[PATCH 00/10] shrinker debugging, .to_text() report

[PATCH 00/10] shrinker debugging, .to_text() report

[Kent Overstreet]

recently new OOMs have been cropping up, and reclaim is implicated, so
I've had to dust off these patches.

nothing significant has changed since the last time I posted, and they
have been valuable - Dave, think we can get them in?

Kent Overstreet (10):
  seq_buf: seq_buf_human_readable_u64()
  mm: shrinker: Add a .to_text() method for shrinkers
  mm: shrinker: Add new stats for .to_text()
  mm: Centralize & improve oom reporting in show_mem.c
  mm: shrinker: Add shrinker_to_text() to debugfs interface
  bcachefs: shrinker.to_text() methods
  percpu: per_cpu_sum()
  fs: Add super_block->s_inodes_nr
  fs/dcache: Add per-sb accounting for nr dentries
  fs: super_cache_to_text()

 fs/bcachefs/btree_cache.c     | 13 +++++
 fs/bcachefs/btree_key_cache.c | 14 ++++++
 fs/bcachefs/util.h            | 10 ----
 fs/dcache.c                   | 18 ++-----
 fs/inode.c                    |  2 +
 fs/super.c                    | 25 ++++++++++
 include/linux/fs.h            |  2 +
 include/linux/percpu.h        | 10 ++++
 include/linux/seq_buf.h       |  4 ++
 include/linux/shrinker.h      | 13 ++++-
 lib/seq_buf.c                 | 10 ++++
 mm/oom_kill.c                 | 23 ---------
 mm/show_mem.c                 | 43 ++++++++++++++++
 mm/shrinker.c                 | 94 ++++++++++++++++++++++++++++++++++-
 mm/shrinker_debug.c           | 18 +++++++
 mm/slab.h                     |  6 ++-
 mm/slab_common.c              | 52 +++++++++++++++----
 17 files changed, 298 insertions(+), 59 deletions(-)

-- 
2.45.2

[PATCH 01/10] seq_buf: seq_buf_human_readable_u64()

[Kent Overstreet]

This adds a seq_buf wrapper for string_get_size().

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 include/linux/seq_buf.h |  4 ++++
 lib/seq_buf.c           | 10 ++++++++++
 2 files changed, 14 insertions(+)

diff --git a/include/linux/seq_buf.h b/include/linux/seq_buf.h
index fe41da005970..1cba369e1821 100644
--- a/include/linux/seq_buf.h
+++ b/include/linux/seq_buf.h
@@ -173,4 +173,8 @@ seq_buf_bprintf(struct seq_buf *s, const char *fmt, const u32 *binary);
 
 void seq_buf_do_printk(struct seq_buf *s, const char *lvl);
 
+enum string_size_units;
+void seq_buf_human_readable_u64(struct seq_buf *s, u64 v,
+				const enum string_size_units units);
+
 #endif /* _LINUX_SEQ_BUF_H */
diff --git a/lib/seq_buf.c b/lib/seq_buf.c
index f3f3436d60a9..3c41ca83a0c3 100644
--- a/lib/seq_buf.c
+++ b/lib/seq_buf.c
@@ -436,3 +436,13 @@ int seq_buf_hex_dump(struct seq_buf *s, const char *prefix_str, int prefix_type,
 	}
 	return 0;
 }
+
+void seq_buf_human_readable_u64(struct seq_buf *s, u64 v, const enum string_size_units units)
+{
+	char *buf;
+	size_t size = seq_buf_get_buf(s, &buf);
+	int wrote = string_get_size(v, 1, units, buf, size);
+
+	seq_buf_commit(s, wrote);
+}
+EXPORT_SYMBOL(seq_buf_human_readable_u64);
-- 
2.45.2

[PATCH 02/10] mm: shrinker: Add a .to_text() method for shrinkers

[Kent Overstreet]

This adds a new callback method to shrinkers which they can use to
describe anything relevant to memory reclaim about their internal state,
for example object dirtyness.

This patch also adds shrinkers_to_text(), which reports on the top 10
shrinkers - by object count - in sorted order, to be used in OOM
reporting.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: linux-mm@kvack.org
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 include/linux/shrinker.h |  7 +++-
 mm/shrinker.c            | 73 +++++++++++++++++++++++++++++++++++++++-
 2 files changed, 78 insertions(+), 2 deletions(-)

diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h
index 1a00be90d93a..6193612617a1 100644
--- a/include/linux/shrinker.h
+++ b/include/linux/shrinker.h
@@ -24,6 +24,8 @@ struct shrinker_info {
 	struct shrinker_info_unit *unit[];
 };
 
+struct seq_buf;
+
 /*
  * This struct is used to pass information from page reclaim to the shrinkers.
  * We consolidate the values for easier extension later.
@@ -80,10 +82,12 @@ struct shrink_control {
  * @flags determine the shrinker abilities, like numa awareness
  */
 struct shrinker {
+	const char *name;
 	unsigned long (*count_objects)(struct shrinker *,
 				       struct shrink_control *sc);
 	unsigned long (*scan_objects)(struct shrinker *,
 				      struct shrink_control *sc);
+	void (*to_text)(struct seq_buf *, struct shrinker *);
 
 	long batch;	/* reclaim batch size, 0 = default */
 	int seeks;	/* seeks to recreate an obj */
@@ -110,7 +114,6 @@ struct shrinker {
 #endif
 #ifdef CONFIG_SHRINKER_DEBUG
 	int debugfs_id;
-	const char *name;
 	struct dentry *debugfs_entry;
 #endif
 	/* objs pending delete, per node */
@@ -135,6 +138,8 @@ __printf(2, 3)
 struct shrinker *shrinker_alloc(unsigned int flags, const char *fmt, ...);
 void shrinker_register(struct shrinker *shrinker);
 void shrinker_free(struct shrinker *shrinker);
+void shrinker_to_text(struct seq_buf *, struct shrinker *);
+void shrinkers_to_text(struct seq_buf *);
 
 static inline bool shrinker_try_get(struct shrinker *shrinker)
 {
diff --git a/mm/shrinker.c b/mm/shrinker.c
index dc5d2a6fcfc4..ad52c269bb48 100644
--- a/mm/shrinker.c
+++ b/mm/shrinker.c
@@ -1,8 +1,9 @@
 // SPDX-License-Identifier: GPL-2.0
 #include <linux/memcontrol.h>
+#include <linux/rculist.h>
 #include <linux/rwsem.h>
+#include <linux/seq_buf.h>
 #include <linux/shrinker.h>
-#include <linux/rculist.h>
 #include <trace/events/vmscan.h>
 
 #include "internal.h"
@@ -807,3 +808,73 @@ void shrinker_free(struct shrinker *shrinker)
 	call_rcu(&shrinker->rcu, shrinker_free_rcu_cb);
 }
 EXPORT_SYMBOL_GPL(shrinker_free);
+
+void shrinker_to_text(struct seq_buf *out, struct shrinker *shrinker)
+{
+	struct shrink_control sc = { .gfp_mask = GFP_KERNEL, };
+
+	seq_buf_puts(out, shrinker->name);
+	seq_buf_printf(out, " objects: %lu\n", shrinker->count_objects(shrinker, &sc));
+
+	if (shrinker->to_text) {
+		shrinker->to_text(out, shrinker);
+		seq_buf_puts(out, "\n");
+	}
+}
+
+/**
+ * shrinkers_to_text - Report on shrinkers with highest usage
+ *
+ * This reports on the top 10 shrinkers, by object counts, in sorted order:
+ * intended to be used for OOM reporting.
+ */
+void shrinkers_to_text(struct seq_buf *out)
+{
+	struct shrinker *shrinker;
+	struct shrinker_by_mem {
+		struct shrinker	*shrinker;
+		unsigned long	mem;
+	} shrinkers_by_mem[10];
+	int i, nr = 0;
+
+	if (!mutex_trylock(&shrinker_mutex)) {
+		seq_buf_puts(out, "(couldn't take shrinker lock)");
+		return;
+	}
+
+	list_for_each_entry(shrinker, &shrinker_list, list) {
+		struct shrink_control sc = { .gfp_mask = GFP_KERNEL, };
+		unsigned long mem = shrinker->count_objects(shrinker, &sc);
+
+		if (!mem || mem == SHRINK_STOP || mem == SHRINK_EMPTY)
+			continue;
+
+		for (i = 0; i < nr; i++)
+			if (mem < shrinkers_by_mem[i].mem)
+				break;
+
+		if (nr < ARRAY_SIZE(shrinkers_by_mem)) {
+			memmove(&shrinkers_by_mem[i + 1],
+				&shrinkers_by_mem[i],
+				sizeof(shrinkers_by_mem[0]) * (nr - i));
+			nr++;
+		} else if (i) {
+			i--;
+			memmove(&shrinkers_by_mem[0],
+				&shrinkers_by_mem[1],
+				sizeof(shrinkers_by_mem[0]) * i);
+		} else {
+			continue;
+		}
+
+		shrinkers_by_mem[i] = (struct shrinker_by_mem) {
+			.shrinker = shrinker,
+			.mem = mem,
+		};
+	}
+
+	for (i = nr - 1; i >= 0; --i)
+		shrinker_to_text(out, shrinkers_by_mem[i].shrinker);
+
+	mutex_unlock(&shrinker_mutex);
+}
-- 
2.45.2

[PATCH 03/10] mm: shrinker: Add new stats for .to_text()

[Kent Overstreet]

Add a few new shrinker stats.

number of objects requested to free, number of objects freed:

Shrinkers won't necessarily free all objects requested for a variety of
reasons, but if the two counts are wildly different something is likely
amiss.

.scan_objects runtime:

If one shrinker is taking an excessive amount of time to free
objects that will block kswapd from running other shrinkers.

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: linux-mm@kvack.org
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 include/linux/shrinker.h |  6 ++++++
 mm/shrinker.c            | 23 ++++++++++++++++++++++-
 2 files changed, 28 insertions(+), 1 deletion(-)

diff --git a/include/linux/shrinker.h b/include/linux/shrinker.h
index 6193612617a1..106622ddac77 100644
--- a/include/linux/shrinker.h
+++ b/include/linux/shrinker.h
@@ -118,6 +118,12 @@ struct shrinker {
 #endif
 	/* objs pending delete, per node */
 	atomic_long_t *nr_deferred;
+
+	atomic_long_t	objects_requested_to_free;
+	atomic_long_t	objects_freed;
+	unsigned long	last_freed;	/* timestamp, in jiffies */
+	unsigned long	last_scanned;	/* timestamp, in jiffies */
+	atomic64_t	ns_run;
 };
 #define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
 
diff --git a/mm/shrinker.c b/mm/shrinker.c
index ad52c269bb48..feaa8122afc9 100644
--- a/mm/shrinker.c
+++ b/mm/shrinker.c
@@ -430,13 +430,24 @@ static unsigned long do_shrink_slab(struct shrink_control *shrinkctl,
 	       total_scan >= freeable) {
 		unsigned long ret;
 		unsigned long nr_to_scan = min(batch_size, total_scan);
+		u64 start_time = ktime_get_ns();
+
+		atomic_long_add(nr_to_scan, &shrinker->objects_requested_to_free);
 
