From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: kvm guest: hrtimer: interrupt too slow Date: Tue, 29 Sep 2009 15:47:07 +0200 Message-ID: <4AC20FDB.2050607@redhat.com> References: <4AC207B1.7020901@msgid.tls.msk.ru> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: kvm , Glauber Costa To: Michael Tokarev Return-path: Received: from mx1.redhat.com ([209.132.183.28]:52476 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751689AbZI2NrL (ORCPT ); Tue, 29 Sep 2009 09:47:11 -0400 In-Reply-To: <4AC207B1.7020901@msgid.tls.msk.ru> Sender: kvm-owner@vger.kernel.org List-ID: On 09/29/2009 03:12 PM, Michael Tokarev wrote: > Hello. > > I'm having quite an.. unusable system here. > It's not really a regresssion with 0.11.0, > it was something similar before, but with > 0.11.0 and/or 2.6.31 it become much worse. > > The thing is that after some uptime, kvm > guest prints something like this: > > hrtimer: interrupt too slow, forcing clock min delta to 461487495 ns > > after which system (guest) speeed becomes > very slow. The above message is from > 2.6.31 guest running wiht 0.11.0 & 2.6.31 > host. Before I tried it with 0.10.6 and > 2.6.30 or 2.6.27, and the delta were a > bit less than that: > > hrtimer: interrupt too slow, forcing clock min delta to 152222415 ns > hrtimer: interrupt too slow, forcing clock min delta to 93629025 ns > > etc. > > Before, guest was just somewhat slow. But > now, it reached the state when it's almost > unusable. > > Especially "nice" it's visible in xterm > (running on that guest): typing a single > char in xterm results in it being displayed > after 1..2 sec pause. With applications > like web browser (where all the font > rendering is done on the client and X > server receives bitmap instead of a > character code to draw, it isn't that > bad. > > I suspect it's only happening with small > network packets. > > The above message (with large min delta) > occured after about 20 hours uptime. Similar > (give or take) delays were observed previously > as well. That to say, -- it's not something > you notice immediately. > > The problem is quite consistent, that is, > different guests shows it sooner or later. > > Right now I'm running 2.6.31-amd64 host, > 32bit qemu-kvm-0.11.0, 2.6.31-i686 uniprocessor > guest, virtio networking and block devices, > kvm-clock is enabled. Similar configuration > were used before (with different versions). > The hardware is Phenom 9750 (4core) running > on Amd780g/sb700 chipset. > > Any hints on what to do with all this? > What happens if you use hpet or pmtimer as guest clocksource? Please post host /proc/cpuinfo. -- error compiling committee.c: too many arguments to function