1996-01-22 00:23:58 +00:00
|
|
|
/*
|
|
|
|
* Copyright (c) 1995 John Birrell <jb@cimlogic.com.au>.
|
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* Redistribution and use in source and binary forms, with or without
|
|
|
|
* modification, are permitted provided that the following conditions
|
|
|
|
* are met:
|
|
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer.
|
|
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
|
|
* documentation and/or other materials provided with the distribution.
|
|
|
|
* 3. All advertising materials mentioning features or use of this software
|
|
|
|
* must display the following acknowledgement:
|
|
|
|
* This product includes software developed by John Birrell.
|
|
|
|
* 4. Neither the name of the author nor the names of any co-contributors
|
|
|
|
* may be used to endorse or promote products derived from this software
|
|
|
|
* without specific prior written permission.
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY JOHN BIRRELL AND CONTRIBUTORS ``AS IS'' AND
|
|
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
1999-03-23 05:07:56 +00:00
|
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
|
1996-01-22 00:23:58 +00:00
|
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
|
|
* SUCH DAMAGE.
|
|
|
|
*
|
1999-08-28 00:22:10 +00:00
|
|
|
* $FreeBSD$
|
1996-01-22 00:23:58 +00:00
|
|
|
*/
|
|
|
|
#include <stdlib.h>
|
|
|
|
#include <errno.h>
|
1998-06-09 23:21:05 +00:00
|
|
|
#include <string.h>
|
1996-01-22 00:23:58 +00:00
|
|
|
#ifdef _THREAD_SAFE
|
|
|
|
#include <pthread.h>
|
|
|
|
#include "pthread_private.h"
|
|
|
|
|
1999-03-23 05:07:56 +00:00
|
|
|
/*
|
|
|
|
* Prototypes
|
|
|
|
*/
|
|
|
|
static inline pthread_t cond_queue_deq(pthread_cond_t);
|
|
|
|
static inline void cond_queue_remove(pthread_cond_t, pthread_t);
|
|
|
|
static inline void cond_queue_enq(pthread_cond_t, pthread_t);
|
|
|
|
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
/* Reinitialize a condition variable to defaults. */
|
|
|
|
int
|
2000-11-09 05:08:26 +00:00
|
|
|
_cond_reinit(pthread_cond_t *cond)
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
if (cond == NULL)
|
|
|
|
ret = EINVAL;
|
|
|
|
else if (*cond == NULL)
|
|
|
|
ret = pthread_cond_init(cond, NULL);
|
|
|
|
else {
|
|
|
|
/*
|
|
|
|
* Initialize the condition variable structure:
|
|
|
|
*/
|
|
|
|
TAILQ_INIT(&(*cond)->c_queue);
|
|
|
|
(*cond)->c_flags = COND_FLAGS_INITED;
|
|
|
|
(*cond)->c_type = COND_TYPE_FAST;
|
|
|
|
(*cond)->c_mutex = NULL;
|
2000-11-09 05:08:26 +00:00
|
|
|
(*cond)->c_seqno = 0;
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
memset(&(*cond)->lock, 0, sizeof((*cond)->lock));
|
|
|
|
}
|
|
|
|
return (ret);
|
|
|
|
}
|
1999-03-23 05:07:56 +00:00
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
int
|
2000-11-09 05:08:26 +00:00
|
|
|
pthread_cond_init(pthread_cond_t *cond, const pthread_condattr_t *cond_attr)
|
1996-01-22 00:23:58 +00:00
|
|
|
{
|
|
|
|
enum pthread_cond_type type;
|
1996-08-20 08:22:01 +00:00
|
|
|
pthread_cond_t pcond;
|
1996-01-22 00:23:58 +00:00
|
|
|
int rval = 0;
|
|
|
|
|
1998-04-29 09:59:34 +00:00
|
|
|
if (cond == NULL)
|
1997-11-25 01:29:16 +00:00
|
|
|
rval = EINVAL;
|
1998-04-29 09:59:34 +00:00
|
|
|
else {
|
1996-08-20 08:22:01 +00:00
|
|
|
/*
|
1997-04-11 22:38:22 +00:00
|
|
|
* Check if a pointer to a condition variable attribute
|
|
|
|
* structure was passed by the caller:
|
1996-08-20 08:22:01 +00:00
|
|
|
*/
|
|
|
|
if (cond_attr != NULL && *cond_attr != NULL) {
|
|
|
|
/* Default to a fast condition variable: */
|
|
|
|
type = (*cond_attr)->c_type;
|
|
|
|
} else {
|
|
|
|
/* Default to a fast condition variable: */
|
|
|
|
type = COND_TYPE_FAST;
|
|
|
|
}
|
1996-01-22 00:23:58 +00:00
|
|
|
|
1996-08-20 08:22:01 +00:00
|
|
|
/* Process according to condition variable type: */
|
|
|
|
switch (type) {
|
1997-04-11 22:38:22 +00:00
|
|
|
/* Fast condition variable: */
|
1996-08-20 08:22:01 +00:00
|
|
|
case COND_TYPE_FAST:
|
|
|
|
/* Nothing to do here. */
|
|
|
|
break;
|
|
|
|
|
1997-04-11 22:38:22 +00:00
|
|
|
/* Trap invalid condition variable types: */
|
1996-08-20 08:22:01 +00:00
|
|
|
default:
|
|
|
|
/* Return an invalid argument error: */
|
1997-11-25 01:29:16 +00:00
|
|
|
rval = EINVAL;
|
1996-08-20 08:22:01 +00:00
|
|
|
break;
|
|
|
|
}
|
1996-01-22 00:23:58 +00:00
