cpuinfo_min_freq, cpuinfo.min_freq);
show_one(cpuinfo_max_freq, cpuinfo.max_freq);
show_one(cpuinfo_transition_latency, cpuinfo.transition_latency);
show_one(scaling_min_freq, min);
show_one(scaling_max_freq, max);
x86: use common aperfmperf_khz_on_cpu() to calculate KHz using APERF/MPERF The goal of this change is to give users a uniform and meaningful result when they read /sys/...cpufreq/scaling_cur_freq on modern x86 hardware, as compared to what they get today. Modern x86 processors include the hardware needed to accurately calculate frequency over an interval -- APERF, MPERF, and the TSC. Here we provide an x86 routine to make this calculation on supported hardware, and use it in preference to any driver driver-specific cpufreq_driver.get() routine. MHz is computed like so: MHz = base_MHz * delta_APERF / delta_MPERF MHz is the average frequency of the busy processor over a measurement interval. The interval is defined to be the time between successive invocations of aperfmperf_khz_on_cpu(), which are expected to to happen on-demand when users read sysfs attribute cpufreq/scaling_cur_freq. As with previous methods of calculating MHz, idle time is excluded. base_MHz above is from TSC calibration global "cpu_khz". This x86 native method to calculate MHz returns a meaningful result no matter if P-states are controlled by hardware or firmware and/or if the Linux cpufreq sub-system is or is-not installed. When this routine is invoked more frequently, the measurement interval becomes shorter. However, the code limits re-computation to 10ms intervals so that average frequency remains meaningful. Discerning users are encouraged to take advantage of the turbostat(8) utility, which can gracefully handle concurrent measurement intervals of arbitrary length. Signed-off-by: Len Brown <len.brown@intel.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2017-06-24 08:11:52 +03:00
__weak unsigned int arch_freq_get_on_cpu(int cpu)
{
return 0;
}
static ssize_t show_scaling_cur_freq(struct cpufreq_policy *policy, char *buf)
{
ssize_t ret;
x86: use common aperfmperf_khz_on_cpu() to calculate KHz using APERF/MPERF The goal of this change is to give users a uniform and meaningful result when they read /sys/...cpufreq/scaling_cur_freq on modern x86 hardware, as compared to what they get today. Modern x86 processors include the hardware needed to accurately calculate frequency over an interval -- APERF, MPERF, and the TSC. Here we provide an x86 routine to make this calculation on supported hardware, and use it in preference to any driver driver-specific cpufreq_driver.get() routine. MHz is computed like so: MHz = base_MHz * delta_APERF / delta_MPERF MHz is the average frequency of the busy processor over a measurement interval. The interval is defined to be the time between successive invocations of aperfmperf_khz_on_cpu(), which are expected to to happen on-demand when users read sysfs attribute cpufreq/scaling_cur_freq. As with previous methods of calculating MHz, idle time is excluded. base_MHz above is from TSC calibration global "cpu_khz". This x86 native method to calculate MHz returns a meaningful result no matter if P-states are controlled by hardware or firmware and/or if the Linux cpufreq sub-system is or is-not installed. When this routine is invoked more frequently, the measurement interval becomes shorter. However, the code limits re-computation to 10ms intervals so that average frequency remains meaningful. Discerning users are encouraged to take advantage of the turbostat(8) utility, which can gracefully handle concurrent measurement intervals of arbitrary length. Signed-off-by: Len Brown <len.brown@intel.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2017-06-24 08:11:52 +03:00
unsigned int freq;
x86: use common aperfmperf_khz_on_cpu() to calculate KHz using APERF/MPERF The goal of this change is to give users a uniform and meaningful result when they read /sys/...cpufreq/scaling_cur_freq on modern x86 hardware, as compared to what they get today. Modern x86 processors include the hardware needed to accurately calculate frequency over an interval -- APERF, MPERF, and the TSC. Here we provide an x86 routine to make this calculation on supported hardware, and use it in preference to any driver driver-specific cpufreq_driver.get() routine. MHz is computed like so: MHz = base_MHz * delta_APERF / delta_MPERF MHz is the average frequency of the busy processor over a measurement interval. The interval is defined to be the time between successive invocations of aperfmperf_khz_on_cpu(), which are expected to to happen on-demand when users read sysfs attribute cpufreq/scaling_cur_freq. As with previous methods of calculating MHz, idle time is excluded. base_MHz above is from TSC calibration global "cpu_khz". This x86 native method to calculate MHz returns a meaningful result no matter if P-states are controlled by hardware or firmware and/or if the Linux cpufreq sub-system is or is-not installed. When this routine is invoked more frequently, the measurement interval becomes shorter. However, the code limits re-computation to 10ms intervals so that average frequency remains meaningful. Discerning users are encouraged to take advantage of the turbostat(8) utility, which can gracefully handle concurrent measurement intervals of arbitrary length. Signed-off-by: Len Brown <len.brown@intel.com> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2017-06-24 08:11:52 +03:00
freq = arch_freq_get_on_cpu(policy->cpu);
if (freq)
ret = sprintf(buf, "%u\n", freq);
else if (cpufreq_driver && cpufreq_driver->setpolicy &&
cpufreq_driver->get)
ret = sprintf(buf, "%u\n", cpufreq_driver->get(policy->cpu));
else
ret = sprintf(buf, "%u\n", policy->cur);
return ret;
}
/**
* cpufreq_per_cpu_attr_write() / store_##file_name() - sysfs write access
*/
#define store_one(file_name, object) \
static ssize_t store_##file_name \
(struct cpufreq_policy *policy, const char *buf, size_t count) \
{ \
int ret, temp; \
struct cpufreq_policy new_policy; \
\
memcpy(&new_policy, policy, sizeof(*policy)); \
new_policy.min = policy->user_policy.min; \
new_policy.max = policy->user_policy.max; \
\
ret = sscanf(buf, "%u", &new_policy.object); \
if (ret != 1) \
return -EINVAL; \
\
temp = new_policy.object; \
ret = cpufreq_set_policy(policy, &new_policy); \
if (!ret) \
policy->user_policy.object = temp; \
\
return ret ? ret : count; \
}
store_one(scaling_min_freq, min);
store_one(scaling_max_freq, max);
/**
* show_cpuinfo_cur_freq - current CPU frequency as detected by hardware
*/
static ssize_t show_cpuinfo_cur_freq(struct cpufreq_policy *policy,
char *buf)
{
unsigned int cur_freq = __cpufreq_get(policy);
if (cur_freq)
return sprintf(buf, "%u\n", cur_freq);
return sprintf(buf, "<unknown>\n");
}
/**
* show_scaling_governor - show the current policy for the specified CPU
*/
static ssize_t show_scaling_governor(struct cpufreq_policy *policy, char *buf)
{
if (policy->policy == CPUFREQ_POLICY_POWERSAVE)
return sprintf(buf, "powersave\n");
else if (policy->policy == CPUFREQ_POLICY_PERFORMANCE)
return sprintf(buf, "performance\n");
else if (policy->governor)
return scnprintf(buf, CPUFREQ_NAME_PLEN, "%s\n",
policy->governor->name);
return -EINVAL;
}
/**
* store_scaling_governor - store policy for the specified CPU
*/
static ssize_t store_scaling_governor(struct cpufreq_policy *policy,
const char *buf, size_t count)
{
int ret;
char str_governor[16];
struct cpufreq_policy new_policy;
memcpy(&new_policy, policy, sizeof(*policy));
ret = sscanf(buf, "%15s", str_governor);
if (ret != 1)
return -EINVAL;
if (cpufreq_parse_governor(str_governor, &new_policy))
return -EINVAL;
ret = cpufreq_set_policy(policy, &new_policy);
if (new_policy.governor)
module_put(new_policy.governor->owner);
return ret ? ret : count;
}
/**
* show_scaling_driver - show the cpufreq driver currently loaded
*/
static ssize_t show_scaling_driver(struct cpufreq_policy *policy, char *buf)
{
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
return scnprintf(buf, CPUFREQ_NAME_PLEN, "%s\n", cpufreq_driver->name);
}
/**
* show_scaling_available_governors - show the available CPUfreq governors
*/
static ssize_t show_scaling_available_governors(struct cpufreq_policy *policy,
char *buf)
{
ssize_t i = 0;
struct cpufreq_governor *t;
cpufreq: Implement light weight ->target_index() routine Currently, the prototype of cpufreq_drivers target routines is: int target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation); And most of the drivers call cpufreq_frequency_table_target() to get a valid index of their frequency table which is closest to the target_freq. And they don't use target_freq and relation after that. So, it makes sense to just do this work in cpufreq core before calling cpufreq_frequency_table_target() and simply pass index instead. But this can be done only with drivers which expose their frequency table with cpufreq core. For others we need to stick with the old prototype of target() until those drivers are converted to expose frequency tables. This patch implements the new light weight prototype for target_index() routine. It looks like this: int target_index(struct cpufreq_policy *policy, unsigned int index); CPUFreq core will call cpufreq_frequency_table_target() before calling this routine and pass index to it. Because CPUFreq core now requires to call routines present in freq_table.c CONFIG_CPU_FREQ_TABLE must be enabled all the time. This also marks target() interface as deprecated. So, that new drivers avoid using it. And Documentation is updated accordingly. It also converts existing .target() to newly defined light weight .target_index() routine for many driver. Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no> Acked-by: Jesper Nilsson <jesper.nilsson@axis.com> Acked-by: Linus Walleij <linus.walleij@linaro.org> Acked-by: Russell King <linux@arm.linux.org.uk> Acked-by: David S. Miller <davem@davemloft.net> Tested-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rjw@rjwysocki.net>
2013-10-25 18:15:48 +04:00
if (!has_target()) {
i += sprintf(buf, "performance powersave");
goto out;
}
for_each_governor(t) {
if (i >= (ssize_t) ((PAGE_SIZE / sizeof(char))
- (CPUFREQ_NAME_LEN + 2)))
goto out;
i += scnprintf(&buf[i], CPUFREQ_NAME_PLEN, "%s ", t->name);
}
out:
i += sprintf(&buf[i], "\n");
return i;
}
ssize_t cpufreq_show_cpus(const struct cpumask *mask, char *buf)
{
ssize_t i = 0;
unsigned int cpu;
for_each_cpu(cpu, mask) {
if (i)
i += scnprintf(&buf[i], (PAGE_SIZE - i - 2), " ");
i += scnprintf(&buf[i], (PAGE_SIZE - i - 2), "%u", cpu);
if (i >= (PAGE_SIZE - 5))
break;
}
i += sprintf(&buf[i], "\n");
return i;
}
EXPORT_SYMBOL_GPL(cpufreq_show_cpus);
/**
* show_related_cpus - show the CPUs affected by each transition even if
* hw coordination is in use
*/
static ssize_t show_related_cpus(struct cpufreq_policy *policy, char *buf)
{
return cpufreq_show_cpus(policy->related_cpus, buf);
}
/**
* show_affected_cpus - show the CPUs affected by each transition
*/
static ssize_t show_affected_cpus(struct cpufreq_policy *policy, char *buf)
{
return cpufreq_show_cpus(policy->cpus, buf);
}
[CPUFREQ] Eliminate cpufreq_userspace scaling_setspeed deadlock Eliminate cpufreq_userspace scaling_setspeed deadlock. Luming Yu recently uncovered yet another cpufreq related deadlock. One thread that continuously switches the governors and the other thread that repeatedly cats the contents of cpufreq directory causes both these threads to go into a deadlock. Detailed examination of the deadlock showed the exact flow before the deadlock as: Thread 1 Thread 2 ________ ________ cats files under /sys/devices/.../cpufreq/ Set governor to userspace Adds a new sysfs entry for scaling_setspeed cats files under /sys/devices/.../cpufreq/ Set governor to performance Holds cpufreq_rw_sem in write mode Sends a STOP notify to userspace governor cat /sys/devices/.../cpufreq/scaling_setspeed Gets a handle on the above sysfs entry with sysfs_get_active Blocks while trying to get cpufreq_rw_sem in read mode Remove a sysfs entry for scaling_setspeed Blocks on sysfs_deactivate while waiting for earlier get_active (on other thread) to drain At this point both threads go into deadlock and any other thread that tries to do anything with sysfs cpufreq will also block. There seems to be no easy way to avoid this deadlock as long as cpufreq_userspace adds/removes the sysfs entry under same kobject as cpufreq. Below patch moves scaling_setspeed to cpufreq.c, keeping it always and calling back the governor on read/write. This is the cleanest fix I could think of, even though adding two callbacks in governor structure just for this seems unnecessary. Note that the change makes scaling_setspeed under /sys/.../cpufreq permanent and returns <unsupported> when governor is not userspace. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-10-26 21:18:21 +04:00
static ssize_t store_scaling_setspeed(struct cpufreq_policy *policy,
const char *buf, size_t count)
[CPUFREQ] Eliminate cpufreq_userspace scaling_setspeed deadlock Eliminate cpufreq_userspace scaling_setspeed deadlock. Luming Yu recently uncovered yet another cpufreq related deadlock. One thread that continuously switches the governors and the other thread that repeatedly cats the contents of cpufreq directory causes both these threads to go into a deadlock. Detailed examination of the deadlock showed the exact flow before the deadlock as: Thread 1 Thread 2 ________ ________ cats files under /sys/devices/.../cpufreq/ Set governor to userspace Adds a new sysfs entry for scaling_setspeed cats files under /sys/devices/.../cpufreq/ Set governor to performance Holds cpufreq_rw_sem in write mode Sends a STOP notify to userspace governor cat /sys/devices/.../cpufreq/scaling_setspeed Gets a handle on the above sysfs entry with sysfs_get_active Blocks while trying to get cpufreq_rw_sem in read mode Remove a sysfs entry for scaling_setspeed Blocks on sysfs_deactivate while waiting for earlier get_active (on other thread) to drain At this point both threads go into deadlock and any other thread that tries to do anything with sysfs cpufreq will also block. There seems to be no easy way to avoid this deadlock as long as cpufreq_userspace adds/removes the sysfs entry under same kobject as cpufreq. Below patch moves scaling_setspeed to cpufreq.c, keeping it always and calling back the governor on read/write. This is the cleanest fix I could think of, even though adding two callbacks in governor structure just for this seems unnecessary. Note that the change makes scaling_setspeed under /sys/.../cpufreq permanent and returns <unsupported> when governor is not userspace. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-10-26 21:18:21 +04:00
{
unsigned int freq = 0;
unsigned int ret;
if (!policy->governor || !policy->governor->store_setspeed)
[CPUFREQ] Eliminate cpufreq_userspace scaling_setspeed deadlock Eliminate cpufreq_userspace scaling_setspeed deadlock. Luming Yu recently uncovered yet another cpufreq related deadlock. One thread that continuously switches the governors and the other thread that repeatedly cats the contents of cpufreq directory causes both these threads to go into a deadlock. Detailed examination of the deadlock showed the exact flow before the deadlock as: Thread 1 Thread 2 ________ ________ cats files under /sys/devices/.../cpufreq/ Set governor to userspace Adds a new sysfs entry for scaling_setspeed cats files under /sys/devices/.../cpufreq/ Set governor to performance Holds cpufreq_rw_sem in write mode Sends a STOP notify to userspace governor cat /sys/devices/.../cpufreq/scaling_setspeed Gets a handle on the above sysfs entry with sysfs_get_active Blocks while trying to get cpufreq_rw_sem in read mode Remove a sysfs entry for scaling_setspeed Blocks on sysfs_deactivate while waiting for earlier get_active (on other thread) to drain At this point both threads go into deadlock and any other thread that tries to do anything with sysfs cpufreq will also block. There seems to be no easy way to avoid this deadlock as long as cpufreq_userspace adds/removes the sysfs entry under same kobject as cpufreq. Below patch moves scaling_setspeed to cpufreq.c, keeping it always and calling back the governor on read/write. This is the cleanest fix I could think of, even though adding two callbacks in governor structure just for this seems unnecessary. Note that the change makes scaling_setspeed under /sys/.../cpufreq permanent and returns <unsupported> when governor is not userspace. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-10-26 21:18:21 +04:00
return -EINVAL;
ret = sscanf(buf, "%u", &freq);
if (ret != 1)
return -EINVAL;
policy->governor->store_setspeed(policy, freq);
return count;
}
static ssize_t show_scaling_setspeed(struct cpufreq_policy *policy, char *buf)
{
if (!policy->governor || !policy->governor->show_setspeed)
[CPUFREQ] Eliminate cpufreq_userspace scaling_setspeed deadlock Eliminate cpufreq_userspace scaling_setspeed deadlock. Luming Yu recently uncovered yet another cpufreq related deadlock. One thread that continuously switches the governors and the other thread that repeatedly cats the contents of cpufreq directory causes both these threads to go into a deadlock. Detailed examination of the deadlock showed the exact flow before the deadlock as: Thread 1 Thread 2 ________ ________ cats files under /sys/devices/.../cpufreq/ Set governor to userspace Adds a new sysfs entry for scaling_setspeed cats files under /sys/devices/.../cpufreq/ Set governor to performance Holds cpufreq_rw_sem in write mode Sends a STOP notify to userspace governor cat /sys/devices/.../cpufreq/scaling_setspeed Gets a handle on the above sysfs entry with sysfs_get_active Blocks while trying to get cpufreq_rw_sem in read mode Remove a sysfs entry for scaling_setspeed Blocks on sysfs_deactivate while waiting for earlier get_active (on other thread) to drain At this point both threads go into deadlock and any other thread that tries to do anything with sysfs cpufreq will also block. There seems to be no easy way to avoid this deadlock as long as cpufreq_userspace adds/removes the sysfs entry under same kobject as cpufreq. Below patch moves scaling_setspeed to cpufreq.c, keeping it always and calling back the governor on read/write. This is the cleanest fix I could think of, even though adding two callbacks in governor structure just for this seems unnecessary. Note that the change makes scaling_setspeed under /sys/.../cpufreq permanent and returns <unsupported> when governor is not userspace. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-10-26 21:18:21 +04:00
return sprintf(buf, "<unsupported>\n");
return policy->governor->show_setspeed(policy, buf);
}
[ACPI/CPUFREQ] Introduce bios_limit per cpu cpufreq sysfs interface This interface is mainly intended (and implemented) for ACPI _PPC BIOS frequency limitations, but other cpufreq drivers can also use it for similar use-cases. Why is this needed: Currently it's not obvious why cpufreq got limited. People see cpufreq/scaling_max_freq reduced, but this could have happened by: - any userspace prog writing to scaling_max_freq - thermal limitations - hardware (_PPC in ACPI case) limitiations Therefore export bios_limit (in kHz) to: - Point the user that it's the BIOS (broken or intended) which limits frequency - Export it as a sysfs interface for userspace progs. While this was a rarely used feature on laptops, there will appear more and more server implemenations providing "Green IT" features like allowing the service processor to limit the frequency. People want to know about HW/BIOS frequency limitations. All ACPI P-state driven cpufreq drivers are covered with this patch: - powernow-k8 - powernow-k7 - acpi-cpufreq Tested with a patched DSDT which limits the first two cores (_PPC returns 1) via _PPC, exposed by bios_limit: # echo 2200000 >cpu2/cpufreq/scaling_max_freq # cat cpu*/cpufreq/scaling_max_freq 2600000 2600000 2200000 2200000 # #scaling_max_freq shows general user/thermal/BIOS limitations # cat cpu*/cpufreq/bios_limit 2600000 2600000 2800000 2800000 # #bios_limit only shows the HW/BIOS limitation CC: Pallipadi Venkatesh <venkatesh.pallipadi@intel.com> CC: Len Brown <lenb@kernel.org> CC: davej@codemonkey.org.uk CC: linux@dominikbrodowski.net Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-11-19 14:31:01 +03:00
/**
* show_bios_limit - show the current cpufreq HW/BIOS limitation
[ACPI/CPUFREQ] Introduce bios_limit per cpu cpufreq sysfs interface This interface is mainly intended (and implemented) for ACPI _PPC BIOS frequency limitations, but other cpufreq drivers can also use it for similar use-cases. Why is this needed: Currently it's not obvious why cpufreq got limited. People see cpufreq/scaling_max_freq reduced, but this could have happened by: - any userspace prog writing to scaling_max_freq - thermal limitations - hardware (_PPC in ACPI case) limitiations Therefore export bios_limit (in kHz) to: - Point the user that it's the BIOS (broken or intended) which limits frequency - Export it as a sysfs interface for userspace progs. While this was a rarely used feature on laptops, there will appear more and more server implemenations providing "Green IT" features like allowing the service processor to limit the frequency. People want to know about HW/BIOS frequency limitations. All ACPI P-state driven cpufreq drivers are covered with this patch: - powernow-k8 - powernow-k7 - acpi-cpufreq Tested with a patched DSDT which limits the first two cores (_PPC returns 1) via _PPC, exposed by bios_limit: # echo 2200000 >cpu2/cpufreq/scaling_max_freq # cat cpu*/cpufreq/scaling_max_freq 2600000 2600000 2200000 2200000 # #scaling_max_freq shows general user/thermal/BIOS limitations # cat cpu*/cpufreq/bios_limit 2600000 2600000 2800000 2800000 # #bios_limit only shows the HW/BIOS limitation CC: Pallipadi Venkatesh <venkatesh.pallipadi@intel.com> CC: Len Brown <lenb@kernel.org> CC: davej@codemonkey.org.uk CC: linux@dominikbrodowski.net Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-11-19 14:31:01 +03:00
*/
static ssize_t show_bios_limit(struct cpufreq_policy *policy, char *buf)
{
unsigned int limit;
int ret;
ret = cpufreq_driver->bios_limit(policy->cpu, &limit);
if (!ret)
return sprintf(buf, "%u\n", limit);
[ACPI/CPUFREQ] Introduce bios_limit per cpu cpufreq sysfs interface This interface is mainly intended (and implemented) for ACPI _PPC BIOS frequency limitations, but other cpufreq drivers can also use it for similar use-cases. Why is this needed: Currently it's not obvious why cpufreq got limited. People see cpufreq/scaling_max_freq reduced, but this could have happened by: - any userspace prog writing to scaling_max_freq - thermal limitations - hardware (_PPC in ACPI case) limitiations Therefore export bios_limit (in kHz) to: - Point the user that it's the BIOS (broken or intended) which limits frequency - Export it as a sysfs interface for userspace progs. While this was a rarely used feature on laptops, there will appear more and more server implemenations providing "Green IT" features like allowing the service processor to limit the frequency. People want to know about HW/BIOS frequency limitations. All ACPI P-state driven cpufreq drivers are covered with this patch: - powernow-k8 - powernow-k7 - acpi-cpufreq Tested with a patched DSDT which limits the first two cores (_PPC returns 1) via _PPC, exposed by bios_limit: # echo 2200000 >cpu2/cpufreq/scaling_max_freq # cat cpu*/cpufreq/scaling_max_freq 2600000 2600000 2200000 2200000 # #scaling_max_freq shows general user/thermal/BIOS limitations # cat cpu*/cpufreq/bios_limit 2600000 2600000 2800000 2800000 # #bios_limit only shows the HW/BIOS limitation CC: Pallipadi Venkatesh <venkatesh.pallipadi@intel.com> CC: Len Brown <lenb@kernel.org> CC: davej@codemonkey.org.uk CC: linux@dominikbrodowski.net Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-11-19 14:31:01 +03:00
