glibc/nptl/pthread_cond_signal.c

101 lines
4.2 KiB
C
Raw Normal View History

/* Copyright (C) 2003-2019 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Contributed by Martin Schwidefsky <schwidefsky@de.ibm.com>, 2003.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<http://www.gnu.org/licenses/>. */
#include <endian.h>
#include <errno.h>
#include <sysdep.h>
New condvar implementation that provides stronger ordering guarantees. This is a new implementation for condition variables, required after http://austingroupbugs.net/view.php?id=609 to fix bug 13165. In essence, we need to be stricter in which waiters a signal or broadcast is required to wake up; this couldn't be solved using the old algorithm. ISO C++ made a similar clarification, so this also fixes a bug in current libstdc++, for example. We can't use the old algorithm anymore because futexes do not guarantee to wake in FIFO order. Thus, when we wake, we can't simply let any waiter grab a signal, but we need to ensure that one of the waiters happening before the signal is woken up. This is something the previous algorithm violated (see bug 13165). There's another issue specific to condvars: ABA issues on the underlying futexes. Unlike mutexes that have just three states, or semaphores that have no tokens or a limited number of them, the state of a condvar is the *order* of the waiters. A waiter on a semaphore can grab a token whenever one is available; a condvar waiter must only consume a signal if it is eligible to do so as determined by the relative order of the waiter and the signal. Therefore, this new algorithm maintains two groups of waiters: Those eligible to consume signals (G1), and those that have to wait until previous waiters have consumed signals (G2). Once G1 is empty, G2 becomes the new G1. 64b counters are used to avoid ABA issues. This condvar doesn't yet use a requeue optimization (ie, on a broadcast, waking just one thread and requeueing all others on the futex of the mutex supplied by the program). I don't think doing the requeue is necessarily the right approach (but I haven't done real measurements yet): * If a program expects to wake many threads at the same time and make that scalable, a condvar isn't great anyway because of how it requires waiters to operate mutually exclusive (due to the mutex usage). Thus, a thundering herd problem is a scalability problem with or without the optimization. Using something like a semaphore might be more appropriate in such a case. * The scalability problem is actually at the mutex side; the condvar could help (and it tries to with the requeue optimization), but it should be the mutex who decides how that is done, and whether it is done at all. * Forcing all but one waiter into the kernel-side wait queue of the mutex prevents/avoids the use of lock elision on the mutex. Thus, it prevents the only cure against the underlying scalability problem inherent to condvars. * If condvars use short critical sections (ie, hold the mutex just to check a binary flag or such), which they should do ideally, then forcing all those waiter to proceed serially with kernel-based hand-off (ie, futex ops in the mutex' contended state, via the futex wait queues) will be less efficient than just letting a scalable mutex implementation take care of it. Our current mutex impl doesn't employ spinning at all, but if critical sections are short, spinning can be much better. * Doing the requeue stuff requires all waiters to always drive the mutex into the contended state. This leads to each waiter having to call futex_wake after lock release, even if this wouldn't be necessary. [BZ #13165] * nptl/pthread_cond_broadcast.c (__pthread_cond_broadcast): Rewrite to use new algorithm. * nptl/pthread_cond_destroy.c (__pthread_cond_destroy): Likewise. * nptl/pthread_cond_init.c (__pthread_cond_init): Likewise. * nptl/pthread_cond_signal.c (__pthread_cond_signal): Likewise. * nptl/pthread_cond_wait.c (__pthread_cond_wait): Likewise. (__pthread_cond_timedwait): Move here from pthread_cond_timedwait.c. (__condvar_confirm_wakeup, __condvar_cancel_waiting, __condvar_cleanup_waiting, __condvar_dec_grefs, __pthread_cond_wait_common): New. (__condvar_cleanup): Remove. * npt/pthread_condattr_getclock.c (pthread_condattr_getclock): Adapt. * npt/pthread_condattr_setclock.c (pthread_condattr_setclock): Likewise. * npt/pthread_condattr_getpshared.c (pthread_condattr_getpshared): Likewise. * npt/pthread_condattr_init.c (pthread_condattr_init): Likewise. * nptl/tst-cond1.c: Add comment. * nptl/tst-cond20.c (do_test): Adapt. * nptl/tst-cond22.c (do_test): Likewise. * sysdeps/aarch64/nptl/bits/pthreadtypes.h (pthread_cond_t): Adapt structure. * sysdeps/arm/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/ia64/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/m68k/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/microblaze/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/mips/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nios2/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/s390/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/sh/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/tile/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/alpha/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/powerpc/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/x86/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nptl/internaltypes.h (COND_NWAITERS_SHIFT): Remove. (COND_CLOCK_BITS): Adapt. * sysdeps/nptl/pthread.h (PTHREAD_COND_INITIALIZER): Adapt. * nptl/pthreadP.h (__PTHREAD_COND_CLOCK_MONOTONIC_MASK, __PTHREAD_COND_SHARED_MASK): New. * nptl/nptl-printers.py (CLOCK_IDS): Remove. (ConditionVariablePrinter, ConditionVariableAttributesPrinter): Adapt. * nptl/nptl_lock_constants.pysym: Adapt. * nptl/test-cond-printers.py: Adapt. * sysdeps/unix/sysv/linux/hppa/internaltypes.h (cond_compat_clear, cond_compat_check_and_clear): Adapt. * sysdeps/unix/sysv/linux/hppa/pthread_cond_timedwait.c: Remove file ... * sysdeps/unix/sysv/linux/hppa/pthread_cond_wait.c (__pthread_cond_timedwait): ... and move here. * nptl/DESIGN-condvar.txt: Remove file. * nptl/lowlevelcond.sym: Likewise. * nptl/pthread_cond_timedwait.c: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S: Likewise.
2016-05-25 23:43:36 +02:00
#include <futex-internal.h>
#include <pthread.h>
#include <pthreadP.h>
New condvar implementation that provides stronger ordering guarantees. This is a new implementation for condition variables, required after http://austingroupbugs.net/view.php?id=609 to fix bug 13165. In essence, we need to be stricter in which waiters a signal or broadcast is required to wake up; this couldn't be solved using the old algorithm. ISO C++ made a similar clarification, so this also fixes a bug in current libstdc++, for example. We can't use the old algorithm anymore because futexes do not guarantee to wake in FIFO order. Thus, when we wake, we can't simply let any waiter grab a signal, but we need to ensure that one of the waiters happening before the signal is woken up. This is something the previous algorithm violated (see bug 13165). There's another issue specific to condvars: ABA issues on the underlying futexes. Unlike mutexes that have just three states, or semaphores that have no tokens or a limited number of them, the state of a condvar is the *order* of the waiters. A waiter on a semaphore can grab a token whenever one is available; a condvar waiter must only consume a signal if it is eligible to do so as determined by the relative order of the waiter and the signal. Therefore, this new algorithm maintains two groups of waiters: Those eligible to consume signals (G1), and those that have to wait until previous waiters have consumed signals (G2). Once G1 is empty, G2 becomes the new G1. 64b counters are used to avoid ABA issues. This condvar doesn't yet use a requeue optimization (ie, on a broadcast, waking just one thread and requeueing all others on the futex of the mutex supplied by the program). I don't think doing the requeue is necessarily the right approach (but I haven't done real measurements yet): * If a program expects to wake many threads at the same time and make that scalable, a condvar isn't great anyway because of how it requires waiters to operate mutually exclusive (due to the mutex usage). Thus, a thundering herd problem is a scalability problem with or without the optimization. Using something like a semaphore might be more appropriate in such a case. * The scalability problem is actually at the mutex side; the condvar could help (and it tries to with the requeue optimization), but it should be the mutex who decides how that is done, and whether it is done at all. * Forcing all but one waiter into the kernel-side wait queue of the mutex prevents/avoids the use of lock elision on the mutex. Thus, it prevents the only cure against the underlying scalability problem inherent to condvars. * If condvars use short critical sections (ie, hold the mutex just to check a binary flag or such), which they should do ideally, then forcing all those waiter to proceed serially with kernel-based hand-off (ie, futex ops in the mutex' contended state, via the futex wait queues) will be less efficient than just letting a scalable mutex implementation take care of it. Our current mutex impl doesn't employ spinning at all, but if critical sections are short, spinning can be much better. * Doing the requeue stuff requires all waiters to always drive the mutex into the contended state. This leads to each waiter having to call futex_wake after lock release, even if this wouldn't be necessary. [BZ #13165] * nptl/pthread_cond_broadcast.c (__pthread_cond_broadcast): Rewrite to use new algorithm. * nptl/pthread_cond_destroy.c (__pthread_cond_destroy): Likewise. * nptl/pthread_cond_init.c (__pthread_cond_init): Likewise. * nptl/pthread_cond_signal.c (__pthread_cond_signal): Likewise. * nptl/pthread_cond_wait.c (__pthread_cond_wait): Likewise. (__pthread_cond_timedwait): Move here from pthread_cond_timedwait.c. (__condvar_confirm_wakeup, __condvar_cancel_waiting, __condvar_cleanup_waiting, __condvar_dec_grefs, __pthread_cond_wait_common): New. (__condvar_cleanup): Remove. * npt/pthread_condattr_getclock.c (pthread_condattr_getclock): Adapt. * npt/pthread_condattr_setclock.c (pthread_condattr_setclock): Likewise. * npt/pthread_condattr_getpshared.c (pthread_condattr_getpshared): Likewise. * npt/pthread_condattr_init.c (pthread_condattr_init): Likewise. * nptl/tst-cond1.c: Add comment. * nptl/tst-cond20.c (do_test): Adapt. * nptl/tst-cond22.c (do_test): Likewise. * sysdeps/aarch64/nptl/bits/pthreadtypes.h (pthread_cond_t): Adapt structure. * sysdeps/arm/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/ia64/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/m68k/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/microblaze/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/mips/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nios2/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/s390/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/sh/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/tile/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/alpha/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/powerpc/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/x86/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nptl/internaltypes.h (COND_NWAITERS_SHIFT): Remove. (COND_CLOCK_BITS): Adapt. * sysdeps/nptl/pthread.h (PTHREAD_COND_INITIALIZER): Adapt. * nptl/pthreadP.h (__PTHREAD_COND_CLOCK_MONOTONIC_MASK, __PTHREAD_COND_SHARED_MASK): New. * nptl/nptl-printers.py (CLOCK_IDS): Remove. (ConditionVariablePrinter, ConditionVariableAttributesPrinter): Adapt. * nptl/nptl_lock_constants.pysym: Adapt. * nptl/test-cond-printers.py: Adapt. * sysdeps/unix/sysv/linux/hppa/internaltypes.h (cond_compat_clear, cond_compat_check_and_clear): Adapt. * sysdeps/unix/sysv/linux/hppa/pthread_cond_timedwait.c: Remove file ... * sysdeps/unix/sysv/linux/hppa/pthread_cond_wait.c (__pthread_cond_timedwait): ... and move here. * nptl/DESIGN-condvar.txt: Remove file. * nptl/lowlevelcond.sym: Likewise. * nptl/pthread_cond_timedwait.c: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S: Likewise.
2016-05-25 23:43:36 +02:00
#include <atomic.h>
#include <stdint.h>
#include <shlib-compat.h>
#include <stap-probe.h>
New condvar implementation that provides stronger ordering guarantees. This is a new implementation for condition variables, required after http://austingroupbugs.net/view.php?id=609 to fix bug 13165. In essence, we need to be stricter in which waiters a signal or broadcast is required to wake up; this couldn't be solved using the old algorithm. ISO C++ made a similar clarification, so this also fixes a bug in current libstdc++, for example. We can't use the old algorithm anymore because futexes do not guarantee to wake in FIFO order. Thus, when we wake, we can't simply let any waiter grab a signal, but we need to ensure that one of the waiters happening before the signal is woken up. This is something the previous algorithm violated (see bug 13165). There's another issue specific to condvars: ABA issues on the underlying futexes. Unlike mutexes that have just three states, or semaphores that have no tokens or a limited number of them, the state of a condvar is the *order* of the waiters. A waiter on a semaphore can grab a token whenever one is available; a condvar waiter must only consume a signal if it is eligible to do so as determined by the relative order of the waiter and the signal. Therefore, this new algorithm maintains two groups of waiters: Those eligible to consume signals (G1), and those that have to wait until previous waiters have consumed signals (G2). Once G1 is empty, G2 becomes the new G1. 64b counters are used to avoid ABA issues. This condvar doesn't yet use a requeue optimization (ie, on a broadcast, waking just one thread and requeueing all others on the futex of the mutex supplied by the program). I don't think doing the requeue is necessarily the right approach (but I haven't done real measurements yet): * If a program expects to wake many threads at the same time and make that scalable, a condvar isn't great anyway because of how it requires waiters to operate mutually exclusive (due to the mutex usage). Thus, a thundering herd problem is a scalability problem with or without the optimization. Using something like a semaphore might be more appropriate in such a case. * The scalability problem is actually at the mutex side; the condvar could help (and it tries to with the requeue optimization), but it should be the mutex who decides how that is done, and whether it is done at all. * Forcing all but one waiter into the kernel-side wait queue of the mutex prevents/avoids the use of lock elision on the mutex. Thus, it prevents the only cure against the underlying scalability problem inherent to condvars. * If condvars use short critical sections (ie, hold the mutex just to check a binary flag or such), which they should do ideally, then forcing all those waiter to proceed serially with kernel-based hand-off (ie, futex ops in the mutex' contended state, via the futex wait queues) will be less efficient than just letting a scalable mutex implementation take care of it. Our current mutex impl doesn't employ spinning at all, but if critical sections are short, spinning can be much better. * Doing the requeue stuff requires all waiters to always drive the mutex into the contended state. This leads to each waiter having to call futex_wake after lock release, even if this wouldn't be necessary. [BZ #13165] * nptl/pthread_cond_broadcast.c (__pthread_cond_broadcast): Rewrite to use new algorithm. * nptl/pthread_cond_destroy.c (__pthread_cond_destroy): Likewise. * nptl/pthread_cond_init.c (__pthread_cond_init): Likewise. * nptl/pthread_cond_signal.c (__pthread_cond_signal): Likewise. * nptl/pthread_cond_wait.c (__pthread_cond_wait): Likewise. (__pthread_cond_timedwait): Move here from pthread_cond_timedwait.c. (__condvar_confirm_wakeup, __condvar_cancel_waiting, __condvar_cleanup_waiting, __condvar_dec_grefs, __pthread_cond_wait_common): New. (__condvar_cleanup): Remove. * npt/pthread_condattr_getclock.c (pthread_condattr_getclock): Adapt. * npt/pthread_condattr_setclock.c (pthread_condattr_setclock): Likewise. * npt/pthread_condattr_getpshared.c (pthread_condattr_getpshared): Likewise. * npt/pthread_condattr_init.c (pthread_condattr_init): Likewise. * nptl/tst-cond1.c: Add comment. * nptl/tst-cond20.c (do_test): Adapt. * nptl/tst-cond22.c (do_test): Likewise. * sysdeps/aarch64/nptl/bits/pthreadtypes.h (pthread_cond_t): Adapt structure. * sysdeps/arm/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/ia64/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/m68k/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/microblaze/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/mips/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nios2/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/s390/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/sh/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/tile/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/alpha/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/powerpc/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/x86/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nptl/internaltypes.h (COND_NWAITERS_SHIFT): Remove. (COND_CLOCK_BITS): Adapt. * sysdeps/nptl/pthread.h (PTHREAD_COND_INITIALIZER): Adapt. * nptl/pthreadP.h (__PTHREAD_COND_CLOCK_MONOTONIC_MASK, __PTHREAD_COND_SHARED_MASK): New. * nptl/nptl-printers.py (CLOCK_IDS): Remove. (ConditionVariablePrinter, ConditionVariableAttributesPrinter): Adapt. * nptl/nptl_lock_constants.pysym: Adapt. * nptl/test-cond-printers.py: Adapt. * sysdeps/unix/sysv/linux/hppa/internaltypes.h (cond_compat_clear, cond_compat_check_and_clear): Adapt. * sysdeps/unix/sysv/linux/hppa/pthread_cond_timedwait.c: Remove file ... * sysdeps/unix/sysv/linux/hppa/pthread_cond_wait.c (__pthread_cond_timedwait): ... and move here. * nptl/DESIGN-condvar.txt: Remove file. * nptl/lowlevelcond.sym: Likewise. * nptl/pthread_cond_timedwait.c: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S: Likewise.
2016-05-25 23:43:36 +02:00
#include "pthread_cond_common.c"
New condvar implementation that provides stronger ordering guarantees. This is a new implementation for condition variables, required after http://austingroupbugs.net/view.php?id=609 to fix bug 13165. In essence, we need to be stricter in which waiters a signal or broadcast is required to wake up; this couldn't be solved using the old algorithm. ISO C++ made a similar clarification, so this also fixes a bug in current libstdc++, for example. We can't use the old algorithm anymore because futexes do not guarantee to wake in FIFO order. Thus, when we wake, we can't simply let any waiter grab a signal, but we need to ensure that one of the waiters happening before the signal is woken up. This is something the previous algorithm violated (see bug 13165). There's another issue specific to condvars: ABA issues on the underlying futexes. Unlike mutexes that have just three states, or semaphores that have no tokens or a limited number of them, the state of a condvar is the *order* of the waiters. A waiter on a semaphore can grab a token whenever one is available; a condvar waiter must only consume a signal if it is eligible to do so as determined by the relative order of the waiter and the signal. Therefore, this new algorithm maintains two groups of waiters: Those eligible to consume signals (G1), and those that have to wait until previous waiters have consumed signals (G2). Once G1 is empty, G2 becomes the new G1. 64b counters are used to avoid ABA issues. This condvar doesn't yet use a requeue optimization (ie, on a broadcast, waking just one thread and requeueing all others on the futex of the mutex supplied by the program). I don't think doing the requeue is necessarily the right approach (but I haven't done real measurements yet): * If a program expects to wake many threads at the same time and make that scalable, a condvar isn't great anyway because of how it requires waiters to operate mutually exclusive (due to the mutex usage). Thus, a thundering herd problem is a scalability problem with or without the optimization. Using something like a semaphore might be more appropriate in such a case. * The scalability problem is actually at the mutex side; the condvar could help (and it tries to with the requeue optimization), but it should be the mutex who decides how that is done, and whether it is done at all. * Forcing all but one waiter into the kernel-side wait queue of the mutex prevents/avoids the use of lock elision on the mutex. Thus, it prevents the only cure against the underlying scalability problem inherent to condvars. * If condvars use short critical sections (ie, hold the mutex just to check a binary flag or such), which they should do ideally, then forcing all those waiter to proceed serially with kernel-based hand-off (ie, futex ops in the mutex' contended state, via the futex wait queues) will be less efficient than just letting a scalable mutex implementation take care of it. Our current mutex impl doesn't employ spinning at all, but if critical sections are short, spinning can be much better. * Doing the requeue stuff requires all waiters to always drive the mutex into the contended state. This leads to each waiter having to call futex_wake after lock release, even if this wouldn't be necessary. [BZ #13165] * nptl/pthread_cond_broadcast.c (__pthread_cond_broadcast): Rewrite to use new algorithm. * nptl/pthread_cond_destroy.c (__pthread_cond_destroy): Likewise. * nptl/pthread_cond_init.c (__pthread_cond_init): Likewise. * nptl/pthread_cond_signal.c (__pthread_cond_signal): Likewise. * nptl/pthread_cond_wait.c (__pthread_cond_wait): Likewise. (__pthread_cond_timedwait): Move here from pthread_cond_timedwait.c. (__condvar_confirm_wakeup, __condvar_cancel_waiting, __condvar_cleanup_waiting, __condvar_dec_grefs, __pthread_cond_wait_common): New. (__condvar_cleanup): Remove. * npt/pthread_condattr_getclock.c (pthread_condattr_getclock): Adapt. * npt/pthread_condattr_setclock.c (pthread_condattr_setclock): Likewise. * npt/pthread_condattr_getpshared.c (pthread_condattr_getpshared): Likewise. * npt/pthread_condattr_init.c (pthread_condattr_init): Likewise. * nptl/tst-cond1.c: Add comment. * nptl/tst-cond20.c (do_test): Adapt. * nptl/tst-cond22.c (do_test): Likewise. * sysdeps/aarch64/nptl/bits/pthreadtypes.h (pthread_cond_t): Adapt structure. * sysdeps/arm/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/ia64/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/m68k/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/microblaze/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/mips/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nios2/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/s390/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/sh/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/tile/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/alpha/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/powerpc/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/x86/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nptl/internaltypes.h (COND_NWAITERS_SHIFT): Remove. (COND_CLOCK_BITS): Adapt. * sysdeps/nptl/pthread.h (PTHREAD_COND_INITIALIZER): Adapt. * nptl/pthreadP.h (__PTHREAD_COND_CLOCK_MONOTONIC_MASK, __PTHREAD_COND_SHARED_MASK): New. * nptl/nptl-printers.py (CLOCK_IDS): Remove. (ConditionVariablePrinter, ConditionVariableAttributesPrinter): Adapt. * nptl/nptl_lock_constants.pysym: Adapt. * nptl/test-cond-printers.py: Adapt. * sysdeps/unix/sysv/linux/hppa/internaltypes.h (cond_compat_clear, cond_compat_check_and_clear): Adapt. * sysdeps/unix/sysv/linux/hppa/pthread_cond_timedwait.c: Remove file ... * sysdeps/unix/sysv/linux/hppa/pthread_cond_wait.c (__pthread_cond_timedwait): ... and move here. * nptl/DESIGN-condvar.txt: Remove file. * nptl/lowlevelcond.sym: Likewise. * nptl/pthread_cond_timedwait.c: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S: Likewise.
