1996-01-22 00:23:58 +00:00
|
|
|
/*
|
1998-04-29 10:59:34 +01:00
|
|
|
* Copyright (c) 1995-1998 John Birrell <jb@cimlogic.com.au>
|
1996-01-22 00:23:58 +00:00
|
|
|
* 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
|
|
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
|
|
|
|
* 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.
|
|
|
|
*
|
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 09:28:48 +01:00
|
|
|
* $Id: uthread_fd.c,v 1.10 1999/03/23 05:07:55 jb Exp $
|
1997-04-01 23:49:58 +01:00
|
|
|
*
|
1996-01-22 00:23:58 +00:00
|
|
|
*/
|
|
|
|
#include <errno.h>
|
1996-08-20 09:22:01 +01:00
|
|
|
#include <fcntl.h>
|
1996-01-22 00:23:58 +00:00
|
|
|
#include <stdlib.h>
|
1998-06-10 00:16:53 +01:00
|
|
|
#include <string.h>
|
1996-01-22 00:23:58 +00:00
|
|
|
#ifdef _THREAD_SAFE
|
|
|
|
#include <pthread.h>
|
|
|
|
#include "pthread_private.h"
|
|
|
|
|
1998-04-29 10:59:34 +01:00
|
|
|
/* Static variables: */
|
1998-06-10 00:16:53 +01:00
|
|
|
static spinlock_t fd_table_lock = _SPINLOCK_INITIALIZER;
|
1998-04-29 10:59:34 +01:00
|
|
|
|
1998-02-13 01:27:34 +00:00
|
|
|
/*
|
|
|
|
* This function *must* return -1 and set the thread specific errno
|
|
|
|
* as a system call. This is because the error return from this
|
|
|
|
* function is propagated directly back from thread-wrapped system
|
|
|
|
* calls.
|
|
|
|
*/
|
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
int
|
|
|
|
_thread_fd_table_init(int fd)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
1998-06-10 00:16:53 +01:00
|
|
|
struct fd_table_entry *entry;
|
1998-09-13 16:33:42 +01:00
|
|
|
int saved_errno;
|
1996-01-22 00:23:58 +00:00
|
|
|
|
|
|
|
/* Check if the file descriptor is out of range: */
|
1998-02-13 01:27:34 +00:00
|
|
|
if (fd < 0 || fd >= _thread_dtablesize) {
|
1996-01-22 00:23:58 +00:00
|
|
|
/* Return a bad file descriptor error: */
|
1998-02-13 01:27:34 +00:00
|
|
|
errno = EBADF;
|
|
|
|
ret = -1;
|
|
|
|
}
|
1997-02-05 23:26:09 +00:00
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/*
|
|
|
|
* Check if memory has already been allocated for this file
|
|
|
|
* descriptor:
|
|
|
|
*/
|
|
|
|
else if (_thread_fd_table[fd] != NULL) {
|
1997-02-05 23:26:09 +00:00
|
|
|
/* Memory has already been allocated. */
|
1998-06-10 00:16:53 +01:00
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/* Allocate memory for the file descriptor table entry: */
|
1998-06-10 00:16:53 +01:00
|
|
|
} else if ((entry = (struct fd_table_entry *)
|
1998-02-13 01:27:34 +00:00
|
|
|
malloc(sizeof(struct fd_table_entry))) == NULL) {
|
1998-06-10 00:16:53 +01:00
|
|
|
/* Return an insufficient memory error: */
|
|
|
|
errno = ENOMEM;
|
1998-02-13 01:27:34 +00:00
|
|
|
ret = -1;
|
1998-04-29 10:59:34 +01:00
|
|
|
} else {
|
1996-01-22 00:23:58 +00:00
|
|
|
/* Initialise the file locks: */
|
