Skip to contentSkip to conversation

mm/mglru: performance optimizations

PATCH Superseded Yu Zhao 1 participant 7 messages Dec 2, 2024

Subsystems (4)

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Yu Zhao

author
#0 lore raw

cover mm/mglru: performance optimizations

This series improves performance for some previously discussed test cases. Most of the code changes gathered here has been floating on the mailing list [1][2]. They are now properly organized and have gone through various benchmarks on both client and server devices, including Android, FIO, memcached, multiple VMs and MongoDB.

[1] https://lore.kernel.org/CAOUHufahuWcKf5f1Sg3emnqX+cODuR=2TQo7T4Gr-QYLujn4RA@mail.gmail.com/
[2] https://lore.kernel.org/CAOUHufawNerxqLm7L9Yywp3HJFiYVrYO26ePUb1jH-qxNGWzyA@mail.gmail.com/

Yu Zhao (6):
mm/mglru: clean up workingset
mm/mglru: optimize deactivation
mm/mglru: rework aging feedback
mm/mglru: rework type selection
mm/mglru: rework refault detection
mm/mglru: rework workingset protection

 include/linux/mm_inline.h |  94 +++++----
 include/linux/mmzone.h    |  88 +++++---
 mm/swap.c                 |  65 ++++--
 mm/vmscan.c               | 423 +++++++++++++++++---------------------
 mm/workingset.c           |  69 +++----
 5 files changed, 374 insertions(+), 365 deletions(-)
Signature · 1 line
2.47.0.338.g60cca15819-goog
author
#1 lore raw

PATCH 1/6 mm/mglru: clean up workingset

Move VM_BUG_ON_FOLIO() to cover both the default and MGLRU paths. Also use a pair of rcu_read_lock() and rcu_read_unlock() within each path, to improve readability.

This change should not have any side effects.

---
 mm/workingset.c | 25 ++++++++++++-------------
 1 file changed, 12 insertions(+), 13 deletions(-)
mm/workingset.c +12 -13
index a4705e196545..ad181d1b8cf1 100644
@@ -428,17 +428,17 @@ bool workingset_test_recent(void *shadow, bool file, bool *workingset,
 	struct pglist_data *pgdat;
 	unsigned long eviction;
 
-	rcu_read_lock();
-
 	if (lru_gen_enabled()) {
-		bool recent = lru_gen_test_recent(shadow, file,
-				&eviction_lruvec, &eviction, workingset);
+		bool recent;
 
+		rcu_read_lock();
+		recent = lru_gen_test_recent(shadow, file, &eviction_lruvec,
+					     &eviction, workingset);
 		rcu_read_unlock();
 		return recent;
 	}
 
-
+	rcu_read_lock();
 	unpack_shadow(shadow, &memcgid, &pgdat, &eviction, workingset);
 	eviction <<= bucket_order;
 
@@ -459,14 +459,12 @@ bool workingset_test_recent(void *shadow, bool file, bool *workingset,
 	 * configurations instead.
 	 */
 	eviction_memcg = mem_cgroup_from_id(memcgid);
-	if (!mem_cgroup_disabled() &&
-	    (!eviction_memcg || !mem_cgroup_tryget(eviction_memcg))) {
-		rcu_read_unlock();
+	if (!mem_cgroup_tryget(eviction_memcg))
+		eviction_memcg = NULL;
+	rcu_read_unlock();
+
+	if (!mem_cgroup_disabled() && !eviction_memcg)
 		return false;
-	}
-
-	rcu_read_unlock();
-
 	/*
 	 * Flush stats (and potentially sleep) outside the RCU read section.
 	 *
@@ -544,6 +542,8 @@ void workingset_refault(struct folio *folio, void *shadow)
 	bool workingset;
 	long nr;
 
+	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
+
 	if (lru_gen_enabled()) {
 		lru_gen_refault(folio, shadow);
 		return;
@@ -558,7 +558,6 @@ void workingset_refault(struct folio *folio, void *shadow)
 	 * is actually experiencing the refault event. Make sure the folio is
 	 * locked to guarantee folio_memcg() stability throughout.
 	 */
-	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
 	nr = folio_nr_pages(folio);
 	memcg = folio_memcg(folio);
 	pgdat = folio_pgdat(folio);
Signature · 1 line
2.47.0.338.g60cca15819-goog
author
#2 lore raw

PATCH 2/6 mm/mglru: optimize deactivation

In reply to Yu Zhao

Do not shuffle a folio in the deactivation paths if it is already in the oldest generation. This reduces the LRU lock contention.

