Xen / libvirt - "End of file while reading data: Input/output error" when attempting to start a domain

Question

I'm getting the following string of errors when I attempt to start a domain using virsh:

virsh # start vimes
error: Failed to start domain vimes
error: End of file while reading data: Input/output error
error: Failed to reconnect to the hypervisor

Tailing the log files as this happen, I get the following output:

/var/log/libvirt/libxl/vimes.log

libxl: debug: libxl_create.c:1362:do_domain_create: ao 0x7f4e1c0017e0: create: how=(nil) callback=(nil) poller=0x7f4e1c0019c0

dmesg

[  852.559719] libvirtd[2187]: segfault at ffffffffffffffff ip 00007f4e45472614 sp 00007f4e3e6b60d8 error 7 in libc-2.19.so[7f4e453ed000+19e000]

/var/log/libvirt/libxl/libxl-driver.log

xc: debug: hypercall buffer: cache current size:4
xc: debug: hypercall buffer: cache hits:13 misses:4 toobig:1
xc: debug: hypercall buffer: total allocations:18 total releases:18
xc: debug: hypercall buffer: current allocations:0 maximum allocations:4
xc: debug: hypercall buffer: cache current size:4
xc: debug: hypercall buffer: cache hits:13 misses:4 toobig:1
xc: debug: hypercall buffer: total allocations:18 total releases:18
xc: debug: hypercall buffer: current allocations:0 maximum allocations:4
xc: debug: hypercall buffer: cache current size:4
xc: debug: hypercall buffer: cache hits:13 misses:4 toobig:1

I'm at a bit of a loss about how to proceed. It's not clear from the error messages what's actually going wrong. Host and guest are both running OpenSuse 13.2,

Solution

Seems to me, we are facing the same problem.

Please look at Bugzilla – Bug 909547 https://bugzilla.opensuse.org/show_bug.cgi?id=909547

and temporary partial workaround (until host reboot !!) described in opensuse forum https://forums.opensuse.org/showthread.php/502307-openSUSE-13-1-gt-13-2-upgrade-issues-with-xen-libvirt-and-virt-manager

Workaround:

1. As root make sure that there is nothing listed in the Virtual Machine Manager, and confirm there are no vm's running using xl list or virsh list.

2. Go to Yast > Software > Software Management, search for libvirt. Delete libvirt, libvirt-daemon, libvirt-glib-1_0-0, libvirt-python, and perl-Sys-Virt, when prompted with conflicts, use the first option to deinstall the conflicts.

3. Backup or delete /usr/share/virt-manager, and it's contents. (The folder will be recreated upon re-installation.)
4. Reboot using xen hypervisor, and login as root. (As a rule I have not been doing this using the desktop kernel)
5. Follow step 2 expect this time install the listed packages.

6. In Yast > Virtualization > Install Hypervisor and Tools tick only xen tools, then Ok.

   If at this point you would happen to reboot. It will break the system and you get the EOF error trying to install a virtual machine.

7. Go to the Launcher > Applications > System > Virtualization > Virtual Machine Manager. There should be nothing listed except the local host (xen).

8. Now you have two choices to installing a virtual machine by either clicking the "Create a new virtual machine" button which I believe uses virt-install, or clicking on the arrow next to the button, then clicking on vm-install. (I've been using the latter)

Comments

[Michael Hampton]

Edit /etc/libvirt/libvirtd.conf and change log_level to 1. Then reload libvirtd and try again. You should get a much more detailed and hopefully helpful log.

[Adam Luchjenbroers]

I've set that and restarted libvirtd, but it doesn't appear to have any affect on the output. I can get more output with 'LIBVIRT_DEBUG=1 virsh start vimes' though. I've put the dump from that up on Pastebin: pastebin.com/xYQV8n72