qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PULL 22/33] migration: remove the QEMUFileOps 'get_buffer' callback


From: Daniel P . Berrangé
Subject: Re: [PULL 22/33] migration: remove the QEMUFileOps 'get_buffer' callback
Date: Mon, 27 Jun 2022 15:41:57 +0100
User-agent: Mutt/2.2.6 (2022-06-05)

On Thu, Jun 23, 2022 at 03:13:36PM -0400, Peter Xu wrote:
> On Thu, Jun 23, 2022 at 09:47:51AM +0100, Daniel P. Berrangé wrote:
> > > Hmm, when I wanted to run the whole bunch of the migration-test again I
> > > found that precopy tls test hangs 
> > > (/x86_64/migration/precopy/unix/tls/psk).
> > > Though for this time it also hangs for me even with the master branch, so
> > > maybe not anything wrong with this specific pull req but still something
> > > needs fixing..
> > 
> > That pre-existing test has been runnnig by default in CI for a while
> > now, under different OS builds, so I'm surprised. Is there anything
> > especially unusual / different about your setup that could explain
> > why you see hang that we don't get anywhere else ?
> 
> TL;DR: I think it's not run in CI?
> 
> Please see ufd_version_check(), as when uffd not detected we'll skip the
> whole thing.

Our CI should be passing that check for the private runners eg

https://gitlab.com/qemu-project/qemu/-/jobs/2641920502

shows us running 35 tests

  2/178 qemu:qtest+qtest-x86_64 / qtest-x86_64/migration-test         OK        
     65.57s   35 subtests passed

but yes, the container based jobs are all skipping

> 
> We really need to apply this patch, soon-ish..
> 
>   https://lore.kernel.org/all/20210615175523.439830-2-peterx@redhat.com/
> 
> I can easily reproduce the hang on two x86_64 hosts I have, with current
> master commit (2b049d2c8dc01de750410f8f1a4eac498c04c723).  Or am I the only
> one?
> 
> So I think it also means we don't run migration unit tests on non-Linux
> OSes for sure because uffd was never there, meanwhile it also requires
> (mostly) root privilege even for Linux hosts so if the sysctl knob was not
> set properly (on sysctl.unprivileged_userfaultfd=1) the test can be skipped
> too.

It appears Fedora at least set unprivileged_userfaultfd=1 by default,
hence why I've never seen it skipped.

With regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|




reply via email to

[Prev in Thread] Current Thread [Next in Thread]