From: Ingo Molnar <mingo@elte.hu>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>,
Thomas Gleixner <tglx@linutronix.de>,
LKML <linux-kernel@vger.kernel.org>,
Alok Kataria <akataria@vmware.com>,
Arjan van de Veen <arjan@infradead.org>,
"H. Peter Anvin" <hpa@zytor.com>,
Peter Zijlstra <a.p.zijlstra@chello.nl>
Subject: Re: [RFC patch 0/4] TSC calibration improvements
Date: Thu, 4 Sep 2008 23:33:50 +0200 [thread overview]
Message-ID: <20080904213350.GA15678@elte.hu> (raw)
In-Reply-To: <alpine.LFD.1.10.0809041403450.3117@nehalem.linux-foundation.org>
* Linus Torvalds <torvalds@linux-foundation.org> wrote:
> On Thu, 4 Sep 2008, Ingo Molnar wrote:
> >
> > hm, unless i'm missing something i think here we still have a small
> > window for an SMI or some virtualization delay to slip in and cause
> > massive inaccuracy: if the delay happens _after_ the last
> > pit_expect_msb() and _before_ the external get_cycles() call. Right?
>
> Yes. I had the extra pit_expect_msb() originally, but decided that
> basically a single-instruction race for somethign that ran without any
> MSI for 15ms was a bit pointless.
the race is wider than that i think: all it takes an SMI at the last PIO
access, so the window should be 1 usec, against a 15000 usecs period.
That's 1 out of 15,000 boxes coming up with totally incorrect
calibration.
we also might have a very theoretical race of an SMI taking exactly 65
msecs so that the whole PIT wraps around and fools the fastpath - the
chance for that would be around 1:300 - assuming we only have to hit the
right MSB with a ~200 usecs precision). That assumes equal distribution
of SMI costs which they certainly dont have - most of them are much less
than 60 msecs. So i dont think it's an issue in practice - on real hw.
But it's still a possibility unless i'm missing something. We could
protect against that case by reading the IRQ0-pending bit and making
sure it's not pending after we have done the closing TSC readout.
> But adding another pit_expect_msb() is certainly not wrong.
ok, i kept that bit.
Ingo
next prev parent reply other threads:[~2008-09-04 21:34 UTC|newest]
Thread overview: 52+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-09-04 15:18 [RFC patch 0/4] TSC calibration improvements Thomas Gleixner
2008-09-04 15:18 ` [RFC patch 1/4] x86: TSC: define the PIT latch value separate Thomas Gleixner
2008-09-04 15:18 ` [RFC patch 2/4] x86: TSC: separate hpet/pmtimer calculation out Thomas Gleixner
2008-09-04 15:18 ` [RFC patch 3/4] x86: TSC: use one set of reference variables Thomas Gleixner
2008-09-04 15:18 ` [RFC patch 4/4] x86: TSC make the calibration loop smarter Thomas Gleixner
2008-09-04 15:36 ` [RFC patch 0/4] TSC calibration improvements Ingo Molnar
2008-09-04 15:45 ` Linus Torvalds
2008-09-04 16:00 ` Ingo Molnar
2008-09-04 16:21 ` Linus Torvalds
2008-09-04 16:36 ` Ingo Molnar
2008-09-04 17:41 ` Linus Torvalds
2008-09-04 18:07 ` Alan Cox
2008-09-04 18:26 ` Linus Torvalds
2008-09-04 18:30 ` H. Peter Anvin
2008-09-04 20:09 ` Linus Torvalds
2008-09-04 20:43 ` Ingo Molnar
2008-09-04 20:52 ` Ingo Molnar
2008-09-04 21:09 ` Linus Torvalds
2008-09-04 21:21 ` Ingo Molnar
2008-09-04 21:30 ` Linus Torvalds
2008-09-04 21:34 ` Linus Torvalds
2008-09-04 21:39 ` Ingo Molnar
2008-09-04 21:33 ` Ingo Molnar [this message]
2008-09-05 22:18 ` Alok Kataria
2008-09-05 22:34 ` Linus Torvalds
2008-09-06 20:03 ` Thomas Gleixner
2008-09-06 20:29 ` Linus Torvalds
2008-09-06 20:37 ` Thomas Gleixner
2008-09-06 20:50 ` Linus Torvalds
2008-09-06 20:55 ` Linus Torvalds
2008-09-06 21:15 ` Thomas Gleixner
2008-09-06 21:22 ` Linus Torvalds
2008-09-06 21:30 ` Thomas Gleixner
2008-09-06 22:40 ` Ingo Molnar
2008-09-06 20:58 ` Thomas Gleixner
2008-09-06 21:10 ` Linus Torvalds
2008-09-07 6:01 ` Willy Tarreau
2008-09-06 20:52 ` Thomas Gleixner
2008-09-06 20:59 ` Linus Torvalds
2008-09-06 21:07 ` Thomas Gleixner
2008-09-06 21:15 ` Linus Torvalds
2008-09-06 21:26 ` Thomas Gleixner
2008-09-06 21:32 ` Linus Torvalds
2008-09-04 20:53 ` Linus Torvalds
2008-09-04 21:38 ` Alok Kataria
2008-09-04 21:52 ` Linus Torvalds
2008-09-04 22:09 ` Alok Kataria
2008-09-04 17:39 ` Alok Kataria
2008-09-04 17:53 ` Linus Torvalds
2008-09-04 18:31 ` Alok Kataria
2008-09-04 18:34 ` H. Peter Anvin
2008-09-04 21:00 ` Valdis.Kletnieks
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=20080904213350.GA15678@elte.hu \
--to=mingo@elte.hu \
--cc=a.p.zijlstra@chello.nl \
--cc=akataria@vmware.com \
--cc=alan@lxorguk.ukuu.org.uk \
--cc=arjan@infradead.org \
--cc=hpa@zytor.com \
--cc=linux-kernel@vger.kernel.org \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.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