Re: [PATCH 0/4] powerpc/topology: Introduce llc_override debugfs

[Shrikanth Hegde <sshegde@linux.ibm.com>]

On 10/14/26 9:26 AM, Srikar Dronamraju wrote:
> * Shrikanth Hegde <sshegde@linux.ibm.com> [2026-10-14 08:46:11]:
> 
>> On 10/14/26 2:14 AM, Tim Chen wrote:
>>> On Tue, 2026-10-13 at 15:28 +0530, Shrikanth Hegde wrote:
>>>>
>>>
>>> - SMT: patch 3 adds SD_SHARE_LLC only to MC/PKG, so for "smt" the
>>>     flags are as before. I think it is possible for P9 to have
>>>     shared_caches between two cores:
>>>
>>>     /*
>>>      * P9 has a slightly odd architecture where pairs of cores share an L2 cache.
>>>      * This topology makes it *much* cheaper to migrate tasks between adjacent cores
>>>      * since the migrated task remains cache hot. We want to take advantage of this
>>>      * at the scheduler level so an extra topology level is required.
>>>      */
>>>
>>>     the CACHE level also has SD_SHARE_LLC, and highest_flag_domain() picks it, so sd_llc spans
>>>     cpu_l2_cache_map. shared_caches is only set when that map is wider
>>>     than the core's threads, e.g. a P9 core pair sharing an L2:
>>>
>>>       CPU:               0 1 2 3 | 4 5 6 7
>>>       cpu_smt_mask():    core A  | core B
>>>       sd_llc (CACHE):    0-7
>>>
>>>     The ids follow cpu_smt_mask(), so 0-3 and 4-7 get different ids
>>>     while sd_llc and sd_llc_size cover all eight, and
>>>     cpus_share_cache(0, 4) is false, which is incorrect.
>>>
>>
>> Yes. I doubt anyone wants to try this experiment on Power9.
>> But yes, functionally updating the flags is needed.
>>
>> @srikar,
>> That is also the reason for having the disabled. By default on P9,
>> llc will be at CACHE domain. But, one may want to experiment with =smt
>> which i doubt anyone will ever do.
>>
> 
> I think P9 is a very valid point, but I don't think we should look at disable
> as an option just for this.
> 
> I would say, lets keep it simple now, just have an option rather a toggle to
> have LLC as MC/PKG for experimentation purpose. Switch the toggle and it
> does back to regular LLC mode. Also we don't know how the future Power
> Systems LLCs will look like. So keeping it a simple toggle, will reduce pain
> and increase our flexibility for the future.
> 
> Whether the toggle will switch to MC or PKG could be decided based on the
> experiment results.
> 

Ok. So kernel parameter "llc_override=mc/pkg" exclusively for ppc.

llc_override=mc means LLC is at MC.
llc_override=pkg means LLC is at PKG.
No kernel parameter means no change.

Hope that sounds good.

Also this new kernel parameter sounds a distant relative of split_llc.