Re: [xen-4.3-testing test] 63948: regressions - FAIL

[Ian Campbell <ian.campbell@citrix.com>]

On Wed, 2015-11-11 at 03:58 -0700, Jan Beulich wrote:
> > > > On 10.11.15 at 18:59, <osstest-admin@xenproject.org> wrote:
> > flight 63948 xen-4.3-testing real [real]
> > http://logs.test-lab.xenproject.org/osstest/logs/63948/ 
> > 
> > Regressions :-(
> > 
> > Tests which did not succeed and are blocking,
> > including tests which could not be run:
> >  test-amd64-amd64-migrupgrade 21 guest-migrate/src_host/dst_host fail
> > REGR. vs. 63212
> 
> This having failed for quite some time, I've finally looked more closely
> and found
> 
> Nov 10 14:36:16.949051 (XEN) vmce.c:88: PV restore: unsupported MCA
> capabilities 0x1000802 for d1:v0 (supported: 0)
> 
> to be the reason for the EPERM here
> 
> xc: error: Couldn't set extended vcpu0 info (1 = Operation not
> permitted): Internal error
> 
> Taking apart the value, it is MCG_SER_P | MCG_TES_P (the low 8 bits
> get masked out anyway), which is in line with 4.2's GUEST_MCG_CAP.
> Hence I would guess that previous successful runs of this test would
> have been on Intel systems only; I can't see how this test would ever
> succeed on AMD ones.

FWIW you can find the history of any given test at a URL like:
http://logs.test-lab.xenproject.org/osstest/results/history/test-amd64-amd64-migrupgrade/xen-4.3-testing.html

Figuring out the arch of the machines is a bit of a faff, especially since
some of the relevant logs no longer exist. From
http://logs.test-lab.xenproject.org/osstest/results/host/huxelrebe0.html
http://logs.test-lab.xenproject.org/osstest/results/host/godello0.html
http://logs.test-lab.xenproject.org/osstest/results/host/pinot0.html

I found recent logs which confirm (via the serial log):
Huxelrebe:
    CPU0: Intel(R) Xeon(R) CPU E3-1225 v3 @ 3.20GHz stepping 03
Godello:
    CPU0: Intel(R) Xeon(R) CPU E3-1220 v3 @ 3.10GHz stepping 03
Pinot:
    CPU0: AMD Opteron(tm) Processor 3350 HE               stepping 00

With Huxelrebe passing and godello and pinot failing this doesn't seem to
correlate with your investigation though.

>  Considering that 4.3 is out of maintenance, I
> think the only reasonable change to avoid endless failure here is to
> limit this test to Intel systems for this version.

Aside from the above I don't think osstest is currently aware of the vendor
of the processors (although I can certainly think of several reasons it
should be).

But given this is a new test case I would be happy, I think, to restrict it
to only go back as far as the earliest release which was in maintenance at
the time the test was introduced (August this year), or maybe (if something
just dropped out of maintenance recently) just the ones maintained today
(since it took a while for the test case to "bed in" and be made working on
some of the older ones). FWIW the last related fix I see in osstest was
early October.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel