Bugzilla will be upgraded to version 5.0 on December 2, 2018. The outage period for the upgrade will start at 0:00 UTC and have a duration of 12 hours
Bug 1055002 - RFE: default to cpu host-model (when it's ready)
RFE: default to cpu host-model (when it's ready)
Product: Virtualization Tools
Classification: Community
Component: virt-manager (Show other bugs)
Unspecified Unspecified
unspecified Severity unspecified
: ---
: ---
Assigned To: Cole Robinson
: 1468016 (view as bug list)
Depends On:
  Show dependency treegraph
Reported: 2014-01-17 17:51 EST by Cole Robinson
Modified: 2018-10-04 14:46 EDT (History)
11 users (show)

See Also:
Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2018-10-04 14:46:36 EDT
Type: Bug
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)

  None (edit)
Description Cole Robinson 2014-01-17 17:51:10 EST
Eventually I think we should default to cpu mode=host-model for new VMs, or at least provide the option. Upstream now has a 'default CPU' section in the preferences dialog, but it's only doing 'copy host cpu' and not host-model at the moment. Reason being, host-model implementation has known issues:

host-model doesn't match reality, can cause guest panics: https://bugzilla.redhat.com/show_bug.cgi?id=870071
host-model persistence: https://bugzilla.redhat.com/show_bug.cgi?id=1054935

RHEL bugs about the steps needed to get to a proper working order:

libvirt: host-model avx crash: https://bugzilla.redhat.com/show_bug.cgi?id=1014682
libvirt: query qemu: https://bugzilla.redhat.com/show_bug.cgi?id=824989
qemu: CPU probe interface: https://bugzilla.redhat.com/show_bug.cgi?id=824987
libvirt: use -cpu enforce: https://bugzilla.redhat.com/show_bug.cgi?id=822148

And some misc libvirt bugs that come into play for virt-manager:

cpu map API support: https://bugzilla.redhat.com/show_bug.cgi?id=916786
cpu model checking busted for tcg: https://bugzilla.redhat.com/show_bug.cgi?id=1002066

So basically, we need to wait for a lot of that to be solved before we can safely change the default.
Comment 1 Frank Ch. Eigler 2015-05-06 13:01:41 EDT
Would this be a good place to ask that virt-manager's preferences permit "host-passthrough" as a valid option for cpu-model?
Comment 2 Cole Robinson 2015-05-06 13:08:40 EDT
You can type 'host-passthrough' into the UI box and virt-manager will do the right thing. However host-passthrough has certain supportability issues with libvirt, so we don't expose it as a proper option in the UI
Comment 3 Pavel Valena 2016-11-22 12:48:31 EST
Hello, what is the status on this?

Also, is 'host-passthrough' safe with libvirt now?
Comment 4 Cole Robinson 2017-07-10 19:47:47 EDT
*** Bug 1468016 has been marked as a duplicate of this bug. ***
Comment 5 Eduardo Habkost 2017-07-10 19:58:52 EDT
(In reply to Pavel Valena from comment #3)
> Hello, what is the status on this?
> Also, is 'host-passthrough' safe with libvirt now?

(Replying 8 months later)

I consider host-passthrough/"-cpu host" safe today, but it will prevent live-migration.  If live-migration is not required, it is the best way to make of use all features supported by the host CPU + hypervisor.

"host-model" is now supposed to be safe (at least the bugs mentioned in comment #0 are all fixed now).
Comment 6 Pavel Valena 2017-07-11 18:09:22 EDT
Thanks for information!
Comment 7 Cole Robinson 2018-10-04 14:46:36 EDT
If libvirt/qemu combo is new enough, we use host-model as the default now:

commit 469fed08a04777d72cbb73e1dc948cffe31f7858
Author: Cole Robinson <crobinso@redhat.com>
Date:   Thu Oct 4 12:23:32 2018 -0400

    guest: Default to cpu mode=host-model if libvirt is new enough

Note You need to log in before you can comment on or make changes to this bug.