[Qemu-devel] MTTCG Call Summary 10/8/2015

[Alex Bennée]

Hi,

Here is the summary of today's MTTCG call.

Fred:

Fred talked about the current status of his v7 patches. They are ready
to go save for two bugs that were triggered when running Alex's Debian
Jessie images. After discussion it seems there may be multiple bug
manifestations. They are:

  - Bug with non-DTB activated CPUs on the vexpress model

It seems in this case the CPUs not in the DTB will spin forever on a
region of memory waiting to be started up. The work around is to disable
the call to memory_region_rom_device_set_romd(). Alex expressed concern
there may be a deeper issue misunderstanding the locking requirements of
the memory dispatch code.

  - A segfault in the QEMU code when booting the image

This was not the failure case Alex had seen (he boots OK but see's segs
when running apt operations). This seems to be triggered from the TCG
generated code.

  - A crash in the startup code

The boot fails to get to the prompt

After some discussion Alex said he'd send details to Alvise to see which
bugs manifest in his setup. Alex encouraged Fred to post the v7 patches
to the list anyway in case something obvious was spotted by the
reviewers.

Alvise:

The non-MTCG slow-path patch set was posted to the list last week and
has already attracted some comments including a potential solution that
avoids updating all the backends. Alvise hopes to have the MTTCG version
out later this week based on Fred's latest patch series.

Alex:

Alex has sent out some drafts of his presentation to Fred and Alvise and
will mostly be spending this week polishing the presentation and getting
it ready for next week. There was some discussion about the barrier
tests which he's re-writing and finding the best way to trigger the
bugs (ARM TCG on ARM rather than x86 backend).

KVM Forum:

Alex, Claudio and Fred will be at KVM Forum next week. It was suggested
they meet up on the Tuesday to flesh out the demo options for the talk.
The consensus seemed to be to keep it simple but show host CPUs being
utilised and the improvement in performance.

If I've missed anything out then please reply and mention it.

-- 
Alex Bennée