[PATCH v5 0/2] mm/damon/core: iterate over installed probes instead of DAMON_MAX_PROBES

[PATCH v5 0/2] mm/damon/core: iterate over installed probes instead of DAMON_MAX_PROBES

[SJ Park]

From: Suhaas Joshi <suhaas@s-joshi.in>

There were two "todo: " items of the same nature in mm/damon/core.c.
Both of these suggested that we should iterate over only installed
probes while updating (last_)probe_hits[] during merging/splitting
regions, instead of iterating over the entire array, which may have
empty/unused indices.

So this patch series does exactly that. It replaces the for-loop over
[0, DAMON_MAX_PROBES) with a damon_for_each_probe() over the probe list
of the context.

Reviewer note.

These patches are based off mm-new. They pass all KUnit tests related to
damon. I have boot tested (but nothing beyond) on qemu-x86_64. Please do
let me know if I should do any specific tests and link the logs here.

The series has been tested with mm-new fetched on Oct 14, 2026.

Changes from v4
- v4: https://lore.kernel.org/20261015135533.18761-1-suhaas@s-joshi.in
- Collect R-b: from SJ.
- Remove/separate messages in cover letter that not suitable for commit
  message.
- Wrap cover letter for 75 columns length.
- Rebase to latest mm-new.

Changes v3 -> v4:
1. In some places, kunit_skip() was being called before freeing
   resources.  This is correctly ordered now.
2. In some places, regions were not being freed. In v4, they are being
   freed.
Link to v3: https://lore.kernel.org/damon/20261012145636.168436-1-suhaas@s-joshi.in/T/#t

Changes v2 -> v3:
1. In test functions, new probe and context allocations happen after
   target, region etc have been allocated. Further, target is added to
   the context.  These 2 changes help in reducing allocation code and
   improving simplicity and readability of the patches. This change is
   true for both patches.  (on SJ suggestion)
2. Earlier, if a function's argument list required multiple lines, I was
   aligning these lines to the opening paranthesis of the function. In
   v3, these begin after 2 tab spaces. (On SJ suggestion)
3. damon_test_last_region_quota_reset() uses the older version of
   damos_skip_charged_region().  Make it use the new version.
Link to v2: https://lore.kernel.org/damon/20261008151944.113714-1-suhaas@s-joshi.in/

Changes v1 -> v2:
1. simplified commit message of patch 1 (on Liew's suggestion)
2. wrapped all lines at 80 characters (on SJ's suggestion)
3. added comments for new argument in function (on SJ's suggestion)
4. added damon_ctx_destroy() in r2 failure case in
   damon_test_merge_two() (on SJ/Sashiko/Liew suggestion)
Link to v1: https://lore.kernel.org/damon/20261007154851.45368-1-suhaas@s-joshi.in/

SJ Park (1):
  mm/damon/core: copy only installed probe in damon_split_region_at()

Suhaas Joshi (1):
  mm/damon/core: use only installed probe in damon_merge_two_regions()

 mm/damon/core.c             | 57 +++++++++++++++----------
 mm/damon/tests/core-kunit.h | 84 +++++++++++++++++++++++++++++++------
 2 files changed, 106 insertions(+), 35 deletions(-)


base-commit: 79e2e91bd5a4b213cf40819c2711a9364b2c0f86
-- 
2.47.3

[PATCH v5 1/2] mm/damon/core: use only installed probe in damon_merge_two_regions()

[SJ Park]

From: Suhaas Joshi <suhaas@s-joshi.in>

While merging 2 regions, we iterate over the entire probe_hits[] array,
whose size is determined by the DAMON_MAX_PROBES macro. However, it is
possible that we have fewer probes installed than DAMON_MAX_PROBES. In such
cases, we end up making redundant iterations. Therefore, to remedy this,
iterate over the list of installed probes instead of iterating over the
entire array. For doing this, start accepting a struct damon_ctx in
damon_merge_two_regions(), and update calling functions to pass this
argument.

Update the damon_test_merge_two() test to use this new signature for
damon_merge_two_regions() as well.

Signed-off-by: Suhaas Joshi <suhaas@s-joshi.in>
Reviewed-by: SJ Park <sj@kernel.org>
Signed-off-by: SJ Park <sj@kernel.org>
---
 mm/damon/core.c             | 15 ++++++++++-----
 mm/damon/tests/core-kunit.h | 20 ++++++++++++++++++--
 2 files changed, 28 insertions(+), 7 deletions(-)

diff --git a/mm/damon/core.c b/mm/damon/core.c
index 9d54899004026..36d60f3e1de62 100644
--- a/mm/damon/core.c
+++ b/mm/damon/core.c
@@ -3549,20 +3549,25 @@ static void damon_verify_merge_two_regions(
 /*
  * Merge two adjacent regions into one region
  */
-static void damon_merge_two_regions(struct damon_target *t,
-		struct damon_region *l, struct damon_region *r)
+static void damon_merge_two_regions(struct damon_ctx *ctx,
+		struct damon_target *t, struct damon_region *l,
+		struct damon_region *r)
 {
 	unsigned long sz_l = damon_sz_region(l), sz_r = damon_sz_region(r);
 	int i;
+	struct damon_probe *p;
 
 	l->nr_accesses = (l->nr_accesses * sz_l + r->nr_accesses * sz_r) /
 			(sz_l + sz_r);
 	l->age = (l->age * sz_l + r->age * sz_r) / (sz_l + sz_r);
 	l->ar.end = r->ar.end;
-	/* todo: do this for only installed probes */
-	for (i = 0; i < DAMON_MAX_PROBES; i++)
+
+	i = 0;
+	damon_for_each_probe(p, ctx) {
 		l->probe_hits[i] = (l->probe_hits[i] * sz_l + r->probe_hits[i]
 				* sz_r) / (sz_l + sz_r);
+		++i;
+	}
 	damon_verify_merge_two_regions(l, r);
 	damon_destroy_region(r, t);
 }
@@ -3615,7 +3620,7 @@ static void damon_merge_regions_of(struct damon_target *t, unsigned int thres,
 			goto set_prev_continue;
 		if (damon_sz_region(prev) + damon_sz_region(r) > sz_limit)
 			goto set_prev_continue;
-		damon_merge_two_regions(t, prev, r);
+		damon_merge_two_regions(ctx, t, prev, r);
 		continue;
 set_prev_continue:
 		prev = r;
diff --git a/mm/damon/tests/core-kunit.h b/mm/damon/tests/core-kunit.h
index 43e2f950f8ff4..ca10d450f7e23 100644
--- a/mm/damon/tests/core-kunit.h
+++ b/mm/damon/tests/core-kunit.h
@@ -182,6 +182,8 @@ static void damon_test_merge_two(struct kunit *test)
 {
 	struct damon_target *t;
 	struct damon_region *r, *r2, *r3;
+	struct damon_probe *p;
+	struct damon_ctx *ctx;
 	int i;
 
 	t = damon_new_target();
@@ -206,7 +208,21 @@ static void damon_test_merge_two(struct kunit *test)
 	r2->age = 21;
 	damon_add_region(r2, t);
 
-	damon_merge_two_regions(t, r, r2);
+	p = damon_new_probe();
+	if (!p) {
+		damon_free_target(t);
+		kunit_skip(test, "probe alloc fail");
+	}
+	ctx = damon_new_ctx();
+	if (!ctx) {
+		damon_destroy_probe(p);
+		damon_free_target(t);
+		kunit_skip(test, "context alloc fail");
+	}
+	damon_add_probe(ctx, p);
+	damon_add_target(ctx, t);
+
+	damon_merge_two_regions(ctx, t, r, r2);
 	KUNIT_EXPECT_EQ(test, r->ar.start, 0ul);
 	KUNIT_EXPECT_EQ(test, r->ar.end, 300ul);
 	KUNIT_EXPECT_EQ(test, r->nr_accesses, 16u);
@@ -220,7 +236,7 @@ static void damon_test_merge_two(struct kunit *test)
 	}
 	KUNIT_EXPECT_EQ(test, i, 1);
 
-	damon_free_target(t);
+	damon_destroy_ctx(ctx);
 }
 
 static struct damon_region *__nth_region_of(struct damon_target *t, int idx)
-- 
2.47.3

[PATCH v5 2/2] mm/damon/core: copy only installed probe in damon_split_region_at()

[SJ Park]

In damon_split_region_at(), we copy the entire probe_hits[] and
last_probe_hits[] arrays of DAMON_MAX_PROBES size. However it is possible
that the number of probes installed is less than DAMON_MAX_PROBES, in which
case one or more iterations is/are redundant. To remedy this, iterate over
only as many times as the number of installed probes. To do so, accept a
struct damon_ctx pointer in damon_split_region_at(), and update all
invocations of the function to pass the new argument.

Also update damon_test_split_at() and damos_test_filter_out() in
core-kunit.h to supply a context to function paths that call
damon_split_region_at().

Signed-off-by: Suhaas Joshi <suhaas@s-joshi.in>
Reviewed-by: SJ Park <sj@kernel.org>
Signed-off-by: SJ Park <sj@kernel.org>
---
 mm/damon/core.c             | 42 +++++++++++++-----------
 mm/damon/tests/core-kunit.h | 64 +++++++++++++++++++++++++++++++------
 2 files changed, 78 insertions(+), 28 deletions(-)

diff --git a/mm/damon/core.c b/mm/damon/core.c
index 36d60f3e1de62..d36823f4d0928 100644
--- a/mm/damon/core.c
+++ b/mm/damon/core.c
@@ -2046,8 +2046,8 @@ static unsigned long damon_region_sz_limit(struct damon_ctx *ctx)
 	return sz;
 }
 
-static int damon_split_region_at(struct damon_target *t,
-				  struct damon_region *r, unsigned long sz_r);
+static int damon_split_region_at(struct damon_ctx *ctx, struct damon_target *t,
+		struct damon_region *r, unsigned long sz_r);
 
 /*
  * damon_apply_min_nr_regions() - Make effect of min_nr_regions parameter.
@@ -2072,7 +2072,7 @@ static unsigned long damon_apply_min_nr_regions(struct damon_ctx *ctx)
 	damon_for_each_target(t, ctx) {
 		damon_for_each_region_safe(r, next, t) {
 			while (damon_sz_region(r) > max_region_sz) {
-				if (damon_split_region_at(t, r, max_region_sz))
+				if (damon_split_region_at(ctx, t, r, max_region_sz))
 					goto out;
 				r = damon_next_region(r);
 			}
@@ -2469,6 +2469,7 @@ static bool damos_valid_target(struct damon_ctx *c, struct damon_region *r,
 /*
  * damos_skip_charged_region() - Check if the given region or starting part of
  * it is already charged for the DAMOS quota.
+ * @ctx: The pointer to the context.
  * @t:	The target of the region.
  * @rp:	The pointer to the region.
  * @s:	The scheme to be applied.
@@ -2491,8 +2492,8 @@ static bool damos_valid_target(struct damon_ctx *c, struct damon_region *r,
  *
  * Return: true if the region should be skipped, false otherwise.
  */
-static bool damos_skip_charged_region(struct damon_target *t,
-		struct damon_region *r, struct damos *s,
+static bool damos_skip_charged_region(struct damon_ctx *ctx,
+		struct damon_target *t, struct damon_region *r, struct damos *s,
 		unsigned long min_region_sz)
 {
 	struct damos_quota *quota = &s->quota;
@@ -2520,7 +2521,7 @@ static bool damos_skip_charged_region(struct damon_target *t,
 				}
 				sz_to_skip = min_region_sz;
 			}
-			damon_split_region_at(t, r, sz_to_skip);
+			damon_split_region_at(ctx, t, r, sz_to_skip);
 			skip = true;
 		}
 	}
@@ -2579,13 +2580,13 @@ static bool damos_filter_match(struct damon_ctx *ctx, struct damon_target *t,
 		}
 		/* start before the range and overlap */
 		if (r->ar.start < start) {
-			*err = damon_split_region_at(t, r,
+			*err = damon_split_region_at(ctx, t, r,
 					start - r->ar.start);
 			matched = false;
 			break;
 		}
 		/* start inside the range */
-		*err = damon_split_region_at(t, r, end - r->ar.start);
+		*err = damon_split_region_at(ctx, t, r, end - r->ar.start);
 		matched = true;
 		break;
 	case DAMOS_FILTER_TYPE_PROBE_HITS_WSUM:
@@ -2784,7 +2785,7 @@ static void damos_apply_scheme(struct damon_ctx *c, struct damon_target *t,
 					c->min_region_sz);
 			if (!sz)
 				goto update_stat;
-			if (damon_split_region_at(t, r, sz))
+			if (damon_split_region_at(c, t, r, sz))
 				goto update_stat;
 		}
 		if (damos_core_filter_out(c, t, r, s))
@@ -2831,7 +2832,7 @@ static void damon_do_apply_schemes(struct damon_ctx *c,
 		if (damos_quota_is_full(quota, c->min_region_sz))
 			continue;
 
-		if (damos_skip_charged_region(t, r, s, c->min_region_sz))
+		if (damos_skip_charged_region(c, t, r, s, c->min_region_sz))
 			continue;
 
 		if (s->max_nr_snapshots &&
@@ -3690,15 +3691,18 @@ static void damon_verify_split_region_at(struct damon_region *r,
 /*
  * Split a region in two
  *
+ * ctx		pointer to the region's context
  * r		the region to be split
  * sz_r		size of the first sub-region that will be made
  *
  * Return: 0 on success, negative error code otherwise.
  */
-static int damon_split_region_at(struct damon_target *t,
-				  struct damon_region *r, unsigned long sz_r)
+static int damon_split_region_at(struct damon_ctx *ctx, struct damon_target *t,
+		struct damon_region *r, unsigned long sz_r)
 {
 	struct damon_region *new;
+	struct damon_probe *p;
+	int i;
 
 	damon_verify_split_region_at(r, sz_r);
 	new = damon_new_region(r->ar.start + sz_r, r->ar.end);
@@ -3710,11 +3714,13 @@ static int damon_split_region_at(struct damon_target *t,
 	new->age = r->age;
 	new->last_nr_accesses = r->last_nr_accesses;
 	new->nr_accesses = r->nr_accesses;
-	/* todo: do this for only installed probes */
-	memcpy(new->probe_hits, r->probe_hits, sizeof(r->probe_hits));
-	memcpy(new->last_probe_hits, r->last_probe_hits,
-			sizeof(r->last_probe_hits));
 
+	i = 0;
+	damon_for_each_probe(p, ctx) {
+		new->probe_hits[i] = r->probe_hits[i];
+		new->last_probe_hits[i] = r->last_probe_hits[i];
+		++i;
+	}
 	damon_insert_region(new, r, damon_next_region(r), t);
 	return 0;
 }
@@ -3743,7 +3749,7 @@ static void damon_split_regions_of(struct damon_ctx *ctx,
 			if (sz_sub == 0 || sz_sub >= sz_region)
 				continue;
 
-			damon_split_region_at(t, r, sz_sub);
+			damon_split_region_at(ctx, t, r, sz_sub);
 			sz_region = sz_sub;
 		}
 	}
@@ -3776,7 +3782,7 @@ static void damon_split_some_regions(struct damon_ctx *ctx,
 			if (sz_sub == 0 || sz_sub >= sz_region)
 				continue;
 
-			damon_split_region_at(t, r, sz_sub);
+			damon_split_region_at(ctx, t, r, sz_sub);
 		}
 	}
 }
diff --git a/mm/damon/tests/core-kunit.h b/mm/damon/tests/core-kunit.h
index ca10d450f7e23..a5ca4bfe2f0db 100644
--- a/mm/damon/tests/core-kunit.h
+++ b/mm/damon/tests/core-kunit.h
@@ -141,6 +141,8 @@ static void damon_test_split_at(struct kunit *test)
 {
 	struct damon_target *t;
 	struct damon_region *r, *r_new;
+	struct damon_ctx *ctx;
+	struct damon_probe *p;
 
 	t = damon_new_target();
 	if (!t)
@@ -150,13 +152,29 @@ static void damon_test_split_at(struct kunit *test)
 		damon_free_target(t);
 		kunit_skip(test, "region alloc fail");
 	}
+	p = damon_new_probe();
+	if (!p) {
+		damon_free_region(r);
+		damon_free_target(t);
+		kunit_skip(test, "probe alloc fail");
+	}
+	ctx = damon_new_ctx();
+	if (!ctx) {
+		damon_destroy_probe(p);
+		damon_free_region(r);
+		damon_free_target(t);
+		kunit_skip(test, "context alloc fail");
+	}
+	damon_add_probe(ctx, p);
+	damon_add_target(ctx, t);
+
 	r->nr_accesses = 42;
 	r->last_nr_accesses = 15;
 	r->probe_hits[0] = 7;
 	r->last_probe_hits[0] = 3;
 	r->age = 10;
 	damon_add_region(r, t);
-	damon_split_region_at(t, r, 25);
+	damon_split_region_at(ctx, t, r, 25);
 	KUNIT_EXPECT_EQ(test, damon_nr_regions(t), 2);
 	if (damon_nr_regions(t) != 2)
 		goto out;
@@ -175,7 +193,7 @@ static void damon_test_split_at(struct kunit *test)
 	KUNIT_EXPECT_EQ(test, r->age, r_new->age);
 
 out:
-	damon_free_target(t);
+	damon_destroy_ctx(ctx);
 }
 
 static void damon_test_merge_two(struct kunit *test)
@@ -1799,6 +1817,8 @@ static void damos_test_filter_out(struct kunit *test)
 	struct damon_target *t;
 	struct damon_region *r, *r2;
 	struct damos_filter *f;
+	struct damon_ctx *ctx;
+	struct damon_probe *p;
 	int err;
 
 	f = damos_new_filter(DAMOS_FILTER_TYPE_ADDR, true, false);
@@ -1819,28 +1839,44 @@ static void damos_test_filter_out(struct kunit *test)
 	}
 	damon_add_region(r, t);
 
