d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
3952015f40
freebsd-dev/lib/libkse/sys/lock.c

345 lines
9.1 KiB
C
Raw Normal View History

Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
/*-
Sorry folks; I accidentally committed a patch from what I was working on a couple of days ago. This should be the most recent changes. Noticed by: davidxu
2003-04-18 07:09:43 +00:00
* Copyright (c) 2001, 2003 Daniel Eischen <deischen@freebsd.org>.
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +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.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHORS 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 AUTHOR 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.
*
* $FreeBSD$
*/
#include <sys/types.h>
#include <machine/atomic.h>
#include <assert.h>
#include <stdlib.h>
#include "atomic_ops.h"
#include "lock.h"
Only generate code for _LCK_ASSERT if _LCK_DEBUG is defined.
2003-10-02 03:24:26 +00:00
#ifdef _LOCK_DEBUG
#define LCK_ASSERT(e) assert(e)
#else
#define LCK_ASSERT(e)
#endif
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
#define MAX_SPINS 500
void
_lock_destroy(struct lock *lck)
{
if ((lck != NULL) && (lck->l_head != NULL)) {
free(lck->l_head);
lck->l_head = NULL;
lck->l_tail = NULL;
}
}
int
_lock_init(struct lock *lck, enum lock_type ltype,
lock_handler_t *waitfunc, lock_handler_t *wakeupfunc)
{
if (lck == NULL)
return (-1);
else if ((lck->l_head = malloc(sizeof(struct lockreq))) == NULL)
return (-1);
else {
lck->l_type = ltype;
lck->l_wait = waitfunc;
lck->l_wakeup = wakeupfunc;
lck->l_head->lr_locked = 0;
lck->l_head->lr_watcher = NULL;
lck->l_head->lr_owner = NULL;
Add a method of yielding the current thread with the scheduler lock held (_thr_sched_switch_unlocked()) and use this to avoid dropping the scheduler lock and having the scheduler retake the same lock again. Add a better way of detecting if a low-level lock is in use. When switching out a thread due to blocking in the UTS, don't switch to the KSE's scheduler stack only to switch back to another thread. If possible switch to the new thread directly from the old thread and avoid the overhead of the extra context switch. Check for pending signals on a thread when entering the scheduler and add them to the threads signal frame. This includes some other minor signal fixes. Most of this was a joint effor between davidxu and myself. Reviewed by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-16 19:58:30 +00:00
lck->l_head->lr_active = 1;
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
lck->l_tail = lck->l_head;
}
return (0);
}
int
Add the ability to reinitialize libpthread's internal FIFO-queueing locks. Reviewed by: davidxu
2003-11-04 20:01:38 +00:00
_lock_reinit(struct lock *lck, enum lock_type ltype,
lock_handler_t *waitfunc, lock_handler_t *wakeupfunc)
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
{
Add the ability to reinitialize libpthread's internal FIFO-queueing locks. Reviewed by: davidxu
2003-11-04 20:01:38 +00:00
if (lck == NULL)
return (-1);
else if (lck->l_head == NULL)
return (_lock_init(lck, ltype, waitfunc, wakeupfunc));
else {
lck->l_head->lr_locked = 0;
lck->l_head->lr_watcher = NULL;
lck->l_head->lr_owner = NULL;
lck->l_head->lr_active = 1;
lck->l_tail = lck->l_head;
}
return (0);
}
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
Add the ability to reinitialize libpthread's internal FIFO-queueing locks. Reviewed by: davidxu
2003-11-04 20:01:38 +00:00
int
_lockuser_init(struct lockuser *lu, void *priv)
{
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
if (lu == NULL)
return (-1);
else if ((lu->lu_myreq == NULL) &&
((lu->lu_myreq = malloc(sizeof(struct lockreq))) == NULL))
return (-1);
else {
lu->lu_myreq->lr_locked = 1;
lu->lu_myreq->lr_watcher = NULL;
lu->lu_myreq->lr_owner = lu;
Add a method of yielding the current thread with the scheduler lock held (_thr_sched_switch_unlocked()) and use this to avoid dropping the scheduler lock and having the scheduler retake the same lock again. Add a better way of detecting if a low-level lock is in use. When switching out a thread due to blocking in the UTS, don't switch to the KSE's scheduler stack only to switch back to another thread. If possible switch to the new thread directly from the old thread and avoid the overhead of the extra context switch. Check for pending signals on a thread when entering the scheduler and add them to the threads signal frame. This includes some other minor signal fixes. Most of this was a joint effor between davidxu and myself. Reviewed by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-16 19:58:30 +00:00
lu->lu_myreq->lr_active = 0;
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
lu->lu_watchreq = NULL;
lu->lu_priority = 0;
lu->lu_private = priv;
lu->lu_private2 = NULL;
}
return (0);
}
Add the ability to reinitialize libpthread's internal FIFO-queueing locks. Reviewed by: davidxu
2003-11-04 20:01:38 +00:00
int
_lockuser_reinit(struct lockuser *lu, void *priv)
{
if (lu == NULL)
return (-1);
/*
* All lockusers keep their watch request and drop their
* own (lu_myreq) request. Their own request is either
* some other lockuser's watch request or is the head of
* the lock.
*/
lu->lu_myreq = lu->lu_watchreq;
if (lu->lu_myreq == NULL)
return (_lockuser_init(lu, priv));
else {
lu->lu_myreq->lr_locked = 1;
lu->lu_myreq->lr_watcher = NULL;
lu->lu_myreq->lr_owner = lu;
lu->lu_myreq->lr_active = 0;
lu->lu_watchreq = NULL;
lu->lu_priority = 0;
lu->lu_private = priv;
lu->lu_private2 = NULL;
}
return (0);
}
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
void
_lockuser_destroy(struct lockuser *lu)
{
if ((lu != NULL) && (lu->lu_myreq != NULL))
free(lu->lu_myreq);
}
/*
* Acquire a lock waiting (spin or sleep) for it to become available.
*/
void
_lock_acquire(struct lock *lck, struct lockuser *lu, int prio)
{
int i;
Don't assume sizeof(long) = sizeof(int) on x86; use int instead of long types for low-level locks. Add prototypes for some internal libc functions that are wrapped by the library as cancellation points. Add memory barriers to alpha atomic swap functions (submitted by davidxu). Requested by: bde
2003-09-03 17:56:26 +00:00
int lval;
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
/**
* XXX - We probably want to remove these checks to optimize
* performance. It is also a bug if any one of the
* checks fail, so it's probably better to just let it
* SEGV and fix it.
*/
#if 0
if (lck == NULL || lu == NULL || lck->l_head == NULL)
return;
#endif
Add the ability to reinitialize libpthread's internal FIFO-queueing locks. Reviewed by: davidxu
2003-11-04 20:01:38 +00:00
if ((lck->l_type & LCK_PRIORITY) != 0) {
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
LCK_ASSERT(lu->lu_myreq->lr_locked == 1);
LCK_ASSERT(lu->lu_myreq->lr_watcher == NULL);
LCK_ASSERT(lu->lu_myreq->lr_owner == lu);
LCK_ASSERT(lu->lu_watchreq == NULL);
lu->lu_priority = prio;
}
Add the ability to reinitialize libpthread's internal FIFO-queueing locks. Reviewed by: davidxu
2003-11-04 20:01:38 +00:00
/*
* Atomically swap the head of the lock request with
* this request.
*/
atomic_swap_ptr(&lck->l_head, lu->lu_myreq, &lu->lu_watchreq);
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
if (lu->lu_watchreq->lr_locked != 0) {
atomic_store_rel_ptr(&lu->lu_watchreq->lr_watcher, lu);
if ((lck->l_wait == NULL) ||
((lck->l_type & LCK_ADAPTIVE) == 0)) {
while (lu->lu_watchreq->lr_locked == 0)
; /* spin, then yield? */
} else {
/*
* Spin for a bit before invoking the wait function.