1998-06-10 00:16:53 +01:00
|
|
|
memset(&entry->lock, 0, sizeof(entry->lock));
|
|
|
|
entry->r_owner = NULL;
|
|
|
|
entry->w_owner = NULL;
|
|
|
|
entry->r_fname = NULL;
|
|
|
|
entry->w_fname = NULL;
|
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 09:28:48 +01:00
|
|
|
entry->r_lineno = 0;
|
|
|
|
entry->w_lineno = 0;
|
|
|
|
entry->r_lockcount = 0;
|
|
|
|
entry->w_lockcount = 0;
|
1996-01-22 00:23:58 +00:00
|
|
|
|
|
|
|
/* Initialise the read/write queues: */
|
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 09:28:48 +01:00
|
|
|
TAILQ_INIT(&entry->r_queue);
|
|
|
|
TAILQ_INIT(&entry->w_queue);
|
1996-08-20 09:22:01 +01:00
|
|
|
|
|
|
|
/* Get the flags for the file: */
|
1998-06-10 00:16:53 +01:00
|
|
|
if (fd >= 3 && (entry->flags =
|
1998-02-13 01:27:34 +00:00
|
|
|
_thread_sys_fcntl(fd, F_GETFL, 0)) == -1) {
|
|
|
|
ret = -1;
|
|
|
|
}
|
1997-02-05 23:26:09 +00:00
|
|
|
else {
|
|
|
|
/* Check if a stdio descriptor: */
|
|
|
|
if (fd < 3)
|
|
|
|
/*
|
|
|
|
* Use the stdio flags read by
|
|
|
|
* _pthread_init() to avoid
|
|
|
|
* mistaking the non-blocking
|
|
|
|
* flag that, when set on one
|
|
|
|
* stdio fd, is set on all stdio
|
|
|
|
* fds.
|
|
|
|
*/
|
1998-06-10 00:16:53 +01:00
|
|
|
entry->flags = _pthread_stdio_flags[fd];
|
1996-08-20 09:22:01 +01:00
|
|
|
|
1998-05-27 01:41:22 +01:00
|
|
|
/*
|
|
|
|
* Make the file descriptor non-blocking.
|
|
|
|
* This might fail if the device driver does
|
|
|
|
* not support non-blocking calls, or if the
|
|
|
|
* driver is naturally non-blocking.
|
|
|
|
*/
|
1998-09-13 16:33:42 +01:00
|
|
|
saved_errno = errno;
|
1998-05-27 01:41:22 +01:00
|
|
|
_thread_sys_fcntl(fd, F_SETFL,
|
1998-06-10 00:16:53 +01:00
|
|
|
entry->flags | O_NONBLOCK);
|
1998-09-13 16:33:42 +01:00
|
|
|
errno = saved_errno;
|
1998-06-10 00:16:53 +01:00
|
|
|
|
|
|
|
/* Lock the file descriptor table: */
|
|
|
|
_SPINLOCK(&fd_table_lock);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if another thread allocated the
|
|
|
|
* file descriptor entry while this thread
|
|
|
|
* was doing the same thing. The table wasn't
|
|
|
|
* kept locked during this operation because
|
|
|
|
* it has the potential to recurse.
|
|
|
|
*/
|
|
|
|
if (_thread_fd_table[fd] == NULL) {
|
|
|
|
/* This thread wins: */
|
|
|
|
_thread_fd_table[fd] = entry;
|
|
|
|
entry = NULL;
|
|
|
|
}
|
1996-08-20 09:22:01 +01:00
|
|
|
|
1998-06-10 00:16:53 +01:00
|
|
|
/* Unlock the file descriptor table: */
|
|
|
|
_SPINUNLOCK(&fd_table_lock);
|
1996-08-20 09:22:01 +01:00
|
|
|
}
|
1996-01-22 00:23:58 +00:00
|
|
|
|
1998-06-10 00:16:53 +01:00
|
|
|
/*
|
|
|
|
* Check if another thread initialised the table entry
|
|
|
|
* before this one could:
|
|
|
|
*/
|
|
|
|
if (entry != NULL)
|
|
|
|
/*
|
|
|
|
* Throw away the table entry that this thread
|
|
|
|
* prepared. The other thread wins.
