Re: [PATCH] x86/Xen: swap NX determination and GDT setup on BSP

[Jan Beulich <jbeulich@xxxxxxxx>]

On 21.05.2021 15:12, Boris Ostrovsky wrote:
> 
> On 5/21/21 3:45 AM, Juergen Gross wrote:
>> On 21.05.21 09:26, Jan Beulich wrote:
>>> On 21.05.2021 09:18, Juergen Gross wrote:
>>>> On 20.05.21 14:08, Jan Beulich wrote:
>>>>> On 20.05.2021 13:57, Juergen Gross wrote:
>>>>>> On 20.05.21 13:42, Jan Beulich wrote:
>>>>>>> xen_setup_gdt(), via xen_load_gdt_boot(), wants to adjust page tables.
>>>>>>> For this to work when NX is not available, x86_configure_nx() needs 
>> to
>>>>>>> be called first.
>>>>>>>
>>>>>>> Reported-by: Olaf Hering <olaf@xxxxxxxxx>
>>>>>>> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
>>>>>>
>>>>>> Reviewed-by: Juergen Gross <jgross@xxxxxxxx>
>>>>>
>>>>> Thanks. I guess I forgot
>>>>>
>>>>> Cc: stable@xxxxxxxxxxxxxxx
>>>>>
>>>>> If you agree, can you please add this before pushing to Linus?
>>>>
>>>> Uh, just had a look why x86_configure_nx() was called after
>>>> xen_setup_gdt().
>>>>
>>>> Upstream your patch will be fine, but before kernel 5.9 it will
>>>> break running as 32-bit PV guest (see commit 36104cb9012a82e7).
>>>
>>> Oh, indeed. That commit then actually introduced the issue here,
>>> and hence a Fixes: tag may be warranted.
>>
>> Added it already. :-)
>>
>> And I've limited the backport to happen not for 5.8 and older, of
>> course.
> 
> 
> Did something changed recently that this became a problem? That commit has 
> been there for 3 years.

He happened to try on a system where NX was turned off in the BIOS. That's
not a setting you would usually find in use.


> Didn't Olaf report this to be a problem only on SLES kernels?

Which I assume have backports of the problematic change.

Jan