Re: [PATCH 0/5] gfx: improvements when using multiboot2 and EFI + misc

["Marek Marczykowski-Górecki" <marmarek@invisiblethingslab.com>]

[-- Attachment #1: Type: text/plain, Size: 2065 bytes --]

On Thu, Nov 24, 2022 at 10:56:33AM +0100, Roger Pau Monné wrote:
> On Thu, Nov 24, 2022 at 09:59:25AM +0100, Roger Pau Monné wrote:
> > On Thu, Nov 24, 2022 at 06:15:15AM +0100, Marek Marczykowski-Górecki wrote:
> > > But to get dom0 display image from BGRT, it seems something else is
> > > needed too. Linux complains "Incorrect checksum in table [BGRT]". The
> > > only relevant google result I get is this: https://support.citrix.com/article/CTX460227/citrix-hypervisor-acpi-warning-incorrect-checksum-in-table-bgrt
> > > It blames firmware. But then, it's suspicious that it's also about Xen.
> > > And also, native Linux on the same hw does not complain about the
> > > checksum. So, I think it's rather Xen to blame...
> > > The table lives in area marked as EfiACPIReclaimMemory in memory map, so
> > > I think it shouldn't be clobbered by Xen, at least in theory. I'll look
> > > into it later. It's getting off-topic for this thread anyway.
> > 
> > See commit 89238ef7797023f318f82f4f9dddef59c435b8bd.  I wonder whether
> > the BGRT image region is marked as EFI_MEMORY_RUNTIME, I will have to
> > check on my system.
> 
> Just checked on my system, and the BGRT image is placed in a
> EfiBootServicesData section with no EFI_MEMORY_RUNTIME attribute.

Right, while the BGRT table itself is in EfiACPIReclaimMemory, the image
it points to lives in EfiBootServicesData. And no EFI_MEMORY_RUNTIME
attribute there either.

> To fix this we would need to change efi_arch_process_memory_map() so
> it takes the BGRT image address into account and marks the region
> where it's placed as reserved.  I'm not aware of anyway to get such
> address from EFI data, so we would likely need to parse the BGRT in
> efi_arch_process_memory_map().

Since Xen has code to do that already, moving it earlier shouldn't be
too much issue. Can `acpi_boot_table_init()` be called that early? And
then, it sounds very similar to the issue we have with the ESRT table.

-- 
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]