Re: [TEST/REVIEW] cpu time accounting patch, step 2

From: Kris Kennaway <kris_at_obsecurity.org>
Date: Fri, 10 Feb 2006 19:45:40 -0500
On Fri, Feb 10, 2006 at 07:24:51PM -0500, Kris Kennaway wrote:
> On Fri, Feb 10, 2006 at 08:33:01AM +0000, Poul-Henning Kamp wrote:
> > In message <43EC2C4C.5000709_at_rogers.com>, Mike Jakubik writes:
> > >Poul-Henning Kamp wrote:
> > >> 	http://phk.freebsd.dk/patch/cputime.patch 
> > >>
> > >> This patch enables the new cpu time accounting method for i386 with
> > >> TSC, all amd64 and sparc64 (?) and i386 with Geode cpu.
> > >>   
> > >
> > >I got a bunch of these during the first bootup.
> > >
> > >--
> > >calcru: runtime went backwards from 1227916 usec to 1227883 usec for pid 
> > >30 (pagezero)
> > 
> > Yes, that will happen right now until the calibration of the TSC's
> > max frequency is completed.
> > 
> > I'm analyzing various solutions to that.
> 
> I'm getting it with your committed patch too..are you aware of it?

In addition to the continuous spam of these calcru messages on the
console (>10/sec), I'm getting the following:

Feb 11 09:41:00 dosirak kernel: pid 8831 (bsdtar), uid 0, was killed: exceeded maximum CPU limit
Feb 11 09:41:06 dosirak kernel: pid 8850 (bsdtar), uid 0, was killed: exceeded maximum CPU limit
Feb 11 09:41:09 dosirak kernel: pid 9479 (umount), uid 0, was killed: exceeded maximum CPU limit

which suggests CPU time limits are also hosed (these processes ran for
a few seconds at best, not an hour, which is the ulimit -t value).

Again, this is with the code you already committed to CVS.

Kris


Received on Fri Feb 10 2006 - 23:45:42 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:52 UTC