Re: [PATCH 01/20] perf: Add PERF_EVENT_STATE_EXIT state for events with exited task

From: Peter Zijlstra <peterz@infradead.org>
To: Jiri Olsa <jolsa@kernel.org>
Cc: linux-kernel@vger.kernel.org,
	Adrian Hunter <adrian.hunter@intel.com>,
	Arnaldo Carvalho de Melo <acme@kernel.org>,
	Corey Ashford <cjashfor@linux.vnet.ibm.com>,
	David Ahern <dsahern@gmail.com>,
	Frederic Weisbecker <fweisbec@gmail.com>,
	Ingo Molnar <mingo@kernel.org>,
	Jean Pihet <jean.pihet@linaro.org>,
	Namhyung Kim <namhyung@kernel.org>,
	Paul Mackerras <paulus@samba.org>
Subject: Re: [PATCH 01/20] perf: Add PERF_EVENT_STATE_EXIT state for events with exited task
Date: Mon, 11 Aug 2014 14:01:02 +0200	[thread overview]
Message-ID: <20140811120102.GY9918@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <1407747014-18394-2-git-send-email-jolsa@kernel.org>

[-- Attachment #1: Type: text/plain, Size: 2402 bytes --]

On Mon, Aug 11, 2014 at 10:49:55AM +0200, Jiri Olsa wrote:
> ---
>  include/linux/perf_event.h |  1 +
>  kernel/events/core.c       | 12 ++++++++++--
>  2 files changed, 11 insertions(+), 2 deletions(-)
> 
> diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
> index 707617a8c0f6..54f3a6241386 100644
> --- a/include/linux/perf_event.h
> +++ b/include/linux/perf_event.h
> @@ -268,6 +268,7 @@ struct pmu {
>   * enum perf_event_active_state - the states of a event
>   */
>  enum perf_event_active_state {
> +	PERF_EVENT_STATE_EXIT		= -3,
>  	PERF_EVENT_STATE_ERROR		= -2,
>  	PERF_EVENT_STATE_OFF		= -1,
>  	PERF_EVENT_STATE_INACTIVE	=  0,
> diff --git a/kernel/events/core.c b/kernel/events/core.c
> index 14086e45c5c4..dde0eefa410a 100644
> --- a/kernel/events/core.c
> +++ b/kernel/events/core.c
> @@ -3499,7 +3499,8 @@ perf_read_hw(struct perf_event *event, char __user *buf, size_t count)
>  	 * error state (i.e. because it was pinned but it couldn't be
>  	 * scheduled on to the CPU at some point).
>  	 */
> -	if (event->state == PERF_EVENT_STATE_ERROR)
> +	if ((event->state == PERF_EVENT_STATE_ERROR) ||
> +	    (event->state == PERF_EVENT_STATE_EXIT))
>  		return 0;
>  
>  	if (count < event->read_size)
> @@ -3526,9 +3527,13 @@ static unsigned int perf_poll(struct file *file, poll_table *wait)
>  {
>  	struct perf_event *event = file->private_data;
>  	struct ring_buffer *rb;
> -	unsigned int events = POLL_HUP;
> +	unsigned int events = POLLHUP;

Should not that be an independent bugfix? It is a silly little thing
indeed, but it does change behaviour.

>  	poll_wait(file, &event->waitq, wait);
> +
> +	if (event->state == PERF_EVENT_STATE_EXIT)
> +		return POLLHUP;
> +

So, seeing how events is already POLLHUP here, why not return that?

>  	/*
>  	 * Pin the event->rb by taking event->mmap_mutex; otherwise
>  	 * perf_event_set_output() can swizzle our rb and make us miss wakeups.
> @@ -7484,6 +7489,9 @@ __perf_event_exit_task(struct perf_event *child_event,
>  	if (child_event->parent) {
>  		sync_child_event(child_event, child);
>  		free_event(child_event);
> +	} else {
> +		child_event->state = PERF_EVENT_STATE_EXIT;
> +		perf_event_wakeup(child_event);
>  	}
>  }

In any case, ACK on this patch, I'll assume you want to take the lot
through acme seeing how its mostly tools bits.

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]