|
|
|
|
1996-08-20 08:22:01 +00:00
|
|
|
/* Check for no errors: */
|
|
|
|
if (rval == 0) {
|
1997-04-11 22:38:22 +00:00
|
|
|
if ((pcond = (pthread_cond_t)
|
|
|
|
malloc(sizeof(struct pthread_cond))) == NULL) {
|
1997-11-25 01:29:16 +00:00
|
|
|
rval = ENOMEM;
|
1996-08-20 08:22:01 +00:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Initialise the condition variable
|
|
|
|
* structure:
|
|
|
|
*/
|
1999-03-23 05:07:56 +00:00
|
|
|
TAILQ_INIT(&pcond->c_queue);
|
1996-08-20 08:22:01 +00:00
|
|
|
pcond->c_flags |= COND_FLAGS_INITED;
|
|
|
|
pcond->c_type = type;
|
1999-03-23 05:07:56 +00:00
|
|
|
pcond->c_mutex = NULL;
|
2000-11-09 05:08:26 +00:00
|
|
|
pcond->c_seqno = 0;
|
1998-06-09 23:21:05 +00:00
|
|
|
memset(&pcond->lock,0,sizeof(pcond->lock));
|
1996-08-20 08:22:01 +00:00
|
|
|
*cond = pcond;
|
|
|
|
}
|
|
|
|
}
|
1996-01-22 00:23:58 +00:00
|
|
|
}
|
|
|
|
/* Return the completion status: */
|
|
|
|
return (rval);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
2000-11-09 05:08:26 +00:00
|
|
|
pthread_cond_destroy(pthread_cond_t *cond)
|
1996-01-22 00:23:58 +00:00
|
|
|
{
|
|
|
|
int rval = 0;
|
|
|
|
|
1998-04-29 09:59:34 +00:00
|
|
|
if (cond == NULL || *cond == NULL)
|
1997-11-25 01:29:16 +00:00
|
|
|
rval = EINVAL;
|
1998-04-29 09:59:34 +00:00
|
|
|
else {
|
|
|
|
/* Lock the condition variable structure: */
|
1998-06-09 23:21:05 +00:00
|
|
|
_SPINLOCK(&(*cond)->lock);
|
1996-08-20 08:22:01 +00:00
|
|
|
|
1998-04-29 09:59:34 +00:00
|
|
|
/*
|
|
|
|
* Free the memory allocated for the condition
|
|
|
|
* variable structure:
|
|
|
|
*/
|
|
|
|
free(*cond);
|
1996-01-22 00:23:58 +00:00
|
|
|
|
1998-04-29 09:59:34 +00:00
|
|
|
/*
|
|
|
|
* NULL the caller's pointer now that the condition
|
|
|
|
* variable has been destroyed:
|
|
|
|
*/
|
|
|
|
*cond = NULL;
|
1996-01-22 00:23:58 +00:00
|
|
|
}
|
|
|
|
/* Return the completion status: */
|
|
|
|
return (rval);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
2000-11-09 05:08:26 +00:00
|
|
|
pthread_cond_wait(pthread_cond_t *cond, pthread_mutex_t *mutex)
|
1996-01-22 00:23:58 +00:00
|
|
|
{
|
2000-01-19 07:04:50 +00:00
|
|
|
int rval = 0;
|
2000-11-09 05:08:26 +00:00
|
|
|
int done = 0;
|
2000-01-19 07:04:50 +00:00
|
|
|
int interrupted = 0;
|
2000-11-09 05:08:26 +00:00
|
|
|
int unlock_mutex = 1;
|
|
|
|
int seqno;
|
1996-01-22 00:23:58 +00:00
|
|
|
|
Simplify sytem call renaming. Instead of _foo() <-- _libc_foo <-- foo(),
just use _foo() <-- foo(). In the case of a libpthread that doesn't do
call conversion (such as linuxthreads and our upcoming libpthread), this
is adequate. In the case of libc_r, we still need three names, which are
now _thread_sys_foo() <-- _foo() <-- foo().
Convert all internal libc usage of: aio_suspend(), close(), fsync(), msync(),
nanosleep(), open(), fcntl(), read(), and write() to _foo() instead of foo().
Remove all internal libc usage of: creat(), pause(), sleep(), system(),
tcdrain(), wait(), and waitpid().
Make thread cancellation fully POSIX-compliant.
Suggested by: deischen
2000-01-27 23:07:25 +00:00
|
|
|
_thread_enter_cancellation_point();
|
|
|
|
|
1998-04-04 11:33:01 +00:00
|
|
|
if (cond == NULL)
|
2000-11-09 05:08:26 +00:00
|
|
|
return (EINVAL);
|
1998-04-04 11:33:01 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If the condition variable is statically initialized,
|
|
|
|
* perform the dynamic initialization:
|
|
|
|
*/
|
2000-11-09 05:08:26 +00:00
|
|
|
if (*cond == NULL &&
|
|
|
|
(rval = pthread_cond_init(cond, NULL)) != 0)
|
|
|
|
return (rval);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Enter a loop waiting for a condition signal or broadcast
|
|
|
|
* to wake up this thread. A loop is needed in case the waiting
|
|
|
|
* thread is interrupted by a signal to execute a signal handler.
|
|
|
|
* It is not (currently) possible to remain in the waiting queue
|
|
|
|
* while running a handler. Instead, the thread is interrupted
|
|
|
|
* and backed out of the waiting queue prior to executing the
|
|
|
|
* signal handler.
|
|
|
|
*/
|
|
|
|
do {
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
/* Lock the condition variable structure: */
|
|
|
|
_SPINLOCK(&(*cond)->lock);
|
|
|
|
|
1999-05-23 10:55:33 +00:00
|
|
|
/*
|
|
|
|
* If the condvar was statically allocated, properly
|
|
|
|
* initialize the tail queue.
