]> git.sesse.net Git - bcachefs-tools-debian/blobdiff - libbcachefs/sysfs.c
Move c_src dirs back to toplevel
[bcachefs-tools-debian] / libbcachefs / sysfs.c
diff --git a/libbcachefs/sysfs.c b/libbcachefs/sysfs.c
new file mode 100644 (file)
index 0000000..8ed5231
--- /dev/null
@@ -0,0 +1,1026 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * bcache sysfs interfaces
+ *
+ * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
+ * Copyright 2012 Google, Inc.
+ */
+
+#ifndef NO_BCACHEFS_SYSFS
+
+#include "bcachefs.h"
+#include "alloc_background.h"
+#include "alloc_foreground.h"
+#include "sysfs.h"
+#include "btree_cache.h"
+#include "btree_io.h"
+#include "btree_iter.h"
+#include "btree_key_cache.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 "movinggc.h"
+#include "nocow_locking.h"
+#include "opts.h"
+#include "rebalance.h"
+#include "replicas.h"
+#include "super-io.h"
+#include "tests.h"
+
+#include <linux/blkdev.h>
+#include <linux/sort.h>
+#include <linux/sched/clock.h>
+
+#include "util.h"
+
+#define SYSFS_OPS(type)                                                        \
+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 *);    \
+                                                                       \
+static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
+                          char *buf)                                   \
+{                                                                      \
+       struct printbuf out = PRINTBUF;                                 \
+       ssize_t ret = fn ## _to_text(&out, kobj, attr);                 \
+                                                                       \
+       if (out.pos && out.buf[out.pos - 1] != '\n')                    \
+               prt_newline(&out);                                      \
+                                                                       \
+       if (!ret && out.allocation_failure)                             \
+               ret = -ENOMEM;                                          \
+                                                                       \
+       if (!ret) {                                                     \
+               ret = min_t(size_t, out.pos, PAGE_SIZE - 1);            \
+               memcpy(buf, out.buf, ret);                              \
+       }                                                               \
+       printbuf_exit(&out);                                            \
+       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, 0200)
+#define read_attribute(n)      __sysfs_attribute(n, 0444)
+#define rw_attribute(n)                __sysfs_attribute(n, 0644)
+
+#define sysfs_printf(file, fmt, ...)                                   \
+do {                                                                   \
+       if (attr == &sysfs_ ## file)                                    \
+               prt_printf(out, fmt "\n", __VA_ARGS__);                 \
+} while (0)
+
+#define sysfs_print(file, var)                                         \
+do {                                                                   \
+       if (attr == &sysfs_ ## file)                                    \
+               snprint(out, var);                                      \
+} while (0)
+
+#define sysfs_hprint(file, val)                                                \
+do {                                                                   \
+       if (attr == &sysfs_ ## file)                                    \
+               prt_human_readable_s64(out, val);                       \
+} while (0)
+
+#define sysfs_strtoul(file, var)                                       \
+do {                                                                   \
+       if (attr == &sysfs_ ## file)                                    \
+               return strtoul_safe(buf, var) ?: (ssize_t) size;        \
+} while (0)
+
+#define sysfs_strtoul_clamp(file, var, min, max)                       \
+do {                                                                   \
+       if (attr == &sysfs_ ## file)                                    \
+               return strtoul_safe_clamp(buf, var, min, max)           \
+                       ?: (ssize_t) size;                              \
+} while (0)
+
+#define strtoul_or_return(cp)                                          \
+({                                                                     \
+       unsigned long _v;                                               \
+       int _r = kstrtoul(cp, 10, &_v);                                 \
+       if (_r)                                                         \
+               return _r;                                              \
+       _v;                                                             \
+})
+
+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);
+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);
+read_attribute(io_latency_stats_read);
+read_attribute(io_latency_stats_write);
+read_attribute(congested);
+
+read_attribute(btree_write_stats);
+
+read_attribute(btree_cache_size);
+read_attribute(compression_stats);
+read_attribute(journal_debug);
+read_attribute(btree_updates);
+read_attribute(btree_cache);
+read_attribute(btree_key_cache);
+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);
+
+#define x(t, n, ...) read_attribute(t);
+BCH_PERSISTENT_COUNTERS()
+#undef x
+
+rw_attribute(discard);
+rw_attribute(label);
+
+rw_attribute(copy_gc_enabled);
+read_attribute(copy_gc_wait);
+
+rw_attribute(rebalance_enabled);
+sysfs_pd_controller_attribute(rebalance);
+read_attribute(rebalance_status);
+rw_attribute(promote_whole_extents);
+
+read_attribute(new_stripes);
+
+read_attribute(io_timers_read);
+read_attribute(io_timers_write);
+
+read_attribute(moving_ctxts);
+
+#ifdef CONFIG_BCACHEFS_TESTS
+write_attribute(perf_test);
+#endif /* CONFIG_BCACHEFS_TESTS */
+
+#define x(_name)                                               \
+       static struct attribute sysfs_time_stat_##_name =               \
+               { .name = #_name, .mode = 0444 };
+       BCH_TIME_STATS()
+#undef x
+
+static struct attribute sysfs_state_rw = {
+       .name = "state",
+       .mode =  0444,
+};
+
+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.live, list)
+               ret += btree_bytes(c);
+
+       mutex_unlock(&c->btree_cache.lock);
+       return ret;
+}
+
+static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c)
+{
+       struct btree_trans *trans;
+       enum btree_id id;
+       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))
+               return -EPERM;
+
+       trans = bch2_trans_get(c);
+
+       for (id = 0; id < BTREE_ID_NR; id++) {
+               if (!btree_type_has_ptrs(id))
+                       continue;
+
+               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;
+                       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;
+                               }
+                       }
+
+                       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_put(trans);
+
+       if (ret)
+               return ret;
+
+       prt_str(out, "type");
+       printbuf_tabstop_push(out, 12);
+       prt_tab(out);
+
+       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++) {
+               prt_str(out, bch2_compression_types[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);
+       }
+
+       return 0;
+}
+
+static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
+{
+       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);
+
+       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));
+
+       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);
+
+       if (attr == &sysfs_gc_gens_pos)
+               bch2_gc_gens_pos_to_text(out, c);
+
+       sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
+
+       sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
+       sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
+
+       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);
+
+       /* Debugging: */
+
+       if (attr == &sysfs_journal_debug)
+               bch2_journal_debug_to_text(out, &c->journal);
+
+       if (attr == &sysfs_btree_updates)
+               bch2_btree_updates_to_text(out, c);
+
+       if (attr == &sysfs_btree_cache)
+               bch2_btree_cache_to_text(out, c);
+
+       if (attr == &sysfs_btree_key_cache)
+               bch2_btree_key_cache_to_text(out, &c->btree_key_cache);
+
+       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);
+
+       if (attr == &sysfs_new_stripes)
+               bch2_new_stripes_to_text(out, c);
+
+       if (attr == &sysfs_io_timers_read)
+               bch2_io_timers_to_text(out, &c->io_clock[READ]);
+
+       if (attr == &sysfs_io_timers_write)
+               bch2_io_timers_to_text(out, &c->io_clock[WRITE]);
+
+       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;
+}
+
+STORE(bch2_fs)
+{
+       struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
+
