[PATCH] Revert "efi_memory: do not add U-Boot memory to the memory map"

[Simon Glass]

Hi Tom,

On Sun, 16 Feb 2025 at 08:04, Tom Rini <trini at konsulko.com> wrote:
>
> On Sun, Feb 16, 2025 at 07:50:48AM -0700, Simon Glass wrote:
> > Hi Sughosh,
> >
> > On Wed, 27 Nov 2024 at 04:48, Sughosh Ganu <sughosh.ganu at linaro.org> wrote:
> > >
> > > On Tue, 26 Nov 2024 at 21:09, Simon Glass <sjg at chromium.org> wrote:
> > > >
> > > > Hi Sughosh,
> > > >
> > > > On Thu, 21 Nov 2024 at 00:53, Sughosh Ganu <sughosh.ganu at linaro.org> wrote:
> > > > >
> > > > > On Tue, 12 Nov 2024 at 18:48, Simon Glass <sjg at chromium.org> wrote:
> > > > > >
> > > > > > A bisect of Ubuntu 2022.04 boot-failure on qemu-x86_64 resulted in this
> > > > > > patch. I am not sure how to investigate it.
> > > > > >
> > > > > > The boot hangs at some point during booting of the install image, before
> > > > > > the Ubuntu logo appears.
> > > > > >
> > > > > > I will sent a series with a script showing how it is run.
> > > > > >
> > > > > > This reverts commit a68c9ac5d8afc51c619c5d3e865fcf147bea90cb.
> > > > > >
> > > > > > Signed-off-by: Simon Glass <sjg at chromium.org>
> > > > > > ---
> > > > >
> > > > > I did some digging into this issue, and it turns out that the revert
> > > > > is only masking the real issue. What this revert does is mark the
> > > > > memory region occupied by U-Boot as EFI_BOOT_SERVICES_CODE. The x86
> > > > > kernel seems to mark memory regions other than
> > > > > EFI_{BOOT,RUNTIME}_SERVICES_CODE as not executable -- and this is
> > > > > precisely why this crash shows up only on x86. The actual cause of
> > > > > this issue is in the efi runtime functionality of U-Boot. The kernel
> > > > > is calling the SetVirtualAddressMap() runtime service function, and
> > > > > that seems to be calling some of the boot service code of U-Boot,
> > > > > which it shouldn't. We do not get the crash after disabling the
> > > > > ConvertPointer() and SetVirtualAddressMap() runtime functions.
> > > >
> > > > OK, so what is the solution here? Did you send a patch?
> > >
> > > I am trying to check if the installation can be done by disabling the
> > > ConvertPointer() and SetVirtualAddressMap() runtime service calls. If
> > > so, then I will send a patch which disables them temporarily till the
> > > actual issue is triaged and fixed. If the installation does not work
> > > without these runtime calls, I would say putting this change in with
> > > an explanation of what is happening (similar to the above
> > > explanation). Please give me some more days, and I will let you know.
> > > Thanks.
> >
> > I forgot what happened with this. Could you update this thread, please?
>
> After much further investigation to understand why the revert worked,
> the revert was applied with a different commit message.

OK, thanks.

I have found another problem in that it breaks sandbox (bootflow_efi
test) when it calls exit-boot-services. But you have not applied that
series[1] so we can wait to see if you hit that problem in general
usage. Hopefully not!

Regards,
Simon

[1] https://patchwork.ozlabs.org/project/uboot/list/?series=439021