Re: [Xen-devel] [xen-4.6-testing test] 138333: regressions - FAIL

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 25.06.19 at 16:12, <ian.jackson@xxxxxxxxxx> wrote:
> Ian Jackson writes ("Re: [xen-4.6-testing test] 138333: regressions - FAIL"):
>> Ian Jackson writes ("Re: [xen-4.6-testing test] 138333: regressions - FAIL"):
>> > Ian Jackson writes ("Re: [xen-4.6-testing test] 138333: regressions - 
>> > FAIL"):
>> > > Ian Jackson writes ("Re: [xen-4.6-testing test] 138333: regressions - 
>> > > FAIL"):
>> > > > These all have `qemut' in common.
>> ...
>> > I'm trying a test with 4.7's version of qemu trad.
>> 
>> This does not work.  4.7's qemu trad doesn't build because of tools
>> library reorganisation.  Reverting those changes to 4.7 produces a
>> qemu trad that is identical to 4.6's.  So the regression is not in
>> qemu.
>> 
>> I suspect a firmware or hvmloader problem.
>> 
>> This is blocking us getting a push for the Xen 4.8 stable branches:
> 
> These have not had a push for, in the case of 4.9, 133 days.

Yes, I had noticed this too. Embarrassing.

> Unless someone explains to me a plan for how to get 4.6 to actually
> work again well enough to test 4.7, or some other better proposal, I
> intend to force push 4.6 at the end of this week.

I've taken a look. The guests now triple fault during BIOS initialization:

(d1218) 18124 bytes of ROMBIOS high-memory extensions:
(d1218)   Relocating to 0xfc001000-0xfc0056cc ... done
...
(XEN) d1218v0 Triple fault - invoking HVM shutdown action 1
(XEN) *** Dumping Dom1218 vcpu#0 state: ***
(XEN) ----[ Xen-4.6.6  x86_64  debug=y  Not tainted ]----
(XEN) CPU:    0
(XEN) RIP:    0008:[<00000000fc00424c>]

[Note in particular this address.]

(XEN) RFLAGS: 0000000000010086   CONTEXT: hvm guest (d1218v0)
(XEN) rax: 00000000fc004369   rbx: 00000000fc0040e9   rcx: 0000000000000002
(XEN) rdx: 00000000fc004307   rsi: 0000000000000000   rdi: 0000000000000020
(XEN) rbp: 000000000009eed2   rsp: 000000000009ee96   r8:  0000000000000000
(XEN) r9:  0000000000000000   r10: 0000000000000000   r11: 0000000000000000
(XEN) r12: 0000000000000000   r13: 0000000000000000   r14: 0000000000000000
(XEN) r15: 0000000000000000   cr0: 0000000000000011   cr4: 0000000000000000
(XEN) cr3: 0000000000000000   cr2: 0000000000000000
(XEN) ds: 0018   es: 0018   fs: 0000   gs: c900   ss: 0018   cs: 0008

I wouldn't be surprised if the rombios build is broken - did you happen
to compare those binaries? Otoh I can't seem to spot any fixes in
master that would look like possibly addressing build issues with a
newer tool chain (other than cases where the build itself would fail).

Irrespective of this I'm not really opposed to a force push as you've
suggested, despite being afraid that this may hide an actual issue.
That's even more so that by now 4.7 has gone out of security
support, and hence we only need it now for 4.8's -prev tests.

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel