kvm hangs on AMD
Bug #380650 reported by
Kurt Harders
This bug report is a duplicate of:
Bug #361754: guest needs to boot with clock=acpi_pm (older AMD freq scaling issues).
Edit
Remove
This bug affects 2 people
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| kvm (Ubuntu) |
New
|
Undecided
|
Unassigned | ||
Bug Description
Binary package hint: kvm
Host is Ubuntu server 9.04 with only ssh and kvm installed.
When running kvm on an AMD processor one virtual machine runs OK, the next machine(s) run an arbitrary time from some seconfs to several minutes. After that the process of kvm needs 100% cpu and does not react to any communication, neither ssh nor vnc. This happens when started from the command line or via virsh. If started via virsh no shutdown is possible. This happens with 9.04 guest, 8.04.1 guest, gentoo guest and xp guest. It does not depend on the driver for the network nor on bridging or user space NAT.
The problem disappeared after installing an Intel CPU with no changes to host or guests, just changing the mainboard.
Revision history for this message
| Bradley Hook (bdhook) wrote | #1 |
To post a comment you must log in.
This same behavior is happening for me with 9.04 hosts and vms. Using hardware acceleration on an AMD dual core processor. I had one running over the weekend that started doing this, and I can confirm that the VM doesn't completely hang, it just shoots to 100% CPU (in my case 100% of the one core it's allowed to use) and continues very very slowly (presumably with 120 second timeout for every operation). It does dump error messages to the console, but I neglected to record them before killing the instance. The first VM on the system has been running the whole time with no issues, only the second and third (which are running the same software) are having issues.