]> git.sesse.net Git - bcachefs-tools-debian/blobdiff - libbcachefs/sysfs.c
Disable pristine-tar option in gbp.conf, since there is no pristine-tar branch.
[bcachefs-tools-debian] / libbcachefs / sysfs.c
index d72ec0629a376b24f22deeb7e66b58969b0f72fe..c86a93a8d8fc81bbe373efcbec74f3e2563e6da5 100644 (file)
 #include "btree_gc.h"
 #include "buckets.h"
 #include "clock.h"
+#include "compress.h"
 #include "disk_groups.h"
 #include "ec.h"
 #include "inode.h"
 #include "journal.h"
 #include "keylist.h"
 #include "move.h"
+#include "movinggc.h"
+#include "nocow_locking.h"
 #include "opts.h"
 #include "rebalance.h"
 #include "replicas.h"
 #include "tests.h"
 
 #include <linux/blkdev.h>
-#include <linux/pretty-printers.h>
 #include <linux/sort.h>
 #include <linux/sched/clock.h>
 
 #include "util.h"
 
 #define SYSFS_OPS(type)                                                        \
-const struct sysfs_ops type ## _sysfs_ops = {                                  \
+const struct sysfs_ops type ## _sysfs_ops = {                          \
        .show   = type ## _show,                                        \
        .store  = type ## _store                                        \
 }
 
 #define SHOW(fn)                                                       \
 static ssize_t fn ## _to_text(struct printbuf *,                       \
-                             struct kobject *, struct attribute *);\
+                             struct kobject *, struct attribute *);    \
                                                                        \
 static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
                           char *buf)                                   \
@@ -67,23 +69,32 @@ static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
                memcpy(buf, out.buf, ret);                              \
        }                                                               \
        printbuf_exit(&out);                                            \
-       return ret;                                                     \
+       return bch2_err_class(ret);                                     \
 }                                                                      \
                                                                        \
 static ssize_t fn ## _to_text(struct printbuf *out, struct kobject *kobj,\
                              struct attribute *attr)
 
 #define STORE(fn)                                                      \
+static ssize_t fn ## _store_inner(struct kobject *, struct attribute *,\
+                           const char *, size_t);                      \
+                                                                       \
 static ssize_t fn ## _store(struct kobject *kobj, struct attribute *attr,\
                            const char *buf, size_t size)               \
+{                                                                      \
+       return bch2_err_class(fn##_store_inner(kobj, attr, buf, size)); \
+}                                                                      \
+                                                                       \
+static ssize_t fn ## _store_inner(struct kobject *kobj, struct attribute *attr,\
+                                 const char *buf, size_t size)
 
 #define __sysfs_attribute(_name, _mode)                                        \
        static struct attribute sysfs_##_name =                         \
                { .name = #_name, .mode = _mode }
 
-#define write_attribute(n)     __sysfs_attribute(n, S_IWUSR)
-#define read_attribute(n)      __sysfs_attribute(n, S_IRUGO)
-#define rw_attribute(n)                __sysfs_attribute(n, S_IRUGO|S_IWUSR)
+#define write_attribute(n)     __sysfs_attribute(n, 0200)
+#define read_attribute(n)      __sysfs_attribute(n, 0444)
+#define rw_attribute(n)                __sysfs_attribute(n, 0644)
 
 #define sysfs_printf(file, fmt, ...)                                   \
 do {                                                                   \
@@ -103,10 +114,6 @@ do {                                                                       \
                prt_human_readable_s64(out, val);                       \
 } while (0)
 
-#define var_printf(_var, fmt)  sysfs_printf(_var, fmt, var(_var))
-#define var_print(_var)                sysfs_print(_var, var(_var))
-#define var_hprint(_var)       sysfs_hprint(_var, var(_var))
-
 #define sysfs_strtoul(file, var)                                       \
 do {                                                                   \
        if (attr == &sysfs_ ## file)                                    \
@@ -129,44 +136,24 @@ do {                                                                      \
        _v;                                                             \
 })
 
