| 
    
 [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [PATCH] libxl/CPUID: drop two more feature flag table entries
 On 23.08.2023 15:45, Anthony PERARD wrote:
> On Wed, Aug 23, 2023 at 09:21:26AM +0200, Jan Beulich wrote:
>> Two entries were left in place by d638fe233cb3 ("libxl: use the cpuid
>> feature names from cpufeatureset.h"), despite matching the generated
>> names.
>>
>> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> 
> Acked-by: Anthony PERARD <anthony.perard@xxxxxxxxxx>
Thanks.
>> ---
>> Permitting "psn", "ia64, "cntxid", and "perfctr_*" when the hypervisor
>> doesn't even know of the bits (perhaps wrongly so) is kind of odd as
>> well. Permitting bits like "est", which the hypervisor knows of but
>> doesn't expose to guests, is also questionable.
> 
> I think those are just aliases, to a specific bit in a bitmap. Even if
> we remove those aliases, it is still possible to instruct libxl to do
> something with those bits. So there presence isn't permission, I don't
> think.
It's not permission, but recognizing the field name when its use then
(in the best case) doesn't do anything is at least misleading.
> Looks like "est" is just an alias for "eist", so it doesn't seems useful
> to remove it either.
Well, I'm effectively also questioning the exposure of "eist". Using
INIT_FEATURE_NAMES here was likely okay as a quick first approach, but
I think we want to limit what is recognized to what actually is useful
in at least some cases (yet better would of course be to also not
recognize e.g. HVM-only options when we're dealing with a PV guest).
Jan
 
  | 
  
![]()  | 
            
         Lists.xenproject.org is hosted with RackSpace, monitoring our  |