Re: [PATCH v6 00/42] x86/resctrl: Move the resctrl filesystem code to /fs/resctrl

[Reinette Chatre <reinette.chatre@intel.com>]

Hi James,

On 2/11/25 10:37 AM, James Morse wrote:
> Hi Reinette,
> 
> On 10/02/2025 17:24, Reinette Chatre wrote:
>> I'd like to check in on what you said in [1]. It sounded as though you were
>> planning to look at the assignable counter work from an Arm/MPAM
>> perspective but that work has since progressed (now at V11 [2]) without 
>> input from Arm/MPAM perspective. As I understand assignable counters may benefit
>> MPAM and looking close to settled but it is difficult to gain confidence
>> in an interface that may (may not?) be used for MPAM without any feedback
>> from Arm/MPAM. I am trying to prevent future issues when/if MPAM needs to use
>> this new interface and find it confusing that there does not seem to be
>> any input from MPAM side. What am I missing?
> 
> Shortly after that some 'new' Spectre issue turned up - unfortunately those rapidly
> consume all the time available, and predicting them is, er, the nature of the problem.

I understand how these things go. At the same time the new version you sent hinted to
me that you were able to come up for air to give some attention to resctrl upstream and
I grasped the opportunity to get your input on this item that moved along while you were
busy elsewhere.  

> 
> This is still on my todo list, I think Dave is planning to look through the ABMC series
> too. I had previously sent some comments, (words to the effect "works for me"), and shared
> a branch with the MPAM tree rebased on top.

To me the primary concern is the new resctrl files introduced by this work. To get an idea
of the work and the new files you need only consider Babu's detailed cover letter. These files
have changed since your previous comments so a fresh "works for me" will be appreciated.

> 
> I'm part way though rebasing the MPAM tree on top of Babu's latest version, and still hope
> to give some feedback based on testing it...

We tried to separate the arch and fs code logically but I am sure that it can be improved once
another architecture actually tries to use it. To me these internals seem a problem that
can be solved after merge of the assignable counter work, but addressing issues with the new
resctrl fs user interface after merge will be hard.

Reinette