-#define strtoul_restrict_or_return(cp, min, max)                       \
-({                                                                     \
-       unsigned long __v = 0;                                          \
-       int _r = strtoul_safe_restrict(cp, __v, min, max);              \
-       if (_r)                                                         \
-               return _r;                                              \
-       __v;                                                            \
-})
-
-#define strtoi_h_or_return(cp)                                         \
-({                                                                     \
-       u64 _v;                                                         \
-       int _r = strtoi_h(cp, &_v);                                     \
-       if (_r)                                                         \
-               return _r;                                              \
-       _v;                                                             \
-})
-
-#define sysfs_hatoi(file, var)                                         \
-do {                                                                   \
-       if (attr == &sysfs_ ## file)                                    \
-               return strtoi_h(buf, &var) ?: (ssize_t) size;           \
-} while (0)
-
 write_attribute(trigger_gc);
 write_attribute(trigger_discards);
 write_attribute(trigger_invalidates);
 write_attribute(prune_cache);
+write_attribute(btree_wakeup);
 rw_attribute(btree_gc_periodic);
 rw_attribute(gc_gens_pos);
 
 read_attribute(uuid);
 read_attribute(minor);
+read_attribute(flags);
 read_attribute(bucket_size);
 read_attribute(first_bucket);
 read_attribute(nbuckets);
-read_attribute(durability);
-read_attribute(iodone);
+rw_attribute(durability);
+read_attribute(io_done);
+read_attribute(io_errors);
+write_attribute(io_errors_reset);
 
 read_attribute(io_latency_read);
 read_attribute(io_latency_write);
@@ -174,7 +161,7 @@ read_attribute(io_latency_stats_read);
 read_attribute(io_latency_stats_write);
 read_attribute(congested);
 
-read_attribute(btree_avg_write_size);
+read_attribute(btree_write_stats);
 
 read_attribute(btree_cache_size);
 read_attribute(compression_stats);
@@ -182,20 +169,41 @@ read_attribute(journal_debug);
 read_attribute(btree_updates);
 read_attribute(btree_cache);
 read_attribute(btree_key_cache);
-read_attribute(btree_transactions);
 read_attribute(stripes_heap);
 read_attribute(open_buckets);
+read_attribute(open_buckets_partial);
+read_attribute(write_points);
+read_attribute(nocow_lock_table);
+
+#ifdef BCH_WRITE_REF_DEBUG
+read_attribute(write_refs);
+
+static const char * const bch2_write_refs[] = {
+#define x(n)   #n,
+       BCH_WRITE_REFS()
+#undef x
+       NULL
+};
+
+static void bch2_write_refs_to_text(struct printbuf *out, struct bch_fs *c)
+{
+       bch2_printbuf_tabstop_push(out, 24);
+
+       for (unsigned i = 0; i < ARRAY_SIZE(c->writes); i++) {
+               prt_str(out, bch2_write_refs[i]);
+               prt_tab(out);
+               prt_printf(out, "%li", atomic_long_read(&c->writes[i]));
+               prt_newline(out);
+       }
+}
+#endif
 
 read_attribute(internal_uuid);
+read_attribute(disk_groups);
 
 read_attribute(has_data);
 read_attribute(alloc_debug);
 
-read_attribute(read_realloc_races);
-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
@@ -208,7 +216,7 @@ read_attribute(copy_gc_wait);
 
 rw_attribute(rebalance_enabled);
 sysfs_pd_controller_attribute(rebalance);
-read_attribute(rebalance_work);
+read_attribute(rebalance_status);
 rw_attribute(promote_whole_extents);
 
 read_attribute(new_stripes);
@@ -216,7 +224,7 @@ read_attribute(new_stripes);
 read_attribute(io_timers_read);
 read_attribute(io_timers_write);
 
-read_attribute(data_jobs);
+read_attribute(moving_ctxts);
 
 #ifdef CONFIG_BCACHEFS_TESTS
 write_attribute(perf_test);
@@ -224,13 +232,13 @@ write_attribute(perf_test);
 
 #define x(_name)                                               \
        static struct attribute sysfs_time_stat_##_name =               \
-               { .name = #_name, .mode = S_IRUGO };
+               { .name = #_name, .mode = 0444 };
        BCH_TIME_STATS()
 #undef x
 
 static struct attribute sysfs_state_rw = {
        .name = "state",
-       .mode = S_IRUGO
+       .mode =  0444,
 };
 
 static size_t bch2_btree_cache_size(struct bch_fs *c)
@@ -240,135 +248,135 @@ static size_t bch2_btree_cache_size(struct bch_fs *c)
 
        mutex_lock(&c->btree_cache.lock);
        list_for_each_entry(b, &c->btree_cache.live, list)
-               ret += btree_bytes(c);
+               ret += btree_buf_bytes(b);
 
        mutex_unlock(&c->btree_cache.lock);
        return ret;
 }
 
-static size_t bch2_btree_avg_write_size(struct bch_fs *c)
-{
-       u64 nr = atomic64_read(&c->btree_writes_nr);
-       u64 sectors = atomic64_read(&c->btree_writes_sectors);
-
-       return nr ? div64_u64(sectors, nr) : 0;
-}
-
-static long data_progress_to_text(struct printbuf *out, struct bch_fs *c)
-{
-       long ret = 0;
-       struct bch_move_stats *stats;
-
-       mutex_lock(&c->data_progress_lock);
-       list_for_each_entry(stats, &c->data_progress_list, list) {
-               prt_printf(out, "%s: data type %s btree_id %s position: ",
-                      stats->name,
-                      bch2_data_types[stats->data_type],
-                      bch2_btree_ids[stats->btree_id]);
-               bch2_bpos_to_text(out, stats->pos);
-               prt_printf(out, "%s", "\n");
-       }
-
-       mutex_unlock(&c->data_progress_lock);
-       return ret;
-}
-
 static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c)
 {
-       struct btree_trans trans;
-       struct btree_iter iter;
-       struct bkey_s_c k;
+       struct btree_trans *trans;
        enum btree_id id;
-       u64 nr_uncompressed_extents = 0,
-           nr_compressed_extents = 0,
-           nr_incompressible_extents = 0,
-           uncompressed_sectors = 0,
-           incompressible_sectors = 0,
-           compressed_sectors_compressed = 0,
-           compressed_sectors_uncompressed = 0;
-       int ret;
+       struct compression_type_stats {
+               u64             nr_extents;
+               u64             sectors_compressed;
+               u64             sectors_uncompressed;
+       } s[BCH_COMPRESSION_TYPE_NR];
+       u64 compressed_incompressible = 0;
+       int ret = 0;
+
+       memset(s, 0, sizeof(s));
 
-       if (!test_bit(BCH_FS_STARTED, &c->flags))
+       if (!test_bit(BCH_FS_started, &c->flags))
                return -EPERM;
 
-       bch2_trans_init(&trans, c, 0, 0);
+       trans = bch2_trans_get(c);
 
        for (id = 0; id < BTREE_ID_NR; id++) {
-               if (!((1U << id) & BTREE_ID_HAS_PTRS))
+               if (!btree_type_has_ptrs(id))
                        continue;
 
-               for_each_btree_key(&trans, iter, id, POS_MIN,
-                                  BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
+               ret = for_each_btree_key(trans, iter, id, POS_MIN,
+                                        BTREE_ITER_ALL_SNAPSHOTS, k, ({
                        struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+                       struct bch_extent_crc_unpacked crc;
                        const union bch_extent_entry *entry;
-                       struct extent_ptr_decoded p;
-                       bool compressed = false, uncompressed = false, incompressible = false;
-
-                       bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
-                               switch (p.crc.compression_type) {
-                               case BCH_COMPRESSION_TYPE_none:
-                                       uncompressed = true;
-                                       uncompressed_sectors += k.k->size;
-                                       break;
-                               case BCH_COMPRESSION_TYPE_incompressible:
-                                       incompressible = true;
-                                       incompressible_sectors += k.k->size;
-                                       break;
-                               default:
-                                       compressed_sectors_compressed +=
-                                               p.crc.compressed_size;
-                                       compressed_sectors_uncompressed +=
-                                               p.crc.uncompressed_size;
-                                       compressed = true;
-                                       break;
+                       bool compressed = false, incompressible = false;
+
+                       bkey_for_each_crc(k.k, ptrs, crc, entry) {
+                               incompressible  |= crc.compression_type == BCH_COMPRESSION_TYPE_incompressible;
+                               compressed      |= crc_is_compressed(crc);
+
+                               if (crc_is_compressed(crc)) {
+                                       s[crc.compression_type].nr_extents++;
+                                       s[crc.compression_type].sectors_compressed += crc.compressed_size;
+                                       s[crc.compression_type].sectors_uncompressed += crc.uncompressed_size;
                                }
                        }
 
-                       if (incompressible)
-                               nr_incompressible_extents++;
-                       else if (uncompressed)
-                               nr_uncompressed_extents++;
-                       else if (compressed)
-                               nr_compressed_extents++;
-               }
-               bch2_trans_iter_exit(&trans, &iter);
+                       compressed_incompressible += compressed && incompressible;
+
+                       if (!compressed) {
+                               unsigned t = incompressible ? BCH_COMPRESSION_TYPE_incompressible : 0;
+
+                               s[t].nr_extents++;
+                               s[t].sectors_compressed += k.k->size;
+                               s[t].sectors_uncompressed += k.k->size;
+                       }
+                       0;
+               }));
        }
 
-       bch2_trans_exit(&trans);
+       bch2_trans_put(trans);
 
        if (ret)
                return ret;
 
-       prt_printf(out, "uncompressed:\n");
-       prt_printf(out, "       nr extents:             %llu\n", nr_uncompressed_extents);
-       prt_printf(out, "       size:                   ");
-       prt_human_readable_u64(out, uncompressed_sectors << 9);
-       prt_printf(out, "\n");
+       prt_str(out, "type");
+       printbuf_tabstop_push(out, 12);
+       prt_tab(out);
 
-       prt_printf(out, "compressed:\n");
-       prt_printf(out, "       nr extents:             %llu\n", nr_compressed_extents);
-       prt_printf(out, "       compressed size:        ");
-       prt_human_readable_u64(out, compressed_sectors_compressed << 9);
-       prt_printf(out, "\n");
-       prt_printf(out, "       uncompressed size:      ");
-       prt_human_readable_u64(out, compressed_sectors_uncompressed << 9);
-       prt_printf(out, "\n");
+       prt_str(out, "compressed");
+       printbuf_tabstop_push(out, 16);
+       prt_tab_rjust(out);
+
+       prt_str(out, "uncompressed");
+       printbuf_tabstop_push(out, 16);
+       prt_tab_rjust(out);
+
+       prt_str(out, "average extent size");
+       printbuf_tabstop_push(out, 24);
+       prt_tab_rjust(out);
+       prt_newline(out);
+
+       for (unsigned i = 0; i < ARRAY_SIZE(s); i++) {
+               bch2_prt_compression_type(out, i);
+               prt_tab(out);
+
+               prt_human_readable_u64(out, s[i].sectors_compressed << 9);
+               prt_tab_rjust(out);
+
+               prt_human_readable_u64(out, s[i].sectors_uncompressed << 9);
+               prt_tab_rjust(out);
+
+               prt_human_readable_u64(out, s[i].nr_extents
+                                      ? div_u64(s[i].sectors_uncompressed << 9, s[i].nr_extents)
+                                      : 0);
+               prt_tab_rjust(out);
+               prt_newline(out);
+       }
+
+       if (compressed_incompressible) {
+               prt_printf(out, "%llu compressed & incompressible extents", compressed_incompressible);
+               prt_newline(out);
+       }
 
-       prt_printf(out, "incompressible:\n");
-       prt_printf(out, "       nr extents:             %llu\n", nr_incompressible_extents);
-       prt_printf(out, "       size:                   ");
-       prt_human_readable_u64(out, incompressible_sectors << 9);
-       prt_printf(out, "\n");
        return 0;
 }
 
 static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
 {
-       prt_printf(out, "%s: ", bch2_btree_ids[c->gc_gens_btree]);
+       prt_printf(out, "%s: ", bch2_btree_id_str(c->gc_gens_btree));
        bch2_bpos_to_text(out, c->gc_gens_pos);
        prt_printf(out, "\n");
 }
 
+static void bch2_btree_wakeup_all(struct bch_fs *c)
+{
+       struct btree_trans *trans;
+
+       seqmutex_lock(&c->btree_trans_lock);
+       list_for_each_entry(trans, &c->btree_trans_list, list) {
+               struct btree_bkey_cached_common *b = READ_ONCE(trans->locking);
+
+               if (b)
+                       six_lock_wakeup_all(&b->lock);
+
+       }
+       seqmutex_unlock(&c->btree_trans_lock);
+}
+
 SHOW(bch2_fs)
 {
        struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
@@ -376,17 +384,13 @@ SHOW(bch2_fs)
        sysfs_print(minor,                      c->minor);
        sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 
+       if (attr == &sysfs_flags)
+               prt_bitflags(out, bch2_fs_flag_strs, c->flags);
+
        sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
-       sysfs_hprint(btree_avg_write_size,      bch2_btree_avg_write_size(c));
 
-       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_print(bucket_alloc_fail,
-                   atomic_long_read(&c->bucket_alloc_fail));
+       if (attr == &sysfs_btree_write_stats)
+               bch2_btree_write_stats_to_text(out, c);
 
        sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 
@@ -397,12 +401,12 @@ SHOW(bch2_fs)
 
        sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
        sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
-       sysfs_hprint(copy_gc_wait,
-                    max(0LL, c->copygc_wait -
-                        atomic64_read(&c->io_clock[WRITE].now)) << 9);
 
-       if (attr == &sysfs_rebalance_work)
-               bch2_rebalance_work_to_text(out, c);
+       if (attr == &sysfs_copy_gc_wait)
+               bch2_copygc_wait_to_text(out, c);
+
+       if (attr == &sysfs_rebalance_status)
+               bch2_rebalance_status_to_text(out, c);
 
        sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 
@@ -420,15 +424,18 @@ SHOW(bch2_fs)
        if (attr == &sysfs_btree_key_cache)
                bch2_btree_key_cache_to_text(out, &c->btree_key_cache);
 
-       if (attr == &sysfs_btree_transactions)
-               bch2_btree_trans_to_text(out, c);
-
        if (attr == &sysfs_stripes_heap)
                bch2_stripes_heap_to_text(out, c);
 
        if (attr == &sysfs_open_buckets)
                bch2_open_buckets_to_text(out, c);
 
+       if (attr == &sysfs_open_buckets_partial)
+               bch2_open_buckets_partial_to_text(out, c);
+
+       if (attr == &sysfs_write_points)
+               bch2_write_points_to_text(out, c);
+
        if (attr == &sysfs_compression_stats)
                bch2_compression_stats_to_text(out, c);
 
@@ -441,8 +448,19 @@ SHOW(bch2_fs)
        if (attr == &sysfs_io_timers_write)
                bch2_io_timers_to_text(out, &c->io_clock[WRITE]);
 
-       if (attr == &sysfs_data_jobs)
-               data_progress_to_text(out, c);
+       if (attr == &sysfs_moving_ctxts)
+               bch2_fs_moving_ctxts_to_text(out, c);
+
+#ifdef BCH_WRITE_REF_DEBUG
+       if (attr == &sysfs_write_refs)
+               bch2_write_refs_to_text(out, c);
+#endif
+
+       if (attr == &sysfs_nocow_lock_table)
+               bch2_nocow_locks_to_text(out, &c->nocow_locks);
+
+       if (attr == &sysfs_disk_groups)
+               bch2_disk_groups_to_text(out, c);
 
        return 0;
 }
@@ -482,12 +500,12 @@ STORE(bch2_fs)
 
        /* Debugging: */
 
-       if (!test_bit(BCH_FS_STARTED, &c->flags))
+       if (!test_bit(BCH_FS_started, &c->flags))
                return -EPERM;
 
        /* Debugging: */
 
-       if (!test_bit(BCH_FS_RW, &c->flags))
+       if (!test_bit(BCH_FS_rw, &c->flags))
                return -EROFS;
 
        if (attr == &sysfs_prune_cache) {
@@ -495,9 +513,12 @@ STORE(bch2_fs)
 
                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);
        }
 
+       if (attr == &sysfs_btree_wakeup)
+               bch2_btree_wakeup_all(c);
+
        if (attr == &sysfs_trigger_gc) {
                /*
                 * Full gc is currently incompatible with btree key cache:
@@ -544,7 +565,7 @@ SYSFS_OPS(bch2_fs);
 struct attribute *bch2_fs_files[] = {
        &sysfs_minor,
        &sysfs_btree_cache_size,
-       &sysfs_btree_avg_write_size,
+       &sysfs_btree_write_stats,
 
        &sysfs_promote_whole_extents,
 
@@ -564,19 +585,20 @@ SHOW(bch2_fs_counters)
        u64 counter = 0;
        u64 counter_since_mount = 0;
 
-       out->tabstops[0] = 32;
+       printbuf_tabstop_push(out, 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];\
                        prt_printf(out, "since mount:");                                \
                        prt_tab(out);                                           \
-                       prt_human_readable_u64(out, counter_since_mount << 9);  \
+                       prt_human_readable_u64(out, counter_since_mount);       \
                        prt_newline(out);                                       \
                                                                                \
                        prt_printf(out, "since filesystem creation:");          \
                        prt_tab(out);                                           \
-                       prt_human_readable_u64(out, counter << 9);              \
+                       prt_human_readable_u64(out, counter);                   \
                        prt_newline(out);                                       \
                }
        BCH_PERSISTENT_COUNTERS()
@@ -602,25 +624,33 @@ struct attribute *bch2_fs_counters_files[] = {
 SHOW(bch2_fs_internal)
 {
        struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
+
        return bch2_fs_to_text(out, &c->kobj, attr);
 }
 
 STORE(bch2_fs_internal)
 {
        struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
+
        return bch2_fs_store(&c->kobj, attr, buf, size);
 }
 SYSFS_OPS(bch2_fs_internal);
 
 struct attribute *bch2_fs_internal_files[] = {
+       &sysfs_flags,
        &sysfs_journal_debug,
        &sysfs_btree_updates,
        &sysfs_btree_cache,
        &sysfs_btree_key_cache,
-       &sysfs_btree_transactions,
        &sysfs_new_stripes,
        &sysfs_stripes_heap,
        &sysfs_open_buckets,
+       &sysfs_open_buckets_partial,
+       &sysfs_write_points,
+#ifdef BCH_WRITE_REF_DEBUG
+       &sysfs_write_refs,
+#endif
+       &sysfs_nocow_lock_table,
        &sysfs_io_timers_read,
        &sysfs_io_timers_write,
 
@@ -628,11 +658,7 @@ struct attribute *bch2_fs_internal_files[] = {
        &sysfs_trigger_discards,
        &sysfs_trigger_invalidates,
        &sysfs_prune_cache,
-
-       &sysfs_read_realloc_races,
-       &sysfs_extent_migrate_done,
-       &sysfs_extent_migrate_raced,
-       &sysfs_bucket_alloc_fail,
+       &sysfs_btree_wakeup,
 
        &sysfs_gc_gens_pos,
 
@@ -640,12 +666,14 @@ struct attribute *bch2_fs_internal_files[] = {
        &sysfs_copy_gc_wait,
 
        &sysfs_rebalance_enabled,
-       &sysfs_rebalance_work,
+       &sysfs_rebalance_status,
        sysfs_pd_controller_files(rebalance),
 
-       &sysfs_data_jobs,
+       &sysfs_moving_ctxts,
 
        &sysfs_internal_uuid,
+
+       &sysfs_disk_groups,
        NULL
 };
 
@@ -676,7 +704,7 @@ STORE(bch2_fs_opts_dir)
         * We don't need to take c->writes for correctness, but it eliminates an
         * unsightly error message in the dmesg log when we're RO:
         */
-       if (unlikely(!percpu_ref_tryget(&c->writes)))
+       if (unlikely(!bch2_write_ref_tryget(c, BCH_WRITE_REF_sysfs)))
                return -EROFS;
 
        tmp = kstrdup(buf, GFP_KERNEL);
@@ -698,15 +726,15 @@ STORE(bch2_fs_opts_dir)
        bch2_opt_set_sb(c, opt, v);
        bch2_opt_set_by_id(&c->opts, id, v);
 
-       if ((id == Opt_background_target ||
-            id == Opt_background_compression) && v) {
-               bch2_rebalance_add_work(c, S64_MAX);
-               rebalance_wakeup(c);
-       }
+       if (v &&
+           (id == Opt_background_target ||
+            id == Opt_background_compression ||
+            (id == Opt_compression && !c->opts.background_compression)))
+               bch2_set_rebalance_needs_scan(c, 0);
 
        ret = size;
 err:
-       percpu_ref_put(&c->writes);
+       bch2_write_ref_put(c, BCH_WRITE_REF_sysfs);
        return ret;
 }
 SYSFS_OPS(bch2_fs_opts_dir);
@@ -772,38 +800,75 @@ static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
        for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
                nr[c->open_buckets[i].data_type]++;
 
