Hi Keir,
I did a new round of test for save/restore w/ r14854 for bug verification, and
your fixes have helped hvm save/restore a lot actually from my testing result.
Two critical issues for hvm save/restore have been fixed, they are:
1) restore hvm guest will cause dom0 crash. Now the dom0 crash issue has been
fixed.
2) cp or scp files in guest with VBD or VNIF inserted and save&restore
many times on pae host have problems. Now cp or scp will continue to work after
restore many times.
However, there are some issues remained for save/restore, they are:
1) save ia32e hvm guest with pv drivers has problem, the guest is 512M, but xm
save command does not return, the guest is still alive and xm dmesg shows:
(XEN) hvm.c:754:d4 HVM vcpu 4:0 bad hypercall 29.
And s/r an hvm guest without pv drivers is okay.
2) sdl guest will be restored as vnc one.
Best regards,
Fan
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
Keir Fraser
2007-Apr-16 10:45 UTC
Re: [Xen-devel] save/restore quality status report w/ r14854
On 16/4/07 08:19, "Zhao, Fan" <fan.zhao@intel.com> wrote:> 1) save ia32e hvm guest with pv drivers has problem, the guest is 512M, but > xm save command does not return, the guest is still alive and xm dmesg shows: > (XEN) hvm.c:754:d4 HVM vcpu 4:0 bad hypercall 29. > And s/r an hvm guest without pv drivers is okay.Hopefully fixed by r14861. Dexuan''s fix for PV guest destroy is also checked in, as r14860.> 2) sdl guest will be restored as vnc one.We don''t usually test with sdl enabled, so I won''t be able to fix this one myself. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Zhao, Fan
2007-Apr-17 07:46 UTC
[Xen-devel] live migration hvm guest with pv drivers for ia32/pae status report
Hi Keir, Thanks for your reply, I haven''t verified the 1) issue because r14861 is still in the staging tree, and I will test it after it has been in xen-unstable. I did some further test for hvm guest with pv drivers on pae platform live migration w/ r14854 since I believe s/r hvm guest with pv drivers is stable from my previous test. The detail report is: 1) live migrate ia32 up hvm guest with vbd while doing kernel build for many times -- pass 2) live migrate ia32 up hvm guest with vnif while scping files from another machine from many times -- pass 3) live migrate ia32 smp hvm guest with vbd while doing kernel build for many times -- pass 4) live migrate ia32 smp hvm guest with vnif while scping files from another machine from many times -- fail After several times'' live migration, it is possible that the guest is still alive on the original machine, and there is a domain whose state is -bp--- on the dest machine, and I also met the failure that the scp is stalled after several times live migration in other conditions. BTW, I found that the scp speed is very slow(about 3MB/s) for me after vnif is inserted, not sure if the live migration failure relates with this. Logs are attached. Best regards, Fan -----Original Message----- From: Keir Fraser [mailto:Keir.Fraser@cl.cam.ac.uk] Sent: 2007年4月16日 18:45 To: Zhao, Fan; xen-devel@lists.xensource.com Subject: Re: [Xen-devel] save/restore quality status report w/ r14854 On 16/4/07 08:19, "Zhao, Fan" <fan.zhao@intel.com> wrote:> 1) save ia32e hvm guest with pv drivers has problem, the guest is 512M, but > xm save command does not return, the guest is still alive and xm dmesg shows: > (XEN) hvm.c:754:d4 HVM vcpu 4:0 bad hypercall 29. > And s/r an hvm guest without pv drivers is okay.Hopefully fixed by r14861. Dexuan''s fix for PV guest destroy is also checked in, as r14860.> 2) sdl guest will be restored as vnc one.We don''t usually test with sdl enabled, so I won''t be able to fix this one myself. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel