All of lore.kernel.org
 help / color / mirror / Atom feed
From: Frederic Weisbecker <frederic@kernel.org>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>,
	LKML <linux-kernel@vger.kernel.org>,
	"Eric W . Biederman" <ebiederm@xmission.com>,
	Oleg Nesterov <oleg@redhat.com>, Ingo Molnar <mingo@kernel.org>
Subject: Re: [PATCH 5/6] posix-cpu-timers: Force next expiration recalc after early timer firing
Date: Wed, 16 Jun 2021 13:59:23 +0200	[thread overview]
Message-ID: <20210616115923.GC801071@lothringen> (raw)
In-Reply-To: <YMnHnUcufPhtnDZP@hirez.programming.kicks-ass.net>

On Wed, Jun 16, 2021 at 11:42:53AM +0200, Peter Zijlstra wrote:
> On Fri, Jun 04, 2021 at 01:31:58PM +0200, Frederic Weisbecker wrote:
> > diff --git a/kernel/time/posix-cpu-timers.c b/kernel/time/posix-cpu-timers.c
> > index 0b5715c8db04..d8325a906314 100644
> > --- a/kernel/time/posix-cpu-timers.c
> > +++ b/kernel/time/posix-cpu-timers.c
> > @@ -405,6 +405,21 @@ static int posix_cpu_timer_create(struct k_itimer *new_timer)
> >  	return 0;
> >  }
> >  
> > +static void __disarm_timer(struct k_itimer *timer, struct task_struct *p,
> > +			   u64 old_expires)
> > +{
> > +	int clkidx = CPUCLOCK_WHICH(timer->it_clock);
> > +	struct posix_cputimer_base *base;
> > +
> > +	if (CPUCLOCK_PERTHREAD(timer->it_clock))
> > +		base = p->posix_cputimers.bases + clkidx;
> > +	else
> > +		base = p->signal->posix_cputimers.bases + clkidx;
> > +
> > +	if (old_expires == base->nextevt)
> > +		base->nextevt = 0;
> > +}
> > +
> >  /*
> >   * Dequeue the timer and reset the base if it was its earliest expiration.
> >   * It makes sure the next tick recalculates the base next expiration so we
> > @@ -415,24 +430,14 @@ static void disarm_timer(struct k_itimer *timer, struct task_struct *p)
> >  {
> >  	struct cpu_timer *ctmr = &timer->it.cpu;
> >  	u64 old_expires = cpu_timer_getexpires(ctmr);
> > -	struct posix_cputimer_base *base;
> >  	bool queued;
> > -	int clkidx;
> >  
> >  	queued = cpu_timer_dequeue(ctmr);
> >  	cpu_timer_setexpires(ctmr, 0);
> >  	if (!queued)
> >  		return;
> >  
> > -	clkidx = CPUCLOCK_WHICH(timer->it_clock);
> > -
> > -	if (CPUCLOCK_PERTHREAD(timer->it_clock))
> > -		base = p->posix_cputimers.bases + clkidx;
> > -	else
> > -		base = p->signal->posix_cputimers.bases + clkidx;
> > -
> > -	if (old_expires == base->nextevt)
> > -		base->nextevt = 0;
> > +	__disarm_timer(timer, p, old_expires);
> >  }
> >  
> >  
> > @@ -686,8 +691,7 @@ static int posix_cpu_timer_set(struct k_itimer *timer, int timer_flags,
> >  			u64 exp = bump_cpu_timer(timer, val);
> >  
> >  			if (val < exp) {
> > -				old_expires = exp - val;
> > -				old->it_value = ns_to_timespec64(old_expires);
> > +				old->it_value = ns_to_timespec64(exp - val);
> >  			} else {
> >  				old->it_value.tv_nsec = 1;
> >  				old->it_value.tv_sec = 0;
> > @@ -748,9 +752,28 @@ static int posix_cpu_timer_set(struct k_itimer *timer, int timer_flags,
> >  		 * accumulate more time on this clock.
> >  		 */
> >  		cpu_timer_fire(timer);
> > +
> > +		sighand = lock_task_sighand(p, &flags);
> > +		if (sighand == NULL)
> > +			goto out;
> > +		if (!cpu_timer_queued(&timer->it.cpu)) {
> > +			/*
> > +			 * Disarm the previous timer to deactivate the tick
> > +			 * dependency and process wide cputime counter if
> > +			 * necessary.
> > +			 */
> > +			__disarm_timer(timer, p, old_expires);
> > +			/*
> > +			 * If the previous timer was deactivated, we might have
> > +			 * just started the process wide cputime counter. Make
> > +			 * sure we poke the tick to deactivate it then.
> > +			 */
> > +			if (!old_expires && !CPUCLOCK_PERTHREAD(timer->it_clock))
> > +				p->signal->posix_cputimers.bases[clkid].nextevt = 0;
> > +		}
> > +		unlock_task_sighand(p, &flags);
> >  	}
> 
> I'm thinking this is a better fix than patch #2. AFAICT you can now go
> back to unconditionally doing start, and then if we fire it early, we'll
> disarm the thing.
> 
> That would avoid the disconnect between the start condition and the fire
> condition.

Right but the drawback is that we unconditionally start the threadgroup
counter while initializing the timer to 0 (deactivated).

Then in the next tick at least one thread will need to lock the sighand
and re-evaluate the whole list.

  reply	other threads:[~2021-06-16 11:59 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-06-04 11:31 [PATCH 0/6] posix-cpu-timers: Bunch of fixes Frederic Weisbecker
2021-06-04 11:31 ` [PATCH 1/6] posix-cpu-timers: Fix rearm racing against process tick Frederic Weisbecker
2021-06-09 11:54   ` Frederic Weisbecker
2021-06-11 11:49     ` Peter Zijlstra
2021-06-11 12:37       ` Frederic Weisbecker
2021-06-04 11:31 ` [PATCH 2/6] posix-cpu-timers: Don't start process wide cputime counter if timer is disabled Frederic Weisbecker
2021-06-09 12:18   ` Frederic Weisbecker
2021-06-10 10:24     ` Frederic Weisbecker
2021-06-16  8:51   ` Peter Zijlstra
2021-06-16 10:51     ` Frederic Weisbecker
2021-06-16 11:26       ` Peter Zijlstra
2021-06-16 11:50         ` Peter Zijlstra
2021-06-04 11:31 ` [PATCH 3/6] posix-cpu-timers: Force next_expiration recalc after timer deletion Frederic Weisbecker
2021-06-16  9:16   ` Peter Zijlstra
2021-06-04 11:31 ` [PATCH 4/6] posix-cpu-timers: Force next_expiration recalc after timer reset Frederic Weisbecker
2021-06-16  9:23   ` Peter Zijlstra
2021-06-16 11:21     ` Frederic Weisbecker
2021-06-16 11:33       ` Peter Zijlstra
2021-06-04 11:31 ` [PATCH 5/6] posix-cpu-timers: Force next expiration recalc after early timer firing Frederic Weisbecker
2021-06-16  9:42   ` Peter Zijlstra
2021-06-16 11:59     ` Frederic Weisbecker [this message]
2021-06-16 13:23       ` Peter Zijlstra
2021-06-16 14:53         ` Frederic Weisbecker
2021-06-04 11:31 ` [PATCH 6/6] posix-cpu-timers: Force next expiration recalc after itimer reset 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=20210616115923.GC801071@lothringen \
    --to=frederic@kernel.org \
    --cc=ebiederm@xmission.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@kernel.org \
    --cc=oleg@redhat.com \
    --cc=peterz@infradead.org \
    --cc=tglx@linutronix.de \
    /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.