Before this patch, the contention is reproducible by FIO, e.g.,

  fio -filename=/dev/nvme1n1p2 -direct=0 -thread -size=1024G \
      -rwmixwrite=30  --norandommap --randrepeat=0 -ioengine=sync \
      -bs=4k -numjobs=400 -runtime=25000 --time_based \
      -group_reporting -name=mglru

98.96%--_raw_spin_lock_irqsave

          folio_lruvec_lock_irqsave
          |
           --98.78%--folio_batch_move_lru
               |
                --98.63%--deactivate_file_folio
                          mapping_try_invalidate
                          invalidate_mapping_pages
                          invalidate_bdev
                          blkdev_common_ioctl
                          blkdev_ioctl

After this patch, deactivate_file_folio() bails out early without taking the LRU lock.

A side effect is that a folio can be left at the head of the oldest generation, rather than the tail. If reclaim happens at the same time, it cannot reclaim this folio immediately. Since there is no known correlation between truncation and reclaim, this side effect is considered insignificant.

---
 mm/swap.c | 44 +++++++++++++++++++++++++++++++++++++-------
 1 file changed, 37 insertions(+), 7 deletions(-)
mm/swap.c +37 -7
index 3a01acfd5a89..15a94be8b0af 100644
@@ -379,7 +379,8 @@ static void __lru_cache_activate_folio(struct folio *folio)
 }
 
 #ifdef CONFIG_LRU_GEN
-static void folio_inc_refs(struct folio *folio)
+
+static void lru_gen_inc_refs(struct folio *folio)
 {
 	unsigned long new_flags, old_flags = READ_ONCE(folio->flags);
 
@@ -406,10 +407,30 @@ static void folio_inc_refs(struct folio *folio)
 		new_flags |= old_flags & ~LRU_REFS_MASK;
 	} while (!try_cmpxchg(&folio->flags, &old_flags, new_flags));
 }
-#else
-static void folio_inc_refs(struct folio *folio)
+
+static bool lru_gen_clear_refs(struct folio *folio)
 {
+	int gen = folio_lru_gen(folio);
+	int type = folio_is_file_lru(folio);
+	struct lru_gen_folio *lrugen = &folio_lruvec(folio)->lrugen;
+
+	set_mask_bits(&folio->flags, LRU_REFS_MASK | LRU_REFS_FLAGS, 0);
+
+	/* whether can do without shuffling under the LRU lock */
+	return gen == lru_gen_from_seq(READ_ONCE(lrugen->min_seq[type]));
 }
+
+#else /* !CONFIG_LRU_GEN */
+
+static void lru_gen_inc_refs(struct folio *folio)
+{
+}
+
+static bool lru_gen_clear_refs(struct folio *folio)
+{
+	return false;
+}
+
 #endif /* CONFIG_LRU_GEN */
 
 /**
@@ -428,7 +449,7 @@ static void folio_inc_refs(struct folio *folio)
 void folio_mark_accessed(struct folio *folio)
 {
 	if (lru_gen_enabled()) {
-		folio_inc_refs(folio);
+		lru_gen_inc_refs(folio);
 		return;
 	}
 
@@ -524,7 +545,7 @@ void folio_add_lru_vma(struct folio *folio, struct vm_area_struct *vma)
  */
 static void lru_deactivate_file(struct lruvec *lruvec, struct folio *folio)
 {
-	bool active = folio_test_active(folio);
+	bool active = folio_test_active(folio) || lru_gen_enabled();
 	long nr_pages = folio_nr_pages(folio);
 
 	if (folio_test_unevictable(folio))
@@ -589,7 +610,10 @@ static void lru_lazyfree(struct lruvec *lruvec, struct folio *folio)
 
 	lruvec_del_folio(lruvec, folio);
 	folio_clear_active(folio);
-	folio_clear_referenced(folio);
+	if (lru_gen_enabled())
+		lru_gen_clear_refs(folio);
+	else
+		folio_clear_referenced(folio);
 	/*
 	 * Lazyfree folios are clean anonymous folios.  They have
 	 * the swapbacked flag cleared, to distinguish them from normal
@@ -657,6 +681,9 @@ void deactivate_file_folio(struct folio *folio)
 	if (folio_test_unevictable(folio))
 		return;
 
+	if (lru_gen_enabled() && lru_gen_clear_refs(folio))
+		return;
+
 	folio_batch_add_and_move(folio, lru_deactivate_file, true);
 }
 
@@ -670,7 +697,10 @@ void deactivate_file_folio(struct folio *folio)
  */
 void folio_deactivate(struct folio *folio)
 {
-	if (folio_test_unevictable(folio) || !(folio_test_active(folio) || lru_gen_enabled()))
+	if (folio_test_unevictable(folio))
+		return;
+
+	if (lru_gen_enabled() ? lru_gen_clear_refs(folio) : !folio_test_active(folio))
 		return;
 
 	folio_batch_add_and_move(folio, lru_deactivate, true);
Signature · 1 line
2.47.0.338.g60cca15819-goog
author
#3 lore raw

PATCH 3/6 mm/mglru: rework aging feedback

In reply to Yu Zhao

The aging feedback is based on both the number of generations and the distribution of folios in each generation. The number of generations is currently the distance between max_seq and anon min_seq. This is because anon min_seq is not allowed to move past file min_seq. The rationale for that is that file is always evictable whereas anon is not. However, for use cases where anon is a lot cheaper than file:
1. Anon in the second oldest generation can be a better choice than file in the oldest generation.
2. A large amount of file in the oldest generation can skew the distribution, making should_run_aging() return false negative.

Allow anon and file min_seq to move independently, and use solely the number of generations as the feedback for aging. Specifically, when both anon and file are evictable, anon min_seq can now be greater than file min_seq, and therefore the number of generations becomes the distance between max_seq and min(min_seq[0],min_seq[1]). And should_run_aging() returns true if and only if the number of generations is less than MAX_NR_GENS.

As the first step to the final optimization, this change by itself should not have userspace-visiable effects beyond performance. The next twos patch will take advantage of this change; the last patch in this series will better distribute folios across MAX_NR_GENS.

---
 include/linux/mmzone.h |   6 +-
 mm/vmscan.c            | 217 +++++++++++++++++------------------------
 2 files changed, 91 insertions(+), 132 deletions(-)
include/linux/mmzone.h +3 -3
index b36124145a16..b998ccc5c341 100644
@@ -446,8 +446,8 @@ struct lru_gen_folio {
 	unsigned long avg_refaulted[ANON_AND_FILE][MAX_NR_TIERS];
 	/* the exponential moving average of evicted+protected */
 	unsigned long avg_total[ANON_AND_FILE][MAX_NR_TIERS];
-	/* the first tier doesn't need protection, hence the minus one */
-	unsigned long protected[NR_HIST_GENS][ANON_AND_FILE][MAX_NR_TIERS - 1];
+	/* can only be modified under the LRU lock */
+	unsigned long protected[NR_HIST_GENS][ANON_AND_FILE][MAX_NR_TIERS];
 	/* can be modified without holding the LRU lock */
 	atomic_long_t evicted[NR_HIST_GENS][ANON_AND_FILE][MAX_NR_TIERS];
 	atomic_long_t refaulted[NR_HIST_GENS][ANON_AND_FILE][MAX_NR_TIERS];
@@ -498,7 +498,7 @@ struct lru_gen_mm_walk {
 	int mm_stats[NR_MM_STATS];
 	/* total batched items */
 	int batched;
-	bool can_swap;
+	int swappiness;
 	bool force_scan;
 };
 
mm/vmscan.c +88 -129

519 lines not on this page · raw message

Signature · 1 line
2.47.0.338.g60cca15819-goog
author
#4 lore raw

PATCH 4/6 mm/mglru: rework type selection

In reply to Yu Zhao

With anon and file min_seq being able to move independently, rework type selection so that it is based on the total refaults from all tiers of each type. Also allow a type to be selected until that type reaches MIN_NR_GENS, and therefore abs_diff(min_seq[0],min_seq[1]) now can be 2 (MAX_NR_GENS-MIN_NR_GENS) instead of 1.

Since some tiers of a selected type can have higher refaults than the first tier of the other type, use a less larger gain factor 2:3 instead of 1:2, in order for those tiers in the selected type to be better protected.

As an intermediate step to the final optimization, this change by itself should not have userspace-visiable effects beyond performance.

---
 mm/vmscan.c | 82 +++++++++++++++++------------------------------------
 1 file changed, 26 insertions(+), 56 deletions(-)
mm/vmscan.c +26 -56
index 00a5aff3db42..307ae8b03d0d 100644
@@ -3089,15 +3089,20 @@ struct ctrl_pos {
 static void read_ctrl_pos(struct lruvec *lruvec, int type, int tier, int gain,
 			  struct ctrl_pos *pos)
 {
+	int i;
 	struct lru_gen_folio *lrugen = &lruvec->lrugen;
 	int hist = lru_hist_from_seq(lrugen->min_seq[type]);
 
-	pos->refaulted = lrugen->avg_refaulted[type][tier] +
-			 atomic_long_read(&lrugen->refaulted[hist][type][tier]);
-	pos->total = lrugen->avg_total[type][tier] +
-		     lrugen->protected[hist][type][tier] +
-		     atomic_long_read(&lrugen->evicted[hist][type][tier]);
+	pos->refaulted = pos->total = 0;
 	pos->gain = gain;
+
+	for (i = tier % MAX_NR_TIERS; i <= min(tier, MAX_NR_TIERS - 1); i++) {
+		pos->refaulted += lrugen->avg_refaulted[type][i] +
+				  atomic_long_read(&lrugen->refaulted[hist][type][i]);
+		pos->total += lrugen->avg_total[type][i] +
+			      lrugen->protected[hist][type][i] +
+			      atomic_long_read(&lrugen->evicted[hist][type][i]);
+	}
 }
 
 static void reset_ctrl_pos(struct lruvec *lruvec, int type, bool carryover)
@@ -4480,13 +4485,13 @@ static int get_tier_idx(struct lruvec *lruvec, int type)
 	struct ctrl_pos sp, pv;
 
 	/*
-	 * To leave a margin for fluctuations, use a larger gain factor (1:2).
+	 * To leave a margin for fluctuations, use a larger gain factor (2:3).
 	 * This value is chosen because any other tier would have at least twice
 	 * as many refaults as the first tier.
 	 */
-	read_ctrl_pos(lruvec, type, 0, 1, &sp);
+	read_ctrl_pos(lruvec, type, 0, 2, &sp);
 	for (tier = 1; tier < MAX_NR_TIERS; tier++) {
-		read_ctrl_pos(lruvec, type, tier, 2, &pv);
+		read_ctrl_pos(lruvec, type, tier, 3, &pv);
 		if (!positive_ctrl_err(&sp, &pv))
 			break;
 	}
@@ -4494,68 +4499,34 @@ static int get_tier_idx(struct lruvec *lruvec, int type)
 	return tier - 1;
 }
 
-static int get_type_to_scan(struct lruvec *lruvec, int swappiness, int *tier_idx)
+static int get_type_to_scan(struct lruvec *lruvec, int swappiness)
 {
-	int type, tier;
 	struct ctrl_pos sp, pv;
-	int gain[ANON_AND_FILE] = { swappiness, MAX_SWAPPINESS - swappiness };
 
+	if (!swappiness)
+		return LRU_GEN_FILE;
+
+	if (swappiness == MAX_SWAPPINESS)
+		return LRU_GEN_ANON;
 	/*
-	 * Compare the first tier of anon with that of file to determine which
-	 * type to scan. Also need to compare other tiers of the selected type
-	 * with the first tier of the other type to determine the last tier (of
-	 * the selected type) to evict.
+	 * Compare the sum of all tiers of anon with that of file to determine
+	 * which type to scan.
 	 */
-	read_ctrl_pos(lruvec, LRU_GEN_ANON, 0, gain[LRU_GEN_ANON], &sp);
-	read_ctrl_pos(lruvec, LRU_GEN_FILE, 0, gain[LRU_GEN_FILE], &pv);
-	type = positive_ctrl_err(&sp, &pv);
+	read_ctrl_pos(lruvec, LRU_GEN_ANON, MAX_NR_TIERS, swappiness, &sp);
+	read_ctrl_pos(lruvec, LRU_GEN_FILE, MAX_NR_TIERS, MAX_SWAPPINESS - swappiness, &pv);
 
-	read_ctrl_pos(lruvec, !type, 0, gain[!type], &sp);
-	for (tier = 1; tier < MAX_NR_TIERS; tier++) {
-		read_ctrl_pos(lruvec, type, tier, gain[type], &pv);
-		if (!positive_ctrl_err(&sp, &pv))
-			break;
-	}
-
-	*tier_idx = tier - 1;
-
-	return type;
+	return positive_ctrl_err(&sp, &pv);
 }
 
 static int isolate_folios(struct lruvec *lruvec, struct scan_control *sc, int swappiness,
 			  int *type_scanned, struct list_head *list)
 {
 	int i;
-	int type;
-	int tier = -1;
-	DEFINE_MIN_SEQ(lruvec);
-
-	/*
-	 * Try to make the obvious choice first, and if anon and file are both
-	 * available from the same generation,
-	 * 1. Interpret swappiness 1 as file first and MAX_SWAPPINESS as anon
-	 *    first.
-	 * 2. If !__GFP_IO, file first since clean pagecache is more likely to
-	 *    exist than clean swapcache.
-	 */
-	if (!swappiness)
-		type = LRU_GEN_FILE;
-	else if (min_seq[LRU_GEN_ANON] < min_seq[LRU_GEN_FILE])
-		type = LRU_GEN_ANON;
-	else if (swappiness == 1)
-		type = LRU_GEN_FILE;
-	else if (swappiness == MAX_SWAPPINESS)
-		type = LRU_GEN_ANON;
-	else if (!(sc->gfp_mask & __GFP_IO))
-		type = LRU_GEN_FILE;
-	else
-		type = get_type_to_scan(lruvec, swappiness, &tier);
+	int type = get_type_to_scan(lruvec, swappiness);
 
 	for_each_evictable_type(i, swappiness) {
 		int scanned;
-
-		if (tier < 0)
-			tier = get_tier_idx(lruvec, type);
+		int tier = get_tier_idx(lruvec, type);
 
 		*type_scanned = type;
 
@@ -4564,7 +4535,6 @@ static int isolate_folios(struct lruvec *lruvec, struct scan_control *sc, int sw
 			return scanned;
 
 		type = !type;
-		tier = -1;
 	}
 
 	return 0;
Signature · 1 line
2.47.0.338.g60cca15819-goog
author
#5 lore raw

PATCH 5/6 mm/mglru: rework refault detection

In reply to Yu Zhao

With anon and file min_seq being able to move independently, rework workingset protection as well so that the comparison of refaults between anon and file is always on an equal footing.

