Messages in this thread Patch in this message | | | Date | Mon, 22 May 2006 02:10:10 +0200 | From | Rene Herman <> | Subject | Re: 2.6.17-rc2+ regression -- audio skipping |
| |
Lee Revell wrote:
> On Sun, 2006-05-21 at 22:24 +0200, Rene Herman wrote:
>> 2.6.17-rc2 (and 3 and 4) make my audio skip. Audio player is ogg123 >> running in an xterm. Browsing heavy sites (say, eBay) with Firefox >> 1.5.0.3 gets me audio underruns quickly. This does not happen on >> 2.6.17-rc1 and earlier (I just tested extensively; quite impossible to >> generate underruns on -rc1, quickly on -rc2 and later). >> >> It's not ALSA; reverted */sound/* from the rc1-rc2 interdiff. It's also >> not cfq-iosched.c. Any likely culprits in there? (I'm not a GIT user). >> > > I would suspect the scheduler interactivity patches. Please confirm > this by running ogg123 at nice -20 - do the underruns persist?
They do persist. Thanks for the hint though -- "sched: fix interactive task starvation" is the culprit:
http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=5ce74abe788a26698876e66b9c9ce7e7acc25413
Added author and acked-by's to the CC. Mike, this patch is no good for me. Audio underruns galore, with only ogg123 and firefox (browsing the GIT tree online is also a nice trigger by the way).
If I back it out, everything is fine for me again. Back-out attached as a patch against -rc4. This also backs out your follow-up "don't awaken RT tasks on expired array" as it was dependant:
http://www.kernel.org/git/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=8a5bc075b8d8cf7a87b3f08fad2fba0f5d13295e
While looking at the patch I noticed there was +1 difference in the "limit" value between the macro and the static inline version of expired_starving() so I experimented with adding that back but that wasn't it unfortunately.
I can test patches (preferably versus -rc4) although possibly not quickly.
Rene.
Index: local/kernel/sched.c =================================================================== --- local.orig/kernel/sched.c 2006-05-08 20:47:06.000000000 +0200 +++ local/kernel/sched.c 2006-05-22 01:03:12.000000000 +0200 @@ -665,55 +665,13 @@ static int effective_prio(task_t *p) } /* - * We place interactive tasks back into the active array, if possible. - * - * To guarantee that this does not starve expired tasks we ignore the - * interactivity of a task if the first expired task had to wait more - * than a 'reasonable' amount of time. This deadline timeout is - * load-dependent, as the frequency of array switched decreases with - * increasing number of running tasks. We also ignore the interactivity - * if a better static_prio task has expired, and switch periodically - * regardless, to ensure that highly interactive tasks do not starve - * the less fortunate for unreasonably long periods. - */ -static inline int expired_starving(runqueue_t *rq) -{ - int limit; - - /* - * Arrays were recently switched, all is well - */ - if (!rq->expired_timestamp) - return 0; - - limit = STARVATION_LIMIT * rq->nr_running; - - /* - * It's time to switch arrays - */ - if (jiffies - rq->expired_timestamp >= limit) - return 1; - - /* - * There's a better selection in the expired array - */ - if (rq->curr->static_prio > rq->best_expired_prio) - return 1; - - /* - * All is well - */ - return 0; -} - -/* * __activate_task - move a task to the runqueue. */ static void __activate_task(task_t *p, runqueue_t *rq) { prio_array_t *target = rq->active; - if (unlikely(batch_task(p) || (expired_starving(rq) && !rt_task(p)))) + if (batch_task(p)) target = rq->expired; enqueue_task(p, target); rq->nr_running++; @@ -2532,6 +2490,22 @@ unsigned long long current_sched_time(co } /* + * We place interactive tasks back into the active array, if possible. + * + * To guarantee that this does not starve expired tasks we ignore the + * interactivity of a task if the first expired task had to wait more + * than a 'reasonable' amount of time. This deadline timeout is + * load-dependent, as the frequency of array switched decreases with + * increasing number of running tasks. We also ignore the interactivity + * if a better static_prio task has expired: + */ +#define EXPIRED_STARVING(rq) \ + ((STARVATION_LIMIT && ((rq)->expired_timestamp && \ + (jiffies - (rq)->expired_timestamp >= \ + STARVATION_LIMIT * ((rq)->nr_running) + 1))) || \ + ((rq)->curr->static_prio > (rq)->best_expired_prio)) + +/* * Account user cpu time to a process. * @p: the process that the cpu time gets accounted to * @hardirq_offset: the offset to subtract from hardirq_count() @@ -2666,7 +2640,7 @@ void scheduler_tick(void) if (!rq->expired_timestamp) rq->expired_timestamp = jiffies; - if (!TASK_INTERACTIVE(p) || expired_starving(rq)) { + if (!TASK_INTERACTIVE(p) || EXPIRED_STARVING(rq)) { enqueue_task(p, rq->expired); if (p->static_prio < rq->best_expired_prio) rq->best_expired_prio = p->static_prio; | |