X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fsysfs.c;h=d78ffcc0e8a457cfef0790908938bda9bce765b0;hb=2aed686c235a7d08adf601f147f823eb62e697a1;hp=ba04bbad9af98af77ea1069d61c091b8b0bf9066;hpb=03bc9d71b13e6f8e879894f93ea16f1f4a8280c9;p=bcachefs-tools-debian

diff --git a/libbcachefs/sysfs.c b/libbcachefs/sysfs.c
index ba04bba..d78ffcc 100644
--- a/libbcachefs/sysfs.c
+++ b/libbcachefs/sysfs.c
@@ -1,3 +1,4 @@
+// SPDX-License-Identifier: GPL-2.0
 /*
  * bcache sysfs interfaces
  *
@@ -5,25 +6,34 @@
  * Copyright 2012 Google, Inc.
  */
 
+#ifndef NO_BCACHEFS_SYSFS
+
 #include "bcachefs.h"
-#include "alloc.h"
-#include "compress.h"
+#include "alloc_background.h"
 #include "sysfs.h"
 #include "btree_cache.h"
+#include "btree_io.h"
 #include "btree_iter.h"
 #include "btree_update.h"
+#include "btree_update_interior.h"
 #include "btree_gc.h"
 #include "buckets.h"
+#include "clock.h"
+#include "disk_groups.h"
+#include "ec.h"
 #include "inode.h"
 #include "journal.h"
 #include "keylist.h"
 #include "move.h"
 #include "opts.h"
+#include "rebalance.h"
+#include "replicas.h"
 #include "super-io.h"
-#include "tier.h"
+#include "tests.h"
 
 #include <linux/blkdev.h>
 #include <linux/sort.h>
+#include <linux/sched/clock.h>
 
 #include "util.h"
 
@@ -52,7 +62,7 @@ static ssize_t fn ## _store(struct kobject *kobj, struct attribute *attr,\
 #define sysfs_printf(file, fmt, ...)					\
 do {									\
 	if (attr == &sysfs_ ## file)					\
-		return snprintf(buf, PAGE_SIZE, fmt "\n", __VA_ARGS__);	\
+		return scnprintf(buf, PAGE_SIZE, fmt "\n", __VA_ARGS__);\
 } while (0)
 
 #define sysfs_print(file, var)						\
@@ -64,9 +74,10 @@ do {									\
 #define sysfs_hprint(file, val)						\
 do {									\
 	if (attr == &sysfs_ ## file) {					\
-		ssize_t ret = bch2_hprint(buf, val);			\
-		strcat(buf, "\n");					\
-		return ret + 1;						\
+		struct printbuf out = _PBUF(buf, PAGE_SIZE);		\
+		bch2_hprint(&out, val);					\
+		pr_buf(&out, "\n");					\
+		return out.pos - buf;					\
 	}								\
 } while (0)
 
@@ -123,7 +134,9 @@ do {									\
 write_attribute(trigger_journal_flush);
 write_attribute(trigger_btree_coalesce);
 write_attribute(trigger_gc);
+write_attribute(trigger_alloc_write);
 write_attribute(prune_cache);
+rw_attribute(btree_gc_periodic);
 
 read_attribute(uuid);
 read_attribute(minor);
@@ -132,79 +145,78 @@ read_attribute(block_size);
 read_attribute(btree_node_size);
 read_attribute(first_bucket);
 read_attribute(nbuckets);
-read_attribute(read_priority_stats);
-read_attribute(write_priority_stats);
-read_attribute(fragmentation_stats);
-read_attribute(oldest_gen_stats);
+read_attribute(durability);
+read_attribute(iodone);
+
+read_attribute(io_latency_read);
+read_attribute(io_latency_write);
+read_attribute(io_latency_stats_read);
+read_attribute(io_latency_stats_write);
+read_attribute(congested);
+
+read_attribute(bucket_quantiles_last_read);
+read_attribute(bucket_quantiles_last_write);
+read_attribute(bucket_quantiles_fragmentation);
+read_attribute(bucket_quantiles_oldest_gen);
+
 read_attribute(reserve_stats);
 read_attribute(btree_cache_size);
 read_attribute(compression_stats);
-read_attribute(written);
-read_attribute(btree_written);
-read_attribute(metadata_written);
 read_attribute(journal_debug);
 read_attribute(journal_pins);
+read_attribute(btree_updates);
+read_attribute(dirty_btree_nodes);
 
 read_attribute(internal_uuid);
 
-read_attribute(available_buckets);
-read_attribute(free_buckets);
-read_attribute(dirty_data);
-read_attribute(dirty_bytes);
-read_attribute(dirty_buckets);
-read_attribute(cached_data);
-read_attribute(cached_bytes);
-read_attribute(cached_buckets);
-read_attribute(meta_buckets);
-read_attribute(alloc_buckets);
 read_attribute(has_data);
-read_attribute(has_metadata);
 read_attribute(alloc_debug);
+write_attribute(wake_allocator);
 
 read_attribute(read_realloc_races);
+read_attribute(extent_migrate_done);
+read_attribute(extent_migrate_raced);
 
 rw_attribute(journal_write_delay_ms);
 rw_attribute(journal_reclaim_delay_ms);
 
 rw_attribute(discard);
 rw_attribute(cache_replacement_policy);
+rw_attribute(label);
 
-rw_attribute(foreground_write_ratelimit_enabled);
 rw_attribute(copy_gc_enabled);
 sysfs_pd_controller_attribute(copy_gc);
 
-rw_attribute(tier);
-rw_attribute(tiering_enabled);
-rw_attribute(tiering_percent);
-sysfs_pd_controller_attribute(tiering);
+rw_attribute(rebalance_enabled);
+sysfs_pd_controller_attribute(rebalance);
+read_attribute(rebalance_work);
+rw_attribute(promote_whole_extents);
 
