Skip to contentSkip to conversation

mm/damon: introduce probe_hits_wsum DAMOS core filter

RFC PATCH Superseded SJ Park 1 participant 9 messages Sep 7

Subsystems (4)
lore
Share & tools
lore

SJ Park

author maintainer
#0 lore raw

v2 cover mm/damon: introduce probe_hits_wsum DAMOS core filter

DAMON can do flexible data attributes monitoring. The classical data access monitoring can also be done using the attributes monitoring. The access event is just one of the data attributes that DAMON supports. Users do monitoring to make some actions based on it. DAMOS is a feature for automating that. However, DAMOS cannot utilize the data attributes monitoring results. It is still Data "Access" Monitoring-based Operation Schemes. It requires users to set the target "access" pattern.

DAMOS core filter is effectively the same as the target access pattern. It is just a more generalized and flexible way of describing the operation action target region. Introduce a new DAMOS core filter type, probe_hits_wsum. It specifies the filter target based on a range of the probe hits weighted sum. Using this, users can apply DAMOS actions to regions of specific data attributes pattern.

Note that the classic target access pattern still works. Hence the target nr_accesses range should still be properly configured. The new filter would be used in only data attributes-only mode. In the mode, classic access monitoring is just turned off, and therefore nr_accesses of regions are always zero. Users could simply set the target nr_accesses range to include the zero nr_Accesses regions.

Patches Sequence
================

Patch 1 updates the DAMON kernel API for the new filter type. Patch 2 extends damon_probe_hits_wsum() to do the calculation based on moving sum. Patch 3 implements the filter type in the core layer. Patch 4 refactors DAMON sysfs interface internal data structure for efficient reuse of data structure for the probe hits weighted sum range user inputs. Patch 5 updates DAMON sysfs interface to support the new filter type. Patches 6 and 7 update design and usage documents for the new filter type, respectively.

Changelog
=========

Changes from RFC
- RFC: https://lore.kernel.org/[email protected]
- Fix kernel-doc typo.
- Use moving sum based weighted sum.
- Fix wrong range saving field names in sysfs.

SJ Park (7):
mm/damon/api: introduce DAMOS_FILTER_TYPE_PROBE_HITS_WSUM mm/damon/core: extend probe_hits_wsum() for moving sum based
calculation
mm/damon/core: support probe_hits_wsum damos core filter mm/damon/sysfs-schemes: rename sysfs_filter->sz_range to

    range_{min,max}

mm/damon/sysfs-schemes: support probe_hits_wsum damos core filter Docs/mm/damon/design: update for probe_hits_wsum DAMOS core filter Docs/admin-guide/mm/damon/usage: update for probe_hits_wsum DAMOS
filter

 Documentation/admin-guide/mm/damon/usage.rst |  4 +++
 Documentation/mm/damon/design.rst            |  3 ++
 include/linux/damon.h                        | 10 ++++++-
 mm/damon/core.c                              | 20 +++++++++++--
 mm/damon/paddr.c                             |  2 +-
 mm/damon/sysfs-schemes.c                     | 30 ++++++++++++++------
 mm/damon/vaddr.c                             |  2 +-
 7 files changed, 57 insertions(+), 14 deletions(-)

base-commit: 48856ec5bf2f614ff3ef647bbf7cc605faf8c977

Signature · 1 line
2.47.3
author maintainer
#1 lore raw

RFC PATCH v2 1/7 mm/damon/api: introduce DAMOS_FILTER_TYPE_PROBE_HITS_WSUM

Update DAMON kernel API to introduce new DAMOS core filter type, PROBE_HITS_WSUM. It will allow API callers to describe the DAMOS action target regions based on their probe_hits weighted sum. For describing the filtering target weighted sum range, add two type-dependent union fields to damos_filter.

---
 include/linux/damon.h | 8 ++++++++
 1 file changed, 8 insertions(+)
include/linux/damon.h +8 -0
index 871d26adf6ae5..1deda16bb70b5 100644
@@ -399,6 +399,7 @@ struct damos_stat {
  * @DAMOS_FILTER_TYPE_UNMAPPED:	Unmapped pages.
  * @DAMOS_FILTER_TYPE_ADDR:	Address range.
  * @DAMOS_FILTER_TYPE_TARGET:	Data Access Monitoring target.
+ * @DAMOS_FILTER_TYPE_PROBE_HITS_WSUM:	probe_hits weighted sum range.
  * @NR_DAMOS_FILTER_TYPES:	Number of filter types.
  *
  * All types except &DAMOS_FILTER_TYPE_ADDR and &DAMOS_FILTER_TYPE_TARGET
@@ -420,6 +421,7 @@ enum damos_filter_type {
 	DAMOS_FILTER_TYPE_UNMAPPED,
 	DAMOS_FILTER_TYPE_ADDR,
 	DAMOS_FILTER_TYPE_TARGET,
+	DAMOS_FILTER_TYPE_PROBE_HITS_WSUM,
 	NR_DAMOS_FILTER_TYPES,
 };
 
@@ -434,6 +436,8 @@ enum damos_filter_type {
  *		&damon_ctx->adaptive_targets if @type is
  *		DAMOS_FILTER_TYPE_TARGET.
  * @sz_range:	Size range if @type is DAMOS_FILTER_TYPE_HUGEPAGE_SIZE.
+ * @range_min:	Minimum value of range arguments.
+ * @range_max:	Maximum value of range arguments.
  *
  * Before applying the &damos->action to a memory region, DAMOS checks if each
  * byte of the region matches to this given condition and avoid applying the
@@ -450,6 +454,10 @@ struct damos_filter {
 		struct damon_addr_range addr_range;
 		int target_idx;
 		struct damon_size_range sz_range;
+		struct {
+			unsigned long range_min;
+			unsigned long range_max;
+		};
 	};
 /* private: */
 	/* List head for siblings. */
Signature · 1 line
2.47.3
author maintainer
#2 lore raw

RFC PATCH v2 2/7 mm/damon/core: extend probe_hits_wsum() for moving sum based calculation

In reply to SJ Park

damon_probe_hits_wsum() is being called only in aggregation time. In future, it could also be used by DAMOS. In this case, since DAMOS uses its own apply_interval, it could be called in sampling time. Then using the not yet fully aggregated probe_hits could result in suboptimum outcomes. Extend damon_probe_hits_wsum() to get the weighted sum based on moving sum to prepare the DAMOS usage.

---
 include/linux/damon.h | 2 +-
 mm/damon/core.c       | 8 ++++++--
 mm/damon/paddr.c      | 2 +-
 mm/damon/vaddr.c      | 2 +-
 4 files changed, 9 insertions(+), 5 deletions(-)
include/linux/damon.h +1 -1
index 1deda16bb70b5..9fa99f92a2e94 100644
@@ -1073,7 +1073,7 @@ unsigned int damon_nr_accesses_mvsum(struct damon_region *r,
 		struct damon_ctx *ctx);
 unsigned char damon_probe_hits_mvsum(int probe_idx, struct damon_region *r,
 		struct damon_ctx *ctx);
-unsigned int damon_probe_hits_wsum(struct damon_region *r, bool last,
+unsigned int damon_probe_hits_wsum(struct damon_region *r, bool last, bool mv,
 		struct damon_ctx *ctx);
 
 int damon_set_regions(struct damon_target *t, struct damon_addr_range *ranges,
mm/damon/core.c +6 -2
index 1752afede01e2..30af5bbcc90e7 100644
@@ -469,11 +469,12 @@ static bool damon_is_last_region(struct damon_region *r,
  * damon_probe_hits_wsum() - Returns probe hits weighted sum of a region.
  * @r:		region to get the weighted sum of.
  * @last:	if the request is for last-window aggregated probe hits.
+ * @mv:		use moving sum.
  * @ctx:	context of &r.
  *
  * Return: the weighted sum of probe hits of the region.
  */
-unsigned int damon_probe_hits_wsum(struct damon_region *r, bool last,
+unsigned int damon_probe_hits_wsum(struct damon_region *r, bool last, bool mv,
 		struct damon_ctx *ctx)
 {
 	struct damon_probe *probe;
@@ -483,6 +484,9 @@ unsigned int damon_probe_hits_wsum(struct damon_region *r, bool last,
 	damon_for_each_probe(probe, ctx) {
 		if (last)
 			sum += r->last_probe_hits[i++] * probe->weight;
+		else if (mv)
+			sum += damon_probe_hits_mvsum(i++, r, ctx) *
+				probe->weight;
 		else
 			sum += r->probe_hits[i++] * probe->weight;
 	}
@@ -3467,7 +3471,7 @@ static unsigned int damon_merge_score(struct damon_region *r, bool last,
 		struct damon_ctx *ctx, bool use_probe_hits)
 {
 	if (use_probe_hits)
-		return damon_probe_hits_wsum(r, last, ctx);
+		return damon_probe_hits_wsum(r, last, false, ctx);
 	if (last)
 		return r->last_nr_accesses;
 	return r->nr_accesses;
mm/damon/paddr.c +1 -1
index d7c81829445ba..b8f4d28165dd7 100644
@@ -202,7 +202,7 @@ static unsigned int damon_pa_apply_probes(struct damon_ctx *ctx,
 				folio_put(folio);
 			if (return_max_wsum)
 				max_wsum = max(damon_probe_hits_wsum(r, false,
-							ctx), max_wsum);
+							false, ctx), max_wsum);
 		}
 	}
 	return max_wsum;
mm/damon/vaddr.c +1 -1
index 9a38dc89a156e..4aecf34e39be2 100644
@@ -704,7 +704,7 @@ static unsigned int damon_va_apply_probes(struct damon_ctx *ctx,
 			__damon_va_apply_probes(ctx, mm, r);
 			if (return_max_wsum)
 				max_wsum = max(damon_probe_hits_wsum(r, false,
-							ctx), max_wsum);
+							false, ctx), max_wsum);
 		}
 		if (mm)
 			mmput(mm);
Signature · 1 line
2.47.3
author maintainer
#3 lore raw

Re: [RFC PATCH v2 2/7] mm/damon/core: extend probe_hits_wsum() for moving sum based calculation

Quoted 6 lines · On Mon, 7 Sep 2026 10:12:11 -0700 SJ Park <[email protected]> wrote:

damon_probe_hits_wsum() is being called only in aggregation time. In future, it could also be used by DAMOS. In this case, since DAMOS uses its own apply_interval, it could be called in sampling time. Then using the not yet fully aggregated probe_hits could result in suboptimum outcomes. Extend damon_probe_hits_wsum() to get the weighted sum based on moving sum to prepare the DAMOS usage.

[...]

mm/damon/vaddr.c static unsigned int damon_va_apply_probes(struct damon_ctx *ctx, · line 710

Quoted 11 lines
mm/damon/vaddr.c +1 -1
@@ -704,7 +704,7 @@ static unsigned int damon_va_apply_probes(struct damon_ctx *ctx,
 			__damon_va_apply_probes(ctx, mm, r);
 			if (return_max_wsum)
 				max_wsum = max(damon_probe_hits_wsum(r, false,
-							ctx), max_wsum);
+							false, ctx), max_wsum);
 		}
 		if (mm)
 			mmput(mm);

This cannot be applied on latest mm-new, because this relies on a not-yet-merged patch series [1]. Hence Sashiko also failed [2] at reviewing this series. I will post another RFC after rebasing to latest mm-new.

[1] https://lore.kernel.org/[email protected]
[2] https://sashiko.dev/#/patchset/20260907171218.101430-1-sj%40kernel.org

Thanks,
SJ

[...]

author maintainer
#4 lore raw

RFC PATCH v2 3/7 mm/damon/core: support probe_hits_wsum damos core filter

In reply to SJ Park

Implement probe_hits_wsum DAMOS core filter support in the core layer. Make three small changes for the support. First, update damos_filter_for_ops() to treat probe_hits_wsum filter as core filter. Second, Update destination damos_filter->range_{min,max} for probe_hits_wsum type damos filter commits. Third, extend damos_filter_match() to handle probe_hits_wsum type filter. Calculate the weighted sum of the given region and compare it with the given filter's target weighted sum range.

---
 mm/damon/core.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
mm/damon/core.c +11 -1
index 30af5bbcc90e7..98efc6e2876f4 100644
@@ -658,6 +658,7 @@ bool damos_filter_for_ops(enum damos_filter_type type)
 	switch (type) {
 	case DAMOS_FILTER_TYPE_ADDR:
 	case DAMOS_FILTER_TYPE_TARGET:
+	case DAMOS_FILTER_TYPE_PROBE_HITS_WSUM:
 		return false;
 	default:
 		break;
@@ -1332,6 +1333,10 @@ static void damos_commit_filter_arg(
 	case DAMOS_FILTER_TYPE_HUGEPAGE_SIZE:
 		dst->sz_range = src->sz_range;
 		break;
+	case DAMOS_FILTER_TYPE_PROBE_HITS_WSUM:
+		dst->range_min = src->range_min;
+		dst->range_max = src->range_max;
+		break;
 	default:
 		break;
 	}
@@ -2510,7 +2515,7 @@ static bool damos_filter_match(struct damon_ctx *ctx, struct damon_target *t,
 	bool matched = false;
 	struct damon_target *ti;
 	int target_idx = 0;
-	unsigned long start, end;
+	unsigned long start, end, wsum;
 
 	switch (filter->type) {
 	case DAMOS_FILTER_TYPE_TARGET:
@@ -2545,6 +2550,11 @@ static bool damos_filter_match(struct damon_ctx *ctx, struct damon_target *t,
 		damon_split_region_at(t, r, end - r->ar.start);
 		matched = true;
 		break;
+	case DAMOS_FILTER_TYPE_PROBE_HITS_WSUM:
+		wsum = damon_probe_hits_wsum(r, false, true, ctx);
+		matched = filter->range_min <= wsum &&
+			wsum <= filter->range_max;
+		break;
 	default:
 		return false;
 	}
Signature · 1 line
2.47.3
author maintainer
#5 lore raw

RFC PATCH v2 4/7 mm/damon/sysfs-schemes: rename sysfs_filter->sz_range to range_{min,max}

In reply to SJ Park

DAMON sysfs interface provides 'min' and 'max' files under the DAMOS filter directory. The purpose is setting the general range arguments for hugepage_size like filters that require range arguments. So far, hugepage_size was the only filter using it. Hence sz_range field of damon_sysfs_scheme_filter struct was connected to the files.

In future, we could add a new filter that can reuse the 'min' and 'max' files. And the filter might use a range of a type that is not size. For example, probe_hits weighted sum. In this case, simply reusing the sz_range field would make it a little confusing. Rename sz_range to range_{min,max} to avoid such confusion.

---
 mm/damon/sysfs-schemes.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)
mm/damon/sysfs-schemes.c +11 -8
index d9b81d7b5910e..4d9147d2a269e 100644
@@ -534,7 +534,8 @@ struct damon_sysfs_scheme_filter {
 	bool allow;
 	char *memcg_path;
 	struct damon_addr_range addr_range;
-	struct damon_size_range sz_range;
+	unsigned long range_min;
+	unsigned long range_max;
 	int target_idx;
 };
 
@@ -588,6 +589,7 @@ damos_sysfs_filter_type_names[] = {
 		.type = DAMOS_FILTER_TYPE_TARGET,
 		.name = "target",
 	},
+
 };
 
 static ssize_t type_show(struct kobject *kobj,
@@ -778,7 +780,7 @@ static ssize_t min_show(struct kobject *kobj,
 	struct damon_sysfs_scheme_filter *filter = container_of(kobj,
 			struct damon_sysfs_scheme_filter, kobj);
 
-	return sysfs_emit(buf, "%lu\n", filter->sz_range.min);
+	return sysfs_emit(buf, "%lu\n", filter->range_min);
 }
 
 static ssize_t min_store(struct kobject *kobj,
@@ -786,7 +788,7 @@ static ssize_t min_store(struct kobject *kobj,
 {
 	struct damon_sysfs_scheme_filter *filter = container_of(kobj,
 			struct damon_sysfs_scheme_filter, kobj);
-	int err = kstrtoul(buf, 0, &filter->sz_range.min);
+	int err = kstrtoul(buf, 0, &filter->range_min);
 
 	return err ? err : count;
 }
@@ -797,7 +799,7 @@ static ssize_t max_show(struct kobject *kobj,
 	struct damon_sysfs_scheme_filter *filter = container_of(kobj,
 			struct damon_sysfs_scheme_filter, kobj);
 
-	return sysfs_emit(buf, "%lu\n", filter->sz_range.max);
+	return sysfs_emit(buf, "%lu\n", filter->range_max);
 }
 
 static ssize_t max_store(struct kobject *kobj,
@@ -805,7 +807,7 @@ static ssize_t max_store(struct kobject *kobj,
 {
 	struct damon_sysfs_scheme_filter *filter = container_of(kobj,
 			struct damon_sysfs_scheme_filter, kobj);
-	int err = kstrtoul(buf, 0, &filter->sz_range.max);
+	int err = kstrtoul(buf, 0, &filter->range_max);
 
 	return err ? err : count;
 }
@@ -2835,12 +2837,13 @@ static int damon_sysfs_add_scheme_filters(struct damos *scheme,
 		} else if (filter->type == DAMOS_FILTER_TYPE_TARGET) {
 			filter->target_idx = sysfs_filter->target_idx;
 		} else if (filter->type == DAMOS_FILTER_TYPE_HUGEPAGE_SIZE) {
-			if (sysfs_filter->sz_range.min >
-					sysfs_filter->sz_range.max) {
+			if (sysfs_filter->range_min >
+					sysfs_filter->range_max) {
 				damos_destroy_filter(filter);
 				return -EINVAL;
 			}
-			filter->sz_range = sysfs_filter->sz_range;
+			filter->sz_range.min = sysfs_filter->range_min;
+			filter->sz_range.max = sysfs_filter->range_max;
 		}
 
 		damos_add_filter(scheme, filter);
Signature · 1 line
2.47.3
author maintainer
#6 lore raw

RFC PATCH v2 5/7 mm/damon/sysfs-schemes: support probe_hits_wsum damos core filter

In reply to SJ Park

Extend DAMON sysfs interface to support probe_hits_wsum input. Also update sysfs input based scheme build logic to setup the min/max probe hits weighted sum range as user provided.

---
 mm/damon/sysfs-schemes.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)
mm/damon/sysfs-schemes.c +12 -1
index 4d9147d2a269e..d9dd4ee0442c7 100644
@@ -589,7 +589,10 @@ damos_sysfs_filter_type_names[] = {
 		.type = DAMOS_FILTER_TYPE_TARGET,
 		.name = "target",
 	},
-
+	{
+		.type = DAMOS_FILTER_TYPE_PROBE_HITS_WSUM,
+		.name = "probe_hits_wsum",
+	},
 };
 
 static ssize_t type_show(struct kobject *kobj,
@@ -2844,6 +2847,14 @@ static int damon_sysfs_add_scheme_filters(struct damos *scheme,
 			}
 			filter->sz_range.min = sysfs_filter->range_min;
 			filter->sz_range.max = sysfs_filter->range_max;
+		} else if (filter->type == DAMOS_FILTER_TYPE_PROBE_HITS_WSUM) {
+			if (sysfs_filter->range_min >
+					sysfs_filter->range_max) {
+				damos_destroy_filter(filter);
+				return -EINVAL;
+			}
+			filter->range_min = sysfs_filter->range_min;
+			filter->range_max = sysfs_filter->range_max;
 		}
 
 		damos_add_filter(scheme, filter);
Signature · 1 line
2.47.3
author maintainer
#7 lore raw

RFC PATCH v2 6/7 Docs/mm/damon/design: update for probe_hits_wsum DAMOS core filter

In reply to SJ Park

Update DAMON design document for the newly added probe hits weighted sum based DAMOS core filter type.

---
 Documentation/mm/damon/design.rst | 3 +++
 1 file changed, 3 insertions(+)
Documentation/mm/damon/design.rst +3 -0
index aac84de261aa8..8e980dde362de 100644
@@ -822,6 +822,9 @@ Below ``type`` of filters are currently supported.
         - Applied to pages that belonging to a given address range.
     - target
         - Applied to pages that belonging to a given DAMON monitoring target.
+    - probe_hits_wsum
+        - Matches to monitoring regions having a given range of :ref:`probe
+          hits weighted sum <damon_design_attrs_only_monitoring>` value.
 - Operations layer handled, supported by only ``paddr`` operations set.
     - anon
         - Applied to pages that containing data that not stored in files.
Signature · 1 line
2.47.3
author maintainer
#8 lore raw

RFC PATCH v2 7/7 Docs/admin-guide/mm/damon/usage: update for probe_hits_wsum DAMOS filter

In reply to SJ Park

Update DAMON usage document for the newly added probe hits weighted sum based DAMOS core filter type.

---
 Documentation/admin-guide/mm/damon/usage.rst | 4 ++++
 1 file changed, 4 insertions(+)
Documentation/admin-guide/mm/damon/usage.rst +4 -0
index 023c6334024f8..d3e37400367bd 100644
@@ -551,6 +551,10 @@ and ``damon_target_idx``.  To ``type`` file, you can write the type of the
 filter.  Refer to :ref:`the design doc <damon_design_damos_filters>` for
 available type names, their meaning and on what layer those are handled.
 
+For ``probe_hits_wsum`` type, you can specify the minimum and maximum probe
+hits weighted sum value for the filter to ``min`` and ``max`` files,
+respectively.
+
 For ``memcg`` type, you can specify the memory cgroup of the interest by
 writing the path of the memory cgroup from the cgroups mount point to
 ``memcg_path`` file.  For ``addr`` type, you can specify the start and end
Signature · 1 line
2.47.3