All of lore.kernel.org
 help / color / mirror / Atom feed
From: Ingo Molnar <mingo@kernel.org>
To: Andy Lutomirski <luto@amacapital.net>
Cc: fweisbec@redhat.com, Paolo Bonzini <pbonzini@redhat.com>,
	X86 ML <x86@kernel.org>, Thomas Gleixner <tglx@linutronix.de>,
	Peter Zijlstra <peterz@infradead.org>,
	Heiko Carstens <heiko.carstens@de.ibm.com>,
	Ingo Molnar <mingo@redhat.com>,
	Mike Galbraith <umgwanakikbuti@gmail.com>,
	Rik van Riel <riel@redhat.com>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	williams@redhat.com
Subject: Re: [PATCH 3/3] context_tracking,x86: remove extraneous irq disable & enable from context tracking on syscall entry
Date: Fri, 8 May 2015 08:37:11 +0200	[thread overview]
Message-ID: <20150508063711.GA4934@gmail.com> (raw)
In-Reply-To: <CALCETrX8zEcAXBAsut8MGU7CwEMd_dFYDsRyysY0qR18qAC0rg@mail.gmail.com>


* Andy Lutomirski <luto@amacapital.net> wrote:

> > So do you mean:
> >
> >    this_cpu_set(rcu_state) = IN_KERNEL;
> >    ...
> >    this_cpu_inc(rcu_qs_ctr);
> >    this_cpu_set(rcu_state) = IN_USER;
> >
> > ?
> >
> > So in your proposal we'd have an INC and two MOVs. I think we can make
> > it just two simple stores into a byte flag, one on entry and one on
> > exit:
> >
> >    this_cpu_set(rcu_state) = IN_KERNEL;
> >    ...
> >    this_cpu_set(rcu_state) = IN_USER;
> >
> 
> I was thinking that either a counter or a state flag could make sense.
> Doing both would be pointless.  The counter could use the low bits to
> indicate the state.  The benefit of the counter would be that the
> RCU-waiting CPU could observe that the counter has incremented and
> that therefore a grace period has elapsed.  Getting it right would
> require lots of care.

So if you mean:

       <syscall entry>
       ...
       this_cpu_inc(rcu_qs_ctr);
       <syscall exit>

I don't see how this would work reliably: how do you handle the case 
of a SCHED_FIFO task never returning from user-space (common technique 
in RT apps)? synchronize_rcu() would block indefinitely as it would 
never see rcu_qs_ctr increase.

We have to be able to observe user-mode anyway, for system-time 
statistics purposes, and that flag could IMHO also drive the RCU GC 
machinery.

> > > The problem is that I don't see how TIF_RCU_THINGY can work 
> > > reliably. If the remote CPU sets it, it'll be too late and we'll 
> > > still enter user mode without seeing it.  If it's just an 
> > > optimization, though, then it should be fine.
> >
> > Well, after setting it, the remote CPU has to re-check whether the 
> > RT CPU has entered user-mode - before it goes to wait.
> 
> How?
> 
> Suppose the exit path looked like:
> 
> this_cpu_write(rcu_state, IN_USER);
> 
> if (ti->flags & _TIF_RCU_NOTIFY) {
>     if (test_and_clear_bit(TIF_RCU_NOTIFY, &ti->flags))
>         slow_notify_rcu_that_we_are_exiting();
> }
> 
> iret or sysret;

No, it would look like this:

   this_cpu_write(rcu_state, IN_USER);
   iret or sysret;

I.e. IN_USER is set well after all notifications are checked. No 
kernel execution happens afterwards. (No extra checks added - the 
regular return-to-user-work checks would handle TIF_RCU_NOTIFY.)

( Same goes for idle: we just mark it IN_IDLE and move it back to 
  IN_KERNEL after the idling ends. )

> The RCU-waiting CPU sees that rcu_state == IN_KERNEL and sets 
> _TIF_RCU_NOTIFY.  This could happen arbitrarily late before IRET 
> because stores can be delayed.  (It could even happen after sysret, 
> IIRC, but IRET is serializing.)

All it has to do in the synchronize_rcu() slowpath is something like:

	if (per_cpu(rcu_state, rt_cpu) == IN_KERNEL) {
		smp_mb__before_atomic();
		set_tsk_thread_flag(remote_task, TIF_RCU_NOTIFY);
		smp_rmb();
		if (per_cpu(rcu_state, rt_cpu) == IN_KERNEL)
			... go wait ...
	}
	/* Cool, we observed quiescent state: */

The cost of the trivial barrier is nothing compared to the 'go wait' 
cost which we will pay in 99.9% of the cases!

> If we put an mfence after this_cpu_set or did an unconditional 
> test_and_clear_bit on ti->flags then this problem goes away, but 
> that would probably be slower than we'd like.

We are talking about a dozen cycles, while a typical synchronize_rcu() 
will wait millions (sometimes billions) of cycles. There's absolutely 
zero performance concern here and it's all CPU local in any case.

In fact a user-mode/kernel-mode flag speeds up naive implementations 
of synchronize_rcu(): because it's able to observe extended quiescent 
state immediately, without having to wait for a counter to increase 
(which was always the classic way to observe grace periods).

If all CPUs are in user mode or are idle (which is rather common!) 
then synchronize_rcu() could return almost immediately - while 
previously it had to wait for scheduling or periodic timer irqs to 
trigger on all CPUs - adding many millisecs of delay even in the best 
of cases.

Thanks,

	Ingo

  reply	other threads:[~2015-05-08  6:37 UTC|newest]

Thread overview: 83+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-04-30 21:23 [PATCH 0/3] reduce nohz_full syscall overhead by 10% riel
2015-04-30 21:23 ` [PATCH 1/3] reduce indentation in __acct_update_integrals riel
2015-04-30 21:23 ` [PATCH 2/3] remove local_irq_save from __acct_update_integrals riel
2015-04-30 21:23 ` [PATCH 3/3] context_tracking,x86: remove extraneous irq disable & enable from context tracking on syscall entry riel
2015-04-30 21:56   ` Andy Lutomirski
2015-05-01  6:40   ` Ingo Molnar
2015-05-01 15:20     ` Rik van Riel
2015-05-01 15:59       ` Ingo Molnar
2015-05-01 16:03         ` Andy Lutomirski
2015-05-01 16:21           ` Ingo Molnar
2015-05-01 16:26             ` Rik van Riel
2015-05-01 16:34               ` Ingo Molnar
2015-05-01 18:05                 ` Rik van Riel
2015-05-01 18:40                   ` Ingo Molnar
2015-05-01 19:11                     ` Rik van Riel
2015-05-01 19:37                       ` Andy Lutomirski
2015-05-02  5:27                         ` Ingo Molnar
2015-05-02 18:27                           ` Rik van Riel
2015-05-03 18:41                           ` Andy Lutomirski
2015-05-07 10:35                             ` Ingo Molnar
2015-05-04  9:26                           ` Paolo Bonzini
2015-05-04 13:30                             ` Rik van Riel
2015-05-04 14:06                             ` Rik van Riel
2015-05-04 14:19                             ` Rik van Riel
2015-05-04 15:59                             ` question about RCU dynticks_nesting Rik van Riel
2015-05-04 18:39                               ` Paul E. McKenney
2015-05-04 19:39                                 ` Rik van Riel
2015-05-04 20:02                                   ` Paul E. McKenney
2015-05-04 20:13                                     ` Rik van Riel
2015-05-04 20:38                                       ` Paul E. McKenney
2015-05-04 20:53                                         ` Rik van Riel
2015-05-05  5:54                                           ` Paul E. McKenney
2015-05-06  1:49                                             ` Mike Galbraith
2015-05-06  3:44                                               ` Mike Galbraith
2015-05-06  6:06                                                 ` Paul E. McKenney
2015-05-06  6:52                                                   ` Mike Galbraith
2015-05-06  7:01                                                     ` Mike Galbraith
2015-05-07  0:59                                           ` Frederic Weisbecker
2015-05-07 15:44                                             ` Rik van Riel
2015-05-04 19:00                               ` Rik van Riel
2015-05-04 19:39                                 ` Paul E. McKenney
2015-05-04 19:59                                   ` Rik van Riel
2015-05-04 20:40                                     ` Paul E. McKenney
2015-05-05 10:53                                   ` Peter Zijlstra
2015-05-05 12:34                                     ` Paul E. McKenney
2015-05-05 13:00                                       ` Peter Zijlstra
2015-05-05 18:35                                         ` Paul E. McKenney
2015-05-05 21:09                                           ` Rik van Riel
2015-05-06  5:41                                             ` Paul E. McKenney
2015-05-05 10:48                                 ` Peter Zijlstra
2015-05-05 10:51                                   ` Peter Zijlstra
2015-05-05 12:30                                     ` Paul E. McKenney
2015-05-02  4:06                   ` [PATCH 3/3] context_tracking,x86: remove extraneous irq disable & enable from context tracking on syscall entry Mike Galbraith
2015-05-01 16:37             ` Ingo Molnar
2015-05-01 16:40               ` Rik van Riel
2015-05-01 16:45                 ` Ingo Molnar
2015-05-01 16:54                   ` Rik van Riel
2015-05-01 17:12                     ` Ingo Molnar
2015-05-01 17:22                       ` Rik van Riel
2015-05-01 17:59                         ` Ingo Molnar
2015-05-01 16:22           ` Rik van Riel
2015-05-01 16:27             ` Ingo Molnar
2015-05-03 13:23       ` Mike Galbraith
2015-05-03 17:30         ` Rik van Riel
2015-05-03 18:24           ` Andy Lutomirski
2015-05-03 18:52             ` Rik van Riel
2015-05-07 10:48               ` Ingo Molnar
2015-05-07 12:18                 ` Frederic Weisbecker
2015-05-07 12:29                   ` Ingo Molnar
2015-05-07 15:47                     ` Rik van Riel
2015-05-08  7:58                       ` Ingo Molnar
2015-05-07 12:22                 ` Andy Lutomirski
2015-05-07 12:44                   ` Ingo Molnar
2015-05-07 12:49                     ` Ingo Molnar
2015-05-08  6:17                       ` Paul E. McKenney
2015-05-07 12:52                     ` Andy Lutomirski
2015-05-07 15:08                       ` Ingo Molnar
2015-05-07 17:47                         ` Andy Lutomirski
2015-05-08  6:37                           ` Ingo Molnar [this message]
2015-05-08 10:59                             ` Andy Lutomirski
2015-05-08 11:27                               ` Ingo Molnar
2015-05-08 12:56                                 ` Andy Lutomirski
2015-05-08 13:27                                   ` Ingo Molnar

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=20150508063711.GA4934@gmail.com \
    --to=mingo@kernel.org \
    --cc=fweisbec@redhat.com \
    --cc=heiko.carstens@de.ibm.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=luto@amacapital.net \
    --cc=mingo@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=peterz@infradead.org \
    --cc=riel@redhat.com \
    --cc=tglx@linutronix.de \
    --cc=umgwanakikbuti@gmail.com \
    --cc=williams@redhat.com \
    --cc=x86@kernel.org \
    /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.