[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <alpine.LFD.1.10.0809022239220.3243@apollo.tec.linutronix.de>
Date: Tue, 2 Sep 2008 23:13:22 +0200 (CEST)
From: Thomas Gleixner <tglx@...utronix.de>
To: Linus Torvalds <torvalds@...ux-foundation.org>
cc: Larry Finger <Larry.Finger@...inger.net>,
LKML <linux-kernel@...r.kernel.org>,
"Rafael J. Wysocki" <rjw@...k.pl>,
Alok Kataria <akataria@...are.com>,
Michael Buesch <mb@...sch.de>
Subject: Re: Regression in 2.6.27 caused by commit bfc0f59
On Tue, 2 Sep 2008, Linus Torvalds wrote:
>
> You're definitely right that this could easily be the _real_ problem.
> Especially as your TSC min value of 2160 is (a) pretty close to the
> expected time of a microsecond and (b) so stable that I actually do not
> believe that the PIT itself is at all emulated or the problem.
On that box, the PIT is probably real hardware or a damned good
emulation. When you look at the 10 loop values you see that it does
50% perfectly fine calibration loops. The others are just SMI
interruptions caused by random unknown crap in the BIOS.
> Btw - as to caring about the average value: that's pointless. If you only
> look at the average time the PIT read takes place, then it is going to
> approximate that "pit_count" thing in the end that I already did.
>
> Why? Because the average value should essentially end up being "(end_tsc -
> start_tsc) / pit_count". And if you just compare that to "min_tsc", then
> that should always be about a microsecond (on normal machines where the
> PIT is essentially on the old emulated internal "ISA" bus on the
> southbridge). So you end up with what I already posted, and you already
> dismissed.
>
> So average TSC is not any more interesting than "pit_count".
Yeah, you're right. Math is hard :)
Thanks,
tglx
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists