[Bug 704537] New: qemu-kvm crashes the first time a guest VM is started after booting
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c0 Summary: qemu-kvm crashes the first time a guest VM is started after booting Classification: openSUSE Product: openSUSE 11.4 Version: Final Platform: x86-64 OS/Version: openSUSE 11.4 Status: NEW Severity: Critical Priority: P5 - None Component: Other AssignedTo: bnc-team-screening@forge.provo.novell.com ReportedBy: gvenkat@attglobal.net QAContact: qa@suse.de Found By: --- Blocker: --- Created an attachment (id=438879) --> (http://bugzilla.novell.com/attachment.cgi?id=438879) Log User-Agent: Mozilla/5.0 (X11; Linux i686; rv:5.0) Gecko/20100101 Firefox/5.0 I use virsh to start Windows guest and use SPICE client to access the desktop. If I start the Windows VM domain the first time after a reboot and connect to it via SPICE, it crashes within a minute or two without doing anything. A restart of the domain in virsh works without any problems. Logs and version info given below. This problem started to appear since one of the builds of kvm in the last few months Reproducible: Always Steps to Reproduce: 1. Reboot host 2. Start Windows XP (or 7) VM domain from virsh 3. Connect from remote SPICE client 4. Wait for the crash/shutdown of guest VM Actual Results: The guest VM is shutdown. There are no errors in the guest VM logs. It is like a forced power off as far as the guest is concerned. Expected Results: KVM should not crash the guest VM KVM compiled from source at http://download.opensuse.org/repositories/Virtualization/openSUSE_11.4/src/ using just --with spice parameter to rpmbuild Current software versions: Kernel 2.6.37.6-0.5-desktop KVM kvm-0.14.1.0-108.1.x86_64 libvirt libvirt-0.9.3-138.2.x86_64 spice-client spice-client-0.8.1-1.1.x86_64 soice server libspice-server1-0.8.1-1.1.x86_64 Note the very large memory allocation to the qxl graphics driver in the qemu-kvm command line in the attachment. Don't know if this is relevant. Please do keep in mind that this always happens ONLY on the first start of a VM, not subsequent starts so the bug can only be in things that persist over multiple invocations which rules out many of the components. The only log entry tied to this event is the virtio trying to map MMIO memory which doesn't appear on subsequent starts where it doesn't crash. This is the most obvious clue. Perhaps there is a timeout error when this is attempted the first time but is done and available for subsequent starts which may indicate a problem with upstream qemu in the kernel. I am just guessing here. Don't have any experience with the code or architecture of these components. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c zj jia <zjjia@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |zjjia@novell.com AssignedTo|bnc-team-screening@forge.pr |brogers@novell.com |ovo.novell.com | -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c Bruce Rogers <brogers@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P3 - Medium -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c1 --- Comment #1 from G Venk <gvenkat@attglobal.net> 2011-09-30 01:39:36 UTC --- Just updating the report to indicate that the bug persists in the latest software builds as indicate below. Also, it is not necessary to connect to the guest for it to crash. Just starting the VMs as part of the host boot and not doing anything with it will make the VMs crash within a minute or two. Subsequent starts of the VM show absolutely no crash problem and has been tested with up-times of weeks. The last log message when it crashes is always qemu-kvm: virtio: trying to map MMIO memory This message does not appear otherwise. Current software versions: Kernel 2.6.37.6-0.7-desktop KVM kvm-0.15.0-119.3.x86_64 libvirt libvirt-0.9.6-167.1.x86_64 spice-client spice-client-0.8.2-4.1.x86_64 soice server libspice-server1-0.8.2-4.1.x86_64 -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c2 --- Comment #2 from G Venk <gvenkat@attglobal.net> 2011-09-30 02:21:18 UTC --- The bug has been reported in other distros as well, see https://bugs.launchpad.net/qemu/+bug/818673 https://bugzilla.redhat.com/show_bug.cgi?id=727034 for additional information of the exact same bug on a different distro. So it doesn't appear to be a distro issue. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c3 --- Comment #3 from Andreas Färber <afaerber@suse.com> 2011-10-12 14:35:13 UTC --- Thanks a lot for reporting this. Apparently Ubuntu and Red Hat are still investigating the root causes. As soon as things are more clear we'll look into the issue again. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c Bruce Rogers <brogers@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|brogers@suse.com |kvm-bugs@forge.provo.novell | |.com -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c4 Bruce Rogers <brogers@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO CC| |brogers@suse.com InfoProvider| |gvenkat@attglobal.net --- Comment #4 from Bruce Rogers <brogers@suse.com> 2012-02-15 13:26:45 UTC --- Could you retest with the packages in Virtualization. We've updated the versions of the related packages, and things should be substantially different. Thanks. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c5 G Venk <gvenkat@attglobal.net> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW InfoProvider|gvenkat@attglobal.net | --- Comment #5 from G Venk <gvenkat@attglobal.net> 2012-02-15 21:41:07 UTC --- I just updated all packages and can confirm the problem exists. It is not just on boot up of host, it happens on every alternate boot of the Guest! Kernel 2.6.37.6-0.11-desktop kvm-1.0-139.4.x86_64 libvirt-0.9.9-196.4.x86_64 libspice-server1-0.10.0-31.1.x86_64 On client (but this doesn't really matter as the Guest crashes even when never connected) spice-client-0.10.0-31.1.i586 The problem from what I have been able to dig up is likely with the virtio drivers provided for Windows guests (this problem happens only with Windows guests - XP to Windows 7). Virtio drivers are used in my configuration for the disk, serial-io and network. Most likely it is the disk virtio that is causing the problem but can also be the serial io driver. Not sure when and how the Novell virtio drivers for Windows provided with the distro are updated. They look outdated in my install. Somewhere in the redhat bugzilla area, I read that many problems have been fixed in the windows virtio drivers but it is not clear whether these fixes are in the open source domain or proprietary to RedHat. Not sure who propagates these Windows binaries to OpenSuse if they are opensource or who updates them for opensuse and where any such updates are to be found (other than being included in the original distro). If anyone knows of a place with updated virtio drivers for Windows, I can try them out and see if they fix the problem. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c6 --- Comment #6 from G Venk <gvenkat@attglobal.net> 2012-05-05 03:02:58 UTC --- Just wanted to update that the bug was in the para-virtualized disk driver for Windows as I suspected above and has been found and fixed by Redhat in their distribution. It turns out that Avast anti-virus is the one that triggers the bug. It explains why I saw it in alternate reboots. Avast does more extensive checks at startup based on when it last did the checks so the second reboot will skip those checks and avoid the bug. https://bugzilla.redhat.com/show_bug.cgi?id=771390 Can anyone confirm whether this fix (and future versions of para-virtualization drivers for Windows guests) will be available for OpenSuse? This is important since without support for these drivers, support for Window guests with para-virtualization is pretty much dead and so OpenSuse can no longer be considered a candidate as an open source linux host for KVM based virtualization. Would appreciate if someone with information or ability to find out can respond as to where OpenSuse stands regarding these drivers. Thanks -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c7 --- Comment #7 from G Venk <gvenkat@attglobal.net> 2012-05-05 03:13:03 UTC --- If anyone is interested I found the opensource code for these drivers with the fix at http://alt.fedoraproject.org/pub/alt/virtio-win/latest/images/src/ Compiled binaries are also available there for users who may have the same problem. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c9 Kirk Allan <kallan@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO InfoProvider| |gvenkat@attglobal.net --- Comment #9 from Kirk Allan <kallan@suse.com> 2012-05-09 23:22:42 UTC --- Is there a consistent way to reproduce this problem? From the RedHat bug, it sounds like it is an issue with how avast antivirus is making requests larger than what the virtio block driver is expecting. If you have some good steps on how to reproduce, that would be helpful. Thanks. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c10 G Venk <gvenkat@attglobal.net> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW InfoProvider|gvenkat@attglobal.net | --- Comment #10 from G Venk <gvenkat@attglobal.net> 2012-05-10 00:06:40 UTC --- The steps reported in the first post IS the consistent way to reproduce it except for the additional information that Avast be running in the guest. This is why the crash would happen only in Windows guests. My VM would consistently crash every alternate boot after a period of time varying from 1 min to 10 min or so from start. As I understand it, Avast just tickles the bug in the virtio block driver not that it is doing anything wrong. The virtio block driver needs to handle it when guest applications make large requests. Not sure what the protocol is but I suspect it is to return some correct error code so the guest application can then ask for a smaller block. Avast just happens to THE application in this case that was making the large request. And the block driver crashed on such requests bringing down the whole VM. I can also confirm that after I installed the "fixed" driver in the link above in the Guest, the VM is no longer crashing. The diff in the source code for the driver from the previous version should indicate what the fix was in the block driver. Suspect it is a fairly simple fix. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c11 --- Comment #11 from Kirk Allan <kallan@suse.com> 2012-05-17 22:07:11 UTC --- Created an attachment (id=491437) --> (http://bugzilla.novell.com/attachment.cgi?id=491437) virtio block driver for windows xp I have been able to reproduce this issue on Windows XP but not on Windows 7 32 bit vm. Please install VMDP-WIN-2.0 in the winxp and win7 vms: http://download.novell.com/Download?buildid=31EiPNl27zQ~ For the winxp vm, copy the attached vrtioblk.sys file into \windows\system32\drivers. Reboot the vms and see if the issue still happens or not in either vm and report. Thanks. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c Kirk Allan <kallan@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO InfoProvider| |gvenkat@attglobal.net -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c12 G Venk <gvenkat@attglobal.net> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW InfoProvider|gvenkat@attglobal.net | --- Comment #12 from G Venk <gvenkat@attglobal.net> 2012-05-24 03:08:13 UTC --- Thanks for the followup. The VMDP-WIN-2.0 is commercial software with it's own EULA tied to SuSE commercial subscription, not part of OpenSuse distribution and so I will not be able to install it to try. Can you let me know if this set of drivers is available as part of any OpenSuse distribution version under a non-commercial license? Is there a plan to update the drivers provided with the OpenSuse distributions (which are really old)? If not, then I am afraid I will not be able to help you with the debugging/testing. Whatever Novell decides to do is fine with me but I can only help maintain the non-commercial OpenSuse distribution. Hope you understand. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c13 Bruno Friedmann <bruno@ioda-net.ch> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO InfoProvider| |gvenkat@attglobal.net --- Comment #13 from Bruno Friedmann <bruno@ioda-net.ch> 2013-10-02 08:17:42 UTC --- VMDP-WIN-2.0 ( now 2.1 are freely available, the support is commercial ) In the meantime you could also use the fedora iso driver http://alt.fedoraproject.org/pub/alt/virtio-win/stable/ or win-virtio-drivers.iso present in /usr/share/qemu-kvm I didn't see any trouble with all of them and windows xp or windows 7 (32 & 64bits) Are you still able to reproduce it with new builds ? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=704537 https://bugzilla.novell.com/show_bug.cgi?id=704537#c14 Kirk Allan <kallan@suse.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |RESOLVED InfoProvider|gvenkat@attglobal.net | Resolution| |FIXED --- Comment #14 from Kirk Allan <kallan@suse.com> 2014-03-31 15:45:42 UTC --- The fixed windows xp virtio block driver is now in vmdp 2.1. Since vmdp, as noted above is a commercial package, the open source drivers from comment 13 can be used instead. Marking bug as fixed. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply@novell.com