|
|
|
|
*/
|
|
|
|
free(entry);
|
|
|
|
}
|
1996-01-22 00:23:58 +00:00
|
|
|
|
|
|
|
/* Return the completion status: */
|
|
|
|
return (ret);
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
_thread_fd_unlock(int fd, int lock_type)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check that the file descriptor table is initialised for this
|
|
|
|
* entry:
|
|
|
|
*/
|
1998-04-29 10:59:34 +01:00
|
|
|
if ((ret = _thread_fd_table_init(fd)) == 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 09:28:48 +01:00
|
|
|
/*
|
|
|
|
* Defer signals to protect the scheduling queues from
|
|
|
|
* access by the signal handler:
|
|
|
|
*/
|
|
|
|
_thread_kern_sig_defer();
|
|
|
|
|
1998-04-29 10:59:34 +01:00
|
|
|
/*
|
|
|
|
* Lock the file descriptor table entry to prevent
|
|
|
|
* other threads for clashing with the current
|
|
|
|
* thread's accesses:
|
|
|
|
*/
|
1998-06-10 00:16:53 +01:00
|
|
|
_SPINLOCK(&_thread_fd_table[fd]->lock);
|
1998-04-29 10:59:34 +01:00
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/* Check if the running thread owns the read lock: */
|
|
|
|
if (_thread_fd_table[fd]->r_owner == _thread_run) {
|
|
|
|
/* Check the file descriptor and lock types: */
|
|
|
|
if (lock_type == FD_READ || lock_type == FD_RDWR) {
|
|
|
|
/*
|
|
|
|
* Decrement the read lock count for the
|
|
|
|
* running thread:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->r_lockcount--;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if the running thread still has read
|
|
|
|
* locks on this file descriptor:
|
|
|
|
*/
|
|
|
|
if (_thread_fd_table[fd]->r_lockcount != 0) {
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Get the next thread in the queue for a
|
|
|
|
* read lock on this file descriptor:
|
|
|
|
*/
|
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 09:28:48 +01:00
|
|
|
else if ((_thread_fd_table[fd]->r_owner = TAILQ_FIRST(&_thread_fd_table[fd]->r_queue)) == NULL) {
|
1996-01-22 00:23:58 +00:00
|
|
|
} else {
|
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 09:28:48 +01:00
|
|
|
/* Remove this thread from the queue: */
|
|
|
|
TAILQ_REMOVE(&_thread_fd_table[fd]->r_queue,
|
|
|
|
_thread_fd_table[fd]->r_owner, qe);
|
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/*
|
|
|
|
* Set the state of the new owner of
|
1999-03-23 05:07:56 +00:00
|
|
|
* the thread to running:
|
1996-01-22 00:23:58 +00:00
|
|
|
*/
|
1997-02-05 23:26:09 +00:00
|
|
|
PTHREAD_NEW_STATE(_thread_fd_table[fd]->r_owner,PS_RUNNING);
|
1996-01-22 00:23:58 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the number of read locks.
|
|
|
|
* This will be incremented by the
|
|
|
|
* new owner of the lock when it sees
|
|
|
|
* that it has the lock.
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->r_lockcount = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
/* Check if the running thread owns the write lock: */
|
|
|
|
if (_thread_fd_table[fd]->w_owner == _thread_run) {
|
|
|
|
/* Check the file descriptor and lock types: */
|
|
|
|
if (lock_type == FD_WRITE || lock_type == FD_RDWR) {
|
|
|
|
/*
|
|
|
|
* Decrement the write lock count for the
|
|
|
|
* running thread:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->w_lockcount--;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if the running thread still has
|
|
|
|
* write locks on this file descriptor:
|
|
|
|
*/
|
|
|
|
if (_thread_fd_table[fd]->w_lockcount != 0) {
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Get the next thread in the queue for a
|
|
|
|
* write lock on this file descriptor:
|
|
|
|
*/
|
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 09:28:48 +01:00
|
|
|
else if ((_thread_fd_table[fd]->w_owner = TAILQ_FIRST(&_thread_fd_table[fd]->w_queue)) == NULL) {
|
1996-01-22 00:23:58 +00:00
|
|
|
} else {
|
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 09:28:48 +01:00
|
|
|
/* Remove this thread from the queue: */
|
|
|
|
TAILQ_REMOVE(&_thread_fd_table[fd]->w_queue,
|
|
|
|
_thread_fd_table[fd]->w_owner, qe);
|
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/*
|
|
|
|
* Set the state of the new owner of
|
|
|
|
* the thread to running:
|
|
|
|
*/
|
1997-02-05 23:26:09 +00:00
|
|
|
PTHREAD_NEW_STATE(_thread_fd_table[fd]->w_owner,PS_RUNNING);
|
1996-01-22 00:23:58 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the number of write locks.
|
|
|
|
* This will be incremented by the
|
|
|
|
* new owner of the lock when it
|
1998-06-10 00:16:53 +01:00
|
|
|
* sees that it has the lock.
|
1996-01-22 00:23:58 +00:00
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->w_lockcount = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
1998-04-29 10:59:34 +01:00
|
|
|
/* Unlock the file descriptor table entry: */
|
1998-06-10 00:16:53 +01:00
|
|
|
_SPINUNLOCK(&_thread_fd_table[fd]->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 09:28:48 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Undefer and handle pending signals, yielding if
|
|
|
|
* necessary:
|
|
|
|
*/
|
|
|
|
_thread_kern_sig_undefer();
|
1998-06-10 00:16:53 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Nothing to return. */
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
|
|
|
_thread_fd_lock(int fd, int lock_type, struct timespec * timeout)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check that the file descriptor table is initialised for this
|
|
|
|
* entry:
|
|
|
|
*/
|
|
|
|
if ((ret = _thread_fd_table_init(fd)) == 0) {
|
|
|
|
/*
|
|
|
|
* Lock the file descriptor table entry to prevent
|
|
|
|
* other threads for clashing with the current
|
|
|
|
* thread's accesses:
|
|
|
|
*/
|
|
|
|
_SPINLOCK(&_thread_fd_table[fd]->lock);
|
|
|
|
|
|
|
|
/* Check the file descriptor and lock types: */
|
|
|
|
if (lock_type == FD_READ || lock_type == FD_RDWR) {
|
|
|
|
/*
|
|
|
|
* Enter a loop to wait for the file descriptor to be
|
|
|
|
* locked for read for the current thread:
|
|
|
|
*/
|
|
|
|
while (_thread_fd_table[fd]->r_owner != _thread_run) {
|
|
|
|
/*
|
|
|
|
* Check if the file descriptor is locked by
|
|
|
|
* another thread:
|
|
|
|
*/
|
|
|
|
if (_thread_fd_table[fd]->r_owner != NULL) {
|
|
|
|
/*
|
|
|
|
* Another thread has locked the file
|
|
|
|
* descriptor for read, so join the
|
|
|
|
* queue of threads waiting for a
|
|
|
|
* read lock on this file descriptor:
|
|
|
|
*/
|
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 09:28:48 +01:00
|
|
|
TAILQ_INSERT_TAIL(&_thread_fd_table[fd]->r_queue, _thread_run, qe);
|
1998-06-10 00:16:53 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Save the file descriptor details
|
|
|
|
* in the thread structure for the
|
|
|
|
* running thread:
|
|
|
|
*/
|
|
|
|
_thread_run->data.fd.fd = fd;
|
|
|
|
|
|
|
|
/* Set the timeout: */
|
|
|
|
_thread_kern_set_timeout(timeout);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Unlock the file descriptor
|
|
|
|
* table entry:
|
|
|
|
*/
|
|
|
|
_SPINUNLOCK(&_thread_fd_table[fd]->lock);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Schedule this thread to wait on
|
|
|
|
* the read lock. It will only be
|
|
|
|
* woken when it becomes the next in
|
|
|
|
* the queue and is granted access
|
|
|
|
* to the lock by the thread
|
|
|
|
* that is unlocking the file
|
|
|
|
* descriptor.
