[PATCH 0/4] mm/damon: add lru-active probe filter type

[PATCH 0/4] mm/damon: add lru-active probe filter type

[SJ Park]

The LRU-activeness of folios is useful information for understanding and
optimizing the system.  DAMOS ops filter therefore supports filtering
memory based on the information.  Using it, the information can be
extracted in the folio granularity.  It is precise but expensive.  DAMON
probe filter is the complementary way for getting the information in
region granularity with much cheaper cost.  DAMON filter does not
support LRU-activeness based filtering, though.  Add the support.

Define new DAMON filter type, DAMON_FILTER_TYPE_ACTIVE, on core API
(patch 1).  Add support for it on both vaddr and paddr operation set
(patch 2).  Extend DAMON filter sysfs ABI to let users configure DAMON
filters of the type using a new keyword, "active" (patch 3).  Finally,
update the DAMON design document to explain the new type (patch 4).

Test
====

Confirmed the functionality like below.

First, start masim [1] to generate a ~476 MiB hot memory region.

    $ ./masim.py run --config_file ./configs/stairs-500mb.cfg --repeat 10

Second, start DAMON to monitor data access and active memory ratio of
each region using DAMON user-space tool (damo) [2].

    $ sudo ./damo start --target_pid $(pidof masim) \
            --probe --probe_prep set_pgidle \
                --probe_filter allow pgidle_unset \
            --probe --probe_filter allow active \
            --probe_weight 1 0

Show the monitoring results.

    $ sudo ./damo report attrs
    heatmap: 00008[...]82222222222222222222222222222222222222222222222257777789999998888888874444445[...]8
    # min/max temperatures: -1,580,000,000, 100,002,000, column size: 62.841 MiB
    probes
    - prep: set_pgidle, filter: allow pgidle_unset (weight: 1)
    - allow active (weight: 0)
    intervals: sample 5 ms aggr 100 ms (max probe hits 20)
    #   size         address       age           probe_hits
    0   253.047 MiB  85.984 TiB    15.800 s      0 0
    1   4.000 KiB    127.984 TiB   15.400 s      0 20
    2   120.000 KiB  127.984 TiB   15.300 s      0 0
    3   1.450 GiB    127.407 TiB   11.300 s      0 0
    4   498.738 MiB  127.409 TiB   11.200 s      0 0
    5   500.273 MiB  127.409 TiB   11.100 s      0 0
    6   502.055 MiB  127.409 TiB   10.900 s      0 0
    7   461.926 MiB  127.411 TiB   6.500 s       0 0
    8   351.754 MiB  127.410 TiB   1.600 s       0 0
    9   497.930 MiB  127.411 TiB   0 ns          1 0
    10  8.000 KiB    127.984 TiB   0 ns          6 0
    11  476.340 MiB  127.410 TiB   1 s           20 0
    memory bw estimate: 0 B per second
    total size: 4.909 GiB
    record DAMON intervals: sample 5 ms, aggr 100 ms

The second (idx 1) 4 KiB memory region is not accessed for 15 seconds
and the page for the region is on an active LRU list.  The last 476.340
MiB memory region is accessed for each sample for the last one second as
expected.  And nearly all pages of the region are not on an active LRU
list.

[1] https://github.com/sjp38/masim
[2] https://github.com/damonitor/damo

Changes from RFC
- RFC: https://lore.kernel.org/20261108191615.113362-1-sj@kernel.org
- Add test results on the cover letter.
- Rebase to the latest mm-new.
- Drop RFC tag.

The test section and the changelog can be removed from the final commit
message.

SJ Park (4):
  mm/damon/api: define DAMON_FILTER_TYPE_ACTIVE
  mm/damon/ops-common: implement DAMON_FILTER_TYPE_ACTIVE operation
  mm/damon/sysfs: support LRU-activeness based DAMON filter setup
  Docs/mm/damon/design: update for DAMON_FILTER_TYPE_ACTIVE

 Documentation/mm/damon/design.rst | 1 +
 include/linux/damon.h             | 2 ++
 mm/damon/ops-common.c             | 7 +++++++
 mm/damon/sysfs.c                  | 4 ++++
 4 files changed, 14 insertions(+)


base-commit: e055da336413b651d1e23d438bb5430f9350d3f8
-- 
2.47.3

[PATCH 1/4] mm/damon/api: define DAMON_FILTER_TYPE_ACTIVE

[SJ Park]

Knowing which memory is on Active LRU lists is useful for understanding
and optimally operating systems.  Extend the DAMON core layer to be
ready for support of that.  For that, define a new damon_filter_type,
DAMON_FILTER_TYPE_ACTIVE.  It has no additional arguments and therefore
no additional change in the core layer is needed.  Testing if a given
memory is on an LRU list for the type of filter should be implemented in
the operation set layer.

Signed-off-by: SJ Park <sj@kernel.org>
---
 include/linux/damon.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/include/linux/damon.h b/include/linux/damon.h
index c32ef79c3acea..06bd92981af21 100644
--- a/include/linux/damon.h
+++ b/include/linux/damon.h
@@ -780,6 +780,7 @@ struct damon_prep {
  * @DAMON_FILTER_TYPE_PGIDLE_UNSET:	Pgidle is unset.
  * @DAMON_FILTER_TYPE_PGIDLE_SET:	Pgidle is set.
  * @DAMON_FILTER_TYPE_HUGEPAGE_SIZE:	Page is part of a hugepage.
+ * @DAMON_FILTER_TYPE_ACTIVE:		LRU-active pages.
  */
 enum damon_filter_type {
 	DAMON_FILTER_TYPE_ANON,
@@ -787,6 +788,7 @@ enum damon_filter_type {
 	DAMON_FILTER_TYPE_PGIDLE_UNSET,
 	DAMON_FILTER_TYPE_PGIDLE_SET,
 	DAMON_FILTER_TYPE_HUGEPAGE_SIZE,
+	DAMON_FILTER_TYPE_ACTIVE,
 };
 
 /**
-- 
2.47.3

[PATCH 2/4] mm/damon/ops-common: implement DAMON_FILTER_TYPE_ACTIVE operation

[SJ Park]

Implement DAMON_FILTER_TYPE_ACTIVE operation.  Use folio_test_active()
to show if the folio of the given memory is on an active LRU list.  The
implementation is made on ops-common, so both vaddr and paddr gain the
support of the filter type.

Signed-off-by: SJ Park <sj@kernel.org>
---
 mm/damon/ops-common.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/mm/damon/ops-common.c b/mm/damon/ops-common.c
index b9571f53e538c..2216a0023154f 100644
--- a/mm/damon/ops-common.c
+++ b/mm/damon/ops-common.c
@@ -574,6 +574,13 @@ bool damon_ops_filter_match(struct damon_filter *filter, struct folio *folio)
 		matched = filter->range_min <= folio_sz &&
 			folio_sz <= filter->range_max;
 		break;
+	case DAMON_FILTER_TYPE_ACTIVE:
+		if (!folio) {
+			matched = false;
+			break;
+		}
+		matched = folio_test_active(folio);
+		break;
 	default:
 		break;
 	}
-- 
2.47.3

[PATCH 3/4] mm/damon/sysfs: support LRU-activeness based DAMON filter setup

[SJ Park]

Support DAMON sysfs interface based setup of DAMON_FILTER_TYPE_ACTIVE
type DAMON filters.  For that, wire the type with a new user keyword,
"active".  Users can set up the new type of filter using the existing
DAMON filter setup interface with the new keyword.

Signed-off-by: SJ Park <sj@kernel.org>
---
 mm/damon/sysfs.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/mm/damon/sysfs.c b/mm/damon/sysfs.c
index 70a4b64fb8ea9..779c98205a94b 100644
--- a/mm/damon/sysfs.c
+++ b/mm/damon/sysfs.c
@@ -1011,6 +1011,10 @@ damon_sysfs_filter_type_names[] = {
 		.type = DAMON_FILTER_TYPE_HUGEPAGE_SIZE,
 		.name = "hugepage_size",
 	},
+	{
+		.type = DAMON_FILTER_TYPE_ACTIVE,
+		.name = "active",
+	},
 };
 
 static ssize_t type_show(struct kobject *kobj,
-- 
2.47.3

[PATCH 4/4] Docs/mm/damon/design: update for DAMON_FILTER_TYPE_ACTIVE

[SJ Park]

Update DAMON design document for the newly added LRU-activeness DAMON
filter type.

Signed-off-by: SJ Park <sj@kernel.org>
---
 Documentation/mm/damon/design.rst | 1 +
 1 file changed, 1 insertion(+)

diff --git a/Documentation/mm/damon/design.rst b/Documentation/mm/damon/design.rst
index 8440c4954090e..b676be6971589 100644
--- a/Documentation/mm/damon/design.rst
+++ b/Documentation/mm/damon/design.rst
@@ -302,6 +302,7 @@ filter types.  Currently below filter types are supported.
 - ``pgidle_set``: Matches if the page for the memory is marked as access-idle.
 - ``hugepage_size``: Matches if the page for the memory is a part of a hugepage
   of a given size range.
+- ``active``: Same to that for DAMOS filters.
 
 If such probes are registered, DAMON executes the probes for each region's
 sampling memory when it does the access :ref:`sampling
-- 
2.47.3

Re: [PATCH 4/4] Docs/mm/damon/design: update for DAMON_FILTER_TYPE_ACTIVE

[sashiko-bot]

> Update DAMON design document for the newly added LRU-activeness DAMON
> filter type.
> 
> Signed-off-by: SJ Park <sj@kernel.org>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261113051711.89274-1-sj@kernel.org?part=4

Re: [PATCH 1/4] mm/damon/api: define DAMON_FILTER_TYPE_ACTIVE

[sashiko-bot]

> Knowing which memory is on Active LRU lists is useful for understanding
> and optimally operating systems.  Extend the DAMON core layer to be
> ready for support of that.  For that, define a new damon_filter_type,
> DAMON_FILTER_TYPE_ACTIVE.  It has no additional arguments and therefore
> no additional change in the core layer is needed.  Testing if a given
> memory is on an LRU list for the type of filter should be implemented in
> the operation set layer.
> 
> Signed-off-by: SJ Park <sj@kernel.org>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261113051711.89274-1-sj@kernel.org?part=1

Re: [PATCH 3/4] mm/damon/sysfs: support LRU-activeness based DAMON filter setup

[sashiko-bot]

> Support DAMON sysfs interface based setup of DAMON_FILTER_TYPE_ACTIVE
> type DAMON filters.  For that, wire the type with a new user keyword,
> "active".  Users can set up the new type of filter using the existing
> DAMON filter setup interface with the new keyword.
> 
> Signed-off-by: SJ Park <sj@kernel.org>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261113051711.89274-1-sj@kernel.org?part=3

Re: [PATCH 2/4] mm/damon/ops-common: implement DAMON_FILTER_TYPE_ACTIVE operation

[sashiko-bot]

> Implement DAMON_FILTER_TYPE_ACTIVE operation.  Use folio_test_active()
> to show if the folio of the given memory is on an active LRU list.  The
> implementation is made on ops-common, so both vaddr and paddr gain the
> support of the filter type.
> 
> Signed-off-by: SJ Park <sj@kernel.org>

Sashiko has reviewed this patch and found no issues.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261113051711.89274-1-sj@kernel.org?part=2

Re: [PATCH 4/4] Docs/mm/damon/design: update for DAMON_FILTER_TYPE_ACTIVE

[Enze Li]

Hi SJ,

On 11/13/26 1:17 PM, SJ Park wrote:
> Update DAMON design document for the newly added LRU-activeness DAMON
> filter type.
> 
> Signed-off-by: SJ Park <sj@kernel.org>
> ---
>  Documentation/mm/damon/design.rst | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/Documentation/mm/damon/design.rst b/Documentation/mm/damon/design.rst
> index 8440c4954090e..b676be6971589 100644
> --- a/Documentation/mm/damon/design.rst
> +++ b/Documentation/mm/damon/design.rst
> @@ -302,6 +302,7 @@ filter types.  Currently below filter types are supported.
>  - ``pgidle_set``: Matches if the page for the memory is marked as access-idle.
>  - ``hugepage_size``: Matches if the page for the memory is a part of a hugepage
>    of a given size range.
> +- ``active``: Same to that for DAMOS filters.

This reads a bit awkward to me.  How about:

  "Same as that for DAMOS filters."

The ``anon`` and ``memcg`` entries above use the same wording, so
it would be nice to fix those at the same time.

WDYT?

Thanks,
Enze

>  
>  If such probes are registered, DAMON executes the probes for each region's
>  sampling memory when it does the access :ref:`sampling

Re: [PATCH 4/4] Docs/mm/damon/design: update for DAMON_FILTER_TYPE_ACTIVE

[SJ Park]

On Fri, 13 Nov 2026 17:54:37 +0800 Enze Li <lienze@kylinos.cn> wrote:

> Hi SJ,
> 
> On 11/13/26 1:17 PM, SJ Park wrote:
> > Update DAMON design document for the newly added LRU-activeness DAMON
> > filter type.
> > 
> > Signed-off-by: SJ Park <sj@kernel.org>
> > ---
> >  Documentation/mm/damon/design.rst | 1 +
> >  1 file changed, 1 insertion(+)
> > 
> > diff --git a/Documentation/mm/damon/design.rst b/Documentation/mm/damon/design.rst
> > index 8440c4954090e..b676be6971589 100644
> > --- a/Documentation/mm/damon/design.rst
> > +++ b/Documentation/mm/damon/design.rst
> > @@ -302,6 +302,7 @@ filter types.  Currently below filter types are supported.
> >  - ``pgidle_set``: Matches if the page for the memory is marked as access-idle.
> >  - ``hugepage_size``: Matches if the page for the memory is a part of a hugepage
> >    of a given size range.
> > +- ``active``: Same to that for DAMOS filters.
> 
> This reads a bit awkward to me.  How about:
> 
>   "Same as that for DAMOS filters."
> 
> The ``anon`` and ``memcg`` entries above use the same wording, so
> it would be nice to fix those at the same time.
> 
> WDYT?

Makes sense to me, thank you for sharing the finding, Enze!  I also find more
"same to" usages in the doc.  Would you like to send a patch for fixing those
at once?


Thanks,
SJ

[...]