Re: [PATCH RFC] kexec: add option to print previous kernel release after kexec

[Breno Leitao <leitao@debian.org>]

Hello Pasha,


On Tue, Dec 30, 2025 at 10:03:38AM -0500, Pasha Tatashin wrote:
> On Tue, Dec 30, 2025 at 9:13 AM Breno Leitao <leitao@debian.org> wrote:
> >
> > Introduce a new Kconfig option, CONFIG_KEXEC_SHOW_PREVIOUS_RELEASE,
> 
> I don't think we need a config for this. Let's just show the previous
> kernel version unconditionally when KHO is enabled.
> 
> > which enables the storage and display of the previous kernel's release
> > version across kexec boots when using Kexec Handover (KHO).
> >
> > Motivation
> > ==========
> >
> > Kernel bugs that depend on the version of the previously running kernel
> > are not uncommon, especially in scenarios where a buggy kernel kexecs
> > into a new one and the issue only manifests in the subsequent kernel.
> >
> > Recent examples include:
> >
> >  * eb2266312507 ("x86/boot: Fix page table access in 5-level to 4-level paging transition")
> >  * 77d48d39e991 ("efistub/tpm: Use ACPI reclaim memory for event log to avoid corruption")
> >  * 64b45dd46e15 ("x86/efi: skip memattr table on kexec boot")
> >
> > As kexec-based reboots become more widespread among Linux users, these
> > issues are becoming increasingly prominent.
> >
> > When such crashes occur, correlating them to the previous kernel version
> > is challenging, especially at scale, where the problem may only appear in
> > specific cases and it is hard to correlate.
> >
> > With the introduction of Kexec Handover (KHO), we now have a reliable
> > mechanism to pass information between kernels. This presents an
> > opportunity to address the problem by carrying the previous kernel's
> > release version to the next kernel and printing it at boot time.
> 
> LUO added the number of times a machine has been live updated to help
> debugging problems that are related to live update, perhaps, in
> addition to the previous kernel version, we should also add a number
> of times the machine has been kexec'd. It is useful when a bug
> manifests only after a number of kexec's which is also sometimes a
> case.

Thanks for bringing this up, I am on the same page as you here.

In fact, this was on my TODO list, but I didn't want to send in this RFC
to avoid too much changes/code.

I will wait a bit more and add the number of kexec'ed kernels in the
next revision.

Thanks for your input,
--breno