2016-05-25 23:43:36 +02:00
/* See __pthread_cond_wait for a high-level description of the algorithm. */
int
Convert 703 function definitions to prototype style. This automatically-generated patch converts 703 function definitions in glibc from old-style K&R to prototype-style. This conversion is deliberately simplistic, excluding any tricky cases as even a patch covering only simple cases is still very large. Currently excluded are: sysdeps files (to improve test coverage for the initial patch); files containing assertions (to avoid line number changes so that generated libraries can be compared); any cases where the generated function declaration would involve lines over 79 characters and so need to be wrapped; any cases with array parameters or other cases where parameter declarators don't end with the parameter name; any other cases that my script didn't parse. I didn't try to make the ChangeLog generation indicate when function definitions are conditional; it just lists the functions changed without regard to that. Tested for x86_64 and x86 (testsuite, and that installed stripped shared libraries are unchanged by the patch). * crypt/cert.c (good_bye): Convert to prototype-style function definition. (get8): Likewise. (put8): Likewise. * crypt/crypt-entry.c (crypt): Likewise. (__fcrypt): Likewise. * crypt/crypt_util.c (_ufc_prbits): Likewise. (_ufc_set_bits): Likewise. (_ufc_clearmem): Likewise. (__init_des_r): Likewise. (shuffle_sb): Likewise. (shuffle_sb): Likewise. (_ufc_setup_salt_r): Likewise. (_ufc_mk_keytab_r): Likewise. (_ufc_dofinalperm_r): Likewise. (encrypt): Likewise. (__setkey_r): Likewise. (setkey): Likewise. * crypt/md5.c (md5_init_ctx): Likewise. (md5_read_ctx): Likewise. (md5_finish_ctx): Likewise. (md5_stream): Likewise. (md5_buffer): Likewise. (md5_process_bytes): Likewise. * crypt/sha256.c (__sha256_init_ctx): Likewise. (__sha256_finish_ctx): Likewise. (__sha256_process_bytes): Likewise. * crypt/sha512.c (__sha512_init_ctx): Likewise. (__sha512_finish_ctx): Likewise. (__sha512_process_bytes): Likewise. * ctype/isctype.c (__isctype): Likewise. * debug/backtrace.c (__backtrace): Likewise. * debug/backtracesymsfd.c (__backtrace_symbols_fd): Likewise. * debug/fgets_chk.c (__fgets_chk): Likewise. * debug/fgets_u_chk.c (__fgets_unlocked_chk): Likewise. * debug/memcpy_chk.c (__memcpy_chk): Likewise. * debug/memmove_chk.c (MEMMOVE_CHK): Likewise. * debug/mempcpy_chk.c (__mempcpy_chk): Likewise. * debug/memset_chk.c (__memset_chk): Likewise. * debug/strcat_chk.c (__strcat_chk): Likewise. * debug/strncat_chk.c (__strncat_chk): Likewise. * debug/strncpy_chk.c (__strncpy_chk): Likewise. * debug/vsprintf_chk.c (_IO_str_chk_overflow): Likewise. * dirent/dirfd.c (dirfd): Likewise. * dirent/getdents.c (__getdirentries): Likewise. * dirent/getdents64.c (getdirentries64): Likewise. * dirent/rewinddir.c (__rewinddir): Likewise. * dirent/seekdir.c (seekdir): Likewise. * dirent/telldir.c (telldir): Likewise. * elf/sln.c (makesymlinks): Likewise. (makesymlink): Likewise. * gmon/gmon.c (__moncontrol): Likewise. (__monstartup): Likewise. (write_hist): Likewise. (write_call_graph): Likewise. (write_bb_counts): Likewise. * grp/setgroups.c (setgroups): Likewise. * inet/inet_lnaof.c (inet_lnaof): Likewise. * inet/inet_net.c (inet_network): Likewise. * inet/inet_netof.c (inet_netof): Likewise. * inet/rcmd.c (rresvport_af): Likewise. (rresvport): Likewise. * io/access.c (__access): Likewise. * io/chdir.c (__chdir): Likewise. * io/chmod.c (__chmod): Likewise. * io/chown.c (__chown): Likewise. * io/close.c (__close): Likewise. * io/creat.c (creat): Likewise. * io/creat64.c (creat64): Likewise. * io/dup.c (__dup): Likewise. * io/dup2.c (__dup2): Likewise. * io/dup3.c (__dup3): Likewise. * io/euidaccess.c (__euidaccess): Likewise. * io/faccessat.c (faccessat): Likewise. * io/fchmod.c (__fchmod): Likewise. * io/fchmodat.c (fchmodat): Likewise. * io/fchown.c (__fchown): Likewise. * io/fchownat.c (fchownat): Likewise. * io/fcntl.c (__fcntl): Likewise. * io/flock.c (__flock): Likewise. * io/fts.c (fts_load): Likewise. (fts_close): Likewise. (fts_read): Likewise. (fts_set): Likewise. (fts_children): Likewise. (fts_build): Likewise. (fts_stat): Likewise. (fts_sort): Likewise. (fts_alloc): Likewise. (fts_lfree): Likewise. (fts_palloc): Likewise. (fts_padjust): Likewise. (fts_maxarglen): Likewise. (fts_safe_changedir): Likewise. * io/getwd.c (getwd): Likewise. * io/isatty.c (__isatty): Likewise. * io/lchown.c (__lchown): Likewise. * io/link.c (__link): Likewise. * io/linkat.c (linkat): Likewise. * io/lseek.c (__libc_lseek): Likewise. * io/mkdir.c (__mkdir): Likewise. * io/mkdirat.c (mkdirat): Likewise. * io/mkfifo.c (mkfifo): Likewise. * io/mkfifoat.c (mkfifoat): Likewise. * io/open.c (__libc_open): Likewise. * io/open64.c (__libc_open64): Likewise. * io/readlink.c (__readlink): Likewise. * io/readlinkat.c (readlinkat): Likewise. * io/rmdir.c (__rmdir): Likewise. * io/symlink.c (__symlink): Likewise. * io/symlinkat.c (symlinkat): Likewise. * io/ttyname.c (ttyname): Likewise. * io/ttyname_r.c (__ttyname_r): Likewise. * io/umask.c (__umask): Likewise. * io/unlink.c (__unlink): Likewise. * io/unlinkat.c (unlinkat): Likewise. * io/utime.c (utime): Likewise. * libio/clearerr.c (clearerr): Likewise. * libio/clearerr_u.c (clearerr_unlocked): Likewise. * libio/feof.c (_IO_feof): Likewise. * libio/feof_u.c (feof_unlocked): Likewise. * libio/ferror.c (_IO_ferror): Likewise. * libio/ferror_u.c (ferror_unlocked): Likewise. * libio/filedoalloc.c (_IO_file_doallocate): Likewise. * libio/fileno.c (__fileno): Likewise. * libio/fputc.c (fputc): Likewise. * libio/fputc_u.c (fputc_unlocked): Likewise. * libio/fputwc.c (fputwc): Likewise. * libio/fputwc_u.c (fputwc_unlocked): Likewise. * libio/freopen.c (freopen): Likewise. * libio/freopen64.c (freopen64): Likewise. * libio/fseek.c (fseek): Likewise. * libio/fseeko.c (fseeko): Likewise. * libio/fseeko64.c (fseeko64): Likewise. * libio/ftello.c (__ftello): Likewise. * libio/ftello64.c (ftello64): Likewise. * libio/fwide.c (fwide): Likewise. * libio/genops.c (_IO_un_link): Likewise. (_IO_link_in): Likewise. (_IO_least_marker): Likewise. (_IO_switch_to_main_get_area): Likewise. (_IO_switch_to_backup_area): Likewise. (_IO_switch_to_get_mode): Likewise. (_IO_free_backup_area): Likewise. (_IO_switch_to_put_mode): Likewise. (__overflow): Likewise. (__underflow): Likewise. (__uflow): Likewise. (_IO_setb): Likewise. (_IO_doallocbuf): Likewise. (_IO_default_underflow): Likewise. (_IO_default_uflow): Likewise. (_IO_default_xsputn): Likewise. (_IO_sgetn): Likewise. (_IO_default_xsgetn): Likewise. (_IO_sync): Likewise. (_IO_default_setbuf): Likewise. (_IO_default_seekpos): Likewise. (_IO_default_doallocate): Likewise. (_IO_init): Likewise. (_IO_old_init): Likewise. (_IO_default_sync): Likewise. (_IO_default_finish): Likewise. (_IO_default_seekoff): Likewise. (_IO_sputbackc): Likewise. (_IO_sungetc): Likewise. (_IO_set_column): Likewise. (_IO_set_column): Likewise. (_IO_adjust_column): Likewise. (_IO_get_column): Likewise. (_IO_init_marker): Likewise. (_IO_remove_marker): Likewise. (_IO_marker_difference): Likewise. (_IO_marker_delta): Likewise. (_IO_seekmark): Likewise. (_IO_unsave_markers): Likewise. (_IO_nobackup_pbackfail): Likewise. (_IO_default_pbackfail): Likewise. (_IO_default_seek): Likewise. (_IO_default_stat): Likewise. (_IO_default_read): Likewise. (_IO_default_write): Likewise. (_IO_default_showmanyc): Likewise. (_IO_default_imbue): Likewise. (_IO_iter_next): Likewise. (_IO_iter_file): Likewise. * libio/getc.c (_IO_getc): Likewise. * libio/getwc.c (_IO_getwc): Likewise. * libio/iofclose.c (_IO_new_fclose): Likewise. * libio/iofdopen.c (_IO_new_fdopen): Likewise. * libio/iofflush.c (_IO_fflush): Likewise. * libio/iofflush_u.c (__fflush_unlocked): Likewise. * libio/iofgetpos.c (_IO_new_fgetpos): Likewise. * libio/iofgetpos64.c (_IO_new_fgetpos64): Likewise. * libio/iofgets.c (_IO_fgets): Likewise. * libio/iofgets_u.c (__fgets_unlocked): Likewise. * libio/iofgetws.c (fgetws): Likewise. * libio/iofgetws_u.c (fgetws_unlocked): Likewise. * libio/iofopen64.c (_IO_fopen64): Likewise. * libio/iofopncook.c (_IO_cookie_read): Likewise. (_IO_cookie_write): Likewise. (_IO_cookie_seek): Likewise. (_IO_cookie_close): Likewise. (_IO_cookie_seekoff): Likewise. (_IO_old_cookie_seek): Likewise. * libio/iofputs.c (_IO_fputs): Likewise. * libio/iofputs_u.c (__fputs_unlocked): Likewise. * libio/iofputws.c (fputws): Likewise. * libio/iofputws_u.c (fputws_unlocked): Likewise. * libio/iofread.c (_IO_fread): Likewise. * libio/iofread_u.c (__fread_unlocked): Likewise. * libio/iofsetpos.c (_IO_new_fsetpos): Likewise. * libio/iofsetpos64.c (_IO_new_fsetpos64): Likewise. * libio/ioftell.c (_IO_ftell): Likewise. * libio/iofwrite.c (_IO_fwrite): Likewise. * libio/iogetdelim.c (_IO_getdelim): Likewise. * libio/iogets.c (_IO_gets): Likewise. * libio/iopadn.c (_IO_padn): Likewise. * libio/iopopen.c (_IO_new_proc_open): Likewise. (_IO_new_popen): Likewise. (_IO_new_proc_close): Likewise. * libio/ioputs.c (_IO_puts): Likewise. * libio/ioseekoff.c (_IO_seekoff_unlocked): Likewise. (_IO_seekoff): Likewise. * libio/ioseekpos.c (_IO_seekpos_unlocked): Likewise. (_IO_seekpos): Likewise. * libio/iosetbuffer.c (_IO_setbuffer): Likewise. * libio/iosetvbuf.c (_IO_setvbuf): Likewise. * libio/ioungetc.c (_IO_ungetc): Likewise. * libio/ioungetwc.c (ungetwc): Likewise. * libio/iovdprintf.c (_IO_vdprintf): Likewise. * libio/iovsscanf.c (_IO_vsscanf): Likewise. * libio/iowpadn.c (_IO_wpadn): Likewise. * libio/libc_fatal.c (__libc_fatal): Likewise. * libio/memstream.c (__open_memstream): Likewise. (_IO_mem_sync): Likewise. (_IO_mem_finish): Likewise. * libio/oldfileops.c (_IO_old_file_init): Likewise. (_IO_old_file_close_it): Likewise. (_IO_old_file_finish): Likewise. (_IO_old_file_fopen): Likewise. (_IO_old_file_attach): Likewise. (_IO_old_file_setbuf): Likewise. (_IO_old_do_write): Likewise. (old_do_write): Likewise. (_IO_old_file_underflow): Likewise. (_IO_old_file_overflow): Likewise. (_IO_old_file_sync): Likewise. (_IO_old_file_seekoff): Likewise. (_IO_old_file_write): Likewise. (_IO_old_file_xsputn): Likewise. * libio/oldiofclose.c (_IO_old_fclose): Likewise. * libio/oldiofdopen.c (_IO_old_fdopen): Likewise. * libio/oldiofgetpos.c (_IO_old_fgetpos): Likewise. * libio/oldiofgetpos64.c (_IO_old_fgetpos64): Likewise. * libio/oldiofopen.c (_IO_old_fopen): Likewise. * libio/oldiofsetpos.c (_IO_old_fsetpos): Likewise. * libio/oldiofsetpos64.c (_IO_old_fsetpos64): Likewise. * libio/oldiopopen.c (_IO_old_proc_open): Likewise. (_IO_old_popen): Likewise. (_IO_old_proc_close): Likewise. * libio/oldpclose.c (__old_pclose): Likewise. * libio/pclose.c (__new_pclose): Likewise. * libio/peekc.c (_IO_peekc_locked): Likewise. * libio/putc.c (_IO_putc): Likewise. * libio/putc_u.c (putc_unlocked): Likewise. * libio/putchar.c (putchar): Likewise. * libio/putchar_u.c (putchar_unlocked): Likewise. * libio/putwc.c (putwc): Likewise. * libio/putwc_u.c (putwc_unlocked): Likewise. * libio/putwchar.c (putwchar): Likewise. * libio/putwchar_u.c (putwchar_unlocked): Likewise. * libio/rewind.c (rewind): Likewise. * libio/setbuf.c (setbuf): Likewise. * libio/setlinebuf.c (setlinebuf): Likewise. * libio/vasprintf.c (_IO_vasprintf): Likewise. * libio/vscanf.c (_IO_vscanf): Likewise. * libio/vsnprintf.c (_IO_strn_overflow): Likewise. * libio/vswprintf.c (_IO_wstrn_overflow): Likewise. * libio/wfiledoalloc.c (_IO_wfile_doallocate): Likewise. * libio/wgenops.c (_IO_least_wmarker): Likewise. (_IO_switch_to_main_wget_area): Likewise. (_IO_switch_to_wbackup_area): Likewise. (_IO_wsetb): Likewise. (_IO_wdefault_pbackfail): Likewise. (_IO_wdefault_finish): Likewise. (_IO_wdefault_uflow): Likewise. (__woverflow): Likewise. (__wuflow): Likewise. (__wunderflow): Likewise. (_IO_wdefault_xsputn): Likewise. (_IO_wdefault_xsgetn): Likewise. (_IO_wdoallocbuf): Likewise. (_IO_wdefault_doallocate): Likewise. (_IO_switch_to_wget_mode): Likewise. (_IO_free_wbackup_area): Likewise. (_IO_switch_to_wput_mode): Likewise. (_IO_sputbackwc): Likewise. (_IO_sungetwc): Likewise. (_IO_adjust_wcolumn): Likewise. (_IO_init_wmarker): Likewise. (_IO_wmarker_delta): Likewise. (_IO_seekwmark): Likewise. (_IO_unsave_wmarkers): Likewise. * libio/wmemstream.c (open_wmemstream): Likewise. (_IO_wmem_sync): Likewise. (_IO_wmem_finish): Likewise. * locale/nl_langinfo.c (nl_langinfo): Likewise. * locale/nl_langinfo_l.c (__nl_langinfo_l): Likewise. * locale/programs/simple-hash.c (init_hash): Likewise. (delete_hash): Likewise. (insert_entry): Likewise. (set_entry): Likewise. (next_prime): Likewise. (is_prime): Likewise. * locale/programs/xmalloc.c (fixup_null_alloc): Likewise. (xmalloc): Likewise. (xrealloc): Likewise. * locale/programs/xstrdup.c (xstrdup): Likewise. * localedata/collate-test.c (xstrcoll): Likewise. * localedata/xfrm-test.c (xstrcmp): Likewise. * login/getlogin_r.c (__getlogin_r): Likewise. * login/getpt.c (__posix_openpt): Likewise. * login/login_tty.c (login_tty): Likewise. * login/setlogin.c (setlogin): Likewise. * mach/msg-destroy.c (__mach_msg_destroy): Likewise. (mach_msg_destroy_port): Likewise. (mach_msg_destroy_memory): Likewise. * malloc/mcheck.c (flood): Likewise. * misc/acct.c (acct): Likewise. * misc/brk.c (__brk): Likewise. * misc/chflags.c (chflags): Likewise. * misc/chroot.c (chroot): Likewise. * misc/fchflags.c (fchflags): Likewise. * misc/fstab.c (getfsspec): Likewise. (getfsfile): Likewise. * misc/fsync.c (fsync): Likewise. * misc/ftruncate.c (__ftruncate): Likewise. * misc/ftruncate64.c (__ftruncate64): Likewise. * misc/getdomain.c (getdomainname): Likewise. (getdomainname): Likewise. * misc/gethostname.c (__gethostname): Likewise. * misc/getpass.c (getpass): Likewise. * misc/getttyent.c (skip): Likewise. (value): Likewise. * misc/gtty.c (gtty): Likewise. * misc/hsearch.c (hsearch): Likewise. (hcreate): Likewise. * misc/hsearch_r.c (__hcreate_r): Likewise. (__hdestroy_r): Likewise. * misc/ioctl.c (__ioctl): Likewise. * misc/mkdtemp.c (mkdtemp): Likewise. * misc/mkostemp.c (mkostemp): Likewise. * misc/mkostemp64.c (mkostemp64): Likewise. * misc/mkostemps.c (mkostemps): Likewise. * misc/mkostemps64.c (mkostemps64): Likewise. * misc/mkstemp.c (mkstemp): Likewise. * misc/mkstemp64.c (mkstemp64): Likewise. * misc/mkstemps.c (mkstemps): Likewise. * misc/mkstemps64.c (mkstemps64): Likewise. * misc/mktemp.c (__mktemp): Likewise. * misc/preadv.c (preadv): Likewise. * misc/preadv64.c (preadv64): Likewise. * misc/pwritev.c (pwritev): Likewise. * misc/pwritev64.c (pwritev64): Likewise. * misc/readv.c (__readv): Likewise. * misc/revoke.c (revoke): Likewise. * misc/setdomain.c (setdomainname): Likewise. * misc/setegid.c (setegid): Likewise. * misc/seteuid.c (seteuid): Likewise. * misc/sethostid.c (sethostid): Likewise. * misc/sethostname.c (sethostname): Likewise. * misc/setregid.c (__setregid): Likewise. * misc/setreuid.c (__setreuid): Likewise. * misc/sstk.c (sstk): Likewise. * misc/stty.c (stty): Likewise. * misc/syscall.c (syscall): Likewise. * misc/syslog.c (setlogmask): Likewise. * misc/truncate.c (__truncate): Likewise. * misc/truncate64.c (truncate64): Likewise. * misc/ualarm.c (ualarm): Likewise. * misc/usleep.c (usleep): Likewise. * misc/ustat.c (ustat): Likewise. * misc/writev.c (__writev): Likewise. * nptl/cleanup_compat.c (_pthread_cleanup_pop): Likewise. * nptl/old_pthread_cond_broadcast.c (__pthread_cond_broadcast_2_0): Likewise. * nptl/old_pthread_cond_destroy.c (__pthread_cond_destroy_2_0): Likewise. * nptl/old_pthread_cond_signal.c (__pthread_cond_signal_2_0): Likewise. * nptl/old_pthread_cond_wait.c (__pthread_cond_wait_2_0): Likewise. * nptl/pt-raise.c (raise): Likewise. * nptl/pthread_barrier_destroy.c (pthread_barrier_destroy): Likewise. * nptl/pthread_barrier_wait.c (__pthread_barrier_wait): Likewise. * nptl/pthread_barrierattr_destroy.c (pthread_barrierattr_destroy): Likewise. * nptl/pthread_barrierattr_init.c (pthread_barrierattr_init): Likewise. * nptl/pthread_barrierattr_setpshared.c (pthread_barrierattr_setpshared): Likewise. * nptl/pthread_cond_broadcast.c (__pthread_cond_broadcast): Likewise. * nptl/pthread_cond_destroy.c (__pthread_cond_destroy): Likewise. * nptl/pthread_cond_init.c (__pthread_cond_init): Likewise. * nptl/pthread_cond_signal.c (__pthread_cond_signal): Likewise. * nptl/pthread_condattr_destroy.c (__pthread_condattr_destroy): Likewise. * nptl/pthread_condattr_getclock.c (pthread_condattr_getclock): Likewise. * nptl/pthread_condattr_getpshared.c (pthread_condattr_getpshared): Likewise. * nptl/pthread_condattr_init.c (__pthread_condattr_init): Likewise. * nptl/pthread_condattr_setpshared.c (pthread_condattr_setpshared): Likewise. * nptl/pthread_detach.c (pthread_detach): Likewise. * nptl/pthread_equal.c (__pthread_equal): Likewise. * nptl/pthread_getcpuclockid.c (pthread_getcpuclockid): Likewise. * nptl/pthread_getspecific.c (__pthread_getspecific): Likewise. * nptl/pthread_key_delete.c (pthread_key_delete): Likewise. * nptl/pthread_mutex_consistent.c (pthread_mutex_consistent): Likewise. * nptl/pthread_mutex_destroy.c (__pthread_mutex_destroy): Likewise. * nptl/pthread_mutex_getprioceiling.c (pthread_mutex_getprioceiling): Likewise. * nptl/pthread_mutexattr_destroy.c (__pthread_mutexattr_destroy): Likewise. * nptl/pthread_mutexattr_getprotocol.c (pthread_mutexattr_getprotocol): Likewise. * nptl/pthread_mutexattr_getpshared.c (pthread_mutexattr_getpshared): Likewise. * nptl/pthread_mutexattr_getrobust.c (pthread_mutexattr_getrobust): Likewise. * nptl/pthread_mutexattr_gettype.c (pthread_mutexattr_gettype): Likewise. * nptl/pthread_mutexattr_init.c (__pthread_mutexattr_init): Likewise. * nptl/pthread_mutexattr_setprioceiling.c (pthread_mutexattr_setprioceiling): Likewise. * nptl/pthread_mutexattr_setprotocol.c (pthread_mutexattr_setprotocol): Likewise. * nptl/pthread_mutexattr_setpshared.c (pthread_mutexattr_setpshared): Likewise. * nptl/pthread_mutexattr_setrobust.c (pthread_mutexattr_setrobust): Likewise. * nptl/pthread_mutexattr_settype.c (__pthread_mutexattr_settype): Likewise. * nptl/pthread_rwlock_destroy.c (__pthread_rwlock_destroy): Likewise. * nptl/pthread_rwlockattr_destroy.c (pthread_rwlockattr_destroy): Likewise. * nptl/pthread_rwlockattr_getkind_np.c (pthread_rwlockattr_getkind_np): Likewise. * nptl/pthread_rwlockattr_getpshared.c (pthread_rwlockattr_getpshared): Likewise. * nptl/pthread_rwlockattr_init.c (pthread_rwlockattr_init): Likewise. * nptl/pthread_rwlockattr_setkind_np.c (pthread_rwlockattr_setkind_np): Likewise. * nptl/pthread_rwlockattr_setpshared.c (pthread_rwlockattr_setpshared): Likewise. * nptl/pthread_setcancelstate.c (__pthread_setcancelstate): Likewise. * nptl/pthread_setcanceltype.c (__pthread_setcanceltype): Likewise. * nptl/pthread_setconcurrency.c (pthread_setconcurrency): Likewise. * nptl/pthread_setschedprio.c (pthread_setschedprio): Likewise. * nptl/pthread_setspecific.c (__pthread_setspecific): Likewise. * nptl/pthread_spin_destroy.c (pthread_spin_destroy): Likewise. * nptl/pthread_tryjoin.c (pthread_tryjoin_np): Likewise. * nptl/sem_close.c (sem_close): Likewise. * nptl/sem_destroy.c (__new_sem_destroy): Likewise. * nptl/sem_init.c (__old_sem_init): Likewise. * nptl/sigaction.c (__sigaction): Likewise. * nptl/unregister-atfork.c (__unregister_atfork): Likewise. * posix/_exit.c (_exit): Likewise. * posix/alarm.c (alarm): Likewise. * posix/confstr.c (confstr): Likewise. * posix/fpathconf.c (__fpathconf): Likewise. * posix/getgroups.c (__getgroups): Likewise. * posix/getpgid.c (__getpgid): Likewise. * posix/group_member.c (__group_member): Likewise. * posix/pathconf.c (__pathconf): Likewise. * posix/sched_getaffinity.c (sched_getaffinity): Likewise. * posix/sched_setaffinity.c (sched_setaffinity): Likewise. * posix/setgid.c (__setgid): Likewise. * posix/setpgid.c (__setpgid): Likewise. * posix/setuid.c (__setuid): Likewise. * posix/sleep.c (__sleep): Likewise. * posix/sysconf.c (__sysconf): Likewise. * posix/times.c (__times): Likewise. * posix/uname.c (__uname): Likewise. * posix/waitid.c (__waitid): Likewise. * pwd/getpw.c (__getpw): Likewise. * resolv/base64.c (b64_pton): Likewise. * resolv/gai_sigqueue.c (__gai_sigqueue): Likewise. * resolv/gethnamaddr.c (Dprintf): Likewise. (gethostbyname): Likewise. (gethostbyname2): Likewise. (gethostbyaddr): Likewise. (_sethtent): Likewise. (_gethtbyname): Likewise. (_gethtbyname2): Likewise. (_gethtbyaddr): Likewise. (map_v4v6_address): Likewise. (map_v4v6_hostent): Likewise. (addrsort): Likewise. (ht_sethostent): Likewise. (ht_gethostbyname): Likewise. (ht_gethostbyaddr): Likewise. * resolv/inet_net_ntop.c (inet_net_ntop): Likewise. (inet_net_ntop_ipv4): Likewise. * resolv/inet_neta.c (inet_neta): Likewise. * resolv/inet_ntop.c (inet_ntop): Likewise. (inet_ntop4): Likewise. (inet_ntop6): Likewise. * resolv/inet_pton.c (__inet_pton): Likewise. (inet_pton4): Likewise. (inet_pton6): Likewise. * resolv/res_debug.c (loc_aton): Likewise. (loc_ntoa): Likewise. * resource/getpriority.c (__getpriority): Likewise. * resource/getrusage.c (__getrusage): Likewise. * resource/nice.c (nice): Likewise. * resource/setpriority.c (__setpriority): Likewise. * resource/setrlimit64.c (setrlimit64): Likewise. * resource/vlimit.c (vlimit): Likewise. * resource/vtimes.c (vtimes): Likewise. * rt/aio_error.c (aio_error): Likewise. * rt/aio_return.c (aio_return): Likewise. * rt/aio_sigqueue.c (__aio_sigqueue): Likewise. * signal/kill.c (__kill): Likewise. * signal/killpg.c (killpg): Likewise. * signal/raise.c (raise): Likewise. * signal/sigaction.c (__sigaction): Likewise. * signal/sigaddset.c (sigaddset): Likewise. * signal/sigaltstack.c (sigaltstack): Likewise. * signal/sigandset.c (sigandset): Likewise. * signal/sigblock.c (__sigblock): Likewise. * signal/sigdelset.c (sigdelset): Likewise. * signal/sigempty.c (sigemptyset): Likewise. * signal/sigfillset.c (sigfillset): Likewise. * signal/sighold.c (sighold): Likewise. * signal/sigignore.c (sigignore): Likewise. * signal/sigintr.c (siginterrupt): Likewise. * signal/sigisempty.c (sigisemptyset): Likewise. * signal/sigismem.c (sigismember): Likewise. * signal/signal.c (signal): Likewise. * signal/sigorset.c (sigorset): Likewise. * signal/sigpause.c (__sigpause): Likewise. * signal/sigpending.c (sigpending): Likewise. * signal/sigprocmask.c (__sigprocmask): Likewise. * signal/sigrelse.c (sigrelse): Likewise. * signal/sigreturn.c (__sigreturn): Likewise. * signal/sigset.c (sigset): Likewise. * signal/sigsetmask.c (__sigsetmask): Likewise. * signal/sigstack.c (sigstack): Likewise. * signal/sigsuspend.c (__sigsuspend): Likewise. * signal/sigvec.c (sigvec_wrapper_handler): Likewise. * signal/sysv_signal.c (__sysv_signal): Likewise. * socket/accept.c (accept): Likewise. * socket/accept4.c (__libc_accept4): Likewise. * socket/bind.c (__bind): Likewise. * socket/connect.c (__connect): Likewise. * socket/getpeername.c (getpeername): Likewise. * socket/getsockname.c (__getsockname): Likewise. * socket/getsockopt.c (getsockopt): Likewise. * socket/listen.c (__listen): Likewise. * socket/recv.c (__recv): Likewise. * socket/recvmsg.c (__recvmsg): Likewise. * socket/send.c (__send): Likewise. * socket/sendmsg.c (__sendmsg): Likewise. * socket/shutdown.c (shutdown): Likewise. * socket/sockatmark.c (sockatmark): Likewise. * socket/socket.c (__socket): Likewise. * stdio-common/ctermid.c (ctermid): Likewise. * stdio-common/cuserid.c (cuserid): Likewise. * stdio-common/printf-prs.c (parse_printf_format): Likewise. * stdio-common/remove.c (remove): Likewise. * stdio-common/rename.c (rename): Likewise. * stdio-common/renameat.c (renameat): Likewise. * stdio-common/tempname.c (__gen_tempname): Likewise. * stdio-common/xbug.c (InitBuffer): Likewise. (AppendToBuffer): Likewise. (ReadFile): Likewise. * stdlib/a64l.c (a64l): Likewise. * stdlib/drand48_r.c (drand48_r): Likewise. * stdlib/getcontext.c (getcontext): Likewise. * stdlib/getenv.c (getenv): Likewise. * stdlib/l64a.c (l64a): Likewise. * stdlib/llabs.c (llabs): Likewise. * stdlib/lldiv.c (lldiv): Likewise. * stdlib/lrand48_r.c (lrand48_r): Likewise. * stdlib/mrand48_r.c (mrand48_r): Likewise. * stdlib/putenv.c (putenv): Likewise. * stdlib/random.c (__srandom): Likewise. (__initstate): Likewise. (__setstate): Likewise. * stdlib/random_r.c (__srandom_r): Likewise. (__setstate_r): Likewise. (__random_r): Likewise. * stdlib/secure-getenv.c (__libc_secure_getenv): Likewise. * stdlib/setcontext.c (setcontext): Likewise. * stdlib/setenv.c (setenv): Likewise. (unsetenv): Likewise. * stdlib/srand48.c (srand48): Likewise. * stdlib/srand48_r.c (__srand48_r): Likewise. * stdlib/swapcontext.c (swapcontext): Likewise. * stdlib/system.c (__libc_system): Likewise. * stdlib/tst-strtod.c (expand): Likewise. * stdlib/tst-strtol.c (expand): Likewise. * stdlib/tst-strtoll.c (expand): Likewise. * streams/fattach.c (fattach): Likewise. * streams/fdetach.c (fdetach): Likewise. * streams/getmsg.c (getmsg): Likewise. * streams/isastream.c (isastream): Likewise. * string/ffs.c (__ffs): Likewise. * string/ffsll.c (ffsll): Likewise. * string/memcmp.c (memcmp_common_alignment): Likewise. (memcmp_not_common_alignment): Likewise. (MEMCMP): Likewise. * string/memcpy.c (memcpy): Likewise. * string/memmove.c (MEMMOVE): Likewise. * string/memset.c (memset): Likewise. * string/rawmemchr.c (RAWMEMCHR): Likewise. * string/strchrnul.c (STRCHRNUL): Likewise. * string/strerror.c (strerror): Likewise. * string/strndup.c (__strndup): Likewise. * string/strverscmp.c (__strverscmp): Likewise. * sunrpc/clnt_raw.c (clntraw_freeres): Likewise. * sunrpc/clnt_tcp.c (clnttcp_geterr): Likewise. (clnttcp_freeres): Likewise. * sunrpc/clnt_unix.c (clntunix_freeres): Likewise. * sunrpc/pmap_prot.c (xdr_pmap): Likewise. * sunrpc/pmap_prot2.c (xdr_pmaplist): Likewise. * sunrpc/pmap_rmt.c (xdr_rmtcallres): Likewise. * sunrpc/rpc_prot.c (xdr_replymsg): Likewise. (xdr_callhdr): Likewise. * sunrpc/rpcinfo.c (udpping): Likewise. (tcpping): Likewise. (pstatus): Likewise. (pmapdump): Likewise. (brdcst): Likewise. (deletereg): Likewise. (getprognum): Likewise. (getvers): Likewise. (get_inet_address): Likewise. * sunrpc/svc_raw.c (svcraw_recv): Likewise. * sunrpc/svc_udp.c (svcudp_create): Likewise. (svcudp_stat): Likewise. (svcudp_recv): Likewise. (svcudp_reply): Likewise. (svcudp_getargs): Likewise. (svcudp_freeargs): Likewise. (svcudp_destroy): Likewise. * sunrpc/xdr.c (xdr_bytes): Likewise. (xdr_netobj): Likewise. (xdr_string): Likewise. (xdr_wrapstring): Likewise. * sunrpc/xdr_float.c (xdr_float): Likewise. (xdr_double): Likewise. * sunrpc/xdr_mem.c (xdrmem_setpos): Likewise. * sunrpc/xdr_ref.c (xdr_pointer): Likewise. * sysvipc/ftok.c (ftok): Likewise. * sysvipc/msgctl.c (msgctl): Likewise. * sysvipc/msgget.c (msgget): Likewise. * sysvipc/msgrcv.c (msgrcv): Likewise. * sysvipc/msgsnd.c (msgsnd): Likewise. * sysvipc/semget.c (semget): Likewise. * sysvipc/semop.c (semop): Likewise. * sysvipc/shmat.c (shmat): Likewise. * sysvipc/shmctl.c (shmctl): Likewise. * sysvipc/shmdt.c (shmdt): Likewise. * sysvipc/shmget.c (shmget): Likewise. * termios/cfmakeraw.c (cfmakeraw): Likewise. * termios/speed.c (cfgetospeed): Likewise. (cfgetispeed): Likewise. (cfsetospeed): Likewise. (cfsetispeed): Likewise. * termios/tcflow.c (tcflow): Likewise. * termios/tcflush.c (tcflush): Likewise. * termios/tcgetattr.c (__tcgetattr): Likewise. * termios/tcgetpgrp.c (tcgetpgrp): Likewise. * termios/tcgetsid.c (tcgetsid): Likewise. * termios/tcsendbrk.c (tcsendbreak): Likewise. * termios/tcsetpgrp.c (tcsetpgrp): Likewise. * time/adjtime.c (__adjtime): Likewise. * time/dysize.c (dysize): Likewise. * time/ftime.c (ftime): Likewise. * time/getitimer.c (__getitimer): Likewise. * time/gettimeofday.c (__gettimeofday): Likewise. * time/gmtime.c (__gmtime_r): Likewise. (gmtime): Likewise. * time/localtime.c (__localtime_r): Likewise. (localtime): Likewise. * time/offtime.c (__offtime): Likewise. * time/settimeofday.c (__settimeofday): Likewise. * time/stime.c (stime): Likewise. * time/strftime_l.c (tm_diff): Likewise. (iso_week_days): Likewise. * time/strptime.c (strptime): Likewise. * time/time.c (time): Likewise. * time/timespec_get.c (timespec_get): Likewise. * time/tzset.c (tzset_internal): Likewise. (compute_change): Likewise. (__tz_compute): Likewise. * wcsmbs/btowc.c (__btowc): Likewise. * wcsmbs/mbrlen.c (__mbrlen): Likewise. * wcsmbs/mbsinit.c (__mbsinit): Likewise. * wcsmbs/mbsrtowcs.c (__mbsrtowcs): Likewise. * wcsmbs/wcpcpy.c (__wcpcpy): Likewise. * wcsmbs/wcpncpy.c (__wcpncpy): Likewise. * wcsmbs/wcscat.c (__wcscat): Likewise. * wcsmbs/wcschrnul.c (__wcschrnul): Likewise. * wcsmbs/wcscmp.c (WCSCMP): Likewise. * wcsmbs/wcscpy.c (WCSCPY): Likewise. * wcsmbs/wcscspn.c (wcscspn): Likewise. * wcsmbs/wcsdup.c (wcsdup): Likewise. * wcsmbs/wcslen.c (__wcslen): Likewise. * wcsmbs/wcsncat.c (WCSNCAT): Likewise. * wcsmbs/wcsncmp.c (WCSNCMP): Likewise. * wcsmbs/wcsncpy.c (__wcsncpy): Likewise. * wcsmbs/wcsnlen.c (__wcsnlen): Likewise. * wcsmbs/wcspbrk.c (wcspbrk): Likewise. * wcsmbs/wcsrchr.c (WCSRCHR): Likewise. * wcsmbs/wcsspn.c (wcsspn): Likewise. * wcsmbs/wcsstr.c (wcsstr): Likewise. * wcsmbs/wcstok.c (wcstok): Likewise. * wcsmbs/wctob.c (wctob): Likewise. * wcsmbs/wmemchr.c (__wmemchr): Likewise. * wcsmbs/wmemcmp.c (WMEMCMP): Likewise. * wcsmbs/wmemcpy.c (__wmemcpy): Likewise. * wcsmbs/wmemmove.c (__wmemmove): Likewise. * wcsmbs/wmempcpy.c (__wmempcpy): Likewise. * wcsmbs/wmemset.c (__wmemset): Likewise. * wctype/wcfuncs.c (__towlower): Likewise. (__towupper): Likewise.
