]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
x86/sev/vc: Fix EFI runtime instruction emulation
authorGerd Hoffmann <kraxel@redhat.com>
Thu, 26 Jun 2025 11:40:11 +0000 (13:40 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 20 Aug 2025 16:41:01 +0000 (18:41 +0200)
commite88695515ea9d7514f52e9439b49d6ab399bc26c
tree898cfa9d49977d3d7b008ce6adcf9ee0ad037c4d
parent6e7004b0d80b5311ff119dd488c92fc287a3392b
x86/sev/vc: Fix EFI runtime instruction emulation

[ Upstream commit 7b22e0432981c2fa230f1b493082b7e67112c4aa ]

In case efi_mm is active go use the userspace instruction decoder which
supports fetching instructions from active_mm.  This is needed to make
instruction emulation work for EFI runtime code, so it can use CPUID and
RDMSR.

EFI runtime code uses the CPUID instruction to gather information about
the environment it is running in, such as SEV being enabled or not, and
choose (if needed) the SEV code path for ioport access.

EFI runtime code uses the RDMSR instruction to get the location of the
CAA page (see SVSM spec, section 4.2 - "Post Boot").

The big picture behind this is that the kernel needs to be able to
properly handle #VC exceptions that come from EFI runtime services.
Since EFI runtime services have a special page table mapping for the EFI
virtual address space, the efi_mm context must be used when decoding
instructions during #VC handling.

  [ bp: Massage. ]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Link: https://lore.kernel.org/20250626114014.373748-2-kraxel@redhat.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
arch/x86/coco/sev/vc-handle.c