mm/damon/core: split a fraction of regions when nr_regions exceeds max/2

Patch series "mm/damon/core: detect internal variation above
max_nr_regions/2", v3.

kdamond_split_regions() bails out early when nr_regions is already above
max_nr_regions / 2.  A large region that picks up new internal variation
after that point never gets split, so we lose visibility into its hot/cold
structure.

We hit this with damon-paddr on hugepage workloads and damon-vaddr on
processes that mmap a large anonymous range.

Example with max_nr_regions == 1500.  A target ends up with 799 small
hot/cold regions plus one big region (an earlier merge collapsed a
uniformly-accessed range into a single piece):

H:hot
C:cold

      r1     r2     r3                 r800
    HHHHHH|CCCCCC|HHHHHH|...|HHHHHH..........................|

    nr_regions = 800  >  max_nr_regions / 2 = 750

Now a cold subarea shows up inside r800:

      r1     r2     r3                 r800
    HHHHHH|CCCCCC|HHHHHH|...|HHHHHH........CCCCCC.............|

The small regions can't merge with each other (their access counts
differ), so budget never frees up.  r800 can't be split because nr_regions
> max_nr_regions / 2 returns early.  The cold subarea stays invisible.

Patch 1 keeps refining on this path: when nr_regions is above
max_nr_regions / 2 but still under the maximum, it splits a fraction of
the regions instead of returning.  The fraction shrinks as the remaining
budget shrinks, so the count approaches max_nr_regions smoothly.  A
useless split is undone by the next merge cycle.

Patch 2 adds a KUnit test for the case where nr_regions is already above
max_nr_regions / 2.

Thanks to SJ for the suggestion to drive the split fraction from the
remaining budget rather than an age-based filter.


This patch (of 2):

kdamond_split_regions() returns early when nr_regions is above
max_nr_regions / 2, leaving internal access variation inside a large
region undetected.

Such a layout is common with damon-paddr on hugepage workloads or
damon-vaddr on processes with a large anonymous mmap.

For example, with max_nr_regions == 1500, a target may end up with 799
small alternating-temperature regions plus one large region that absorbed
a uniformly-accessed range during an earlier merge:

H:hot
C:cold

      r1     r2     r3                 r800
    HHHHHH|CCCCCC|HHHHHH|...|HHHHHH..........................|

    nr_regions = 800  >  max_nr_regions / 2 = 750

If a cold subarea later emerges inside r800:

      r1     r2     r3                 r800
    HHHHHH|CCCCCC|HHHHHH|...|HHHHHH........CCCCCC.............|

The small regions cannot merge with each other (different access counts),
so the budget stays full.  r800 cannot be split because nr_regions >
max_nr_regions / 2 causes an early return.  The cold subarea is never
discovered.

When nr_regions is above max_nr_regions / 2 but still under the maximum,
split only a fraction of the regions instead of returning.  One region in
every 'max_nr_regions / budget' regions is split, where budget is the
remaining room (max_nr_regions - nr_regions), starting from a rotating
offset so different regions get picked over time.  The fraction shrinks as
the budget shrinks, so the region count keeps refining while approaching
max_nr_regions smoothly rather than overshooting it.  An unnecessary split
is reverted by the next kdamond_merge_regions().

Link: https://lore.kernel.org/20260629145630.134891-1-sj@kernel.org
Link: https://lore.kernel.org/20260626085851.70754-2-jiayuan.chen@linux.dev
Link: https://lore.kernel.org/20260629145630.134891-2-sj@kernel.org
Signed-off-by: Jiayuan Chen <jiayuan.chen@shopee.com>
Signed-off-by: SJ Park <sj@kernel.org>
Reviewed-by: SJ Park <sj@kernel.org>
Cc: Shu Anzai <shu17az@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/core.c

@@ -3247,6 +3247,37 @@ static void damon_split_regions_of(struct damon_ctx *ctx,
 	}
 }
 
+/* Split one in every @split_step regions into two, from a rotating offset */
+static void damon_split_some_regions(struct damon_ctx *ctx,
+				     unsigned long split_step)
+{
+	static unsigned long rotation;
+	struct damon_target *t;
+	struct damon_region *r, *next;
+	unsigned long offset = rotation++ % split_step;
+	unsigned long idx = 0;
+
+	damon_for_each_target(t, ctx) {
+		damon_for_each_region_safe(r, next, t) {
+			unsigned long sz_region, sz_sub;
+
+			if (idx++ % split_step != offset)
+				continue;
+			sz_region = damon_sz_region(r);
+			if (sz_region < 2 * ctx->min_region_sz)
+				continue;
+
+			sz_sub = ALIGN_DOWN(damon_rand(ctx, 1, 10) *
+					sz_region / 10, ctx->min_region_sz);
+			/* Do not allow blank region */
+			if (sz_sub == 0 || sz_sub >= sz_region)
+				continue;
+
+			damon_split_region_at(t, r, sz_sub);
+		}
+	}
+}
+
 /*
  * Split every target region into randomly-sized small regions
  *
@@ -3260,25 +3291,33 @@ static void damon_split_regions_of(struct damon_ctx *ctx,
 static void kdamond_split_regions(struct damon_ctx *ctx)
 {
 	struct damon_target *t;
-	unsigned int nr_regions = 0;
-	static unsigned int last_nr_regions;
+	unsigned long nr_regions = 0;
+	unsigned long max_nr_regions = ctx->attrs.max_nr_regions;
+	static unsigned long last_nr_regions;
 	int nr_subregions = 2;
 
 	damon_for_each_target(t, ctx)
 		nr_regions += damon_nr_regions(t);
 
-	if (nr_regions > ctx->attrs.max_nr_regions / 2)
-		return;
+	if (nr_regions >= max_nr_regions)
+		goto done;
+
+	if (nr_regions > max_nr_regions / 2) {
+		damon_split_some_regions(ctx,
+			max_nr_regions / (max_nr_regions - nr_regions));
+		goto done;
+	}
 
 	/* Maybe the middle of the region has different access frequency */
 	if (last_nr_regions == nr_regions &&
-			nr_regions < ctx->attrs.max_nr_regions / 3)
+			nr_regions < max_nr_regions / 3)
 		nr_subregions = 3;
 
 	damon_for_each_target(t, ctx)
 		damon_split_regions_of(ctx, t, nr_subregions,
 				       ctx->min_region_sz);
 
+done:
 	last_nr_regions = nr_regions;
 }