-       prt_printf(out,
-              "\t\t\t buckets\t sectors      fragmented\n"
-              "capacity\t%16llu\n",
-              ca->mi.nbuckets - ca->mi.first_bucket);
-
-       for (i = 0; i < BCH_DATA_NR; i++)
-               prt_printf(out, "%-16s%16llu%16llu%16llu\n",
-                      bch2_data_types[i], stats.d[i].buckets,
-                      stats.d[i].sectors, stats.d[i].fragmented);
-
-       prt_printf(out,
-              "ec\t\t%16llu\n"
-              "\n"
-              "freelist_wait\t\t%s\n"
-              "open buckets allocated\t%u\n"
-              "open buckets this dev\t%u\n"
-              "open buckets total\t%u\n"
-              "open_buckets_wait\t%s\n"
-              "open_buckets_btree\t%u\n"
-              "open_buckets_user\t%u\n"
-              "buckets_to_invalidate\t%llu\n"
-              "btree reserve cache\t%u\n",
-              stats.buckets_ec,
-              c->freelist_wait.list.first              ? "waiting" : "empty",
-              OPEN_BUCKETS_COUNT - c->open_buckets_nr_free,
-              ca->nr_open_buckets,
-              OPEN_BUCKETS_COUNT,
-              c->open_buckets_wait.list.first          ? "waiting" : "empty",
-              nr[BCH_DATA_btree],
-              nr[BCH_DATA_user],
-              should_invalidate_buckets(ca, stats),
-              c->btree_reserve_cache_nr);
+       printbuf_tabstop_push(out, 8);
+       printbuf_tabstop_push(out, 16);
+       printbuf_tabstop_push(out, 16);
+       printbuf_tabstop_push(out, 16);
+       printbuf_tabstop_push(out, 16);
+
+       bch2_dev_usage_to_text(out, &stats);
+
+       prt_newline(out);
+
+       prt_printf(out, "reserves:");
+       prt_newline(out);
+       for (i = 0; i < BCH_WATERMARK_NR; i++) {
+               prt_str(out, bch2_watermarks[i]);
+               prt_tab(out);
+               prt_u64(out, bch2_dev_buckets_reserved(ca, i));
+               prt_tab_rjust(out);
+               prt_newline(out);
+       }
+
+       prt_newline(out);
+
+       printbuf_tabstops_reset(out);
+       printbuf_tabstop_push(out, 24);
+
+       prt_str(out, "freelist_wait");
+       prt_tab(out);
+       prt_str(out, c->freelist_wait.list.first ? "waiting" : "empty");
+       prt_newline(out);
+
+       prt_str(out, "open buckets allocated");
+       prt_tab(out);
+       prt_u64(out, OPEN_BUCKETS_COUNT - c->open_buckets_nr_free);
+       prt_newline(out);
+
+       prt_str(out, "open buckets this dev");
+       prt_tab(out);
+       prt_u64(out, ca->nr_open_buckets);
+       prt_newline(out);
+
+       prt_str(out, "open buckets total");
+       prt_tab(out);
+       prt_u64(out, OPEN_BUCKETS_COUNT);
+       prt_newline(out);
+
+       prt_str(out, "open_buckets_wait");
+       prt_tab(out);
+       prt_str(out, c->open_buckets_wait.list.first ? "waiting" : "empty");
+       prt_newline(out);
+
+       prt_str(out, "open_buckets_btree");
+       prt_tab(out);
+       prt_u64(out, nr[BCH_DATA_btree]);
+       prt_newline(out);
+
+       prt_str(out, "open_buckets_user");
+       prt_tab(out);
+       prt_u64(out, nr[BCH_DATA_user]);
+       prt_newline(out);
+
+       prt_str(out, "buckets_to_invalidate");
+       prt_tab(out);
+       prt_u64(out, should_invalidate_buckets(ca, stats));
+       prt_newline(out);
+
+       prt_str(out, "btree reserve cache");
+       prt_tab(out);
+       prt_u64(out, c->btree_reserve_cache_nr);
+       prt_newline(out);
 }
 
 static const char * const bch2_rw[] = {
@@ -812,7 +877,7 @@ static const char * const bch2_rw[] = {
        NULL
 };
 