|
|
|
|
*/
|
|
|
|
if (((*cond)->c_flags & COND_FLAGS_INITED) == 0) {
|
|
|
|
TAILQ_INIT(&(*cond)->c_queue);
|
|
|
|
(*cond)->c_flags |= COND_FLAGS_INITED;
|
|
|
|
}
|
|
|
|
|
1996-08-20 08:22:01 +00:00
|
|
|
/* Process according to condition variable type: */
|
|
|
|
switch ((*cond)->c_type) {
|
1997-04-11 22:38:22 +00:00
|
|
|
/* Fast condition variable: */
|
1996-08-20 08:22:01 +00:00
|
|
|
case COND_TYPE_FAST:
|
1999-03-23 05:07:56 +00:00
|
|
|
if ((mutex == NULL) || (((*cond)->c_mutex != NULL) &&
|
|
|
|
((*cond)->c_mutex != *mutex))) {
|
1998-11-06 21:04:02 +00:00
|
|
|
/* Unlock the condition variable structure: */
|
|
|
|
_SPINUNLOCK(&(*cond)->lock);
|
1999-03-23 05:07:56 +00:00
|
|
|
|
|
|
|
/* Return invalid argument error: */
|
|
|
|
rval = EINVAL;
|
1998-11-06 21:04:02 +00:00
|
|
|
} else {
|
1999-11-28 05:38:13 +00:00
|
|
|
/* Reset the timeout and interrupted flags: */
|
1999-03-23 05:07:56 +00:00
|
|
|
_thread_run->timeout = 0;
|
1999-11-28 05:38:13 +00:00
|
|
|
_thread_run->interrupted = 0;
|
1996-01-22 00:23:58 +00:00
|
|
|
|
1999-03-23 05:07:56 +00:00
|
|
|
/*
|
|
|
|
* Queue the running thread for the condition
|
|
|
|
* variable:
|
|
|
|
*/
|
|
|
|
cond_queue_enq(*cond, _thread_run);
|
|
|
|
|
2000-11-09 05:08:26 +00:00
|
|
|
/* Remember the mutex and sequence number: */
|
1999-03-23 05:07:56 +00:00
|
|
|
(*cond)->c_mutex = *mutex;
|
2000-11-09 05:08:26 +00:00
|
|
|
seqno = (*cond)->c_seqno;
|
1999-03-23 05:07:56 +00:00
|
|
|
|
|
|
|
/* Wait forever: */
|
|
|
|
_thread_run->wakeup_time.tv_sec = -1;
|
|
|
|
|
|
|
|
/* Unlock the mutex: */
|
2000-11-09 05:08:26 +00:00
|
|
|
if ((unlock_mutex != 0) &&
|
|
|
|
((rval = _mutex_cv_unlock(mutex)) != 0)) {
|
1999-03-23 05:07:56 +00:00
|
|
|
/*
|
|
|
|
* Cannot unlock the mutex, so remove
|
|
|
|
* the running thread from the condition
|
|
|
|
* variable queue:
|
|
|
|
*/
|
|
|
|
cond_queue_remove(*cond, _thread_run);
|
|
|
|
|
|
|
|
/* Check for no more waiters: */
|
|
|
|
if (TAILQ_FIRST(&(*cond)->c_queue) ==
|
|
|
|
NULL)
|
|
|
|
(*cond)->c_mutex = NULL;
|
|
|
|
|
|
|
|
/* Unlock the condition variable structure: */
|
|
|
|
_SPINUNLOCK(&(*cond)->lock);
|
|
|
|
}
|
|
|
|
else {
|
2000-11-09 05:08:26 +00:00
|
|
|
/*
|
|
|
|
* Don't unlock the mutex in the event
|
|
|
|
* this thread has to be requeued in
|
|
|
|
* condition variable queue:
|
|
|
|
*/
|
|
|
|
unlock_mutex = 0;
|
|
|
|
|
1999-03-23 05:07:56 +00:00
|
|
|
/*
|
|
|
|
* Schedule the next thread and unlock
|
|
|
|
* the condition variable structure:
|
|
|
|
*/
|
|
|
|
_thread_kern_sched_state_unlock(PS_COND_WAIT,
|
|
|
|
&(*cond)->lock, __FILE__, __LINE__);
|
|
|
|
|
2000-11-09 05:08:26 +00:00
|
|
|
done = (seqno != (*cond)->c_seqno);
|
2000-01-19 07:04:50 +00:00
|
|
|
|
2000-11-09 05:08:26 +00:00
|
|
|
if ((_thread_run->flags &
|
|
|
|
PTHREAD_FLAGS_IN_CONDQ) != 0) {
|
1999-11-28 05:38:13 +00:00
|
|
|
/*
|
|
|
|
* Lock the condition variable
|
|
|
|
* while removing the thread.
|
|
|
|
*/
|
|
|
|
_SPINLOCK(&(*cond)->lock);
|
|
|
|
|
|
|
|
cond_queue_remove(*cond,
|
|
|
|
_thread_run);
|
|
|
|
|
|
|
|
/* Check for no more waiters: */
|
|
|
|
if (TAILQ_FIRST(&(*cond)->c_queue) == NULL)
|
|
|
|
(*cond)->c_mutex = NULL;
|
|
|
|
|
|
|
|
_SPINUNLOCK(&(*cond)->lock);
|
|
|
|
}
|
|
|
|
|
2000-11-09 05:08:26 +00:00
|
|
|
/*
|
|
|
|
* Save the interrupted flag; locking
|
|
|
|
* the mutex will destroy it.
|
|
|
|
*/
|
|
|
|
interrupted = _thread_run->interrupted;
|
|
|
|
|
1999-11-28 05:38:13 +00:00
|
|
|
/*
|
|
|
|
* Note that even though this thread may have
|
|
|
|
* been canceled, POSIX requires that the mutex
|
|
|
|
* be reaquired prior to cancellation.
|
|
|
|
*/
|
1999-03-23 05:07:56 +00:00
|
|
|
rval = _mutex_cv_lock(mutex);
|
|
|
|
}
|
1998-11-06 21:04:02 +00:00
|
|
|
}
|
1996-08-20 08:22:01 +00:00
|
|
|
break;
|
|
|
|
|
1997-04-11 22:38:22 +00:00
|
|
|
/* Trap invalid condition variable types: */
|
1996-08-20 08:22:01 +00:00
|
|
|
default:
|
1998-11-06 21:04:02 +00:00
|
|
|
/* Unlock the condition variable structure: */
|
|
|
|
_SPINUNLOCK(&(*cond)->lock);
|
|
|
|
|
1996-08-20 08:22:01 +00:00
|
|
|
/* Return an invalid argument error: */
|
1997-11-25 01:29:16 +00:00
|
|
|
rval = EINVAL;
|
1996-08-20 08:22:01 +00:00
|
|
|
break;
|
|
|
|
}
|
1999-11-28 05:38:13 +00:00
|
|
|
|
2000-11-09 05:08:26 +00:00
|
|
|
if ((interrupted != 0) && (_thread_run->continuation != NULL))
|
|
|
|
_thread_run->continuation((void *) _thread_run);
|
|
|
|
} while ((done == 0) && (rval == 0));
|
1996-01-22 00:23:58 +00:00
|
|
|
|
Simplify sytem call renaming. Instead of _foo() <-- _libc_foo <-- foo(),
just use _foo() <-- foo(). In the case of a libpthread that doesn't do
call conversion (such as linuxthreads and our upcoming libpthread), this
is adequate. In the case of libc_r, we still need three names, which are
now _thread_sys_foo() <-- _foo() <-- foo().
Convert all internal libc usage of: aio_suspend(), close(), fsync(), msync(),
nanosleep(), open(), fcntl(), read(), and write() to _foo() instead of foo().
Remove all internal libc usage of: creat(), pause(), sleep(), system(),
tcdrain(), wait(), and waitpid().
Make thread cancellation fully POSIX-compliant.
Suggested by: deischen
2000-01-27 23:07:25 +00:00
|
|
|
_thread_leave_cancellation_point();
|
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/* Return the completion status: */
|
|
|
|
return (rval);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
pthread_cond_timedwait(pthread_cond_t * cond, pthread_mutex_t * mutex,
|
|
|
|
const struct timespec * abstime)
|
|
|
|
{
|
2000-01-19 07:04:50 +00:00
|
|
|
int rval = 0;
|
2000-11-09 05:08:26 +00:00
|
|
|
int done = 0;
|
2000-01-19 07:04:50 +00:00
|
|
|
int interrupted = 0;
|
2000-11-09 05:08:26 +00:00
|
|
|
int unlock_mutex = 1;
|
|
|
|
int seqno;
|
1996-01-22 00:23:58 +00:00
|
|
|
|
Simplify sytem call renaming. Instead of _foo() <-- _libc_foo <-- foo(),
just use _foo() <-- foo(). In the case of a libpthread that doesn't do
call conversion (such as linuxthreads and our upcoming libpthread), this
is adequate. In the case of libc_r, we still need three names, which are
now _thread_sys_foo() <-- _foo() <-- foo().
Convert all internal libc usage of: aio_suspend(), close(), fsync(), msync(),
nanosleep(), open(), fcntl(), read(), and write() to _foo() instead of foo().
Remove all internal libc usage of: creat(), pause(), sleep(), system(),
tcdrain(), wait(), and waitpid().
Make thread cancellation fully POSIX-compliant.
Suggested by: deischen
2000-01-27 23:07:25 +00:00
|
|
|
_thread_enter_cancellation_point();
|
|
|
|
|
2000-07-17 22:55:05 +00:00
|
|
|
if (abstime == NULL || abstime->tv_sec < 0 || abstime->tv_nsec < 0 ||
|
|
|
|
abstime->tv_nsec >= 1000000000)
|
2000-11-09 05:08:26 +00:00
|
|
|
return (EINVAL);
|
1998-04-04 11:33:01 +00:00
|
|
|
/*
|
2000-07-17 22:55:05 +00:00
|
|
|
* If the condition variable is statically initialized, perform dynamic
|
|
|
|
* initialization.
|
1998-04-04 11:33:01 +00:00
|
|
|
*/
|
2000-11-09 05:08:26 +00:00
|
|
|
if (*cond == NULL && (rval = pthread_cond_init(cond, NULL)) != 0)
|
|
|
|
return (rval);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Enter a loop waiting for a condition signal or broadcast
|
|
|
|
* to wake up this thread. A loop is needed in case the waiting
|
|
|
|
* thread is interrupted by a signal to execute a signal handler.
|
|
|
|
* It is not (currently) possible to remain in the waiting queue
|
|
|
|
* while running a handler. Instead, the thread is interrupted
|
|
|
|
* and backed out of the waiting queue prior to executing the
|
|
|
|
* signal handler.
|
|
|
|
*/
|
|
|
|
do {
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
/* Lock the condition variable structure: */
|
|
|
|
_SPINLOCK(&(*cond)->lock);
|
|
|
|
|
1999-05-23 10:55:33 +00:00
|
|
|
/*
|
|
|
|
* If the condvar was statically allocated, properly
|
|
|
|
* initialize the tail queue.
|
|
|
|
*/
|
|
|
|
if (((*cond)->c_flags & COND_FLAGS_INITED) == 0) {
|
|
|
|
TAILQ_INIT(&(*cond)->c_queue);
|
|
|
|
(*cond)->c_flags |= COND_FLAGS_INITED;
|
|
|
|
}
|
|
|
|
|
1996-08-20 08:22:01 +00:00
|
|
|
/* Process according to condition variable type: */
|
|
|
|
switch ((*cond)->c_type) {
|
1997-04-11 22:38:22 +00:00
|
|
|
/* Fast condition variable: */
|
1996-08-20 08:22:01 +00:00
|
|
|
case COND_TYPE_FAST:
|
1999-03-23 05:07:56 +00:00
|
|
|
if ((mutex == NULL) || (((*cond)->c_mutex != NULL) &&
|
|
|
|
((*cond)->c_mutex != *mutex))) {
|
|
|
|
/* Return invalid argument error: */
|
|
|
|
rval = EINVAL;
|
1998-11-06 21:04:02 +00:00
|
|
|
|
|
|
|
/* Unlock the condition variable structure: */
|
|
|
|
_SPINUNLOCK(&(*cond)->lock);
|
1996-08-20 08:22:01 +00:00
|
|
|
} else {
|
1999-03-23 05:07:56 +00:00
|
|
|
/* Set the wakeup time: */
|
|
|
|
_thread_run->wakeup_time.tv_sec =
|
|
|
|
abstime->tv_sec;
|
|
|
|
_thread_run->wakeup_time.tv_nsec =
|
|
|
|
abstime->tv_nsec;
|
1996-08-20 08:22:01 +00:00
|
|
|
|
1999-11-28 05:38:13 +00:00
|
|
|
/* Reset the timeout and interrupted flags: */
|
1999-03-23 05:07:56 +00:00
|
|
|
_thread_run->timeout = 0;
|
1999-11-28 05:38:13 +00:00
|
|
|
_thread_run->interrupted = 0;
|
1999-03-23 05:07:56 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Queue the running thread for the condition
|
|
|
|
* variable:
|
|
|
|
*/
|
|
|
|
