diff options
author | Stephane Eranian <eranian@google.com> | 2010-10-20 09:25:01 -0400 |
---|---|---|
committer | Ingo Molnar <mingo@elte.hu> | 2010-10-22 08:18:27 -0400 |
commit | 9ffcfa6f1f63eeac15555b745c292eb9f59130f6 (patch) | |
tree | 803cc11f312da98713fdfc366b7ec79351dffa55 /kernel/perf_event.c | |
parent | 96681fc3c9e7d1f89ab64e5eec40b6467c97680f (diff) |
perf_events: Revert: Fix transaction recovery in group_sched_in()
This patch reverts commit 8e5fc1a (perf_events: Fix transaction
recovery in group_sched_in()) because it had one flaw in case the
group could never be scheduled. It would cause time_enabled to get
negative.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <4cbeeeb7.0aefd80a.6e40.0e2f@mx.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Diffstat (limited to 'kernel/perf_event.c')
-rw-r--r-- | kernel/perf_event.c | 76 |
1 files changed, 13 insertions, 63 deletions
diff --git a/kernel/perf_event.c b/kernel/perf_event.c index f309e8014c78..39afdb07d758 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c | |||
@@ -417,8 +417,8 @@ event_filter_match(struct perf_event *event) | |||
417 | return event->cpu == -1 || event->cpu == smp_processor_id(); | 417 | return event->cpu == -1 || event->cpu == smp_processor_id(); |
418 | } | 418 | } |
419 | 419 | ||
420 | static int | 420 | static void |
421 | __event_sched_out(struct perf_event *event, | 421 | event_sched_out(struct perf_event *event, |
422 | struct perf_cpu_context *cpuctx, | 422 | struct perf_cpu_context *cpuctx, |
423 | struct perf_event_context *ctx) | 423 | struct perf_event_context *ctx) |
424 | { | 424 | { |
@@ -437,13 +437,14 @@ __event_sched_out(struct perf_event *event, | |||
437 | } | 437 | } |
438 | 438 | ||
439 | if (event->state != PERF_EVENT_STATE_ACTIVE) | 439 | if (event->state != PERF_EVENT_STATE_ACTIVE) |
440 | return 0; | 440 | return; |
441 | 441 | ||
442 | event->state = PERF_EVENT_STATE_INACTIVE; | 442 | event->state = PERF_EVENT_STATE_INACTIVE; |
443 | if (event->pending_disable) { | 443 | if (event->pending_disable) { |
444 | event->pending_disable = 0; | 444 | event->pending_disable = 0; |
445 | event->state = PERF_EVENT_STATE_OFF; | 445 | event->state = PERF_EVENT_STATE_OFF; |
446 | } | 446 | } |
447 | event->tstamp_stopped = ctx->time; | ||
447 | event->pmu->del(event, 0); | 448 | event->pmu->del(event, 0); |
448 | event->oncpu = -1; | 449 | event->oncpu = -1; |
449 | 450 | ||
@@ -452,19 +453,6 @@ __event_sched_out(struct perf_event *event, | |||
452 | ctx->nr_active--; | 453 | ctx->nr_active--; |
453 | if (event->attr.exclusive || !cpuctx->active_oncpu) | 454 | if (event->attr.exclusive || !cpuctx->active_oncpu) |
454 | cpuctx->exclusive = 0; | 455 | cpuctx->exclusive = 0; |
455 | return 1; | ||
456 | } | ||
457 | |||
458 | static void | ||
459 | event_sched_out(struct perf_event *event, | ||
460 | struct perf_cpu_context *cpuctx, | ||
461 | struct perf_event_context *ctx) | ||
462 | { | ||
463 | int ret; | ||
464 | |||
465 | ret = __event_sched_out(event, cpuctx, ctx); | ||
466 | if (ret) | ||
467 | event->tstamp_stopped = ctx->time; | ||
468 | } | 456 | } |
469 | 457 | ||
470 | static void | 458 | static void |
@@ -664,7 +652,7 @@ retry: | |||
664 | } | 652 | } |
665 | 653 | ||
666 | static int | 654 | static int |
667 | __event_sched_in(struct perf_event *event, | 655 | event_sched_in(struct perf_event *event, |
668 | struct perf_cpu_context *cpuctx, | 656 | struct perf_cpu_context *cpuctx, |
669 | struct perf_event_context *ctx) | 657 | struct perf_event_context *ctx) |
670 | { | 658 | { |
@@ -684,6 +672,8 @@ __event_sched_in(struct perf_event *event, | |||
684 | return -EAGAIN; | 672 | return -EAGAIN; |
685 | } | 673 | } |
686 | 674 | ||
675 | event->tstamp_running += ctx->time - event->tstamp_stopped; | ||
676 | |||
687 | if (!is_software_event(event)) | 677 | if (!is_software_event(event)) |
688 | cpuctx->active_oncpu++; | 678 | cpuctx->active_oncpu++; |
689 | ctx->nr_active++; | 679 | ctx->nr_active++; |
@@ -694,35 +684,6 @@ __event_sched_in(struct perf_event *event, | |||
694 | return 0; | 684 | return 0; |
695 | } | 685 | } |
696 | 686 | ||
697 | static inline int | ||
698 | event_sched_in(struct perf_event *event, | ||
699 | struct perf_cpu_context *cpuctx, | ||
700 | struct perf_event_context *ctx) | ||
701 | { | ||
702 | int ret = __event_sched_in(event, cpuctx, ctx); | ||
703 | if (ret) | ||
704 | return ret; | ||
705 | event->tstamp_running += ctx->time - event->tstamp_stopped; | ||
706 | return 0; | ||
707 | } | ||
708 | |||
709 | static void | ||
710 | group_commit_event_sched_in(struct perf_event *group_event, | ||
711 | struct perf_cpu_context *cpuctx, | ||
712 | struct perf_event_context *ctx) | ||
713 | { | ||
714 | struct perf_event *event; | ||
715 | u64 now = ctx->time; | ||
716 | |||
717 | group_event->tstamp_running += now - group_event->tstamp_stopped; | ||
718 | /* | ||
719 | * Schedule in siblings as one group (if any): | ||
720 | */ | ||
721 | list_for_each_entry(event, &group_event->sibling_list, group_entry) { | ||
722 | event->tstamp_running += now - event->tstamp_stopped; | ||
723 | } | ||
724 | } | ||
725 | |||
726 | static int | 687 | static int |
727 | group_sched_in(struct perf_event *group_event, | 688 | group_sched_in(struct perf_event *group_event, |
728 | struct perf_cpu_context *cpuctx, | 689 | struct perf_cpu_context *cpuctx, |
@@ -736,13 +697,7 @@ group_sched_in(struct perf_event *group_event, | |||
736 | 697 | ||
737 | pmu->start_txn(pmu); | 698 | pmu->start_txn(pmu); |
738 | 699 | ||
739 | /* | 700 | if (event_sched_in(group_event, cpuctx, ctx)) { |
740 | * use __event_sched_in() to delay updating tstamp_running | ||
741 | * until the transaction is committed. In case of failure | ||
742 | * we will keep an unmodified tstamp_running which is a | ||
743 | * requirement to get correct timing information | ||
744 | */ | ||
745 | if (__event_sched_in(group_event, cpuctx, ctx)) { | ||
746 | pmu->cancel_txn(pmu); | 701 | pmu->cancel_txn(pmu); |
747 | return -EAGAIN; | 702 | return -EAGAIN; |
748 | } | 703 | } |
@@ -751,31 +706,26 @@ group_sched_in(struct perf_event *group_event, | |||
751 | * Schedule in siblings as one group (if any): | 706 | * Schedule in siblings as one group (if any): |
752 | */ | 707 | */ |
753 | list_for_each_entry(event, &group_event->sibling_list, group_entry) { | 708 | list_for_each_entry(event, &group_event->sibling_list, group_entry) { |
754 | if (__event_sched_in(event, cpuctx, ctx)) { | 709 | if (event_sched_in(event, cpuctx, ctx)) { |
755 | partial_group = event; | 710 | partial_group = event; |
756 | goto group_error; | 711 | goto group_error; |
757 | } | 712 | } |
758 | } | 713 | } |
759 | 714 | ||
760 | if (!pmu->commit_txn(pmu)) { | 715 | if (!pmu->commit_txn(pmu)) |
761 | /* commit tstamp_running */ | ||
762 | group_commit_event_sched_in(group_event, cpuctx, ctx); | ||
763 | return 0; | 716 | return 0; |
764 | } | 717 | |
765 | group_error: | 718 | group_error: |
766 | /* | 719 | /* |
767 | * Groups can be scheduled in as one unit only, so undo any | 720 | * Groups can be scheduled in as one unit only, so undo any |
768 | * partial group before returning: | 721 | * partial group before returning: |
769 | * | ||
770 | * use __event_sched_out() to avoid updating tstamp_stopped | ||
771 | * because the event never actually ran | ||
772 | */ | 722 | */ |
773 | list_for_each_entry(event, &group_event->sibling_list, group_entry) { | 723 | list_for_each_entry(event, &group_event->sibling_list, group_entry) { |
774 | if (event == partial_group) | 724 | if (event == partial_group) |
775 | break; | 725 | break; |
776 | __event_sched_out(event, cpuctx, ctx); | 726 | event_sched_out(event, cpuctx, ctx); |
777 | } | 727 | } |
778 | __event_sched_out(group_event, cpuctx, ctx); | 728 | event_sched_out(group_event, cpuctx, ctx); |
779 | 729 | ||
780 | pmu->cancel_txn(pmu); | 730 | pmu->cancel_txn(pmu); |
781 | 731 | ||