return sprintf(buf, "%u\n", policy->cpuinfo.max_freq);
}
cpufreq_freq_attr_ro_perm(cpuinfo_cur_freq, 0400);
cpufreq_freq_attr_ro(cpuinfo_min_freq);
cpufreq_freq_attr_ro(cpuinfo_max_freq);
cpufreq_freq_attr_ro(cpuinfo_transition_latency);
cpufreq_freq_attr_ro(scaling_available_governors);
cpufreq_freq_attr_ro(scaling_driver);
cpufreq_freq_attr_ro(scaling_cur_freq);
cpufreq_freq_attr_ro(bios_limit);
cpufreq_freq_attr_ro(related_cpus);
cpufreq_freq_attr_ro(affected_cpus);
cpufreq_freq_attr_rw(scaling_min_freq);
cpufreq_freq_attr_rw(scaling_max_freq);
cpufreq_freq_attr_rw(scaling_governor);
cpufreq_freq_attr_rw(scaling_setspeed);
static struct attribute *default_attrs[] = {
&cpuinfo_min_freq.attr,
&cpuinfo_max_freq.attr,
&cpuinfo_transition_latency.attr,
&scaling_min_freq.attr,
&scaling_max_freq.attr,
&affected_cpus.attr,
&related_cpus.attr,
&scaling_governor.attr,
&scaling_driver.attr,
&scaling_available_governors.attr,
[CPUFREQ] Eliminate cpufreq_userspace scaling_setspeed deadlock Eliminate cpufreq_userspace scaling_setspeed deadlock. Luming Yu recently uncovered yet another cpufreq related deadlock. One thread that continuously switches the governors and the other thread that repeatedly cats the contents of cpufreq directory causes both these threads to go into a deadlock. Detailed examination of the deadlock showed the exact flow before the deadlock as: Thread 1 Thread 2 ________ ________ cats files under /sys/devices/.../cpufreq/ Set governor to userspace Adds a new sysfs entry for scaling_setspeed cats files under /sys/devices/.../cpufreq/ Set governor to performance Holds cpufreq_rw_sem in write mode Sends a STOP notify to userspace governor cat /sys/devices/.../cpufreq/scaling_setspeed Gets a handle on the above sysfs entry with sysfs_get_active Blocks while trying to get cpufreq_rw_sem in read mode Remove a sysfs entry for scaling_setspeed Blocks on sysfs_deactivate while waiting for earlier get_active (on other thread) to drain At this point both threads go into deadlock and any other thread that tries to do anything with sysfs cpufreq will also block. There seems to be no easy way to avoid this deadlock as long as cpufreq_userspace adds/removes the sysfs entry under same kobject as cpufreq. Below patch moves scaling_setspeed to cpufreq.c, keeping it always and calling back the governor on read/write. This is the cleanest fix I could think of, even though adding two callbacks in governor structure just for this seems unnecessary. Note that the change makes scaling_setspeed under /sys/.../cpufreq permanent and returns <unsupported> when governor is not userspace. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-10-26 21:18:21 +04:00
&scaling_setspeed.attr,
NULL
};
#define to_policy(k) container_of(k, struct cpufreq_policy, kobj)
#define to_attr(a) container_of(a, struct freq_attr, attr)
static ssize_t show(struct kobject *kobj, struct attribute *attr, char *buf)
{
struct cpufreq_policy *policy = to_policy(kobj);
struct freq_attr *fattr = to_attr(attr);
ssize_t ret;
down_read(&policy->rwsem);
ret = fattr->show(policy, buf);
up_read(&policy->rwsem);
return ret;
}
static ssize_t store(struct kobject *kobj, struct attribute *attr,
const char *buf, size_t count)
{
struct cpufreq_policy *policy = to_policy(kobj);
struct freq_attr *fattr = to_attr(attr);
ssize_t ret = -EINVAL;
cpufreq: Fix a circular lock dependency problem With lockdep turned on, the following circular lock dependency problem was reported: [ 57.470040] ====================================================== [ 57.502900] WARNING: possible circular locking dependency detected [ 57.535208] 4.18.0-0.rc3.1.el8+7.x86_64+debug #1 Tainted: G [ 57.577761] ------------------------------------------------------ [ 57.609714] tuned/1505 is trying to acquire lock: [ 57.633808] 00000000559deec5 (cpu_hotplug_lock.rw_sem){++++}, at: store+0x27/0x120 [ 57.672880] [ 57.672880] but task is already holding lock: [ 57.702184] 000000002136ca64 (kn->count#118){++++}, at: kernfs_fop_write+0x1d0/0x410 [ 57.742176] [ 57.742176] which lock already depends on the new lock. [ 57.742176] [ 57.785220] [ 57.785220] the existing dependency chain (in reverse order) is: : [ 58.932512] other info that might help us debug this: [ 58.932512] [ 58.973344] Chain exists of: [ 58.973344] cpu_hotplug_lock.rw_sem --> subsys mutex#5 --> kn->count#118 [ 58.973344] [ 59.030795] Possible unsafe locking scenario: [ 59.030795] [ 59.061248] CPU0 CPU1 [ 59.085377] ---- ---- [ 59.108160] lock(kn->count#118); [ 59.124935] lock(subsys mutex#5); [ 59.156330] lock(kn->count#118); [ 59.186088] lock(cpu_hotplug_lock.rw_sem); [ 59.208541] [ 59.208541] *** DEADLOCK *** In the cpufreq_register_driver() function, the lock sequence is: cpus_read_lock --> kn->count For the cpufreq sysfs store method, the lock sequence is: kn->count --> cpus_read_lock These sequences are actually safe as they are taking a share lock on cpu_hotplug_lock. However, the current lockdep code doesn't check for share locking when detecting circular lock dependency. Fixing that could be a substantial effort. Instead, we can work around this problem by using cpus_read_trylock() in the store method which is much simpler. The chance of not getting the read lock is very small. If that happens, the userspace application that writes the sysfs file will get an error. Signed-off-by: Waiman Long <longman@redhat.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2018-07-24 21:26:05 +03:00
/*
* cpus_read_trylock() is used here to work around a circular lock
* dependency problem with respect to the cpufreq_register_driver().
*/
if (!cpus_read_trylock())
return -EBUSY;
cpufreq: Synchronize the cpufreq store_*() routines with CPU hotplug The functions that are used to write to cpufreq sysfs files (such as store_scaling_max_freq()) are not hotplug safe. They can race with CPU hotplug tasks and lead to problems such as trying to acquire an already destroyed timer-mutex etc. Eg: __cpufreq_remove_dev() __cpufreq_governor(policy, CPUFREQ_GOV_STOP); policy->governor->governor(policy, CPUFREQ_GOV_STOP); cpufreq_governor_dbs() case CPUFREQ_GOV_STOP: mutex_destroy(&cpu_cdbs->timer_mutex) cpu_cdbs->cur_policy = NULL; <PREEMPT> store() __cpufreq_set_policy() __cpufreq_governor(policy, CPUFREQ_GOV_LIMITS); policy->governor->governor(policy, CPUFREQ_GOV_LIMITS); case CPUFREQ_GOV_LIMITS: mutex_lock(&cpu_cdbs->timer_mutex); <-- Warning (destroyed mutex) if (policy->max < cpu_cdbs->cur_policy->cur) <- cur_policy == NULL So use get_online_cpus()/put_online_cpus() in the store_*() functions, to synchronize with CPU hotplug. However, there is an additional point to note here: some parts of the CPU teardown in the cpufreq subsystem are done in the CPU_POST_DEAD stage, with cpu_hotplug.lock *released*. So, using the get/put_online_cpus() functions alone is insufficient; we should also ensure that we don't race with those latter steps in the hotplug sequence. We can easily achieve this by checking if the CPU is online before proceeding with the store, since the CPU would have been marked offline by the time the CPU_POST_DEAD notifiers are executed. Reported-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-09-06 23:53:43 +04:00
if (cpu_online(policy->cpu)) {
down_write(&policy->rwsem);
ret = fattr->store(policy, buf, count);
up_write(&policy->rwsem);
}
cpus_read_unlock();
cpufreq: Synchronize the cpufreq store_*() routines with CPU hotplug The functions that are used to write to cpufreq sysfs files (such as store_scaling_max_freq()) are not hotplug safe. They can race with CPU hotplug tasks and lead to problems such as trying to acquire an already destroyed timer-mutex etc. Eg: __cpufreq_remove_dev() __cpufreq_governor(policy, CPUFREQ_GOV_STOP); policy->governor->governor(policy, CPUFREQ_GOV_STOP); cpufreq_governor_dbs() case CPUFREQ_GOV_STOP: mutex_destroy(&cpu_cdbs->timer_mutex) cpu_cdbs->cur_policy = NULL; <PREEMPT> store() __cpufreq_set_policy() __cpufreq_governor(policy, CPUFREQ_GOV_LIMITS); policy->governor->governor(policy, CPUFREQ_GOV_LIMITS); case CPUFREQ_GOV_LIMITS: mutex_lock(&cpu_cdbs->timer_mutex); <-- Warning (destroyed mutex) if (policy->max < cpu_cdbs->cur_policy->cur) <- cur_policy == NULL So use get_online_cpus()/put_online_cpus() in the store_*() functions, to synchronize with CPU hotplug. However, there is an additional point to note here: some parts of the CPU teardown in the cpufreq subsystem are done in the CPU_POST_DEAD stage, with cpu_hotplug.lock *released*. So, using the get/put_online_cpus() functions alone is insufficient; we should also ensure that we don't race with those latter steps in the hotplug sequence. We can easily achieve this by checking if the CPU is online before proceeding with the store, since the CPU would have been marked offline by the time the CPU_POST_DEAD notifiers are executed. Reported-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-09-06 23:53:43 +04:00
return ret;
}
static void cpufreq_sysfs_release(struct kobject *kobj)
{
struct cpufreq_policy *policy = to_policy(kobj);
pr_debug("last reference is dropped\n");
complete(&policy->kobj_unregister);
}
static const struct sysfs_ops sysfs_ops = {
.show = show,
.store = store,
};
static struct kobj_type ktype_cpufreq = {
.sysfs_ops = &sysfs_ops,
.default_attrs = default_attrs,
.release = cpufreq_sysfs_release,
};
static void add_cpu_dev_symlink(struct cpufreq_policy *policy, unsigned int cpu)
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
{
struct device *dev = get_cpu_device(cpu);
if (!dev)
return;
if (cpumask_test_and_set_cpu(cpu, policy->real_cpus))
return;
cpufreq: create link to policy only for registered CPUs If a cpufreq driver is registered very early in the boot stage (e.g. registered from postcore_initcall()), then cpufreq core may generate kernel warnings for it. In this case, the CPUs are brought online, then the cpufreq driver is registered, and then the CPU topology devices are registered. However, by the time cpufreq_add_dev() gets called, the cpu device isn't stored in the per-cpu variable (cpu_sys_devices,) which is read by get_cpu_device(). So the cpufreq core fails to get device for the CPU, for which cpufreq_add_dev() was called in the first place and we will hit a WARN_ON(!cpu_dev). Even if we reuse the 'dev' parameter passed to cpufreq_add_dev() to avoid that warning, there might be other CPUs online that share the policy with the cpu for which cpufreq_add_dev() is called. Eventually get_cpu_device() will return NULL for them as well, and we will hit the same WARN_ON() again. In order to fix these issues, change cpufreq core to create links to the policy for a cpu only when cpufreq_add_dev() is called for that CPU. Reuse the 'real_cpus' mask to track that as well. Note that cpufreq_remove_dev() already handles removal of the links for individual CPUs and cpufreq_add_dev() has aligned with that now. Reported-by: Russell King <rmk+kernel@arm.linux.org.uk> Tested-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-09-12 09:37:05 +03:00
dev_dbg(dev, "%s: Adding symlink\n", __func__);
if (sysfs_create_link(&dev->kobj, &policy->kobj, "cpufreq"))
dev_err(dev, "cpufreq symlink creation failed\n");
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
}
cpufreq: create link to policy only for registered CPUs If a cpufreq driver is registered very early in the boot stage (e.g. registered from postcore_initcall()), then cpufreq core may generate kernel warnings for it. In this case, the CPUs are brought online, then the cpufreq driver is registered, and then the CPU topology devices are registered. However, by the time cpufreq_add_dev() gets called, the cpu device isn't stored in the per-cpu variable (cpu_sys_devices,) which is read by get_cpu_device(). So the cpufreq core fails to get device for the CPU, for which cpufreq_add_dev() was called in the first place and we will hit a WARN_ON(!cpu_dev). Even if we reuse the 'dev' parameter passed to cpufreq_add_dev() to avoid that warning, there might be other CPUs online that share the policy with the cpu for which cpufreq_add_dev() is called. Eventually get_cpu_device() will return NULL for them as well, and we will hit the same WARN_ON() again. In order to fix these issues, change cpufreq core to create links to the policy for a cpu only when cpufreq_add_dev() is called for that CPU. Reuse the 'real_cpus' mask to track that as well. Note that cpufreq_remove_dev() already handles removal of the links for individual CPUs and cpufreq_add_dev() has aligned with that now. Reported-by: Russell King <rmk+kernel@arm.linux.org.uk> Tested-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-09-12 09:37:05 +03:00
static void remove_cpu_dev_symlink(struct cpufreq_policy *policy,
struct device *dev)
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
{
cpufreq: create link to policy only for registered CPUs If a cpufreq driver is registered very early in the boot stage (e.g. registered from postcore_initcall()), then cpufreq core may generate kernel warnings for it. In this case, the CPUs are brought online, then the cpufreq driver is registered, and then the CPU topology devices are registered. However, by the time cpufreq_add_dev() gets called, the cpu device isn't stored in the per-cpu variable (cpu_sys_devices,) which is read by get_cpu_device(). So the cpufreq core fails to get device for the CPU, for which cpufreq_add_dev() was called in the first place and we will hit a WARN_ON(!cpu_dev). Even if we reuse the 'dev' parameter passed to cpufreq_add_dev() to avoid that warning, there might be other CPUs online that share the policy with the cpu for which cpufreq_add_dev() is called. Eventually get_cpu_device() will return NULL for them as well, and we will hit the same WARN_ON() again. In order to fix these issues, change cpufreq core to create links to the policy for a cpu only when cpufreq_add_dev() is called for that CPU. Reuse the 'real_cpus' mask to track that as well. Note that cpufreq_remove_dev() already handles removal of the links for individual CPUs and cpufreq_add_dev() has aligned with that now. Reported-by: Russell King <rmk+kernel@arm.linux.org.uk> Tested-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-09-12 09:37:05 +03:00
dev_dbg(dev, "%s: Removing symlink\n", __func__);
sysfs_remove_link(&dev->kobj, "cpufreq");
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
}
static int cpufreq_add_dev_interface(struct cpufreq_policy *policy)
{
struct freq_attr **drv_attr;
int ret = 0;
/* set up files for this cpu device */
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
drv_attr = cpufreq_driver->attr;
while (drv_attr && *drv_attr) {
ret = sysfs_create_file(&policy->kobj, &((*drv_attr)->attr));
if (ret)
cpufreq: Ref the policy object sooner Do it before it's assigned to cpufreq_cpu_data, otherwise when a driver tries to get the cpu frequency during initialization the policy kobj is referenced and we get this warning: ------------[ cut here ]------------ WARNING: CPU: 1 PID: 64 at include/linux/kref.h:47 kobject_get+0x64/0x70() Modules linked in: CPU: 1 PID: 64 Comm: irq/77-tegra-ac Not tainted 3.18.0-rc4-next-20141114ccu-00050-g3eff942 #326 [<c0016fac>] (unwind_backtrace) from [<c001272c>] (show_stack+0x10/0x14) [<c001272c>] (show_stack) from [<c06085d8>] (dump_stack+0x98/0xd8) [<c06085d8>] (dump_stack) from [<c002892c>] (warn_slowpath_common+0x84/0xb4) [<c002892c>] (warn_slowpath_common) from [<c00289f8>] (warn_slowpath_null+0x1c/0x24) [<c00289f8>] (warn_slowpath_null) from [<c0220290>] (kobject_get+0x64/0x70) [<c0220290>] (kobject_get) from [<c03e944c>] (cpufreq_cpu_get+0x88/0xc8) [<c03e944c>] (cpufreq_cpu_get) from [<c03e9500>] (cpufreq_get+0xc/0x64) [<c03e9500>] (cpufreq_get) from [<c0285288>] (actmon_thread_isr+0x134/0x198) [<c0285288>] (actmon_thread_isr) from [<c0069008>] (irq_thread_fn+0x1c/0x40) [<c0069008>] (irq_thread_fn) from [<c0069324>] (irq_thread+0x134/0x174) [<c0069324>] (irq_thread) from [<c0040290>] (kthread+0xdc/0xf4) [<c0040290>] (kthread) from [<c000f4b8>] (ret_from_fork+0x14/0x3c) ---[ end trace b7bd64a81b340c59 ]--- Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-11-24 12:08:03 +03:00
return ret;
drv_attr++;
}
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
if (cpufreq_driver->get) {
ret = sysfs_create_file(&policy->kobj, &cpuinfo_cur_freq.attr);
if (ret)
cpufreq: Ref the policy object sooner Do it before it's assigned to cpufreq_cpu_data, otherwise when a driver tries to get the cpu frequency during initialization the policy kobj is referenced and we get this warning: ------------[ cut here ]------------ WARNING: CPU: 1 PID: 64 at include/linux/kref.h:47 kobject_get+0x64/0x70() Modules linked in: CPU: 1 PID: 64 Comm: irq/77-tegra-ac Not tainted 3.18.0-rc4-next-20141114ccu-00050-g3eff942 #326 [<c0016fac>] (unwind_backtrace) from [<c001272c>] (show_stack+0x10/0x14) [<c001272c>] (show_stack) from [<c06085d8>] (dump_stack+0x98/0xd8) [<c06085d8>] (dump_stack) from [<c002892c>] (warn_slowpath_common+0x84/0xb4) [<c002892c>] (warn_slowpath_common) from [<c00289f8>] (warn_slowpath_null+0x1c/0x24) [<c00289f8>] (warn_slowpath_null) from [<c0220290>] (kobject_get+0x64/0x70) [<c0220290>] (kobject_get) from [<c03e944c>] (cpufreq_cpu_get+0x88/0xc8) [<c03e944c>] (cpufreq_cpu_get) from [<c03e9500>] (cpufreq_get+0xc/0x64) [<c03e9500>] (cpufreq_get) from [<c0285288>] (actmon_thread_isr+0x134/0x198) [<c0285288>] (actmon_thread_isr) from [<c0069008>] (irq_thread_fn+0x1c/0x40) [<c0069008>] (irq_thread_fn) from [<c0069324>] (irq_thread+0x134/0x174) [<c0069324>] (irq_thread) from [<c0040290>] (kthread+0xdc/0xf4) [<c0040290>] (kthread) from [<c000f4b8>] (ret_from_fork+0x14/0x3c) ---[ end trace b7bd64a81b340c59 ]--- Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-11-24 12:08:03 +03:00
return ret;
}
ret = sysfs_create_file(&policy->kobj, &scaling_cur_freq.attr);
if (ret)
cpufreq: Ref the policy object sooner Do it before it's assigned to cpufreq_cpu_data, otherwise when a driver tries to get the cpu frequency during initialization the policy kobj is referenced and we get this warning: ------------[ cut here ]------------ WARNING: CPU: 1 PID: 64 at include/linux/kref.h:47 kobject_get+0x64/0x70() Modules linked in: CPU: 1 PID: 64 Comm: irq/77-tegra-ac Not tainted 3.18.0-rc4-next-20141114ccu-00050-g3eff942 #326 [<c0016fac>] (unwind_backtrace) from [<c001272c>] (show_stack+0x10/0x14) [<c001272c>] (show_stack) from [<c06085d8>] (dump_stack+0x98/0xd8) [<c06085d8>] (dump_stack) from [<c002892c>] (warn_slowpath_common+0x84/0xb4) [<c002892c>] (warn_slowpath_common) from [<c00289f8>] (warn_slowpath_null+0x1c/0x24) [<c00289f8>] (warn_slowpath_null) from [<c0220290>] (kobject_get+0x64/0x70) [<c0220290>] (kobject_get) from [<c03e944c>] (cpufreq_cpu_get+0x88/0xc8) [<c03e944c>] (cpufreq_cpu_get) from [<c03e9500>] (cpufreq_get+0xc/0x64) [<c03e9500>] (cpufreq_get) from [<c0285288>] (actmon_thread_isr+0x134/0x198) [<c0285288>] (actmon_thread_isr) from [<c0069008>] (irq_thread_fn+0x1c/0x40) [<c0069008>] (irq_thread_fn) from [<c0069324>] (irq_thread+0x134/0x174) [<c0069324>] (irq_thread) from [<c0040290>] (kthread+0xdc/0xf4) [<c0040290>] (kthread) from [<c000f4b8>] (ret_from_fork+0x14/0x3c) ---[ end trace b7bd64a81b340c59 ]--- Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-11-24 12:08:03 +03:00
return ret;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
if (cpufreq_driver->bios_limit) {
[ACPI/CPUFREQ] Introduce bios_limit per cpu cpufreq sysfs interface This interface is mainly intended (and implemented) for ACPI _PPC BIOS frequency limitations, but other cpufreq drivers can also use it for similar use-cases. Why is this needed: Currently it's not obvious why cpufreq got limited. People see cpufreq/scaling_max_freq reduced, but this could have happened by: - any userspace prog writing to scaling_max_freq - thermal limitations - hardware (_PPC in ACPI case) limitiations Therefore export bios_limit (in kHz) to: - Point the user that it's the BIOS (broken or intended) which limits frequency - Export it as a sysfs interface for userspace progs. While this was a rarely used feature on laptops, there will appear more and more server implemenations providing "Green IT" features like allowing the service processor to limit the frequency. People want to know about HW/BIOS frequency limitations. All ACPI P-state driven cpufreq drivers are covered with this patch: - powernow-k8 - powernow-k7 - acpi-cpufreq Tested with a patched DSDT which limits the first two cores (_PPC returns 1) via _PPC, exposed by bios_limit: # echo 2200000 >cpu2/cpufreq/scaling_max_freq # cat cpu*/cpufreq/scaling_max_freq 2600000 2600000 2200000 2200000 # #scaling_max_freq shows general user/thermal/BIOS limitations # cat cpu*/cpufreq/bios_limit 2600000 2600000 2800000 2800000 # #bios_limit only shows the HW/BIOS limitation CC: Pallipadi Venkatesh <venkatesh.pallipadi@intel.com> CC: Len Brown <lenb@kernel.org> CC: davej@codemonkey.org.uk CC: linux@dominikbrodowski.net Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-11-19 14:31:01 +03:00
ret = sysfs_create_file(&policy->kobj, &bios_limit.attr);
if (ret)
cpufreq: Ref the policy object sooner Do it before it's assigned to cpufreq_cpu_data, otherwise when a driver tries to get the cpu frequency during initialization the policy kobj is referenced and we get this warning: ------------[ cut here ]------------ WARNING: CPU: 1 PID: 64 at include/linux/kref.h:47 kobject_get+0x64/0x70() Modules linked in: CPU: 1 PID: 64 Comm: irq/77-tegra-ac Not tainted 3.18.0-rc4-next-20141114ccu-00050-g3eff942 #326 [<c0016fac>] (unwind_backtrace) from [<c001272c>] (show_stack+0x10/0x14) [<c001272c>] (show_stack) from [<c06085d8>] (dump_stack+0x98/0xd8) [<c06085d8>] (dump_stack) from [<c002892c>] (warn_slowpath_common+0x84/0xb4) [<c002892c>] (warn_slowpath_common) from [<c00289f8>] (warn_slowpath_null+0x1c/0x24) [<c00289f8>] (warn_slowpath_null) from [<c0220290>] (kobject_get+0x64/0x70) [<c0220290>] (kobject_get) from [<c03e944c>] (cpufreq_cpu_get+0x88/0xc8) [<c03e944c>] (cpufreq_cpu_get) from [<c03e9500>] (cpufreq_get+0xc/0x64) [<c03e9500>] (cpufreq_get) from [<c0285288>] (actmon_thread_isr+0x134/0x198) [<c0285288>] (actmon_thread_isr) from [<c0069008>] (irq_thread_fn+0x1c/0x40) [<c0069008>] (irq_thread_fn) from [<c0069324>] (irq_thread+0x134/0x174) [<c0069324>] (irq_thread) from [<c0040290>] (kthread+0xdc/0xf4) [<c0040290>] (kthread) from [<c000f4b8>] (ret_from_fork+0x14/0x3c) ---[ end trace b7bd64a81b340c59 ]--- Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-11-24 12:08:03 +03:00
return ret;
[ACPI/CPUFREQ] Introduce bios_limit per cpu cpufreq sysfs interface This interface is mainly intended (and implemented) for ACPI _PPC BIOS frequency limitations, but other cpufreq drivers can also use it for similar use-cases. Why is this needed: Currently it's not obvious why cpufreq got limited. People see cpufreq/scaling_max_freq reduced, but this could have happened by: - any userspace prog writing to scaling_max_freq - thermal limitations - hardware (_PPC in ACPI case) limitiations Therefore export bios_limit (in kHz) to: - Point the user that it's the BIOS (broken or intended) which limits frequency - Export it as a sysfs interface for userspace progs. While this was a rarely used feature on laptops, there will appear more and more server implemenations providing "Green IT" features like allowing the service processor to limit the frequency. People want to know about HW/BIOS frequency limitations. All ACPI P-state driven cpufreq drivers are covered with this patch: - powernow-k8 - powernow-k7 - acpi-cpufreq Tested with a patched DSDT which limits the first two cores (_PPC returns 1) via _PPC, exposed by bios_limit: # echo 2200000 >cpu2/cpufreq/scaling_max_freq # cat cpu*/cpufreq/scaling_max_freq 2600000 2600000 2200000 2200000 # #scaling_max_freq shows general user/thermal/BIOS limitations # cat cpu*/cpufreq/bios_limit 2600000 2600000 2800000 2800000 # #bios_limit only shows the HW/BIOS limitation CC: Pallipadi Venkatesh <venkatesh.pallipadi@intel.com> CC: Len Brown <lenb@kernel.org> CC: davej@codemonkey.org.uk CC: linux@dominikbrodowski.net Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-11-19 14:31:01 +03:00
}
cpufreq: create link to policy only for registered CPUs If a cpufreq driver is registered very early in the boot stage (e.g. registered from postcore_initcall()), then cpufreq core may generate kernel warnings for it. In this case, the CPUs are brought online, then the cpufreq driver is registered, and then the CPU topology devices are registered. However, by the time cpufreq_add_dev() gets called, the cpu device isn't stored in the per-cpu variable (cpu_sys_devices,) which is read by get_cpu_device(). So the cpufreq core fails to get device for the CPU, for which cpufreq_add_dev() was called in the first place and we will hit a WARN_ON(!cpu_dev). Even if we reuse the 'dev' parameter passed to cpufreq_add_dev() to avoid that warning, there might be other CPUs online that share the policy with the cpu for which cpufreq_add_dev() is called. Eventually get_cpu_device() will return NULL for them as well, and we will hit the same WARN_ON() again. In order to fix these issues, change cpufreq core to create links to the policy for a cpu only when cpufreq_add_dev() is called for that CPU. Reuse the 'real_cpus' mask to track that as well. Note that cpufreq_remove_dev() already handles removal of the links for individual CPUs and cpufreq_add_dev() has aligned with that now. Reported-by: Russell King <rmk+kernel@arm.linux.org.uk> Tested-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-09-12 09:37:05 +03:00
return 0;
}
__weak struct cpufreq_governor *cpufreq_default_governor(void)
{
return NULL;
}
static int cpufreq_init_policy(struct cpufreq_policy *policy)
{
struct cpufreq_governor *gov = NULL;
struct cpufreq_policy new_policy;
memcpy(&new_policy, policy, sizeof(*policy));
/* Update governor of new_policy to the governor used before hotplug */
gov = find_governor(policy->last_governor);
if (gov) {
pr_debug("Restoring governor %s for cpu %d\n",
policy->governor->name, policy->cpu);
} else {
gov = cpufreq_default_governor();
if (!gov)
return -ENODATA;
}
new_policy.governor = gov;
/* Use the default policy if there is no last_policy. */
if (cpufreq_driver->setpolicy) {
if (policy->last_policy)
new_policy.policy = policy->last_policy;
else
cpufreq_parse_governor(gov->name, &new_policy);
}
/* set default policy */
return cpufreq_set_policy(policy, &new_policy);
}
static int cpufreq_add_policy_cpu(struct cpufreq_policy *policy, unsigned int cpu)
{
cpufreq: Implement light weight ->target_index() routine Currently, the prototype of cpufreq_drivers target routines is: int target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation); And most of the drivers call cpufreq_frequency_table_target() to get a valid index of their frequency table which is closest to the target_freq. And they don't use target_freq and relation after that. So, it makes sense to just do this work in cpufreq core before calling cpufreq_frequency_table_target() and simply pass index instead. But this can be done only with drivers which expose their frequency table with cpufreq core. For others we need to stick with the old prototype of target() until those drivers are converted to expose frequency tables. This patch implements the new light weight prototype for target_index() routine. It looks like this: int target_index(struct cpufreq_policy *policy, unsigned int index); CPUFreq core will call cpufreq_frequency_table_target() before calling this routine and pass index to it. Because CPUFreq core now requires to call routines present in freq_table.c CONFIG_CPU_FREQ_TABLE must be enabled all the time. This also marks target() interface as deprecated. So, that new drivers avoid using it. And Documentation is updated accordingly. It also converts existing .target() to newly defined light weight .target_index() routine for many driver. Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no> Acked-by: Jesper Nilsson <jesper.nilsson@axis.com> Acked-by: Linus Walleij <linus.walleij@linaro.org> Acked-by: Russell King <linux@arm.linux.org.uk> Acked-by: David S. Miller <davem@davemloft.net> Tested-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rjw@rjwysocki.net>
2013-10-25 18:15:48 +04:00
int ret = 0;
/* Has this CPU been taken care of already? */
if (cpumask_test_cpu(cpu, policy->cpus))
return 0;
down_write(&policy->rwsem);
if (has_target())
cpufreq_stop_governor(policy);
cpumask_set_cpu(cpu, policy->cpus);
cpufreq: Implement light weight ->target_index() routine Currently, the prototype of cpufreq_drivers target routines is: int target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation); And most of the drivers call cpufreq_frequency_table_target() to get a valid index of their frequency table which is closest to the target_freq. And they don't use target_freq and relation after that. So, it makes sense to just do this work in cpufreq core before calling cpufreq_frequency_table_target() and simply pass index instead. But this can be done only with drivers which expose their frequency table with cpufreq core. For others we need to stick with the old prototype of target() until those drivers are converted to expose frequency tables. This patch implements the new light weight prototype for target_index() routine. It looks like this: int target_index(struct cpufreq_policy *policy, unsigned int index); CPUFreq core will call cpufreq_frequency_table_target() before calling this routine and pass index to it. Because CPUFreq core now requires to call routines present in freq_table.c CONFIG_CPU_FREQ_TABLE must be enabled all the time. This also marks target() interface as deprecated. So, that new drivers avoid using it. And Documentation is updated accordingly. It also converts existing .target() to newly defined light weight .target_index() routine for many driver. Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no> Acked-by: Jesper Nilsson <jesper.nilsson@axis.com> Acked-by: Linus Walleij <linus.walleij@linaro.org> Acked-by: Russell King <linux@arm.linux.org.uk> Acked-by: David S. Miller <davem@davemloft.net> Tested-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rjw@rjwysocki.net>
2013-10-25 18:15:48 +04:00
if (has_target()) {
ret = cpufreq_start_governor(policy);
if (ret)
pr_err("%s: Failed to start governor\n", __func__);
}
up_write(&policy->rwsem);
return ret;
}
static void handle_update(struct work_struct *work)
{
struct cpufreq_policy *policy =
container_of(work, struct cpufreq_policy, update);
unsigned int cpu = policy->cpu;
pr_debug("handle_update for cpu %u called\n", cpu);
cpufreq_update_policy(cpu);
}
static struct cpufreq_policy *cpufreq_policy_alloc(unsigned int cpu)
{
struct cpufreq_policy *policy;
int ret;
policy = kzalloc(sizeof(*policy), GFP_KERNEL);
if (!policy)
return NULL;
if (!alloc_cpumask_var(&policy->cpus, GFP_KERNEL))
goto err_free_policy;
if (!zalloc_cpumask_var(&policy->related_cpus, GFP_KERNEL))
goto err_free_cpumask;
cpufreq: Avoid attempts to create duplicate symbolic links After commit 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) there is a problem with CPUs that share cpufreq policy objects with other CPUs and are initially offline. Say CPU1 shares a policy with CPU0 which is online and is registered first. As part of the registration process, cpufreq_add_dev() is called for it. It creates the policy object and a symbolic link to it from the CPU1's sysfs directory. If CPU1 is registered subsequently and it is offline at that time, cpufreq_add_dev() will attempt to create a symbolic link to the policy object for it, but that link is present already, so a warning about that will be triggered. To avoid that warning, make cpufreq use an additional CPU mask containing related CPUs that are actually present for each policy object. That mask is initialized when the policy object is populated after its creation (for the first online CPU using it) and it includes CPUs from the "policy CPUs" mask returned by the cpufreq driver's ->init() callback that are physically present at that time. Symbolic links to the policy are created only for the CPUs in that mask. If cpufreq_add_dev() is invoked for an offline CPU, it checks the new mask and only creates the symlink if the CPU was not in it (the CPU is added to the mask at the same time). In turn, cpufreq_remove_dev() drops the given CPU from the new mask, removes its symlink to the policy object and returns, unless it is the CPU owning the policy object. In that case, the policy object is moved to a new CPU's sysfs directory or deleted if the CPU being removed was the last user of the policy. While at it, notice that cpufreq_remove_dev() can't fail, because its return value is ignored, so make it ignore return values from __cpufreq_remove_dev_prepare() and __cpufreq_remove_dev_finish() and prevent these functions from aborting on errors returned by __cpufreq_governor(). Also drop the now unused sif argument from them. Fixes: 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reported-and-tested-by: Russell King <linux@arm.linux.org.uk> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2015-07-26 03:07:47 +03:00
if (!zalloc_cpumask_var(&policy->real_cpus, GFP_KERNEL))
goto err_free_rcpumask;
ret = kobject_init_and_add(&policy->kobj, &ktype_cpufreq,
cpufreq_global_kobject, "policy%u", cpu);
if (ret) {
pr_err("%s: failed to init policy->kobj: %d\n", __func__, ret);
goto err_free_real_cpus;
}
INIT_LIST_HEAD(&policy->policy_list);
init_rwsem(&policy->rwsem);
cpufreq: Make sure frequency transitions are serialized Whenever we change the frequency of a CPU, we call the PRECHANGE and POSTCHANGE notifiers. They must be serialized, i.e. PRECHANGE and POSTCHANGE notifiers should strictly alternate, thereby preventing two different sets of PRECHANGE or POSTCHANGE notifiers from interleaving arbitrarily. The following examples illustrate why this is important: Scenario 1: ----------- A thread reading the value of cpuinfo_cur_freq, will call __cpufreq_cpu_get()->cpufreq_out_of_sync()->cpufreq_notify_transition() The ondemand governor can decide to change the frequency of the CPU at the same time and hence it can end up sending the notifications via ->target(). If the notifiers are not serialized, the following sequence can occur: - PRECHANGE Notification for freq A (from cpuinfo_cur_freq) - PRECHANGE Notification for freq B (from target()) - Freq changed by target() to B - POSTCHANGE Notification for freq B - POSTCHANGE Notification for freq A We can see from the above that the last POSTCHANGE Notification happens for freq A but the hardware is set to run at freq B. Where would we break then?: adjust_jiffies() in cpufreq.c & cpufreq_callback() in arch/arm/kernel/smp.c (which also adjusts the jiffies). All the loops_per_jiffy calculations will get messed up. Scenario 2: ----------- The governor calls __cpufreq_driver_target() to change the frequency. At the same time, if we change scaling_{min|max}_freq from sysfs, it will end up calling the governor's CPUFREQ_GOV_LIMITS notification, which will also call __cpufreq_driver_target(). And hence we end up issuing concurrent calls to ->target(). Typically, platforms have the following logic in their ->target() routines: (Eg: cpufreq-cpu0, omap, exynos, etc) A. If new freq is more than old: Increase voltage B. Change freq C. If new freq is less than old: decrease voltage Now, if the two concurrent calls to ->target() are X and Y, where X is trying to increase the freq and Y is trying to decrease it, we get the following race condition: X.A: voltage gets increased for larger freq Y.A: nothing happens Y.B: freq gets decreased Y.C: voltage gets decreased X.B: freq gets increased X.C: nothing happens Thus we can end up setting a freq which is not supported by the voltage we have set. That will probably make the clock to the CPU unstable and the system might not work properly anymore. This patch introduces a set of synchronization primitives to serialize frequency transitions, which are to be used as shown below: cpufreq_freq_transition_begin(); //Perform the frequency change cpufreq_freq_transition_end(); The _begin() call sends the PRECHANGE notification whereas the _end() call sends the POSTCHANGE notification. Also, all the necessary synchronization is handled within these calls. In particular, even drivers which set the ASYNC_NOTIFICATION flag can also use these APIs for performing frequency transitions (ie., you can call _begin() from one task, and call the corresponding _end() from a different task). The actual synchronization underneath is not that complicated: The key challenge is to allow drivers to begin the transition from one thread and end it in a completely different thread (this is to enable drivers that do asynchronous POSTCHANGE notification from bottom-halves, to also use the same interface). To achieve this, a 'transition_ongoing' flag, a 'transition_lock' spinlock and a wait-queue are added per-policy. The flag and the wait-queue are used in conjunction to create an "uninterrupted flow" from _begin() to _end(). The spinlock is used to ensure that only one such "flow" is in flight at any given time. Put together, this provides us all the necessary synchronization. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-03-24 12:05:44 +04:00
spin_lock_init(&policy->transition_lock);
init_waitqueue_head(&policy->transition_wait);
init_completion(&policy->kobj_unregister);
INIT_WORK(&policy->update, handle_update);
policy->cpu = cpu;
return policy;
err_free_real_cpus:
free_cpumask_var(policy->real_cpus);
err_free_rcpumask:
free_cpumask_var(policy->related_cpus);
err_free_cpumask:
free_cpumask_var(policy->cpus);
err_free_policy:
kfree(policy);
return NULL;
}
static void cpufreq_policy_put_kobj(struct cpufreq_policy *policy)
cpufreq: remove sysfs files for CPUs which failed to come back after resume There are cases where cpufreq_add_dev() may fail for some CPUs during system resume. With the current code we will still have sysfs cpufreq files for those CPUs and struct cpufreq_policy would be already freed for them. Hence any operation on those sysfs files would result in kernel warnings. Example of problems resulting from resume errors (from Bjørn Mork): WARNING: CPU: 0 PID: 6055 at fs/sysfs/file.c:343 sysfs_open_file+0x77/0x212() missing sysfs attribute operations for kobject: (null) Modules linked in: [stripped as irrelevant] CPU: 0 PID: 6055 Comm: grep Tainted: G D 3.13.0-rc2 #153 Hardware name: LENOVO 2776LEG/2776LEG, BIOS 6EET55WW (3.15 ) 12/19/2011 0000000000000009 ffff8802327ebb78 ffffffff81380b0e 0000000000000006 ffff8802327ebbc8 ffff8802327ebbb8 ffffffff81038635 0000000000000000 ffffffff811823c7 ffff88021a19e688 ffff88021a19e688 ffff8802302f9310 Call Trace: [<ffffffff81380b0e>] dump_stack+0x55/0x76 [<ffffffff81038635>] warn_slowpath_common+0x7c/0x96 [<ffffffff811823c7>] ? sysfs_open_file+0x77/0x212 [<ffffffff810386e3>] warn_slowpath_fmt+0x41/0x43 [<ffffffff81182dec>] ? sysfs_get_active+0x6b/0x82 [<ffffffff81182382>] ? sysfs_open_file+0x32/0x212 [<ffffffff811823c7>] sysfs_open_file+0x77/0x212 [<ffffffff81182350>] ? sysfs_schedule_callback+0x1ac/0x1ac [<ffffffff81122562>] do_dentry_open+0x17c/0x257 [<ffffffff8112267e>] finish_open+0x41/0x4f [<ffffffff81130225>] do_last+0x80c/0x9ba [<ffffffff8112dbbd>] ? inode_permission+0x40/0x42 [<ffffffff81130606>] path_openat+0x233/0x4a1 [<ffffffff81130b7e>] do_filp_open+0x35/0x85 [<ffffffff8113b787>] ? __alloc_fd+0x172/0x184 [<ffffffff811232ea>] do_sys_open+0x6b/0xfa [<ffffffff811233a7>] SyS_openat+0xf/0x11 [<ffffffff8138c812>] system_call_fastpath+0x16/0x1b To fix this, remove those sysfs files or put the associated kobject in case of such errors. Also, to make it simple, remove the cpufreq sysfs links from all the CPUs (except for the policy->cpu) during suspend, as that operation won't result in a loss of sysfs file permissions and we can create those links during resume just fine. Fixes: 5302c3fb2e62 ("cpufreq: Perform light-weight init/teardown during suspend/resume") Reported-and-tested-by: Bjørn Mork <bjorn@mork.no> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Cc: 3.12+ <stable@vger.kernel.org> # 3.12+ [rjw: Changelog] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-12-20 19:56:02 +04:00
{
struct kobject *kobj;
struct completion *cmp;
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
down_write(&policy->rwsem);
cpufreq_stats_free_table(policy);
cpufreq: remove sysfs files for CPUs which failed to come back after resume There are cases where cpufreq_add_dev() may fail for some CPUs during system resume. With the current code we will still have sysfs cpufreq files for those CPUs and struct cpufreq_policy would be already freed for them. Hence any operation on those sysfs files would result in kernel warnings. Example of problems resulting from resume errors (from Bjørn Mork): WARNING: CPU: 0 PID: 6055 at fs/sysfs/file.c:343 sysfs_open_file+0x77/0x212() missing sysfs attribute operations for kobject: (null) Modules linked in: [stripped as irrelevant] CPU: 0 PID: 6055 Comm: grep Tainted: G D 3.13.0-rc2 #153 Hardware name: LENOVO 2776LEG/2776LEG, BIOS 6EET55WW (3.15 ) 12/19/2011 0000000000000009 ffff8802327ebb78 ffffffff81380b0e 0000000000000006 ffff8802327ebbc8 ffff8802327ebbb8 ffffffff81038635 0000000000000000 ffffffff811823c7 ffff88021a19e688 ffff88021a19e688 ffff8802302f9310 Call Trace: [<ffffffff81380b0e>] dump_stack+0x55/0x76 [<ffffffff81038635>] warn_slowpath_common+0x7c/0x96 [<ffffffff811823c7>] ? sysfs_open_file+0x77/0x212 [<ffffffff810386e3>] warn_slowpath_fmt+0x41/0x43 [<ffffffff81182dec>] ? sysfs_get_active+0x6b/0x82 [<ffffffff81182382>] ? sysfs_open_file+0x32/0x212 [<ffffffff811823c7>] sysfs_open_file+0x77/0x212 [<ffffffff81182350>] ? sysfs_schedule_callback+0x1ac/0x1ac [<ffffffff81122562>] do_dentry_open+0x17c/0x257 [<ffffffff8112267e>] finish_open+0x41/0x4f [<ffffffff81130225>] do_last+0x80c/0x9ba [<ffffffff8112dbbd>] ? inode_permission+0x40/0x42 [<ffffffff81130606>] path_openat+0x233/0x4a1 [<ffffffff81130b7e>] do_filp_open+0x35/0x85 [<ffffffff8113b787>] ? __alloc_fd+0x172/0x184 [<ffffffff811232ea>] do_sys_open+0x6b/0xfa [<ffffffff811233a7>] SyS_openat+0xf/0x11 [<ffffffff8138c812>] system_call_fastpath+0x16/0x1b To fix this, remove those sysfs files or put the associated kobject in case of such errors. Also, to make it simple, remove the cpufreq sysfs links from all the CPUs (except for the policy->cpu) during suspend, as that operation won't result in a loss of sysfs file permissions and we can create those links during resume just fine. Fixes: 5302c3fb2e62 ("cpufreq: Perform light-weight init/teardown during suspend/resume") Reported-and-tested-by: Bjørn Mork <bjorn@mork.no> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Cc: 3.12+ <stable@vger.kernel.org> # 3.12+ [rjw: Changelog] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-12-20 19:56:02 +04:00
kobj = &policy->kobj;
cmp = &policy->kobj_unregister;
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
up_write(&policy->rwsem);
cpufreq: remove sysfs files for CPUs which failed to come back after resume There are cases where cpufreq_add_dev() may fail for some CPUs during system resume. With the current code we will still have sysfs cpufreq files for those CPUs and struct cpufreq_policy would be already freed for them. Hence any operation on those sysfs files would result in kernel warnings. Example of problems resulting from resume errors (from Bjørn Mork): WARNING: CPU: 0 PID: 6055 at fs/sysfs/file.c:343 sysfs_open_file+0x77/0x212() missing sysfs attribute operations for kobject: (null) Modules linked in: [stripped as irrelevant] CPU: 0 PID: 6055 Comm: grep Tainted: G D 3.13.0-rc2 #153 Hardware name: LENOVO 2776LEG/2776LEG, BIOS 6EET55WW (3.15 ) 12/19/2011 0000000000000009 ffff8802327ebb78 ffffffff81380b0e 0000000000000006 ffff8802327ebbc8 ffff8802327ebbb8 ffffffff81038635 0000000000000000 ffffffff811823c7 ffff88021a19e688 ffff88021a19e688 ffff8802302f9310 Call Trace: [<ffffffff81380b0e>] dump_stack+0x55/0x76 [<ffffffff81038635>] warn_slowpath_common+0x7c/0x96 [<ffffffff811823c7>] ? sysfs_open_file+0x77/0x212 [<ffffffff810386e3>] warn_slowpath_fmt+0x41/0x43 [<ffffffff81182dec>] ? sysfs_get_active+0x6b/0x82 [<ffffffff81182382>] ? sysfs_open_file+0x32/0x212 [<ffffffff811823c7>] sysfs_open_file+0x77/0x212 [<ffffffff81182350>] ? sysfs_schedule_callback+0x1ac/0x1ac [<ffffffff81122562>] do_dentry_open+0x17c/0x257 [<ffffffff8112267e>] finish_open+0x41/0x4f [<ffffffff81130225>] do_last+0x80c/0x9ba [<ffffffff8112dbbd>] ? inode_permission+0x40/0x42 [<ffffffff81130606>] path_openat+0x233/0x4a1 [<ffffffff81130b7e>] do_filp_open+0x35/0x85 [<ffffffff8113b787>] ? __alloc_fd+0x172/0x184 [<ffffffff811232ea>] do_sys_open+0x6b/0xfa [<ffffffff811233a7>] SyS_openat+0xf/0x11 [<ffffffff8138c812>] system_call_fastpath+0x16/0x1b To fix this, remove those sysfs files or put the associated kobject in case of such errors. Also, to make it simple, remove the cpufreq sysfs links from all the CPUs (except for the policy->cpu) during suspend, as that operation won't result in a loss of sysfs file permissions and we can create those links during resume just fine. Fixes: 5302c3fb2e62 ("cpufreq: Perform light-weight init/teardown during suspend/resume") Reported-and-tested-by: Bjørn Mork <bjorn@mork.no> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Cc: 3.12+ <stable@vger.kernel.org> # 3.12+ [rjw: Changelog] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-12-20 19:56:02 +04:00
kobject_put(kobj);
/*
* We need to make sure that the underlying kobj is
* actually not referenced anymore by anybody before we
* proceed with unloading.
*/
pr_debug("waiting for dropping of refcount\n");
wait_for_completion(cmp);
pr_debug("wait complete\n");
}
static void cpufreq_policy_free(struct cpufreq_policy *policy)
{
unsigned long flags;
int cpu;
/* Remove policy from list */
write_lock_irqsave(&cpufreq_driver_lock, flags);
list_del(&policy->policy_list);
for_each_cpu(cpu, policy->related_cpus)
per_cpu(cpufreq_cpu_data, cpu) = NULL;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
cpufreq_policy_put_kobj(policy);
cpufreq: Avoid attempts to create duplicate symbolic links After commit 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) there is a problem with CPUs that share cpufreq policy objects with other CPUs and are initially offline. Say CPU1 shares a policy with CPU0 which is online and is registered first. As part of the registration process, cpufreq_add_dev() is called for it. It creates the policy object and a symbolic link to it from the CPU1's sysfs directory. If CPU1 is registered subsequently and it is offline at that time, cpufreq_add_dev() will attempt to create a symbolic link to the policy object for it, but that link is present already, so a warning about that will be triggered. To avoid that warning, make cpufreq use an additional CPU mask containing related CPUs that are actually present for each policy object. That mask is initialized when the policy object is populated after its creation (for the first online CPU using it) and it includes CPUs from the "policy CPUs" mask returned by the cpufreq driver's ->init() callback that are physically present at that time. Symbolic links to the policy are created only for the CPUs in that mask. If cpufreq_add_dev() is invoked for an offline CPU, it checks the new mask and only creates the symlink if the CPU was not in it (the CPU is added to the mask at the same time). In turn, cpufreq_remove_dev() drops the given CPU from the new mask, removes its symlink to the policy object and returns, unless it is the CPU owning the policy object. In that case, the policy object is moved to a new CPU's sysfs directory or deleted if the CPU being removed was the last user of the policy. While at it, notice that cpufreq_remove_dev() can't fail, because its return value is ignored, so make it ignore return values from __cpufreq_remove_dev_prepare() and __cpufreq_remove_dev_finish() and prevent these functions from aborting on errors returned by __cpufreq_governor(). Also drop the now unused sif argument from them. Fixes: 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reported-and-tested-by: Russell King <linux@arm.linux.org.uk> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2015-07-26 03:07:47 +03:00
free_cpumask_var(policy->real_cpus);
free_cpumask_var(policy->related_cpus);
free_cpumask_var(policy->cpus);
kfree(policy);
}
static int cpufreq_online(unsigned int cpu)
{
struct cpufreq_policy *policy;
bool new_policy;
unsigned long flags;
unsigned int j;
int ret;
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
pr_debug("%s: bringing CPU%u online\n", __func__, cpu);
/* Check if this CPU already has a policy to manage it */
policy = per_cpu(cpufreq_cpu_data, cpu);
if (policy) {
WARN_ON(!cpumask_test_cpu(cpu, policy->related_cpus));
if (!policy_is_inactive(policy))
return cpufreq_add_policy_cpu(policy, cpu);
/* This is the only online CPU for the policy. Start over. */
new_policy = false;
down_write(&policy->rwsem);
policy->cpu = cpu;
policy->governor = NULL;
up_write(&policy->rwsem);
} else {
new_policy = true;
policy = cpufreq_policy_alloc(cpu);
if (!policy)
return -ENOMEM;
}
cpufreq: Fix crash in cpufreq-stats during suspend/resume Stephen Warren reported that the cpufreq-stats code hits a NULL pointer dereference during the second attempt to suspend a system. He also pin-pointed the problem to commit 5302c3f "cpufreq: Perform light-weight init/teardown during suspend/resume". That commit actually ensured that the cpufreq-stats table and the cpufreq-stats sysfs entries are *not* torn down (ie., not freed) during suspend/resume, which makes it all the more surprising. However, it turns out that the root-cause is not that we access an already freed memory, but that the reference to the allocated memory gets moved around and we lose track of that during resume, leading to the reported crash in a subsequent suspend attempt. In the suspend path, during CPU offline, the value of policy->cpu is updated by choosing one of the surviving CPUs in that policy, as long as there is atleast one CPU in that policy. And cpufreq_stats_update_policy_cpu() is invoked to update the reference to the stats structure by assigning it to the new CPU. However, in the resume path, during CPU online, we end up assigning a fresh CPU as the policy->cpu, without letting cpufreq-stats know about this. Thus the reference to the stats structure remains (incorrectly) associated with the old CPU. So, in a subsequent suspend attempt, during CPU offline, we end up accessing an incorrect location to get the stats structure, which eventually leads to the NULL pointer dereference. Fix this by letting cpufreq-stats know about the update of the policy->cpu during CPU online in the resume path. (Also, move the update_policy_cpu() function higher up in the file, so that __cpufreq_add_dev() can invoke it). Reported-and-tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-09-12 00:12:59 +04:00
if (!new_policy && cpufreq_driver->online) {
ret = cpufreq_driver->online(policy);
if (ret) {
pr_debug("%s: %d: initialization failed\n", __func__,
__LINE__);
goto out_exit_policy;
}
/* Recover policy->cpus using related_cpus */
cpumask_copy(policy->cpus, policy->related_cpus);
} else {
cpumask_copy(policy->cpus, cpumask_of(cpu));
/*
* Call driver. From then on the cpufreq must be able
* to accept all calls to ->verify and ->setpolicy for this CPU.
*/
ret = cpufreq_driver->init(policy);
if (ret) {
pr_debug("%s: %d: initialization failed\n", __func__,
__LINE__);
goto out_free_policy;
}
ret = cpufreq_table_validate_and_sort(policy);
if (ret)
goto out_exit_policy;
cpufreq: Ref the policy object sooner Do it before it's assigned to cpufreq_cpu_data, otherwise when a driver tries to get the cpu frequency during initialization the policy kobj is referenced and we get this warning: ------------[ cut here ]------------ WARNING: CPU: 1 PID: 64 at include/linux/kref.h:47 kobject_get+0x64/0x70() Modules linked in: CPU: 1 PID: 64 Comm: irq/77-tegra-ac Not tainted 3.18.0-rc4-next-20141114ccu-00050-g3eff942 #326 [<c0016fac>] (unwind_backtrace) from [<c001272c>] (show_stack+0x10/0x14) [<c001272c>] (show_stack) from [<c06085d8>] (dump_stack+0x98/0xd8) [<c06085d8>] (dump_stack) from [<c002892c>] (warn_slowpath_common+0x84/0xb4) [<c002892c>] (warn_slowpath_common) from [<c00289f8>] (warn_slowpath_null+0x1c/0x24) [<c00289f8>] (warn_slowpath_null) from [<c0220290>] (kobject_get+0x64/0x70) [<c0220290>] (kobject_get) from [<c03e944c>] (cpufreq_cpu_get+0x88/0xc8) [<c03e944c>] (cpufreq_cpu_get) from [<c03e9500>] (cpufreq_get+0xc/0x64) [<c03e9500>] (cpufreq_get) from [<c0285288>] (actmon_thread_isr+0x134/0x198) [<c0285288>] (actmon_thread_isr) from [<c0069008>] (irq_thread_fn+0x1c/0x40) [<c0069008>] (irq_thread_fn) from [<c0069324>] (irq_thread+0x134/0x174) [<c0069324>] (irq_thread) from [<c0040290>] (kthread+0xdc/0xf4) [<c0040290>] (kthread) from [<c000f4b8>] (ret_from_fork+0x14/0x3c) ---[ end trace b7bd64a81b340c59 ]--- Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-11-24 12:08:03 +03:00
/* related_cpus should at least include policy->cpus. */
cpumask_copy(policy->related_cpus, policy->cpus);
}
cpufreq: Avoid attempts to create duplicate symbolic links After commit 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) there is a problem with CPUs that share cpufreq policy objects with other CPUs and are initially offline. Say CPU1 shares a policy with CPU0 which is online and is registered first. As part of the registration process, cpufreq_add_dev() is called for it. It creates the policy object and a symbolic link to it from the CPU1's sysfs directory. If CPU1 is registered subsequently and it is offline at that time, cpufreq_add_dev() will attempt to create a symbolic link to the policy object for it, but that link is present already, so a warning about that will be triggered. To avoid that warning, make cpufreq use an additional CPU mask containing related CPUs that are actually present for each policy object. That mask is initialized when the policy object is populated after its creation (for the first online CPU using it) and it includes CPUs from the "policy CPUs" mask returned by the cpufreq driver's ->init() callback that are physically present at that time. Symbolic links to the policy are created only for the CPUs in that mask. If cpufreq_add_dev() is invoked for an offline CPU, it checks the new mask and only creates the symlink if the CPU was not in it (the CPU is added to the mask at the same time). In turn, cpufreq_remove_dev() drops the given CPU from the new mask, removes its symlink to the policy object and returns, unless it is the CPU owning the policy object. In that case, the policy object is moved to a new CPU's sysfs directory or deleted if the CPU being removed was the last user of the policy. While at it, notice that cpufreq_remove_dev() can't fail, because its return value is ignored, so make it ignore return values from __cpufreq_remove_dev_prepare() and __cpufreq_remove_dev_finish() and prevent these functions from aborting on errors returned by __cpufreq_governor(). Also drop the now unused sif argument from them. Fixes: 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reported-and-tested-by: Russell King <linux@arm.linux.org.uk> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2015-07-26 03:07:47 +03:00
down_write(&policy->rwsem);
/*
* affected cpus must always be the one, which are online. We aren't
* managing offline cpus here.
*/
cpumask_and(policy->cpus, policy->cpus, cpu_online_mask);
if (new_policy) {
policy->user_policy.min = policy->min;
policy->user_policy.max = policy->max;
cpufreq: Ref the policy object sooner Do it before it's assigned to cpufreq_cpu_data, otherwise when a driver tries to get the cpu frequency during initialization the policy kobj is referenced and we get this warning: ------------[ cut here ]------------ WARNING: CPU: 1 PID: 64 at include/linux/kref.h:47 kobject_get+0x64/0x70() Modules linked in: CPU: 1 PID: 64 Comm: irq/77-tegra-ac Not tainted 3.18.0-rc4-next-20141114ccu-00050-g3eff942 #326 [<c0016fac>] (unwind_backtrace) from [<c001272c>] (show_stack+0x10/0x14) [<c001272c>] (show_stack) from [<c06085d8>] (dump_stack+0x98/0xd8) [<c06085d8>] (dump_stack) from [<c002892c>] (warn_slowpath_common+0x84/0xb4) [<c002892c>] (warn_slowpath_common) from [<c00289f8>] (warn_slowpath_null+0x1c/0x24) [<c00289f8>] (warn_slowpath_null) from [<c0220290>] (kobject_get+0x64/0x70) [<c0220290>] (kobject_get) from [<c03e944c>] (cpufreq_cpu_get+0x88/0xc8) [<c03e944c>] (cpufreq_cpu_get) from [<c03e9500>] (cpufreq_get+0xc/0x64) [<c03e9500>] (cpufreq_get) from [<c0285288>] (actmon_thread_isr+0x134/0x198) [<c0285288>] (actmon_thread_isr) from [<c0069008>] (irq_thread_fn+0x1c/0x40) [<c0069008>] (irq_thread_fn) from [<c0069324>] (irq_thread+0x134/0x174) [<c0069324>] (irq_thread) from [<c0040290>] (kthread+0xdc/0xf4) [<c0040290>] (kthread) from [<c000f4b8>] (ret_from_fork+0x14/0x3c) ---[ end trace b7bd64a81b340c59 ]--- Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-11-24 12:08:03 +03:00
for_each_cpu(j, policy->related_cpus) {
per_cpu(cpufreq_cpu_data, j) = policy;
add_cpu_dev_symlink(policy, j);
}
} else {
policy->min = policy->user_policy.min;
policy->max = policy->user_policy.max;
}
cpufreq: Skip current frequency initialization for ->setpolicy drivers After commit da60ce9f2fac (cpufreq: call cpufreq_driver->get() after calling ->init()) __cpufreq_add_dev() sometimes fails for CPUs handled by intel_pstate, because that driver may return 0 from its ->get() callback if it has not run long enough to collect enough samples on the given CPU. That didn't happen before commit da60ce9f2fac which added policy->cur initialization to __cpufreq_add_dev() to help reduce code duplication in other cpufreq drivers. However, the code added by commit da60ce9f2fac need not be executed for cpufreq drivers having the ->setpolicy callback defined, because the subsequent invocation of cpufreq_set_policy() will use that callback to initialize the policy anyway and it doesn't need policy->cur to be initialized upfront. The analogous code in cpufreq_update_policy() is also unnecessary for cpufreq drivers having ->setpolicy set and may be skipped for them as well. Since intel_pstate provides ->setpolicy, skipping the upfront policy->cur initialization for cpufreq drivers with that callback set will cover intel_pstate and the problem it's been having after commit da60ce9f2fac will be addressed. Fixes: da60ce9f2fac (cpufreq: call cpufreq_driver->get() after calling ->init()) References: https://bugzilla.kernel.org/show_bug.cgi?id=71931 Reported-and-tested-by: Patrik Lundquist <patrik.lundquist@gmail.com> Acked-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Cc: 3.13+ <stable@vger.kernel.org> # 3.13+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-03-13 00:49:33 +04:00
if (cpufreq_driver->get && !cpufreq_driver->setpolicy) {
policy->cur = cpufreq_driver->get(policy->cpu);
if (!policy->cur) {
pr_err("%s: ->get() failed\n", __func__);
goto out_destroy_policy;
}
}
/*
* Sometimes boot loaders set CPU frequency to a value outside of
* frequency table present with cpufreq core. In such cases CPU might be
* unstable if it has to run on that frequency for long duration of time
* and so its better to set it to a frequency which is specified in
* freq-table. This also makes cpufreq stats inconsistent as
* cpufreq-stats would fail to register because current frequency of CPU
* isn't found in freq-table.
*
* Because we don't want this change to effect boot process badly, we go
* for the next freq which is >= policy->cur ('cur' must be set by now,
* otherwise we will end up setting freq to lowest of the table as 'cur'
* is initialized to zero).
*
* We are passing target-freq as "policy->cur - 1" otherwise
* __cpufreq_driver_target() would simply fail, as policy->cur will be
* equal to target-freq.
*/
if ((cpufreq_driver->flags & CPUFREQ_NEED_INITIAL_FREQ_CHECK)
&& has_target()) {
/* Are we running at unknown frequency ? */
ret = cpufreq_frequency_table_get_index(policy, policy->cur);
if (ret == -EINVAL) {
/* Warn user and fix it */
pr_warn("%s: CPU%d: Running at unlisted freq: %u KHz\n",
__func__, policy->cpu, policy->cur);
ret = __cpufreq_driver_target(policy, policy->cur - 1,
CPUFREQ_RELATION_L);
/*
* Reaching here after boot in a few seconds may not
* mean that system will remain stable at "unknown"
* frequency for longer duration. Hence, a BUG_ON().
