Re: [PATCH for-4.22 0/4] x86/kexec: Improvements for FRED

[Oleksii Kurochko <oleksii.kurochko@xxxxxxxxx>]

On 6/22/26 7:26 PM, Andrew Cooper wrote:
> On 19/03/2026 12:25 pm, Andrew Cooper wrote:
> > Kexec with FRED happens to already function.
> >
> > Looking at the logic, several unsafe actions are occuring.  Fix the worst of
> > them.
How critical are the remaining actions?

> > Andrew Cooper (4):
> >    x86/kexec: Stop hooking NMIs with trap_nop()
> >    x86/kexec: Fix and expands comments for kexec_reloc()
> >    x86/kexec: Invalidate the IDT earlier in kexec_reloc()
> >    x86/kexec: Disable FRED earlier in kexec_reloc()
> >
> >   xen/arch/x86/crash.c              | 20 +++++-----
> >   xen/arch/x86/x86_64/kexec_reloc.S | 62 ++++++++++++++++++++++---------
> >   2 files changed, 55 insertions(+), 27 deletions(-)
> This is fully reviewed/acked but fell through the cracks an I didn't get
> it committed in time.  Sorry.
>
> Please could I request a release ack?
>
> Now that FRED is active by default on AMD hardware in 4.22, a poorly
> timed NMI will break transitioning into the crash environment.  i.e.
> it's fixing a corner-case bug in a new feature for 4.22.
Sounds reasonable to me:

Release-Acked-by: Oleksii Kurochko <oleksii.kurochko@xxxxxxxxx>

Thanks.

~ Oleksii