*
* We should be a little smarter here. If we're
* running on a single processor, then the lock
* owner got preempted and spinning will accomplish
* nothing but waste time. If we're running on
* multiple processors, the owner could be running
* on another CPU and we might acquire the lock if
* we spin for a bit.
*
* The other thing to keep in mind is that threads
* acquiring these locks are considered to be in
* critical regions; they will not be preempted by
* the _UTS_ until they release the lock. It is
* therefore safe to assume that if a lock can't
* be acquired, it is currently held by a thread
* running in another KSE.
*/
for (i = 0; i < MAX_SPINS; i++) {
if (lu->lu_watchreq->lr_locked == 0)
return;
Add a method of yielding the current thread with the scheduler lock held (_thr_sched_switch_unlocked()) and use this to avoid dropping the scheduler lock and having the scheduler retake the same lock again. Add a better way of detecting if a low-level lock is in use. When switching out a thread due to blocking in the UTS, don't switch to the KSE's scheduler stack only to switch back to another thread. If possible switch to the new thread directly from the old thread and avoid the overhead of the extra context switch. Check for pending signals on a thread when entering the scheduler and add them to the threads signal frame. This includes some other minor signal fixes. Most of this was a joint effor between davidxu and myself. Reviewed by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-16 19:58:30 +00:00
if (lu->lu_watchreq->lr_active == 0)
break;
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
}
Don't assume sizeof(long) = sizeof(int) on x86; use int instead of long types for low-level locks. Add prototypes for some internal libc functions that are wrapped by the library as cancellation points. Add memory barriers to alpha atomic swap functions (submitted by davidxu). Requested by: bde
2003-09-03 17:56:26 +00:00
atomic_swap_int((int *)&lu->lu_watchreq->lr_locked,
Change low-level locking a bit so that we can tell if a lock is being waitied on. Fix a races in join and cancellation. When trying to wait on a CV and the library is not yet threaded, make it threaded so that waiting actually works. When trying to nanosleep() and we're not threaded, just call the system call nanosleep instead of adding the thread to the wait queue. Clean up adding/removing new threads to the "all threads queue", assigning them unique ids, and tracking how many active threads there are. Do it all when the thread is added to the scheduling queue instead of making pthread_create() know how to do it. Fix a race where a thread could be marked for signal delivery but it could be exited before we actually add the signal to it. Other minor cleanups and bug fixes. Submitted by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-24 02:29:25 +00:00
2, &lval);
if (lval == 0)
lu->lu_watchreq->lr_locked = 0;
else
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
lck->l_wait(lck, lu);
Change low-level locking a bit so that we can tell if a lock is being waitied on. Fix a races in join and cancellation. When trying to wait on a CV and the library is not yet threaded, make it threaded so that waiting actually works. When trying to nanosleep() and we're not threaded, just call the system call nanosleep instead of adding the thread to the wait queue. Clean up adding/removing new threads to the "all threads queue", assigning them unique ids, and tracking how many active threads there are. Do it all when the thread is added to the scheduling queue instead of making pthread_create() know how to do it. Fix a race where a thread could be marked for signal delivery but it could be exited before we actually add the signal to it. Other minor cleanups and bug fixes. Submitted by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-24 02:29:25 +00:00
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
}
}
Add a method of yielding the current thread with the scheduler lock held (_thr_sched_switch_unlocked()) and use this to avoid dropping the scheduler lock and having the scheduler retake the same lock again. Add a better way of detecting if a low-level lock is in use. When switching out a thread due to blocking in the UTS, don't switch to the KSE's scheduler stack only to switch back to another thread. If possible switch to the new thread directly from the old thread and avoid the overhead of the extra context switch. Check for pending signals on a thread when entering the scheduler and add them to the threads signal frame. This includes some other minor signal fixes. Most of this was a joint effor between davidxu and myself. Reviewed by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-16 19:58:30 +00:00
lu->lu_myreq->lr_active = 1;
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
}
/*
* Release a lock.
*/
void
_lock_release(struct lock *lck, struct lockuser *lu)
{
struct lockuser *lu_tmp, *lu_h;
struct lockreq *myreq;
int prio_h;
Don't assume sizeof(long) = sizeof(int) on x86; use int instead of long types for low-level locks. Add prototypes for some internal libc functions that are wrapped by the library as cancellation points. Add memory barriers to alpha atomic swap functions (submitted by davidxu). Requested by: bde
2003-09-03 17:56:26 +00:00
int lval;
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
/**
* XXX - We probably want to remove these checks to optimize
* performance. It is also a bug if any one of the
* checks fail, so it's probably better to just let it
* SEGV and fix it.
*/
#if 0
if ((lck == NULL) || (lu == NULL))
return;
#endif
if ((lck->l_type & LCK_PRIORITY) != 0) {
prio_h = 0;
lu_h = NULL;
/* Update tail if our request is last. */
if (lu->lu_watchreq->lr_owner == NULL) {
atomic_store_rel_ptr(&lck->l_tail, lu->lu_myreq);
atomic_store_rel_ptr(&lu->lu_myreq->lr_owner, NULL);
} else {
/* Remove ourselves from the list. */
atomic_store_rel_ptr(&lu->lu_myreq->lr_owner,
lu->lu_watchreq->lr_owner);
atomic_store_rel_ptr(
&lu->lu_watchreq->lr_owner->lu_myreq, lu->lu_myreq);
}
/*
* The watch request now becomes our own because we've
* traded away our previous request. Save our previous
* request so that we can grant the lock.
*/
myreq = lu->lu_myreq;
lu->lu_myreq = lu->lu_watchreq;
lu->lu_watchreq = NULL;
lu->lu_myreq->lr_locked = 1;
lu->lu_myreq->lr_owner = lu;
lu->lu_myreq->lr_watcher = NULL;
/*
* Traverse the list of lock requests in reverse order
* looking for the user with the highest priority.
*/
for (lu_tmp = lck->l_tail->lr_watcher; lu_tmp != NULL;
lu_tmp = lu_tmp->lu_myreq->lr_watcher) {
if (lu_tmp->lu_priority > prio_h) {
lu_h = lu_tmp;
prio_h = lu_tmp->lu_priority;
}
}
if (lu_h != NULL) {
/* Give the lock to the highest priority user. */
Change low-level locking a bit so that we can tell if a lock is being waitied on. Fix a races in join and cancellation. When trying to wait on a CV and the library is not yet threaded, make it threaded so that waiting actually works. When trying to nanosleep() and we're not threaded, just call the system call nanosleep instead of adding the thread to the wait queue. Clean up adding/removing new threads to the "all threads queue", assigning them unique ids, and tracking how many active threads there are. Do it all when the thread is added to the scheduling queue instead of making pthread_create() know how to do it. Fix a race where a thread could be marked for signal delivery but it could be exited before we actually add the signal to it. Other minor cleanups and bug fixes. Submitted by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-24 02:29:25 +00:00
if (lck->l_wakeup != NULL) {
Don't assume sizeof(long) = sizeof(int) on x86; use int instead of long types for low-level locks. Add prototypes for some internal libc functions that are wrapped by the library as cancellation points. Add memory barriers to alpha atomic swap functions (submitted by davidxu). Requested by: bde
2003-09-03 17:56:26 +00:00
atomic_swap_int(
(int *)&lu_h->lu_watchreq->lr_locked,
Change low-level locking a bit so that we can tell if a lock is being waitied on. Fix a races in join and cancellation. When trying to wait on a CV and the library is not yet threaded, make it threaded so that waiting actually works. When trying to nanosleep() and we're not threaded, just call the system call nanosleep instead of adding the thread to the wait queue. Clean up adding/removing new threads to the "all threads queue", assigning them unique ids, and tracking how many active threads there are. Do it all when the thread is added to the scheduling queue instead of making pthread_create() know how to do it. Fix a race where a thread could be marked for signal delivery but it could be exited before we actually add the signal to it. Other minor cleanups and bug fixes. Submitted by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-24 02:29:25 +00:00
0, &lval);
if (lval == 2)
/* Notify the sleeper */
lck->l_wakeup(lck,
lu_h->lu_myreq->lr_watcher);
}
Add a method of yielding the current thread with the scheduler lock held (_thr_sched_switch_unlocked()) and use this to avoid dropping the scheduler lock and having the scheduler retake the same lock again. Add a better way of detecting if a low-level lock is in use. When switching out a thread due to blocking in the UTS, don't switch to the KSE's scheduler stack only to switch back to another thread. If possible switch to the new thread directly from the old thread and avoid the overhead of the extra context switch. Check for pending signals on a thread when entering the scheduler and add them to the threads signal frame. This includes some other minor signal fixes. Most of this was a joint effor between davidxu and myself. Reviewed by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-16 19:58:30 +00:00
else
Don't assume sizeof(long) = sizeof(int) on x86; use int instead of long types for low-level locks. Add prototypes for some internal libc functions that are wrapped by the library as cancellation points. Add memory barriers to alpha atomic swap functions (submitted by davidxu). Requested by: bde
2003-09-03 17:56:26 +00:00
atomic_store_rel_int(
Change low-level locking a bit so that we can tell if a lock is being waitied on. Fix a races in join and cancellation. When trying to wait on a CV and the library is not yet threaded, make it threaded so that waiting actually works. When trying to nanosleep() and we're not threaded, just call the system call nanosleep instead of adding the thread to the wait queue. Clean up adding/removing new threads to the "all threads queue", assigning them unique ids, and tracking how many active threads there are. Do it all when the thread is added to the scheduling queue instead of making pthread_create() know how to do it. Fix a race where a thread could be marked for signal delivery but it could be exited before we actually add the signal to it. Other minor cleanups and bug fixes. Submitted by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-24 02:29:25 +00:00
&lu_h->lu_watchreq->lr_locked, 0);
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
} else {
Change low-level locking a bit so that we can tell if a lock is being waitied on. Fix a races in join and cancellation. When trying to wait on a CV and the library is not yet threaded, make it threaded so that waiting actually works. When trying to nanosleep() and we're not threaded, just call the system call nanosleep instead of adding the thread to the wait queue. Clean up adding/removing new threads to the "all threads queue", assigning them unique ids, and tracking how many active threads there are. Do it all when the thread is added to the scheduling queue instead of making pthread_create() know how to do it. Fix a race where a thread could be marked for signal delivery but it could be exited before we actually add the signal to it. Other minor cleanups and bug fixes. Submitted by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-24 02:29:25 +00:00
if (lck->l_wakeup != NULL) {
Don't assume sizeof(long) = sizeof(int) on x86; use int instead of long types for low-level locks. Add prototypes for some internal libc functions that are wrapped by the library as cancellation points. Add memory barriers to alpha atomic swap functions (submitted by davidxu). Requested by: bde
2003-09-03 17:56:26 +00:00
atomic_swap_int((int *)&myreq->lr_locked,
Change low-level locking a bit so that we can tell if a lock is being waitied on. Fix a races in join and cancellation. When trying to wait on a CV and the library is not yet threaded, make it threaded so that waiting actually works. When trying to nanosleep() and we're not threaded, just call the system call nanosleep instead of adding the thread to the wait queue. Clean up adding/removing new threads to the "all threads queue", assigning them unique ids, and tracking how many active threads there are. Do it all when the thread is added to the scheduling queue instead of making pthread_create() know how to do it. Fix a race where a thread could be marked for signal delivery but it could be exited before we actually add the signal to it. Other minor cleanups and bug fixes. Submitted by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-24 02:29:25 +00:00
0, &lval);
if (lval == 2)
/* Notify the sleeper */
lck->l_wakeup(lck, myreq->lr_watcher);
}
Add a method of yielding the current thread with the scheduler lock held (_thr_sched_switch_unlocked()) and use this to avoid dropping the scheduler lock and having the scheduler retake the same lock again. Add a better way of detecting if a low-level lock is in use. When switching out a thread due to blocking in the UTS, don't switch to the KSE's scheduler stack only to switch back to another thread. If possible switch to the new thread directly from the old thread and avoid the overhead of the extra context switch. Check for pending signals on a thread when entering the scheduler and add them to the threads signal frame. This includes some other minor signal fixes. Most of this was a joint effor between davidxu and myself. Reviewed by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-16 19:58:30 +00:00
else
/* Give the lock to the previous request. */
Don't assume sizeof(long) = sizeof(int) on x86; use int instead of long types for low-level locks. Add prototypes for some internal libc functions that are wrapped by the library as cancellation points. Add memory barriers to alpha atomic swap functions (submitted by davidxu). Requested by: bde
2003-09-03 17:56:26 +00:00
atomic_store_rel_int(&myreq->lr_locked, 0);
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
}
} else {
/*
* The watch request now becomes our own because we've
* traded away our previous request. Save our previous
* request so that we can grant the lock.
*/
myreq = lu->lu_myreq;
lu->lu_myreq = lu->lu_watchreq;
lu->lu_watchreq = NULL;
lu->lu_myreq->lr_locked = 1;
Change low-level locking a bit so that we can tell if a lock is being waitied on. Fix a races in join and cancellation. When trying to wait on a CV and the library is not yet threaded, make it threaded so that waiting actually works. When trying to nanosleep() and we're not threaded, just call the system call nanosleep instead of adding the thread to the wait queue. Clean up adding/removing new threads to the "all threads queue", assigning them unique ids, and tracking how many active threads there are. Do it all when the thread is added to the scheduling queue instead of making pthread_create() know how to do it. Fix a race where a thread could be marked for signal delivery but it could be exited before we actually add the signal to it. Other minor cleanups and bug fixes. Submitted by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-24 02:29:25 +00:00
if (lck->l_wakeup) {
Don't assume sizeof(long) = sizeof(int) on x86; use int instead of long types for low-level locks. Add prototypes for some internal libc functions that are wrapped by the library as cancellation points. Add memory barriers to alpha atomic swap functions (submitted by davidxu). Requested by: bde
2003-09-03 17:56:26 +00:00
atomic_swap_int((int *)&myreq->lr_locked, 0, &lval);
Change low-level locking a bit so that we can tell if a lock is being waitied on. Fix a races in join and cancellation. When trying to wait on a CV and the library is not yet threaded, make it threaded so that waiting actually works. When trying to nanosleep() and we're not threaded, just call the system call nanosleep instead of adding the thread to the wait queue. Clean up adding/removing new threads to the "all threads queue", assigning them unique ids, and tracking how many active threads there are. Do it all when the thread is added to the scheduling queue instead of making pthread_create() know how to do it. Fix a race where a thread could be marked for signal delivery but it could be exited before we actually add the signal to it. Other minor cleanups and bug fixes. Submitted by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-24 02:29:25 +00:00
if (lval == 2)
/* Notify the sleeper */
lck->l_wakeup(lck, myreq->lr_watcher);
}
Add a method of yielding the current thread with the scheduler lock held (_thr_sched_switch_unlocked()) and use this to avoid dropping the scheduler lock and having the scheduler retake the same lock again. Add a better way of detecting if a low-level lock is in use. When switching out a thread due to blocking in the UTS, don't switch to the KSE's scheduler stack only to switch back to another thread. If possible switch to the new thread directly from the old thread and avoid the overhead of the extra context switch. Check for pending signals on a thread when entering the scheduler and add them to the threads signal frame. This includes some other minor signal fixes. Most of this was a joint effor between davidxu and myself. Reviewed by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-16 19:58:30 +00:00
else
Protect against a race between granting a lock and accessing other parts of the lock. Submitted by: davidxu
2003-05-04 22:29:09 +00:00
/* Give the lock to the previous request. */
Don't assume sizeof(long) = sizeof(int) on x86; use int instead of long types for low-level locks. Add prototypes for some internal libc functions that are wrapped by the library as cancellation points. Add memory barriers to alpha atomic swap functions (submitted by davidxu). Requested by: bde
2003-09-03 17:56:26 +00:00
atomic_store_rel_int(&myreq->lr_locked, 0);
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
}
Add a method of yielding the current thread with the scheduler lock held (_thr_sched_switch_unlocked()) and use this to avoid dropping the scheduler lock and having the scheduler retake the same lock again. Add a better way of detecting if a low-level lock is in use. When switching out a thread due to blocking in the UTS, don't switch to the KSE's scheduler stack only to switch back to another thread. If possible switch to the new thread directly from the old thread and avoid the overhead of the extra context switch. Check for pending signals on a thread when entering the scheduler and add them to the threads signal frame. This includes some other minor signal fixes. Most of this was a joint effor between davidxu and myself. Reviewed by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-16 19:58:30 +00:00
lu->lu_myreq->lr_active = 0;
Add FIFO queueing locking operations based on atomic swap. Modify thread errno for the new libpthread changes. Reviewed by: davidxu
2003-04-18 05:02:39 +00:00
}
Add a method of yielding the current thread with the scheduler lock held (_thr_sched_switch_unlocked()) and use this to avoid dropping the scheduler lock and having the scheduler retake the same lock again. Add a better way of detecting if a low-level lock is in use. When switching out a thread due to blocking in the UTS, don't switch to the KSE's scheduler stack only to switch back to another thread. If possible switch to the new thread directly from the old thread and avoid the overhead of the extra context switch. Check for pending signals on a thread when entering the scheduler and add them to the threads signal frame. This includes some other minor signal fixes. Most of this was a joint effor between davidxu and myself. Reviewed by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-16 19:58:30 +00:00
void
_lock_grant(struct lock *lck /* unused */, struct lockuser *lu)
{
Don't assume sizeof(long) = sizeof(int) on x86; use int instead of long types for low-level locks. Add prototypes for some internal libc functions that are wrapped by the library as cancellation points. Add memory barriers to alpha atomic swap functions (submitted by davidxu). Requested by: bde
2003-09-03 17:56:26 +00:00
atomic_store_rel_int(&lu->lu_watchreq->lr_locked, 3);
Add a method of yielding the current thread with the scheduler lock held (_thr_sched_switch_unlocked()) and use this to avoid dropping the scheduler lock and having the scheduler retake the same lock again. Add a better way of detecting if a low-level lock is in use. When switching out a thread due to blocking in the UTS, don't switch to the KSE's scheduler stack only to switch back to another thread. If possible switch to the new thread directly from the old thread and avoid the overhead of the extra context switch. Check for pending signals on a thread when entering the scheduler and add them to the threads signal frame. This includes some other minor signal fixes. Most of this was a joint effor between davidxu and myself. Reviewed by: davidxu Approved by: re@ (blanket for libpthread)
2003-05-16 19:58:30 +00:00
}
void
_lockuser_setactive(struct lockuser *lu, int active)
{
lu->lu_myreq->lr_active = active;
}
Reference in New Issue Copy Permalink