 		shrinkctl->nr_to_scan = nr_to_scan;
 		shrinkctl->nr_scanned = nr_to_scan;
 		ret = shrinker->scan_objects(shrinker, shrinkctl);
+
+		atomic64_add(ktime_get_ns() - start_time, &shrinker->ns_run);
 		if (ret == SHRINK_STOP)
 			break;
 		freed += ret;
+		unsigned long now = jiffies;
+		if (ret) {
+			atomic_long_add(ret, &shrinker->objects_freed);
+			shrinker->last_freed = now;
+		}
+		shrinker->last_scanned = now;
 
 		count_vm_events(SLABS_SCANNED, shrinkctl->nr_scanned);
 		total_scan -= shrinkctl->nr_scanned;
@@ -812,9 +823,19 @@ EXPORT_SYMBOL_GPL(shrinker_free);
 void shrinker_to_text(struct seq_buf *out, struct shrinker *shrinker)
 {
 	struct shrink_control sc = { .gfp_mask = GFP_KERNEL, };
+	unsigned long nr_freed = atomic_long_read(&shrinker->objects_freed);
 
 	seq_buf_puts(out, shrinker->name);
-	seq_buf_printf(out, " objects: %lu\n", shrinker->count_objects(shrinker, &sc));
+	seq_buf_putc(out, '\n');
+
+	seq_buf_printf(out, "objects:             %lu\n", shrinker->count_objects(shrinker, &sc));
+	seq_buf_printf(out, "requested to free:   %lu\n", atomic_long_read(&shrinker->objects_requested_to_free));
+	seq_buf_printf(out, "objects freed:       %lu\n", nr_freed);
+	seq_buf_printf(out, "last scanned:        %li sec ago\n", (jiffies - shrinker->last_scanned) / HZ);
+	seq_buf_printf(out, "last freed:          %li sec ago\n", (jiffies - shrinker->last_freed) / HZ);
+	seq_buf_printf(out, "ns per object freed: %llu\n", nr_freed
+		       ? div64_ul(atomic64_read(&shrinker->ns_run), nr_freed)
+		       : 0);
 
 	if (shrinker->to_text) {
 		shrinker->to_text(out, shrinker);
-- 
2.45.2

[PATCH 04/10] mm: Centralize & improve oom reporting in show_mem.c

[Kent Overstreet]

This patch:
 - Changes show_mem() to always report on slab usage
 - Instead of reporting on all slabs, we only report on top 10 slabs,
   and in sorted order
 - Also reports on shrinkers, with the new shrinkers_to_text().
   Shrinkers need to be included in OOM/allocation failure reporting
   because they're responsible for memory reclaim - if a shrinker isn't
   giving up its memory, we need to know which one and why.

More OOM reporting can be moved to show_mem.c and improved, this patch
is only a start.

New example output on OOM/memory allocation failure:

00177 Mem-Info:
00177 active_anon:13706 inactive_anon:32266 isolated_anon:16
00177  active_file:1653 inactive_file:1822 isolated_file:0
00177  unevictable:0 dirty:0 writeback:0
00177  slab_reclaimable:6242 slab_unreclaimable:11168
00177  mapped:3824 shmem:3 pagetables:1266 bounce:0
00177  kernel_misc_reclaimable:0
00177  free:4362 free_pcp:35 free_cma:0
00177 Node 0 active_anon:54824kB inactive_anon:129064kB active_file:6612kB inactive_file:7288kB unevictable:0kB isolated(anon):64kB isolated(file):0kB mapped:15296kB dirty:0kB writeback:0kB shmem:12kB writeback_tmp:0kB kernel_stack:3392kB pagetables:5064kB all_unreclaimable? no
00177 DMA free:2232kB boost:0kB min:88kB low:108kB high:128kB reserved_highatomic:0KB active_anon:2924kB inactive_anon:6596kB active_file:428kB inactive_file:384kB unevictable:0kB writepending:0kB present:15992kB managed:15360kB mlocked:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
00177 lowmem_reserve[]: 0 426 426 426
00177 DMA32 free:15092kB boost:5836kB min:8432kB low:9080kB high:9728kB reserved_highatomic:0KB active_anon:52196kB inactive_anon:122392kB active_file:6176kB inactive_file:7068kB unevictable:0kB writepending:0kB present:507760kB managed:441816kB mlocked:0kB bounce:0kB free_pcp:72kB local_pcp:0kB free_cma:0kB
00177 lowmem_reserve[]: 0 0 0 0
00177 DMA: 284*4kB (UM) 53*8kB (UM) 21*16kB (U) 11*32kB (U) 0*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 2248kB
00177 DMA32: 2765*4kB (UME) 375*8kB (UME) 57*16kB (UM) 5*32kB (U) 0*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 15132kB
00177 4656 total pagecache pages
00177 1031 pages in swap cache
00177 Swap cache stats: add 6572399, delete 6572173, find 488603/3286476
00177 Free swap  = 509112kB
00177 Total swap = 2097148kB
00177 130938 pages RAM
00177 0 pages HighMem/MovableOnly
00177 16644 pages reserved
00177 Unreclaimable slab info:
00177 9p-fcall-cache    total: 8.25 MiB active: 8.25 MiB
00177 kernfs_node_cache total: 2.15 MiB active: 2.15 MiB
00177 kmalloc-64        total: 2.08 MiB active: 2.07 MiB
00177 task_struct       total: 1.95 MiB active: 1.95 MiB
00177 kmalloc-4k        total: 1.50 MiB active: 1.50 MiB
00177 signal_cache      total: 1.34 MiB active: 1.34 MiB
00177 kmalloc-2k        total: 1.16 MiB active: 1.16 MiB
00177 bch_inode_info    total: 1.02 MiB active: 922 KiB
00177 perf_event        total: 1.02 MiB active: 1.02 MiB
00177 biovec-max        total: 992 KiB active: 960 KiB
00177 Shrinkers:
00177 super_cache_scan: objects: 127
00177 super_cache_scan: objects: 106
00177 jbd2_journal_shrink_scan: objects: 32
00177 ext4_es_scan: objects: 32
00177 bch2_btree_cache_scan: objects: 8
00177   nr nodes:          24
00177   nr dirty:          0
00177   cannibalize lock:  0000000000000000
00177
00177 super_cache_scan: objects: 8
00177 super_cache_scan: objects: 1

Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: linux-mm@kvack.org
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 mm/oom_kill.c    | 23 ---------------------
 mm/show_mem.c    | 43 +++++++++++++++++++++++++++++++++++++++
 mm/slab.h        |  6 ++++--
 mm/slab_common.c | 52 +++++++++++++++++++++++++++++++++++++++---------
 4 files changed, 90 insertions(+), 34 deletions(-)

diff --git a/mm/oom_kill.c b/mm/oom_kill.c
index 4d7a0004df2c..dc56239ff057 100644
--- a/mm/oom_kill.c
+++ b/mm/oom_kill.c
@@ -169,27 +169,6 @@ static bool oom_unkillable_task(struct task_struct *p)
 	return false;
 }
 
-/*
- * Check whether unreclaimable slab amount is greater than
- * all user memory(LRU pages).
- * dump_unreclaimable_slab() could help in the case that
- * oom due to too much unreclaimable slab used by kernel.
-*/
-static bool should_dump_unreclaim_slab(void)
-{
-	unsigned long nr_lru;
-
-	nr_lru = global_node_page_state(NR_ACTIVE_ANON) +
-		 global_node_page_state(NR_INACTIVE_ANON) +
-		 global_node_page_state(NR_ACTIVE_FILE) +
-		 global_node_page_state(NR_INACTIVE_FILE) +
-		 global_node_page_state(NR_ISOLATED_ANON) +
-		 global_node_page_state(NR_ISOLATED_FILE) +
-		 global_node_page_state(NR_UNEVICTABLE);
-
-	return (global_node_page_state_pages(NR_SLAB_UNRECLAIMABLE_B) > nr_lru);
-}
-
 /**
  * oom_badness - heuristic function to determine which candidate task to kill
  * @p: task struct of which task we should calculate
@@ -464,8 +443,6 @@ static void dump_header(struct oom_control *oc)
 		mem_cgroup_print_oom_meminfo(oc->memcg);
 	else {
 		__show_mem(SHOW_MEM_FILTER_NODES, oc->nodemask, gfp_zone(oc->gfp_mask));
-		if (should_dump_unreclaim_slab())
-			dump_unreclaimable_slab();
 	}
 	if (sysctl_oom_dump_tasks)
 		dump_tasks(oc);
diff --git a/mm/show_mem.c b/mm/show_mem.c
index bdb439551eef..a8ea4c41ced5 100644
--- a/mm/show_mem.c
+++ b/mm/show_mem.c
@@ -7,15 +7,18 @@
 
 #include <linux/blkdev.h>
 #include <linux/cma.h>
+#include <linux/console.h>
 #include <linux/cpuset.h>
 #include <linux/highmem.h>
 #include <linux/hugetlb.h>
 #include <linux/mm.h>
 #include <linux/mmzone.h>
+#include <linux/seq_buf.h>
 #include <linux/swap.h>
 #include <linux/vmstat.h>
 
 #include "internal.h"
+#include "slab.h"
 #include "swap.h"
 
 atomic_long_t _totalram_pages __read_mostly;
@@ -397,10 +400,31 @@ static void show_free_areas(unsigned int filter, nodemask_t *nodemask, int max_z
 	show_swap_cache_info();
 }
 
+static void print_string_as_lines(const char *prefix, const char *lines)
+{
+	if (!lines) {
+		printk("%s (null)\n", prefix);
+		return;
+	}
+
+	bool locked = console_trylock();
+
+	while (1) {
+		const char *p = strchrnul(lines, '\n');
+		printk("%s%.*s\n", prefix, (int) (p - lines), lines);
+		if (!*p)
+			break;
+		lines = p + 1;
+	}
+	if (locked)
+		console_unlock();
+}
+
 void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
 {
 	unsigned long total = 0, reserved = 0, highmem = 0;
 	struct zone *zone;
+	char *buf;
 
 	printk("Mem-Info:\n");
 	show_free_areas(filter, nodemask, max_zone_idx);
@@ -449,4 +473,23 @@ void __show_mem(unsigned int filter, nodemask_t *nodemask, int max_zone_idx)
 		}
 	}
 #endif
+
+	const unsigned buf_size = 8192;
+	buf = kmalloc(buf_size, GFP_ATOMIC);
+	if (buf) {
+		struct seq_buf s;
+
+		printk("Unreclaimable slab info:\n");
+		seq_buf_init(&s, buf, buf_size);
+		dump_unreclaimable_slab(&s);
+		print_string_as_lines(KERN_NOTICE, seq_buf_str(&s));
+
+		printk("Shrinkers:\n");
+		seq_buf_init(&s, buf, buf_size);
+		shrinkers_to_text(&s);
+		print_string_as_lines(KERN_NOTICE, seq_buf_str(&s));
+		/* previous output doesn't get flushed without this - why? */
+
+		kfree(buf);
+	}
 }
diff --git a/mm/slab.h b/mm/slab.h
index dcdb56b8e7f5..b523b3e3d9d3 100644
--- a/mm/slab.h
+++ b/mm/slab.h
@@ -611,10 +611,12 @@ static inline size_t slab_ksize(const struct kmem_cache *s)
 	return s->size;
 }
 
+struct seq_buf;
+
 #ifdef CONFIG_SLUB_DEBUG
-void dump_unreclaimable_slab(void);
+void dump_unreclaimable_slab(struct seq_buf *);
 #else
-static inline void dump_unreclaimable_slab(void)
+static inline void dump_unreclaimable_slab(struct seq_buf *out)
 {
 }
 #endif
diff --git a/mm/slab_common.c b/mm/slab_common.c
index 40b582a014b8..bd50a57161cf 100644
--- a/mm/slab_common.c
+++ b/mm/slab_common.c
@@ -27,6 +27,7 @@
 #include <asm/tlbflush.h>
 #include <asm/page.h>
 #include <linux/memcontrol.h>
+#include <linux/seq_buf.h>
 #include <linux/stackdepot.h>
 
 #include "internal.h"
@@ -1181,10 +1182,15 @@ static int slab_show(struct seq_file *m, void *p)
 	return 0;
 }
 
-void dump_unreclaimable_slab(void)
+void dump_unreclaimable_slab(struct seq_buf *out)
 {
 	struct kmem_cache *s;
 	struct slabinfo sinfo;
+	struct slab_by_mem {
+		struct kmem_cache *s;
+		size_t total, active;
+	} slabs_by_mem[10], n;
+	int i, nr = 0;
 
 	/*
 	 * Here acquiring slab_mutex is risky since we don't prefer to get
@@ -1194,24 +1200,52 @@ void dump_unreclaimable_slab(void)
 	 * without acquiring the mutex.
 	 */
 	if (!mutex_trylock(&slab_mutex)) {
-		pr_warn("excessive unreclaimable slab but cannot dump stats\n");
+		seq_buf_puts(out, "excessive unreclaimable slab but cannot dump stats\n");
 		return;
 	}
 
-	pr_info("Unreclaimable slab info:\n");
-	pr_info("Name                      Used          Total\n");
-
 	list_for_each_entry(s, &slab_caches, list) {
 		if (s->flags & SLAB_RECLAIM_ACCOUNT)
 			continue;
 
 		get_slabinfo(s, &sinfo);
 
-		if (sinfo.num_objs > 0)
-			pr_info("%-17s %10luKB %10luKB\n", s->name,
-				(sinfo.active_objs * s->size) / 1024,
-				(sinfo.num_objs * s->size) / 1024);
+		if (!sinfo.num_objs)
+			continue;
+
+		n.s = s;
+		n.total = sinfo.num_objs * s->size;
+		n.active = sinfo.active_objs * s->size;
+
+		for (i = 0; i < nr; i++)
+			if (n.total < slabs_by_mem[i].total)
+				break;
+
+		if (nr < ARRAY_SIZE(slabs_by_mem)) {
+			memmove(&slabs_by_mem[i + 1],
+				&slabs_by_mem[i],
+				sizeof(slabs_by_mem[0]) * (nr - i));
+			nr++;
+		} else if (i) {
+			i--;
+			memmove(&slabs_by_mem[0],
+				&slabs_by_mem[1],
+				sizeof(slabs_by_mem[0]) * i);
+		} else {
+			continue;
+		}
+
+		slabs_by_mem[i] = n;
 	}
+
+	for (i = nr - 1; i >= 0; --i) {
+		seq_buf_printf(out, "%-17s total: ", slabs_by_mem[i].s->name);
+		seq_buf_human_readable_u64(out, slabs_by_mem[i].total, STRING_UNITS_2);
+		seq_buf_printf(out, " active: ");
+		seq_buf_human_readable_u64(out, slabs_by_mem[i].active, STRING_UNITS_2);
+		seq_buf_putc(out, '\n');
+	}
+
 	mutex_unlock(&slab_mutex);
 }
 
-- 
2.45.2

[PATCH 05/10] mm: shrinker: Add shrinker_to_text() to debugfs interface

[Kent Overstreet]

Previously, we added shrinker_to_text() and hooked it up to the OOM
report - now, the same report is available via debugfs.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 mm/shrinker_debug.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/mm/shrinker_debug.c b/mm/shrinker_debug.c
index 12ea5486a3e9..39342aa9f4ca 100644
--- a/mm/shrinker_debug.c
+++ b/mm/shrinker_debug.c
@@ -2,6 +2,7 @@
 #include <linux/idr.h>
 #include <linux/slab.h>
 #include <linux/debugfs.h>
+#include <linux/seq_buf.h>
 #include <linux/seq_file.h>
 #include <linux/shrinker.h>
 #include <linux/memcontrol.h>
@@ -159,6 +160,21 @@ static const struct file_operations shrinker_debugfs_scan_fops = {
 	.write	 = shrinker_debugfs_scan_write,
 };
 