cond_queue_enq(*cond, _thread_run);
|
|
|
|
|
2000-11-09 05:08:26 +00:00
|
|
|
/* Remember the mutex and sequence number: */
|
1999-03-23 05:07:56 +00:00
|
|
|
(*cond)->c_mutex = *mutex;
|
2000-11-09 05:08:26 +00:00
|
|
|
seqno = (*cond)->c_seqno;
|
1999-03-23 05:07:56 +00:00
|
|
|
|
|
|
|
/* Unlock the mutex: */
|
2000-11-09 05:08:26 +00:00
|
|
|
if ((unlock_mutex != 0) &&
|
|
|
|
((rval = _mutex_cv_unlock(mutex)) != 0)) {
|
1999-03-23 05:07:56 +00:00
|
|
|
/*
|
|
|
|
* Cannot unlock the mutex, so remove
|
|
|
|
* the running thread from the condition
|
|
|
|
* variable queue:
|
|
|
|
*/
|
|
|
|
cond_queue_remove(*cond, _thread_run);
|
|
|
|
|
|
|
|
/* Check for no more waiters: */
|
|
|
|
if (TAILQ_FIRST(&(*cond)->c_queue) == NULL)
|
|
|
|
(*cond)->c_mutex = NULL;
|
|
|
|
|
|
|
|
/* Unlock the condition variable structure: */
|
|
|
|
_SPINUNLOCK(&(*cond)->lock);
|
|
|
|
} else {
|
2000-11-09 05:08:26 +00:00
|
|
|
/*
|
|
|
|
* Don't unlock the mutex in the event
|
|
|
|
* this thread has to be requeued in
|
|
|
|
* condition variable queue:
|
|
|
|
*/
|
|
|
|
unlock_mutex = 0;
|
|
|
|
|
1999-03-23 05:07:56 +00:00
|
|
|
/*
|
|
|
|
* Schedule the next thread and unlock
|
|
|
|
* the condition variable structure:
|
|
|
|
*/
|
|
|
|
_thread_kern_sched_state_unlock(PS_COND_WAIT,
|
|
|
|
&(*cond)->lock, __FILE__, __LINE__);
|
|
|
|
|
2000-11-09 05:08:26 +00:00
|
|
|
done = (seqno != (*cond)->c_seqno);
|
|
|
|
|
1999-11-28 05:38:13 +00:00
|
|
|
/*
|
2000-11-09 05:08:26 +00:00
|
|
|
* Check if the wait timedout, was
|
|
|
|
* interrupted (canceled), or needs to
|
|
|
|
* be resumed after handling a signal.
|
1999-11-28 05:38:13 +00:00
|
|
|
*/
|
|
|
|
if ((_thread_run->timeout == 0) &&
|
2000-11-09 05:08:26 +00:00
|
|
|
(_thread_run->interrupted == 0) &&
|
|
|
|
(done != 0)) {
|
1999-03-23 05:07:56 +00:00
|
|
|
/* Lock the mutex: */
|
|
|
|
rval = _mutex_cv_lock(mutex);
|
1999-11-28 05:38:13 +00:00
|
|
|
} else {
|
2000-11-09 05:08:26 +00:00
|
|
|
/* Lock the CV structure: */
|
1999-03-23 05:07:56 +00:00
|
|
|
_SPINLOCK(&(*cond)->lock);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The wait timed out; remove
|
|
|
|
* the thread from the condition
|
2000-11-09 05:08:26 +00:00
|
|
|
* variable queue:
|
1999-03-23 05:07:56 +00:00
|
|
|
*/
|
|
|
|
cond_queue_remove(*cond,
|
|
|
|
_thread_run);
|
|
|
|
|
|
|
|
/* Check for no more waiters: */
|
|
|
|
if (TAILQ_FIRST(&(*cond)->c_queue) == NULL)
|
|
|
|
(*cond)->c_mutex = NULL;
|
|
|
|
|
2000-11-09 05:08:26 +00:00
|
|
|
/* Unock the CV structure: */
|
1999-03-23 05:07:56 +00:00
|
|
|
_SPINUNLOCK(&(*cond)->lock);
|
|
|
|
|
|
|
|
/* Return a timeout error: */
|
2000-11-09 05:08:26 +00:00
|
|
|
if (_thread_run->timeout != 0)
|
|
|
|
rval = ETIMEDOUT;
|
|
|
|
/*
|
|
|
|
* Save the interrupted flag;
|
|
|
|
* locking the mutex will
|
|
|
|
* destroy it.
|
|
|
|
*/
|
|
|
|
interrupted = _thread_run->interrupted;
|
1999-03-23 05:07:56 +00:00
|
|
|
|
|
|
|
/*
|
1999-11-28 05:38:13 +00:00
|
|
|
* Lock the mutex and ignore any
|
|
|
|
* errors. Note that even though
|
|
|
|
* this thread may have been
|
|
|
|
* canceled, POSIX requires that
|
|
|
|
* the mutex be reaquired prior
|
|
|
|
* to cancellation.
|
1999-03-23 05:07:56 +00:00
|
|
|
*/
|
|
|
|
(void)_mutex_cv_lock(mutex);
|
|
|
|
}
|
1996-08-20 08:22:01 +00:00
|
|
|
}
|
1996-01-22 00:23:58 +00:00
|
|
|
}
|
1996-08-20 08:22:01 +00:00
|
|
|
break;
|
|
|
|
|
1997-04-11 22:38:22 +00:00
|
|
|
/* Trap invalid condition variable types: */
|
1996-08-20 08:22:01 +00:00
|
|
|
default:
|
1998-11-06 21:04:02 +00:00
|
|
|
/* Unlock the condition variable structure: */
|
|
|
|
_SPINUNLOCK(&(*cond)->lock);
|
|
|
|
|
1996-08-20 08:22:01 +00:00
|
|
|
/* Return an invalid argument error: */
|
1997-11-25 01:29:16 +00:00
|
|
|
rval = EINVAL;
|
1996-08-20 08:22:01 +00:00
|
|
|
break;
|
1996-01-22 00:23:58 +00:00
|
|
|
}
|
|
|
|
|
2000-11-09 05:08:26 +00:00
|
|
|
if ((interrupted != 0) && (_thread_run->continuation != NULL))
|
|
|
|
_thread_run->continuation((void *) _thread_run);
|
|
|
|
} while ((done == 0) && (rval == 0));
|
1996-01-22 00:23:58 +00:00
|
|
|
|
Simplify sytem call renaming. Instead of _foo() <-- _libc_foo <-- foo(),
just use _foo() <-- foo(). In the case of a libpthread that doesn't do
call conversion (such as linuxthreads and our upcoming libpthread), this
is adequate. In the case of libc_r, we still need three names, which are
now _thread_sys_foo() <-- _foo() <-- foo().
