Re: [PATCHv13 4/9] x86/boot/compressed: Handle unaccepted memory

[Ard Biesheuvel]

On Fri, 2 Jun 2023 at 18:09, Borislav Petkov <bp@xxxxxxxxx> wrote:
>
> On Fri, Jun 02, 2023 at 06:36:44PM +0300, Kirill A. Shutemov wrote:
..
> > Configuration table suppose to be present, even if unaccepted memory is
> > not supported. Something is very wrong if it is missing.
>
> I am not sure if it is the decompressor's job to do such validation
> - I guess this is something the EFI code should do.
>

'EFI code' is ambiguous here.

Most of the decompressor code is constructed in a way that permits
- booting 'native EFI' via the EFI stub
- booting 'pseudo-EFI' where GRUB or another Linux/x86 specific
bootloader populates boot_params with all the EFI specific information
(system table, memory map, etc)

This distinction has been abstracted away here, and so we might be
dealing with the second case, and booting from a GRUB that does not
understand accepted memory, but simply copied the EFI memory map
(including unaccepted regions) as it normally does. (Note that the
second case also covers kexec boot, so we do need to support it)