Re: [edk2-devel] [PATCH v3 03/16] ArmVirtPkg: make EFI_LOADER_DATA non-executable

["Gerd Hoffmann" <kraxel@redhat.com>]

  Hi,

> > That sounds neat, especially as we can print a big'n'fat warning in that
> > case, so the problem gets attention without actually breaking users.
> >
> 
> That, and a sleep(5)
> 
> > Looking at the efi calls it looks like edk2 doesn't track the owner of
> > an allocation (say by image handle), so I suspect it is not possible to
> > automatically figure who is to blame?
> >
> > > Does GRUB generally load/map executable modules at page granularity?
> >
> > I don't think so, at least the code handles modules not being page
> > aligned.  But I think it's not grub modules, that fix was actually
> > picked up meanwhile.  But there are downstream patches for image
> > loader code which look suspicious to me ...
> 
> OK, so the GRUB PE/COFF loader strikes again :-(

Yep.

> Could someone check/confirm whether shim builds need to be take into
> account here? Thanks.

Tried booting grub.efi directly and via shim.efi, on Fedora 37 GA.

In both cases I get a pagefault on linux kernel boot (before any other
message is printed), which I guess happens because the loader places the
linux kernel efi stub in EfiLoaderData memory.

I'd say that confirms grub.efi being buggy.

Not sure about shim.efi.  It managed to run grub.efi without hitting a
fault, which is good.  But it also installs efi protocols for the boot
loader to call, so it could be involved too.  But maybe that happens
only in case secure boot is active, which is not supported by
ArmVirtPkg.

take care,
  Gerd