From 1a582e6e01744be5a70cdf77fa7ee8f19e6c3a5a Mon Sep 17 00:00:00 2001 From: Jiayuan Chen Date: Mon, 29 Jun 2026 07:56:28 -0700 Subject: [PATCH] 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 Signed-off-by: SJ Park Reviewed-by: SJ Park Cc: Shu Anzai Signed-off-by: Andrew Morton --- mm/damon/core.c | 49 ++++++++++++++++++++++++++++++++++++++++++++----- 1 file changed, 44 insertions(+), 5 deletions(-) diff --git a/mm/damon/core.c b/mm/damon/core.c index 957cd8067bc0..a99458c57851 100644 --- a/mm/damon/core.c +++ b/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; }