-static void dev_iodone_to_text(struct printbuf *out, struct bch_dev *ca)
+static void dev_io_done_to_text(struct printbuf *out, struct bch_dev *ca)
 {
        int rw, i;
 
@@ -821,7 +886,7 @@ static void dev_iodone_to_text(struct printbuf *out, struct bch_dev *ca)
 
                for (i = 1; i < BCH_DATA_NR; i++)
                        prt_printf(out, "%-12s:%12llu\n",
-                              bch2_data_types[i],
+                              bch2_data_type_str(i),
                               percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
        }
 }
@@ -840,18 +905,13 @@ SHOW(bch2_dev)
        sysfs_print(discard,            ca->mi.discard);
 
        if (attr == &sysfs_label) {
-               if (ca->mi.group) {
-                       mutex_lock(&c->sb_lock);
-                       bch2_disk_path_to_text(out, c->disk_sb.sb,
-                                              ca->mi.group - 1);
-                       mutex_unlock(&c->sb_lock);
-               }
-
+               if (ca->mi.group)
+                       bch2_disk_path_to_text(out, c, ca->mi.group - 1);
                prt_char(out, '\n');
        }
 
        if (attr == &sysfs_has_data) {
-               prt_bitflags(out, bch2_data_types, bch2_dev_has_data(c, ca));
+               prt_bitflags(out, __bch2_data_types, bch2_dev_has_data(c, ca));
                prt_char(out, '\n');
        }
 
@@ -860,17 +920,20 @@ SHOW(bch2_dev)
                prt_char(out, '\n');
        }
 
-       if (attr == &sysfs_iodone)
-               dev_iodone_to_text(out, ca);
+       if (attr == &sysfs_io_done)
+               dev_io_done_to_text(out, ca);
+
+       if (attr == &sysfs_io_errors)
+               bch2_dev_io_errors_to_text(out, ca);
 
        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)
-               bch2_time_stats_to_text(out, &ca->io_latency[READ]);
+               bch2_time_stats_to_text(out, &ca->io_latency[READ].stats);
 
        if (attr == &sysfs_io_latency_stats_write)
-               bch2_time_stats_to_text(out, &ca->io_latency[WRITE]);
+               bch2_time_stats_to_text(out, &ca->io_latency[WRITE].stats);
 
        sysfs_printf(congested,                 "%u%%",
                     clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
@@ -892,7 +955,7 @@ STORE(bch2_dev)
                bool v = strtoul_or_return(buf);
 
                mutex_lock(&c->sb_lock);
-               mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
+               mi = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
 
                if (v != BCH_MEMBER_DISCARD(mi)) {
                        SET_BCH_MEMBER_DISCARD(mi, v);
@@ -901,6 +964,19 @@ STORE(bch2_dev)
                mutex_unlock(&c->sb_lock);
        }
 
+       if (attr == &sysfs_durability) {
+               u64 v = strtoul_or_return(buf);
+
+               mutex_lock(&c->sb_lock);
+               mi = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
+
+               if (v + 1 != BCH_MEMBER_DURABILITY(mi)) {
+                       SET_BCH_MEMBER_DURABILITY(mi, v + 1);
+                       bch2_write_super(c);
+               }
+               mutex_unlock(&c->sb_lock);
+       }
+
        if (attr == &sysfs_label) {
                char *tmp;
                int ret;
@@ -915,6 +991,9 @@ STORE(bch2_dev)
                        return ret;
        }
 
+       if (attr == &sysfs_io_errors_reset)
+               bch2_dev_errors_reset(ca);
+
        return size;
 }
 SYSFS_OPS(bch2_dev);
@@ -932,7 +1011,9 @@ struct attribute *bch2_dev_files[] = {
        &sysfs_label,
 
        &sysfs_has_data,
-       &sysfs_iodone,
+       &sysfs_io_done,
+       &sysfs_io_errors,
+       &sysfs_io_errors_reset,
 
        &sysfs_io_latency_read,
        &sysfs_io_latency_write,