stand/efi: Limit where we load the arm64 kernel

[andrew]

Summary

When FEAT_LPA2 is implemented the hardware supports increasing the
physical address space from 48-bit to 52-bit.

As older kernels only support a 48-bit physical address space limit
where the kernel is loaded to this.

Sponsored by: Arm Ltd

[markj]

stand/efi/loader/copy.c

311

The - 1 seems inconsistent with the amd64 case above. Which one is right?

[andrew]

stand/efi/loader/copy.c

311

BS->AllocatePages with AllocateMaxAddress can allocate pages where the uppermost address is less than or equal to this value.

It probably doesn't matter to not have the - 1 as no pages above the 1G or 4G mark should be allocated in the amd64 case as they would have to be zero length.

[markj]

stand/efi/loader/copy.c

311

Then perhaps let's just use (1ul << 48) for consistency's sake?

[kib]

stand/efi/loader/copy.c

330

Does this work? I thought that cpp #if only takes arithmetic expressions + defined().

[andrew]

Adjust the AllocateMaxAddress check

[imp]

Is there any way that we can know if the kernel supports higher loading? Otherwise we'll have this weird quirk forever...

[imp]

We need to start tagging arm64 kernels that can run about 48 bits so we don't have this workaround around forever

stand/efi/loader/copy.c

440

I'd suggest we do this for all architectures. There's no reason to ifdef this. Otherwise new_base is uninitialized.

[andrew]

Not currently as no kernels support it. We could add an elf note when we support LPA2 in locore.S to tell loader the kernel supports executing from a higher address.

[andrew]

stand/efi/loader/copy.c

440

For an allocation type of AllocateAnyPages this value is ignored in AllocatePages

[imp]

stand/efi/loader/copy.c

440

The file has way too many ifdefs. This is an easy way to limit that. The arch code will do the right thing..

[imp]

In D46625#1094783, @andrew wrote:

Not currently as no kernels support it. We could add an elf note when we support LPA2 in locore.S to tell loader the kernel supports executing from a higher address.

That would be great.

[kib]

In D46625#1094801, @imp wrote:

In D46625#1094783, @andrew wrote:

Not currently as no kernels support it. We could add an elf note when we support LPA2 in locore.S to tell loader the kernel supports executing from a higher address.

That would be great.

For amd64, when I needed to mark kernel as supporting load at arbitrary address below 4G (as opposed of 2M), I check for presence of the magic symbol 'kernphys'.