*/
BUG_ON(ret);
pr_warn("%s: CPU%d: Unlisted initial frequency changed to: %u KHz\n",
__func__, policy->cpu, policy->cur);
}
}
if (new_policy) {
ret = cpufreq_add_dev_interface(policy);
if (ret)
goto out_destroy_policy;
cpufreq_stats_create_table(policy);
write_lock_irqsave(&cpufreq_driver_lock, flags);
list_add(&policy->policy_list, &cpufreq_policy_list);
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
}
ret = cpufreq_init_policy(policy);
if (ret) {
pr_err("%s: Failed to initialize policy for cpu: %d (%d)\n",
__func__, cpu, ret);
goto out_destroy_policy;
}
cpufreq: Initialize governor for a new policy under policy->rwsem policy->rwsem is used to lock access to all parts of code modifying struct cpufreq_policy, but it's not used on a new policy created by __cpufreq_add_dev(). Because of that, if cpufreq_update_policy() is called in a tight loop on one CPU in parallel with offline/online of another CPU, then the following crash can be triggered: Unable to handle kernel NULL pointer dereference at virtual address 00000020 pgd = c0003000 [00000020] *pgd=80000000004003, *pmd=00000000 Internal error: Oops: 206 [#1] PREEMPT SMP ARM PC is at __cpufreq_governor+0x10/0x1ac LR is at cpufreq_update_policy+0x114/0x150 ---[ end trace f23a8defea6cd706 ]--- Kernel panic - not syncing: Fatal exception CPU0: stopping CPU: 0 PID: 7136 Comm: mpdecision Tainted: G D W 3.10.0-gd727407-00074-g979ede8 #396 [<c0afe180>] (notifier_call_chain+0x40/0x68) from [<c02a23ac>] (__blocking_notifier_call_chain+0x40/0x58) [<c02a23ac>] (__blocking_notifier_call_chain+0x40/0x58) from [<c02a23d8>] (blocking_notifier_call_chain+0x14/0x1c) [<c02a23d8>] (blocking_notifier_call_chain+0x14/0x1c) from [<c0803c68>] (cpufreq_set_policy+0xd4/0x2b8) [<c0803c68>] (cpufreq_set_policy+0xd4/0x2b8) from [<c0803e7c>] (cpufreq_init_policy+0x30/0x98) [<c0803e7c>] (cpufreq_init_policy+0x30/0x98) from [<c0805a18>] (__cpufreq_add_dev.isra.17+0x4dc/0x7a4) [<c0805a18>] (__cpufreq_add_dev.isra.17+0x4dc/0x7a4) from [<c0805d38>] (cpufreq_cpu_callback+0x58/0x84) [<c0805d38>] (cpufreq_cpu_callback+0x58/0x84) from [<c0afe180>] (notifier_call_chain+0x40/0x68) [<c0afe180>] (notifier_call_chain+0x40/0x68) from [<c02812dc>] (__cpu_notify+0x28/0x44) [<c02812dc>] (__cpu_notify+0x28/0x44) from [<c0aeed90>] (_cpu_up+0xf4/0x1dc) [<c0aeed90>] (_cpu_up+0xf4/0x1dc) from [<c0aeeed4>] (cpu_up+0x5c/0x78) [<c0aeeed4>] (cpu_up+0x5c/0x78) from [<c0aec808>] (store_online+0x44/0x74) [<c0aec808>] (store_online+0x44/0x74) from [<c03a40f4>] (sysfs_write_file+0x108/0x14c) [<c03a40f4>] (sysfs_write_file+0x108/0x14c) from [<c03517d4>] (vfs_write+0xd0/0x180) [<c03517d4>] (vfs_write+0xd0/0x180) from [<c0351ca8>] (SyS_write+0x38/0x68) [<c0351ca8>] (SyS_write+0x38/0x68) from [<c0205de0>] (ret_fast_syscall+0x0/0x30) Fix that by taking locks at appropriate places in __cpufreq_add_dev() as well. Reported-by: Saravana Kannan <skannan@codeaurora.org> Suggested-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> [rjw: Changelog] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-03-04 07:44:01 +04:00
up_write(&policy->rwsem);
kobject_uevent(&policy->kobj, KOBJ_ADD);
/* Callback for handling stuff after policy is ready */
if (cpufreq_driver->ready)
cpufreq_driver->ready(policy);
if (IS_ENABLED(CONFIG_CPU_THERMAL) &&
cpufreq_driver->flags & CPUFREQ_IS_COOLING_DEV)
policy->cdev = of_cpufreq_cooling_register(policy);
pr_debug("initialization complete\n");
return 0;
out_destroy_policy:
for_each_cpu(j, policy->real_cpus)
remove_cpu_dev_symlink(policy, get_cpu_device(j));
cpufreq: release policy->rwsem on error While debugging a cpufreq-related hardware failure on a system I saw the following lockdep warning: ========================= [ BUG: held lock freed! ] 3.17.0-rc4+ #1 Tainted: G E ------------------------- insmod/2247 is freeing memory ffff88006e1b1400-ffff88006e1b17ff, with a lock still held there! (&policy->rwsem){+.+...}, at: [<ffffffff8156d37d>] __cpufreq_add_dev.isra.21+0x47d/0xb80 3 locks held by insmod/2247: #0: (subsys mutex#5){+.+.+.}, at: [<ffffffff81485579>] subsys_interface_register+0x69/0x120 #1: (cpufreq_rwsem){.+.+.+}, at: [<ffffffff8156cf73>] __cpufreq_add_dev.isra.21+0x73/0xb80 #2: (&policy->rwsem){+.+...}, at: [<ffffffff8156d37d>] __cpufreq_add_dev.isra.21+0x47d/0xb80 stack backtrace: CPU: 0 PID: 2247 Comm: insmod Tainted: G E 3.17.0-rc4+ #1 Hardware name: HP ProLiant MicroServer Gen8, BIOS J06 08/24/2013 0000000000000000 000000008f3063c4 ffff88006f87bb30 ffffffff8171b358 ffff88006bcf3750 ffff88006f87bb68 ffffffff810e09e1 ffff88006e1b1400 ffffea0001b86c00 ffffffff8156d327 ffff880073003500 0000000000000246 Call Trace: [<ffffffff8171b358>] dump_stack+0x4d/0x66 [<ffffffff810e09e1>] debug_check_no_locks_freed+0x171/0x180 [<ffffffff8156d327>] ? __cpufreq_add_dev.isra.21+0x427/0xb80 [<ffffffff8121412b>] kfree+0xab/0x2b0 [<ffffffff8156d327>] __cpufreq_add_dev.isra.21+0x427/0xb80 [<ffffffff81724cf7>] ? _raw_spin_unlock+0x27/0x40 [<ffffffffa003517f>] ? pcc_cpufreq_do_osc+0x17f/0x17f [pcc_cpufreq] [<ffffffff8156da8e>] cpufreq_add_dev+0xe/0x10 [<ffffffff814855d1>] subsys_interface_register+0xc1/0x120 [<ffffffff8156bcf2>] cpufreq_register_driver+0x112/0x340 [<ffffffff8121415a>] ? kfree+0xda/0x2b0 [<ffffffffa003517f>] ? pcc_cpufreq_do_osc+0x17f/0x17f [pcc_cpufreq] [<ffffffffa003562e>] pcc_cpufreq_init+0x4af/0xe81 [pcc_cpufreq] [<ffffffffa003517f>] ? pcc_cpufreq_do_osc+0x17f/0x17f [pcc_cpufreq] [<ffffffff81002144>] do_one_initcall+0xd4/0x210 [<ffffffff811f7472>] ? __vunmap+0xd2/0x120 [<ffffffff81127155>] load_module+0x1315/0x1b70 [<ffffffff811222a0>] ? store_uevent+0x70/0x70 [<ffffffff811229d9>] ? copy_module_from_fd.isra.44+0x129/0x180 [<ffffffff81127b86>] SyS_finit_module+0xa6/0xd0 [<ffffffff81725b69>] system_call_fastpath+0x16/0x1b cpufreq: __cpufreq_add_dev: ->get() failed insmod: ERROR: could not insert module pcc-cpufreq.ko: No such device The warning occurs in the __cpufreq_add_dev() code which does down_write(&policy->rwsem); ... if (cpufreq_driver->get && !cpufreq_driver->setpolicy) { policy->cur = cpufreq_driver->get(policy->cpu); if (!policy->cur) { pr_err("%s: ->get() failed\n", __func__); goto err_get_freq; } If cpufreq_driver->get(policy->cpu) returns an error we execute the code at err_get_freq, which does not up the policy->rwsem. This causes the lockdep warning. Trivial patch to up the policy->rwsem in the error path. After the patch has been applied, and an error occurs in the cpufreq_driver->get(policy->cpu) call we will now see cpufreq: __cpufreq_add_dev: ->get() failed cpufreq: __cpufreq_add_dev: ->get() failed modprobe: ERROR: could not insert 'pcc_cpufreq': No such device Fixes: 4e97b631f24c (cpufreq: Initialize governor for a new policy under policy->rwsem) Signed-off-by: Prarit Bhargava <prarit@redhat.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Cc: 3.14+ <stable@vger.kernel.org> # 3.14+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-09-10 18:12:08 +04:00
up_write(&policy->rwsem);
out_exit_policy:
if (cpufreq_driver->exit)
cpufreq_driver->exit(policy);
out_free_policy:
cpufreq_policy_free(policy);
return ret;
}
/**
* cpufreq_add_dev - the cpufreq interface for a CPU device.
* @dev: CPU device.
* @sif: Subsystem interface structure pointer (not used)
*/
static int cpufreq_add_dev(struct device *dev, struct subsys_interface *sif)
{
struct cpufreq_policy *policy;
unsigned cpu = dev->id;
cpufreq: create link to policy only for registered CPUs If a cpufreq driver is registered very early in the boot stage (e.g. registered from postcore_initcall()), then cpufreq core may generate kernel warnings for it. In this case, the CPUs are brought online, then the cpufreq driver is registered, and then the CPU topology devices are registered. However, by the time cpufreq_add_dev() gets called, the cpu device isn't stored in the per-cpu variable (cpu_sys_devices,) which is read by get_cpu_device(). So the cpufreq core fails to get device for the CPU, for which cpufreq_add_dev() was called in the first place and we will hit a WARN_ON(!cpu_dev). Even if we reuse the 'dev' parameter passed to cpufreq_add_dev() to avoid that warning, there might be other CPUs online that share the policy with the cpu for which cpufreq_add_dev() is called. Eventually get_cpu_device() will return NULL for them as well, and we will hit the same WARN_ON() again. In order to fix these issues, change cpufreq core to create links to the policy for a cpu only when cpufreq_add_dev() is called for that CPU. Reuse the 'real_cpus' mask to track that as well. Note that cpufreq_remove_dev() already handles removal of the links for individual CPUs and cpufreq_add_dev() has aligned with that now. Reported-by: Russell King <rmk+kernel@arm.linux.org.uk> Tested-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-09-12 09:37:05 +03:00
int ret;
dev_dbg(dev, "%s: adding CPU%u\n", __func__, cpu);
cpufreq: create link to policy only for registered CPUs If a cpufreq driver is registered very early in the boot stage (e.g. registered from postcore_initcall()), then cpufreq core may generate kernel warnings for it. In this case, the CPUs are brought online, then the cpufreq driver is registered, and then the CPU topology devices are registered. However, by the time cpufreq_add_dev() gets called, the cpu device isn't stored in the per-cpu variable (cpu_sys_devices,) which is read by get_cpu_device(). So the cpufreq core fails to get device for the CPU, for which cpufreq_add_dev() was called in the first place and we will hit a WARN_ON(!cpu_dev). Even if we reuse the 'dev' parameter passed to cpufreq_add_dev() to avoid that warning, there might be other CPUs online that share the policy with the cpu for which cpufreq_add_dev() is called. Eventually get_cpu_device() will return NULL for them as well, and we will hit the same WARN_ON() again. In order to fix these issues, change cpufreq core to create links to the policy for a cpu only when cpufreq_add_dev() is called for that CPU. Reuse the 'real_cpus' mask to track that as well. Note that cpufreq_remove_dev() already handles removal of the links for individual CPUs and cpufreq_add_dev() has aligned with that now. Reported-by: Russell King <rmk+kernel@arm.linux.org.uk> Tested-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-09-12 09:37:05 +03:00
if (cpu_online(cpu)) {
ret = cpufreq_online(cpu);
if (ret)
return ret;
}
cpufreq: create link to policy only for registered CPUs If a cpufreq driver is registered very early in the boot stage (e.g. registered from postcore_initcall()), then cpufreq core may generate kernel warnings for it. In this case, the CPUs are brought online, then the cpufreq driver is registered, and then the CPU topology devices are registered. However, by the time cpufreq_add_dev() gets called, the cpu device isn't stored in the per-cpu variable (cpu_sys_devices,) which is read by get_cpu_device(). So the cpufreq core fails to get device for the CPU, for which cpufreq_add_dev() was called in the first place and we will hit a WARN_ON(!cpu_dev). Even if we reuse the 'dev' parameter passed to cpufreq_add_dev() to avoid that warning, there might be other CPUs online that share the policy with the cpu for which cpufreq_add_dev() is called. Eventually get_cpu_device() will return NULL for them as well, and we will hit the same WARN_ON() again. In order to fix these issues, change cpufreq core to create links to the policy for a cpu only when cpufreq_add_dev() is called for that CPU. Reuse the 'real_cpus' mask to track that as well. Note that cpufreq_remove_dev() already handles removal of the links for individual CPUs and cpufreq_add_dev() has aligned with that now. Reported-by: Russell King <rmk+kernel@arm.linux.org.uk> Tested-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-09-12 09:37:05 +03:00
/* Create sysfs link on CPU registration */
policy = per_cpu(cpufreq_cpu_data, cpu);
if (policy)
add_cpu_dev_symlink(policy, cpu);
cpufreq: create link to policy only for registered CPUs If a cpufreq driver is registered very early in the boot stage (e.g. registered from postcore_initcall()), then cpufreq core may generate kernel warnings for it. In this case, the CPUs are brought online, then the cpufreq driver is registered, and then the CPU topology devices are registered. However, by the time cpufreq_add_dev() gets called, the cpu device isn't stored in the per-cpu variable (cpu_sys_devices,) which is read by get_cpu_device(). So the cpufreq core fails to get device for the CPU, for which cpufreq_add_dev() was called in the first place and we will hit a WARN_ON(!cpu_dev). Even if we reuse the 'dev' parameter passed to cpufreq_add_dev() to avoid that warning, there might be other CPUs online that share the policy with the cpu for which cpufreq_add_dev() is called. Eventually get_cpu_device() will return NULL for them as well, and we will hit the same WARN_ON() again. In order to fix these issues, change cpufreq core to create links to the policy for a cpu only when cpufreq_add_dev() is called for that CPU. Reuse the 'real_cpus' mask to track that as well. Note that cpufreq_remove_dev() already handles removal of the links for individual CPUs and cpufreq_add_dev() has aligned with that now. Reported-by: Russell King <rmk+kernel@arm.linux.org.uk> Tested-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-09-12 09:37:05 +03:00
return 0;
}
static int cpufreq_offline(unsigned int cpu)
{
struct cpufreq_policy *policy;
int ret;
pr_debug("%s: unregistering CPU %u\n", __func__, cpu);
policy = cpufreq_cpu_get_raw(cpu);
if (!policy) {
pr_debug("%s: No cpu_data found\n", __func__);
return 0;
}
down_write(&policy->rwsem);
if (has_target())
cpufreq_stop_governor(policy);
cpumask_clear_cpu(cpu, policy->cpus);
if (policy_is_inactive(policy)) {
if (has_target())
strncpy(policy->last_governor, policy->governor->name,
CPUFREQ_NAME_LEN);
else
policy->last_policy = policy->policy;
} else if (cpu == policy->cpu) {
/* Nominate new CPU */
policy->cpu = cpumask_any(policy->cpus);
}
/* Start governor again for active policy */
if (!policy_is_inactive(policy)) {
if (has_target()) {
ret = cpufreq_start_governor(policy);
if (ret)
pr_err("%s: Failed to start governor\n", __func__);
}
goto unlock;
}
if (IS_ENABLED(CONFIG_CPU_THERMAL) &&
cpufreq_driver->flags & CPUFREQ_IS_COOLING_DEV) {
cpufreq_cooling_unregister(policy->cdev);
policy->cdev = NULL;
}
if (cpufreq_driver->stop_cpu)
cpufreq_driver->stop_cpu(policy);
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
if (has_target())
cpufreq_exit_governor(policy);
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
/*
* Perform the ->offline() during light-weight tear-down, as
* that allows fast recovery when the CPU comes back.
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
*/
if (cpufreq_driver->offline) {
cpufreq_driver->offline(policy);
} else if (cpufreq_driver->exit) {
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
cpufreq_driver->exit(policy);
policy->freq_table = NULL;
}
unlock:
up_write(&policy->rwsem);
return 0;
}
/**
* cpufreq_remove_dev - remove a CPU device
*
* Removes the cpufreq interface for a CPU device.
*/
static void cpufreq_remove_dev(struct device *dev, struct subsys_interface *sif)
{
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 02:29:42 +04:00
unsigned int cpu = dev->id;
cpufreq: Avoid attempts to create duplicate symbolic links After commit 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) there is a problem with CPUs that share cpufreq policy objects with other CPUs and are initially offline. Say CPU1 shares a policy with CPU0 which is online and is registered first. As part of the registration process, cpufreq_add_dev() is called for it. It creates the policy object and a symbolic link to it from the CPU1's sysfs directory. If CPU1 is registered subsequently and it is offline at that time, cpufreq_add_dev() will attempt to create a symbolic link to the policy object for it, but that link is present already, so a warning about that will be triggered. To avoid that warning, make cpufreq use an additional CPU mask containing related CPUs that are actually present for each policy object. That mask is initialized when the policy object is populated after its creation (for the first online CPU using it) and it includes CPUs from the "policy CPUs" mask returned by the cpufreq driver's ->init() callback that are physically present at that time. Symbolic links to the policy are created only for the CPUs in that mask. If cpufreq_add_dev() is invoked for an offline CPU, it checks the new mask and only creates the symlink if the CPU was not in it (the CPU is added to the mask at the same time). In turn, cpufreq_remove_dev() drops the given CPU from the new mask, removes its symlink to the policy object and returns, unless it is the CPU owning the policy object. In that case, the policy object is moved to a new CPU's sysfs directory or deleted if the CPU being removed was the last user of the policy. While at it, notice that cpufreq_remove_dev() can't fail, because its return value is ignored, so make it ignore return values from __cpufreq_remove_dev_prepare() and __cpufreq_remove_dev_finish() and prevent these functions from aborting on errors returned by __cpufreq_governor(). Also drop the now unused sif argument from them. Fixes: 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reported-and-tested-by: Russell King <linux@arm.linux.org.uk> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2015-07-26 03:07:47 +03:00
struct cpufreq_policy *policy = per_cpu(cpufreq_cpu_data, cpu);
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
cpufreq: Avoid attempts to create duplicate symbolic links After commit 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) there is a problem with CPUs that share cpufreq policy objects with other CPUs and are initially offline. Say CPU1 shares a policy with CPU0 which is online and is registered first. As part of the registration process, cpufreq_add_dev() is called for it. It creates the policy object and a symbolic link to it from the CPU1's sysfs directory. If CPU1 is registered subsequently and it is offline at that time, cpufreq_add_dev() will attempt to create a symbolic link to the policy object for it, but that link is present already, so a warning about that will be triggered. To avoid that warning, make cpufreq use an additional CPU mask containing related CPUs that are actually present for each policy object. That mask is initialized when the policy object is populated after its creation (for the first online CPU using it) and it includes CPUs from the "policy CPUs" mask returned by the cpufreq driver's ->init() callback that are physically present at that time. Symbolic links to the policy are created only for the CPUs in that mask. If cpufreq_add_dev() is invoked for an offline CPU, it checks the new mask and only creates the symlink if the CPU was not in it (the CPU is added to the mask at the same time). In turn, cpufreq_remove_dev() drops the given CPU from the new mask, removes its symlink to the policy object and returns, unless it is the CPU owning the policy object. In that case, the policy object is moved to a new CPU's sysfs directory or deleted if the CPU being removed was the last user of the policy. While at it, notice that cpufreq_remove_dev() can't fail, because its return value is ignored, so make it ignore return values from __cpufreq_remove_dev_prepare() and __cpufreq_remove_dev_finish() and prevent these functions from aborting on errors returned by __cpufreq_governor(). Also drop the now unused sif argument from them. Fixes: 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reported-and-tested-by: Russell King <linux@arm.linux.org.uk> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2015-07-26 03:07:47 +03:00
if (!policy)
return;
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
if (cpu_online(cpu))
cpufreq_offline(cpu);
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
cpufreq: Avoid attempts to create duplicate symbolic links After commit 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) there is a problem with CPUs that share cpufreq policy objects with other CPUs and are initially offline. Say CPU1 shares a policy with CPU0 which is online and is registered first. As part of the registration process, cpufreq_add_dev() is called for it. It creates the policy object and a symbolic link to it from the CPU1's sysfs directory. If CPU1 is registered subsequently and it is offline at that time, cpufreq_add_dev() will attempt to create a symbolic link to the policy object for it, but that link is present already, so a warning about that will be triggered. To avoid that warning, make cpufreq use an additional CPU mask containing related CPUs that are actually present for each policy object. That mask is initialized when the policy object is populated after its creation (for the first online CPU using it) and it includes CPUs from the "policy CPUs" mask returned by the cpufreq driver's ->init() callback that are physically present at that time. Symbolic links to the policy are created only for the CPUs in that mask. If cpufreq_add_dev() is invoked for an offline CPU, it checks the new mask and only creates the symlink if the CPU was not in it (the CPU is added to the mask at the same time). In turn, cpufreq_remove_dev() drops the given CPU from the new mask, removes its symlink to the policy object and returns, unless it is the CPU owning the policy object. In that case, the policy object is moved to a new CPU's sysfs directory or deleted if the CPU being removed was the last user of the policy. While at it, notice that cpufreq_remove_dev() can't fail, because its return value is ignored, so make it ignore return values from __cpufreq_remove_dev_prepare() and __cpufreq_remove_dev_finish() and prevent these functions from aborting on errors returned by __cpufreq_governor(). Also drop the now unused sif argument from them. Fixes: 87549141d516 (cpufreq: Stop migrating sysfs files on hotplug) Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reported-and-tested-by: Russell King <linux@arm.linux.org.uk> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2015-07-26 03:07:47 +03:00
cpumask_clear_cpu(cpu, policy->real_cpus);
cpufreq: create link to policy only for registered CPUs If a cpufreq driver is registered very early in the boot stage (e.g. registered from postcore_initcall()), then cpufreq core may generate kernel warnings for it. In this case, the CPUs are brought online, then the cpufreq driver is registered, and then the CPU topology devices are registered. However, by the time cpufreq_add_dev() gets called, the cpu device isn't stored in the per-cpu variable (cpu_sys_devices,) which is read by get_cpu_device(). So the cpufreq core fails to get device for the CPU, for which cpufreq_add_dev() was called in the first place and we will hit a WARN_ON(!cpu_dev). Even if we reuse the 'dev' parameter passed to cpufreq_add_dev() to avoid that warning, there might be other CPUs online that share the policy with the cpu for which cpufreq_add_dev() is called. Eventually get_cpu_device() will return NULL for them as well, and we will hit the same WARN_ON() again. In order to fix these issues, change cpufreq core to create links to the policy for a cpu only when cpufreq_add_dev() is called for that CPU. Reuse the 'real_cpus' mask to track that as well. Note that cpufreq_remove_dev() already handles removal of the links for individual CPUs and cpufreq_add_dev() has aligned with that now. Reported-by: Russell King <rmk+kernel@arm.linux.org.uk> Tested-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2016-09-12 09:37:05 +03:00
remove_cpu_dev_symlink(policy, dev);
cpufreq: Stop migrating sysfs files on hotplug When we hot-unplug a cpu, we remove its sysfs cpufreq directory and if the outgoing cpu was the owner of policy->kobj earlier then we migrate the sysfs directory to under another online cpu. There are few disadvantages this brings: - Code Complexity - Slower hotplug/suspend/resume - sysfs file permissions are reset after all policy->cpus are offlined - CPUFreq stats history lost after all policy->cpus are offlined - Special management of sysfs stuff during suspend/resume To overcome these, this patch modifies the way sysfs directories are managed: - Select sysfs kobjects owner while initializing policy and don't change it during hotplugs. Track it with kobj_cpu created earlier. - Create symlinks for all related CPUs (can be offline) instead of affected CPUs on policy initialization and remove them only when the policy is freed. - Free policy structure only on the removal of cpufreq-driver and not during hotplug/suspend/resume, detected by checking 'struct subsys_interface *' (Valid only when called from subsys_interface_unregister() while unregistering driver). Apart from this, special care is taken to handle physical hoplug of CPUs as we wouldn't remove sysfs links or remove policies on logical hotplugs. Physical hotplug happens in the following sequence. Hot removal: - CPU is offlined first, ~ 'echo 0 > /sys/devices/system/cpu/cpuX/online' - Then its device is removed along with all sysfs files, cpufreq core notified with cpufreq_remove_dev() callback from subsys-interface.. Hot addition: - First the device along with its sysfs files is added, cpufreq core notified with cpufreq_add_dev() callback from subsys-interface.. - CPU is onlined, ~ 'echo 1 > /sys/devices/system/cpu/cpuX/online' We call the same routines with both hotplug and subsys callbacks, and we sense physical hotplug with cpu_offline() check in subsys callback. We can handle most of the stuff with regular hotplug callback paths and add/remove cpufreq sysfs links or free policy from subsys callbacks. Original-by: Saravana Kannan <skannan@codeaurora.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-06-10 03:13:21 +03:00
if (cpumask_empty(policy->real_cpus)) {
/* We did light-weight exit earlier, do full tear down now */
if (cpufreq_driver->offline)
cpufreq_driver->exit(policy);
cpufreq_policy_free(policy);
}
}
/**
* cpufreq_out_of_sync - If actual and saved CPU frequency differs, we're
* in deep trouble.
