ATM pthread_cond_signal is unnecessarily slow, because it wakes one waiter
(which at least on UP usually means an immediate context switch to one of
the waiter threads). This waiter wakes up and after a few instructions it
attempts to acquire the cv internal lock, but that lock is still held by
the thread calling pthread_cond_signal. So it goes to sleep and eventually
the signalling thread is scheduled in, unlocks the internal lock and wakes
the waiter again.
Now, before 2003-09-21 NPTL was using FUTEX_REQUEUE in pthread_cond_signal
to avoid this performance issue, but it was removed when locks were
redesigned to the 3 state scheme (unlocked, locked uncontended, locked
contended).
Following scenario shows why simply using FUTEX_REQUEUE in
pthread_cond_signal together with using lll_mutex_unlock_force in place of
lll_mutex_unlock is not enough and probably why it has been disabled at
that time:
The number is value in cv->__data.__lock.
thr1 thr2 thr3
0 pthread_cond_wait
1 lll_mutex_lock (cv->__data.__lock)
0 lll_mutex_unlock (cv->__data.__lock)
0 lll_futex_wait (&cv->__data.__futex, futexval)
0 pthread_cond_signal
1 lll_mutex_lock (cv->__data.__lock)
1 pthread_cond_signal
2 lll_mutex_lock (cv->__data.__lock)
2 lll_futex_wait (&cv->__data.__lock, 2)
2 lll_futex_requeue (&cv->__data.__futex, 0, 1, &cv->__data.__lock)
# FUTEX_REQUEUE, not FUTEX_CMP_REQUEUE
2 lll_mutex_unlock_force (cv->__data.__lock)
0 cv->__data.__lock = 0
0 lll_futex_wake (&cv->__data.__lock, 1)
1 lll_mutex_lock (cv->__data.__lock)
0 lll_mutex_unlock (cv->__data.__lock)
# Here, lll_mutex_unlock doesn't know there are threads waiting
# on the internal cv's lock
Now, I believe it is possible to use FUTEX_REQUEUE in pthread_cond_signal,
but it will cost us not one, but 2 extra syscalls and, what's worse, one of
these extra syscalls will be done for every single waiting loop in
pthread_cond_*wait.
We would need to use lll_mutex_unlock_force in pthread_cond_signal after
requeue and lll_mutex_cond_lock in pthread_cond_*wait after lll_futex_wait.
Another alternative is to do the unlocking pthread_cond_signal needs to do
(the lock can't be unlocked before lll_futex_wake, as that is racy) in the
kernel.
I have implemented both variants, futex-requeue-glibc.patch is the first
one and futex-wake_op{,-glibc}.patch is the unlocking inside of the kernel.
The kernel interface allows userland to specify how exactly an unlocking
operation should look like (some atomic arithmetic operation with optional
constant argument and comparison of the previous futex value with another
constant).
It has been implemented just for ppc*, x86_64 and i?86, for other
architectures I'm including just a stub header which can be used as a
starting point by maintainers to write support for their arches and ATM
will just return -ENOSYS for FUTEX_WAKE_OP. The requeue patch has been
(lightly) tested just on x86_64, the wake_op patch on ppc64 kernel running
32-bit and 64-bit NPTL and x86_64 kernel running 32-bit and 64-bit NPTL.
With the following benchmark on UP x86-64 I get:
for i in nptl-orig nptl-requeue nptl-wake_op; do echo time elf/ld.so --library-path .:$i /tmp/bench; \
for j in 1 2; do echo ( time elf/ld.so --library-path .:$i /tmp/bench ) 2>&1; done; done
time elf/ld.so --library-path .:nptl-orig /tmp/bench
real 0m0.655s user 0m0.253s sys 0m0.403s
real 0m0.657s user 0m0.269s sys 0m0.388s
time elf/ld.so --library-path .:nptl-requeue /tmp/bench
real 0m0.496s user 0m0.225s sys 0m0.271s
real 0m0.531s user 0m0.242s sys 0m0.288s
time elf/ld.so --library-path .:nptl-wake_op /tmp/bench
real 0m0.380s user 0m0.176s sys 0m0.204s
real 0m0.382s user 0m0.175s sys 0m0.207s
The benchmark is at:
http://sourceware.org/ml/libc-alpha/2005-03/txt00001.txt
Older futex-requeue-glibc.patch version is at:
http://sourceware.org/ml/libc-alpha/2005-03/txt00002.txt
Older futex-wake_op-glibc.patch version is at:
http://sourceware.org/ml/libc-alpha/2005-03/txt00003.txt
Will post a new version (just x86-64 fixes so that the patch
applies against pthread_cond_signal.S) to libc-hacker ml soon.