+static int shrinker_debugfs_report_show(struct seq_file *m, void *v)
+{
+	struct shrinker *shrinker = m->private;
+	char *bufp;
+	size_t buflen = seq_get_buf(m, &bufp);
+	struct seq_buf out;
+
+	seq_buf_init(&out, bufp, buflen);
+	shrinker_to_text(&out, shrinker);
+	seq_commit(m, seq_buf_used(&out));
+
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(shrinker_debugfs_report);
+
 int shrinker_debugfs_add(struct shrinker *shrinker)
 {
 	struct dentry *entry;
@@ -190,6 +206,8 @@ int shrinker_debugfs_add(struct shrinker *shrinker)
 			    &shrinker_debugfs_count_fops);
 	debugfs_create_file("scan", 0220, entry, shrinker,
 			    &shrinker_debugfs_scan_fops);
+	debugfs_create_file("report", 0440, entry, shrinker,
+			    &shrinker_debugfs_report_fops);
 	return 0;
 }
 
-- 
2.45.2

[PATCH 06/10] bcachefs: shrinker.to_text() methods

[Kent Overstreet]

This adds shrinker.to_text() methods for our shrinkers and hooks them up
to our existing to_text() functions.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/btree_cache.c     | 13 +++++++++++++
 fs/bcachefs/btree_key_cache.c | 14 ++++++++++++++
 2 files changed, 27 insertions(+)

diff --git a/fs/bcachefs/btree_cache.c b/fs/bcachefs/btree_cache.c
index 662f0f79b7af..9f096fdcaf9a 100644
--- a/fs/bcachefs/btree_cache.c
+++ b/fs/bcachefs/btree_cache.c
@@ -15,6 +15,7 @@
 
 #include <linux/prefetch.h>
 #include <linux/sched/mm.h>
+#include <linux/seq_buf.h>
 
 #define BTREE_CACHE_NOT_FREED_INCREMENT(counter) \
 do {						 \
@@ -487,6 +488,17 @@ static unsigned long bch2_btree_cache_count(struct shrinker *shrink,
 	return btree_cache_can_free(bc);
 }
 
+static void bch2_btree_cache_shrinker_to_text(struct seq_buf *s, struct shrinker *shrink)
+{
+	struct bch_fs *c = shrink->private_data;
+	char *cbuf;
+	size_t buflen = seq_buf_get_buf(s, &cbuf);
+	struct printbuf out = PRINTBUF_EXTERN(cbuf, buflen);
+
+	bch2_btree_cache_to_text(&out, &c->btree_cache);
+	seq_buf_commit(s, out.pos);
+}
+
 void bch2_fs_btree_cache_exit(struct bch_fs *c)
 {
 	struct btree_cache *bc = &c->btree_cache;
@@ -570,6 +582,7 @@ int bch2_fs_btree_cache_init(struct bch_fs *c)
 	bc->shrink = shrink;
 	shrink->count_objects	= bch2_btree_cache_count;
 	shrink->scan_objects	= bch2_btree_cache_scan;
+	shrink->to_text		= bch2_btree_cache_shrinker_to_text;
 	shrink->seeks		= 4;
 	shrink->private_data	= c;
 	shrinker_register(shrink);
diff --git a/fs/bcachefs/btree_key_cache.c b/fs/bcachefs/btree_key_cache.c
index 2e49ca71194f..af84516fb607 100644
--- a/fs/bcachefs/btree_key_cache.c
+++ b/fs/bcachefs/btree_key_cache.c
@@ -13,6 +13,7 @@
 #include "trace.h"
 
 #include <linux/sched/mm.h>
+#include <linux/seq_buf.h>
 
 static inline bool btree_uses_pcpu_readers(enum btree_id id)
 {
@@ -746,6 +747,18 @@ void bch2_fs_btree_key_cache_init_early(struct btree_key_cache *c)
 {
 }
 
+static void bch2_btree_key_cache_shrinker_to_text(struct seq_buf *s, struct shrinker *shrink)
+{
+	struct bch_fs *c = shrink->private_data;
+	struct btree_key_cache *bc = &c->btree_key_cache;
+	char *cbuf;
+	size_t buflen = seq_buf_get_buf(s, &cbuf);
+	struct printbuf out = PRINTBUF_EXTERN(cbuf, buflen);
+
+	bch2_btree_key_cache_to_text(&out, bc);
+	seq_buf_commit(s, out.pos);
+}
+
 int bch2_fs_btree_key_cache_init(struct btree_key_cache *bc)
 {
 	struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
@@ -770,6 +783,7 @@ int bch2_fs_btree_key_cache_init(struct btree_key_cache *bc)
 	bc->shrink = shrink;
 	shrink->count_objects	= bch2_btree_key_cache_count;
 	shrink->scan_objects	= bch2_btree_key_cache_scan;
+	shrink->to_text		= bch2_btree_key_cache_shrinker_to_text;
 	shrink->batch		= 1 << 14;
 	shrink->seeks		= 0;
 	shrink->private_data	= c;
-- 
2.45.2

[PATCH 07/10] percpu: per_cpu_sum()

[Kent Overstreet]

Add a little helper to replace open coded versions.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: linux-fsdevel@vger.krenel.org
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/bcachefs/util.h     | 10 ----------
 fs/dcache.c            | 16 +++-------------
 include/linux/percpu.h | 10 ++++++++++
 3 files changed, 13 insertions(+), 23 deletions(-)

diff --git a/fs/bcachefs/util.h b/fs/bcachefs/util.h
index fb02c1c36004..e90c2f546007 100644
--- a/fs/bcachefs/util.h
+++ b/fs/bcachefs/util.h
@@ -584,16 +584,6 @@ do {									\
 	}								\
 } while (0)
 