Specifically, make lru_gen_test_recent() return true for refaults happening within the distance of MAX_NR_GENS. For example, if min_seq of a type is max_seq-MIN_NR_GENS, refaults from min_seq-1, i.e., max_seq-MIN_NR_GENS-1, are also considered recent, since the distance max_seq-(max_seq-MIN_NR_GENS-1), i.e., MIN_NR_GENS+1 is less than MAX_NR_GENS.

As an intermediate step to the final optimization, this change by itself should not have userspace-visiable effects beyond performance.

---
 mm/workingset.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)
mm/workingset.c +9 -8
index ad181d1b8cf1..2c310c29f51e 100644
@@ -260,11 +260,11 @@ static void *lru_gen_eviction(struct folio *folio)
  * Tests if the shadow entry is for a folio that was recently evicted.
  * Fills in @lruvec, @token, @workingset with the values unpacked from shadow.
  */
-static bool lru_gen_test_recent(void *shadow, bool file, struct lruvec **lruvec,
+static bool lru_gen_test_recent(void *shadow, struct lruvec **lruvec,
 				unsigned long *token, bool *workingset)
 {
 	int memcg_id;
-	unsigned long min_seq;
+	unsigned long max_seq;
 	struct mem_cgroup *memcg;
 	struct pglist_data *pgdat;
 
@@ -273,8 +273,10 @@ static bool lru_gen_test_recent(void *shadow, bool file, struct lruvec **lruvec,
 	memcg = mem_cgroup_from_id(memcg_id);
 	*lruvec = mem_cgroup_lruvec(memcg, pgdat);
 
-	min_seq = READ_ONCE((*lruvec)->lrugen.min_seq[file]);
-	return (*token >> LRU_REFS_WIDTH) == (min_seq & (EVICTION_MASK >> LRU_REFS_WIDTH));
+	max_seq = READ_ONCE((*lruvec)->lrugen.max_seq);
+	max_seq &= EVICTION_MASK >> LRU_REFS_WIDTH;
+
+	return abs_diff(max_seq, *token >> LRU_REFS_WIDTH) < MAX_NR_GENS;
 }
 
 static void lru_gen_refault(struct folio *folio, void *shadow)
@@ -290,7 +292,7 @@ static void lru_gen_refault(struct folio *folio, void *shadow)
 
 	rcu_read_lock();
 
-	recent = lru_gen_test_recent(shadow, type, &lruvec, &token, &workingset);
+	recent = lru_gen_test_recent(shadow, &lruvec, &token, &workingset);
 	if (lruvec != folio_lruvec(folio))
 		goto unlock;
 
@@ -331,7 +333,7 @@ static void *lru_gen_eviction(struct folio *folio)
 	return NULL;
 }
 
-static bool lru_gen_test_recent(void *shadow, bool file, struct lruvec **lruvec,
+static bool lru_gen_test_recent(void *shadow, struct lruvec **lruvec,
 				unsigned long *token, bool *workingset)
 {
 	return false;
@@ -432,8 +434,7 @@ bool workingset_test_recent(void *shadow, bool file, bool *workingset,
 		bool recent;
 
 		rcu_read_lock();
-		recent = lru_gen_test_recent(shadow, file, &eviction_lruvec,
-					     &eviction, workingset);
+		recent = lru_gen_test_recent(shadow, &eviction_lruvec, &eviction, workingset);
 		rcu_read_unlock();
 		return recent;
 	}
Signature · 1 line
2.47.0.338.g60cca15819-goog
author
#6 lore raw

PATCH 6/6 mm/mglru: rework workingset protection

In reply to Yu Zhao

With the aging feedback no longer considering the distribution of folios in each generation, rework workingset protection to better distribute folios across MAX_NR_GENS. This is achieved by reusing PG_workingset and PG_referenced/LRU_REFS_FLAGS in a slightly different way.

For folios accessed multiple times through file descriptors, make lru_gen_inc_refs() set additional bits of LRU_REFS_WIDTH in folio->flags after PG_referenced, then PG_workingset after LRU_REFS_WIDTH. After all its bits are set, i.e., LRU_REFS_FLAGS|BIT(PG_workingset), a folio is lazily promoted into the second oldest generation (min_seq+1) in the eviction path. And when folio_inc_gen() does that, it clears LRU_REFS_FLAGS so that lru_gen_inc_refs() can start over. For this case, LRU_REFS_MASK is only valid when PG_referenced is set.

For folios accessed multiple times through page tables, folio_update_gen() from a page table walk or lru_gen_set_refs() from a rmap walk sets PG_referenced after the accessed bit is cleared for the first time. Thereafter, those two paths set PG_workingset and promote folios to the youngest generation. Like folio_inc_gen(), folio_update_gen() also clears PG_referenced. For this case, LRU_REFS_MASK is not used.

For both cases above, after PG_workingset is set on a folio, it remains until this folio is either reclaimed, or "deactivated" by lru_gen_clear_refs(). It can be set again if lru_gen_test_recent() returns true upon a refault.

When adding folios to the LRU lists, lru_gen_distance() distributes them as follows:

+---------------------------------+---------------------------------+
|    Accessed thru page tables    | Accessed thru file descriptors  |
+---------------------------------+---------------------------------+
| PG_active (set while isolated)  |                                 |
+----------------+----------------+----------------+----------------+
| PG_workingset  | PG_referenced  | PG_workingset  | LRU_REFS_FLAGS |
+---------------------------------+---------------------------------+
|<--------- MIN_NR_GENS --------->|                                 |
|<-------------------------- MAX_NR_GENS -------------------------->|

After this patch, some typical client and server workloads showed improvements under heavy memory pressure. For example, Python TPC-C, which was used to benchmark a different approach [1] to better detect refault distances, showed a significant decrease in total refaults:

                          Before      After      Change
Time (seconds)            10801       10801      0%
Executed (transactions)   41472       43663      +5%
workingset_nodes          109070      120244     +10%
workingset_refault_anon   5019627     7281831    +45%
workingset_refault_file   1294678786  554855564  -57%
workingset_refault_total  1299698413  562137395  -57%

[1] https://lore.kernel.org/[email protected]/

---
 include/linux/mm_inline.h |  94 +++++++++++++------------
 include/linux/mmzone.h    |  82 +++++++++++++---------
 mm/swap.c                 |  23 +++---
 mm/vmscan.c               | 142 +++++++++++++++++++++++---------------
 mm/workingset.c           |  31 ++++-----
 5 files changed, 211 insertions(+), 161 deletions(-)
include/linux/mm_inline.h +50 -44
index 1b6a917fffa4..adc879c3e19f 100644
@@ -133,31 +133,25 @@ static inline int lru_hist_from_seq(unsigned long seq)
 	return seq % NR_HIST_GENS;
 }
 
-static inline int lru_tier_from_refs(int refs)
+static inline int lru_tier_from_refs(int refs, bool workingset)
 {
 	VM_WARN_ON_ONCE(refs > BIT(LRU_REFS_WIDTH));
 
-	/* see the comment in folio_lru_refs() */
-	return order_base_2(refs + 1);
+	/* see the comment on MAX_NR_TIERS */
+	return workingset ? MAX_NR_TIERS - 1 : order_base_2(refs);
 }
 
 static inline int folio_lru_refs(struct folio *folio)
 {
 	unsigned long flags = READ_ONCE(folio->flags);
-	bool workingset = flags & BIT(PG_workingset);
 
+	if (!(flags & BIT(PG_referenced)))
+		return 0;
 	/*
-	 * Return the number of accesses beyond PG_referenced, i.e., N-1 if the
-	 * total number of accesses is N>1, since N=0,1 both map to the first
-	 * tier. lru_tier_from_refs() will account for this off-by-one. Also see
-	 * the comment on MAX_NR_TIERS.
+	 * Return the total number of accesses including PG_referenced. Also see
+	 * the comment on LRU_REFS_FLAGS.
 	 */
-	return ((flags & LRU_REFS_MASK) >> LRU_REFS_PGOFF) + workingset;
-}
-
-static inline void folio_clear_lru_refs(struct folio *folio)
-{
-	set_mask_bits(&folio->flags, LRU_REFS_MASK | LRU_REFS_FLAGS, 0);
+	return ((flags & LRU_REFS_MASK) >> LRU_REFS_PGOFF) + 1;
 }
 
 static inline int folio_lru_gen(struct folio *folio)
@@ -223,11 +217,46 @@ static inline void lru_gen_update_size(struct lruvec *lruvec, struct folio *foli
 	VM_WARN_ON_ONCE(lru_gen_is_active(lruvec, old_gen) && !lru_gen_is_active(lruvec, new_gen));
 }
 
