RHBA-2009:1488 - Bug Fix Advisory

Topic

Updated kvm packages that resolved two issues are now available.

Description

KVM (Kernel-based Virtual Machine) is a full virtualization solution for
Linux on x86 hardware. KVM can run multiple virtual machines running
unmodified Linux or Windows images. Each virtual machine has private
virtualized hardware: a network card, disk, graphics adapter, etc.

These updated packages fix the following bugs:

• the pthread_cond_timedwait time out was not properly handled.

Consequently, under some loads, some KVM guests stopped responding to
commands from the management interface. (Note: the reproducer was a host
running around 300 KVM guests with each guest consuming around 50% of their
virtual CPU. On this setup, some guests became non-responsive after several
hours.) With this update the time outs are handled properly and KVM guests
remain responsive, as expected. (BZ#526244)

• some Linux-based guests that used virtio virtual block devices aborted

during installation, returning the error message: "unhandled vm exit: 0x31
vcpu_id 0".

Using an interface other than virtio for the guest virtual disk was a work
around documented in the Red Hat Enterprise Linux 5.4 Technical Notes Known
Issues for KVM. The work around was associated with BZ#518081, the original
Bugzilla report for this issue.

http://www.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/5.4/html/Technical_Notes/Known_Issues-kvm.html

With this update, the underlying issue (stale EPTP-tagged mappings possibly
being used when a virtual CPU or vcpu migrated to a different Physical CPU
or pcpu) has been addressed and the work around is no longer necessary:
Linux-based guests using virtio virtual block devices no longer abort
during installation. (BZ#527192)

All users of kvm are advised to upgrade to these updated packages, which
resolve this issue.

Solution

Before applying this update, make sure all previously-released errata
relevant to your system have been applied.

The update is available via the Red Hat Network (RHN). To apply package
updates via the RHN see http://kbase.redhat.com/faq/docs/DOC-11259

After installation, use the following procedure to ensure the fixes take
effect.

1. Stop all KVM guest virtual machines (VMs).

2. Reboot the host (hypervisor) system.

3. Restart the KVM guest VMs.

If re-booting the host system is not possible, an alternative to step 2
above is as follows:

2a. Determine which of the following modules are running using the "lsmod"
command: kvm, ksm, kvm-intel or kvm-amd.

2b. Remove each running module using the "modprobe -r [module]" command.
(If kvm is running, for example, the "modprobe -r kvm" command removes it.)

2c. Reload each running module using the "modprobe [module]" command. (For
example, to reload kvm, use the "modprobe kvm" command.)

Note: removing and reloading a running module can only be done as the
superuser.

Affected Products

• Red Hat Enterprise Linux Server 5 x86_64
• Red Hat Enterprise Linux for x86_64 - Extended Update Support 5.4 x86_64
• Red Hat Enterprise Linux Workstation 5 x86_64

Fixes

• BZ - 526244 - KVM QEMU:RHEL OS VM's became not-responding and stopped answering monitor Commands.
• BZ - 527192 - vmentry fails with ept misconfig

CVEs

(none)

References

• http://www.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/5.4/html/Technical_Notes/Known_Issues-kvm.html