Re: [RFC PATCH 4/4] sched/numa: Don't update mm->numa_next_scan from fault path

[Mel Gorman]

On Tue, Oct 05, 2021 at 02:40:15PM +0530, Bharata B Rao wrote:
> > 
> > Updating via cmpxchg would be ok to avoid potential collisions between
> > threads updating a shared mm.
> 
> Ok, may be I could just resend with changing the scan period update
> to use cmpxchg.
> 
> I also notice that in this case of scan period update, we just return
> without resetting the p->numa_faults_locality[]. Do you think if
> skipping the reset doesn't matter in this case?
> 

If there is no fault activity or migrations are failing, there is no
advantage to clearing numa_faults_locality[]. The information there
is still useful even if the scan period is updated.

-- 
Mel Gorman
SUSE Labs