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.
2002-11-26 Ulrich Drepper <drepper@redhat.com>
* allocatestack.c (queue_stack): Don't remove stack from list here.
Do it in the caller. Correct condition to prematurely terminate
loop to free stacks.
(__deallocate_stack): Remove stack from list here.
2002-11-26 Ulrich Drepper <drepper@redhat.com>
* Makefile (tests): Add tst-stack1.
* tst-stack1.c: New file.
* allocatestack.c (allocate_stack): Initialize the TCB on a user
provided stack.
* pthread_attr_getstack.c: Return bottom of the thread area.
2002-11-25 Ulrich Drepper <drepper@redhat.com>
* Makefile (libpthread-routines): Add pt-allocrtsig and
pthread_kill_other_threads.
* pt-allocrtsig.c: New file.
* pthread_kill_other_threads.c: New file.
* sysdeps/unix/sysv/linux/allocrtsig.c: Add additional aliases for
all three functions.
* sysdeps/unix/sysv/linux/Makefile (sysdep_routines): Remove
allocrtsig.
* sysdeps/unix/sysv/linux/Versions (libc:GLIBC_PRIVATE): Export
__libc_current_sigrtmin_private, __libc_current_sigrtmax_private,
and __libc_allocate_rtsig_private.
* Versions (libpthread): Export pthread_kill_other_threads_np,
__libc_current_sigrtmin, and __libc_current_sigrtmax.
2002-11-24 Ulrich Drepper <drepper@redhat.com>
* allocatestack.c (allocate_stack): stackaddr in attribute points to
the end of the stack. Adjust computations.
When mprotect call fails dequeue stack and free it.
* pthread_attr_setstack.c: Store top of the stack in stackaddr
attribute.
* pthread_getattr_np.c: Likewise.
* descr.h (IS_DETACHED): Add some more parenthesis to prevent
surprises.
2002-11-23 Ulrich Drepper <drepper@redhat.com>
* sysdeps/pthread/pthread.h (pthread_self): __THROW must come before
attribute definitions. Patch by Luca Barbieri <ldb@ldb.ods.org>.
2002-11-22 Ulrich Drepper <drepper@redhat.com>
* pthread_getspecific.c: Optimize access to first 2nd-level array.
* pthread_setspecific.c: Likewise.
2002-11-21 Ulrich Drepper <drepper@redhat.com>
* sysdeps/unix/sysv/linux/i386/createthread.c: Remove CLONE_ flags
definitions. Get them from the official place.
* sysdeps/unix/sysv/linux/i386/fork.c: Likewise.
* sysdeps/unix/sysv/linux/i386/createthread.c: Update CLONE_* flags.
Use new CLONE_ flags in clone() calls.
* sysdeps/unix/sysv/linux/fork.c: Use ARCH_FORK to actually fork.
* sysdeps/unix/sysv/linux/i386/fork.c: New file.
* Versions: Add pthread_* functions for libc.
* forward.c: New file.
* sysdeps/pthread/Makefile (libpthread-sysdeps_routines): Add
errno-loc.
* herrno.c: New file.
* res.c: New file.
* Makefile (libpthread-routines): Remove sem_post, sem_wait,
sem_trywait, and sem_timedwait. Add herrno and res.
* sem_init.c: Don't initialize lock and waiters members.
* sem_open.c: Likewise.
* sem_post.c: Removed.
* sem_wait.c: Removed.
* sem_trywait.c: Removed.
* sem_timedwait.c: Removed.
* sysdeps/unix/sysv/linux/i386/i486/lowlevelsem.S: Complete rewrite.
Includes full implementations of sem_post, sem_wait, sem_trywait,
and sem_timedwait.
* sysdeps/unix/sysv/linux/i386/lowlevelsem.h (lll_sem_post): Adjust
for new implementation.
* sysdeps/unix/sysv/linux/internaltypes.h (struct sem): Remove lock
and waiters fields.
* tst-sem3.c: Improve error message.
* tst-signal3.c: Likewise.
* init.c (__pthread_initialize_minimal): Use set_tid_address syscall
to tell the kernel about the termination futex and to initialize tid
member. Don't initialize main_thread.
* descr.h (struct pthread): Remove main_thread member.
* cancelllation.c (__do_cancel): Remove code handling main thread.
The main thread is not special anymore.
* allocatestack.c (__reclaim_stacks): Mark stacks as unused. Add
size of the stacks to stack_cache_actsize.
* pt-readv.c: Add missing "defined".
* pt-sigwait.c: Likewise.
* pt-writev.c: Likewise.
2002-11-09 Ulrich Drepper <drepper@redhat.com>
* Versions: Export __connect from libpthread.
Patch by Luca Barbieri <ldb@ldb.ods.org>.
* Makefile (libpthread-routines): Add pt-raise.
* sysdeps/unix/sysv/linux/raise.c: New file.
* sysdeps/unix/sysv/linux/pt-raise.c: New file.
* sysdeps/generic/pt-raise.c: New file.
* pthread_cond_init.c: Initialize all data elements of the condvar
structure. Patch by Luca Barbieri <ldb@ldb.ods.org>.
* pthread_attr_init.c: Actually implement 2.0 compatibility version.
* pthread_create.c: Likewise.
* Makefile (tests): Add tst-key1, tst-key2, tst-key3.
* tst-key1.c: New file.
* tst-key2.c: New file.
* tst-key3.c: New file.
* Versions: Export pthread_detach for version GLIBC_2.0.
Reported by Saurabh Desai <sdesai@austin.ibm.com>.
2002-11-08 Ulrich Drepper <drepper@redhat.com>
* pthread_key_create.c: Terminate search after an unused key was found.
Patch by Luca Barbieri <ldb@ldb.ods.org>.
* sysdeps/unix/sysv/linux/i386/pthread_once.S: Return zero.
Patch by Luca Barbieri <ldb@ldb.ods.org>.
2002-10-10 Ulrich Drepper <drepper@redhat.com>
* sysdeps/unix/sysv/linux/i386/i486/lowlevelsem.S: Use slow generic
dynamic lookup for errno in PIC.
* allocatestack.c (get_cached_stack): Rearrange code slightly to
release the stack lock as soon as possible.
Call _dl_allocate_tls_init for TCB from the cache to re-initialize
the static TLS block.
(allocate_stack): Call _dl_allocate_tls_init for user-provided stack.
* cancellation.c: Renamed from cancelation.c.
* Makefile: Adjust accordingly.
* pthreadP.h (CANCELLATION_P): Renamed from CANCELATION_P.
* cleanup_defer.c: Use CANCELLATION_P.
* pthread_testcancel.c: Likewise.
* descr.h: Fix spelling in comments.
* init.c: Likewise.
* pthread_getattr_np.c: Likewise.
* pthread_getschedparam.c: Likewise.
* pthread_setschedparam.c: Likewise.
* Versions: Likewise.
* pt-pselect.c: New file.
* Makefile (libpthread-routines): Add pt-pselect.
* Versions: Add pselect.
* tst-cancel4.c: New file.
* Makefile (tests): Add tst-cancel4.
2002-10-09 Ulrich Drepper <drepper@redhat.com>
* pthread_mutex_lock.c: Always record lock ownership.
* pthread_mutex_timedlock.c: Likewise.
* pthread_mutex_trylock.c: Likewise.
* pt-readv.c: New file.
* pt-writev.c: New file.
* pt-creat.c: New file.
* pt-msgrcv.c: New file.
* pt-msgsnd.c: New file.
* pt-poll.c: New file.
* pt-select.c: New file.
* pt-sigpause.c: New file.
* pt-sigsuspend.c: New file.
* pt-sigwait.c: New file.
* pt-sigwaitinfo.c: New file.
* pt-waitid.c: New file.
* Makefile (libpthread-routines): Add pt-readv, pt-writev, pt-creat,
pt-msgrcv, pt-msgsnd, pt-poll, pt-select, pt-sigpause, pt-sigsuspend,
pt-sigwait, pt-sigwaitinfo, and pt-waitid.
* Versions: Add all the new functions.
* tst-exit1.c: New file.
* Makefile (tests): Add tst-exit1.
* sem_timedwait.c: Minor optimization for more optimal fastpath.
2002-10-08 Ulrich Drepper <drepper@redhat.com>
* pt-fcntl.c: Only enable asynchronous cancellation for F_SETLKW.
* pthread_join.c: Enable asynchronous cancellation around lll_wait_tid
call. pthread_join is an official cancellation point.
* pthread_timedjoin.c: Likewise.
* pthread_cond_wait.c: Revert order in which internal lock are dropped
and the condvar's mutex are retrieved.
* pthread_cond_timedwait.c: Likewise.
Reported by dice@saros.East.Sun.COM.
2002-10-07 Ulrich Drepper <drepper@redhat.com>
* pthreadP.h: Cut out all type definitions and move them...
* sysdeps/unix/sysv/linux/internaltypes.h: ...here. New file.
* pthreadP.h: Include <internaltypes.h>.
* sysdeps/unix/sysv/linux/i386/lowlevelsem.h (lll_sem_post): Little
performance tweaks.
* sem_trywait.c: Shuffle #includes around to get right order.
* sem_timedwait.c: Likewise.
* sem_post.c: Likewise.
* sem_wait.c: Likewise.
* nptl 0.3 released.
* Makefile (tests): Add tst-signal3.
* tst-signal3.c: New file.
2002-10-05 Ulrich Drepper <drepper@redhat.com>
* sysdeps/unix/sysv/linux/i386/lowlevelsem.h: Tell the compiler that
the asms modify the sem object.
(__lll_sem_timedwait): Now takes struct sem* as first parameter.
* sysdeps/unix/sysv/linux/i386/bits/semaphore.h (sem_t): Don't expose
the actual members.
* pthreadP.h (struct sem): New type. Actual semaphore type.
* semaphoreP.h: Include pthreadP.h.
* sem_getvalue.c: Adjust to sem_t change.
* sem_init.c: Likewise.
* sem_open.c: Likewise.
* sem_post.c: Likewise.
* sem_timedwait.c: Likewise.
* sem_trywait.c: Likewise.
* sem_wait.c: Likewise.
2002-10-04 Ulrich Drepper <drepper@redhat.com>
* Makefile (tests): Add tst-basic2, tst-exec1, tst-exec3, tst-exec3.
* tst-basic2.c: New file.
* tst-exec1.c: New file.
* tst-exec2.c: New file.
* tst-exec3.c: New file.
* tst-fork1.c: Remove extra */.
* nptl 0.2 released. The API for IA-32 is complete.
