mutex sleepq chain not owned at
/usr/src/sys/kern/subr_sleepqueue.c
Attilio Rao
attilio at freebsd.org
Tue Jun 10 19:44:39 UTC 2008
2008/6/9, kevinxlinuz <kevinxlinuz at 163.com>:
> Recently I meet a problem in freebsd 8.0/amd64.
> See PR/124200
> http://www.freebsd.org/cgi/query-pr.cgi?pr=124200&cat=
>
> I try to find the reason.
>
> in cv_broadcastpri(...),it call sleepq_lock(cvp),next it call sleepq_broadcast(cvp, SLEEPQ_CONDVAR, pri, 0).
> in sleepq_broadcast(void *wchan, int flags, int pri, int queue),sleepqueue sq = sleepq_lookup(wchan) /* here wchan will be checked,and sq->sq_wchan == wchan == cvp (passed from cv_broadcastpri())*/;
> I add mtx_assert in /usr/src/sys/kern/subr_sleepqueue.c
> sleepq_broadcast(void *wchan, int flags, int pri, int queue)
> {
> struct sleepqueue *sq;
> struct thread *td;
>
> struct sleepqueue_chain *sc;
>
> CTR2(KTR_PROC, "sleepq_broadcast(%p, %d)", wchan, flags);
> KASSERT(wchan != NULL, ("%s: invalid NULL wait channel", __func__));
> MPASS((queue >= 0) && (queue < NR_SLEEPQS));
> sq = sleepq_lookup(wchan); //wchan == cvp, cvp from cv_broadcastpri(...),and sleepq_lock(cvp)
> //here sq->sq_wchan == wchan == cvp
> if (sq == NULL)
> return;
> KASSERT(sq->sq_type == (flags & SLEEPQ_TYPE),
> ("%s: mismatch between sleep/wakeup and cv_*", __func__));
>
> /* Resume all blocked threads on the sleep queue. */
> while (!TAILQ_EMPTY(&sq->sq_blocked[queue])) {
> td = TAILQ_FIRST(&sq->sq_blocked[queue]);
> thread_lock(td);
> /* ------test start---------- */
> sc = SC_LOOKUP(sq->sq_wchan); //sq->sq_wchan should be wchan
> mtx_assert(&sc->sc_lock, MA_OWNED); //panic here,sq->sq_wchan != wchan ? or sleepq_unlock(wchan) was called by others
> /* -----test end----- */
> sleepq_resume_thread(sq, td, pri);
> thread_unlock(td);
> }
> }
>
Hello,
We are trying to track this down but things go very slowly because I
can't reproduce the bug.
I would need you try some diagnostic patches, do you think you can
work on that with me? Can you reproduce easilly the bug?
Thanks,
Attilio
--
Peace can only be achieved by understanding - A. Einstein
More information about the freebsd-current
mailing list