-#define per_cpu_sum(_p)							\
-({									\
-	typeof(*_p) _ret = 0;						\
-									\
-	int cpu;							\
-	for_each_possible_cpu(cpu)					\
-		_ret += *per_cpu_ptr(_p, cpu);				\
-	_ret;								\
-})
-
 static inline u64 percpu_u64_get(u64 __percpu *src)
 {
 	return per_cpu_sum(src);
diff --git a/fs/dcache.c b/fs/dcache.c
index 3d8daaecb6d1..64108cbd52f6 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -151,29 +151,19 @@ static struct dentry_stat_t dentry_stat = {
  */
 static long get_nr_dentry(void)
 {
-	int i;
-	long sum = 0;
-	for_each_possible_cpu(i)
-		sum += per_cpu(nr_dentry, i);
+	long sum = per_cpu_sum(&nr_dentry);
 	return sum < 0 ? 0 : sum;
 }
 
 static long get_nr_dentry_unused(void)
 {
-	int i;
-	long sum = 0;
-	for_each_possible_cpu(i)
-		sum += per_cpu(nr_dentry_unused, i);
+	long sum = per_cpu_sum(&nr_dentry_unused);
 	return sum < 0 ? 0 : sum;
 }
 
 static long get_nr_dentry_negative(void)
 {
-	int i;
-	long sum = 0;
-
-	for_each_possible_cpu(i)
-		sum += per_cpu(nr_dentry_negative, i);
+	long sum = per_cpu_sum(&nr_dentry_negative);
 	return sum < 0 ? 0 : sum;
 }
 
diff --git a/include/linux/percpu.h b/include/linux/percpu.h
index 4b2047b78b67..0df28ff54f66 100644
--- a/include/linux/percpu.h
+++ b/include/linux/percpu.h
@@ -162,4 +162,14 @@ extern phys_addr_t per_cpu_ptr_to_phys(void *addr);
 
 extern unsigned long pcpu_nr_pages(void);
 
+#define per_cpu_sum(_p)							\
+({									\
+	typeof(*(_p)) sum = 0;						\
+	int cpu;							\
+									\
+	for_each_possible_cpu(cpu)					\
+		sum += *per_cpu_ptr(_p, cpu);				\
+	sum;								\
+})
+
 #endif /* __LINUX_PERCPU_H */
-- 
2.45.2

[PATCH 08/10] fs: Add super_block->s_inodes_nr

[Kent Overstreet]

Upcoming shrinker debugging patchset is going to give us a callback for
reporting on all memory owned by a shrinker.

This adds a counter for total number of inodes allocated for a given
superblock, so we can compare with the number of reclaimable inodes we
already have.

Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Dave Chinner <david@fromorbit.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/inode.c         | 2 ++
 fs/super.c         | 7 +++++++
 include/linux/fs.h | 1 +
 3 files changed, 10 insertions(+)

diff --git a/fs/inode.c b/fs/inode.c
index 5e7dcdeedd4d..2650c5ce74e1 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -489,12 +489,14 @@ void inode_sb_list_add(struct inode *inode)
 	spin_lock(&inode->i_sb->s_inode_list_lock);
 	list_add(&inode->i_sb_list, &inode->i_sb->s_inodes);
 	spin_unlock(&inode->i_sb->s_inode_list_lock);
+	this_cpu_inc(*inode->i_sb->s_inodes_nr);
 }
 EXPORT_SYMBOL_GPL(inode_sb_list_add);
 
 static inline void inode_sb_list_del(struct inode *inode)
 {
 	if (!list_empty(&inode->i_sb_list)) {
+		this_cpu_dec(*inode->i_sb->s_inodes_nr);
 		spin_lock(&inode->i_sb->s_inode_list_lock);
 		list_del_init(&inode->i_sb_list);
 		spin_unlock(&inode->i_sb->s_inode_list_lock);
diff --git a/fs/super.c b/fs/super.c
index b7913b55debc..b1b6ae491b6c 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -278,6 +278,7 @@ static void destroy_super_work(struct work_struct *work)
 	security_sb_free(s);
 	put_user_ns(s->s_user_ns);
 	kfree(s->s_subtype);
+	free_percpu(s->s_inodes_nr);
 	for (int i = 0; i < SB_FREEZE_LEVELS; i++)
 		percpu_free_rwsem(&s->s_writers.rw_sem[i]);
 	kfree(s);
@@ -298,6 +299,7 @@ static void destroy_unused_super(struct super_block *s)
 	super_unlock_excl(s);
 	list_lru_destroy(&s->s_dentry_lru);
 	list_lru_destroy(&s->s_inode_lru);
+	free_percpu(s->s_inodes_nr);
 	shrinker_free(s->s_shrink);
 	/* no delays needed */
 	destroy_super_work(&s->destroy_work);
@@ -375,6 +377,10 @@ static struct super_block *alloc_super(struct file_system_type *type, int flags,
 	s->s_time_min = TIME64_MIN;
 	s->s_time_max = TIME64_MAX;
 
+	s->s_inodes_nr = alloc_percpu(size_t);
+	if (!s->s_inodes_nr)
+		goto fail;
+
 	s->s_shrink = shrinker_alloc(SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE,
 				     "sb-%s", type->name);
 	if (!s->s_shrink)
@@ -408,6 +414,7 @@ static void __put_super(struct super_block *s)
 		WARN_ON(s->s_dentry_lru.node);
 		WARN_ON(s->s_inode_lru.node);
 		WARN_ON(!list_empty(&s->s_mounts));
+		WARN_ON(per_cpu_sum(s->s_inodes_nr));
 		call_rcu(&s->rcu, destroy_super_rcu);
 	}
 }
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 8fc4bad3b6ae..86636831b9d0 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1346,6 +1346,7 @@ struct super_block {
 	/* s_inode_list_lock protects s_inodes */
 	spinlock_t		s_inode_list_lock ____cacheline_aligned_in_smp;
 	struct list_head	s_inodes;	/* all inodes */
+	size_t __percpu		*s_inodes_nr;
 
 	spinlock_t		s_inode_wblist_lock;
 	struct list_head	s_inodes_wb;	/* writeback inodes */
-- 
2.45.2

[PATCH 09/10] fs/dcache: Add per-sb accounting for nr dentries

[Kent Overstreet]

Like the previous patch, add a counter for total dentries, so we can
print total vs. reclaimable.

Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Dave Chinner <david@fromorbit.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/dcache.c        | 2 ++
 fs/super.c         | 6 ++++++
 include/linux/fs.h | 1 +
 3 files changed, 9 insertions(+)

diff --git a/fs/dcache.c b/fs/dcache.c
index 64108cbd52f6..4bbb2c87f824 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -601,6 +601,7 @@ static struct dentry *__dentry_kill(struct dentry *dentry)
 	else
 		spin_unlock(&dentry->d_lock);
 	this_cpu_dec(nr_dentry);
+	this_cpu_dec(*dentry->d_sb->s_dentry_nr);
 	if (dentry->d_op && dentry->d_op->d_release)
 		dentry->d_op->d_release(dentry);
 
@@ -1683,6 +1684,7 @@ static struct dentry *__d_alloc(struct super_block *sb, const struct qstr *name)
 	}
 
 	this_cpu_inc(nr_dentry);
+	this_cpu_inc(*sb->s_dentry_nr);
 
 	return dentry;
 }
diff --git a/fs/super.c b/fs/super.c
index b1b6ae491b6c..5b0fea6ff1cd 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -278,6 +278,7 @@ static void destroy_super_work(struct work_struct *work)
 	security_sb_free(s);
 	put_user_ns(s->s_user_ns);
 	kfree(s->s_subtype);
+	free_percpu(s->s_dentry_nr);
 	free_percpu(s->s_inodes_nr);
 	for (int i = 0; i < SB_FREEZE_LEVELS; i++)
 		percpu_free_rwsem(&s->s_writers.rw_sem[i]);
@@ -299,6 +300,7 @@ static void destroy_unused_super(struct super_block *s)
 	super_unlock_excl(s);
 	list_lru_destroy(&s->s_dentry_lru);
 	list_lru_destroy(&s->s_inode_lru);
+	free_percpu(s->s_dentry_nr);
 	free_percpu(s->s_inodes_nr);
 	shrinker_free(s->s_shrink);
 	/* no delays needed */
@@ -381,6 +383,10 @@ static struct super_block *alloc_super(struct file_system_type *type, int flags,
 	if (!s->s_inodes_nr)
 		goto fail;
 
+	s->s_dentry_nr = alloc_percpu(size_t);
+	if (!s->s_dentry_nr)
+		goto fail;
+
 	s->s_shrink = shrinker_alloc(SHRINKER_NUMA_AWARE | SHRINKER_MEMCG_AWARE,
 				     "sb-%s", type->name);
 	if (!s->s_shrink)
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 86636831b9d0..493fb8e72bf0 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1332,6 +1332,7 @@ struct super_block {
 	 * There is no need to put them into separate cachelines.
 	 */
 	struct list_lru		s_dentry_lru;
+	size_t __percpu		*s_dentry_nr;
 	struct list_lru		s_inode_lru;
 	struct rcu_head		rcu;
 	struct work_struct	destroy_work;
-- 
2.45.2

[PATCH 10/10] fs: super_cache_to_text()

[Kent Overstreet]

Implement shrinker.to_text() for the superblock shrinker: print out nr
of dentries and inodes, total and shrinkable.

Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Dave Chinner <david@fromorbit.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
 fs/super.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/fs/super.c b/fs/super.c
index 5b0fea6ff1cd..d3e43127e311 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -36,6 +36,7 @@
 #include <linux/lockdep.h>
 #include <linux/user_namespace.h>
 #include <linux/fs_context.h>
+#include <linux/seq_buf.h>
 #include <uapi/linux/mount.h>
 #include "internal.h"
 
@@ -270,6 +271,16 @@ static unsigned long super_cache_count(struct shrinker *shrink,
 	return total_objects;
 }
 
+static void super_cache_to_text(struct seq_buf *out, struct shrinker *shrink)
+{
+	struct super_block *sb = shrink->private_data;
+
+	seq_buf_printf(out, "inodes:   total %zu shrinkable %lu\n",
+		       per_cpu_sum(sb->s_inodes_nr), list_lru_count(&sb->s_inode_lru));
+	seq_buf_printf(out, "dentries: toal %zu shrinkbale %lu\n",
+		       per_cpu_sum(sb->s_dentry_nr), list_lru_count(&sb->s_dentry_lru));
+}
+
 static void destroy_super_work(struct work_struct *work)
 {
 	struct super_block *s = container_of(work, struct super_block,
@@ -394,6 +405,7 @@ static struct super_block *alloc_super(struct file_system_type *type, int flags,
 
 	s->s_shrink->scan_objects = super_cache_scan;
 	s->s_shrink->count_objects = super_cache_count;
+	s->s_shrink->to_text = super_cache_to_text;
 	s->s_shrink->batch = 1024;
 	s->s_shrink->private_data = s;
 
-- 
2.45.2