* @policy: policy managing CPUs
* @new_freq: CPU frequency the CPU actually runs at
*
* We adjust to current frequency first, and need to clean up later.
* So either call to cpufreq_update_policy() or schedule handle_update()).
*/
static void cpufreq_out_of_sync(struct cpufreq_policy *policy,
unsigned int new_freq)
{
struct cpufreq_freqs freqs;
pr_debug("Warning: CPU frequency out of sync: cpufreq and timing core thinks of %u, is %u kHz\n",
policy->cur, new_freq);
freqs.old = policy->cur;
freqs.new = new_freq;
cpufreq_freq_transition_begin(policy, &freqs);
cpufreq_freq_transition_end(policy, &freqs, 0);
}
/**
* cpufreq_quick_get - get the CPU frequency (in kHz) from policy->cur
* @cpu: CPU number
*
* This is the last known freq, without actually getting it from the driver.
* Return value will be same as what is shown in scaling_cur_freq in sysfs.
*/
unsigned int cpufreq_quick_get(unsigned int cpu)
{
struct cpufreq_policy *policy;
unsigned int ret_freq = 0;
unsigned long flags;
read_lock_irqsave(&cpufreq_driver_lock, flags);
if (cpufreq_driver && cpufreq_driver->setpolicy && cpufreq_driver->get) {
ret_freq = cpufreq_driver->get(cpu);
read_unlock_irqrestore(&cpufreq_driver_lock, flags);
return ret_freq;
}
read_unlock_irqrestore(&cpufreq_driver_lock, flags);
policy = cpufreq_cpu_get(cpu);
if (policy) {
ret_freq = policy->cur;
cpufreq_cpu_put(policy);
}
return ret_freq;
}
EXPORT_SYMBOL(cpufreq_quick_get);
/**
* cpufreq_quick_get_max - get the max reported CPU frequency for this CPU
* @cpu: CPU number
*
* Just return the max possible frequency for a given CPU.
*/
unsigned int cpufreq_quick_get_max(unsigned int cpu)
{
struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
unsigned int ret_freq = 0;
if (policy) {
ret_freq = policy->max;
cpufreq_cpu_put(policy);
}
return ret_freq;
}
EXPORT_SYMBOL(cpufreq_quick_get_max);
static unsigned int __cpufreq_get(struct cpufreq_policy *policy)
{
unsigned int ret_freq = 0;
if (unlikely(policy_is_inactive(policy)))
return ret_freq;
ret_freq = cpufreq_driver->get(policy->cpu);
/*
cpufreq: check if policy is inactive early in __cpufreq_get() cpuinfo_cur_freq gets current CPU frequency as detected by hardware while scaling_cur_freq last known CPU frequency. Some platforms may not allow checking the CPU frequency of an offline CPU or the associated resources may have been released via cpufreq_exit when the CPU gets offlined, in which case the policy would have been invalidated already. If we attempt to get current frequency from the hardware, it may result in hang or crash. For example on Juno, I see: Unable to handle kernel NULL pointer dereference at virtual address 0000000000000188 [0000000000000188] pgd=0000000000000000 Internal error: Oops: 96000004 [#1] PREEMPT SMP Modules linked in: CPU: 5 PID: 4202 Comm: cat Not tainted 4.20.0-08251-ga0f2c0318a15-dirty #87 Hardware name: ARM LTD ARM Juno Development Platform/ARM Juno Development Platform pstate: 40000005 (nZcv daif -PAN -UAO) pc : scmi_cpufreq_get_rate+0x34/0xb0 lr : scmi_cpufreq_get_rate+0x34/0xb0 Call trace: scmi_cpufreq_get_rate+0x34/0xb0 __cpufreq_get+0x34/0xc0 show_cpuinfo_cur_freq+0x24/0x78 show+0x40/0x60 sysfs_kf_seq_show+0xc0/0x148 kernfs_seq_show+0x44/0x50 seq_read+0xd4/0x480 kernfs_fop_read+0x15c/0x208 __vfs_read+0x60/0x188 vfs_read+0x94/0x150 ksys_read+0x6c/0xd8 __arm64_sys_read+0x24/0x30 el0_svc_common+0x78/0x100 el0_svc_handler+0x38/0x78 el0_svc+0x8/0xc ---[ end trace 3d1024e58f77f6b2 ]--- So fix the issue by checking if the policy is invalid early in __cpufreq_get before attempting to get the current frequency. Signed-off-by: Sudeep Holla <sudeep.holla@arm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2019-01-07 21:51:53 +03:00
* If fast frequency switching is used with the given policy, the check
* against policy->cur is pointless, so skip it in that case too.
*/
cpufreq: check if policy is inactive early in __cpufreq_get() cpuinfo_cur_freq gets current CPU frequency as detected by hardware while scaling_cur_freq last known CPU frequency. Some platforms may not allow checking the CPU frequency of an offline CPU or the associated resources may have been released via cpufreq_exit when the CPU gets offlined, in which case the policy would have been invalidated already. If we attempt to get current frequency from the hardware, it may result in hang or crash. For example on Juno, I see: Unable to handle kernel NULL pointer dereference at virtual address 0000000000000188 [0000000000000188] pgd=0000000000000000 Internal error: Oops: 96000004 [#1] PREEMPT SMP Modules linked in: CPU: 5 PID: 4202 Comm: cat Not tainted 4.20.0-08251-ga0f2c0318a15-dirty #87 Hardware name: ARM LTD ARM Juno Development Platform/ARM Juno Development Platform pstate: 40000005 (nZcv daif -PAN -UAO) pc : scmi_cpufreq_get_rate+0x34/0xb0 lr : scmi_cpufreq_get_rate+0x34/0xb0 Call trace: scmi_cpufreq_get_rate+0x34/0xb0 __cpufreq_get+0x34/0xc0 show_cpuinfo_cur_freq+0x24/0x78 show+0x40/0x60 sysfs_kf_seq_show+0xc0/0x148 kernfs_seq_show+0x44/0x50 seq_read+0xd4/0x480 kernfs_fop_read+0x15c/0x208 __vfs_read+0x60/0x188 vfs_read+0x94/0x150 ksys_read+0x6c/0xd8 __arm64_sys_read+0x24/0x30 el0_svc_common+0x78/0x100 el0_svc_handler+0x38/0x78 el0_svc+0x8/0xc ---[ end trace 3d1024e58f77f6b2 ]--- So fix the issue by checking if the policy is invalid early in __cpufreq_get before attempting to get the current frequency. Signed-off-by: Sudeep Holla <sudeep.holla@arm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2019-01-07 21:51:53 +03:00
if (policy->fast_switch_enabled)
return ret_freq;
if (ret_freq && policy->cur &&
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) {
/* verify no discrepancy between actual and
saved value exists */
if (unlikely(ret_freq != policy->cur)) {
cpufreq_out_of_sync(policy, ret_freq);
schedule_work(&policy->update);
}
}
return ret_freq;
}
/**
* cpufreq_get - get the current CPU frequency (in kHz)
* @cpu: CPU number
*
* Get the CPU current (static) CPU frequency
*/
unsigned int cpufreq_get(unsigned int cpu)
{
struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
unsigned int ret_freq = 0;
if (policy) {
down_read(&policy->rwsem);
if (cpufreq_driver->get)
ret_freq = __cpufreq_get(policy);
up_read(&policy->rwsem);
cpufreq_cpu_put(policy);
}
return ret_freq;
}
EXPORT_SYMBOL(cpufreq_get);
static unsigned int cpufreq_update_current_freq(struct cpufreq_policy *policy)
{
unsigned int new_freq;
new_freq = cpufreq_driver->get(policy->cpu);
if (!new_freq)
return 0;
if (!policy->cur) {
pr_debug("cpufreq: Driver did not initialize current freq\n");
policy->cur = new_freq;
} else if (policy->cur != new_freq && has_target()) {
cpufreq_out_of_sync(policy, new_freq);
}
return new_freq;
}
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 02:29:42 +04:00
static struct subsys_interface cpufreq_interface = {
.name = "cpufreq",
.subsys = &cpu_subsys,
.add_dev = cpufreq_add_dev,
.remove_dev = cpufreq_remove_dev,
};
/*
* In case platform wants some specific frequency to be configured
* during suspend..
*/
int cpufreq_generic_suspend(struct cpufreq_policy *policy)
{
int ret;
if (!policy->suspend_freq) {
pr_debug("%s: suspend_freq not defined\n", __func__);
return 0;
}
pr_debug("%s: Setting suspend-freq: %u\n", __func__,
policy->suspend_freq);
ret = __cpufreq_driver_target(policy, policy->suspend_freq,
CPUFREQ_RELATION_H);
if (ret)
pr_err("%s: unable to set suspend-freq: %u. err: %d\n",
__func__, policy->suspend_freq, ret);
return ret;
}
EXPORT_SYMBOL(cpufreq_generic_suspend);
/**
* cpufreq_suspend() - Suspend CPUFreq governors
*
* Called during system wide Suspend/Hibernate cycles for suspending governors
* as some platforms can't change frequency after this point in suspend cycle.
* Because some of the devices (like: i2c, regulators, etc) they use for
* changing frequency are suspended quickly after this point.
*/
void cpufreq_suspend(void)
{
struct cpufreq_policy *policy;
if (!cpufreq_driver)
return;
if (!has_target() && !cpufreq_driver->suspend)
goto suspend;
pr_debug("%s: Suspending Governors\n", __func__);
for_each_active_policy(policy) {
if (has_target()) {
down_write(&policy->rwsem);
cpufreq_stop_governor(policy);
up_write(&policy->rwsem);
}
if (cpufreq_driver->suspend && cpufreq_driver->suspend(policy))
pr_err("%s: Failed to suspend driver: %p\n", __func__,
policy);
}
suspend:
cpufreq_suspended = true;
}
/**
* cpufreq_resume() - Resume CPUFreq governors
*
* Called during system wide Suspend/Hibernate cycle for resuming governors that
* are suspended with cpufreq_suspend().
*/
void cpufreq_resume(void)
{
struct cpufreq_policy *policy;
int ret;
if (!cpufreq_driver)
return;
if (unlikely(!cpufreq_suspended))
return;
cpufreq_suspended = false;
if (!has_target() && !cpufreq_driver->resume)
return;
pr_debug("%s: Resuming Governors\n", __func__);
for_each_active_policy(policy) {
if (cpufreq_driver->resume && cpufreq_driver->resume(policy)) {
pr_err("%s: Failed to resume driver: %p\n", __func__,
policy);
} else if (has_target()) {
down_write(&policy->rwsem);
ret = cpufreq_start_governor(policy);
up_write(&policy->rwsem);
if (ret)
pr_err("%s: Failed to start governor for policy: %p\n",
__func__, policy);
}
}
}
/**
* cpufreq_get_current_driver - return current driver's name
*
* Return the name string of the currently loaded cpufreq driver
* or NULL, if none.
*/
const char *cpufreq_get_current_driver(void)
{
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
if (cpufreq_driver)
return cpufreq_driver->name;
return NULL;
}
EXPORT_SYMBOL_GPL(cpufreq_get_current_driver);
/**
* cpufreq_get_driver_data - return current driver data
*
* Return the private data of the currently loaded cpufreq
* driver, or NULL if no cpufreq driver is loaded.
*/
void *cpufreq_get_driver_data(void)
{
if (cpufreq_driver)
return cpufreq_driver->driver_data;
return NULL;
}
EXPORT_SYMBOL_GPL(cpufreq_get_driver_data);
/*********************************************************************
* NOTIFIER LISTS INTERFACE *
*********************************************************************/
/**
* cpufreq_register_notifier - register a driver with cpufreq
* @nb: notifier function to register
* @list: CPUFREQ_TRANSITION_NOTIFIER or CPUFREQ_POLICY_NOTIFIER
*
* Add a driver to one of two lists: either a list of drivers that
* are notified about clock rate changes (once before and once after
* the transition), or a list of drivers that are notified about
* changes in cpufreq policy.
*
* This function may sleep, and has the same return conditions as
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 13:16:30 +04:00
* blocking_notifier_chain_register.
*/
int cpufreq_register_notifier(struct notifier_block *nb, unsigned int list)
{
int ret;
if (cpufreq_disabled())
return -EINVAL;
switch (list) {
case CPUFREQ_TRANSITION_NOTIFIER:
mutex_lock(&cpufreq_fast_switch_lock);
if (cpufreq_fast_switch_count > 0) {
mutex_unlock(&cpufreq_fast_switch_lock);
return -EBUSY;
}
ret = srcu_notifier_chain_register(
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 13:16:30 +04:00
&cpufreq_transition_notifier_list, nb);
if (!ret)
cpufreq_fast_switch_count--;
mutex_unlock(&cpufreq_fast_switch_lock);
break;
case CPUFREQ_POLICY_NOTIFIER:
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 13:16:30 +04:00
ret = blocking_notifier_chain_register(
&cpufreq_policy_notifier_list, nb);
break;
default:
ret = -EINVAL;
}
return ret;
}
EXPORT_SYMBOL(cpufreq_register_notifier);
/**
* cpufreq_unregister_notifier - unregister a driver with cpufreq
* @nb: notifier block to be unregistered
* @list: CPUFREQ_TRANSITION_NOTIFIER or CPUFREQ_POLICY_NOTIFIER
*
* Remove a driver from the CPU frequency notifier list.
*
* This function may sleep, and has the same return conditions as
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 13:16:30 +04:00
* blocking_notifier_chain_unregister.
*/
int cpufreq_unregister_notifier(struct notifier_block *nb, unsigned int list)
{
int ret;
if (cpufreq_disabled())
return -EINVAL;
switch (list) {
case CPUFREQ_TRANSITION_NOTIFIER:
mutex_lock(&cpufreq_fast_switch_lock);
ret = srcu_notifier_chain_unregister(
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 13:16:30 +04:00
&cpufreq_transition_notifier_list, nb);
if (!ret && !WARN_ON(cpufreq_fast_switch_count >= 0))
cpufreq_fast_switch_count++;
mutex_unlock(&cpufreq_fast_switch_lock);
break;
case CPUFREQ_POLICY_NOTIFIER:
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 13:16:30 +04:00
ret = blocking_notifier_chain_unregister(
&cpufreq_policy_notifier_list, nb);
break;
default:
ret = -EINVAL;
}
return ret;
}
EXPORT_SYMBOL(cpufreq_unregister_notifier);
/*********************************************************************
* GOVERNORS *
*********************************************************************/
/**
* cpufreq_driver_fast_switch - Carry out a fast CPU frequency switch.
* @policy: cpufreq policy to switch the frequency for.
* @target_freq: New frequency to set (may be approximate).
*
* Carry out a fast frequency switch without sleeping.
*
* The driver's ->fast_switch() callback invoked by this function must be
* suitable for being called from within RCU-sched read-side critical sections
* and it is expected to select the minimum available frequency greater than or
* equal to @target_freq (CPUFREQ_RELATION_L).
*
* This function must not be called if policy->fast_switch_enabled is unset.
*
* Governors calling this function must guarantee that it will never be invoked
* twice in parallel for the same policy and that it will never be called in
* parallel with either ->target() or ->target_index() for the same policy.
*
* Returns the actual frequency set for the CPU.
*
* If 0 is returned by the driver's ->fast_switch() callback to indicate an
* error condition, the hardware configuration must be preserved.
*/
unsigned int cpufreq_driver_fast_switch(struct cpufreq_policy *policy,
unsigned int target_freq)
{
target_freq = clamp_val(target_freq, policy->min, policy->max);
return cpufreq_driver->fast_switch(policy, target_freq);
}
EXPORT_SYMBOL_GPL(cpufreq_driver_fast_switch);
cpufreq: add support for intermediate (stable) frequencies Douglas Anderson, recently pointed out an interesting problem due to which udelay() was expiring earlier than it should. While transitioning between frequencies few platforms may temporarily switch to a stable frequency, waiting for the main PLL to stabilize. For example: When we transition between very low frequencies on exynos, like between 200MHz and 300MHz, we may temporarily switch to a PLL running at 800MHz. No CPUFREQ notification is sent for that. That means there's a period of time when we're running at 800MHz but loops_per_jiffy is calibrated at between 200MHz and 300MHz. And so udelay behaves badly. To get this fixed in a generic way, introduce another set of callbacks get_intermediate() and target_intermediate(), only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset. get_intermediate() should return a stable intermediate frequency platform wants to switch to, and target_intermediate() should set CPU to that frequency, before jumping to the frequency corresponding to 'index'. Core will take care of sending notifications and driver doesn't have to handle them in target_intermediate() or target_index(). NOTE: ->target_index() should restore to policy->restore_freq in case of failures as core would send notifications for that. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Doug Anderson <dianders@chromium.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-02 21:19:28 +04:00
/* Must set freqs->new to intermediate frequency */
static int __target_intermediate(struct cpufreq_policy *policy,
struct cpufreq_freqs *freqs, int index)
{
int ret;
freqs->new = cpufreq_driver->get_intermediate(policy, index);
/* We don't need to switch to intermediate freq */
if (!freqs->new)
return 0;
pr_debug("%s: cpu: %d, switching to intermediate freq: oldfreq: %u, intermediate freq: %u\n",
__func__, policy->cpu, freqs->old, freqs->new);
cpufreq_freq_transition_begin(policy, freqs);
ret = cpufreq_driver->target_intermediate(policy, index);
cpufreq_freq_transition_end(policy, freqs, ret);
if (ret)
pr_err("%s: Failed to change to intermediate frequency: %d\n",
__func__, ret);
return ret;
}
static int __target_index(struct cpufreq_policy *policy, int index)
{
cpufreq: add support for intermediate (stable) frequencies Douglas Anderson, recently pointed out an interesting problem due to which udelay() was expiring earlier than it should. While transitioning between frequencies few platforms may temporarily switch to a stable frequency, waiting for the main PLL to stabilize. For example: When we transition between very low frequencies on exynos, like between 200MHz and 300MHz, we may temporarily switch to a PLL running at 800MHz. No CPUFREQ notification is sent for that. That means there's a period of time when we're running at 800MHz but loops_per_jiffy is calibrated at between 200MHz and 300MHz. And so udelay behaves badly. To get this fixed in a generic way, introduce another set of callbacks get_intermediate() and target_intermediate(), only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset. get_intermediate() should return a stable intermediate frequency platform wants to switch to, and target_intermediate() should set CPU to that frequency, before jumping to the frequency corresponding to 'index'. Core will take care of sending notifications and driver doesn't have to handle them in target_intermediate() or target_index(). NOTE: ->target_index() should restore to policy->restore_freq in case of failures as core would send notifications for that. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Doug Anderson <dianders@chromium.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-02 21:19:28 +04:00
struct cpufreq_freqs freqs = {.old = policy->cur, .flags = 0};
unsigned int intermediate_freq = 0;
unsigned int newfreq = policy->freq_table[index].frequency;
int retval = -EINVAL;
bool notify;
if (newfreq == policy->cur)
return 0;
notify = !(cpufreq_driver->flags & CPUFREQ_ASYNC_NOTIFICATION);
if (notify) {
cpufreq: add support for intermediate (stable) frequencies Douglas Anderson, recently pointed out an interesting problem due to which udelay() was expiring earlier than it should. While transitioning between frequencies few platforms may temporarily switch to a stable frequency, waiting for the main PLL to stabilize. For example: When we transition between very low frequencies on exynos, like between 200MHz and 300MHz, we may temporarily switch to a PLL running at 800MHz. No CPUFREQ notification is sent for that. That means there's a period of time when we're running at 800MHz but loops_per_jiffy is calibrated at between 200MHz and 300MHz. And so udelay behaves badly. To get this fixed in a generic way, introduce another set of callbacks get_intermediate() and target_intermediate(), only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset. get_intermediate() should return a stable intermediate frequency platform wants to switch to, and target_intermediate() should set CPU to that frequency, before jumping to the frequency corresponding to 'index'. Core will take care of sending notifications and driver doesn't have to handle them in target_intermediate() or target_index(). NOTE: ->target_index() should restore to policy->restore_freq in case of failures as core would send notifications for that. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Doug Anderson <dianders@chromium.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-02 21:19:28 +04:00
/* Handle switching to intermediate frequency */
if (cpufreq_driver->get_intermediate) {
retval = __target_intermediate(policy, &freqs, index);
if (retval)
return retval;
intermediate_freq = freqs.new;
/* Set old freq to intermediate */
if (intermediate_freq)
freqs.old = freqs.new;
}
freqs.new = newfreq;
pr_debug("%s: cpu: %d, oldfreq: %u, new freq: %u\n",
__func__, policy->cpu, freqs.old, freqs.new);
cpufreq_freq_transition_begin(policy, &freqs);
}
retval = cpufreq_driver->target_index(policy, index);
if (retval)
pr_err("%s: Failed to change cpu frequency: %d\n", __func__,
retval);
cpufreq: add support for intermediate (stable) frequencies Douglas Anderson, recently pointed out an interesting problem due to which udelay() was expiring earlier than it should. While transitioning between frequencies few platforms may temporarily switch to a stable frequency, waiting for the main PLL to stabilize. For example: When we transition between very low frequencies on exynos, like between 200MHz and 300MHz, we may temporarily switch to a PLL running at 800MHz. No CPUFREQ notification is sent for that. That means there's a period of time when we're running at 800MHz but loops_per_jiffy is calibrated at between 200MHz and 300MHz. And so udelay behaves badly. To get this fixed in a generic way, introduce another set of callbacks get_intermediate() and target_intermediate(), only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset. get_intermediate() should return a stable intermediate frequency platform wants to switch to, and target_intermediate() should set CPU to that frequency, before jumping to the frequency corresponding to 'index'. Core will take care of sending notifications and driver doesn't have to handle them in target_intermediate() or target_index(). NOTE: ->target_index() should restore to policy->restore_freq in case of failures as core would send notifications for that. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Doug Anderson <dianders@chromium.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-02 21:19:28 +04:00
if (notify) {
cpufreq_freq_transition_end(policy, &freqs, retval);
cpufreq: add support for intermediate (stable) frequencies Douglas Anderson, recently pointed out an interesting problem due to which udelay() was expiring earlier than it should. While transitioning between frequencies few platforms may temporarily switch to a stable frequency, waiting for the main PLL to stabilize. For example: When we transition between very low frequencies on exynos, like between 200MHz and 300MHz, we may temporarily switch to a PLL running at 800MHz. No CPUFREQ notification is sent for that. That means there's a period of time when we're running at 800MHz but loops_per_jiffy is calibrated at between 200MHz and 300MHz. And so udelay behaves badly. To get this fixed in a generic way, introduce another set of callbacks get_intermediate() and target_intermediate(), only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset. get_intermediate() should return a stable intermediate frequency platform wants to switch to, and target_intermediate() should set CPU to that frequency, before jumping to the frequency corresponding to 'index'. Core will take care of sending notifications and driver doesn't have to handle them in target_intermediate() or target_index(). NOTE: ->target_index() should restore to policy->restore_freq in case of failures as core would send notifications for that. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Doug Anderson <dianders@chromium.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-02 21:19:28 +04:00
/*
* Failed after setting to intermediate freq? Driver should have
* reverted back to initial frequency and so should we. Check
* here for intermediate_freq instead of get_intermediate, in
* case we haven't switched to intermediate freq at all.
