Re: [RFC PATCH 2/4] mm/damon/sysfs-schemes: support DAMOS_FILTER_TYPE_AGE_MS

[SJ Park <sj@kernel.org>]

On Wed, 04 Nov 2026 15:50:23 +0000 sashiko-bot@kernel.org wrote:

> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [Medium] mm/damon: 32-bit overflow calculating age_ms in damos_filter_match()
> --
> 
> commit 9c78f1b31239922167bf213c5406c37c05fd310f
> Author: SJ Park <sj@kernel.org>
> 
> mm/damon/sysfs-schemes: support DAMOS_FILTER_TYPE_AGE_MS
> 
> This commit exposes DAMOS_FILTER_TYPE_AGE_MS to the DAMON sysfs interface
> using the "age_ms" filter type keyword and the existing min and max sysfs
> range files.
> 
> > diff --git a/mm/damon/sysfs-schemes.c b/mm/damon/sysfs-schemes.c
> > index 637f1bd0c950b..230987efac3ac 100644
> > --- a/mm/damon/sysfs-schemes.c
> > +++ b/mm/damon/sysfs-schemes.c
> 
> [ ... ]
> 
> > @@ -2875,6 +2879,7 @@ static int damon_sysfs_add_scheme_filters(struct damos *scheme,
> >  			filter->sz_range.max = sysfs_filter->range_max;
> >  			break;
> >  		case DAMOS_FILTER_TYPE_PROBE_HITS_WSUM:
> > +		case DAMOS_FILTER_TYPE_AGE_MS:
> >  			filter->range_min = sysfs_filter->range_min;
> >  			filter->range_max = sysfs_filter->range_max;
> >  			break;
> 
> [Severity: Medium]
> When userspace configures an age_ms filter with range_min and range_max via
> damon_sysfs_add_scheme_filters() on a 32-bit system, can age_ms overflow in
> damos_filter_match()?
> 
> When kdamond evaluates the filter via
> kdamond_apply_schemes()->damon_do_apply_schemes()->damos_apply_scheme()->damos_core_filter_out()->damos_filter_match():
> 
> mm/damon/core.c:damos_filter_match() {
> 	...
> 	case DAMOS_FILTER_TYPE_AGE_MS:
> 		age_ms = r->age * ctx->attrs.aggr_interval / USEC_PER_MSEC;
> 		matched = filter->range_min <= age_ms &&
> 			age_ms <= filter->range_max;
> 		break;
> 	...
> }
> 
> Here r->age is unsigned int and ctx->attrs.aggr_interval is unsigned long
> (in microseconds). On 32-bit architectures, multiplying r->age by
> ctx->attrs.aggr_interval is performed in 32-bit unsigned arithmetic before
> dividing by USEC_PER_MSEC, which wraps around once a region's age exceeds
> approximately 4,295 seconds (~71.6 minutes) even though the resulting
> millisecond value fits in a 32-bit unsigned long.
> 
> A similar 32-bit overflow on r->age * c->attrs.aggr_interval in
> damon_hot_score() in mm/damon/ops-common.c was fixed in commit 0284dff76e87
> ("mm/damon/ops-common: fix age_in_sec overflow on 32-bit"). Should
> damos_filter_match() use div_u64((u64)r->age * ctx->attrs.aggr_interval,
> USEC_PER_MSEC) as well?

I will fix [1] in the next version.

[1] https://lore.kernel.org/20261104154746.83311-1-sj@kernel.org

> 
> -- 
> Sashiko AI review · https://sashiko.dev/#/patchset/20261104153645.82945-1-sj@kernel.org?part=2
> 


Thanks,
SJ