RHEL 8.8 and derivatives: suspend and live migration issue

RHEL 8.8 and derivatives: suspend and live migration issue

[David Morel]

Hello,

This mail is meant to be more of a summary plus a heads up, I don't think
there is any action needed on it.

A user reported [1] a crash when live migrating a VM running Rocky Linux
8.8. After investigation it appears to be happening:
- for RHEL derivatives when migrating from 8.7 to 8.8.
- on suspend/resume
- on live migration

With a callstack we provided, Roger Pau Monné quickly identified the
issue as being the addition of:
- x86/idt: Remove address operator on function machine_check() [2]
While missing:
- x86/xen: Split HVM vector callback setup and interrupt gate allocation [3]

Version-wise:
- 4.18.0-466.el8 will start having the issue as it integrate [2].
- 4.18.0-477 are versions used in 8.8 releases, these versions will
  therefore have the issue.
- 4.18.0-488 that can be found in CentOS 8 Stream integrates [3] and
  works fine.

According to the RPM changelog I think the bugzilla entry at RH should
be the number 2187810 [4], but that is not publicly available.

The -488 versions are not yet available for the derivative
distributions. We do not think there is much to do, raising this to
Rocky Linux or Alma directly likely won't change their kernel upgrade
plan.


[1] https://xcp-ng.org/forum/topic/7420/live-migrate-of-rocky-linux-8-8-vm-crashes-reboots-vm
[2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/xen/events/events_base.c?id=fbaed278a3cc72a46aadae667b8c6754b78640a6
[3] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/xen/events/events_base.c?id=a0bb51f2638e0810c347024679239fd10a8f7990
[4] https://bugzilla.redhat.com/show_bug.cgi?id=2187810

-- 
David Morel

Re: RHEL 8.8 and derivatives: suspend and live migration issue

[Roger Pau Monné]

On Fri, Jun 09, 2023 at 04:07:54PM +0200, David Morel wrote:
> Hello,
> 
> This mail is meant to be more of a summary plus a heads up, I don't think
> there is any action needed on it.
> 
> A user reported [1] a crash when live migrating a VM running Rocky Linux
> 8.8. After investigation it appears to be happening:
> - for RHEL derivatives when migrating from 8.7 to 8.8.
> - on suspend/resume
> - on live migration
> 
> With a callstack we provided, Roger Pau Monné quickly identified the
> issue as being the addition of:
> - x86/idt: Remove address operator on function machine_check() [2]

No, the above reference is incorrect, the issue is caused by the
backport of [0]:

06184325a1cc x86/idt: Annotate alloc_intr_gate() with __init

Without having backported the mentioned Xen commit (x86/xen: Split HVM
vector callback setup and interrupt gate allocation).

Regards, Roger.

[0] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=06184325a1cce27a02a688d98740f90fe06e0133