arm64: efi: Force the use of SetVirtualAddressMap() on Altra machines
authorArd Biesheuvel <ardb@kernel.org>
Thu, 10 Nov 2022 09:36:20 +0000 (10:36 +0100)
committerArd Biesheuvel <ardb@kernel.org>
Thu, 10 Nov 2022 22:14:14 +0000 (23:14 +0100)
commit550b33cfd445296868a478e8413ffb2e963eed32
tree20293f6a8aa8916a87ef2a433a1a7fc4c3044556
parent23715a26c8d812912a70c6ac1ce67af649b95914
arm64: efi: Force the use of SetVirtualAddressMap() on Altra machines

Ampere Altra machines are reported to misbehave when the SetTime() EFI
runtime service is called after ExitBootServices() but before calling
SetVirtualAddressMap(). Given that the latter is horrid, pointless and
explicitly documented as optional by the EFI spec, we no longer invoke
it at boot if the configured size of the VA space guarantees that the
EFI runtime memory regions can remain mapped 1:1 like they are at boot
time.

On Ampere Altra machines, this results in SetTime() calls issued by the
rtc-efi driver triggering synchronous exceptions during boot.  We can
now recover from those without bringing down the system entirely, due to
commit 23715a26c8d81291 ("arm64: efi: Recover from synchronous
exceptions occurring in firmware"). However, it would be better to avoid
the issue entirely, given that the firmware appears to remain in a funny
state after this.

So attempt to identify these machines based on the 'family' field in the
type #1 SMBIOS record, and call SetVirtualAddressMap() unconditionally
in that case.

Tested-by: Alexandru Elisei <alexandru.elisei@gmail.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
drivers/firmware/efi/libstub/Makefile
drivers/firmware/efi/libstub/arm64-stub.c
drivers/firmware/efi/libstub/efistub.h
drivers/firmware/efi/libstub/smbios.c [new file with mode: 0644]
include/linux/efi.h