------- Comment From dieg...@br.ibm.com 2017-11-27 06:16 EDT-------
(In reply to comment #11)
> The problem seems to be related to how qemu allocates the virtual cores and
> threads to the VM.
>
> Starting the VMs with the following command I cannot reproduce the issue:
>
> $ qemu-system-ppc64le -m 1024 -smp 16,threads=8 -nographic -net
> nic,model=virtio -net user,hostfwd=tcp::10022-:22 -drive
> file=./autopkgtest-xenial-ppc64el.img -enable-kvm
>
> As long the number of cpus and threads provided to the -smp parameter follow
> the topology of the p8 processor, making qemu use more than one core only if
> the number of threads is multiple of 8, I can't reproduce the hang. If the
> parameters provided makes qemu spread the virtual threads on more than one
> physical cpu without using all the 8 threads (e.g. '-smp 2', '-smp
> 4,threads=2', '-smp 8,threads=4') then the guest kernel hangs at some point.
>
> This is the environment being tested:
>
> Machine type-model: 8247-22L
>
> Host: Artful
> kernel 4.13.0-17-generic
> qemu 2.10+dfsg-0ubuntu3
>
> Guests: Xenial, Zesty and Artful with latest kernel from -updates.
>
> Could someone from IBM please confirm if this is a known or expected
> behavior? This issue seems to be around for a long time and it's odd that
> this hasn't come up before.

I was not able to reproduce this issue with the following command and
environment:

$ /usr/libexec/qemu-kvm -hda /home/libvirt-images/diegodo/diegodo-
ubuntu16.qcow2 -boot c -net nic -net user -m 512 -localtime -nographic
-smp 4,threads=2

HOST: Rhel 7.3
kernel: 4.13.0-rc4
qemu: 2.6.0

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1733864

Title:
  kernel 4.10.0-40 is hanging with a CPU soft lock

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu-power-systems/+bug/1733864/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to