David Xu 17ce606321 umtx operation UMTX_OP_MUTEX_WAKE has a side-effect that it accesses ... a mutex after a thread has unlocked it, it event writes data to the mutex memory to clear contention bit, there is a race that other threads can lock it and unlock it, then destroy it, so it should not write data to the mutex memory if there isn't any waiter. The new operation UMTX_OP_MUTEX_WAKE2 try to fix the problem. It requires thread library to clear the lock word entirely, then call the WAKE2 operation to check if there is any waiter in kernel, and try to wake up a thread, if necessary, the contention bit is set again by the operation. This also mitgates the chance that other threads find the contention bit and try to enter kernel to compete with each other to wake up sleeping thread, this is unnecessary. With this change, the mutex owner is no longer holding the mutex until it reaches a point where kernel umtx queue is locked, it releases the mutex as soon as possible. Performance is improved when the mutex is contensted heavily. On Intel i3-2310M, the runtime of a benchmark program is reduced from 26.87 seconds to 2.39 seconds, it even is better than UMTX_OP_MUTEX_WAKE which is deprecated now. http://people.freebsd.org/~davidxu/bench/mutex_perf.c		2012-04-05 02:24:08 +00:00
..
Makefile.inc	Introduce a non-portable function pthread_getthreadid_np(3) to retrieve	2011-02-07 21:26:46 +00:00
thr_affinity.c	Don't compare thread pointers again.	2010-09-13 11:58:42 +00:00
thr_attr.c	Use sysctl kern.sched.cpusetsize to retrieve size of kernel cpuset.	2010-11-02 02:13:13 +00:00
thr_autoinit.c	Make WARNS2 clean. The fixes mostly included:	2003-05-23 09:48:20 +00:00
thr_barrier.c	When destroying a barrier, waiting all threads exit the barrier,	2012-03-16 04:35:52 +00:00
thr_barrierattr.c	WARNS level 4 cleanup.	2006-04-04 02:57:49 +00:00
thr_cancel.c	To support stack unwinding for cancellation points, add -fexceptions flag	2010-09-25 01:57:47 +00:00
thr_clean.c	Only access unwind_disabled when _PTHREAD_FORCED_UNWIND is defined.	2010-09-25 09:43:24 +00:00
thr_concurrency.c	WARNS level 4 cleanup.	2006-04-04 02:57:49 +00:00
thr_cond.c	Because sleepqueue may still being used, we should always check wchan with	2011-01-04 05:35:19 +00:00
thr_condattr.c	Coalesce one more broken line.	2010-05-24 13:44:39 +00:00
thr_create.c	Implement the __pthread_map_stacks_exec() for libthr.	2011-01-09 12:38:40 +00:00
thr_detach.c	Convert thread list lock from mutex to rwlock.	2010-09-13 07:03:01 +00:00
thr_equal.c	Remove 3rd clause, renumber, ok per email	2007-01-12 07:26:21 +00:00
thr_event.c	Allow psaddr_t to be widened by using thr_pread_{int,long,ptr},	2008-09-14 16:07:21 +00:00
thr_exit.c	change code to use unwind.h.	2010-09-30 12:59:56 +00:00
thr_fork.c	In most cases, cancel_point and cancel_async needn't be checked again,	2010-09-24 07:52:07 +00:00
thr_getcpuclockid.c	Add POSIX pthread API pthread_getcpuclockid() to get a thread's cpu	2008-03-22 09:59:20 +00:00
thr_getprio.c	Remove 3rd clause, renumber, ok per email	2007-01-12 07:26:21 +00:00
thr_getschedparam.c	use rtprio_thread system call to get or set thread priority.	2006-09-21 04:21:30 +00:00
thr_getthreadid_np.c	Introduce a non-portable function pthread_getthreadid_np(3) to retrieve	2011-02-07 21:26:46 +00:00
thr_info.c	If a thread who's name is being set is not the current thread, use macros	2007-04-05 07:20:31 +00:00
thr_init.c	Do not set thread name to less than informative 'initial thread'.	2011-06-19 13:35:36 +00:00
thr_join.c	Fix typo.	2010-09-19 08:55:36 +00:00
thr_kern.c	MFp4:	2010-12-22 05:01:52 +00:00
thr_kill.c	Convert thread list lock from mutex to rwlock.	2010-09-13 07:03:01 +00:00
thr_list.c	Make code more stable by checking NULL pointers.	2012-02-11 04:12:12 +00:00
thr_main_np.c	WARNS level 4 cleanup.	2006-04-04 02:57:49 +00:00
thr_multi_np.c	Remove 3rd clause, renumber, ok per email	2007-01-12 07:26:21 +00:00
thr_mutex.c	Return 0 instead of garbage value.	2011-01-06 08:13:30 +00:00
thr_mutexattr.c	pthread_mutexattr_destroy() was accidentally broken in last revision,	2008-03-20 11:47:08 +00:00
thr_once.c	r179417 introduced a bug into pthread_once(). Previously pthread_once()	2011-04-20 14:19:34 +00:00
thr_printf.c	Import my recent 1:1 threading working. some features improved includes:	2005-04-02 01:20:00 +00:00
thr_private.h	umtx operation UMTX_OP_MUTEX_WAKE has a side-effect that it accesses	2012-04-05 02:24:08 +00:00
thr_pspinlock.c	Reverse the logic of UP and SMP.	2007-10-16 07:36:02 +00:00
thr_resume_np.c	Convert thread list lock from mutex to rwlock.	2010-09-13 07:03:01 +00:00
thr_rtld.c	Implement the __pthread_map_stacks_exec() for libthr.	2011-01-09 12:38:40 +00:00
thr_rwlock.c	Follow changes made in revision 232144, pass absolute timeout to kernel,	2012-02-27 13:38:52 +00:00
thr_rwlockattr.c	Revert revision 214007, I realized that MySQL wants to resolve	2010-10-20 02:34:02 +00:00
thr_self.c	Remove 3rd clause, renumber, ok per email	2007-01-12 07:26:21 +00:00
thr_sem.c	Use umtx to implement process sharable semaphore, to make this work,	2010-01-05 02:37:59 +00:00
thr_setprio.c	Remove 3rd clause, renumber, ok per email	2007-01-12 07:26:21 +00:00
thr_setschedparam.c	use rtprio_thread system call to get or set thread priority.	2006-09-21 04:21:30 +00:00
thr_sig.c	libthr: In the atfork handlers for signals, do not skip the last signal.	2012-03-26 17:05:26 +00:00
thr_single_np.c	Remove 3rd clause, renumber, ok per email	2007-01-12 07:26:21 +00:00
thr_sleepq.c	Add sleep queue code.	2010-12-22 05:03:24 +00:00
thr_spec.c	Unregister thread specific data destructor when a corresponding dso	2010-08-27 05:20:22 +00:00
thr_spinlock.c	Avoid various shadowed variables. libthr is now almost WARNS=4 clean except	2008-04-23 21:06:51 +00:00
thr_stack.c	For the process that already loaded libthr but still not initialized	2011-01-10 16:10:25 +00:00
thr_suspend_np.c	Convert thread list lock from mutex to rwlock.	2010-09-13 07:03:01 +00:00
thr_switch_np.c	WARNS level 4 cleanup.	2006-04-04 02:57:49 +00:00
thr_symbols.c	Remove 3rd clause, renumber, ok per email	2007-01-12 07:26:21 +00:00
thr_syscalls.c	Add signal handler wrapper, the reason to add it becauses there are	2010-09-01 02:18:33 +00:00
thr_umtx.c	Use clockid parameter instead of hard-coded CLOCK_REALTIME.	2012-03-19 00:07:10 +00:00
thr_umtx.h	umtx operation UMTX_OP_MUTEX_WAKE has a side-effect that it accesses	2012-04-05 02:24:08 +00:00
thr_yield.c	Remove 3rd clause, renumber, ok per email	2007-01-12 07:26:21 +00:00