bcachefs: Add persistent counters

This adds a new superblock field for persisting counters
and adds a sysfs interface in counters/ exposing these counters.

The superblock field is ignored by older versions letting us avoid
an on disk version bump.

Each sysfs file outputs a counter that tracks since filesystem
creation and a counter for the current mount session.

Signed-off-by: Daniel Hill <daniel@gluo.nz>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/Makefile

@@ -21,6 +21,7 @@ bcachefs-y		:=	\
 	checksum.o		\
 	clock.o			\
 	compress.o		\
+	counters.o		\
 	debug.o			\
 	dirent.o		\
 	disk_groups.o		\

fs/bcachefs/bcachefs.h

@@ -584,6 +584,7 @@ struct bch_fs {
 
 	struct list_head	list;
 	struct kobject		kobj;
+	struct kobject		counters_kobj;
 	struct kobject		internal;
 	struct kobject		opts_dir;
 	struct kobject		time_stats;
@@ -900,12 +901,15 @@ mempool_t		bio_bounce_pages;
 
 	u64			last_bucket_seq_cleanup;
 
-	/* The rest of this all shows up in sysfs */
+	/* TODO rewrite as counters - The rest of this all shows up in sysfs */
 	atomic_long_t		read_realloc_races;
 	atomic_long_t		extent_migrate_done;
 	atomic_long_t		extent_migrate_raced;
 	atomic_long_t		bucket_alloc_fail;
 
+	u64			counters_on_mount[BCH_COUNTER_NR];
+	u64 __percpu		*counters;
+
 	unsigned		btree_gc_periodic:1;
 	unsigned		copy_gc_enabled:1;
 	bool			promote_whole_extents;

fs/bcachefs/bcachefs_format.h

@@ -1090,7 +1090,8 @@ struct bch_sb_field {
 	x(clean,	6)			\
 	x(replicas,	7)			\
 	x(journal_seq_blacklist, 8)		\
-	x(journal_v2,	9)
+	x(journal_v2,	9)			\
+	x(counters,	10)
 
 enum bch_sb_field_type {
 #define x(f, nr)	BCH_SB_FIELD_##f = nr,
@@ -1323,6 +1324,25 @@ struct bch_sb_field_disk_groups {
 	struct bch_disk_group	entries[0];
 } __attribute__((packed, aligned(8)));
 
+/* BCH_SB_FIELD_counters */
+
+#define BCH_PERSISTENT_COUNTERS()	\
+	x(io_read,  0)			\
+	x(io_write, 1)			\
+	x(io_move,  2)
+
+enum bch_persistent_counters {
+#define x(t, n, ...) BCH_COUNTER_##t,
+	BCH_PERSISTENT_COUNTERS()
+#undef x
+	BCH_COUNTER_NR
+};
+
+struct bch_sb_field_counters {
+	struct bch_sb_field	field;
+	__le64			d[0];
+};
+
 /*
  * On clean shutdown, store btree roots and current journal sequence number in
  * the superblock:

fs/bcachefs/counters.c

@@ -0,0 +1,107 @@
+// SPDX-License-Identifier: GPL-2.0
+#include "bcachefs.h"
+#include "super-io.h"
+#include "counters.h"
+
+/* BCH_SB_FIELD_counters */
+
+const char * const bch2_counter_names[] = {
+#define x(t, n, ...) (#t),
+	BCH_PERSISTENT_COUNTERS()
+#undef x
+	NULL
+};
+
+static size_t bch2_sb_counter_nr_entries(struct bch_sb_field_counters *ctrs)
+{
+	if (!ctrs)
+		return 0;
+
+	return (__le64 *) vstruct_end(&ctrs->field) - &ctrs->d[0];
+};
+
+static int bch2_sb_counters_validate(struct bch_sb *sb,
+				     struct bch_sb_field *f,
+				     struct printbuf *err)
+{
+	return 0;
+};
+
+void bch2_sb_counters_to_text(struct printbuf *out, struct bch_sb *sb,
+			      struct bch_sb_field *f)
+{
+	struct bch_sb_field_counters *ctrs = field_to_type(f, counters);
+	unsigned int i;
+	unsigned int nr = bch2_sb_counter_nr_entries(ctrs);
+
+	for (i = 0; i < nr; i++) {
+		if (i < BCH_COUNTER_NR)
+			pr_buf(out, "%s", bch2_counter_names[i]);
+		else
+			pr_buf(out, "(unknown)");
+
+		pr_tab(out);
+		pr_buf(out, "%llu", le64_to_cpu(ctrs->d[i]));
+		pr_newline(out);
+	};
+};
+
+int bch2_sb_counters_to_cpu(struct bch_fs *c)
+{
+	struct bch_sb_field_counters *ctrs = bch2_sb_get_counters(c->disk_sb.sb);
+	unsigned int i;
+	unsigned int nr = bch2_sb_counter_nr_entries(ctrs);
+	u64 val = 0;
+
+	for (i = 0; i < BCH_COUNTER_NR; i++)
+		c->counters_on_mount[i] = 0;
+
+	for (i = 0; i < min_t(unsigned int, nr, BCH_COUNTER_NR); i++) {
+		val = le64_to_cpu(ctrs->d[i]);
+		percpu_u64_set(&c->counters[i], val);
+		c->counters_on_mount[i] = val;
+	}
+	return 0;
+};
+
+int bch2_sb_counters_from_cpu(struct bch_fs *c)
+{
+	struct bch_sb_field_counters *ctrs = bch2_sb_get_counters(c->disk_sb.sb);
+	struct bch_sb_field_counters *ret;
+	unsigned int i;
+	unsigned int nr = bch2_sb_counter_nr_entries(ctrs);
+
+	if (nr < BCH_COUNTER_NR) {
+		ret = bch2_sb_resize_counters(&c->disk_sb,
+					       sizeof(*ctrs) / sizeof(u64) + BCH_COUNTER_NR);
+
+		if (ret) {
+			ctrs = ret;
+			nr = bch2_sb_counter_nr_entries(ctrs);
+		}
+	}
+
+
+	for (i = 0; i < min_t(unsigned int, nr, BCH_COUNTER_NR); i++)
+		ctrs->d[i] = cpu_to_le64(percpu_u64_get(&c->counters[i]));
+	return 0;
+}
+
+void bch2_fs_counters_exit(struct bch_fs *c)
+{
+	free_percpu(c->counters);
+}
+
+int bch2_fs_counters_init(struct bch_fs *c)
+{
+	c->counters = __alloc_percpu(sizeof(u64) * BCH_COUNTER_NR, sizeof(u64));
+	if (!c->counters)
+		return -ENOMEM;
+
+	return bch2_sb_counters_to_cpu(c);
+}
+
+const struct bch_sb_field_ops bch_sb_field_ops_counters = {
+	.validate	= bch2_sb_counters_validate,
+	.to_text	= bch2_sb_counters_to_text,
+};

fs/bcachefs/counters.h

@@ -0,0 +1,17 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _BCACHEFS_COUNTERS_H
+#define _BCACHEFS_COUNTERS_H
+
+#include "bcachefs.h"
+#include "super-io.h"
+
+
+int bch2_sb_counters_to_cpu(struct bch_fs *);
+int bch2_sb_counters_from_cpu(struct bch_fs *);
+
+void bch2_fs_counters_exit(struct bch_fs *);
+int bch2_fs_counters_init(struct bch_fs *);
+
+extern const struct bch_sb_field_ops bch_sb_field_ops_counters;
+
+#endif // _BCACHEFS_COUNTERS_H

fs/bcachefs/io.c

@@ -1403,6 +1403,7 @@ void bch2_write(struct closure *cl)
 		goto err;
 	}
 
+	this_cpu_add(c->counters[BCH_COUNTER_io_write], bio_sectors(bio));
 	bch2_increment_clock(c, bio_sectors(bio), WRITE);
 
 	data_len = min_t(u64, bio->bi_iter.bi_size,
@@ -2310,6 +2311,7 @@ get_bio:
 	if (rbio->bounce)
 		trace_read_bounce(&rbio->bio);
 
+	this_cpu_add(c->counters[BCH_COUNTER_io_read], bio_sectors(&rbio->bio));
 	bch2_increment_clock(c, bio_sectors(&rbio->bio), READ);
 
