Re: [PATCH 3.4 43/91] perf: Fix a race condition in perf_remove_from_context()

[Zefan Li]

> | diff --git a/kernel/events/core.c b/kernel/events/core.c
> | index 685ce46..c958be1 100644
> | --- a/kernel/events/core.c
> | +++ b/kernel/events/core.c
> | @@ -1702,6 +1702,16 @@ retry:
> |  	 */
> |  	if (ctx->is_active) {
> |  		raw_spin_unlock_irq(&ctx->lock);
> | +		/*
> | +		 * Reload the task pointer, it might have been changed by
> | +		 * a concurrent perf_event_context_sched_out().
> | +		 */
> | +		task = ctx->task;
> | +		/*
> | +		 * Reload the task pointer, it might have been changed by
> | +		 * a concurrent perf_event_context_sched_out().
> | +		 */
> | +		task = ctx->task;
> 
> Something wrong in the way the patch was applied ?
> The lines are identical...
> 
> The original commit, 3577af70, has the change applied in two
> places: perf_event_disable() and perf_remove_from_context().
> 

I guess it's because the two parts of the original patch are identical,
so the patch command wasn't smart enough to do the right work.

It looks to me this won't make things worse than bebore, so I'll fix
it in 3.4.106.

Thanks for spotting this!

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/