Re: [PATCH v4 0/4] x86/pvh: fix unbootable VMs again (PVH + KASAN)

[Mauricio Faria de Oliveira <mfo@igalia.com>]

Thanks for having a look at this.

On 2026-05-27 17:47, Borislav Petkov wrote:
> On Tue, May 26, 2026 at 11:52:32AM -0300, Mauricio Faria de Oliveira wrote:
>> The issue of unbootable VMs with CONFIG_PVH due to CONFIG_KASAN is back.
> 
> ... and we care about that particular configuration because...?

PVH and KASAN can be enabled independently for their own purposes
(startup entry point and debugging), but when combined, KASAN might
break booting via the PVH entry point. There are precedents for
disabling KASAN instrumentation in specific code paths, including this
one (see listed commits), so addressing this issue/regression seems
reasonable.

In this instance, the use case is to speed up the dev-test cycle when
debugging on VMs: PVH allows booting directly from vmlinux, avoiding the
need to build/compress and decompress bzImage, while KASAN helps detect
issues early.

> 
>>  arch/x86/boot/string.c               |  6 ++----
>>  arch/x86/include/asm/cpuid/api.h     |  2 +-
>>  arch/x86/include/asm/string.h        | 21 +-------------------
>>  arch/x86/include/asm/string_inline.h | 37 ++++++++++++++++++++++++++++++++++++
>>  arch/x86/platform/pvh/enlighten.c    |  3 ++-
>>  5 files changed, 43 insertions(+), 26 deletions(-)
> 
> I fail to see justification for those gymnastics.

Briefly, this is a side effect of reusing code from 'boot/string.c'
through <asm/string.h>, which introduces additional complications;
please see patch 1 for details.

These complications could be avoided by not reusing that code and
instead adding slightly duplicated code [1] directly to <asm/string.h>,
if that is preferred. 

Please let me know your thoughts.

Thanks,

[1]
https://lore.kernel.org/all/324ef97b16f52e0ccc72f6381d1b5dd2@igalia.com/

-- 
Mauricio