I just built bk-1.1514 (which oddly enough that version I can''t find a reference to on bk-bits anymore, was that changeset backed out?) and I am noticing that domU''s cpus are getting timer ticks at 100Hz even when idle. I know I have observed a "no tick" behavior before. Was this change intentional? Thanks, -Andrew _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
> I just built bk-1.1514 (which oddly enough that version I > can''t find a reference to on bk-bits anymore, was that > changeset backed out?)It''s probably bk renumbering things -- you really need to refer to changesets by MD5KEY.> and I am noticing that domU''s cpus are > getting timer ticks at 100Hz even when idle. I know I have > observed a "no tick" behavior before. Was this change intentional?I bet I know what this is. Please can you try the following to see if it goes away: ''bk cset -x428f2c2a_3sOeZVGhQutbJc15aCU9g'' ChangeSet@1.1159.258.151, 2005-05-21 13:40:10+01:00, cl349@firebug.cl.cam.ac.uk Fix e1000 hang during reboot bug. Kudos to Steven Hand for analyzing. time.c: Fix bug in code which sets our one-shot timer when we decide to block during the idle loop. Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk> Thanks, Ian _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 23 May 2005, at 20:07, Andrew Theurer wrote:> I just built bk-1.1514 (which oddly enough that version I can''t find a > reference to on bk-bits anymore, was that changeset backed out?) and I > am noticing that domU''s cpus are getting timer ticks at 100Hz even when > idle. I know I have observed a "no tick" behavior before. Was this > change intentional?If it''s an SMP domU then it''s because we do not do tick holdoff for SMP guests. There are unresolved issues with how we interact with the RCU mechanisms in Linux 2.6. It is our plan to resolve these and hold off ticks for idle SMP guests just as we do for uniproc ones. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On Monday 23 May 2005 14:51, Ian Pratt wrote:> > I just built bk-1.1514 (which oddly enough that version I > > can''t find a reference to on bk-bits anymore, was that > > changeset backed out?) > > It''s probably bk renumbering things -- you really need to refer to > changesets by MD5KEY.OK, thanks.> > > and I am noticing that domU''s cpus are > > getting timer ticks at 100Hz even when idle. I know I have > > observed a "no tick" behavior before. Was this change intentional? > > I bet I know what this is. Please can you try the following to see if > it goes away: > > ''bk cset -x428f2c2a_3sOeZVGhQutbJc15aCU9g''I''ll give it a shot right now. FWIW, I did tack cpu util with the 100 Hz timer on 12 idle 4-way domU''s (on 2 cpu Xeon with HT). Attached is a text file with the cpu util. Each cpu burns about 1.5% just to maintain the idle domains. Not sure of the impact of a workload running in one of these while many other are idle, but I suspect it could be worse than 1.5% (will find out soon). -Andrew> > ChangeSet@1.1159.258.151, 2005-05-21 13:40:10+01:00, > cl349@firebug.cl.cam.ac.uk > Fix e1000 hang during reboot bug. Kudos to Steven Hand for > analyzing. time.c: > Fix bug in code which sets our one-shot timer when we decide to > block during > the idle loop. > Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk> > > Thanks, > Ian_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On Monday 23 May 2005 15:03, Keir Fraser wrote:> On 23 May 2005, at 20:07, Andrew Theurer wrote: > > I just built bk-1.1514 (which oddly enough that version I can''t > > find a reference to on bk-bits anymore, was that changeset backed > > out?) and I am noticing that domU''s cpus are getting timer ticks at > > 100Hz even when idle. I know I have observed a "no tick" behavior > > before. Was this change intentional? > > If it''s an SMP domU then it''s because we do not do tick holdoff for > SMP guests. There are unresolved issues with how we interact with the > RCU mechanisms in Linux 2.6. It is our plan to resolve these and hold > off ticks for idle SMP guests just as we do for uniproc ones.OK, saw the changes in time.c and now I understand what''s going on, thanks. -Andrew _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel