From: Don Zickus <dzickus@redhat.com>
To: Robert Richter <robert.richter@amd.com>
Cc: Ingo Molnar <mingo@elte.hu>,
Peter Zijlstra <peterz@infradead.org>,
"gorcunov@gmail.com" <gorcunov@gmail.com>,
"fweisbec@gmail.com" <fweisbec@gmail.com>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
"ying.huang@intel.com" <ying.huang@intel.com>,
"ming.m.lin@intel.com" <ming.m.lin@intel.com>,
"yinghai@kernel.org" <yinghai@kernel.org>,
"andi@firstfloor.org" <andi@firstfloor.org>,
"eranian@google.com" <eranian@google.com>
Subject: Re: [PATCH] perf, x86: catch spurious interrupts after disabling counters
Date: Fri, 24 Sep 2010 14:11:43 -0400 [thread overview]
Message-ID: <20100924181143.GQ26290@redhat.com> (raw)
In-Reply-To: <20100924100345.GE13563@erda.amd.com>
On Fri, Sep 24, 2010 at 12:03:45PM +0200, Robert Richter wrote:
> On 23.09.10 23:18:34, Don Zickus wrote:
>
> > > I was able to duplicate the problem and can confirm this patch fixes the
> > > issue for me. I tried poking around (similar to things Robert probably
> > > did) and had no luck. Something just doesn't make sense, but I guess for
> > > now this patch is good enough for me. :-)
> >
> > Ah ha! I figured out what the problem was, need to disable the pmu while
> > processing the nmi. :-) Finally something simple in this crazy unknown
> > NMI spree.
> >
> > Oh yeah and trace_printk is now my new favorite tool!
> >
> > From: Don Zickus <dzickus@redhat.com>
> > Date: Thu, 23 Sep 2010 22:52:09 -0400
> > Subject: [PATCH] x86, perf: disable pmu from counting when processing irq
> >
> > On certain AMD and Intel machines, the pmu was left enabled
> > while the counters were reset during handling of the NMI.
> > After the counters are reset, code continues to process an
> > overflow. During this time another counter overflow interrupt
> > could happen because the counter is still ticking. This leads to
> > an unknown NMI.
>
> I don't like the approach of disabling all counters in the nmi
> handler. First, it stops counting and thus may falsify
> measurement. Second, it introduces much overhead doing a rd-/wrmsrl()
> for each counter.
Ok. good points. Though we may have to re-visit the perf_event_intel.c
code as my patch was based on 'well they did it there, we can do it here'
approach.
>
> Does your patch below solve something that my patch doesn't?
Well, two things I was trying to solve with your approach is the extra NMI
that is generated, and the fact that you may falsely eat an unknown NMI
(because you don't clear cpuc->running except in the special case).
Yeah, I know the heurestics from your patch a couple of weeks ago will
make the case of falsely eating an unknown NMI next to impossible. But I
was trying to limit the number seen.
For example, the back-to-back nmi problem we dealt with a couple of weeks
ago, only had a 'special case' that had to be dealt with every once in a
while (depending on the level of perf activity). Now I can see unknown
NMIs every time there is a pmu_stop() issued. I was trying to figure out
a way to cut down on that noise.
I came up with a couple of approaches but both involve another rdmsrl in
the handle_irq path. I thought I would toss these ideas out for
conversation.
the first approach gets rid of the extra nmi by waiting until the end to
re-write the counter, but adds a second rdmsrl to resync the period in the
case where pmu_stop is not called.
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 48c6d8d..1642f48 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1175,11 +1175,22 @@ static int x86_pmu_handle_irq(struct pt_regs *regs)
handled++;
data.period = event->hw.last_period;
- if (!x86_perf_event_set_period(event))
- continue;
+ /*
+ * if period is over, process the overflow
+ * before reseting the counter, otherwise
+ * a new overflow could occur before the
+ * event is stopped
+ */
+ if (local64_read(&hwc->period_left) <= 0) {
+ if (perf_event_overflow(event, 1, &data, regs)) {
+ x86_pmu_stop(event, 0);
+ continue;
+ }
+ /* if the overflow doesn't stop the event, resync */
+ x86_perf_event_update(event);
+ }
- if (perf_event_overflow(event, 1, &data, regs))
- x86_pmu_stop(event, 0);
+ x86_perf_event_set_period(event);
}
if (handled)
and the second approach, accepts the fact that we will get another unknown
NMI but check for it first (after stopping) and clear the cpuc->running
bit if we didn't overflow yet.
diff --git a/arch/x86/kernel/cpu/perf_event.c b/arch/x86/kernel/cpu/perf_event.c
index 18c8856..bfb05da 100644
--- a/arch/x86/kernel/cpu/perf_event.c
+++ b/arch/x86/kernel/cpu/perf_event.c
@@ -1189,8 +1189,12 @@ static int x86_pmu_handle_irq(struct pt_regs *regs)
if (!x86_perf_event_set_period(event))
continue;
- if (perf_event_overflow(event, 1, &data, regs))
+ if (perf_event_overflow(event, 1, &data, regs)) {
x86_pmu_stop(event, 0);
+ rdmsrl(hwc->event_base + idx, val);
+ if (val & (1ULL << (x86_pmu.cntval_bits - 1)))
+ __clear_bit(idx, cpuc->running);
+ }
}
if (handled)
After dealing with the Intel folks on another thread about how they are
finding more ways to detect and report hardware errors with NMIs, that I was
getting nervous about generating so many false NMIs and accidentally
eating a real one.
Cheers,
Don
next prev parent reply other threads:[~2010-09-24 18:12 UTC|newest]
Thread overview: 117+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-09-02 19:07 [PATCH 0/3 v2] nmi perf fixes Don Zickus
2010-09-02 19:07 ` [PATCH 1/3] perf, x86: Fix accidentally ack'ing a second event on intel perf counter Don Zickus
2010-09-02 19:26 ` Cyrill Gorcunov
2010-09-02 20:00 ` Don Zickus
2010-09-02 20:36 ` Cyrill Gorcunov
2010-09-03 7:10 ` [tip:perf/urgent] " tip-bot for Don Zickus
2010-09-03 7:39 ` Yinghai Lu
2010-09-03 15:00 ` Don Zickus
2010-09-03 17:15 ` Yinghai Lu
2010-09-03 18:35 ` Don Zickus
2010-09-03 19:24 ` Yinghai Lu
2010-09-03 20:10 ` Don Zickus
2010-10-04 23:24 ` Yinghai Lu
2010-10-11 20:25 ` Don Zickus
2010-09-02 19:07 ` [PATCH 2/3] perf, x86: Try to handle unknown nmis with an enabled PMU Don Zickus
2010-09-03 7:11 ` [tip:perf/urgent] " tip-bot for Robert Richter
2010-09-02 19:07 ` [PATCH 3/3] perf, x86: Fix handle_irq return values Don Zickus
2010-09-03 7:10 ` [tip:perf/urgent] " tip-bot for Peter Zijlstra
2010-09-10 11:41 ` [PATCH 0/3 v2] nmi perf fixes Peter Zijlstra
2010-09-10 12:10 ` Stephane Eranian
2010-09-10 12:13 ` Stephane Eranian
2010-09-10 13:27 ` Don Zickus
2010-09-10 14:46 ` Ingo Molnar
2010-09-10 15:17 ` Robert Richter
2010-09-10 15:58 ` Peter Zijlstra
2010-09-10 16:41 ` Ingo Molnar
2010-09-10 16:42 ` Ingo Molnar
2010-09-10 16:37 ` Ingo Molnar
2010-09-10 16:51 ` Ingo Molnar
2010-09-10 15:56 ` [PATCH] x86: fix duplicate calls of the nmi handler Robert Richter
2010-09-10 16:15 ` Peter Zijlstra
2010-09-11 9:41 ` Ingo Molnar
2010-09-11 11:44 ` Robert Richter
2010-09-11 12:45 ` Ingo Molnar
2010-09-12 9:52 ` Robert Richter
2010-09-13 14:37 ` Robert Richter
2010-09-14 17:41 ` Robert Richter
2010-09-15 16:20 ` [PATCH] perf, x86: catch spurious interrupts after disabling counters Robert Richter
2010-09-15 16:36 ` Stephane Eranian
2010-09-15 17:00 ` Robert Richter
2010-09-15 17:32 ` Stephane Eranian
2010-09-15 18:44 ` Robert Richter
2010-09-15 19:34 ` Cyrill Gorcunov
2010-09-15 20:21 ` Stephane Eranian
2010-09-15 20:39 ` Cyrill Gorcunov
2010-09-15 22:27 ` Robert Richter
2010-09-16 14:51 ` Frederic Weisbecker
2010-09-15 16:46 ` Cyrill Gorcunov
2010-09-15 16:47 ` Stephane Eranian
2010-09-15 17:02 ` Cyrill Gorcunov
2010-09-15 17:28 ` Robert Richter
2010-09-15 17:40 ` Cyrill Gorcunov
2010-09-15 22:10 ` Robert Richter
2010-09-16 6:53 ` Cyrill Gorcunov
2010-09-16 17:34 ` Peter Zijlstra
2010-09-17 8:51 ` Robert Richter
2010-09-17 9:14 ` Peter Zijlstra
2010-09-17 13:06 ` Stephane Eranian
2010-09-20 8:41 ` Robert Richter
2010-09-24 0:02 ` Don Zickus
2010-09-24 3:18 ` Don Zickus
2010-09-24 10:03 ` Robert Richter
2010-09-24 13:38 ` Stephane Eranian
2010-09-30 12:33 ` Peter Zijlstra
2010-09-24 18:11 ` Don Zickus [this message]
2010-09-24 10:41 ` [tip:perf/urgent] perf, x86: Catch " tip-bot for Robert Richter
2010-09-29 12:26 ` Stephane Eranian
2010-09-29 12:53 ` Robert Richter
2010-09-29 12:54 ` Robert Richter
2010-09-29 13:13 ` Stephane Eranian
2010-09-29 13:28 ` Stephane Eranian
2010-09-29 15:01 ` Robert Richter
2010-09-29 15:12 ` Robert Richter
2010-09-29 15:27 ` Cyrill Gorcunov
2010-09-29 15:33 ` Stephane Eranian
2010-09-29 15:45 ` Cyrill Gorcunov
2010-09-29 15:51 ` Cyrill Gorcunov
2010-09-29 16:32 ` Robert Richter
2010-09-29 16:48 ` Cyrill Gorcunov
2010-09-29 16:00 ` Stephane Eranian
2010-09-29 17:09 ` Robert Richter
2010-09-29 17:41 ` Cyrill Gorcunov
2010-09-29 18:12 ` Don Zickus
2010-09-29 19:42 ` Stephane Eranian
2010-09-29 20:03 ` Don Zickus
2010-09-30 9:12 ` Robert Richter
2010-09-30 19:44 ` Don Zickus
2010-10-01 7:17 ` Robert Richter
[not found] ` <AANLkTimUyLaVaBigjm0-CwRsdh4UXWDiss2ffX53S+k_@mail.gmail.com>
2010-10-01 11:53 ` Stephane Eranian
2010-10-02 9:35 ` Robert Richter
2010-10-04 8:53 ` Stephane Eranian
2010-10-04 9:07 ` Andi Kleen
2010-10-04 17:28 ` Stephane Eranian
2010-09-29 16:31 ` Robert Richter
2010-09-29 16:22 ` Robert Richter
2010-09-29 19:01 ` Don Zickus
2010-09-29 13:39 ` Robert Richter
2010-09-29 13:56 ` Stephane Eranian
2010-09-29 14:00 ` Stephane Eranian
2010-10-02 9:50 ` Robert Richter
2010-10-02 17:40 ` Stephane Eranian
2010-09-29 15:02 ` Cyrill Gorcunov
2010-09-16 17:42 ` [PATCH] x86: fix duplicate calls of the nmi handler Peter Zijlstra
2010-09-16 20:18 ` Stephane Eranian
2010-09-17 7:09 ` Peter Zijlstra
2010-09-17 0:13 ` Huang Ying
2010-09-17 7:52 ` Peter Zijlstra
2010-09-17 8:13 ` Robert Richter
2010-09-17 8:37 ` Cyrill Gorcunov
2010-09-17 8:47 ` Huang Ying
2010-09-10 13:34 ` [PATCH 0/3 v2] nmi perf fixes Peter Zijlstra
2010-09-10 13:52 ` Peter Zijlstra
2010-09-13 8:55 ` Cyrill Gorcunov
2010-09-13 9:54 ` Stephane Eranian
2010-09-13 10:07 ` Cyrill Gorcunov
2010-09-13 10:10 ` Stephane Eranian
2010-09-13 10:12 ` Cyrill Gorcunov
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=20100924181143.GQ26290@redhat.com \
--to=dzickus@redhat.com \
--cc=andi@firstfloor.org \
--cc=eranian@google.com \
--cc=fweisbec@gmail.com \
--cc=gorcunov@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=ming.m.lin@intel.com \
--cc=mingo@elte.hu \
--cc=peterz@infradead.org \
--cc=robert.richter@amd.com \
--cc=ying.huang@intel.com \
--cc=yinghai@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 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).