Convert all internal libc usage of: aio_suspend(), close(), fsync(), msync(),
nanosleep(), open(), fcntl(), read(), and write() to _foo() instead of foo().
Remove all internal libc usage of: creat(), pause(), sleep(), system(),
tcdrain(), wait(), and waitpid().
Make thread cancellation fully POSIX-compliant.
Suggested by: deischen
2000-01-27 23:07:25 +00:00
|
|
|
_thread_leave_cancellation_point();
|
2000-07-18 01:38:19 +00:00
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/* Return the completion status: */
|
|
|
|
return (rval);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
pthread_cond_signal(pthread_cond_t * cond)
|
|
|
|
{
|
|
|
|
int rval = 0;
|
|
|
|
pthread_t pthread;
|
|
|
|
|
2000-07-17 22:55:05 +00:00
|
|
|
if (cond == NULL)
|
1997-11-25 01:29:16 +00:00
|
|
|
rval = EINVAL;
|
2000-07-17 22:55:05 +00:00
|
|
|
/*
|
|
|
|
* If the condition variable is statically initialized, perform dynamic
|
|
|
|
* initialization.
|
|
|
|
*/
|
2000-11-16 22:50:33 +00:00
|
|
|
else if (*cond != NULL || (rval = pthread_cond_init(cond, NULL)) == 0) {
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
/*
|
|
|
|
* Defer signals to protect the scheduling queues
|
|
|
|
* from access by the signal handler:
|
|
|
|
*/
|
|
|
|
_thread_kern_sig_defer();
|
|
|
|
|
1998-04-29 09:59:34 +00:00
|
|
|
/* Lock the condition variable structure: */
|
1998-06-09 23:21:05 +00:00
|
|
|
_SPINLOCK(&(*cond)->lock);
|
1996-08-20 08:22:01 +00:00
|
|
|
|
|
|
|
/* Process according to condition variable type: */
|
|
|
|
switch ((*cond)->c_type) {
|
1997-04-11 22:38:22 +00:00
|
|
|
/* Fast condition variable: */
|
1996-08-20 08:22:01 +00:00
|
|
|
case COND_TYPE_FAST:
|
2000-11-09 05:08:26 +00:00
|
|
|
/* Increment the sequence number: */
|
|
|
|
(*cond)->c_seqno++;
|
|
|
|
|
2000-06-14 17:17:41 +00:00
|
|
|
if ((pthread = cond_queue_deq(*cond)) != NULL) {
|
|
|
|
/*
|
|
|
|
* Unless the thread is currently suspended,
|
|
|
|
* allow it to run. If the thread is suspended,
|
|
|
|
* make a note that the thread isn't in a wait
|
|
|
|
* queue any more.
|
|
|
|
*/
|
|
|
|
if (pthread->state != PS_SUSPENDED)
|
|
|
|
PTHREAD_NEW_STATE(pthread,PS_RUNNING);
|
|
|
|
else
|
|
|
|
pthread->suspended = SUSP_NOWAIT;
|
|
|
|
}
|
1999-03-23 05:07:56 +00:00
|
|
|
|
|
|
|
/* Check for no more waiters: */
|
|
|
|
if (TAILQ_FIRST(&(*cond)->c_queue) == NULL)
|
|
|
|
(*cond)->c_mutex = NULL;
|
1996-08-20 08:22:01 +00:00
|
|
|
break;
|
|
|
|
|
1997-04-11 22:38:22 +00:00
|
|
|
/* Trap invalid condition variable types: */
|
1996-08-20 08:22:01 +00:00
|
|
|
default:
|
|
|
|
/* Return an invalid argument error: */
|
1997-11-25 01:29:16 +00:00
|
|
|
rval = EINVAL;
|
1996-08-20 08:22:01 +00:00
|
|
|
break;
|
1996-01-22 00:23:58 +00:00
|
|
|
}
|
|
|
|
|
1998-04-29 09:59:34 +00:00
|
|
|
/* Unlock the condition variable structure: */
|
1998-06-09 23:21:05 +00:00
|
|
|
_SPINUNLOCK(&(*cond)->lock);
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Undefer and handle pending signals, yielding if
|
|
|
|
* necessary:
|
|
|
|
*/
|
|
|
|
_thread_kern_sig_undefer();
|
1996-01-22 00:23:58 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Return the completion status: */
|
|
|
|
return (rval);
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
pthread_cond_broadcast(pthread_cond_t * cond)
|
|
|
|
{
|
|
|
|
int rval = 0;
|
|
|
|
pthread_t pthread;
|
|
|
|
|
2000-07-17 22:55:05 +00:00
|
|
|
if (cond == NULL)
|
1998-04-29 09:59:34 +00:00
|
|
|
rval = EINVAL;
|
2000-07-17 22:55:05 +00:00
|
|
|
/*
|
|
|
|
* If the condition variable is statically initialized, perform dynamic
|
|
|
|
* initialization.
|
|
|
|
*/
|
2000-11-16 22:50:33 +00:00
|
|
|
else if (*cond != NULL || (rval = pthread_cond_init(cond, NULL)) == 0) {
|
1999-03-23 05:07:56 +00:00
|
|
|
/*
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
* Defer signals to protect the scheduling queues
|
|
|
|
* from access by the signal handler:
|
1999-03-23 05:07:56 +00:00
|
|
|
*/
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
_thread_kern_sig_defer();
|
1999-03-23 05:07:56 +00:00
|
|
|
|
1998-04-29 09:59:34 +00:00
|
|
|
/* Lock the condition variable structure: */
|
1998-06-09 23:21:05 +00:00
|
|
|
_SPINLOCK(&(*cond)->lock);
|
1996-01-22 00:23:58 +00:00
|
|
|
|
1998-04-29 09:59:34 +00:00
|
|
|
/* Process according to condition variable type: */
|
|
|
|
switch ((*cond)->c_type) {
|
|
|
|
/* Fast condition variable: */
|
|
|
|
case COND_TYPE_FAST:
|
2000-11-09 05:08:26 +00:00
|
|
|
/* Increment the sequence number: */
|
|
|
|
(*cond)->c_seqno++;
|
|
|
|
|
1998-04-29 09:59:34 +00:00
|
|
|
/*
|
|
|
|
* Enter a loop to bring all threads off the
|
|
|
|
* condition queue:
|
|
|
|
*/
|
1999-03-23 05:07:56 +00:00
|
|
|
while ((pthread = cond_queue_deq(*cond)) != NULL) {
|
2000-06-14 17:17:41 +00:00
|
|
|
/*
|
|
|
|
* Unless the thread is currently suspended,
|
|
|
|
* allow it to run. If the thread is suspended,
|
|
|
|
* make a note that the thread isn't in a wait
|
|
|
|
* queue any more.
|
|
|
|
*/
|
|
|
|
if (pthread->state != PS_SUSPENDED)
|
|
|
|
PTHREAD_NEW_STATE(pthread,PS_RUNNING);
|
|
|
|
else
|
|
|
|
pthread->suspended = SUSP_NOWAIT;
|
1998-04-29 09:59:34 +00:00
|
|
|
}
|
1999-03-23 05:07:56 +00:00
|
|
|
|
|
|
|
/* There are no more waiting threads: */
|
|
|
|
(*cond)->c_mutex = NULL;
|
1998-04-29 09:59:34 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
/* Trap invalid condition variable types: */
|
|
|
|
default:
|
|
|
|
/* Return an invalid argument error: */
|
|
|
|
rval = EINVAL;
|
|
|
|
break;
|
1996-01-22 00:23:58 +00:00
|
|
|
}
|
|
|
|
|
1998-04-29 09:59:34 +00:00
|
|
|
/* Unlock the condition variable structure: */
|
1998-06-09 23:21:05 +00:00
|
|
|
_SPINUNLOCK(&(*cond)->lock);
|
1999-03-23 05:07:56 +00:00
|
|
|
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
/*
|
|
|
|
* Undefer and handle pending signals, yielding if
|
|
|
|
* necessary:
|
1999-03-23 05:07:56 +00:00
|
|
|
*/
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
_thread_kern_sig_undefer();
|
1996-01-22 00:23:58 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Return the completion status: */
|
|
|
|
return (rval);
|
|
|
|
}
|
1999-03-23 05:07:56 +00:00
|
|
|
|
2000-10-13 22:12:32 +00:00
|
|
|
void
|
|
|
|
_cond_wait_backout(pthread_t pthread)
|
|
|
|
{
|
|
|
|
pthread_cond_t cond;
|
|
|
|
|
|
|
|
cond = pthread->data.cond;
|
|
|
|
if (cond != NULL) {
|
|
|
|
/*
|
|
|
|
* Defer signals to protect the scheduling queues
|
|
|
|
* from access by the signal handler:
|
|
|
|
*/
|
|
|
|
_thread_kern_sig_defer();
|
|
|
|
|
|
|
|
/* Lock the condition variable structure: */
|
|
|
|
_SPINLOCK(&cond->lock);
|
|
|
|
|
|
|
|
/* Process according to condition variable type: */
|
|
|
|
switch (cond->c_type) {
|
|
|
|
/* Fast condition variable: */
|
|
|
|
case COND_TYPE_FAST:
|
|
|
|
cond_queue_remove(cond, pthread);
|
|
|
|
|
|
|
|
/* Check for no more waiters: */
|
|
|
|
if (TAILQ_FIRST(&cond->c_queue) == NULL)
|
|
|
|
cond->c_mutex = NULL;
|
|
|
|
break;
|
|
|
|
|
|
|
|
default:
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Unlock the condition variable structure: */
|
|
|
|
_SPINUNLOCK(&cond->lock);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Undefer and handle pending signals, yielding if
|
|
|
|
* necessary:
|
|
|
|
*/
|
|
|
|
_thread_kern_sig_undefer();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1999-03-23 05:07:56 +00:00
|
|
|
/*
|
|
|
|
* Dequeue a waiting thread from the head of a condition queue in
|
|
|
|
* descending priority order.
|
|
|
|
*/
|
|
|
|
static inline pthread_t
|
|
|
|
cond_queue_deq(pthread_cond_t cond)
|
|
|
|
{
|
|
|
|
pthread_t pthread;
|
|
|
|
|
1999-11-28 05:38:13 +00:00
|
|
|
while ((pthread = TAILQ_FIRST(&cond->c_queue)) != NULL) {
|
2000-10-13 22:12:32 +00:00
|
|
|
TAILQ_REMOVE(&cond->c_queue, pthread, sqe);
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
pthread->flags &= ~PTHREAD_FLAGS_IN_CONDQ;
|
1999-11-28 05:38:13 +00:00
|
|
|
if ((pthread->timeout == 0) && (pthread->interrupted == 0))
|
|
|
|
/*
|
|
|
|
* Only exit the loop when we find a thread
|
|
|
|
* that hasn't timed out or been canceled;
|
|
|
|
* those threads are already running and don't
|
|
|
|
* need their run state changed.
|
|
|
|
*/
|
|
|
|
break;
|
1999-03-23 05:07:56 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
return(pthread);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Remove a waiting thread from a condition queue in descending priority
|
|
|
|
* order.
|
|
|
|
*/
|
|
|
|
static inline void
|
|
|
|
cond_queue_remove(pthread_cond_t cond, pthread_t pthread)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Because pthread_cond_timedwait() can timeout as well
|
|
|
|
* as be signaled by another thread, it is necessary to
|
|
|
|
* guard against removing the thread from the queue if
|
|
|
|
* it isn't in the queue.
|
|
|
|
*/
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
if (pthread->flags & PTHREAD_FLAGS_IN_CONDQ) {
|
2000-10-13 22:12:32 +00:00
|
|
|
TAILQ_REMOVE(&cond->c_queue, pthread, sqe);
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
pthread->flags &= ~PTHREAD_FLAGS_IN_CONDQ;
|
1999-03-23 05:07:56 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Enqueue a waiting thread to a condition queue in descending priority
|
|
|
|
* order.
|
|
|
|
*/
|
|
|
|
static inline void
|
|
|
|
cond_queue_enq(pthread_cond_t cond, pthread_t pthread)
|
|
|
|
{
|
|
|
|
pthread_t tid = TAILQ_LAST(&cond->c_queue, cond_head);
|
|
|
|
|
2000-10-13 22:12:32 +00:00
|
|
|
PTHREAD_ASSERT_NOT_IN_SYNCQ(pthread);
|
|
|
|
|
1999-03-23 05:07:56 +00:00
|
|
|
/*
|
|
|
|
* For the common case of all threads having equal priority,
|
|
|
|
* we perform a quick check against the priority of the thread
|
|
|
|
* at the tail of the queue.
|
|
|
|
*/
|
|
|
|
if ((tid == NULL) || (pthread->active_priority <= tid->active_priority))
|
2000-10-13 22:12:32 +00:00
|
|
|
TAILQ_INSERT_TAIL(&cond->c_queue, pthread, sqe);
|
1999-03-23 05:07:56 +00:00
|
|
|
else {
|
|
|
|
tid = TAILQ_FIRST(&cond->c_queue);
|
|
|
|
while (pthread->active_priority <= tid->active_priority)
|
2000-10-13 22:12:32 +00:00
|
|
|
tid = TAILQ_NEXT(tid, sqe);
|
|
|
|
TAILQ_INSERT_BEFORE(tid, pthread, sqe);
|
1999-03-23 05:07:56 +00:00
|
|
|
}
|
In the words of the author:
o The polling mechanism for I/O readiness was changed from
select() to poll(). In additon, a wrapped version of poll()
is now provided.
o The wrapped select routine now converts each fd_set to a
poll array so that the thread scheduler doesn't have to
perform a bitwise search for selected fds each time file
descriptors are polled for I/O readiness.
o The thread scheduler was modified to use a new queue (_workq)
for threads that need work. Threads waiting for I/O readiness
and spinblocks are added to the work queue in addition to the
waiting queue. This reduces the time spent forming/searching
the array of file descriptors being polled.
o The waiting queue (_waitingq) is now maintained in order of
thread wakeup time. This allows the thread scheduler to
find the nearest wakeup time by looking at the first thread
in the queue instead of searching the entire queue.
o Removed file descriptor locking for select/poll routines. An
application should not rely on the threads library for providing
this locking; if necessary, the application should use mutexes
to protect selecting/polling of file descriptors.
o Retrieve and use the kernel clock rate/resolution at startup
instead of hardcoding the clock resolution to 10 msec (tested
with kernel running at 1000 HZ).
o All queues have been changed to use queue.h macros. These
include the queues of all threads, dead threads, and threads
waiting for file descriptor locks.
o Added reinitialization of the GC mutex and condition variable
after a fork. Also prevented reallocation of the ready queue
after a fork.
o Prevented the wrapped close routine from closing the thread
kernel pipes.
o Initialized file descriptor table for stdio entries at thread
init.
o Provided additional flags to indicate to what queues threads
belong.
o Moved TAILQ initialization for statically allocated mutex and
condition variables to after the spinlock.
o Added dispatching of signals to pthread_kill. Removing the
dispatching of signals from thread activation broke sigsuspend
when pthread_kill was used to send a signal to a thread.
o Temporarily set the state of a thread to PS_SUSPENDED when it
is first created and placed in the list of threads so that it
will not be accidentally scheduled before becoming a member
of one of the scheduling queues.
o Change the signal handler to queue signals to the thread kernel
pipe if the scheduling queues are protected. When scheduling
queues are unprotected, signals are then dequeued and handled.
o Ensured that all installed signal handlers block the scheduling
signal and that the scheduling signal handler blocks all
other signals. This ensures that the signal handler is only
interruptible for and by non-scheduling signals. An atomic
lock is used to decide which instance of the signal handler
will handle pending signals.
o Removed _lock_thread_list and _unlock_thread_list as they are
no longer used to protect the thread list.
o Added missing RCS IDs to modified files.
o Added checks for appropriate queue membership and activity when
adding, removing, and searching the scheduling queues. These
checks add very little overhead and are enabled when compiled
with _PTHREADS_INVARIANTS defined. Suggested and implemented
by Tor Egge with some modification by me.
o Close a race condition in uthread_close. (Tor Egge)
o Protect the scheduling queues while modifying them in
pthread_cond_signal and _thread_fd_unlock. (Tor Egge)
o Ensure that when a thread gets a mutex, the mutex is on that
threads list of owned mutexes. (Tor Egge)
o Set the kernel-in-scheduler flag in _thread_kern_sched_state
and _thread_kern_sched_state_unlock to prevent a scheduling
signal from calling the scheduler again. (Tor Egge)
o Don't use TAILQ_FOREACH macro while searching the waiting
queue for threads in a sigwait state, because a change of
state destroys the TAILQ link. It is actually safe to do
so, though, because once a sigwaiting thread is found, the
loop ends and the function returns. (Tor Egge)
o When dispatching signals to threads, make the thread inherit
the signal deferral flag of the currently running thread.
(Tor Egge)
Submitted by: Daniel Eischen <eischen@vigrid.com> and
Tor Egge <Tor.Egge@fast.no>
1999-06-20 08:28:48 +00:00
|
|
|
pthread->flags |= PTHREAD_FLAGS_IN_CONDQ;
|
2000-10-13 22:12:32 +00:00
|
|
|
pthread->data.cond = cond;
|
1999-03-23 05:07:56 +00:00
|
|
|
}
|
1996-01-22 00:23:58 +00:00
|
|
|
#endif
|