I started to work on the fix for the PoD, only to discover that just starting a normal HVM on recent -unstable build (no PoD enabled), the domain crashes with invalid guest state. I did a binary search, and found that the change which breaks things was introduced sometime between CS 19089 and 19095 I''m still looking into it, but if anyone else has any ideas, they''d be appreciated. I''m attaching a console log and cpuinfo. The chip is an older one, a dual-core Intel Conroe. -George _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Keir Fraser
2009-Jan-28 11:46 UTC
[Xen-devel] Re: VM Entry fails in changesets since 19095
It''s going to be a slightly subtle one since it works for and for our automated tests. It would be worth making sure you''ve done a ''make clean'' in the xen directory -- I recently changed the dependency system, and I''m still not 100% sure of it yet. I actually got a broken image myself somehow, which was fixed by ''make clean'', but I haven''t tracked down the issue as yet. -- Keir On 28/01/2009 11:05, "George Dunlap" <George.Dunlap@eu.citrix.com> wrote:> I started to work on the fix for the PoD, only to discover that just > starting a normal HVM on recent -unstable build (no PoD enabled), the > domain crashes with invalid guest state. > > I did a binary search, and found that the change which breaks things > was introduced sometime between CS 19089 and 19095 > > I''m still looking into it, but if anyone else has any ideas, they''d be > appreciated. > > I''m attaching a console log and cpuinfo. The chip is an older one, a > dual-core Intel Conroe. > > -George_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
George Dunlap
2009-Jan-28 12:23 UTC
Re: [Xen-devel] Re: VM Entry fails in changesets since 19095
I just did a "make mrproper" and re-built. Still crashes. I assume the automated test suites include a test with a 32-bit dom0 and a 64-bit hypervisor (which is my configuration)? I narrowed it down to cs 18090. With build of 18090 (plus a backport of 18095 so that dom0 boots) the domain will crash, but with a build of 18089 (plus a backport of 18095 for good measure) the domain will boot fine. Going to dig a little deeper... -George On Wed, Jan 28, 2009 at 11:46 AM, Keir Fraser <keir.fraser@eu.citrix.com> wrote:> It''s going to be a slightly subtle one since it works for and for our > automated tests. It would be worth making sure you''ve done a ''make clean'' in > the xen directory -- I recently changed the dependency system, and I''m still > not 100% sure of it yet. I actually got a broken image myself somehow, which > was fixed by ''make clean'', but I haven''t tracked down the issue as yet. > > -- Keir > > On 28/01/2009 11:05, "George Dunlap" <George.Dunlap@eu.citrix.com> wrote: > >> I started to work on the fix for the PoD, only to discover that just >> starting a normal HVM on recent -unstable build (no PoD enabled), the >> domain crashes with invalid guest state. >> >> I did a binary search, and found that the change which breaks things >> was introduced sometime between CS 19089 and 19095 >> >> I''m still looking into it, but if anyone else has any ideas, they''d be >> appreciated. >> >> I''m attaching a console log and cpuinfo. The chip is an older one, a >> dual-core Intel Conroe. >> >> -George > > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
George Dunlap
2009-Jan-28 12:26 UTC
Re: [Xen-devel] Re: VM Entry fails in changesets since 19095
Er sorry... s/18/19/g. So that''s cs 19090, with 19095 backported that doesn''t work, and 19089 that does work. -George On Wed, Jan 28, 2009 at 12:23 PM, George Dunlap <George.Dunlap@eu.citrix.com> wrote:> I just did a "make mrproper" and re-built. Still crashes. > > I assume the automated test suites include a test with a 32-bit dom0 > and a 64-bit hypervisor (which is my configuration)? > > I narrowed it down to cs 18090. With build of 18090 (plus a backport > of 18095 so that dom0 boots) the domain will crash, but with a build > of 18089 (plus a backport of 18095 for good measure) the domain will > boot fine. > > Going to dig a little deeper... > > -George > > On Wed, Jan 28, 2009 at 11:46 AM, Keir Fraser <keir.fraser@eu.citrix.com> wrote: >> It''s going to be a slightly subtle one since it works for and for our >> automated tests. It would be worth making sure you''ve done a ''make clean'' in >> the xen directory -- I recently changed the dependency system, and I''m still >> not 100% sure of it yet. I actually got a broken image myself somehow, which >> was fixed by ''make clean'', but I haven''t tracked down the issue as yet. >> >> -- Keir >> >> On 28/01/2009 11:05, "George Dunlap" <George.Dunlap@eu.citrix.com> wrote: >> >>> I started to work on the fix for the PoD, only to discover that just >>> starting a normal HVM on recent -unstable build (no PoD enabled), the >>> domain crashes with invalid guest state. >>> >>> I did a binary search, and found that the change which breaks things >>> was introduced sometime between CS 19089 and 19095 >>> >>> I''m still looking into it, but if anyone else has any ideas, they''d be >>> appreciated. >>> >>> I''m attaching a console log and cpuinfo. The chip is an older one, a >>> dual-core Intel Conroe. >>> >>> -George >> >> >> >> _______________________________________________ >> Xen-devel mailing list >> Xen-devel@lists.xensource.com >> http://lists.xensource.com/xen-devel >> >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel