On 2020/2/20 17:21, Ard Biesheuvel wrote:

On Thu, 20 Feb 2020 at 04:43, Jing Xiangfeng <jingxiangfeng@huawei.com> wrote:

This series enable finding mirrored memory ranges
functionality on arm64 platform. This feature has been
implemented on the x86 platform, so we move some
functions from x86.

Hello Jing Xiangfeng,

Could you explain your use case a bit better? Usually, the firmware is
a better place to make modifications to the EFI memory map.

The reason I am asking is that currently, on ARM and arm64, we never
make *any* changes to the firmware provided tables (EFI system table,
EFI memory map, DT/ACPI/SMBIOS tables etc), in order to ensure that
kexec is idempotent, i.e., it will always see the exact same state as
far as the firmware is concerned. This is a bit different from x86,
where the memory map is already modified for various other reasons, so
using it for fake memory regions is not such a big deal.

    I am not familiar with the firmware specifications. If we want to support
    this functionality now, does it mean that the UEFI spec needs to be modified?
    kexec related functions also need to be verified under synchronization, right?


Do you see a use case for this in production?

I'm debugging it by fake memory regions. Our server is expected to support it.

Thanks for your comments

Jing Xiangfeng (2):
  efi: allow EFI_FAKE_MEMMAP on arm64 platform
  arm64/efi: support to find mirrored memory ranges

 arch/x86/include/asm/efi.h      |  5 -----
 arch/x86/platform/efi/efi.c     | 39 ---------------------------------------
 drivers/firmware/efi/Kconfig    |  2 +-
 drivers/firmware/efi/arm-init.c |  2 ++
 drivers/firmware/efi/efi.c      | 23 +++++++++++++++++++++++
 drivers/firmware/efi/memmap.c   | 16 ++++++++++++++++
 include/linux/efi.h             |  5 +++++
 7 files changed, 47 insertions(+), 45 deletions(-)

--
1.8.3.1