Re: [PATCH 03/19] mm/damon/paddr: set samples in apply_probes() if requested

[SJ Park <sj@kernel.org>]

On Fri, 10 Jul 2026 14:01:55 +0000 sashiko-bot@kernel.org wrote:

> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> 
> Pre-existing issues:
> - [High] Address advancement using `folio_size()` from an unaligned address within a large folio overshoots the next folio.
> --
> 
> --- Patch [3]: [PATCH 03/19] mm/damon/paddr: set samples in apply_probes() if requested ---
> commit 60a9acbe070a2be34c313d6be62fe7e4acb7ec54
> Author: SJ Park <sj@kernel.org>
> 
> mm/damon/paddr: set samples in apply_probes() if requested
> 
> This commit updates damon_pa_apply_probes() to respect the set_samples
> parameter. It adds logic to randomize the sampling address within the
> region boundaries using damon_rand() when requested.
> 
> > diff --git a/mm/damon/paddr.c b/mm/damon/paddr.c
> > index b13bf7c6eade3..ff4fcca949660 100644
> > --- a/mm/damon/paddr.c
> > +++ b/mm/damon/paddr.c
> > @@ -178,6 +178,10 @@ static void damon_pa_apply_probes(struct damon_ctx *ctx, bool set_samples)
> >  			phys_addr_t pa;
> >  			struct folio *folio;
> >  
> > +			if (set_samples)
> > +				r->sampling_addr = damon_rand(ctx, r->ar.start,
> > +						r->ar.end);
> > +
> >  			pa = damon_pa_phys_addr(r->sampling_addr,
> >  					ctx->addr_unit);
> >  			folio = damon_get_folio(PHYS_PFN(pa));
> 
> [Severity: High]
> This is a pre-existing issue in this file, but can the address advancement
> logic skip over pages during region iteration?
> 
> Looking at iteration functions like damon_pa_pageout(), 
> damon_pa_de_activate(), damon_pa_migrate(), and damon_pa_stat() in 
> mm/damon/paddr.c, they all use a pattern for advancing the address
> similar to this:
> 
> mm/damon/paddr.c:damon_pa_pageout() {
>     ...
>     folio = damon_get_folio(PHYS_PFN(addr));
>     if (!folio) {
>         addr += PAGE_SIZE;
>         continue;
>     }
>     ...
>     addr += folio_size(folio);
>     ...
> }
> 
> If the initial iteration address is not aligned to the folio boundary (for
> example, after a region split or hole), wouldn't adding the full folio size
> to this unaligned address cause it to overshoot the end of the current folio?
> 
> This seems like it would skip over the beginning of the next physical folio,
> causing those pages to incorrectly escape reclaim, migration, or statistics
> gathering.

Known issue.  I'm working on it.  No blocker of this patch.

> 
> -- 
> Sashiko AI review · https://sashiko.dev/#/patchset/20260710134651.18084-1-sj@kernel.org?part=3


Thanks,
SJ