+       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) {
+               ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
+                       ?: (ssize_t) size;
+
+               if (c->copygc_thread)
+                       wake_up_process(c->copygc_thread);
+               return ret;
+       }
+
+       if (attr == &sysfs_rebalance_enabled) {
+               ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
+                       ?: (ssize_t) size;
+
+               rebalance_wakeup(c);
+               return ret;
+       }
+
+       sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
+
+       sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
+
+       /* Debugging: */
+
+       if (!test_bit(BCH_FS_started, &c->flags))
+               return -EPERM;
+
+       /* Debugging: */
+
+       if (!test_bit(BCH_FS_rw, &c->flags))
+               return -EROFS;
+
+       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);
+       }
+
+       if (attr == &sysfs_btree_wakeup)
+               bch2_btree_wakeup_all(c);
+
+       if (attr == &sysfs_trigger_gc) {
+               /*
+                * Full gc is currently incompatible with btree key cache:
+                */
+#if 0
+               down_read(&c->state_lock);
+               bch2_gc(c, false, false);
+               up_read(&c->state_lock);
+#else
+               bch2_gc_gens(c);
+#endif
+       }
+
+       if (attr == &sysfs_trigger_discards)
+               bch2_do_discards(c);
+
+       if (attr == &sysfs_trigger_invalidates)
+               bch2_do_invalidates(c);
+
+#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)))
+                       ret = bch2_btree_perf_test(c, test, nr, threads);
+               kfree(tmp);
+
+               if (ret)
+                       size = ret;
+       }
+#endif
+       return size;
+}
+SYSFS_OPS(bch2_fs);
+
+struct attribute *bch2_fs_files[] = {
+       &sysfs_minor,
+       &sysfs_btree_cache_size,
+       &sysfs_btree_write_stats,
+
+       &sysfs_promote_whole_extents,
+
+       &sysfs_compression_stats,
+
+#ifdef CONFIG_BCACHEFS_TESTS
+       &sysfs_perf_test,
+#endif
+       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;
+
+       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);       \
+                       prt_newline(out);                                       \
+                                                                               \
+                       prt_printf(out, "since filesystem creation:");          \
+                       prt_tab(out);                                           \
+                       prt_human_readable_u64(out, counter);                   \
+                       prt_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)
+{
+       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_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,
+
+       &sysfs_trigger_gc,
+       &sysfs_trigger_discards,
+       &sysfs_trigger_invalidates,
+       &sysfs_prune_cache,
+       &sysfs_btree_wakeup,
+
+       &sysfs_gc_gens_pos,
+
+       &sysfs_copy_gc_enabled,
+       &sysfs_copy_gc_wait,
+
+       &sysfs_rebalance_enabled,
+       &sysfs_rebalance_status,
+       sysfs_pd_controller_files(rebalance),
+
+       &sysfs_moving_ctxts,
+
+       &sysfs_internal_uuid,
+
+       &sysfs_disk_groups,
+       NULL
+};
+
+/* options */
+
+SHOW(bch2_fs_opts_dir)
+{
+       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, c->disk_sb.sb, opt, v, OPT_SHOW_FULL_LIST);
+       prt_char(out, '\n');
+
+       return 0;
+}
+
+STORE(bch2_fs_opts_dir)
+{
+       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 ret, id = opt - bch2_opt_table;
+       char *tmp;
+       u64 v;
+
+       /*
+        * 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(!bch2_write_ref_tryget(c, BCH_WRITE_REF_sysfs)))
+               return -EROFS;
+
+       tmp = kstrdup(buf, GFP_KERNEL);
+       if (!tmp) {
+               ret = -ENOMEM;
+               goto err;
+       }
+
+       ret = bch2_opt_parse(c, opt, strim(tmp), &v, NULL);
+       kfree(tmp);
+
+       if (ret < 0)
+               goto err;
+
+       ret = bch2_opt_check_may_set(c, id, v);
+       if (ret < 0)
+               goto err;
+
+       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_set_rebalance_needs_scan(c, 0);
+
+       ret = size;
+err:
+       bch2_write_ref_put(c, BCH_WRITE_REF_sysfs);
+       return ret;
+}
+SYSFS_OPS(bch2_fs_opts_dir);
+
+struct attribute *bch2_fs_opts_dir_files[] = { NULL };
+
+int bch2_opts_create_sysfs_files(struct kobject *kobj)
+{
+       const struct bch_option *i;
+       int ret;
+
+       for (i = bch2_opt_table;
+            i < bch2_opt_table + bch2_opts_nr;
+            i++) {
+               if (!(i->flags & OPT_FS))
+                       continue;
+
+               ret = sysfs_create_file(kobj, &i->attr);
+               if (ret)
+                       return ret;
+       }
+
+       return 0;
+}
+
+/* time stats */
+
+SHOW(bch2_fs_time_stats)
+{
+       struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
+
+#define x(name)                                                                \
+       if (attr == &sysfs_time_stat_##name)                            \
+               bch2_time_stats_to_text(out, &c->times[BCH_TIME_##name]);
+       BCH_TIME_STATS()
+#undef x
+
+       return 0;
+}
+
+STORE(bch2_fs_time_stats)
+{
+       return size;
+}
+SYSFS_OPS(bch2_fs_time_stats);
+
+struct attribute *bch2_fs_time_stats_files[] = {
+#define x(name)                                                \
+       &sysfs_time_stat_##name,
+       BCH_TIME_STATS()
+#undef x
+       NULL
+};
+
+static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
+{
+       struct bch_fs *c = ca->fs;
+       struct bch_dev_usage stats = bch2_dev_usage_read(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].data_type]++;
+
+       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[] = {
+       "read",
+       "write",
+       NULL
+};
+
+static void dev_io_done_to_text(struct printbuf *out, struct bch_dev *ca)
+{
+       int rw, i;
+
+       for (rw = 0; rw < 2; rw++) {
+               prt_printf(out, "%s:\n", bch2_rw[rw]);
+
+               for (i = 1; i < BCH_DATA_NR; i++)
+                       prt_printf(out, "%-12s:%12llu\n",
+                              bch2_data_types[i],
+                              percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
+       }
+}
+
+SHOW(bch2_dev)
+{
+       struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
+       struct bch_fs *c = ca->fs;
+
+       sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
+
+       sysfs_print(bucket_size,        bucket_bytes(ca));
+       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);
+
+       if (attr == &sysfs_label) {
+               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_char(out, '\n');
+       }
+
+       if (attr == &sysfs_state_rw) {
+               prt_string_option(out, bch2_member_states, ca->mi.state);
+               prt_char(out, '\n');
+       }
+
+       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]);
+
+       if (attr == &sysfs_io_latency_stats_write)
+               bch2_time_stats_to_text(out, &ca->io_latency[WRITE]);
+
+       sysfs_printf(congested,                 "%u%%",
+                    clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
+                    * 100 / CONGESTED_MAX);
+
+       if (attr == &sysfs_alloc_debug)
+               dev_alloc_debug_to_text(out, ca);
+
+       return 0;
+}
+
+STORE(bch2_dev)
+{
+       struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
+       struct bch_fs *c = ca->fs;
+       struct bch_member *mi;
+
+       if (attr == &sysfs_discard) {
+               bool 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 != BCH_MEMBER_DISCARD(mi)) {
+                       SET_BCH_MEMBER_DISCARD(mi, v);
+                       bch2_write_super(c);
+               }
+               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;
+
+               tmp = kstrdup(buf, GFP_KERNEL);
+               if (!tmp)
+                       return -ENOMEM;
+
+               ret = bch2_dev_group_set(c, ca, strim(tmp));
+               kfree(tmp);
+               if (ret)
+                       return ret;
+       }
+
+       if (attr == &sysfs_io_errors_reset)
+               bch2_dev_errors_reset(ca);
+
+       return size;
+}
+SYSFS_OPS(bch2_dev);
+
+struct attribute *bch2_dev_files[] = {
+       &sysfs_uuid,
+       &sysfs_bucket_size,
+       &sysfs_first_bucket,
+       &sysfs_nbuckets,
+       &sysfs_durability,
+
+       /* settings: */
+       &sysfs_discard,
+       &sysfs_state_rw,
+       &sysfs_label,
+
+       &sysfs_has_data,
+       &sysfs_io_done,
+       &sysfs_io_errors,
+       &sysfs_io_errors_reset,
+
+       &sysfs_io_latency_read,
+       &sysfs_io_latency_write,
+       &sysfs_io_latency_stats_read,
+       &sysfs_io_latency_stats_write,
+       &sysfs_congested,
+
+       /* debug: */
+       &sysfs_alloc_debug,
+       NULL
+};
+
+#endif  /* _BCACHEFS_SYSFS_H_ */