+static inline int lru_gen_distance(struct folio *folio, bool reclaiming)
+{
+	/*
+	 * Distance until eviction (larger values provide stronger protection):
+	 * +-------------------------------------+-------------------------------------+
+	 * | Accessed through page tables and    | Accessed through file descriptors   |
+	 * | promoted by folio_update_gen()      | and protected by folio_inc_gen()    |
+	 * +-------------------------------------+-------------------------------------+
+	 * | PG_active (only set while isolated) |                                     |
+	 * +------------------+------------------+------------------+------------------+
+	 * |  PG_workingset   |  PG_referenced   |  PG_workingset   |  LRU_REFS_FLAGS  |
+	 * +-------------------------------------+-------------------------------------+
+	 * |        3         |        2         |        1         |        0         |
+	 * +-------------------------------------+-------------------------------------+
+	 * |<----------- MIN_NR_GENS ----------->|                                     |
+	 * |<------------------------------ MAX_NR_GENS ------------------------------>|
+	 */
+	if (reclaiming)
+		return 0;
+
+	if (folio_test_active(folio))
+		return MIN_NR_GENS + folio_test_workingset(folio);
+
+	if (folio_test_workingset(folio))
+		return MIN_NR_GENS - 1;
+
+	if (!folio_is_file_lru(folio) && !folio_test_swapcache(folio))
+		return MIN_NR_GENS - 1;
+
+	if (folio_test_reclaim(folio) && (folio_test_dirty(folio) || folio_test_writeback(folio)))
+		return MIN_NR_GENS - 1;
+
+	return 0;
+}
+
 static inline bool lru_gen_add_folio(struct lruvec *lruvec, struct folio *folio, bool reclaiming)
 {
+	int dist;
 	unsigned long seq;
 	unsigned long flags;
-	unsigned long mask;
 	int gen = folio_lru_gen(folio);
 	int type = folio_is_file_lru(folio);
 	int zone = folio_zonenum(folio);
@@ -237,40 +266,17 @@ static inline bool lru_gen_add_folio(struct lruvec *lruvec, struct folio *folio,
 
 	if (folio_test_unevictable(folio) || !lrugen->enabled)
 		return false;
-	/*
-	 * There are four common cases for this page:
-	 * 1. If it's hot, i.e., freshly faulted in, add it to the youngest
-	 *    generation, and it's protected over the rest below.
-	 * 2. If it can't be evicted immediately, i.e., a dirty page pending
-	 *    writeback, add it to the second youngest generation.
-	 * 3. If it should be evicted first, e.g., cold and clean from
-	 *    folio_rotate_reclaimable(), add it to the oldest generation.
-	 * 4. Everything else falls between 2 & 3 above and is added to the
-	 *    second oldest generation if it's considered inactive, or the
-	 *    oldest generation otherwise. See lru_gen_is_active().
-	 */
-	if (folio_test_active(folio))
-		seq = lrugen->max_seq;
-	else if ((type == LRU_GEN_ANON && !folio_test_swapcache(folio)) ||
-		 (folio_test_reclaim(folio) &&
-		  (folio_test_dirty(folio) || folio_test_writeback(folio))))
-		seq = lrugen->max_seq - 1;
-	else if (reclaiming || lrugen->min_seq[type] + MIN_NR_GENS >= lrugen->max_seq)
-		seq = lrugen->min_seq[type];
+
+	dist = lru_gen_distance(folio, reclaiming);
+	if (dist < MIN_NR_GENS)
+		seq = lrugen->min_seq[type] + dist;
 	else
-		seq = lrugen->min_seq[type] + 1;
+		seq = lrugen->max_seq + dist - MIN_NR_GENS - 1;
 
 	gen = lru_gen_from_seq(seq);
 	flags = (gen + 1UL) << LRU_GEN_PGOFF;
 	/* see the comment on MIN_NR_GENS about PG_active */
-	mask = LRU_GEN_MASK;
-	/*
-	 * Don't clear PG_workingset here because it can affect PSI accounting
-	 * if the activation is due to workingset refault.
-	 */
-	if (folio_test_active(folio))
-		mask |= LRU_REFS_MASK | BIT(PG_referenced) | BIT(PG_active);
-	set_mask_bits(&folio->flags, mask, flags);
+	set_mask_bits(&folio->flags, LRU_GEN_MASK | BIT(PG_active), flags);
 
 	lru_gen_update_size(lruvec, folio, -1, gen);
 	/* for folio_rotate_reclaimable() */
include/linux/mmzone.h +51 -31

125 lines not on this page · raw message

mm/swap.c +9 -14
index 15a94be8b0af..320b959b74c6 100644
@@ -387,24 +387,19 @@ static void lru_gen_inc_refs(struct folio *folio)
 	if (folio_test_unevictable(folio))
 		return;
 
+	/* see the comment on LRU_REFS_FLAGS */
 	if (!folio_test_referenced(folio)) {
-		folio_set_referenced(folio);
+		set_mask_bits(&folio->flags, LRU_REFS_MASK, BIT(PG_referenced));
 		return;
 	}
 
-	if (!folio_test_workingset(folio)) {
-		folio_set_workingset(folio);
-		return;
-	}
-
-	/* see the comment on MAX_NR_TIERS */
 	do {
-		new_flags = old_flags & LRU_REFS_MASK;
-		if (new_flags == LRU_REFS_MASK)
-			break;
+		if ((old_flags & LRU_REFS_MASK) == LRU_REFS_MASK) {
+			folio_set_workingset(folio);
+			return;
+		}
 