-sysfs_pd_controller_attribute(foreground_write);
+read_attribute(new_stripes);
 
 rw_attribute(pd_controllers_update_seconds);
 
-rw_attribute(foreground_target_percent);
-
 read_attribute(meta_replicas_have);
 read_attribute(data_replicas_have);
 
+read_attribute(io_timers_read);
+read_attribute(io_timers_write);
+
+#ifdef CONFIG_BCACHEFS_TESTS
+write_attribute(perf_test);
+#endif /* CONFIG_BCACHEFS_TESTS */
+
 #define BCH_DEBUG_PARAM(name, description)				\
 	rw_attribute(name);
 
 	BCH_DEBUG_PARAMS()
 #undef BCH_DEBUG_PARAM
 
-#define BCH_OPT(_name, _mode, ...)					\
-	static struct attribute sysfs_opt_##_name = {			\
-		.name = #_name,	.mode = _mode,				\
-	};
-
-	BCH_VISIBLE_OPTS()
-#undef BCH_OPT
-
-#define BCH_TIME_STAT(name, frequency_units, duration_units)		\
-	sysfs_time_stats_attribute(name, frequency_units, duration_units);
+#define x(_name)						\
+	static struct attribute sysfs_time_stat_##_name =		\
+		{ .name = #_name, .mode = S_IRUGO };
 	BCH_TIME_STATS()
-#undef BCH_TIME_STAT
+#undef x
 
 static struct attribute sysfs_state_rw = {
 	.name = "state",
@@ -216,78 +228,75 @@ static size_t bch2_btree_cache_size(struct bch_fs *c)
 	size_t ret = 0;
 	struct btree *b;
 
-	mutex_lock(&c->btree_cache_lock);
-	list_for_each_entry(b, &c->btree_cache, list)
+	mutex_lock(&c->btree_cache.lock);
+	list_for_each_entry(b, &c->btree_cache.live, list)
 		ret += btree_bytes(c);
 
-	mutex_unlock(&c->btree_cache_lock);
+	mutex_unlock(&c->btree_cache.lock);
 	return ret;
 }
 
 static ssize_t show_fs_alloc_debug(struct bch_fs *c, char *buf)
 {
-	struct bch_fs_usage stats = bch2_fs_usage_read(c);
+	struct printbuf out = _PBUF(buf, PAGE_SIZE);
+	struct bch_fs_usage *fs_usage = bch2_fs_usage_read(c);
 
-	return scnprintf(buf, PAGE_SIZE,
-			 "capacity:\t\t%llu\n"
-			 "compressed:\n"
-			 "\tmeta:\t\t%llu\n"
-			 "\tdirty:\t\t%llu\n"
-			 "\tcached:\t\t%llu\n"
-			 "uncompressed:\n"
-			 "\tmeta:\t\t%llu\n"
-			 "\tdirty:\t\t%llu\n"
-			 "\tcached:\t\t%llu\n"
-			 "persistent reserved sectors:\t%llu\n"
-			 "online reserved sectors:\t%llu\n",
-			 c->capacity,
-			 stats.s[S_COMPRESSED][S_META],
-			 stats.s[S_COMPRESSED][S_DIRTY],
-			 stats.s[S_COMPRESSED][S_CACHED],
-			 stats.s[S_UNCOMPRESSED][S_META],
-			 stats.s[S_UNCOMPRESSED][S_DIRTY],
-			 stats.s[S_UNCOMPRESSED][S_CACHED],
-			 stats.persistent_reserved,
-			 stats.online_reserved);
+	if (!fs_usage)
+		return -ENOMEM;
+
+	bch2_fs_usage_to_text(&out, c, fs_usage);
+
+	percpu_up_read(&c->mark_lock);
+
+	kfree(fs_usage);
+
+	return out.pos - buf;
 }
 
 static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 {
-	struct btree_iter iter;
+	struct btree_trans trans;
+	struct btree_iter *iter;
 	struct bkey_s_c k;
 	u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 	    nr_compressed_extents = 0,
 	    compressed_sectors_compressed = 0,
 	    compressed_sectors_uncompressed = 0;
+	int ret;
 
-	if (!bch2_fs_running(c))
+	if (!test_bit(BCH_FS_STARTED, &c->flags))
 		return -EPERM;
 
-	for_each_btree_key(&iter, c, BTREE_ID_EXTENTS, POS_MIN, k)
-		if (k.k->type == BCH_EXTENT) {
+	bch2_trans_init(&trans, c, 0, 0);
+
+	for_each_btree_key(&trans, iter, BTREE_ID_EXTENTS, POS_MIN, 0, k, ret)
+		if (k.k->type == KEY_TYPE_extent) {
 			struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
-			const struct bch_extent_ptr *ptr;
-			const union bch_extent_crc *crc;
+			const union bch_extent_entry *entry;
+			struct extent_ptr_decoded p;
 
-			extent_for_each_ptr_crc(e, ptr, crc) {
-				if (crc_compression_type(crc) == BCH_COMPRESSION_NONE) {
+			extent_for_each_ptr_decode(e, p, entry) {
+				if (!crc_is_compressed(p.crc)) {
 					nr_uncompressed_extents++;
 					uncompressed_sectors += e.k->size;
 				} else {
 					nr_compressed_extents++;
 					compressed_sectors_compressed +=
-						crc_compressed_size(e.k, crc);
+						p.crc.compressed_size;
 					compressed_sectors_uncompressed +=
-						crc_uncompressed_size(e.k, crc);
+						p.crc.uncompressed_size;
 				}
 
 				/* only looking at the first ptr */
 				break;
 			}
 		}
-	bch2_btree_iter_unlock(&iter);
 
-	return snprintf(buf, PAGE_SIZE,
+	ret = bch2_trans_exit(&trans) ?: ret;
+	if (ret)
+		return ret;
+
+	return scnprintf(buf, PAGE_SIZE,
 			"uncompressed data:\n"
 			"	nr extents:			%llu\n"
 			"	size (bytes):			%llu\n"
@@ -302,6 +311,41 @@ static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 			compressed_sectors_uncompressed << 9);
 }
 
+static ssize_t bch2_new_stripes(struct bch_fs *c, char *buf)
+{
+	char *out = buf, *end = buf + PAGE_SIZE;
+	struct ec_stripe_head *h;
+	struct ec_stripe_new *s;
+
+	mutex_lock(&c->ec_new_stripe_lock);
+	list_for_each_entry(h, &c->ec_new_stripe_list, list) {
+		out += scnprintf(out, end - out,
+				 "target %u algo %u redundancy %u:\n",
+				 h->target, h->algo, h->redundancy);
+
+		if (h->s)
+			out += scnprintf(out, end - out,
+					 "\tpending: blocks %u allocated %u\n",
+					 h->s->blocks.nr,
+					 bitmap_weight(h->s->blocks_allocated,
+						       h->s->blocks.nr));
+
+		mutex_lock(&h->lock);
+		list_for_each_entry(s, &h->stripes, list)
+			out += scnprintf(out, end - out,
+					 "\tin flight: blocks %u allocated %u pin %u\n",
+					 s->blocks.nr,
+					 bitmap_weight(s->blocks_allocated,
+						       s->blocks.nr),
+					 atomic_read(&s->pin));
+		mutex_unlock(&h->lock);
+
+	}
+	mutex_unlock(&c->ec_new_stripe_lock);
+
+	return out - buf;
+}
+
 SHOW(bch2_fs)
 {
 	struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
@@ -318,23 +362,28 @@ SHOW(bch2_fs)
 
 	sysfs_print(read_realloc_races,
 		    atomic_long_read(&c->read_realloc_races));
+	sysfs_print(extent_migrate_done,
+		    atomic_long_read(&c->extent_migrate_done));
+	sysfs_print(extent_migrate_raced,
+		    atomic_long_read(&c->extent_migrate_raced));
+
+	sysfs_printf(btree_gc_periodic, "%u",	(int) c->btree_gc_periodic);
 
-	sysfs_printf(foreground_write_ratelimit_enabled, "%i",
-		     c->foreground_write_ratelimit_enabled);
 	sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
-	sysfs_pd_controller_show(foreground_write, &c->foreground_write_pd);
 
 	sysfs_print(pd_controllers_update_seconds,
 		    c->pd_controllers_update_seconds);
-	sysfs_print(foreground_target_percent, c->foreground_target_percent);
 
-	sysfs_printf(tiering_enabled,		"%i", c->tiering_enabled);
-	sysfs_print(tiering_percent,		c->tiering_percent);
+	sysfs_printf(rebalance_enabled,		"%i", c->rebalance.enabled);
+	sysfs_pd_controller_show(rebalance,	&c->rebalance.pd); /* XXX */
+
+	if (attr == &sysfs_rebalance_work)
+		return bch2_rebalance_work_show(c, buf);
 
-	sysfs_pd_controller_show(tiering,	&c->tiers[1].pd); /* XXX */
+	sysfs_print(promote_whole_extents,	c->promote_whole_extents);
 
-	sysfs_printf(meta_replicas_have, "%u",	c->sb.meta_replicas_have);
-	sysfs_printf(data_replicas_have, "%u",	c->sb.data_replicas_have);
+	sysfs_printf(meta_replicas_have, "%i",	bch2_replicas_online(c, true));
+	sysfs_printf(data_replicas_have, "%i",	bch2_replicas_online(c, false));
 
 	/* Debugging: */
 
@@ -347,9 +396,23 @@ SHOW(bch2_fs)
 	if (attr == &sysfs_journal_pins)
 		return bch2_journal_print_pins(&c->journal, buf);
 
+	if (attr == &sysfs_btree_updates)
+		return bch2_btree_updates_print(c, buf);
+
+	if (attr == &sysfs_dirty_btree_nodes)
+		return bch2_dirty_btree_nodes_print(c, buf);
+
 	if (attr == &sysfs_compression_stats)
 		return bch2_compression_stats(c, buf);
 
+	if (attr == &sysfs_new_stripes)
+		return bch2_new_stripes(c, buf);
+
+	if (attr == &sysfs_io_timers_read)
+		return bch2_io_timers_show(&c->io_clock[READ], buf);
+	if (attr == &sysfs_io_timers_write)
+		return bch2_io_timers_show(&c->io_clock[WRITE], buf);
+
 #define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
 	BCH_DEBUG_PARAMS()
 #undef BCH_DEBUG_PARAM
@@ -364,8 +427,13 @@ STORE(__bch2_fs)
 	sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 	sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 
-	sysfs_strtoul(foreground_write_ratelimit_enabled,
-		      c->foreground_write_ratelimit_enabled);
+	if (attr == &sysfs_btree_gc_periodic) {
+		ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
+			?: (ssize_t) size;
+
+		wake_up_process(c->gc_thread);
+		return ret;
+	}
 
 	if (attr == &sysfs_copy_gc_enabled) {
 		struct bch_dev *ca;
@@ -374,27 +442,24 @@ STORE(__bch2_fs)
 			?: (ssize_t) size;
 
 		for_each_member_device(ca, c, i)
-			if (ca->moving_gc_read)
-				wake_up_process(ca->moving_gc_read);
+			if (ca->copygc_thread)
+				wake_up_process(ca->copygc_thread);
 		return ret;
 	}
 
-	if (attr == &sysfs_tiering_enabled) {
-		ssize_t ret = strtoul_safe(buf, c->tiering_enabled)
+	if (attr == &sysfs_rebalance_enabled) {
+		ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 			?: (ssize_t) size;
 
-		bch2_tiering_start(c); /* issue wakeups */
+		rebalance_wakeup(c);
 		return ret;
 	}
 