 	/*

fs/bcachefs/move.c

@@ -575,6 +575,7 @@ static int bch2_move_extent(struct btree_trans *trans,
 
 	atomic64_inc(&ctxt->stats->keys_moved);
 	atomic64_add(k.k->size, &ctxt->stats->sectors_moved);
+	this_cpu_add(c->counters[BCH_COUNTER_io_move], k.k->size);
 
 	trace_move_extent(k.k);
 

fs/bcachefs/super-io.c

@@ -18,6 +18,7 @@
 #include "super.h"
 #include "trace.h"
 #include "vstructs.h"
+#include "counters.h"
 
 #include <linux/backing-dev.h>
 #include <linux/sort.h>
@@ -819,6 +820,8 @@ int bch2_write_super(struct bch_fs *c)
 
 	SET_BCH_SB_BIG_ENDIAN(c->disk_sb.sb, CPU_BIG_ENDIAN);
 
+	bch2_sb_counters_from_cpu(c);
+
 	for_each_online_member(ca, c, i)
 		bch2_sb_from_fs(c, ca);
 

fs/bcachefs/super.c

@@ -21,6 +21,7 @@
 #include "checksum.h"
 #include "clock.h"
 #include "compress.h"
+#include "counters.h"
 #include "debug.h"
 #include "disk_groups.h"
 #include "ec.h"
@@ -78,6 +79,9 @@ static const struct kobj_type type ## _ktype = {			\
 
 static void bch2_fs_release(struct kobject *);
 static void bch2_dev_release(struct kobject *);
+static void bch2_fs_counters_release(struct kobject *k)
+{
+}
 
 static void bch2_fs_internal_release(struct kobject *k)
 {
@@ -92,6 +96,7 @@ static void bch2_fs_time_stats_release(struct kobject *k)
 }
 
 KTYPE(bch2_fs);
+KTYPE(bch2_fs_counters);
 KTYPE(bch2_fs_internal);
 KTYPE(bch2_fs_opts_dir);
 KTYPE(bch2_fs_time_stats);
@@ -416,6 +421,7 @@ static void __bch2_fs_free(struct bch_fs *c)
 	for (i = 0; i < BCH_TIME_STAT_NR; i++)
 		bch2_time_stats_exit(&c->times[i]);
 
+	bch2_fs_counters_exit(c);
 	bch2_fs_snapshots_exit(c);
 	bch2_fs_quota_exit(c);
 	bch2_fs_fsio_exit(c);
@@ -500,6 +506,7 @@ void __bch2_fs_stop(struct bch_fs *c)
 	bch2_fs_debug_exit(c);
 	bch2_fs_chardev_exit(c);
 
+	kobject_put(&c->counters_kobj);
 	kobject_put(&c->time_stats);
 	kobject_put(&c->opts_dir);
 	kobject_put(&c->internal);
@@ -569,6 +576,7 @@ static int bch2_fs_online(struct bch_fs *c)
 	    kobject_add(&c->internal, &c->kobj, "internal") ?:
 	    kobject_add(&c->opts_dir, &c->kobj, "options") ?:
 	    kobject_add(&c->time_stats, &c->kobj, "time_stats") ?:
+	    kobject_add(&c->counters_kobj, &c->kobj, "counters") ?:
 	    bch2_opts_create_sysfs_files(&c->opts_dir);
 	if (ret) {
 		bch_err(c, "error creating sysfs objects");
@@ -617,6 +625,7 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 	kobject_init(&c->internal, &bch2_fs_internal_ktype);
 	kobject_init(&c->opts_dir, &bch2_fs_opts_dir_ktype);
 	kobject_init(&c->time_stats, &bch2_fs_time_stats_ktype);
+	kobject_init(&c->counters_kobj, &bch2_fs_counters_ktype);
 
 	c->minor		= -1;
 	c->disk_sb.fs_sb	= true;
@@ -777,7 +786,8 @@ static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
 	    bch2_fs_encryption_init(c) ?:
 	    bch2_fs_compress_init(c) ?:
 	    bch2_fs_ec_init(c) ?:
-	    bch2_fs_fsio_init(c);
+	    bch2_fs_fsio_init(c) ?:
+	    bch2_fs_counters_init(c);
 	if (ret)
 		goto err;
 

fs/bcachefs/sysfs.c

@@ -40,7 +40,7 @@
 #include "util.h"
 