-		new_flags += BIT(LRU_REFS_PGOFF);
-		new_flags |= old_flags & ~LRU_REFS_MASK;
+		new_flags = old_flags + BIT(LRU_REFS_PGOFF);
 	} while (!try_cmpxchg(&folio->flags, &old_flags, new_flags));
 }
 
@@ -414,7 +409,7 @@ static bool lru_gen_clear_refs(struct folio *folio)
 	int type = folio_is_file_lru(folio);
 	struct lru_gen_folio *lrugen = &folio_lruvec(folio)->lrugen;
 
-	set_mask_bits(&folio->flags, LRU_REFS_MASK | LRU_REFS_FLAGS, 0);
+	set_mask_bits(&folio->flags, LRU_REFS_FLAGS | BIT(PG_workingset), 0);
 
 	/* whether can do without shuffling under the LRU lock */
 	return gen == lru_gen_from_seq(READ_ONCE(lrugen->min_seq[type]));
@@ -495,7 +490,7 @@ void folio_add_lru(struct folio *folio)
 			folio_test_unevictable(folio), folio);
 	VM_BUG_ON_FOLIO(folio_test_lru(folio), folio);
 
-	/* see the comment in lru_gen_add_folio() */
+	/* see the comment in lru_gen_distance() */
 	if (lru_gen_enabled() && !folio_test_unevictable(folio) &&
 	    lru_gen_in_fault() && !(current->flags & PF_MEMALLOC))
 		folio_set_active(folio);
mm/vmscan.c +86 -56

243 lines not on this page · raw message

mm/workingset.c +15 -16
index 2c310c29f51e..2fe142aebd07 100644
@@ -239,7 +239,8 @@ static void *lru_gen_eviction(struct folio *folio)
 	int type = folio_is_file_lru(folio);
 	int delta = folio_nr_pages(folio);
 	int refs = folio_lru_refs(folio);
-	int tier = lru_tier_from_refs(refs);
+	bool workingset = folio_test_workingset(folio);
+	int tier = lru_tier_from_refs(refs, workingset);
 	struct mem_cgroup *memcg = folio_memcg(folio);
 	struct pglist_data *pgdat = folio_pgdat(folio);
 
@@ -253,7 +254,7 @@ static void *lru_gen_eviction(struct folio *folio)
 	hist = lru_hist_from_seq(min_seq);
 	atomic_long_add(delta, &lrugen->evicted[hist][type][tier]);
 
-	return pack_shadow(mem_cgroup_id(memcg), pgdat, token, refs);
+	return pack_shadow(mem_cgroup_id(memcg), pgdat, token, workingset);
 }
 
 /*
@@ -282,8 +283,9 @@ static bool lru_gen_test_recent(void *shadow, struct lruvec **lruvec,
 static void lru_gen_refault(struct folio *folio, void *shadow)
 {
 	bool recent;
-	int hist, tier, refs;
+	int hist, tier;
 	bool workingset;
+	unsigned long refs;
 	unsigned long token;
 	struct lruvec *lruvec;
 	struct lru_gen_folio *lrugen;
@@ -305,23 +307,20 @@ static void lru_gen_refault(struct folio *folio, void *shadow)
 
 	hist = lru_hist_from_seq(READ_ONCE(lrugen->min_seq[type]));
 	/* see the comment in folio_lru_refs() */
-	refs = (token & (BIT(LRU_REFS_WIDTH) - 1)) + workingset;
-	tier = lru_tier_from_refs(refs);
+	refs = (token & (BIT(LRU_REFS_WIDTH) - 1)) + 1;
+	tier = lru_tier_from_refs(refs, workingset);
 
 	atomic_long_add(delta, &lrugen->refaulted[hist][type][tier]);
-	mod_lruvec_state(lruvec, WORKINGSET_ACTIVATE_BASE + type, delta);
 
-	/*
-	 * Count the following two cases as stalls:
-	 * 1. For pages accessed through page tables, hotter pages pushed out
-	 *    hot pages which refaulted immediately.
-	 * 2. For pages accessed multiple times through file descriptors,
-	 *    they would have been protected by sort_folio().
-	 */
-	if (lru_gen_in_fault() || refs >= BIT(LRU_REFS_WIDTH) - 1) {
-		set_mask_bits(&folio->flags, 0, LRU_REFS_MASK | BIT(PG_workingset));
+	/* see folio_add_lru() where folio_set_active() happens */
+	if (lru_gen_in_fault())
+		mod_lruvec_state(lruvec, WORKINGSET_ACTIVATE_BASE + type, delta);
+
+	if (workingset) {
+		folio_set_workingset(folio);
 		mod_lruvec_state(lruvec, WORKINGSET_RESTORE_BASE + type, delta);
-	}
+	} else if (refs)
+		set_mask_bits(&folio->flags, LRU_REFS_MASK, refs << LRU_REFS_PGOFF);
 unlock:
 	rcu_read_unlock();
 }
Signature · 1 line
2.47.0.338.g60cca15819-goog