-	sysfs_pd_controller_store(foreground_write, &c->foreground_write_pd);
-
 	sysfs_strtoul(pd_controllers_update_seconds,
 		      c->pd_controllers_update_seconds);
-	sysfs_strtoul(foreground_target_percent, c->foreground_target_percent);
+	sysfs_pd_controller_store(rebalance,	&c->rebalance.pd);
 
-	sysfs_strtoul(tiering_percent,		c->tiering_percent);
-	sysfs_pd_controller_store(tiering,	&c->tiers[1].pd); /* XXX */
+	sysfs_strtoul(promote_whole_extents,	c->promote_whole_extents);
 
 	/* Debugging: */
 
@@ -402,7 +467,7 @@ STORE(__bch2_fs)
 	BCH_DEBUG_PARAMS()
 #undef BCH_DEBUG_PARAM
 
-	if (!bch2_fs_running(c))
+	if (!test_bit(BCH_FS_STARTED, &c->flags))
 		return -EPERM;
 
 	/* Debugging: */
@@ -414,16 +479,40 @@ STORE(__bch2_fs)
 		bch2_coalesce(c);
 
 	if (attr == &sysfs_trigger_gc)
-		bch2_gc(c);
+		bch2_gc(c, NULL, false, false);
+
+	if (attr == &sysfs_trigger_alloc_write) {
+		bool wrote;
+
+		bch2_alloc_write(c, 0, &wrote);
+	}
 
 	if (attr == &sysfs_prune_cache) {
 		struct shrink_control sc;
 
 		sc.gfp_mask = GFP_KERNEL;
 		sc.nr_to_scan = strtoul_or_return(buf);
-		c->btree_cache_shrink.scan_objects(&c->btree_cache_shrink, &sc);
+		c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 	}
-
+#ifdef CONFIG_BCACHEFS_TESTS
+	if (attr == &sysfs_perf_test) {
+		char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
+		char *test		= strsep(&p, " \t\n");
+		char *nr_str		= strsep(&p, " \t\n");
+		char *threads_str	= strsep(&p, " \t\n");
+		unsigned threads;
+		u64 nr;
+		int ret = -EINVAL;
+
+		if (threads_str &&
+		    !(ret = kstrtouint(threads_str, 10, &threads)) &&
+		    !(ret = bch2_strtoull_h(nr_str, &nr)))
+			bch2_btree_perf_test(c, test, nr, threads);
+		else
+			size = ret;
+		kfree(tmp);
+	}
+#endif
 	return size;
 }
 
@@ -451,10 +540,13 @@ struct attribute *bch2_fs_files[] = {
 	&sysfs_journal_write_delay_ms,
 	&sysfs_journal_reclaim_delay_ms,
 
-	&sysfs_foreground_target_percent,
-	&sysfs_tiering_percent,
+	&sysfs_promote_whole_extents,
 
 	&sysfs_compression_stats,
+
+#ifdef CONFIG_BCACHEFS_TESTS
+	&sysfs_perf_test,
+#endif
 	NULL
 };
 
@@ -477,19 +569,30 @@ struct attribute *bch2_fs_internal_files[] = {
 	&sysfs_alloc_debug,
 	&sysfs_journal_debug,
 	&sysfs_journal_pins,
+	&sysfs_btree_updates,
+	&sysfs_dirty_btree_nodes,
 
 	&sysfs_read_realloc_races,
+	&sysfs_extent_migrate_done,
+	&sysfs_extent_migrate_raced,
 
 	&sysfs_trigger_journal_flush,
 	&sysfs_trigger_btree_coalesce,
 	&sysfs_trigger_gc,
+	&sysfs_trigger_alloc_write,
 	&sysfs_prune_cache,
 
-	&sysfs_foreground_write_ratelimit_enabled,
 	&sysfs_copy_gc_enabled,
-	&sysfs_tiering_enabled,
-	sysfs_pd_controller_files(tiering),
-	sysfs_pd_controller_files(foreground_write),
+
+	&sysfs_rebalance_enabled,
+	&sysfs_rebalance_work,
+	sysfs_pd_controller_files(rebalance),
+
+	&sysfs_new_stripes,
+
+	&sysfs_io_timers_read,
+	&sysfs_io_timers_write,
+
 	&sysfs_internal_uuid,
 
 #define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
@@ -503,56 +606,79 @@ struct attribute *bch2_fs_internal_files[] = {
 
 SHOW(bch2_fs_opts_dir)
 {
+	struct printbuf out = _PBUF(buf, PAGE_SIZE);
 	struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
+	const struct bch_option *opt = container_of(attr, struct bch_option, attr);
+	int id = opt - bch2_opt_table;
+	u64 v = bch2_opt_get_by_id(&c->opts, id);
+
+	bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
+	pr_buf(&out, "\n");
 
-	return bch2_opt_show(&c->opts, attr->name, buf, PAGE_SIZE);
+	return out.pos - buf;
 }
 
 STORE(bch2_fs_opts_dir)
 {
 	struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
-	const struct bch_option *opt;
-	enum bch_opt_id id;
+	const struct bch_option *opt = container_of(attr, struct bch_option, attr);
+	int ret, id = opt - bch2_opt_table;
+	char *tmp;
 	u64 v;
 
-	id = bch2_parse_sysfs_opt(attr->name, buf, &v);
-	if (id < 0)
-		return id;
+	tmp = kstrdup(buf, GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
 
-	opt = &bch2_opt_table[id];
+	ret = bch2_opt_parse(c, opt, strim(tmp), &v);
+	kfree(tmp);
 
-	mutex_lock(&c->sb_lock);
+	if (ret < 0)
+		return ret;
 
-	if (id == Opt_compression) {
-		int ret = bch2_check_set_has_compressed_data(c, v);
-		if (ret) {
-			mutex_unlock(&c->sb_lock);
-			return ret;
-		}
-	}
+	ret = bch2_opt_check_may_set(c, id, v);
+	if (ret < 0)
+		return ret;
 
 	if (opt->set_sb != SET_NO_SB_OPT) {
-		opt->set_sb(c->disk_sb, v);
+		mutex_lock(&c->sb_lock);
+		opt->set_sb(c->disk_sb.sb, v);
 		bch2_write_super(c);
+		mutex_unlock(&c->sb_lock);
 	}
 
-	bch2_opt_set(&c->opts, id, v);
+	bch2_opt_set_by_id(&c->opts, id, v);
 
-	mutex_unlock(&c->sb_lock);
+	if ((id == Opt_background_target ||
+	     id == Opt_background_compression) && v) {
+		bch2_rebalance_add_work(c, S64_MAX);
+		rebalance_wakeup(c);
+	}
 
 	return size;
 }
 SYSFS_OPS(bch2_fs_opts_dir);
 
-struct attribute *bch2_fs_opts_dir_files[] = {
-#define BCH_OPT(_name, ...)						\
-	&sysfs_opt_##_name,
+struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 
-	BCH_VISIBLE_OPTS()
-#undef BCH_OPT
+int bch2_opts_create_sysfs_files(struct kobject *kobj)
+{
+	const struct bch_option *i;
+	int ret;
 
-	NULL
-};
+	for (i = bch2_opt_table;
+	     i < bch2_opt_table + bch2_opts_nr;
+	     i++) {
+		if (!(i->mode & (OPT_FORMAT|OPT_MOUNT|OPT_RUNTIME)))
+			continue;
+
+		ret = sysfs_create_file(kobj, &i->attr);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
 
 /* time stats */
 
@@ -560,78 +686,84 @@ SHOW(bch2_fs_time_stats)
 {
 	struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 
-#define BCH_TIME_STAT(name, frequency_units, duration_units)		\
-	sysfs_print_time_stats(&c->name##_time, name,			\
-			       frequency_units, duration_units);
+#define x(name)						\
+	if (attr == &sysfs_time_stat_##name)				\
+		return bch2_time_stats_print(&c->times[BCH_TIME_##name],\
+					     buf, PAGE_SIZE);
 	BCH_TIME_STATS()
-#undef BCH_TIME_STAT
+#undef x
 
 	return 0;
 }
 
 STORE(bch2_fs_time_stats)
 {
-	struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
-
-#define BCH_TIME_STAT(name, frequency_units, duration_units)		\
-	sysfs_clear_time_stats(&c->name##_time, name);
-	BCH_TIME_STATS()
-#undef BCH_TIME_STAT
-
 	return size;
 }
 SYSFS_OPS(bch2_fs_time_stats);
 
 struct attribute *bch2_fs_time_stats_files[] = {
-#define BCH_TIME_STAT(name, frequency_units, duration_units)		\
-	sysfs_time_stats_attribute_list(name, frequency_units, duration_units)
+#define x(name)						\
+	&sysfs_time_stat_##name,
 	BCH_TIME_STATS()
-#undef BCH_TIME_STAT
-
+#undef x
 	NULL
 };
 
-typedef unsigned (bucket_map_fn)(struct bch_dev *, struct bucket *, void *);
+typedef unsigned (bucket_map_fn)(struct bch_fs *, struct bch_dev *,
+				 size_t, void *);
 
-static unsigned bucket_priority_fn(struct bch_dev *ca, struct bucket *g,
-				   void *private)
+static unsigned bucket_last_io_fn(struct bch_fs *c, struct bch_dev *ca,
+				  size_t b, void *private)
 {
 	int rw = (private ? 1 : 0);
 
-	return ca->fs->prio_clock[rw].hand - g->prio[rw];
+	return bucket_last_io(c, bucket(ca, b), rw);
 }
 
-static unsigned bucket_sectors_used_fn(struct bch_dev *ca, struct bucket *g,
-				       void *private)
+static unsigned bucket_sectors_used_fn(struct bch_fs *c, struct bch_dev *ca,
+				       size_t b, void *private)
 {
-	return bucket_sectors_used(g);
+	struct bucket *g = bucket(ca, b);
+	return bucket_sectors_used(g->mark);
 }
 
-static unsigned bucket_oldest_gen_fn(struct bch_dev *ca, struct bucket *g,
-				     void *private)
+static unsigned bucket_oldest_gen_fn(struct bch_fs *c, struct bch_dev *ca,
+				     size_t b, void *private)
 {
-	return bucket_gc_gen(ca, g);
+	return bucket_gc_gen(ca, b);
 }
 
-static ssize_t show_quantiles(struct bch_dev *ca, char *buf,
-			      bucket_map_fn *fn, void *private)
+static int unsigned_cmp(const void *_l, const void *_r)
 {
-	int cmp(const void *l, const void *r)
-	{	return *((unsigned *) r) - *((unsigned *) l); }
+	const unsigned *l = _l;
+	const unsigned *r = _r;
 
-	size_t n = ca->mi.nbuckets, i;
+	return cmp_int(*l, *r);
+}
+
+static ssize_t show_quantiles(struct bch_fs *c, struct bch_dev *ca,
+			      char *buf, bucket_map_fn *fn, void *private)
+{
+	size_t i, n;
 	/* Compute 31 quantiles */
 	unsigned q[31], *p;
 	ssize_t ret = 0;
 
-	p = vzalloc(ca->mi.nbuckets * sizeof(unsigned));
-	if (!p)
+	down_read(&ca->bucket_lock);
+	n = ca->mi.nbuckets;
+
+	p = vzalloc(n * sizeof(unsigned));
+	if (!p) {
+		up_read(&ca->bucket_lock);
 		return -ENOMEM;
+	}
 
 	for (i = ca->mi.first_bucket; i < n; i++)
-		p[i] = fn(ca, &ca->buckets[i], private);
+		p[i] = fn(c, ca, i, private);
 
-	sort(p, n, sizeof(unsigned), cmp, NULL);
+	sort(p, n, sizeof(unsigned), unsigned_cmp, NULL);
+	up_read(&ca->bucket_lock);
 
 	while (n &&
 	       !p[n - 1])
@@ -648,80 +780,126 @@ static ssize_t show_quantiles(struct bch_dev *ca, char *buf,
 	buf[ret - 1] = '\n';
 
 	return ret;
-
 }
 
 static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
 {
+	struct printbuf out = _PBUF(buf, PAGE_SIZE);
 	enum alloc_reserve i;
-	ssize_t ret;
 
-	spin_lock(&ca->freelist_lock);
+	spin_lock(&ca->fs->freelist_lock);
 
-	ret = scnprintf(buf, PAGE_SIZE,
-			"free_inc:\t%zu\t%zu\n",
-			fifo_used(&ca->free_inc),
-			ca->free_inc.size);
+	pr_buf(&out, "free_inc:\t%zu\t%zu\n",
+	       fifo_used(&ca->free_inc),
+	       ca->free_inc.size);
 
 	for (i = 0; i < RESERVE_NR; i++)
-		ret += scnprintf(buf + ret, PAGE_SIZE - ret,
-				 "free[%u]:\t%zu\t%zu\n", i,
-				 fifo_used(&ca->free[i]),
-				 ca->free[i].size);
+		pr_buf(&out, "free[%u]:\t%zu\t%zu\n", i,
+		       fifo_used(&ca->free[i]),
+		       ca->free[i].size);
 
-	spin_unlock(&ca->freelist_lock);
+	spin_unlock(&ca->fs->freelist_lock);
 
-	return ret;
+	return out.pos - buf;
 }
 
 static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
 {
 	struct bch_fs *c = ca->fs;
-	struct bch_dev_usage stats = bch2_dev_usage_read(ca);
+	struct bch_dev_usage stats = bch2_dev_usage_read(c, ca);
+	unsigned i, nr[BCH_DATA_NR];
+
+	memset(nr, 0, sizeof(nr));
+
+	for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
+		nr[c->open_buckets[i].type]++;
 
 	return scnprintf(buf, PAGE_SIZE,
 		"free_inc:               %zu/%zu\n"
-		"free[RESERVE_PRIO]:     %zu/%zu\n"
 		"free[RESERVE_BTREE]:    %zu/%zu\n"
 		"free[RESERVE_MOVINGGC]: %zu/%zu\n"
 		"free[RESERVE_NONE]:     %zu/%zu\n"
-		"alloc:                  %llu/%llu\n"
-		"meta:                   %llu/%llu\n"
-		"dirty:                  %llu/%llu\n"
-		"available:              %llu/%llu\n"
+		"buckets:\n"
+		"    capacity:           %llu\n"
+		"    alloc:              %llu\n"
+		"    sb:                 %llu\n"
+		"    journal:            %llu\n"
+		"    meta:               %llu\n"
+		"    user:               %llu\n"
+		"    cached:             %llu\n"
+		"    erasure coded:      %llu\n"
+		"    available:          %lli\n"
+		"sectors:\n"
+		"    sb:                 %llu\n"
+		"    journal:            %llu\n"
+		"    meta:               %llu\n"
+		"    user:               %llu\n"
+		"    cached:             %llu\n"
+		"    fragmented:         %llu\n"
+		"    copygc threshold:   %llu\n"
 		"freelist_wait:          %s\n"
 		"open buckets:           %u/%u (reserved %u)\n"
-		"open_buckets_wait:      %s\n",
+		"open_buckets_wait:      %s\n"
+		"open_buckets_btree:     %u\n"
+		"open_buckets_user:      %u\n"
+		"btree reserve cache:    %u\n",
 		fifo_used(&ca->free_inc),		ca->free_inc.size,
-		fifo_used(&ca->free[RESERVE_PRIO]),	ca->free[RESERVE_PRIO].size,
 		fifo_used(&ca->free[RESERVE_BTREE]),	ca->free[RESERVE_BTREE].size,
 		fifo_used(&ca->free[RESERVE_MOVINGGC]),	ca->free[RESERVE_MOVINGGC].size,
 		fifo_used(&ca->free[RESERVE_NONE]),	ca->free[RESERVE_NONE].size,
-		stats.buckets_alloc,			ca->mi.nbuckets - ca->mi.first_bucket,
-		stats.buckets_meta,			ca->mi.nbuckets - ca->mi.first_bucket,
-		stats.buckets_dirty,			ca->mi.nbuckets - ca->mi.first_bucket,
-		__dev_buckets_available(ca, stats),	ca->mi.nbuckets - ca->mi.first_bucket,
+		ca->mi.nbuckets - ca->mi.first_bucket,
+		stats.buckets_alloc,
+		stats.buckets[BCH_DATA_SB],
+		stats.buckets[BCH_DATA_JOURNAL],
+		stats.buckets[BCH_DATA_BTREE],
+		stats.buckets[BCH_DATA_USER],
+		stats.buckets[BCH_DATA_CACHED],
+		stats.buckets_ec,
+		ca->mi.nbuckets - ca->mi.first_bucket - stats.buckets_unavailable,
+		stats.sectors[BCH_DATA_SB],
+		stats.sectors[BCH_DATA_JOURNAL],
+		stats.sectors[BCH_DATA_BTREE],
+		stats.sectors[BCH_DATA_USER],
+		stats.sectors[BCH_DATA_CACHED],
+		stats.sectors_fragmented,
+		ca->copygc_threshold,
 		c->freelist_wait.list.first		? "waiting" : "empty",
-		c->open_buckets_nr_free, OPEN_BUCKETS_COUNT, BTREE_NODE_RESERVE,
-		c->open_buckets_wait.list.first		? "waiting" : "empty");
+		c->open_buckets_nr_free, OPEN_BUCKETS_COUNT,
+		BTREE_NODE_OPEN_BUCKET_RESERVE,
+		c->open_buckets_wait.list.first		? "waiting" : "empty",
+		nr[BCH_DATA_BTREE],
+		nr[BCH_DATA_USER],
+		c->btree_reserve_cache_nr);
 }
 