cpufreq: add support for intermediate (stable) frequencies Douglas Anderson, recently pointed out an interesting problem due to which udelay() was expiring earlier than it should. While transitioning between frequencies few platforms may temporarily switch to a stable frequency, waiting for the main PLL to stabilize. For example: When we transition between very low frequencies on exynos, like between 200MHz and 300MHz, we may temporarily switch to a PLL running at 800MHz. No CPUFREQ notification is sent for that. That means there's a period of time when we're running at 800MHz but loops_per_jiffy is calibrated at between 200MHz and 300MHz. And so udelay behaves badly. To get this fixed in a generic way, introduce another set of callbacks get_intermediate() and target_intermediate(), only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset. get_intermediate() should return a stable intermediate frequency platform wants to switch to, and target_intermediate() should set CPU to that frequency, before jumping to the frequency corresponding to 'index'. Core will take care of sending notifications and driver doesn't have to handle them in target_intermediate() or target_index(). NOTE: ->target_index() should restore to policy->restore_freq in case of failures as core would send notifications for that. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Doug Anderson <dianders@chromium.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-02 21:19:28 +04:00
*/
if (unlikely(retval && intermediate_freq)) {
freqs.old = intermediate_freq;
freqs.new = policy->restore_freq;
cpufreq_freq_transition_begin(policy, &freqs);
cpufreq_freq_transition_end(policy, &freqs, 0);
}
}
return retval;
}
int __cpufreq_driver_target(struct cpufreq_policy *policy,
unsigned int target_freq,
unsigned int relation)
{
unsigned int old_target_freq = target_freq;
int index;
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 01:59:54 +03:00
if (cpufreq_disabled())
return -ENODEV;
/* Make sure that target_freq is within supported range */
target_freq = clamp_val(target_freq, policy->min, policy->max);
pr_debug("target for CPU %u: %u kHz, relation %u, requested %u kHz\n",
policy->cpu, target_freq, relation, old_target_freq);
cpufreq: Implement light weight ->target_index() routine Currently, the prototype of cpufreq_drivers target routines is: int target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation); And most of the drivers call cpufreq_frequency_table_target() to get a valid index of their frequency table which is closest to the target_freq. And they don't use target_freq and relation after that. So, it makes sense to just do this work in cpufreq core before calling cpufreq_frequency_table_target() and simply pass index instead. But this can be done only with drivers which expose their frequency table with cpufreq core. For others we need to stick with the old prototype of target() until those drivers are converted to expose frequency tables. This patch implements the new light weight prototype for target_index() routine. It looks like this: int target_index(struct cpufreq_policy *policy, unsigned int index); CPUFreq core will call cpufreq_frequency_table_target() before calling this routine and pass index to it. Because CPUFreq core now requires to call routines present in freq_table.c CONFIG_CPU_FREQ_TABLE must be enabled all the time. This also marks target() interface as deprecated. So, that new drivers avoid using it. And Documentation is updated accordingly. It also converts existing .target() to newly defined light weight .target_index() routine for many driver. Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no> Acked-by: Jesper Nilsson <jesper.nilsson@axis.com> Acked-by: Linus Walleij <linus.walleij@linaro.org> Acked-by: Russell King <linux@arm.linux.org.uk> Acked-by: David S. Miller <davem@davemloft.net> Tested-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rjw@rjwysocki.net>
2013-10-25 18:15:48 +04:00
/*
* This might look like a redundant call as we are checking it again
* after finding index. But it is left intentionally for cases where
* exactly same freq is called again and so we can save on few function
* calls.
*/
if (target_freq == policy->cur)
return 0;
cpufreq: add support for intermediate (stable) frequencies Douglas Anderson, recently pointed out an interesting problem due to which udelay() was expiring earlier than it should. While transitioning between frequencies few platforms may temporarily switch to a stable frequency, waiting for the main PLL to stabilize. For example: When we transition between very low frequencies on exynos, like between 200MHz and 300MHz, we may temporarily switch to a PLL running at 800MHz. No CPUFREQ notification is sent for that. That means there's a period of time when we're running at 800MHz but loops_per_jiffy is calibrated at between 200MHz and 300MHz. And so udelay behaves badly. To get this fixed in a generic way, introduce another set of callbacks get_intermediate() and target_intermediate(), only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset. get_intermediate() should return a stable intermediate frequency platform wants to switch to, and target_intermediate() should set CPU to that frequency, before jumping to the frequency corresponding to 'index'. Core will take care of sending notifications and driver doesn't have to handle them in target_intermediate() or target_index(). NOTE: ->target_index() should restore to policy->restore_freq in case of failures as core would send notifications for that. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Doug Anderson <dianders@chromium.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-02 21:19:28 +04:00
/* Save last value to restore later on errors */
policy->restore_freq = policy->cur;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
if (cpufreq_driver->target)
return cpufreq_driver->target(policy, target_freq, relation);
cpufreq: Implement light weight ->target_index() routine Currently, the prototype of cpufreq_drivers target routines is: int target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation); And most of the drivers call cpufreq_frequency_table_target() to get a valid index of their frequency table which is closest to the target_freq. And they don't use target_freq and relation after that. So, it makes sense to just do this work in cpufreq core before calling cpufreq_frequency_table_target() and simply pass index instead. But this can be done only with drivers which expose their frequency table with cpufreq core. For others we need to stick with the old prototype of target() until those drivers are converted to expose frequency tables. This patch implements the new light weight prototype for target_index() routine. It looks like this: int target_index(struct cpufreq_policy *policy, unsigned int index); CPUFreq core will call cpufreq_frequency_table_target() before calling this routine and pass index to it. Because CPUFreq core now requires to call routines present in freq_table.c CONFIG_CPU_FREQ_TABLE must be enabled all the time. This also marks target() interface as deprecated. So, that new drivers avoid using it. And Documentation is updated accordingly. It also converts existing .target() to newly defined light weight .target_index() routine for many driver. Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no> Acked-by: Jesper Nilsson <jesper.nilsson@axis.com> Acked-by: Linus Walleij <linus.walleij@linaro.org> Acked-by: Russell King <linux@arm.linux.org.uk> Acked-by: David S. Miller <davem@davemloft.net> Tested-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rjw@rjwysocki.net>
2013-10-25 18:15:48 +04:00
if (!cpufreq_driver->target_index)
return -EINVAL;
cpufreq: Implement light weight ->target_index() routine Currently, the prototype of cpufreq_drivers target routines is: int target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation); And most of the drivers call cpufreq_frequency_table_target() to get a valid index of their frequency table which is closest to the target_freq. And they don't use target_freq and relation after that. So, it makes sense to just do this work in cpufreq core before calling cpufreq_frequency_table_target() and simply pass index instead. But this can be done only with drivers which expose their frequency table with cpufreq core. For others we need to stick with the old prototype of target() until those drivers are converted to expose frequency tables. This patch implements the new light weight prototype for target_index() routine. It looks like this: int target_index(struct cpufreq_policy *policy, unsigned int index); CPUFreq core will call cpufreq_frequency_table_target() before calling this routine and pass index to it. Because CPUFreq core now requires to call routines present in freq_table.c CONFIG_CPU_FREQ_TABLE must be enabled all the time. This also marks target() interface as deprecated. So, that new drivers avoid using it. And Documentation is updated accordingly. It also converts existing .target() to newly defined light weight .target_index() routine for many driver. Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no> Acked-by: Jesper Nilsson <jesper.nilsson@axis.com> Acked-by: Linus Walleij <linus.walleij@linaro.org> Acked-by: Russell King <linux@arm.linux.org.uk> Acked-by: David S. Miller <davem@davemloft.net> Tested-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rjw@rjwysocki.net>
2013-10-25 18:15:48 +04:00
index = cpufreq_frequency_table_target(policy, target_freq, relation);
return __target_index(policy, index);
}
EXPORT_SYMBOL_GPL(__cpufreq_driver_target);
int cpufreq_driver_target(struct cpufreq_policy *policy,
unsigned int target_freq,
unsigned int relation)
{
int ret = -EINVAL;
down_write(&policy->rwsem);
ret = __cpufreq_driver_target(policy, target_freq, relation);
up_write(&policy->rwsem);
return ret;
}
EXPORT_SYMBOL_GPL(cpufreq_driver_target);
__weak struct cpufreq_governor *cpufreq_fallback_governor(void)
{
return NULL;
}
static int cpufreq_init_governor(struct cpufreq_policy *policy)
{
int ret;
/* Don't start any governor operations if we are entering suspend */
if (cpufreq_suspended)
return 0;
cpufreq: fix a NULL pointer dereference in __cpufreq_governor() If ACPI _PPC changed notification happens before governor was initiated while kernel is booting, a NULL pointer dereference will be triggered: BUG: unable to handle kernel NULL pointer dereference at 0000000000000030 IP: [<ffffffff81470453>] __cpufreq_governor+0x23/0x1e0 PGD 0 Oops: 0000 [#1] SMP ... ... RIP: 0010:[<ffffffff81470453>] [<ffffffff81470453>] __cpufreq_governor+0x23/0x1e0 RSP: 0018:ffff881fcfbcfbb8 EFLAGS: 00010286 RAX: 0000000000000000 RBX: ffff881fd11b3980 RCX: ffff88407fc20000 RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff881fd11b3980 RBP: ffff881fcfbcfbd8 R08: 0000000000000000 R09: 000000000000000f R10: ffffffff818068d0 R11: 0000000000000043 R12: 0000000000000004 R13: 0000000000000000 R14: ffffffff8196cae0 R15: 0000000000000000 FS: 0000000000000000(0000) GS:ffff881fffc00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000030 CR3: 00000000018ae000 CR4: 00000000000407f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process kworker/0:3 (pid: 750, threadinfo ffff881fcfbce000, task ffff881fcf556400) Stack: ffff881fffc17d00 ffff881fcfbcfc18 ffff881fd11b3980 0000000000000000 ffff881fcfbcfc08 ffffffff81470d08 ffff881fd11b3980 0000000000000007 ffff881fcfbcfc18 ffff881fffc17d00 ffff881fcfbcfd28 ffffffff81472e9a Call Trace: [<ffffffff81470d08>] __cpufreq_set_policy+0x1b8/0x2e0 [<ffffffff81472e9a>] cpufreq_update_policy+0xca/0x150 [<ffffffff81472f20>] ? cpufreq_update_policy+0x150/0x150 [<ffffffff81324a96>] acpi_processor_ppc_has_changed+0x71/0x7b [<ffffffff81320bcd>] acpi_processor_notify+0x55/0x115 [<ffffffff812f9c29>] acpi_device_notify+0x19/0x1b [<ffffffff813084ca>] acpi_ev_notify_dispatch+0x41/0x5f [<ffffffff812f64a4>] acpi_os_execute_deferred+0x27/0x34 The root cause is a race conditon -- cpufreq core and acpi-cpufreq driver were initiated, but cpufreq_governor wasn't and _PPC changed notification happened, __cpufreq_governor() was called within acpi_os_execute_deferred kernel thread context. To fix this panic issue, add pointer checking code in __cpufreq_governor() before pointer policy->governor is to be dereferenced. Signed-off-by: Ethan Zhao <ethan.zhao@oracle.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-12-18 09:28:19 +03:00
/*
* Governor might not be initiated here if ACPI _PPC changed
* notification happened, so check it.
*/
if (!policy->governor)
return -EINVAL;
/* Platform doesn't want dynamic frequency switching ? */
if (policy->governor->dynamic_switching &&
cpufreq_driver->flags & CPUFREQ_NO_AUTO_DYNAMIC_SWITCHING) {
struct cpufreq_governor *gov = cpufreq_fallback_governor();
if (gov) {
pr_warn("Can't use %s governor as dynamic switching is disallowed. Fallback to %s governor\n",
policy->governor->name, gov->name);
policy->governor = gov;
} else {
return -EINVAL;
}
}
if (!try_module_get(policy->governor->owner))
return -EINVAL;
cpufreq: Fix governor start/stop race condition Cpufreq governors' stop and start operations should be carried out in sequence. Otherwise, there will be unexpected behavior, like in the example below. Suppose there are 4 CPUs and policy->cpu=CPU0, CPU1/2/3 are linked to CPU0. The normal sequence is: 1) Current governor is userspace. An application tries to set the governor to ondemand. It will call __cpufreq_set_policy() in which it will stop the userspace governor and then start the ondemand governor. 2) Current governor is userspace. The online of CPU3 runs on CPU0. It will call cpufreq_add_policy_cpu() in which it will first stop the userspace governor, and then start it again. If the sequence of the above two cases interleaves, it becomes: 1) Application stops userspace governor 2) Hotplug stops userspace governor which is a problem, because the governor shouldn't be stopped twice in a row. What happens next is: 3) Application starts ondemand governor 4) Hotplug starts a governor In step 4, the hotplug is supposed to start the userspace governor, but now the governor has been changed by the application to ondemand, so the ondemand governor is started once again, which is incorrect. The solution is to prevent policy governors from being stopped multiple times in a row. A governor should only be stopped once for one policy. After it has been stopped, no more governor stop operations should be executed. Also add a mutex to serialize governor operations. [rjw: Changelog. And you owe me a beverage of my choice.] Signed-off-by: Xiaoguang Chen <chenxg@marvell.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-06-19 11:00:07 +04:00
pr_debug("%s: for CPU %u\n", __func__, policy->cpu);
if (policy->governor->init) {
ret = policy->governor->init(policy);
if (ret) {
module_put(policy->governor->owner);
return ret;
}
}
return 0;
}
static void cpufreq_exit_governor(struct cpufreq_policy *policy)
{
if (cpufreq_suspended || !policy->governor)
return;
pr_debug("%s: for CPU %u\n", __func__, policy->cpu);
if (policy->governor->exit)
policy->governor->exit(policy);
module_put(policy->governor->owner);
}
static int cpufreq_start_governor(struct cpufreq_policy *policy)
{
int ret;
if (cpufreq_suspended)
return 0;
if (!policy->governor)
return -EINVAL;
pr_debug("%s: for CPU %u\n", __func__, policy->cpu);
if (cpufreq_driver->get && !cpufreq_driver->setpolicy)
cpufreq_update_current_freq(policy);
if (policy->governor->start) {
ret = policy->governor->start(policy);
if (ret)
return ret;
}
if (policy->governor->limits)
policy->governor->limits(policy);
return 0;
}
static void cpufreq_stop_governor(struct cpufreq_policy *policy)
{
if (cpufreq_suspended || !policy->governor)
return;
pr_debug("%s: for CPU %u\n", __func__, policy->cpu);
if (policy->governor->stop)
policy->governor->stop(policy);
}
static void cpufreq_governor_limits(struct cpufreq_policy *policy)
{
if (cpufreq_suspended || !policy->governor)
return;
pr_debug("%s: for CPU %u\n", __func__, policy->cpu);
if (policy->governor->limits)
policy->governor->limits(policy);
}
int cpufreq_register_governor(struct cpufreq_governor *governor)
{
int err;
if (!governor)
return -EINVAL;
if (cpufreq_disabled())
return -ENODEV;
mutex_lock(&cpufreq_governor_mutex);
err = -EBUSY;
if (!find_governor(governor->name)) {
err = 0;
list_add(&governor->governor_list, &cpufreq_governor_list);
}
mutex_unlock(&cpufreq_governor_mutex);
return err;
}
EXPORT_SYMBOL_GPL(cpufreq_register_governor);
void cpufreq_unregister_governor(struct cpufreq_governor *governor)
{
struct cpufreq_policy *policy;
unsigned long flags;
if (!governor)
return;
if (cpufreq_disabled())
return;
/* clear last_governor for all inactive policies */
read_lock_irqsave(&cpufreq_driver_lock, flags);
for_each_inactive_policy(policy) {
if (!strcmp(policy->last_governor, governor->name)) {
policy->governor = NULL;
strcpy(policy->last_governor, "\0");
}
}
read_unlock_irqrestore(&cpufreq_driver_lock, flags);
mutex_lock(&cpufreq_governor_mutex);
list_del(&governor->governor_list);
mutex_unlock(&cpufreq_governor_mutex);
}
EXPORT_SYMBOL_GPL(cpufreq_unregister_governor);
/*********************************************************************
* POLICY INTERFACE *
*********************************************************************/
/**
* cpufreq_get_policy - get the current cpufreq_policy
* @policy: struct cpufreq_policy into which the current cpufreq_policy
* is written
*
* Reads the current cpufreq policy.
*/
int cpufreq_get_policy(struct cpufreq_policy *policy, unsigned int cpu)
{
struct cpufreq_policy *cpu_policy;
if (!policy)
return -EINVAL;
cpu_policy = cpufreq_cpu_get(cpu);
if (!cpu_policy)
return -EINVAL;
memcpy(policy, cpu_policy, sizeof(*policy));
cpufreq_cpu_put(cpu_policy);
return 0;
}
EXPORT_SYMBOL(cpufreq_get_policy);
/**
* cpufreq_set_policy - Modify cpufreq policy parameters.
* @policy: Policy object to modify.
* @new_policy: New policy data.
*
* Pass @new_policy to the cpufreq driver's ->verify() callback, run the
* installed policy notifiers for it with the CPUFREQ_ADJUST value, pass it to
* the driver's ->verify() callback again and run the notifiers for it again
* with the CPUFREQ_NOTIFY value. Next, copy the min and max parameters
* of @new_policy to @policy and either invoke the driver's ->setpolicy()
* callback (if present) or carry out a governor update for @policy. That is,
* run the current governor's ->limits() callback (if the governor field in
* @new_policy points to the same object as the one in @policy) or replace the
* governor for @policy with the new one stored in @new_policy.
*
* The cpuinfo part of @policy is not updated by this function.
*/
int cpufreq_set_policy(struct cpufreq_policy *policy,
struct cpufreq_policy *new_policy)
{
struct cpufreq_governor *old_gov;
int ret;
pr_debug("setting new policy for CPU %u: %u - %u kHz\n",
new_policy->cpu, new_policy->min, new_policy->max);
memcpy(&new_policy->cpuinfo, &policy->cpuinfo, sizeof(policy->cpuinfo));
cpufreq: Correct a freq check in cpufreq_set_policy() This check was originally added by commit 9c9a43ed2734 ("[CPUFREQ] return error when failing to set minfreq").It attempt to return an error on obviously incorrect limits when we echo xxx >.../scaling_max,min_freq Actually we just need check if new_policy->min > new_policy->max. Because at least one of max/min is copied from cpufreq_get_policy(). For example, when we echo xxx > .../scaling_min_freq, new_policy is copied from policy in cpufreq_get_policy. new_policy->max is same with policy->max. new_policy->min is set to a new value. Let me explain it in deduction method, first statement in if (): new_policy->min > policy->max policy->max == new_policy->max ==> new_policy->min > new_policy->max second statement in if(): new_policy->max < policy->min policy->max < policy->min ==>new_policy->min > new_policy->max (induction method) So we have proved that we only need check if new_policy->min > new_policy->max. After apply this patch, we can also modify ->min and ->max at same time if new freq range is very much different from current freq range. For example, if current freq range is 480000-960000, then we want to set this range to 1120000-2240000, we would fail in the past because new_policy->min > policy->max. As long as the cpufreq range is valid, we has no reason to reject the user. So correct the check to avoid such case. Signed-off-by: Pan Xinhui <xinhuix.pan@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-07-30 13:10:40 +03:00
/*
* This check works well when we store new min/max freq attributes,
* because new_policy is a copy of policy with one field updated.
*/
if (new_policy->min > new_policy->max)
return -EINVAL;
/* verify the cpu speed can be set within this limit */
ret = cpufreq_driver->verify(new_policy);
if (ret)
return ret;
/* adjust if necessary - all reasons */
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 13:16:30 +04:00
blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
CPUFREQ_ADJUST, new_policy);
/*
* verify the cpu speed can be set within this limit, which might be
* different to the first one
*/
ret = cpufreq_driver->verify(new_policy);
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 13:16:30 +04:00
if (ret)
return ret;
/* notification of the new policy */
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 13:16:30 +04:00
blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
CPUFREQ_NOTIFY, new_policy);
policy->min = new_policy->min;
policy->max = new_policy->max;
trace_cpu_frequency_limits(policy);
policy->cached_target_freq = UINT_MAX;
pr_debug("new min and max freqs are %u - %u kHz\n",
policy->min, policy->max);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
if (cpufreq_driver->setpolicy) {
policy->policy = new_policy->policy;
pr_debug("setting range\n");
return cpufreq_driver->setpolicy(policy);
}
if (new_policy->governor == policy->governor) {
pr_debug("governor limits update\n");
cpufreq_governor_limits(policy);
return 0;
}
pr_debug("governor switch\n");
/* save old, working values */
old_gov = policy->governor;
/* end old governor */
if (old_gov) {
cpufreq_stop_governor(policy);
cpufreq_exit_governor(policy);
}
/* start new governor */
policy->governor = new_policy->governor;
ret = cpufreq_init_governor(policy);
if (!ret) {
ret = cpufreq_start_governor(policy);
if (!ret) {
pr_debug("governor change\n");
sched/topology: Make Energy Aware Scheduling depend on schedutil Energy Aware Scheduling (EAS) is designed with the assumption that frequencies of CPUs follow their utilization value. When using a CPUFreq governor other than schedutil, the chances of this assumption being true are small, if any. When schedutil is being used, EAS' predictions are at least consistent with the frequency requests. Although those requests have no guarantees to be honored by the hardware, they should at least guide DVFS in the right direction and provide some hope in regards to the EAS model being accurate. To make sure EAS is only used in a sane configuration, create a strong dependency on schedutil being used. Since having sugov compiled-in does not provide that guarantee, make CPUFreq call a scheduler function on governor changes hence letting it rebuild the scheduling domains, check the governors of the online CPUs, and enable/disable EAS accordingly. Signed-off-by: Quentin Perret <quentin.perret@arm.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Rafael J. Wysocki <rjw@rjwysocki.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: adharmap@codeaurora.org Cc: chris.redpath@arm.com Cc: currojerez@riseup.net Cc: dietmar.eggemann@arm.com Cc: edubezval@gmail.com Cc: gregkh@linuxfoundation.org Cc: javi.merino@kernel.org Cc: joel@joelfernandes.org Cc: juri.lelli@redhat.com Cc: morten.rasmussen@arm.com Cc: patrick.bellasi@arm.com Cc: pkondeti@codeaurora.org Cc: skannan@codeaurora.org Cc: smuckle@google.com Cc: srinivas.pandruvada@linux.intel.com Cc: thara.gopinath@linaro.org Cc: tkjos@google.com Cc: valentin.schneider@arm.com Cc: vincent.guittot@linaro.org Cc: viresh.kumar@linaro.org Link: https://lkml.kernel.org/r/20181203095628.11858-9-quentin.perret@arm.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2018-12-03 12:56:21 +03:00
sched_cpufreq_governor_change(policy, old_gov);
return 0;
}
cpufreq_exit_governor(policy);
}
/* new governor failed, so re-start old one */
pr_debug("starting governor %s failed\n", policy->governor->name);
if (old_gov) {
policy->governor = old_gov;
if (cpufreq_init_governor(policy))
policy->governor = NULL;
else
cpufreq_start_governor(policy);
}
return ret;
}
/**
* cpufreq_update_policy - Re-evaluate an existing cpufreq policy.
* @cpu: CPU to re-evaluate the policy for.
*
* Update the current frequency for the cpufreq policy of @cpu and use
* cpufreq_set_policy() to re-apply the min and max limits saved in the
* user_policy sub-structure of that policy, which triggers the evaluation
* of policy notifiers and the cpufreq driver's ->verify() callback for the
* policy in question, among other things.
*/
void cpufreq_update_policy(unsigned int cpu)
{
struct cpufreq_policy *policy = cpufreq_cpu_acquire(cpu);
struct cpufreq_policy new_policy;
if (!policy)
return;
/*
* BIOS might change freq behind our back
* -> ask driver for current freq and notify governors about a change
*/
if (cpufreq_driver->get && !cpufreq_driver->setpolicy &&
(cpufreq_suspended || WARN_ON(!cpufreq_update_current_freq(policy))))
goto unlock;
pr_debug("updating policy for CPU %u\n", cpu);
memcpy(&new_policy, policy, sizeof(*policy));
new_policy.min = policy->user_policy.min;
new_policy.max = policy->user_policy.max;
cpufreq_set_policy(policy, &new_policy);
unlock:
cpufreq_cpu_release(policy);
}
EXPORT_SYMBOL(cpufreq_update_policy);
/**
* cpufreq_update_limits - Update policy limits for a given CPU.
* @cpu: CPU to update the policy limits for.
*
* Invoke the driver's ->update_limits callback if present or call
* cpufreq_update_policy() for @cpu.
*/
void cpufreq_update_limits(unsigned int cpu)
{
if (cpufreq_driver->update_limits)
cpufreq_driver->update_limits(cpu);
else
cpufreq_update_policy(cpu);
}
EXPORT_SYMBOL_GPL(cpufreq_update_limits);
/*********************************************************************
* BOOST *
*********************************************************************/
static int cpufreq_boost_set_sw(int state)
{
struct cpufreq_policy *policy;
int ret = -EINVAL;
for_each_active_policy(policy) {
if (!policy->freq_table)
continue;
ret = cpufreq_frequency_table_cpuinfo(policy,
policy->freq_table);
if (ret) {
pr_err("%s: Policy frequency update failed\n",
__func__);
break;
}
down_write(&policy->rwsem);
policy->user_policy.max = policy->max;
cpufreq_governor_limits(policy);
up_write(&policy->rwsem);
}
return ret;
}
int cpufreq_boost_trigger_state(int state)
{
unsigned long flags;
int ret = 0;
if (cpufreq_driver->boost_enabled == state)
return 0;
write_lock_irqsave(&cpufreq_driver_lock, flags);
cpufreq_driver->boost_enabled = state;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
ret = cpufreq_driver->set_boost(state);
if (ret) {
write_lock_irqsave(&cpufreq_driver_lock, flags);
cpufreq_driver->boost_enabled = !state;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
pr_err("%s: Cannot %s BOOST\n",
__func__, state ? "enable" : "disable");
}
return ret;
}
static bool cpufreq_boost_supported(void)
{
return cpufreq_driver->set_boost;
}
static int create_boost_sysfs_file(void)
{
int ret;
ret = sysfs_create_file(cpufreq_global_kobject, &boost.attr);
if (ret)
pr_err("%s: cannot register global BOOST sysfs file\n",
__func__);
return ret;
}
static void remove_boost_sysfs_file(void)
{
if (cpufreq_boost_supported())
sysfs_remove_file(cpufreq_global_kobject, &boost.attr);
}
int cpufreq_enable_boost_support(void)
{
if (!cpufreq_driver)
return -EINVAL;
if (cpufreq_boost_supported())
return 0;
cpufreq_driver->set_boost = cpufreq_boost_set_sw;
/* This will get removed on driver unregister */
return create_boost_sysfs_file();
}
EXPORT_SYMBOL_GPL(cpufreq_enable_boost_support);
int cpufreq_boost_enabled(void)
{
return cpufreq_driver->boost_enabled;
}
EXPORT_SYMBOL_GPL(cpufreq_boost_enabled);
/*********************************************************************
* REGISTER / UNREGISTER CPUFREQ DRIVER *
*********************************************************************/
static enum cpuhp_state hp_online;
cpufreq: Bring CPUs up even if cpufreq_online() failed There is a report that after commit 27622b061eb4 ("cpufreq: Convert to hotplug state machine"), the normal CPU offline/online cycle fails on some platforms. According to the ftrace result, this problem was triggered on platforms using acpi-cpufreq as the default cpufreq driver, and due to the lack of some ACPI freq method (eg. _PCT), cpufreq_online() failed and returned a negative value, so the CPU hotplug state machine rolled back the CPU online process. Actually, from the user's perspective, the failure of cpufreq_online() should not prevent that CPU from being brought up, although cpufreq might not work on that CPU. BTW, during system startup cpufreq_online() is not invoked via CPU online but by the cpufreq device creation process, so the APs can be brought up even though cpufreq_online() fails in that stage. This patch ignores the return value of cpufreq_online/offline() and lets the cpufreq framework deal with the failure. cpufreq_online() itself will do a proper rollback in that case and if _PCT is missing, the ACPI cpufreq driver will print a warning if the corresponding debug options have been enabled. Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=194581 Fixes: 27622b061eb4 ("cpufreq: Convert to hotplug state machine") Reported-and-tested-by: Tomasz Maciej Nowak <tmn505@gmail.com> Signed-off-by: Chen Yu <yu.c.chen@intel.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Cc: 4.9+ <stable@vger.kernel.org> # 4.9+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2017-04-09 08:45:16 +03:00
static int cpuhp_cpufreq_online(unsigned int cpu)
{
cpufreq_online(cpu);
return 0;
}
static int cpuhp_cpufreq_offline(unsigned int cpu)
{
cpufreq_offline(cpu);
return 0;
}
/**
* cpufreq_register_driver - register a CPU Frequency driver
* @driver_data: A struct cpufreq_driver containing the values#
* submitted by the CPU Frequency driver.
*
* Registers a CPU Frequency driver to this core code. This code
* returns zero on success, -EEXIST when another driver got here first
* (and isn't unregistered in the meantime).
*
*/
int cpufreq_register_driver(struct cpufreq_driver *driver_data)
{
unsigned long flags;
int ret;
if (cpufreq_disabled())
return -ENODEV;
if (!driver_data || !driver_data->verify || !driver_data->init ||
cpufreq: Implement light weight ->target_index() routine Currently, the prototype of cpufreq_drivers target routines is: int target(struct cpufreq_policy *policy, unsigned int target_freq, unsigned int relation); And most of the drivers call cpufreq_frequency_table_target() to get a valid index of their frequency table which is closest to the target_freq. And they don't use target_freq and relation after that. So, it makes sense to just do this work in cpufreq core before calling cpufreq_frequency_table_target() and simply pass index instead. But this can be done only with drivers which expose their frequency table with cpufreq core. For others we need to stick with the old prototype of target() until those drivers are converted to expose frequency tables. This patch implements the new light weight prototype for target_index() routine. It looks like this: int target_index(struct cpufreq_policy *policy, unsigned int index); CPUFreq core will call cpufreq_frequency_table_target() before calling this routine and pass index to it. Because CPUFreq core now requires to call routines present in freq_table.c CONFIG_CPU_FREQ_TABLE must be enabled all the time. This also marks target() interface as deprecated. So, that new drivers avoid using it. And Documentation is updated accordingly. It also converts existing .target() to newly defined light weight .target_index() routine for many driver. Acked-by: Hans-Christian Egtvedt <egtvedt@samfundet.no> Acked-by: Jesper Nilsson <jesper.nilsson@axis.com> Acked-by: Linus Walleij <linus.walleij@linaro.org> Acked-by: Russell King <linux@arm.linux.org.uk> Acked-by: David S. Miller <davem@davemloft.net> Tested-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rjw@rjwysocki.net>
2013-10-25 18:15:48 +04:00
!(driver_data->setpolicy || driver_data->target_index ||
driver_data->target) ||
(driver_data->setpolicy && (driver_data->target_index ||
cpufreq: add support for intermediate (stable) frequencies Douglas Anderson, recently pointed out an interesting problem due to which udelay() was expiring earlier than it should. While transitioning between frequencies few platforms may temporarily switch to a stable frequency, waiting for the main PLL to stabilize. For example: When we transition between very low frequencies on exynos, like between 200MHz and 300MHz, we may temporarily switch to a PLL running at 800MHz. No CPUFREQ notification is sent for that. That means there's a period of time when we're running at 800MHz but loops_per_jiffy is calibrated at between 200MHz and 300MHz. And so udelay behaves badly. To get this fixed in a generic way, introduce another set of callbacks get_intermediate() and target_intermediate(), only for drivers with target_index() and CPUFREQ_ASYNC_NOTIFICATION unset. get_intermediate() should return a stable intermediate frequency platform wants to switch to, and target_intermediate() should set CPU to that frequency, before jumping to the frequency corresponding to 'index'. Core will take care of sending notifications and driver doesn't have to handle them in target_intermediate() or target_index(). NOTE: ->target_index() should restore to policy->restore_freq in case of failures as core would send notifications for that. Tested-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Reviewed-by: Doug Anderson <dianders@chromium.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2014-06-02 21:19:28 +04:00
driver_data->target)) ||
(!driver_data->get_intermediate != !driver_data->target_intermediate) ||
(!driver_data->online != !driver_data->offline))
return -EINVAL;
pr_debug("trying to register driver %s\n", driver_data->name);
/* Protect against concurrent CPU online/offline. */
cpus_read_lock();
write_lock_irqsave(&cpufreq_driver_lock, flags);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
if (cpufreq_driver) {
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
ret = -EEXIST;
goto out;
}
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
cpufreq_driver = driver_data;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
if (driver_data->setpolicy)
driver_data->flags |= CPUFREQ_CONST_LOOPS;
if (cpufreq_boost_supported()) {
ret = create_boost_sysfs_file();
if (ret)
goto err_null_driver;
}
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 02:29:42 +04:00
ret = subsys_interface_register(&cpufreq_interface);
[CPUFREQ] fix BUG on cpufreq policy init failure cpufreq_register_driver sets cpufreq_driver to a structure owned (and placed) in the caller's memory. If cpufreq policy fails in its ->init function, sysdev_driver_register returns nonzero in cpufreq_register_driver. Now, cpufreq_register_driver returns an error without setting cpufreq_driver back to NULL. Usually cpufreq policy modules are unloaded because they propagate the error to the module init function and return that. So a later access to any member of cpufreq_driver causes bugs like: BUG: unable to handle kernel paging request at ffffffffa00270a0 IP: [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 PGD 1805067 PUD 1809063 PMD 1c3f90067 PTE 0 Oops: 0000 [#1] SMP last sysfs file: /sys/devices/virtual/net/tun0/statistics/collisions CPU 0 Modules linked in: ... Pid: 5677, comm: thunderbird-bin Tainted: G W 2.6.38-rc4-mm1_64+ #1389 To be filled by O.E.M./To Be Filled By O.E.M. RIP: 0010:[<ffffffff8145eca3>] [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 RSP: 0018:ffff8801aec37d98 EFLAGS: 00010086 RAX: 0000000000000202 RBX: 0000000000000000 RCX: 0000000000000001 RDX: ffffffffa00270a0 RSI: 0000000000001000 RDI: ffffffff8199ece8 ... Call Trace: [<ffffffff8145f490>] cpufreq_quick_get+0x10/0x30 [<ffffffff8103f12b>] show_cpuinfo+0x2ab/0x300 [<ffffffff81136292>] seq_read+0xf2/0x3f0 [<ffffffff8126c5d3>] ? __strncpy_from_user+0x33/0x60 [<ffffffff8116850d>] proc_reg_read+0x6d/0xa0 [<ffffffff81116e53>] vfs_read+0xc3/0x180 [<ffffffff81116f5c>] sys_read+0x4c/0x90 [<ffffffff81030dbb>] system_call_fastpath+0x16/0x1b ... It's all cause by weird fail path handling in cpufreq_register_driver. To fix that, shuffle the code to do proper handling with gotos. Signed-off-by: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Dave Jones <davej@redhat.com>
2011-03-01 19:41:10 +03:00
if (ret)
goto err_boost_unreg;
if (!(cpufreq_driver->flags & CPUFREQ_STICKY) &&
list_empty(&cpufreq_policy_list)) {
/* if all ->init() calls failed, unregister */
ret = -ENODEV;
pr_debug("%s: No CPU initialized for driver %s\n", __func__,
driver_data->name);
goto err_if_unreg;
}
ret = cpuhp_setup_state_nocalls_cpuslocked(CPUHP_AP_ONLINE_DYN,
"cpufreq:online",
cpuhp_cpufreq_online,
cpuhp_cpufreq_offline);
if (ret < 0)
goto err_if_unreg;
hp_online = ret;
ret = 0;
pr_debug("driver %s up and running\n", driver_data->name);
goto out;
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 02:29:42 +04:00
err_if_unreg:
subsys_interface_unregister(&cpufreq_interface);
err_boost_unreg:
remove_boost_sysfs_file();
[CPUFREQ] fix BUG on cpufreq policy init failure cpufreq_register_driver sets cpufreq_driver to a structure owned (and placed) in the caller's memory. If cpufreq policy fails in its ->init function, sysdev_driver_register returns nonzero in cpufreq_register_driver. Now, cpufreq_register_driver returns an error without setting cpufreq_driver back to NULL. Usually cpufreq policy modules are unloaded because they propagate the error to the module init function and return that. So a later access to any member of cpufreq_driver causes bugs like: BUG: unable to handle kernel paging request at ffffffffa00270a0 IP: [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 PGD 1805067 PUD 1809063 PMD 1c3f90067 PTE 0 Oops: 0000 [#1] SMP last sysfs file: /sys/devices/virtual/net/tun0/statistics/collisions CPU 0 Modules linked in: ... Pid: 5677, comm: thunderbird-bin Tainted: G W 2.6.38-rc4-mm1_64+ #1389 To be filled by O.E.M./To Be Filled By O.E.M. RIP: 0010:[<ffffffff8145eca3>] [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 RSP: 0018:ffff8801aec37d98 EFLAGS: 00010086 RAX: 0000000000000202 RBX: 0000000000000000 RCX: 0000000000000001 RDX: ffffffffa00270a0 RSI: 0000000000001000 RDI: ffffffff8199ece8 ... Call Trace: [<ffffffff8145f490>] cpufreq_quick_get+0x10/0x30 [<ffffffff8103f12b>] show_cpuinfo+0x2ab/0x300 [<ffffffff81136292>] seq_read+0xf2/0x3f0 [<ffffffff8126c5d3>] ? __strncpy_from_user+0x33/0x60 [<ffffffff8116850d>] proc_reg_read+0x6d/0xa0 [<ffffffff81116e53>] vfs_read+0xc3/0x180 [<ffffffff81116f5c>] sys_read+0x4c/0x90 [<ffffffff81030dbb>] system_call_fastpath+0x16/0x1b ... It's all cause by weird fail path handling in cpufreq_register_driver. To fix that, shuffle the code to do proper handling with gotos. Signed-off-by: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Dave Jones <davej@redhat.com>
2011-03-01 19:41:10 +03:00
err_null_driver:
write_lock_irqsave(&cpufreq_driver_lock, flags);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
cpufreq_driver = NULL;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
out:
cpus_read_unlock();
return ret;
}
EXPORT_SYMBOL_GPL(cpufreq_register_driver);
/**
* cpufreq_unregister_driver - unregister the current CPUFreq driver
*
* Unregister the current CPUFreq driver. Only call this if you have
* the right to do so, i.e. if you have succeeded in initialising before!
* Returns zero if successful, and -EINVAL if the cpufreq_driver is
* currently not initialised.
*/
int cpufreq_unregister_driver(struct cpufreq_driver *driver)
{
unsigned long flags;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
if (!cpufreq_driver || (driver != cpufreq_driver))
return -EINVAL;
pr_debug("unregistering driver %s\n", driver->name);
cpufreq: Remove cpufreq_rwsem cpufreq_rwsem was introduced in commit 6eed9404ab3c4 ("cpufreq: Use rwsem for protecting critical sections) in order to replace try_module_get() on the cpu-freq driver. That try_module_get() worked well until the refcount was so heavily used that module removal became more or less impossible. Though when looking at the various (undocumented) protection mechanisms in that code, the randomly sprinkeled around cpufreq_rwsem locking sites are superfluous. The policy, which is acquired in cpufreq_cpu_get() and released in cpufreq_cpu_put() is sufficiently protected already. cpufreq_cpu_get(cpu) /* Protects against concurrent driver removal */ read_lock_irqsave(&cpufreq_driver_lock, flags); policy = per_cpu(cpufreq_cpu_data, cpu); kobject_get(&policy->kobj); read_unlock_irqrestore(&cpufreq_driver_lock, flags); The reference on the policy serializes versus module unload already: cpufreq_unregister_driver() subsys_interface_unregister() __cpufreq_remove_dev_finish() per_cpu(cpufreq_cpu_data) = NULL; cpufreq_policy_put_kobj() If there is a reference held on the policy, i.e. obtained prior to the unregister call, then cpufreq_policy_put_kobj() will wait until that reference is dropped. So once subsys_interface_unregister() returns there is no policy pointer in flight and no new reference can be obtained. So that rwsem protection is useless. The other usage of cpufreq_rwsem in show()/store() of the sysfs interface is redundant as well because sysfs already does the proper kobject_get()/put() pairs. That leaves CPU hotplug versus module removal. The current down_write() around the write_lock() in cpufreq_unregister_driver() is silly at best as it protects actually nothing. The trivial solution to this is to prevent hotplug across cpufreq_unregister_driver completely. Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2015-07-22 18:59:11 +03:00
/* Protect against concurrent cpu hotplug */
cpus_read_lock();
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 02:29:42 +04:00
subsys_interface_unregister(&cpufreq_interface);
remove_boost_sysfs_file();
cpuhp_remove_state_nocalls_cpuslocked(hp_online);
write_lock_irqsave(&cpufreq_driver_lock, flags);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 02:08:16 +04:00
cpufreq_driver = NULL;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
cpus_read_unlock();
return 0;
}
EXPORT_SYMBOL_GPL(cpufreq_unregister_driver);
/*
* Stop cpufreq at shutdown to make sure it isn't holding any locks
* or mutexes when secondary CPUs are halted.
*/
static struct syscore_ops cpufreq_syscore_ops = {
.shutdown = cpufreq_suspend,
};
struct kobject *cpufreq_global_kobject;
EXPORT_SYMBOL(cpufreq_global_kobject);
static int __init cpufreq_core_init(void)
{
if (cpufreq_disabled())
return -ENODEV;
cpufreq_global_kobject = kobject_create_and_add("cpufreq", &cpu_subsys.dev_root->kobj);
BUG_ON(!cpufreq_global_kobject);
register_syscore_ops(&cpufreq_syscore_ops);
return 0;
}
module_param(off, int, 0444);
core_initcall(cpufreq_core_init);