DAMON resets the age of a region if its nr_accesses value has
significantly changed. Specifically, the threshold is calculated as 20%
of largest nr_accesses of the current snapshot.  This means that regions
changing the nr_accesses from zero to small non-zero value or from a
small non-zero value to zero will keep the age.  Since many users treat
zero nr_accesses regions special, this can be confusing.  Kernel code
including DAMOS' regions priority calculation and DAMON_STAT's idle time
calculation also treat zero nr_accesses regions special.  Make it
unconfusing by resetting the age when the nr_accesses changes between
zero and a non-zero value.

Signed-off-by: SeongJae Park <sj@kernel.org>
---
 mm/damon/core.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/mm/damon/core.c b/mm/damon/core.c
index be5942435d78..996647caca02 100644
--- a/mm/damon/core.c
+++ b/mm/damon/core.c
@@ -2261,6 +2261,9 @@ static void damon_merge_regions_of(struct damon_target *t, unsigned int thres,
 	damon_for_each_region_safe(r, next, t) {
 		if (abs(r->nr_accesses - r->last_nr_accesses) > thres)
 			r->age = 0;
+		else if ((!r->nr_accesses && r->last_nr_accesses) ||
+				(r->nr_accesses && !r->last_nr_accesses))
+			r->age = 0;
 		else
 			r->age++;
 
-- 
2.39.5

On Sun, 14 Sep 2025 18:58:02 -0700 SeongJae Park <sj@kernel.org> wrote:

> DAMON resets the age of a region if its nr_accesses value has
> significantly changed. Specifically, the threshold is calculated as 20%
> of largest nr_accesses of the current snapshot.  This means that regions
> changing the nr_accesses from zero to small non-zero value or from a
> small non-zero value to zero will keep the age.  Since many users treat
> zero nr_accesses regions special, this can be confusing.  Kernel code
> including DAMOS' regions priority calculation and DAMON_STAT's idle time
> calculation also treat zero nr_accesses regions special.  Make it
> unconfusing by resetting the age when the nr_accesses changes between
> zero and a non-zero value.

Hi SJ,

Thank you for the patch, I think the goal of the patch makes sesne to me.
I have a small nit / idea which I think makes the code a bit clearer, at least
for me. It seems that we basically want to XOR the two values's zero-ness, so
maybe something like 

(!!r->nr_accesses) ^ (!!r->last_nr_access) or
(r->nr_accesses == 0) ^ (r->last_nr_access == 0)

Can achieve the goal? I know bitwise operations are sometimes harder to
understand, so I am just throwing the idea out there : -) 


Anyways, the rest of it looks good to me, please feel free to add my review!

Reviewed-by: Joshua Hahn <joshua.hahnjy@gmail.com>

> Signed-off-by: SeongJae Park <sj@kernel.org>
> ---
>  mm/damon/core.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/mm/damon/core.c b/mm/damon/core.c
> index be5942435d78..996647caca02 100644
> --- a/mm/damon/core.c
> +++ b/mm/damon/core.c
> @@ -2261,6 +2261,9 @@ static void damon_merge_regions_of(struct damon_target *t, unsigned int thres,
>  	damon_for_each_region_safe(r, next, t) {
>  		if (abs(r->nr_accesses - r->last_nr_accesses) > thres)
>  			r->age = 0;
> +		else if ((!r->nr_accesses && r->last_nr_accesses) ||
> +				(r->nr_accesses && !r->last_nr_accesses))
> +			r->age = 0;
>  		else
>  			r->age++;
>  
> -- 
> 2.39.5

Sent using hkml (https://github.com/sjp38/hackermail)

On Mon, 15 Sep 2025 07:51:57 -0700 Joshua Hahn <joshua.hahnjy@gmail.com> wrote:

> On Sun, 14 Sep 2025 18:58:02 -0700 SeongJae Park <sj@kernel.org> wrote:
> 
> > DAMON resets the age of a region if its nr_accesses value has
> > significantly changed. Specifically, the threshold is calculated as 20%
> > of largest nr_accesses of the current snapshot.  This means that regions
> > changing the nr_accesses from zero to small non-zero value or from a
> > small non-zero value to zero will keep the age.  Since many users treat
> > zero nr_accesses regions special, this can be confusing.  Kernel code
> > including DAMOS' regions priority calculation and DAMON_STAT's idle time
> > calculation also treat zero nr_accesses regions special.  Make it
> > unconfusing by resetting the age when the nr_accesses changes between
> > zero and a non-zero value.
> 
> Hi SJ,
> 
> Thank you for the patch, I think the goal of the patch makes sesne to me.
> I have a small nit / idea which I think makes the code a bit clearer, at least
> for me. It seems that we basically want to XOR the two values's zero-ness, so
> maybe something like 
> 
> (!!r->nr_accesses) ^ (!!r->last_nr_access) or
> (r->nr_accesses == 0) ^ (r->last_nr_access == 0)
> 
> Can achieve the goal?

Thank you for the idea, this makes sense!

> I know bitwise operations are sometimes harder to
> understand, so I am just throwing the idea out there : -) 

To be honest I'm one of people who are not familiar with XOR.  I had to spend a
minute to understand the above.  Maybe we can replace '^' with '!=', and it is
easier to read for me.  If you don't mind, I will use below in the next
version:

   else if ((r->nr_accesses == 0) != (r->last_nr_accesses == 0))

Please let me know if I'm missing something or you have other opinions.

> 
> 
> Anyways, the rest of it looks good to me, please feel free to add my review!
> 
> Reviewed-by: Joshua Hahn <joshua.hahnjy@gmail.com>

Thank you!

> 
> > Signed-off-by: SeongJae Park <sj@kernel.org>
> > ---
> >  mm/damon/core.c | 3 +++
> >  1 file changed, 3 insertions(+)
> > 
> > diff --git a/mm/damon/core.c b/mm/damon/core.c
> > index be5942435d78..996647caca02 100644
> > --- a/mm/damon/core.c
> > +++ b/mm/damon/core.c
> > @@ -2261,6 +2261,9 @@ static void damon_merge_regions_of(struct damon_target *t, unsigned int thres,
> >  	damon_for_each_region_safe(r, next, t) {
> >  		if (abs(r->nr_accesses - r->last_nr_accesses) > thres)
> >  			r->age = 0;
> > +		else if ((!r->nr_accesses && r->last_nr_accesses) ||
> > +				(r->nr_accesses && !r->last_nr_accesses))
> > +			r->age = 0;
> >  		else
> >  			r->age++;
> >  
> > -- 
> > 2.39.5
> 
> Sent using hkml (https://github.com/sjp38/hackermail)


Thanks,
SJ

[...]

On Mon, 15 Sep 2025 11:26:51 -0700 SeongJae Park <sj@kernel.org> wrote:

> > Thank you for the patch, I think the goal of the patch makes sesne to me.
> > I have a small nit / idea which I think makes the code a bit clearer, at least
> > for me. It seems that we basically want to XOR the two values's zero-ness, so
> > maybe something like 
> > 
> > (!!r->nr_accesses) ^ (!!r->last_nr_access) or
> > (r->nr_accesses == 0) ^ (r->last_nr_access == 0)
> > 
> > Can achieve the goal?
> 
> Thank you for the idea, this makes sense!
> 
> > I know bitwise operations are sometimes harder to
> > understand, so I am just throwing the idea out there : -) 
> 
> To be honest I'm one of people who are not familiar with XOR.  I had to spend a
> minute to understand the above.  Maybe we can replace '^' with '!=', and it is
> easier to read for me.  If you don't mind, I will use below in the next
> version:
> 
>    else if ((r->nr_accesses == 0) != (r->last_nr_accesses == 0))
> 
> Please let me know if I'm missing something or you have other opinions.

I have to say, using xor as a shorthand for what-was-really-intended
always bursts my brain.  I have to stop, think about it and mentally
turn the implementation back into what-was-really-intended.  Maybe
that's just me.

Ditto less-than-utterly-trivial :?  expressions.  Perhaps it's because
if-then-else best suits how our minds work.