Attached is the kernel FUTEX_WAKE_OP patch as well as a simple-minded
testcase that will not test the atomicity of the operation, but at least
check if the threads that should have been woken up are woken up and
whether the arithmetic operation in the kernel gave the expected results.
Acked-by: Ingo Molnar <mingo@redhat.com>
Cc: Ulrich Drepper <drepper@redhat.com>
Cc: Jamie Lokier <jamie@shareable.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Yoichi Yuasa <yuasa@hh.iij4u.or.jp>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch adds some missing pci-related calls to the suspend and resume
routines of the 3c59x driver. It also makes the driver free/request IRQ on
suspend/resume, in accordance with the proposal at:
http://lists.osdl.org/pipermail/linux-pm/2005-May/000955.html
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This updates documentation a bit (mostly removing obsolete stuff), and
marks swsusp as no longer experimental in config.
Signed-off-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This allows a valid iommu placed immediately after memory to work, to be
recognized as after the last byte of memory and not overlapping it.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Need to create sysfs only for cpus that are present. Without which we see
NR_CPUS entries created when we have CONFIG_HOTPLUG and CONFIG_HOTPLUG_CPU
enabled.
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Need to ensure we dont get prempted when we clear ourself from mask when using
clustered mode genapic code.
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This is a multi-part message in MIME format. If the cpu lacks 3DNOW
feature, we can use a normal prefetcht0 instruction instead of NOP5.
"prefetchw (%rxx)" and "prefetcht0 (%rxx)" have the same length, ranging
from 3 to 5 bytes depending on the register. So this patch even helps
AMD64, shortening the length of the code.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Up to date I've been using the GS value to determine the processor number
in dumps from show_regs, however this can be cumbersome to do if you don't
have the vmlinux to verify with the address of cpu_pda, how about the
following? I considered using hard_smp_processor_id for robustness but we
already dereference current so we're already relying on MSR_GS_BASE being
sane.
Signed-off-by: Zwane Mwaikambo <zwane@arm.linux.org.uk>
Acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
When handling writes to /proc/irq, current code is re-programming rte
entries directly. This is not recommended and could potentially cause
chipset's to lockup, or cause missing interrupts.
CONFIG_IRQ_BALANCE does this correctly, where it re-programs only when the
interrupt is pending. The same needs to be done for /proc/irq handling as well.
Otherwise user space irq balancers are really not doing the right thing.
- Changed pending_irq_balance_cpumask to pending_irq_migrate_cpumask for
lack of a generic name.
- added move_irq out of IRQ_BALANCE, and added this same to X86_64
- Added new proc handler for write, so we can do deferred write at irq
handling time.
- Display of /proc/irq/XX/smp_affinity used to display CPU_MASKALL, instead
it now shows only active cpu masks, or exactly what was set.
- Provided a common move_irq implementation, instead of duplicating
when using generic irq framework.
Tested on i386/x86_64 and ia64 with CONFIG_PCI_MSI turned on and off.
Tested UP builds as well.
MSI testing: tbd: I have cards, need to look for a x-over cable, although I
did test an earlier version of this patch. Will test in a couple days.
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Acked-by: Zwane Mwaikambo <zwane@holomorphy.com>
Grudgingly-acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Coywolf Qi Hunt <coywolf@lovecn.org>
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Changing CONFIG_LOCALVERSION rebuilds too much, for no apparent reason.
Use system_utsname for progress and debug header.
Signed-off-by: Olaf Hering <olh@suse.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Currently, we set the class bit in kernel SLB entries, and clear it on
user SLB entries. On POWER5, ERAT entries created in real mode have
the class bit clear. So to avoid flushing kernel ERAT entries on each
context switch, this patch inverts our usage of the class bit, setting
it on user SLB entries and clearing it on kernel SLB entries.
Booted on POWER5 and G5.
