freebsd-ports/devel/linuxthreads/files/patch-spinlock.c
Tor Egge 8f82445315 Backport 2002-04-24 fix from newer linuxthreads versions to avoid hangs
or busy wait loop due to race between timeout and pthread_cond_signal.
2006-01-24 01:05:01 +00:00

59 lines
1.9 KiB
C

--- spinlock.c.orig Tue Mar 27 04:52:56 2001
+++ spinlock.c Tue Jan 10 09:44:39 2006
@@ -72,8 +72,6 @@
#endif
#if defined HAS_COMPARE_AND_SWAP
-again:
-
/* On SMP, try spinning to get the lock. */
if (__pthread_smp_kernel) {
@@ -94,6 +92,8 @@
lock->__spinlock += (spin_count - lock->__spinlock) / 8;
}
+again:
+
/* No luck, try once more or suspend. */
do {
@@ -110,7 +110,7 @@
}
if (self != NULL) {
- THREAD_SETMEM(self, p_nextlock, (pthread_descr) (oldstatus & ~1L));
+ THREAD_SETMEM(self, p_nextlock, (pthread_descr) oldstatus);
/* Make sure the store in p_nextlock completes before performing
the compare-and-swap */
MEMORY_BARRIER();
@@ -188,7 +188,7 @@
multiprocessor Alphas) could perform such reordering even though
the loads are dependent. */
READ_MEMORY_BARRIER();
- thr = *ptr;
+ thr = (pthread_descr)((long)(thr->p_nextlock) & ~1L);
}
/* Prevent reordering of the load of lock->__status above and
thr->p_nextlock below */
@@ -198,17 +198,16 @@
/* If max prio thread is at head, remove it with compare-and-swap
to guard against concurrent lock operation. This removal
also has the side effect of marking the lock as released
- because the new status comes from thr->p_nextlock whose
- least significant bit is clear. */
+ by clearing the least significant bit. */
thr = (pthread_descr) (oldstatus & ~1L);
if (! __compare_and_swap_with_release_semantics
- (&lock->__status, oldstatus, (long)(thr->p_nextlock)))
+ (&lock->__status, oldstatus, (long)(thr->p_nextlock) & ~1L))
goto again;
} else {
/* No risk of concurrent access, remove max prio thread normally.
But in this case we must also flip the least significant bit
of the status to mark the lock as released. */
- thr = *maxptr;
+ thr = (pthread_descr)((long)*maxptr & ~1L);
*maxptr = thr->p_nextlock;
do {