Re: [PATCH v5 00/11] x86,fs/resctrl: Fix long-standing issues

[Reinette Chatre]

Hi Boris,

On 7/1/26 1:37 PM, Borislav Petkov wrote:
> On Wed, Jul 01, 2026 at 12:40:27PM -0700, Borislav Petkov wrote:
>> On Wed, Jul 01, 2026 at 10:55:38AM -0700, Reinette Chatre wrote:
>>> I agree there are serious issues here. What prevents me from advocating for them
>>> all to go to stable is that the fixes to the intricate races required changes to
>>> resctrl's core locking flows. I absolutely tested and scrutinized these changes but
>>> having them go straight to all the stable kernels feels like a big jump to me.
>>
>> Makes sense to me.
>>
>>> The fixes that do not involve locking changes are below. They could be considered
>>> for stable.
>>> 1/11: x86,fs/resctrl: Prevent out-of-bounds access while offlining CPU when SNC enabled
>>> 4/11: fs/resctrl: Free mon_data structures on rdt_get_tree() failure
>>> 5/11: fs/resctrl: Fix use-after-free during unmount
>>> 8/11: fs/resctrl: Fix double-add of pseudo-locked region's RMID to free list
>>
>> Ok, lemme hoist those up into urgent first.
> 
> Yeah, not easy, see below.
> 
> This is because 3/11 which is moving the functions, is missing now.
> 
> We can do a forward declaration but I'd let you reshuffle things the way you
> want them and send me the urgent fixes by placing them at the beginning of the
> set so that I can stick them into urgent.

Will do. Since these patches are now targeting urgent and stable, a forward declaration
would make the backporting easier. I do prefer to keep 3/11 on top of that though. I will
prepare the next version of this series to do this. 

> 
> I'll queue 1/11 only now.

Thank you very much.

Reinette