On Wed, 2011-11-23 at 22:42 +0100, Alexandre Vaissière wrote: > Le 23/11/2011 15:08, Ian Campbell a écrit : > >>> If you boot the same kernel natively what happens when you run pygrub? > >>> (this should work natively, although obviously you won't be able to > >>> actually start a VM). > >> Yes it does work: I see the grub interface. > > What happens if you pass "noxsave" to the native kernel? > > > > [...] > > If I am right then using "noxsave" on native should demonstrate the > > issue as well and show that there is a bug in libc. > > > you are right. With noxsave, I have same behaviour of pygrub under a > native kernel.
Good. Well not "good" but you know what I mean. > > However that doesn't rule out a bug in the hypervisor as well since it > > should support xsave... except I've just noticed that xsave is > > deliberately disabled in Xen 4.1! > > > > It has been re-enabled in xen-unstable with the comment "re-enable xsave > > by default now that it supports live migration" (so it will work in 4.2 > > when it is released). This suggests that you could try with "xsave=1" on > > your hypervisor command line (but that migration won't work). > > > With xsave=1 pygrub works ! However xend segfaults later while starting > the domU :/ > Will try to get a stack, but as you state below: > > > Which seems to already be reported and fixed upstream: > > http://sourceware.org/bugzilla/show_bug.cgi?id=13007 > > as well as against glibc in Debian http://bugs.debian.org/646549. The > > fix is fixed in glibc-package SVN so ought to be fixed with the next > > upload (2.13-22). > > > Maybe I should wait for this libc release before going further. I think that would be a good idea. I think this bug can be closed and a new one opened if the xend seg fault persists after the libc upgrade so I am doing that now. Ian. -- Ian Campbell Novinson's Revolutionary Discovery: When comes the revolution, things will be different -- not better, just different. -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
