Module: xenomai-2.6
Branch: master
Commit: b03e3d08379f236eb75b34c1b705e127c6e3b2e5
URL:
http://git.xenomai.org/?p=xenomai-2.6.git;a=commit;h=b03e3d08379f236eb75b34c1b705e127c6e3b2e5
Author: Philippe Gerum <r...@xenomai.org>
Date: Fri Jun 26 15:50:41 2015 +0200
nucleus/pod: fix missed rescheduling in SMP
During parallel multi-core operations involving two threads, the
following scenario may happen:
CPU0: thread A
xnpod_suspend_thread() on self
=> XNSCHED raised via xnsched_set_resched(), local scheduler (no IPI)
=> suspend eventually denied (bad timeout)
=> no rescheduling
CPU1: thread B
xnpod_resume_thread() on A
=> XNSCHED seen pending by xnsched_set_resched(), no resched IPI marked
=> __xnpod_schedule() does not send a resched IPI to CPU0
The outcome is that CPU1 won't send any resched IPI to CPU0 although
it should have done so, which practically delays wakeup of thread A
until some operation on CPU0 eventually causes a local rescheduling.
The bottom line is that any code raising XNSCHED __must__ trigger the
rescheduling procedure.
---
ksrc/nucleus/pod.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/ksrc/nucleus/pod.c b/ksrc/nucleus/pod.c
index 0a361e9..ef6fb41 100644
--- a/ksrc/nucleus/pod.c
+++ b/ksrc/nucleus/pod.c
@@ -1361,9 +1361,6 @@ void xnpod_suspend_thread(xnthread_t *thread, xnflags_t
mask,
sched = thread->sched;
oldstate = thread->state;
- if (thread == sched->curr)
- xnsched_set_resched(sched);
-
/* Is the thread ready to run? */
if (!xnthread_test_state(thread, XNTHREAD_BLOCK_BITS)) {
#ifdef CONFIG_XENO_OPT_PERVASIVE
@@ -1435,6 +1432,7 @@ void xnpod_suspend_thread(xnthread_t *thread, xnflags_t
mask,
#endif /* __XENO_SIM__ */
if (thread == sched->curr) {
+ xnsched_set_resched(sched);
__clrbits(sched->lflags, XNINLOCK);
/*
* If the current thread is being relaxed, we must
_______________________________________________
Xenomai-git mailing list
Xenomai-git@xenomai.org
http://xenomai.org/mailman/listinfo/xenomai-git
