linux-kernel - Re: [RFC PATCH 6/9] sched: Treat try_to_block

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <fb998d03b4ecc51834bf4383a71932ca877900cd.camel@redhat.com>
Date: Mon, 14 Apr 2025 12:31:12 +0200
From: Gabriele Monaco <gmonaco@...hat.com>
To: Nam Cao <namcao@...utronix.de>
Cc: linux-kernel@...r.kernel.org, Ingo Molnar <mingo@...hat.com>, Peter
 Zijlstra <peterz@...radead.org>, Tomas Glozar <tglozar@...hat.com>, Juri
 Lelli <jlelli@...hat.com>
Subject: Re: [RFC PATCH 6/9] sched: Treat try_to_block_task with pending
 signal as wakeup



On Sun, 2025-04-13 at 17:05 +0200, Nam Cao wrote:
> On Fri, Apr 04, 2025 at 10:45:19AM +0200, Gabriele Monaco wrote:
> > If a task sets itself to interruptible and schedules, the
> > __schedule
> > function checks whether there's a pending signal and, if that's the
> > case, updates the state of the task to runnable instead of
> > dequeuing.
> > By looking at the tracepoints, we see the task enters the scheduler
> > while sleepable but exits as runnable. From a modelling
> > perspective,
> > this is equivalent to a wakeup and the tracepoints should reflect
> > that.
> > 
> > Add the waking/wakeup tracepoints in the try_to_block_task function
> > and
> > set the prev_state used by the sched_switch tracepoint to
> > TASK_RUNNING
> > if the task had a pending signal and was not blocked.
> > 
> > Signed-off-by: Gabriele Monaco <gmonaco@...hat.com>
> > ---
> >  kernel/sched/core.c | 11 +++++++++--
> >  1 file changed, 9 insertions(+), 2 deletions(-)
> > 
> > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > index f2f79236d5811..48cb32abce01a 100644
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -6584,7 +6584,12 @@ static bool try_to_block_task(struct rq *rq,
> > struct task_struct *p,
> >  	int flags = DEQUEUE_NOCLOCK;
> >  
> >  	if (signal_pending_state(task_state, p)) {
> > -		WRITE_ONCE(p->__state, TASK_RUNNING);
> > +		/*
> > +		 * From a modelling perspective, this is
> > equivalent to a wakeup
> > +		 * before dequeuing the task: trace accordingly.
> > +		 */
> > +		trace_sched_waking(p);
> > +		ttwu_do_wakeup(p);
> 
> I don't think we should put trace_sched_waking() here.
> trace_sched_waking()
> "is guaranteed to be called from the waking context", and this is not
> the
> waking context.
> 
> There is already a trace_sched_waking() in signal_wake_up_state().
> This is
> duplicating that, in the wrong context.
> 
> ttwu_do_wakeup() alone should be sufficient?

Mmh, that's a good point.
The thing is: this happens when the signal is generated while we are
scheduling (on different CPUs), so we take a short-cut and put the task
to running directly.
This thing is already racy, so we may or may not see the waking/wakeup.

Now probably waking shouldn't be there for the reason you said, but I'm
not sure a wakeup not following a waking would be correct either.
I might be missing something here, though.

Thanks,
Gabriele