From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1763023AbZEHTDV (ORCPT ); Fri, 8 May 2009 15:03:21 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1757870AbZEHTDF (ORCPT ); Fri, 8 May 2009 15:03:05 -0400 Received: from e35.co.us.ibm.com ([32.97.110.153]:48868 "EHLO e35.co.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754799AbZEHTDD (ORCPT ); Fri, 8 May 2009 15:03:03 -0400 Subject: Re: [RFC][PATCH] tsc_khz= boot option to avoid TSC calibration variance From: john stultz To: George Spelvin Cc: linux-kernel@vger.kernel.org, mingo@elte.hu, tglx@linutronix.de, ulrich.windl@rz.uni-regensburg.de, williams@redhat.com, zippel@linux-m68k.org In-Reply-To: <20090508121913.9039.qmail@science.horizon.com> References: <20090508121913.9039.qmail@science.horizon.com> Content-Type: text/plain Date: Fri, 08 May 2009 12:02:52 -0700 Message-Id: <1241809372.7297.3.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.24.3 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 2009-05-08 at 08:19 -0400, George Spelvin wrote: > > To mitigate this, I wanted to provide a tsc_khz= boot option. This would > > allow users to set the tsc_khz value at boot-up, assuming they are > > within 1Mhz of the calibrated value (to protect against bad values). > > Once the tsc_khz value is set in grub, the box will always boot with the > > same value, so the NTP drift value prior to reboot will still be correct > > after rebooting. > > A run-time adjustable would be more convenient, but this is simple and works. > > The 1 MHz tolerance, however, isn't implemented right. I can't quote > figure out what you were trying to do; was that (x + (x/2))/1000 trying > to round the /1000 properly? That should be (x + 1000/2)/1000 in that > case, which I normally write as (x/500 + 1)/2; > > But in any case, no equality comparison can possibly work; there's > always a case where the measured value rounds to k and the correct > value rounds to k+1. Also, 1 MHz is a pretty wide tolerance on > sub-GHz processors. I'd suggest the following: Ah. Indeed you're right. Thanks for catching this. Your version is much better. -john > diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c > index d57de05..d7ab640 100644 > --- a/arch/x86/kernel/tsc.c > +++ b/arch/x86/kernel/tsc.c > @@ -825,15 +825,42 @@ static void __init init_tsc_clocksource(void) > clocksource_register(&clocksource_tsc); > } > > +unsigned long tsc_khz_specified; > +static int __init tsc_khz_specified_setup(char *str) > +{ > + tsc_khz_specified = simple_strtoul(str, NULL, 0); > + return 1; > +} > + > +__setup("tsc_khz=", tsc_khz_specified_setup); > + > + > void __init tsc_init(void) > { > u64 lpj; > int cpu; > + long difference; > > if (!cpu_has_tsc) > return; > > tsc_khz = calibrate_tsc(); > + > + /* > * * If the calibrated TSC freq and user specified TSC freq > * * are close enough, pick the what the user told us. Reject > * * obviously bogus values to make the option safe to use. > + */ > + difference = tsc_khz - tsc_khz_specified; > + if (difference < 0) > + difference = -difference; > + if (difference <= tsc_khz >> 10) { /* 1/1024 = 976 ppm */ > + printk(KERN_INFO "Using user defined TSC freq: %lu.%03lu MHz\n", > + tsc_khz_specified/1000, > + tsc_khz_specified%1000); > + tsc_khz = tsc_khz_specified; > + } > + > cpu_khz = tsc_khz; > > if (!tsc_khz) {