+	p = damon_new_probe();
+	if (!p) {
+		damon_free_target(t);
+		damos_destroy_filter(f);
+		kunit_skip(test, "probe alloc fail");
+	}
+	ctx = damon_new_ctx();
+	if (!ctx) {
+		damon_destroy_probe(p);
+		damon_free_target(t);
+		damos_destroy_filter(f);
+		kunit_skip(test, "context alloc fail");
+	}
+	damon_add_probe(ctx, p);
+	damon_add_target(ctx, t);
+
 	/* region in the range */
-	KUNIT_EXPECT_TRUE(test, damos_filter_match(NULL, t, r, f, 1, &err));
+	KUNIT_EXPECT_TRUE(test, damos_filter_match(ctx, t, r, f, 1, &err));
 	KUNIT_EXPECT_EQ(test, damon_nr_regions(t), 1);
 
 	/* region before the range */
 	r->ar.start = 1;
 	r->ar.end = 2;
 	KUNIT_EXPECT_FALSE(test,
-			damos_filter_match(NULL, t, r, f, 1, &err));
+			damos_filter_match(ctx, t, r, f, 1, &err));
 	KUNIT_EXPECT_EQ(test, damon_nr_regions(t), 1);
 
 	/* region after the range */
 	r->ar.start = 6;
 	r->ar.end = 8;
 	KUNIT_EXPECT_FALSE(test,
-			damos_filter_match(NULL, t, r, f, 1, &err));
+			damos_filter_match(ctx, t, r, f, 1, &err));
 	KUNIT_EXPECT_EQ(test, damon_nr_regions(t), 1);
 
 	/* region started before the range */
 	r->ar.start = 1;
 	r->ar.end = 4;
-	KUNIT_EXPECT_FALSE(test, damos_filter_match(NULL, t, r, f, 1, &err));
+	KUNIT_EXPECT_FALSE(test, damos_filter_match(ctx, t, r, f, 1, &err));
 	/* filter should have split the region */
 	KUNIT_EXPECT_EQ(test, r->ar.start, 1);
 	KUNIT_EXPECT_EQ(test, r->ar.end, 2);
@@ -1856,7 +1892,7 @@ static void damos_test_filter_out(struct kunit *test)
 	r->ar.start = 2;
 	r->ar.end = 8;
 	KUNIT_EXPECT_TRUE(test,
-			damos_filter_match(NULL, t, r, f, 1, &err));
+			damos_filter_match(ctx, t, r, f, 1, &err));
 	/* filter should have split the region */
 	KUNIT_EXPECT_EQ(test, r->ar.start, 2);
 	KUNIT_EXPECT_EQ(test, r->ar.end, 6);
@@ -1869,7 +1905,7 @@ static void damos_test_filter_out(struct kunit *test)
 	damon_destroy_region(r2, t);
 
 out:
-	damon_free_target(t);
+	damon_destroy_ctx(ctx);
 	damos_free_filter(f);
 }
 
@@ -2459,6 +2495,7 @@ static void damos_test_esz_goal_temporal(struct kunit *test)
 static void damon_test_last_region_quota_reset(struct kunit *test)
 {
 	struct damos s = {0};
+	struct damon_ctx *ctx;
 	struct damon_target *t;
 	struct damon_region *r1, *r2;
 	unsigned long min_region_sz = 1;
@@ -2485,7 +2522,14 @@ static void damon_test_last_region_quota_reset(struct kunit *test)
 	s.quota.charge_target_from = t;
 	s.quota.charge_addr_from = r1->ar.end;
 
-	skipped = damos_skip_charged_region(t, r2, &s, min_region_sz);
+	ctx = damon_new_ctx();
+	if (!ctx) {
+		damon_free_target(t);
+		kunit_skip(test, "ctx alloc failed");
+	}
+	damon_add_target(ctx, t);
+
+	skipped = damos_skip_charged_region(ctx, t, r2, &s, min_region_sz);
 
 	/* 'r2' is not processed, it should not skip */
 	KUNIT_EXPECT_EQ(test, skipped, false);
@@ -2494,7 +2538,7 @@ static void damon_test_last_region_quota_reset(struct kunit *test)
 	KUNIT_EXPECT_PTR_EQ(test, s.quota.charge_target_from, NULL);
 	KUNIT_EXPECT_EQ(test, s.quota.charge_addr_from, 0);
 
-	damon_free_target(t);
+	damon_destroy_ctx(ctx);
 }
 
 static struct kunit_case damon_test_cases[] = {
-- 
2.47.3