Re: [Xen-devel] [PATCH 3/3] x86/EFI: don't accept 64-bit base relocations on page tables

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 31/08/16 13:28, Jan Beulich wrote:
>>>> On 30.08.16 at 18:35, <andrew.cooper3@xxxxxxxxxx> wrote:
>> On 19/08/16 13:57, Jan Beulich wrote:
>>>>>> On 19.08.16 at 14:39, <andrew.cooper3@xxxxxxxxxx> wrote:
>>>> On 19/08/16 08:52, Jan Beulich wrote:
>>>>> Page tables get pre-populated with physical addresses which, due to
>>>>> living inside the Xen image, will never exceed 32 bits in width. That
>>>>> in turn results in the tool generating the relocations to produce
>>>>> 32-bit relocations for them instead of the 64-bit ones needed for
>>>>> relocating virtual addresses. Hence instead of special casing page
>>>>> tables in the processing of 64-bit relocations, let's be more rigid
>>>>> and refuse them (as being indicative of something else having gone
>>>>> wrong in the build process).
>>>>>
>>>>> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
>>>> Is it an ABI requirement to use the minimal available relocation?  It is
>>>> certainly suboptimal to use a 64bit relocation when a 32bit one would
>>>> do, but I wouldn't bet that it is unconditional avoided by all toolchains.
>>> What ABI? The tool in question is one of our own making. And the
>>> way relocations get generated it's hard to tell those that have to
>>> be 32-bit (in early boot code and trampoline code) from those that
>>> may as well be 64-bit ones (in page tables).
>>>
>>>> It is currently the case that Xen needs to live below 4GB physical, so
>>>> from that point of view a 64bit relocation will not be required in the
>>>> pagetables.
>>> And even if Xen didn't itself have this requirement, the EFI loader
>>> would always put us below 4Gb.
>> Why is this necessarily true?
>>
>> xen.efi is built as a 64bit PE, not 32bit.
> The file format doesn't matter here. And we'd have other problems
> to solve if we got loaded above 4Gb.

Right, but that doesn't prohibit the generation of 64bit relocations in
the pagetables.

This patch still seems conceptually incorrect.

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel