All of lore.kernel.org
 help / color / mirror / Atom feed
From: Frederic Weisbecker <fweisbec@gmail.com>
To: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Oleg Nesterov <oleg@redhat.com>, Ingo Molnar <mingo@elte.hu>,
	Alan Stern <stern@rowland.harvard.edu>,
	Arnaldo Carvalho de Melo <acme@redhat.com>,
	Paul Mackerras <paulus@samba.org>,
	Prasad <prasad@linux.vnet.ibm.com>,
	Roland McGrath <roland@redhat.com>,
	linux-kernel@vger.kernel.org
Subject: Re: Q: perf_install_in_context/perf_event_enable are racy?
Date: Fri, 21 Jan 2011 21:40:17 +0100	[thread overview]
Message-ID: <20110121204014.GA2870@nowhere> (raw)
In-Reply-To: <1295622304.28776.293.camel@laptop>

On Fri, Jan 21, 2011 at 04:05:04PM +0100, Peter Zijlstra wrote:
> On Fri, 2011-01-21 at 15:26 +0100, Oleg Nesterov wrote:
> > 
> > > Ah, I think I see how that works:
> > 
> > Hmm. I don't...
> > 
> > >
> > >   __perf_event_task_sched_out()
> > >     perf_event_context_sched_out()
> > >       if (do_switch)
> > >         cpuctx->task_ctx = NULL;
> > 
> > exactly, this clears ->task_ctx
> > 
> > > vs
> > >
> > >   __perf_install_in_context()
> > >    if (cpu_ctx->task_ctx != ctx)
> > 
> > And then __perf_install_in_context() sets cpuctx->task_ctx = ctx,
> > because ctx->task == current && cpuctx->task_ctx == NULL.
> 
> Hrm,. right, so the comment suggests it should do what it doesn't :-)
> 
> It looks like Paul's a63eaf34ae60bd (perf_counter: Dynamically allocate
> tasks' perf_counter_context struct), relevant hunk below, wrecked it:
> 
> @@ -568,11 +582,17 @@ static void __perf_install_in_context(void *info)
>          * If this is a task context, we need to check whether it is
>          * the current task context of this cpu. If not it has been
>          * scheduled out before the smp call arrived.
> +        * Or possibly this is the right context but it isn't
> +        * on this cpu because it had no counters.
>          */
> -       if (ctx->task && cpuctx->task_ctx != ctx)
> -               return;
> +       if (ctx->task && cpuctx->task_ctx != ctx) {
> +               if (cpuctx->task_ctx || ctx->task != current)
> +                       return;
> +               cpuctx->task_ctx = ctx;
> +       }
>  
>         spin_lock_irqsave(&ctx->lock, flags);
> +       ctx->is_active = 1;
>         update_context_time(ctx);
>  
>         /*
> 
> 
> I can't really seem to come up with a sane test that isn't racy with
> something, my cold seems to have clogged not only my nose :/


What do you think about the following (only compile tested yet), it
probably needs more comments, factorizing the checks betwee, perf_event_enable()
and perf_install_in_context(), build-cond against __ARCH_WANT_INTERRUPTS_ON_CTXSW,
but the (good or bad) idea is there.


diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index c5fa717..e97472b 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -928,6 +928,8 @@ static void add_event_to_ctx(struct perf_event *event,
 	event->tstamp_stopped = tstamp;
 }
 
+static DEFINE_PER_CPU(int, task_events_schedulable);
+
 /*
  * Cross CPU call to install and enable a performance event
  *
@@ -949,7 +951,8 @@ static void __perf_install_in_context(void *info)
 	 * on this cpu because it had no events.
 	 */
 	if (ctx->task && cpuctx->task_ctx != ctx) {
-		if (cpuctx->task_ctx || ctx->task != current)
+		if (cpuctx->task_ctx || ctx->task != current
+		    || !__get_cpu_var(task_events_schedulable))
 			return;
 		cpuctx->task_ctx = ctx;
 	}
@@ -1091,7 +1094,8 @@ static void __perf_event_enable(void *info)
 	 * event's task is the current task on this cpu.
 	 */
 	if (ctx->task && cpuctx->task_ctx != ctx) {
-		if (cpuctx->task_ctx || ctx->task != current)
+		if (cpuctx->task_ctx || ctx->task != current
+		    || !__get_cpu_var(task_events_schedulable))
 			return;
 		cpuctx->task_ctx = ctx;
 	}
@@ -1414,6 +1418,9 @@ void __perf_event_task_sched_out(struct task_struct *task,
 {
 	int ctxn;
 
