Re: [PATCH,RFC] perf: panic due to inclied cpu context task_ctx value

[Oleg Nesterov <oleg@redhat.com>]

On 03/24, Jiri Olsa wrote:
>
> - close is called on event on CPU 0:
> 	- the task is scheduled on CPU 0
> 	- __perf_event_task_sched_in is called
> 	- cpuctx->task_ctx is set
> 	- perf_sched_events jump label is decremented and == 0
> 	- __perf_event_task_sched_out is not called
> 	- cpuctx->task_ctx on CPU 0 stays set

I think you are right.

And this is already wrong afaics, even if we add the workaround into
free_ctx/etc.

For example, suppose that we attach another PERF_ATTACH_TASK counter
to this task later. perf_sched_events will be incremented again, but
perf_install_in_context() should hang retrying until this task runs
again, ->is_active == T.

Or, even if sys_perf_event_open() succeeds the next
perf_event_context_sched_in() will do nothing because it checks
cpuctx->task_ctx != ctx (unless another task has a counter and
schedules in on that CPU).

This should be fixed somehow, but so far I have no ideas.

Oleg.