 #define SYSFS_OPS(type)							\
-struct sysfs_ops type ## _sysfs_ops = {					\
+const struct sysfs_ops type ## _sysfs_ops = {					\
 	.show	= type ## _show,					\
 	.store	= type ## _store					\
 }
@@ -195,6 +195,10 @@ read_attribute(extent_migrate_done);
 read_attribute(extent_migrate_raced);
 read_attribute(bucket_alloc_fail);
 
+#define x(t, n, ...) read_attribute(t);
+BCH_PERSISTENT_COUNTERS()
+#undef x
+
 rw_attribute(discard);
 rw_attribute(label);
 
@@ -551,6 +555,47 @@ struct attribute *bch2_fs_files[] = {
 	NULL
 };
 
+/* counters dir */
+
+SHOW(bch2_fs_counters)
+{
+	struct bch_fs *c = container_of(kobj, struct bch_fs, counters_kobj);
+	u64 counter = 0;
+	u64 counter_since_mount = 0;
+
+	out->tabstops[0] = 32;
+	#define x(t, ...) \
+		if (attr == &sysfs_##t) {					\
+			counter             = percpu_u64_get(&c->counters[BCH_COUNTER_##t]);\
+			counter_since_mount = counter - c->counters_on_mount[BCH_COUNTER_##t];\
+			pr_buf(out, "since mount:");				\
+			pr_tab(out);						\
+			bch2_hprint(out, counter_since_mount << 9);		\
+			pr_newline(out);					\
+										\
+			pr_buf(out, "since filesystem creation:");		\
+			pr_tab(out);						\
+			bch2_hprint(out, counter << 9);				\
+			pr_newline(out);					\
+		}
+	BCH_PERSISTENT_COUNTERS()
+	#undef x
+	return 0;
+}
+
+STORE(bch2_fs_counters) {
+	return 0;
+}
+
+SYSFS_OPS(bch2_fs_counters);
+
+struct attribute *bch2_fs_counters_files[] = {
+#define x(t, ...) \
+	&sysfs_##t,
+	BCH_PERSISTENT_COUNTERS()
+#undef x
+	NULL
+};
 /* internal dir - just a wrapper */
 
 SHOW(bch2_fs_internal)

fs/bcachefs/sysfs.h

@@ -10,28 +10,32 @@ struct attribute;
 struct sysfs_ops;
 
 extern struct attribute *bch2_fs_files[];
+extern struct attribute *bch2_fs_counters_files[];
 extern struct attribute *bch2_fs_internal_files[];
 extern struct attribute *bch2_fs_opts_dir_files[];
 extern struct attribute *bch2_fs_time_stats_files[];
 extern struct attribute *bch2_dev_files[];
 
-extern struct sysfs_ops bch2_fs_sysfs_ops;
-extern struct sysfs_ops bch2_fs_internal_sysfs_ops;
-extern struct sysfs_ops bch2_fs_opts_dir_sysfs_ops;
-extern struct sysfs_ops bch2_fs_time_stats_sysfs_ops;
-extern struct sysfs_ops bch2_dev_sysfs_ops;
+extern const struct sysfs_ops bch2_fs_sysfs_ops;
+extern const struct sysfs_ops bch2_fs_counters_sysfs_ops;
+extern const struct sysfs_ops bch2_fs_internal_sysfs_ops;
+extern const struct sysfs_ops bch2_fs_opts_dir_sysfs_ops;
+extern const struct sysfs_ops bch2_fs_time_stats_sysfs_ops;
+extern const struct sysfs_ops bch2_dev_sysfs_ops;
 
 int bch2_opts_create_sysfs_files(struct kobject *);
 
 #else
 
 static struct attribute *bch2_fs_files[] = {};
+static struct attribute *bch2_fs_counters_files[] = {};
 static struct attribute *bch2_fs_internal_files[] = {};
 static struct attribute *bch2_fs_opts_dir_files[] = {};
 static struct attribute *bch2_fs_time_stats_files[] = {};
 static struct attribute *bch2_dev_files[] = {};
 
 static const struct sysfs_ops bch2_fs_sysfs_ops;
+static const struct sysfs_ops bch2_fs_counters_sysfs_ops;
 static const struct sysfs_ops bch2_fs_internal_sysfs_ops;
 static const struct sysfs_ops bch2_fs_opts_dir_sysfs_ops;
 static const struct sysfs_ops bch2_fs_time_stats_sysfs_ops;