Signed-off-by: David Gibson <dwg@au1.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Yet another architecture not coverd by GEN_RTC - sparc64 never picked
it until now and it doesn't have asm/rtc.h to go with it, so it
wouldn't compile anyway (or have these ioctls in the user-visible
headers, for that matter).
FWIW, I'm very tempted to introduce ARCH_HAS_GEN_RTC and have it set
in arch/*/Kconfig for architectures that know what to do with this
stuff - for something supposedly generic the list of architectures
where it doesn't work is getting too long...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
sunsu had been broken by ->stop_tx/->start_tx API changes.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Actually, proper fix of that breakage is embarrassingly simple - it's yet
another gratitious leftover include of asm/segment.h, so incremental to the
previos would be removal of that BROKEN and removal of bogus include from
mxser.c itself.
Signed-off-by: David S. Miller <davem@davemloft.net>
Move oprofile_impl.h into include/asm-ppc64 in preparation for moving
oprofile_model into cpu feature struct.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Change oprofile to use num_pmcs from the cpu feature struct.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Remove the CPU_FTR_PMC8 feature now we encode the number of PMCs
directly.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Add a field in the cputable struct to store the number of PMCs.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
I would like to be able to read the lparcfg data from any user so we
can make "intelligent" decisions based on underlying attributes when
running in lpars. Yes there's software that likes to do this :) and
runs as non-root.
It's very similar to say VM where you can get CP to provide feedback
of the real hardware inside a VM guest.
Signed-off-by: Wim Coekaerts <wim.coekaerts@oracle.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Removed PPC64 architecture specific users of asm/segment.h.
Signed-off-by: Kumar Gala <kumar.gala@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This patch merges several include files from
asm-ppc and asm-ppc64 into the new asm-powerpc.
Signed-off-by: Jon Loeliger <jdl@freescale.com>
Signed-off-by: Kumar Gala <kumar.gala@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Merged several nearly-identical header files from asm-ppc and asm-ppc64
into asm-powerpc.
Signed-off-by: Kumar Gala <kumar.gala@freescale.com>
Signed-off-by: Becky Bruce <becky.bruce@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
cmpxchg has the following code:
__typeof__(*(ptr)) _o_ = (o);
__typeof__(*(ptr)) _n_ = (n);
Unfortunately it makes gcc 4.0 store and load the variables to the stack.
Eg in atomic_dec_and_test we get:
stw r10,112(r1)
stw r9,116(r1)
lwz r9,112(r1)
lwz r0,116(r1)
x86 is just casting the values so do that instead. Also change __xchg*
and __cmpxchg* to take unsigned values, removing a few sign extensions.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Poison initmem after we free it so we catch use after free issues.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The following patch fixes 2 issues:
1) use PLATFORM_LPAR bit to test if running in LPAR mode
2) systemcfg pointer is assigned from static data in
arch/ppc64/kernel/pacaData.c. The file arch/ppc64/kernel/head.S
now refers to is using the GOT binding to the pointer and hence
must deref it.
Signed-off-by: Jimi Xenidis <jimix@watson.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Consolidate the early console and PPCDBG code in udbg.c
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Trim some no longer needed includes from udbg.c and friends.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Take udbg out of ppc_md. Allows us to not overwrite early udbg inits
when assigning ppc_md.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Split scc and 15550 functions from udbg each into their own file.
This makes them more symetric with the lpar and btext code.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
make udbg_init_uart set the ppc_md udbg methods.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Make the 16550 and real mode 16550 use tail recursion like the scc code
instead of repeating the routine except for the character sent.
Gcc recoginizes the tail recursion and handles it efficently without
stack allocations. The maple real putc shrinks from 188 to 104 bytes
of instructions. udbg_putc drops from 188 to 140 bytes.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Now that xmon is fixed we should not need the dummy getc routines.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
udbg_getc_poll is a ppc_md function. don't call directly into udbg.c
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
inline pmac_call_feature references ppc_md so include asm/machdep.h
in asm/pmac_feature.h
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Jakub Jelinek
Rafael J. Wysocki
Pavel Machek
Eric W. Biederman
Ashok Raj
Eric Dumazet
Zwane Mwaikambo
Eugene Surovegin
Linus Torvalds
Olaf Hering
David Gibson
Al Viro
David S. Miller
Anton Blanchard
Wim Coekaerts
Kumar Gala
Jon Loeliger
Becky Bruce
Jimi Xenidis
Milton Miller