|
|
|
|
*/
|
|
|
|
_thread_kern_sched_state(PS_FDLR_WAIT, __FILE__, __LINE__);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lock the file descriptor
|
|
|
|
* table entry again:
|
|
|
|
*/
|
|
|
|
_SPINLOCK(&_thread_fd_table[fd]->lock);
|
|
|
|
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* The running thread now owns the
|
|
|
|
* read lock on this file descriptor:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->r_owner = _thread_run;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the number of read locks for
|
|
|
|
* this file descriptor:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->r_lockcount = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Increment the read lock count: */
|
|
|
|
_thread_fd_table[fd]->r_lockcount++;
|
|
|
|
}
|
1996-01-22 00:23:58 +00:00
|
|
|
|
1998-06-10 00:16:53 +01:00
|
|
|
/* Check the file descriptor and lock types: */
|
|
|
|
if (lock_type == FD_WRITE || lock_type == FD_RDWR) {
|
|
|
|
/*
|
|
|
|
* Enter a loop to wait for the file descriptor to be
|
|
|
|
* locked for write for the current thread:
|
|
|
|
*/
|
|
|
|
while (_thread_fd_table[fd]->w_owner != _thread_run) {
|
|
|
|
/*
|
|
|
|
* Check if the file descriptor is locked by
|
|
|
|
* another thread:
|
|
|
|
*/
|
|
|
|
if (_thread_fd_table[fd]->w_owner != NULL) {
|
|
|
|
/*
|
|
|
|
* Another thread has locked the file
|
|
|
|
* descriptor for write, so join the
|
|
|
|
* queue of threads waiting for a
|
|
|
|
* write lock on this file
|
|
|
|
* descriptor:
|
|
|
|
*/
|
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 09:28:48 +01:00
|
|
|
TAILQ_INSERT_TAIL(&_thread_fd_table[fd]->w_queue, _thread_run, qe);
|
1998-06-10 00:16:53 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Save the file descriptor details
|
|
|
|
* in the thread structure for the
|
|
|
|
* running thread:
|
|
|
|
*/
|
|
|
|
_thread_run->data.fd.fd = fd;
|
|
|
|
|
|
|
|
/* Set the timeout: */
|
|
|
|
_thread_kern_set_timeout(timeout);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Unlock the file descriptor
|
|
|
|
* table entry:
|
|
|
|
*/
|
|
|
|
_SPINUNLOCK(&_thread_fd_table[fd]->lock);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Schedule this thread to wait on
|
|
|
|
* the write lock. It will only be
|
|
|
|
* woken when it becomes the next in
|
|
|
|
* the queue and is granted access to
|
|
|
|
* the lock by the thread that is
|
|
|
|
* unlocking the file descriptor.
|
|
|
|
*/
|
|
|
|
_thread_kern_sched_state(PS_FDLW_WAIT, __FILE__, __LINE__);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lock the file descriptor
|
|
|
|
* table entry again:
|
|
|
|
*/
|
|
|
|
_SPINLOCK(&_thread_fd_table[fd]->lock);
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* The running thread now owns the
|
|
|
|
* write lock on this file
|
|
|
|
* descriptor:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->w_owner = _thread_run;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the number of write locks
|
|
|
|
* for this file descriptor:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->w_lockcount = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Increment the write lock count: */
|
|
|
|
_thread_fd_table[fd]->w_lockcount++;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Unlock the file descriptor table entry: */
|
|
|
|
_SPINUNLOCK(&_thread_fd_table[fd]->lock);
|
1998-04-29 10:59:34 +01:00
|
|
|
}
|
1998-06-10 00:16:53 +01:00
|
|
|
|
|
|
|
/* Return the completion status: */
|
|
|
|
return (ret);
|
|
|
|
}
|
|
|
|
|
|
|
|
void
|
|
|
|
_thread_fd_unlock_debug(int fd, int lock_type, char *fname, int lineno)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check that the file descriptor table is initialised for this
|
|
|
|
* entry:
|
|
|
|
*/
|
|
|
|
if ((ret = _thread_fd_table_init(fd)) == 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 09:28:48 +01:00
|
|
|
/*
|
|
|
|
* Defer signals to protect the scheduling queues from
|
|
|
|
* access by the signal handler:
|
|
|
|
*/
|
|
|
|
_thread_kern_sig_defer();
|
|
|
|
|
1998-06-10 00:16:53 +01:00
|
|
|
/*
|
|
|
|
* Lock the file descriptor table entry to prevent
|
|
|
|
* other threads for clashing with the current
|
|
|
|
* thread's accesses:
|
|
|
|
*/
|
|
|
|
_spinlock_debug(&_thread_fd_table[fd]->lock, fname, lineno);
|
|
|
|
|
|
|
|
/* Check if the running thread owns the read lock: */
|
|
|
|
if (_thread_fd_table[fd]->r_owner == _thread_run) {
|
|
|
|
/* Check the file descriptor and lock types: */
|
|
|
|
if (lock_type == FD_READ || lock_type == FD_RDWR) {
|
|
|
|
/*
|
|
|
|
* Decrement the read lock count for the
|
|
|
|
* running thread:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->r_lockcount--;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if the running thread still has read
|
|
|
|
* locks on this file descriptor:
|
|
|
|
*/
|
|
|
|
if (_thread_fd_table[fd]->r_lockcount != 0) {
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Get the next thread in the queue for a
|
|
|
|
* read lock on this file descriptor:
|
|
|
|
*/
|
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 09:28:48 +01:00
|
|
|
else if ((_thread_fd_table[fd]->r_owner = TAILQ_FIRST(&_thread_fd_table[fd]->r_queue)) == NULL) {
|
1998-06-10 00:16:53 +01:00
|
|
|
} else {
|
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 09:28:48 +01:00
|
|
|
/* Remove this thread from the queue: */
|
|
|
|
TAILQ_REMOVE(&_thread_fd_table[fd]->r_queue,
|
|
|
|
_thread_fd_table[fd]->r_owner, qe);
|
|
|
|
|
1998-06-10 00:16:53 +01:00
|
|
|
/*
|
|
|
|
* Set the state of the new owner of
|
|
|
|
* the thread to running:
|
|
|
|
*/
|
|
|
|
PTHREAD_NEW_STATE(_thread_fd_table[fd]->r_owner,PS_RUNNING);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the number of read locks.
|
|
|
|
* This will be incremented by the
|
|
|
|
* new owner of the lock when it sees
|
|
|
|
* that it has the lock.
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->r_lockcount = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
/* Check if the running thread owns the write lock: */
|
|
|
|
if (_thread_fd_table[fd]->w_owner == _thread_run) {
|
|
|
|
/* Check the file descriptor and lock types: */
|
|
|
|
if (lock_type == FD_WRITE || lock_type == FD_RDWR) {
|
|
|
|
/*
|
|
|
|
* Decrement the write lock count for the
|
|
|
|
* running thread:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->w_lockcount--;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if the running thread still has
|
|
|
|
* write locks on this file descriptor:
|
|
|
|
*/
|
|
|
|
if (_thread_fd_table[fd]->w_lockcount != 0) {
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Get the next thread in the queue for a
|
|
|
|
* write lock on this file descriptor:
|
|
|
|
*/
|
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 09:28:48 +01:00
|
|
|
else if ((_thread_fd_table[fd]->w_owner = TAILQ_FIRST(&_thread_fd_table[fd]->w_queue)) == NULL) {
|
1998-06-10 00:16:53 +01:00
|
|
|
} else {
|
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 09:28:48 +01:00
|
|
|
/* Remove this thread from the queue: */
|
|
|
|
TAILQ_REMOVE(&_thread_fd_table[fd]->w_queue,
|
|
|
|
_thread_fd_table[fd]->w_owner, qe);
|
|
|
|
|
1998-06-10 00:16:53 +01:00
|
|
|
/*
|
|
|
|
* Set the state of the new owner of
|
|
|
|
* the thread to running:
|
|
|
|
*/
|
|
|
|
PTHREAD_NEW_STATE(_thread_fd_table[fd]->w_owner,PS_RUNNING);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the number of write locks.
|
|
|
|
* This will be incremented by the
|
|
|
|
* new owner of the lock when it
|
|
|
|
* sees that it has the lock.
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->w_lockcount = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Unlock the file descriptor table entry: */
|
|
|
|
_SPINUNLOCK(&_thread_fd_table[fd]->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 09:28:48 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Undefer and handle pending signals, yielding if
|
|
|
|
* necessary.