+	__get_cpu_var(task_events_schedulable) = 0;
+	barrier(); /* Must be visible by enable/install_in_context IPI */
+
 	for_each_task_context_nr(ctxn)
 		perf_event_context_sched_out(task, ctxn, next);
 }
@@ -1587,6 +1594,8 @@ void __perf_event_task_sched_in(struct task_struct *task)
 	struct perf_event_context *ctx;
 	int ctxn;
 
+	__get_cpu_var(task_events_schedulable) = 1;
+
 	for_each_task_context_nr(ctxn) {
 		ctx = task->perf_event_ctxp[ctxn];
 		if (likely(!ctx))
@@ -5964,6 +5973,18 @@ SYSCALL_DEFINE5(perf_event_open,
 	WARN_ON_ONCE(ctx->parent_ctx);
 	mutex_lock(&ctx->mutex);
 
+	/*
+	 * Every pending sched switch must finish so that
+	 * we ensure every pending calls to perf_event_task_sched_in/out are
+	 * finished. We ensure the next ones will correctly handle the
+	 * perf_task_events label and then the task_events_schedulable
+	 * state. So perf_install_in_context() won't install events
+	 * in the tiny race window between perf_event_task_sched_out()
+	 * and perf_event_task_sched_in() in the __ARCH_WANT_INTERRUPTS_ON_CTXSW
+	 * case.
+	 */
+	synchronize_sched();
+
 	if (move_group) {
 		perf_install_in_context(ctx, group_leader, cpu);
 		get_ctx(ctx);


  reply	other threads:[~2011-01-21 20:40 UTC|newest]

Thread overview: 91+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-11-08 14:56 Q: perf_event && task->ptrace_bps[] Oleg Nesterov
2010-11-08 14:57 ` Q: sys_perf_event_open() && PF_EXITING Oleg Nesterov
2011-01-19 18:21   ` [PATCH 0/2] Was: " Oleg Nesterov
2011-01-19 18:22     ` [PATCH 1/2] perf: fix find_get_context() vs perf_event_exit_task() race Oleg Nesterov
2011-01-19 18:49       ` Peter Zijlstra
2011-01-19 19:18       ` [tip:perf/urgent] perf: Fix " tip-bot for Oleg Nesterov
2011-01-21 15:29         ` Ingo Molnar
2011-01-21 15:53           ` Oleg Nesterov
2011-01-21 17:45             ` [PATCH] perf: perf_event_exit_task_context: s/rcu_dereference/rcu_dereference_raw/ Oleg Nesterov
2011-01-21 17:53               ` Oleg Nesterov
2011-01-21 21:50                 ` Paul E. McKenney
2011-01-24 11:51                   ` Oleg Nesterov
2011-01-21 22:12               ` [tip:perf/urgent] " tip-bot for Oleg Nesterov
2011-01-19 18:22     ` [PATCH 2/2] perf: fix perf_event_init_task()/perf_event_free_task() interaction Oleg Nesterov
2011-01-19 18:51       ` Peter Zijlstra
2011-01-19 19:19       ` [tip:perf/urgent] perf: Fix " tip-bot for Oleg Nesterov
2011-01-20 19:30     ` Q: perf_install_in_context/perf_event_enable are racy? Oleg Nesterov
2011-01-21 12:11       ` Peter Zijlstra
2011-01-21 13:03         ` Ingo Molnar
2011-01-21 13:39           ` Peter Zijlstra
2011-01-21 14:26             ` Oleg Nesterov
2011-01-21 15:05               ` Peter Zijlstra
2011-01-21 20:40                 ` Frederic Weisbecker [this message]
2011-01-24 11:42                   ` Oleg Nesterov
2011-01-26 17:53                     ` Oleg Nesterov
2011-01-26 18:49                       ` Oleg Nesterov
2011-01-26 18:51                         ` [PATCH] fix the theoretical task_cpu/task_curr problem in kick_process/task_oncpu_function_call Oleg Nesterov
2011-01-26 19:05                         ` Q: perf_install_in_context/perf_event_enable are racy? Peter Zijlstra
2011-01-26 19:33                           ` Peter Zijlstra
2011-01-26 19:38                             ` Peter Zijlstra
2011-01-26 21:19                             ` Oleg Nesterov
2011-01-26 21:33                               ` Oleg Nesterov
2011-01-27 10:32                                 ` Peter Zijlstra
2011-01-27 12:29                                   ` Peter Zijlstra
2011-01-27 16:10                                     ` Oleg Nesterov
2011-01-27 16:27                                       ` Peter Zijlstra
2011-01-27 16:59                                         ` Oleg Nesterov
2011-01-27 15:52                                   ` Oleg Nesterov
2011-01-27 13:14                       ` Peter Zijlstra
2011-01-27 14:28                         ` Peter Zijlstra
2011-01-27 14:58                           ` Peter Zijlstra
2011-01-27 16:57                         ` Oleg Nesterov
2011-01-27 17:11                           ` Peter Zijlstra
2011-01-27 22:18                             ` Oleg Nesterov
2011-01-28 11:52                               ` Peter Zijlstra
2011-01-28 14:57                                 ` Peter Zijlstra
2011-01-28 16:28                                   ` Oleg Nesterov
2011-01-28 18:11                                     ` Peter Zijlstra
2011-01-31 17:26                                       ` Oleg Nesterov
2011-01-31 18:23                                         ` Peter Zijlstra
2011-01-31 19:11                                           ` Oleg Nesterov
2011-01-31 19:29                                             ` Peter Zijlstra
2011-02-01 14:03                                               ` [PATCH] perf: Cure task_oncpu_function_call() races Peter Zijlstra
2011-02-01 17:27                                                 ` Oleg Nesterov
2011-02-01 18:08                                                   ` Peter Zijlstra
2011-02-01 18:18                                                     ` Peter Zijlstra
2011-02-01 21:00                                                     ` Peter Zijlstra
2010-11-08 14:57 ` Q: perf_event && event->owner Oleg Nesterov
2010-11-08 20:11   ` Frederic Weisbecker
2010-11-08 20:41     ` Peter Zijlstra
2010-11-09 16:18       ` Oleg Nesterov
2010-11-09 15:57     ` Oleg Nesterov
2010-11-09 16:56       ` Peter Zijlstra
2010-11-09 16:58         ` Oleg Nesterov
2010-11-09 17:07           ` Peter Zijlstra
2010-11-09 17:42             ` Oleg Nesterov
2010-11-09 18:01               ` Peter Zijlstra
2010-11-09 18:57                 ` Oleg Nesterov
2010-11-09 19:16                   ` Peter Zijlstra
2010-11-10 15:17                   ` Peter Zijlstra
2010-11-10 15:44                     ` Oleg Nesterov
2010-11-12 15:48                       ` Peter Zijlstra
2010-11-12 18:49                         ` Oleg Nesterov
2010-11-18 14:09                         ` [tip:perf/urgent] perf: Fix owner-list vs exit tip-bot for Peter Zijlstra
2010-11-08 18:41 ` Q: perf_event && task->ptrace_bps[] Frederic Weisbecker
2010-11-08 19:18   ` Oleg Nesterov
2011-01-17 23:58     ` Frederic Weisbecker
2011-01-18  1:16       ` Roland McGrath
2011-01-17 20:34 ` Oleg Nesterov
2011-01-17 20:52   ` Peter Zijlstra
2011-01-17 21:01     ` Frederic Weisbecker
2011-01-18 16:09     ` [PATCH 0/2] perf: event->cpu checking fixes Oleg Nesterov
2011-01-18 16:10       ` [PATCH 1/2] perf: find_get_context: fix the per-cpu-counter check Oleg Nesterov
2011-01-18 19:06         ` [tip:perf/urgent] perf: Find_get_context: " tip-bot for Oleg Nesterov
2011-01-18 16:10       ` [PATCH 2/2] perf: validate cpu early in perf_event_alloc() Oleg Nesterov
2011-01-18 19:07         ` [tip:perf/urgent] perf: Validate " tip-bot for Oleg Nesterov
2011-01-18 18:42   ` Q: perf_event && task->ptrace_bps[] Frederic Weisbecker
2011-01-19 15:37     ` Oleg Nesterov
2011-01-19 20:05       ` Frederic Weisbecker
2011-01-20 17:28         ` Oleg Nesterov
2011-01-28 17:41           ` Frederic Weisbecker

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20110121204014.GA2870@nowhere \
    --to=fweisbec@gmail.com \
    --cc=a.p.zijlstra@chello.nl \
    --cc=acme@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@elte.hu \
    --cc=oleg@redhat.com \
    --cc=paulus@samba.org \
    --cc=prasad@linux.vnet.ibm.com \
    --cc=roland@redhat.com \
    --cc=stern@rowland.harvard.edu \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.