Re: [PATCH 4/5] x86/xen: Avoid relocatable quantities in Xen ELF notes

From: Ard Biesheuvel
Date: Fri Sep 27 2024 - 01:49:45 EST


On Fri, 27 Sept 2024 at 03:47, Jason Andryuk <jason.andryuk@xxxxxxx> wrote:
>
> On 2024-09-26 06:41, Ard Biesheuvel wrote:
> > From: Ard Biesheuvel <ardb@xxxxxxxxxx>
> >
> > Xen puts virtual and physical addresses into ELF notes that are treated
> > by the linker as relocatable by default. Doing so is not only pointless,
> > given that the ELF notes are only intended for consumption by Xen before
> > the kernel boots. It is also a KASLR leak, given that the kernel's ELF
> > notes are exposed via the world readable /sys/kernel/notes.
> >
> > So emit these constants in a way that prevents the linker from marking
> > them as relocatable. This involves place-relative relocations (which
> > subtract their own virtual address from the symbol value) and linker
> > provided absolute symbols that add the address of the place to the
> > desired value.
> >
> > While at it, switch to a 32-bit field for XEN_ELFNOTE_PHYS32_ENTRY,
> > which better matches the intent as well as the Xen documentation and
> > source code.
>
> QEMU parses this according to the ELF bitness. It looks like this reads
> 8 bytes on 64bit, and 4 on 32. So I think this change would break its
> parsing.
>

Indeed, well spotted.

> (I don't use QEMU PVH and I'm not that familiar with its implementation.)
>

This is what I used for testing, and it worked fine.

But looking at the code, it does dereference a size_t*, which seems
bizarre but will clearly produce garbage in the upper bits if the note
is 32-bits only and followed by unrelated non-zero data.

I'll just back out this part of the change - it isn't really necessary anyway.