|
|
|
|
*/
|
|
|
|
_thread_kern_sig_undefer();
|
1998-06-10 00:16:53 +01:00
|
|
|
}
|
|
|
|
|
1998-04-29 10:59:34 +01:00
|
|
|
/* Nothing to return. */
|
1996-01-22 00:23:58 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
int
|
1998-06-10 00:16:53 +01:00
|
|
|
_thread_fd_lock_debug(int fd, int lock_type, struct timespec * timeout,
|
1996-01-22 00:23:58 +00:00
|
|
|
char *fname, int lineno)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check that the file descriptor table is initialised for this
|
|
|
|
* entry:
|
|
|
|
*/
|
1998-04-29 10:59:34 +01:00
|
|
|
if ((ret = _thread_fd_table_init(fd)) == 0) {
|
|
|
|
/*
|
|
|
|
* Lock the file descriptor table entry to prevent
|
|
|
|
* other threads for clashing with the current
|
|
|
|
* thread's accesses:
|
|
|
|
*/
|
1998-06-10 00:16:53 +01:00
|
|
|
_spinlock_debug(&_thread_fd_table[fd]->lock, fname, lineno);
|
1998-04-29 10:59:34 +01:00
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/* Check the file descriptor and lock types: */
|
|
|
|
if (lock_type == FD_READ || lock_type == FD_RDWR) {
|
|
|
|
/*
|
|
|
|
* Enter a loop to wait for the file descriptor to be
|
|
|
|
* locked for read for the current thread:
|
|
|
|
*/
|
|
|
|
while (_thread_fd_table[fd]->r_owner != _thread_run) {
|
|
|
|
/*
|
|
|
|
* Check if the file descriptor is locked by
|
|
|
|
* another thread:
|
|
|
|
*/
|
|
|
|
if (_thread_fd_table[fd]->r_owner != NULL) {
|
|
|
|
/*
|
|
|
|
* Another thread has locked the file
|
|
|
|
* descriptor for read, so join the
|
|
|
|
* queue of threads waiting for a
|
|
|
|
* read lock on this file descriptor:
|
|
|
|
*/
|
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 09:28:48 +01:00
|
|
|
TAILQ_INSERT_TAIL(&_thread_fd_table[fd]->r_queue, _thread_run, qe);
|
1996-01-22 00:23:58 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Save the file descriptor details
|
|
|
|
* in the thread structure for the
|
|
|
|
* running thread:
|
|
|
|
*/
|
|
|
|
_thread_run->data.fd.fd = fd;
|
|
|
|
_thread_run->data.fd.branch = lineno;
|
|
|
|
_thread_run->data.fd.fname = fname;
|
|
|
|
|
|
|
|
/* Set the timeout: */
|
|
|
|
_thread_kern_set_timeout(timeout);
|
|
|
|
|
1998-04-29 10:59:34 +01:00
|
|
|
/*
|
|
|
|
* Unlock the file descriptor
|
|
|
|
* table entry:
|
|
|
|
*/
|
1998-06-10 00:16:53 +01:00
|
|
|
_SPINUNLOCK(&_thread_fd_table[fd]->lock);
|
1998-04-29 10:59:34 +01:00
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/*
|
|
|
|
* Schedule this thread to wait on
|
|
|
|
* the read lock. It will only be
|
|
|
|
* woken when it becomes the next in
|
|
|
|
* the queue and is granted access
|
|
|
|
* to the lock by the thread
|
|
|
|
* that is unlocking the file
|
|
|
|
* descriptor.
|
|
|
|
*/
|
|
|
|
_thread_kern_sched_state(PS_FDLR_WAIT, __FILE__, __LINE__);
|
|
|
|
|
|
|
|
/*
|
1998-04-29 10:59:34 +01:00
|
|
|
* Lock the file descriptor
|
|
|
|
* table entry again:
|
1996-01-22 00:23:58 +00:00
|
|
|
*/
|
1998-06-10 00:16:53 +01:00
|
|
|
_SPINLOCK(&_thread_fd_table[fd]->lock);
|
1998-04-29 10:59:34 +01:00
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* The running thread now owns the
|
|
|
|
* read lock on this file descriptor:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->r_owner = _thread_run;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the number of read locks for
|
|
|
|
* this file descriptor:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->r_lockcount = 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Save the source file details for
|
|
|
|
* debugging:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->r_fname = fname;
|
|
|
|
_thread_fd_table[fd]->r_lineno = lineno;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Increment the read lock count: */
|
|
|
|
_thread_fd_table[fd]->r_lockcount++;
|
|
|
|
}
|
1998-06-10 00:16:53 +01:00
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/* Check the file descriptor and lock types: */
|
|
|
|
if (lock_type == FD_WRITE || lock_type == FD_RDWR) {
|
|
|
|
/*
|
|
|
|
* Enter a loop to wait for the file descriptor to be
|
|
|
|
* locked for write for the current thread:
|
|
|
|
*/
|
|
|
|
while (_thread_fd_table[fd]->w_owner != _thread_run) {
|
|
|
|
/*
|
|
|
|
* Check if the file descriptor is locked by
|
|
|
|
* another thread:
|
|
|
|
*/
|
|
|
|
if (_thread_fd_table[fd]->w_owner != NULL) {
|
|
|
|
/*
|
|
|
|
* Another thread has locked the file
|
|
|
|
* descriptor for write, so join the
|
|
|
|
* queue of threads waiting for a
|
|
|
|
* write lock on this file
|
|
|
|
* descriptor:
|
|
|
|
*/
|
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 09:28:48 +01:00
|
|
|
TAILQ_INSERT_TAIL(&_thread_fd_table[fd]->w_queue, _thread_run, qe);
|
1996-01-22 00:23:58 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Save the file descriptor details
|
|
|
|
* in the thread structure for the
|
|
|
|
* running thread:
|
|
|
|
*/
|
|
|
|
_thread_run->data.fd.fd = fd;
|
|
|
|
_thread_run->data.fd.branch = lineno;
|
|
|
|
_thread_run->data.fd.fname = fname;
|
|
|
|
|
|
|
|
/* Set the timeout: */
|
|
|
|
_thread_kern_set_timeout(timeout);
|
|
|
|
|
1998-04-29 10:59:34 +01:00
|
|
|
/*
|
|
|
|
* Unlock the file descriptor
|
|
|
|
* table entry:
|
|
|
|
*/
|
1998-06-10 00:16:53 +01:00
|
|
|
_SPINUNLOCK(&_thread_fd_table[fd]->lock);
|
1998-04-29 10:59:34 +01:00
|
|
|
|
1996-01-22 00:23:58 +00:00
|
|
|
/*
|
|
|
|
* Schedule this thread to wait on
|
|
|
|
* the write lock. It will only be
|
|
|
|
* woken when it becomes the next in
|
|
|
|
* the queue and is granted access to
|
|
|
|
* the lock by the thread that is
|
|
|
|
* unlocking the file descriptor.
|
|
|
|
*/
|
|
|
|
_thread_kern_sched_state(PS_FDLW_WAIT, __FILE__, __LINE__);
|
|
|
|
|
|
|
|
/*
|
1998-04-29 10:59:34 +01:00
|
|
|
* Lock the file descriptor
|
|
|
|
* table entry again:
|
1996-01-22 00:23:58 +00:00
|
|
|
*/
|
1998-06-10 00:16:53 +01:00
|
|
|
_SPINLOCK(&_thread_fd_table[fd]->lock);
|
1996-01-22 00:23:58 +00:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* The running thread now owns the
|
|
|
|
* write lock on this file
|
|
|
|
* descriptor:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->w_owner = _thread_run;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reset the number of write locks
|
|
|
|
* for this file descriptor:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->w_lockcount = 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Save the source file details for
|
|
|
|
* debugging:
|
|
|
|
*/
|
|
|
|
_thread_fd_table[fd]->w_fname = fname;
|
|
|
|
_thread_fd_table[fd]->w_lineno = lineno;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Increment the write lock count: */
|
|
|
|
_thread_fd_table[fd]->w_lockcount++;
|
|
|
|
}
|
|
|
|
|
1998-04-29 10:59:34 +01:00
|
|
|
/* Unlock the file descriptor table entry: */
|
1998-06-10 00:16:53 +01:00
|
|
|
_SPINUNLOCK(&_thread_fd_table[fd]->lock);
|
1998-04-29 10:59:34 +01:00
|
|
|
}
|
1996-01-22 00:23:58 +00:00
|
|
|
|
|
|
|
/* Return the completion status: */
|
|
|
|
return (ret);
|
|
|
|
}
|
|
|
|
#endif
|