Re: [PATCH 3/6] tracing/ftrace: Fix a race condition in sched_switch tracer

[Frédéric Weisbecker]

2008/10/10 Steven Rostedt <rostedt@xxxxxxxxxxx>:
>
> On Fri, 10 Oct 2008, Frederic Weisbecker wrote:
>>
>> This patch fixes a race condition in the sched_switch tracer.
>> If several tasks (IE: concurrent initcalls) are playing with
>> tracing_start_cmdline_record() and tracing_stop_cmdline_record(),
>> the following situation could happen:
>>
>>
>> _ Task A and B are using the same tracepoint probe. Task A holds it. Task B is
>> sleeping and doesn't hold it.
>> _ Task A frees the sched tracer, then sched_ref is decremented to 0.
>> _ Task A is preempted and hadn't yet unregistered its tracepoint probe,
>> then B runs.
>> _ B increments sched_ref, sees it's 1 and then guess it has to register its probe. But it has not been freed by task A.
>> _ A lot of bad things can happen after that...
>
> OK, I see the issue.
>
>>
>> Signed-off-by: Frederic Weisbecker <fweisbec@xxxxxxxxx>
>> CC: Steven Rostedt <rostedt@xxxxxxxxxxx>
>> ---
>>  kernel/trace/trace_sched_switch.c |    5 +++++
>>  1 files changed, 5 insertions(+), 0 deletions(-)
>>
>> diff --git a/kernel/trace/trace_sched_switch.c b/kernel/trace/trace_sched_switch.c
>> index b8f56be..59de514 100644
>> --- a/kernel/trace/trace_sched_switch.c
>> +++ b/kernel/trace/trace_sched_switch.c
>> @@ -17,6 +17,7 @@
>>  static struct trace_array    *ctx_trace;
>>  static int __read_mostly     tracer_enabled;
>>  static atomic_t                      sched_ref;
>> +static DEFINE_MUTEX(tracepoint_mutex);
>>
>>  static void
>>  probe_sched_switch(struct rq *__rq, struct task_struct *prev,
>> @@ -125,18 +126,22 @@ static void tracing_start_sched_switch(void)
>>  {
>>       long ref;
>>
>> +     mutex_lock(&tracepoint_mutex);
>>       ref = atomic_inc_return(&sched_ref);
>>       if (ref == 1)
>>               tracing_sched_register();
>
> Could you make another patch to convert sched_ref to an int. With a mutex
> lock, there's no reason for this to be atomic.
>
> We can simply do:
>
>        if (!(sched_ref++))
>                tracing_sched_register();
>
>> +     mutex_unlock(&tracepoint_mutex);
>>  }
>>
>>  static void tracing_stop_sched_switch(void)
>>  {
>>       long ref;
>>
>> +     mutex_lock(&tracepoint_mutex);
>>       ref = atomic_dec_and_test(&sched_ref);
>>       if (ref)
>>               tracing_sched_unregister();
>
> and
>
>        if (!(--sched_ref))
>                tracing_sched_unregister();
>
>> +     mutex_unlock(&tracepoint_mutex);
>
> Also, lets change the name. sched_register_mutex?
>
>>  }
>>
>>  void tracing_start_cmdline_record(void)

I didn't want to change it into an int because probe_sched_switch().
But actually that shouldn't be a problem.
Ok I will apply your comments in the next days (I can't until this week-end)....
--
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/