2015-10-16 22:21:49 +02:00
__pthread_cond_signal (pthread_cond_t *cond)
{
LIBC_PROBE (cond_signal, 1, cond);
New condvar implementation that provides stronger ordering guarantees. This is a new implementation for condition variables, required after http://austingroupbugs.net/view.php?id=609 to fix bug 13165. In essence, we need to be stricter in which waiters a signal or broadcast is required to wake up; this couldn't be solved using the old algorithm. ISO C++ made a similar clarification, so this also fixes a bug in current libstdc++, for example. We can't use the old algorithm anymore because futexes do not guarantee to wake in FIFO order. Thus, when we wake, we can't simply let any waiter grab a signal, but we need to ensure that one of the waiters happening before the signal is woken up. This is something the previous algorithm violated (see bug 13165). There's another issue specific to condvars: ABA issues on the underlying futexes. Unlike mutexes that have just three states, or semaphores that have no tokens or a limited number of them, the state of a condvar is the *order* of the waiters. A waiter on a semaphore can grab a token whenever one is available; a condvar waiter must only consume a signal if it is eligible to do so as determined by the relative order of the waiter and the signal. Therefore, this new algorithm maintains two groups of waiters: Those eligible to consume signals (G1), and those that have to wait until previous waiters have consumed signals (G2). Once G1 is empty, G2 becomes the new G1. 64b counters are used to avoid ABA issues. This condvar doesn't yet use a requeue optimization (ie, on a broadcast, waking just one thread and requeueing all others on the futex of the mutex supplied by the program). I don't think doing the requeue is necessarily the right approach (but I haven't done real measurements yet): * If a program expects to wake many threads at the same time and make that scalable, a condvar isn't great anyway because of how it requires waiters to operate mutually exclusive (due to the mutex usage). Thus, a thundering herd problem is a scalability problem with or without the optimization. Using something like a semaphore might be more appropriate in such a case. * The scalability problem is actually at the mutex side; the condvar could help (and it tries to with the requeue optimization), but it should be the mutex who decides how that is done, and whether it is done at all. * Forcing all but one waiter into the kernel-side wait queue of the mutex prevents/avoids the use of lock elision on the mutex. Thus, it prevents the only cure against the underlying scalability problem inherent to condvars. * If condvars use short critical sections (ie, hold the mutex just to check a binary flag or such), which they should do ideally, then forcing all those waiter to proceed serially with kernel-based hand-off (ie, futex ops in the mutex' contended state, via the futex wait queues) will be less efficient than just letting a scalable mutex implementation take care of it. Our current mutex impl doesn't employ spinning at all, but if critical sections are short, spinning can be much better. * Doing the requeue stuff requires all waiters to always drive the mutex into the contended state. This leads to each waiter having to call futex_wake after lock release, even if this wouldn't be necessary. [BZ #13165] * nptl/pthread_cond_broadcast.c (__pthread_cond_broadcast): Rewrite to use new algorithm. * nptl/pthread_cond_destroy.c (__pthread_cond_destroy): Likewise. * nptl/pthread_cond_init.c (__pthread_cond_init): Likewise. * nptl/pthread_cond_signal.c (__pthread_cond_signal): Likewise. * nptl/pthread_cond_wait.c (__pthread_cond_wait): Likewise. (__pthread_cond_timedwait): Move here from pthread_cond_timedwait.c. (__condvar_confirm_wakeup, __condvar_cancel_waiting, __condvar_cleanup_waiting, __condvar_dec_grefs, __pthread_cond_wait_common): New. (__condvar_cleanup): Remove. * npt/pthread_condattr_getclock.c (pthread_condattr_getclock): Adapt. * npt/pthread_condattr_setclock.c (pthread_condattr_setclock): Likewise. * npt/pthread_condattr_getpshared.c (pthread_condattr_getpshared): Likewise. * npt/pthread_condattr_init.c (pthread_condattr_init): Likewise. * nptl/tst-cond1.c: Add comment. * nptl/tst-cond20.c (do_test): Adapt. * nptl/tst-cond22.c (do_test): Likewise. * sysdeps/aarch64/nptl/bits/pthreadtypes.h (pthread_cond_t): Adapt structure. * sysdeps/arm/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/ia64/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/m68k/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/microblaze/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/mips/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nios2/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/s390/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/sh/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/tile/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/alpha/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/powerpc/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/x86/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nptl/internaltypes.h (COND_NWAITERS_SHIFT): Remove. (COND_CLOCK_BITS): Adapt. * sysdeps/nptl/pthread.h (PTHREAD_COND_INITIALIZER): Adapt. * nptl/pthreadP.h (__PTHREAD_COND_CLOCK_MONOTONIC_MASK, __PTHREAD_COND_SHARED_MASK): New. * nptl/nptl-printers.py (CLOCK_IDS): Remove. (ConditionVariablePrinter, ConditionVariableAttributesPrinter): Adapt. * nptl/nptl_lock_constants.pysym: Adapt. * nptl/test-cond-printers.py: Adapt. * sysdeps/unix/sysv/linux/hppa/internaltypes.h (cond_compat_clear, cond_compat_check_and_clear): Adapt. * sysdeps/unix/sysv/linux/hppa/pthread_cond_timedwait.c: Remove file ... * sysdeps/unix/sysv/linux/hppa/pthread_cond_wait.c (__pthread_cond_timedwait): ... and move here. * nptl/DESIGN-condvar.txt: Remove file. * nptl/lowlevelcond.sym: Likewise. * nptl/pthread_cond_timedwait.c: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S: Likewise.
2016-05-25 23:43:36 +02:00
/* First check whether there are waiters. Relaxed MO is fine for that for
the same reasons that relaxed MO is fine when observing __wseq (see
below). */
unsigned int wrefs = atomic_load_relaxed (&cond->__data.__wrefs);
if (wrefs >> 3 == 0)
return 0;
int private = __condvar_get_private (wrefs);
__condvar_acquire_lock (cond, private);
/* Load the waiter sequence number, which represents our relative ordering
to any waiters. Relaxed MO is sufficient for that because:
1) We can pick any position that is allowed by external happens-before
constraints. In particular, if another __pthread_cond_wait call
happened before us, this waiter must be eligible for being woken by
us. The only way do establish such a happens-before is by signaling
while having acquired the mutex associated with the condvar and
ensuring that the signal's critical section happens after the waiter.
Thus, the mutex ensures that we see that waiter's __wseq increase.
2) Once we pick a position, we do not need to communicate this to the
program via a happens-before that we set up: First, any wake-up could
be a spurious wake-up, so the program must not interpret a wake-up as
an indication that the waiter happened before a particular signal;
second, a program cannot detect whether a waiter has not yet been
woken (i.e., it cannot distinguish between a non-woken waiter and one
that has been woken but hasn't resumed execution yet), and thus it
cannot try to deduce that a signal happened before a particular
waiter. */
unsigned long long int wseq = __condvar_load_wseq_relaxed (cond);
unsigned int g1 = (wseq & 1) ^ 1;
wseq >>= 1;
bool do_futex_wake = false;
/* If G1 is still receiving signals, we put the signal there. If not, we
check if G2 has waiters, and if so, quiesce and switch G1 to the former
G2; if this results in a new G1 with waiters (G2 might have cancellations
already, see __condvar_quiesce_and_switch_g1), we put the signal in the
new G1. */
if ((cond->__data.__g_size[g1] != 0)
|| __condvar_quiesce_and_switch_g1 (cond, wseq, &g1, private))
{
New condvar implementation that provides stronger ordering guarantees. This is a new implementation for condition variables, required after http://austingroupbugs.net/view.php?id=609 to fix bug 13165. In essence, we need to be stricter in which waiters a signal or broadcast is required to wake up; this couldn't be solved using the old algorithm. ISO C++ made a similar clarification, so this also fixes a bug in current libstdc++, for example. We can't use the old algorithm anymore because futexes do not guarantee to wake in FIFO order. Thus, when we wake, we can't simply let any waiter grab a signal, but we need to ensure that one of the waiters happening before the signal is woken up. This is something the previous algorithm violated (see bug 13165). There's another issue specific to condvars: ABA issues on the underlying futexes. Unlike mutexes that have just three states, or semaphores that have no tokens or a limited number of them, the state of a condvar is the *order* of the waiters. A waiter on a semaphore can grab a token whenever one is available; a condvar waiter must only consume a signal if it is eligible to do so as determined by the relative order of the waiter and the signal. Therefore, this new algorithm maintains two groups of waiters: Those eligible to consume signals (G1), and those that have to wait until previous waiters have consumed signals (G2). Once G1 is empty, G2 becomes the new G1. 64b counters are used to avoid ABA issues. This condvar doesn't yet use a requeue optimization (ie, on a broadcast, waking just one thread and requeueing all others on the futex of the mutex supplied by the program). I don't think doing the requeue is necessarily the right approach (but I haven't done real measurements yet): * If a program expects to wake many threads at the same time and make that scalable, a condvar isn't great anyway because of how it requires waiters to operate mutually exclusive (due to the mutex usage). Thus, a thundering herd problem is a scalability problem with or without the optimization. Using something like a semaphore might be more appropriate in such a case. * The scalability problem is actually at the mutex side; the condvar could help (and it tries to with the requeue optimization), but it should be the mutex who decides how that is done, and whether it is done at all. * Forcing all but one waiter into the kernel-side wait queue of the mutex prevents/avoids the use of lock elision on the mutex. Thus, it prevents the only cure against the underlying scalability problem inherent to condvars. * If condvars use short critical sections (ie, hold the mutex just to check a binary flag or such), which they should do ideally, then forcing all those waiter to proceed serially with kernel-based hand-off (ie, futex ops in the mutex' contended state, via the futex wait queues) will be less efficient than just letting a scalable mutex implementation take care of it. Our current mutex impl doesn't employ spinning at all, but if critical sections are short, spinning can be much better. * Doing the requeue stuff requires all waiters to always drive the mutex into the contended state. This leads to each waiter having to call futex_wake after lock release, even if this wouldn't be necessary. [BZ #13165] * nptl/pthread_cond_broadcast.c (__pthread_cond_broadcast): Rewrite to use new algorithm. * nptl/pthread_cond_destroy.c (__pthread_cond_destroy): Likewise. * nptl/pthread_cond_init.c (__pthread_cond_init): Likewise. * nptl/pthread_cond_signal.c (__pthread_cond_signal): Likewise. * nptl/pthread_cond_wait.c (__pthread_cond_wait): Likewise. (__pthread_cond_timedwait): Move here from pthread_cond_timedwait.c. (__condvar_confirm_wakeup, __condvar_cancel_waiting, __condvar_cleanup_waiting, __condvar_dec_grefs, __pthread_cond_wait_common): New. (__condvar_cleanup): Remove. * npt/pthread_condattr_getclock.c (pthread_condattr_getclock): Adapt. * npt/pthread_condattr_setclock.c (pthread_condattr_setclock): Likewise. * npt/pthread_condattr_getpshared.c (pthread_condattr_getpshared): Likewise. * npt/pthread_condattr_init.c (pthread_condattr_init): Likewise. * nptl/tst-cond1.c: Add comment. * nptl/tst-cond20.c (do_test): Adapt. * nptl/tst-cond22.c (do_test): Likewise. * sysdeps/aarch64/nptl/bits/pthreadtypes.h (pthread_cond_t): Adapt structure. * sysdeps/arm/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/ia64/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/m68k/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/microblaze/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/mips/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nios2/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/s390/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/sh/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/tile/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/alpha/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/powerpc/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/x86/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nptl/internaltypes.h (COND_NWAITERS_SHIFT): Remove. (COND_CLOCK_BITS): Adapt. * sysdeps/nptl/pthread.h (PTHREAD_COND_INITIALIZER): Adapt. * nptl/pthreadP.h (__PTHREAD_COND_CLOCK_MONOTONIC_MASK, __PTHREAD_COND_SHARED_MASK): New. * nptl/nptl-printers.py (CLOCK_IDS): Remove. (ConditionVariablePrinter, ConditionVariableAttributesPrinter): Adapt. * nptl/nptl_lock_constants.pysym: Adapt. * nptl/test-cond-printers.py: Adapt. * sysdeps/unix/sysv/linux/hppa/internaltypes.h (cond_compat_clear, cond_compat_check_and_clear): Adapt. * sysdeps/unix/sysv/linux/hppa/pthread_cond_timedwait.c: Remove file ... * sysdeps/unix/sysv/linux/hppa/pthread_cond_wait.c (__pthread_cond_timedwait): ... and move here. * nptl/DESIGN-condvar.txt: Remove file. * nptl/lowlevelcond.sym: Likewise. * nptl/pthread_cond_timedwait.c: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S: Likewise.
2016-05-25 23:43:36 +02:00
/* Add a signal. Relaxed MO is fine because signaling does not need to
establish a happens-before relation (see above). We do not mask the
release-MO store when initializing a group in
__condvar_quiesce_and_switch_g1 because we use an atomic
read-modify-write and thus extend that store's release sequence. */
atomic_fetch_add_relaxed (cond->__data.__g_signals + g1, 2);
cond->__data.__g_size[g1]--;
/* TODO Only set it if there are indeed futex waiters. */
do_futex_wake = true;
}
New condvar implementation that provides stronger ordering guarantees. This is a new implementation for condition variables, required after http://austingroupbugs.net/view.php?id=609 to fix bug 13165. In essence, we need to be stricter in which waiters a signal or broadcast is required to wake up; this couldn't be solved using the old algorithm. ISO C++ made a similar clarification, so this also fixes a bug in current libstdc++, for example. We can't use the old algorithm anymore because futexes do not guarantee to wake in FIFO order. Thus, when we wake, we can't simply let any waiter grab a signal, but we need to ensure that one of the waiters happening before the signal is woken up. This is something the previous algorithm violated (see bug 13165). There's another issue specific to condvars: ABA issues on the underlying futexes. Unlike mutexes that have just three states, or semaphores that have no tokens or a limited number of them, the state of a condvar is the *order* of the waiters. A waiter on a semaphore can grab a token whenever one is available; a condvar waiter must only consume a signal if it is eligible to do so as determined by the relative order of the waiter and the signal. Therefore, this new algorithm maintains two groups of waiters: Those eligible to consume signals (G1), and those that have to wait until previous waiters have consumed signals (G2). Once G1 is empty, G2 becomes the new G1. 64b counters are used to avoid ABA issues. This condvar doesn't yet use a requeue optimization (ie, on a broadcast, waking just one thread and requeueing all others on the futex of the mutex supplied by the program). I don't think doing the requeue is necessarily the right approach (but I haven't done real measurements yet): * If a program expects to wake many threads at the same time and make that scalable, a condvar isn't great anyway because of how it requires waiters to operate mutually exclusive (due to the mutex usage). Thus, a thundering herd problem is a scalability problem with or without the optimization. Using something like a semaphore might be more appropriate in such a case. * The scalability problem is actually at the mutex side; the condvar could help (and it tries to with the requeue optimization), but it should be the mutex who decides how that is done, and whether it is done at all. * Forcing all but one waiter into the kernel-side wait queue of the mutex prevents/avoids the use of lock elision on the mutex. Thus, it prevents the only cure against the underlying scalability problem inherent to condvars. * If condvars use short critical sections (ie, hold the mutex just to check a binary flag or such), which they should do ideally, then forcing all those waiter to proceed serially with kernel-based hand-off (ie, futex ops in the mutex' contended state, via the futex wait queues) will be less efficient than just letting a scalable mutex implementation take care of it. Our current mutex impl doesn't employ spinning at all, but if critical sections are short, spinning can be much better. * Doing the requeue stuff requires all waiters to always drive the mutex into the contended state. This leads to each waiter having to call futex_wake after lock release, even if this wouldn't be necessary. [BZ #13165] * nptl/pthread_cond_broadcast.c (__pthread_cond_broadcast): Rewrite to use new algorithm. * nptl/pthread_cond_destroy.c (__pthread_cond_destroy): Likewise. * nptl/pthread_cond_init.c (__pthread_cond_init): Likewise. * nptl/pthread_cond_signal.c (__pthread_cond_signal): Likewise. * nptl/pthread_cond_wait.c (__pthread_cond_wait): Likewise. (__pthread_cond_timedwait): Move here from pthread_cond_timedwait.c. (__condvar_confirm_wakeup, __condvar_cancel_waiting, __condvar_cleanup_waiting, __condvar_dec_grefs, __pthread_cond_wait_common): New. (__condvar_cleanup): Remove. * npt/pthread_condattr_getclock.c (pthread_condattr_getclock): Adapt. * npt/pthread_condattr_setclock.c (pthread_condattr_setclock): Likewise. * npt/pthread_condattr_getpshared.c (pthread_condattr_getpshared): Likewise. * npt/pthread_condattr_init.c (pthread_condattr_init): Likewise. * nptl/tst-cond1.c: Add comment. * nptl/tst-cond20.c (do_test): Adapt. * nptl/tst-cond22.c (do_test): Likewise. * sysdeps/aarch64/nptl/bits/pthreadtypes.h (pthread_cond_t): Adapt structure. * sysdeps/arm/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/ia64/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/m68k/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/microblaze/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/mips/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nios2/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/s390/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/sh/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/tile/nptl/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/alpha/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/unix/sysv/linux/powerpc/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/x86/bits/pthreadtypes.h (pthread_cond_t): Likewise. * sysdeps/nptl/internaltypes.h (COND_NWAITERS_SHIFT): Remove. (COND_CLOCK_BITS): Adapt. * sysdeps/nptl/pthread.h (PTHREAD_COND_INITIALIZER): Adapt. * nptl/pthreadP.h (__PTHREAD_COND_CLOCK_MONOTONIC_MASK, __PTHREAD_COND_SHARED_MASK): New. * nptl/nptl-printers.py (CLOCK_IDS): Remove. (ConditionVariablePrinter, ConditionVariableAttributesPrinter): Adapt. * nptl/nptl_lock_constants.pysym: Adapt. * nptl/test-cond-printers.py: Adapt. * sysdeps/unix/sysv/linux/hppa/internaltypes.h (cond_compat_clear, cond_compat_check_and_clear): Adapt. * sysdeps/unix/sysv/linux/hppa/pthread_cond_timedwait.c: Remove file ... * sysdeps/unix/sysv/linux/hppa/pthread_cond_wait.c (__pthread_cond_timedwait): ... and move here. * nptl/DESIGN-condvar.txt: Remove file. * nptl/lowlevelcond.sym: Likewise. * nptl/pthread_cond_timedwait.c: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i486/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i586/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/i386/i686/pthread_cond_wait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_broadcast.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_signal.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_timedwait.S: Likewise. * sysdeps/unix/sysv/linux/x86_64/pthread_cond_wait.S: Likewise.
2016-05-25 23:43:36 +02:00
__condvar_release_lock (cond, private);
if (do_futex_wake)
futex_wake (cond->__data.__g_signals + g1, 1, private);
return 0;
}
versioned_symbol (libpthread, __pthread_cond_signal, pthread_cond_signal,
GLIBC_2_3_2);