From: "Paul E. McKenney" <paulmck@kernel.org>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
rostedt <rostedt@goodmis.org>,
linux-kernel <linux-kernel@vger.kernel.org>,
Ingo Molnar <mingo@kernel.org>,
"Joel Fernandes, Google" <joel@joelfernandes.org>,
Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
"Gustavo A. R. Silva" <gustavo@embeddedor.com>,
Thomas Gleixner <tglx@linutronix.de>,
Josh Triplett <josh@joshtriplett.org>,
Lai Jiangshan <jiangshanlai@gmail.com>
Subject: Re: [PATCH] tracing/perf: Move rcu_irq_enter/exit_irqson() to perf trace point hook
Date: Tue, 11 Feb 2020 05:03:01 -0800 [thread overview]
Message-ID: <20200211130301.GH2935@paulmck-ThinkPad-P72> (raw)
In-Reply-To: <20200211120015.GL14914@hirez.programming.kicks-ass.net>
On Tue, Feb 11, 2020 at 01:00:15PM +0100, Peter Zijlstra wrote:
> On Mon, Feb 10, 2020 at 07:30:32PM -0500, Mathieu Desnoyers wrote:
>
> > > because perf only uses rcu to synchronize trace points.
> >
> > That last part seems inaccurate. The tracepoint synchronization is two-fold:
> > one part is internal to tracepoint.c (see rcu_free_old_probes()), and the other
> > is only needed if the probes are within modules which can be unloaded (see
> > tracepoint_synchronize_unregister()). AFAIK, perf never implements probe callbacks
> > within modules, so the latter is not needed by perf.
> >
> > The culprit of the problem here is that perf issues "rcu_read_lock()" and
> > "rcu_read_unlock()" within the probe callbacks it registers to the tracepoints,
> > including the rcuidle ones. Those require that RCU is "watching", which is
> > triggering the regression when we remove the calls to rcu_irq_enter/exit_irqson()
> > from the rcuidle tracepoint instrumentation sites.
>
> It is not the fact that perf issues rcu_read_lock() that is the problem.
> As we established yesterday, I can probably remove most rcu_read_lock()
> calls from perf today (yay RCU flavour unification).
Glad some aspect of this unification is actually helping you. ;-)
> The problem is that the core perf code uses RCU managed data; and we
> need an existence guarantee for it. It would be BAD (TM) if the
> ring-buffer we're writing data to were to suddenly dissapear under our
> feet etc..
>
> > Which brings a question about handling of NMIs: in the proposed patch, if
> > a NMI nests over rcuidle context, AFAIU it will be in a state
> > !rcu_is_watching() && in_nmi(), which is handled by this patch with a simple
> > "return", meaning important NMIs doing hardware event sampling can be
> > completely lost.
> >
> > Considering that we cannot use rcu_irq_enter/exit_irqson() from NMI context,
> > is it at all valid to use rcu_read_lock/unlock() as perf does from NMI handlers,
>
> Again, rcu_read_lock() itself really isn't the problem. But we need
> NMIs, just like regular interrupts, to imply rcu_read_lock(). That is,
> any observable (RCU managed) pointer must stay valid during the NMI/IRQ
> execution.
>
> > considering that those can be nested on top of rcuidle context ?
>
> As per nmi_enter() calling rcu_nmi_enter() I've always assumed that NMIs
> are fully covered by RCU.
>
> If this isn't so, RCU it terminally broken :-)
All RCU can do is respond to calls to rcu_nmi_enter() and rcu_nmi_exit().
It has not yet figured out how to force people to add these calls where
they are needed. ;-)
But yes, it would be very nice if architectures arranged things so
that all NMI handlers were visible to RCU. And we no longer have
half-interrupts, so maybe there is hope...
Thanx, Paul
next prev parent reply other threads:[~2020-02-11 13:03 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-02-10 22:06 [PATCH] tracing/perf: Move rcu_irq_enter/exit_irqson() to perf trace point hook Steven Rostedt
2020-02-11 0:30 ` Mathieu Desnoyers
2020-02-11 2:22 ` Steven Rostedt
2020-02-11 2:32 ` joel
2020-02-11 15:19 ` Mathieu Desnoyers
2020-02-11 12:00 ` Peter Zijlstra
2020-02-11 13:03 ` Paul E. McKenney [this message]
2020-02-11 13:16 ` Peter Zijlstra
2020-02-11 13:23 ` Paul E. McKenney
2020-02-11 14:10 ` Steven Rostedt
2020-02-11 11:49 ` Peter Zijlstra
2020-02-11 12:59 ` Paul E. McKenney
2020-02-11 13:10 ` Peter Zijlstra
2020-02-11 13:20 ` Paul E. McKenney
2020-02-11 14:05 ` Steven Rostedt
2020-02-11 15:05 ` Peter Zijlstra
2020-02-11 15:29 ` Peter Zijlstra
2020-02-11 16:16 ` Steven Rostedt
2020-02-11 15:06 ` Paul E. McKenney
2020-02-11 15:31 ` Peter Zijlstra
2020-02-11 15:40 ` Paul E. McKenney
2020-02-11 12:21 ` Peter Zijlstra
2020-02-11 14:10 ` Steven Rostedt
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=20200211130301.GH2935@paulmck-ThinkPad-P72 \
--to=paulmck@kernel.org \
--cc=gregkh@linuxfoundation.org \
--cc=gustavo@embeddedor.com \
--cc=jiangshanlai@gmail.com \
--cc=joel@joelfernandes.org \
--cc=josh@joshtriplett.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mathieu.desnoyers@efficios.com \
--cc=mingo@kernel.org \
--cc=peterz@infradead.org \
--cc=rostedt@goodmis.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).