-static u64 sectors_written(struct bch_dev *ca)
+static const char * const bch2_rw[] = {
+	"read",
+	"write",
+	NULL
+};
+
+static ssize_t show_dev_iodone(struct bch_dev *ca, char *buf)
 {
-	u64 ret = 0;
-	int cpu;
+	struct printbuf out = _PBUF(buf, PAGE_SIZE);
+	int rw, i;
 
-	for_each_possible_cpu(cpu)
-		ret += *per_cpu_ptr(ca->sectors_written, cpu);
+	for (rw = 0; rw < 2; rw++) {
+		pr_buf(&out, "%s:\n", bch2_rw[rw]);
 
-	return ret;
+		for (i = 1; i < BCH_DATA_NR; i++)
+			pr_buf(&out, "%-12s:%12llu\n",
+			       bch2_data_types[i],
+			       percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
+	}
+
+	return out.pos - buf;
 }
 
 SHOW(bch2_dev)
 {
 	struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 	struct bch_fs *c = ca->fs;
-	struct bch_dev_usage stats = bch2_dev_usage_read(ca);
+	struct printbuf out = _PBUF(buf, PAGE_SIZE);
 
 	sysfs_printf(uuid,		"%pU\n", ca->uuid.b);
 
@@ -729,49 +907,69 @@ SHOW(bch2_dev)
 	sysfs_print(block_size,		block_bytes(c));
 	sysfs_print(first_bucket,	ca->mi.first_bucket);
 	sysfs_print(nbuckets,		ca->mi.nbuckets);
+	sysfs_print(durability,		ca->mi.durability);
 	sysfs_print(discard,		ca->mi.discard);
-	sysfs_hprint(written, sectors_written(ca) << 9);
-	sysfs_hprint(btree_written,
-		     atomic64_read(&ca->btree_sectors_written) << 9);
-	sysfs_hprint(metadata_written,
-		     (atomic64_read(&ca->meta_sectors_written) +
-		      atomic64_read(&ca->btree_sectors_written)) << 9);
-
-	sysfs_hprint(dirty_data,	stats.sectors[S_DIRTY] << 9);
-	sysfs_print(dirty_bytes,	stats.sectors[S_DIRTY] << 9);
-	sysfs_print(dirty_buckets,	stats.buckets_dirty);
-	sysfs_hprint(cached_data,	stats.sectors[S_CACHED] << 9);
-	sysfs_print(cached_bytes,	stats.sectors[S_CACHED] << 9);
-	sysfs_print(cached_buckets,	stats.buckets_cached);
-	sysfs_print(meta_buckets,	stats.buckets_meta);
-	sysfs_print(alloc_buckets,	stats.buckets_alloc);
-	sysfs_print(available_buckets,	dev_buckets_available(ca));
-	sysfs_print(free_buckets,	dev_buckets_free(ca));
-	sysfs_print(has_data,		ca->mi.has_data);
-	sysfs_print(has_metadata,	ca->mi.has_metadata);
-
-	sysfs_pd_controller_show(copy_gc, &ca->moving_gc_pd);
-
-	if (attr == &sysfs_cache_replacement_policy)
-		return bch2_snprint_string_list(buf, PAGE_SIZE,
-						bch2_cache_replacement_policies,
-						ca->mi.replacement);
-
-	sysfs_print(tier,		ca->mi.tier);
-
-	if (attr == &sysfs_state_rw)
-		return bch2_snprint_string_list(buf, PAGE_SIZE,
-						bch2_dev_state,
-						ca->mi.state);
-
-	if (attr == &sysfs_read_priority_stats)
-		return show_quantiles(ca, buf, bucket_priority_fn, (void *) 0);
-	if (attr == &sysfs_write_priority_stats)
-		return show_quantiles(ca, buf, bucket_priority_fn, (void *) 1);
-	if (attr == &sysfs_fragmentation_stats)
-		return show_quantiles(ca, buf, bucket_sectors_used_fn, NULL);
-	if (attr == &sysfs_oldest_gen_stats)
-		return show_quantiles(ca, buf, bucket_oldest_gen_fn, NULL);
+
+	if (attr == &sysfs_label) {
+		if (ca->mi.group) {
+			mutex_lock(&c->sb_lock);
+			bch2_disk_path_to_text(&out, &c->disk_sb,
+					       ca->mi.group - 1);
+			mutex_unlock(&c->sb_lock);
+		}
+
+		pr_buf(&out, "\n");
+		return out.pos - buf;
+	}
+
+	if (attr == &sysfs_has_data) {
+		bch2_flags_to_text(&out, bch2_data_types,
+				   bch2_dev_has_data(c, ca));
+		pr_buf(&out, "\n");
+		return out.pos - buf;
+	}
+
+	sysfs_pd_controller_show(copy_gc, &ca->copygc_pd);
+
+	if (attr == &sysfs_cache_replacement_policy) {
+		bch2_string_opt_to_text(&out,
+					bch2_cache_replacement_policies,
+					ca->mi.replacement);
+		pr_buf(&out, "\n");
+		return out.pos - buf;
+	}
+
+	if (attr == &sysfs_state_rw) {
+		bch2_string_opt_to_text(&out, bch2_dev_state,
+					ca->mi.state);
+		pr_buf(&out, "\n");
+		return out.pos - buf;
+	}
+
+	if (attr == &sysfs_iodone)
+		return show_dev_iodone(ca, buf);
+
+	sysfs_print(io_latency_read,		atomic64_read(&ca->cur_latency[READ]));
+	sysfs_print(io_latency_write,		atomic64_read(&ca->cur_latency[WRITE]));
+
+	if (attr == &sysfs_io_latency_stats_read)
+		return bch2_time_stats_print(&ca->io_latency[READ], buf, PAGE_SIZE);
+	if (attr == &sysfs_io_latency_stats_write)
+		return bch2_time_stats_print(&ca->io_latency[WRITE], buf, PAGE_SIZE);
+
+	sysfs_printf(congested,			"%u%%",
+		     clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
+		     * 100 / CONGESTED_MAX);
+
+	if (attr == &sysfs_bucket_quantiles_last_read)
+		return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 0);
+	if (attr == &sysfs_bucket_quantiles_last_write)
+		return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 1);
+	if (attr == &sysfs_bucket_quantiles_fragmentation)
+		return show_quantiles(c, ca, buf, bucket_sectors_used_fn, NULL);
+	if (attr == &sysfs_bucket_quantiles_oldest_gen)
+		return show_quantiles(c, ca, buf, bucket_oldest_gen_fn, NULL);
+
 	if (attr == &sysfs_reserve_stats)
 		return show_reserve_stats(ca, buf);
 	if (attr == &sysfs_alloc_debug)
@@ -786,13 +984,13 @@ STORE(bch2_dev)
 	struct bch_fs *c = ca->fs;
 	struct bch_member *mi;
 
-	sysfs_pd_controller_store(copy_gc, &ca->moving_gc_pd);
+	sysfs_pd_controller_store(copy_gc, &ca->copygc_pd);
 
 	if (attr == &sysfs_discard) {
 		bool v = strtoul_or_return(buf);
 
 		mutex_lock(&c->sb_lock);
-		mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
+		mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 
 		if (v != BCH_MEMBER_DISCARD(mi)) {
 			SET_BCH_MEMBER_DISCARD(mi, v);
@@ -802,13 +1000,13 @@ STORE(bch2_dev)
 	}
 
 	if (attr == &sysfs_cache_replacement_policy) {
-		ssize_t v = bch2_read_string_list(buf, bch2_cache_replacement_policies);
+		ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
 
 		if (v < 0)
 			return v;
 
 		mutex_lock(&c->sb_lock);
-		mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
+		mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 
 		if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
 			SET_BCH_MEMBER_REPLACEMENT(mi, v);
@@ -817,31 +1015,23 @@ STORE(bch2_dev)
 		mutex_unlock(&c->sb_lock);
 	}
 
-	if (attr == &sysfs_tier) {
-		unsigned prev_tier;
-		unsigned v = strtoul_restrict_or_return(buf,
-					0, BCH_TIER_MAX - 1);
-
-		mutex_lock(&c->sb_lock);
-		prev_tier = ca->mi.tier;
-
-		if (v == ca->mi.tier) {
-			mutex_unlock(&c->sb_lock);
-			return size;
-		}
-
-		mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
-		SET_BCH_MEMBER_TIER(mi, v);
-		bch2_write_super(c);
+	if (attr == &sysfs_label) {
+		char *tmp;
+		int ret;
 
-		bch2_dev_group_remove(&c->tiers[prev_tier].devs, ca);
-		bch2_dev_group_add(&c->tiers[ca->mi.tier].devs, ca);
-		mutex_unlock(&c->sb_lock);
+		tmp = kstrdup(buf, GFP_KERNEL);
+		if (!tmp)
+			return -ENOMEM;
 
-		bch2_recalc_capacity(c);
-		bch2_tiering_start(c);
+		ret = bch2_dev_group_set(c, ca, strim(tmp));
+		kfree(tmp);
+		if (ret)
+			return ret;
 	}
 
+	if (attr == &sysfs_wake_allocator)
+		bch2_wake_allocator(ca);
+
 	return size;
 }
 SYSFS_OPS(bch2_dev);
@@ -852,45 +1042,37 @@ struct attribute *bch2_dev_files[] = {
 	&sysfs_block_size,
 	&sysfs_first_bucket,
 	&sysfs_nbuckets,
+	&sysfs_durability,
 
 	/* settings: */
 	&sysfs_discard,
 	&sysfs_cache_replacement_policy,
-	&sysfs_tier,
 	&sysfs_state_rw,
+	&sysfs_label,
 
 	&sysfs_has_data,
-	&sysfs_has_metadata,
-
-	/* io stats: */
-	&sysfs_written,
-	&sysfs_btree_written,
-	&sysfs_metadata_written,
-
-	/* alloc info - data: */
-	&sysfs_dirty_data,
-	&sysfs_dirty_bytes,
-	&sysfs_cached_data,
-	&sysfs_cached_bytes,
-
-	/* alloc info - buckets: */
-	&sysfs_available_buckets,
-	&sysfs_free_buckets,
-	&sysfs_dirty_buckets,
-	&sysfs_cached_buckets,
-	&sysfs_meta_buckets,
-	&sysfs_alloc_buckets,
+	&sysfs_iodone,
+
+	&sysfs_io_latency_read,
+	&sysfs_io_latency_write,
+	&sysfs_io_latency_stats_read,
+	&sysfs_io_latency_stats_write,
+	&sysfs_congested,
 
 	/* alloc info - other stats: */
-	&sysfs_read_priority_stats,
-	&sysfs_write_priority_stats,
-	&sysfs_fragmentation_stats,
-	&sysfs_oldest_gen_stats,
+	&sysfs_bucket_quantiles_last_read,
+	&sysfs_bucket_quantiles_last_write,
+	&sysfs_bucket_quantiles_fragmentation,
+	&sysfs_bucket_quantiles_oldest_gen,
+
 	&sysfs_reserve_stats,
 
 	/* debug: */
 	&sysfs_alloc_debug,
+	&sysfs_wake_allocator,
 
 	sysfs_pd_controller_files(copy_gc),
 	NULL
 };
+
+#endif  /* _BCACHEFS_SYSFS_H_ */