Re: [patch 04/18] x86, bts: wait until traced task has been scheduled out

[Ingo Molnar <mingo@elte.hu>]

* Metzger, Markus T <markus.t.metzger@intel.com> wrote:

> >-----Original Message-----
> >From: Ingo Molnar [mailto:mingo@elte.hu]
> >Sent: Thursday, April 02, 2009 9:18 PM
> >To: Metzger, Markus T
> >Cc: tglx@linutronix.de; hpa@zytor.com; markus.t.metzger@gmail.com; roland@redhat.com;
> >eranian@googlemail.com; oleg@redhat.com; Villacis, Juan; ak@linux.jf.intel.com; linux-
> >kernel@vger.kernel.org
> >Subject: Re: [patch 04/18] x86, bts: wait until traced task has been scheduled out
> >
> >
> >* markus.t.metzger@intel.com <markus.t.metzger@intel.com> wrote:
> >
> >> In order to stop branch tracing for a running task, we need to
> >> first clear the branch tracing control bits before we may free the
> >> tracing buffer. If the traced task is running, the cpu might still
> >> trace that task after the branch trace control bits have cleared.
> >>
> >> Wait until the traced task has been scheduled out before
> >> proceeding.
> >>
> >> A similar problem affects the task debug store context. We first
> >> remove the context, then we need to wait until the task has been
> >> scheduled out before we can free the context memory.
> >>
> >>
> >> Reviewed-by: Oleg Nesterov <oleg@redhat.com>
> >> Signed-off-by: Markus Metzger <markus.t.metzger@intel.com>
> >> ---
> >>  arch/x86/kernel/ds.c |   40 	40 +	0 -	0 !
> >>  1 file changed, 40 insertions(+)
> >>
> >> Index: b/arch/x86/kernel/ds.c
> >> ===================================================================
> >> --- a/arch/x86/kernel/ds.c
> >> +++ b/arch/x86/kernel/ds.c
> >> @@ -250,6 +250,40 @@ static DEFINE_PER_CPU(struct ds_context
> >>  #define system_context per_cpu(system_context_array, smp_processor_id())
> >>
> >>
> >> +/*
> >> + * Wait for the traced task to unschedule.
> >> + *
> >> + * This guarantees that the bts trace configuration has been
> >> + * synchronized with the cpu executing the task.
> >> + */
> >> +static void wait_to_unschedule(struct task_struct *task)
> >> +{
> >
> >this should be in sched.c and task_is_running() should not be
> >exported from there.
> >
> >I.e. your original patch which i objected to is probably the right
> >one, but should be named something like "task_wait_context_switch()"
> >- which signals its purpose: that it is to wait for the task to
> >context-switch at least once, so that its ptrace state is installed
> >(or deinstalled) for sure.
> 
> OK.
> 
> I'll move it to sched.c.
> 
> In that case, I would use task_running() without holding the rq 
> lock, since we don't really care whether we read an old value or 
> not. Would that be OK with you?

i'd have to see that in full context - reading non-locked results 
can in essence result in stale old values being read out, regardless 
of current reality. task_running() is normally used within the rq 
lock.

	Ingo