X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fsysfs.c;h=4fc5777ecfb09d8959db36a0fa3c60719b2ddd13;hb=9fce394ca6d0082ced3612a627cd16e06d84244a;hp=edfa85b0561461cbeb89edbf088c4a55b3afb768;hpb=f9395eeca59290b210bc2b79f7bf2e9cb779cf3f;p=bcachefs-tools-debian diff --git a/libbcachefs/sysfs.c b/libbcachefs/sysfs.c index edfa85b..4fc5777 100644 --- a/libbcachefs/sysfs.c +++ b/libbcachefs/sysfs.c @@ -1,3 +1,4 @@ +// SPDX-License-Identifier: GPL-2.0 /* * bcache sysfs interfaces * @@ -5,22 +6,31 @@ * 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_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 "opts.h" +#include "rebalance.h" +#include "replicas.h" #include "super-io.h" -#include "tier.h" +#include "tests.h" #include #include @@ -53,7 +63,7 @@ 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) \ @@ -65,9 +75,9 @@ do { \ #define sysfs_hprint(file, val) \ do { \ if (attr == &sysfs_ ## file) { \ - ssize_t ret = bch2_hprint(buf, val); \ - strcat(buf, "\n"); \ - return ret + 1; \ + bch2_hprint(&out, val); \ + pr_buf(&out, "\n"); \ + return out.pos - buf; \ } \ } while (0) @@ -134,79 +144,76 @@ 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(btree_cache); +read_attribute(btree_key_cache); +read_attribute(btree_transactions); +read_attribute(stripes_heap); 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); -#define BCH_DEBUG_PARAM(name, description) \ - rw_attribute(name); +read_attribute(io_timers_read); +read_attribute(io_timers_write); - BCH_DEBUG_PARAMS() -#undef BCH_DEBUG_PARAM +#ifdef CONFIG_BCACHEFS_TESTS +write_attribute(perf_test); +#endif /* CONFIG_BCACHEFS_TESTS */ -#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", @@ -218,95 +225,92 @@ 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) +static int fs_alloc_debug_to_text(struct printbuf *out, struct bch_fs *c) { - struct bch_fs_usage stats = 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); + struct bch_fs_usage *fs_usage = bch2_fs_usage_read(c); + + if (!fs_usage) + return -ENOMEM; + + bch2_fs_usage_to_text(out, c, fs_usage); + + percpu_up_read(&c->mark_lock); + + kfree(fs_usage); + return 0; } -static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf) +static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c) { - 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, 0, 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, - "uncompressed data:\n" - " nr extents: %llu\n" - " size (bytes): %llu\n" - "compressed data:\n" - " nr extents: %llu\n" - " compressed size (bytes): %llu\n" - " uncompressed size (bytes): %llu\n", - nr_uncompressed_extents, - uncompressed_sectors << 9, - nr_compressed_extents, - compressed_sectors_compressed << 9, - compressed_sectors_uncompressed << 9); + + ret = bch2_trans_exit(&trans) ?: ret; + if (ret) + return ret; + + pr_buf(out, + "uncompressed data:\n" + " nr extents: %llu\n" + " size (bytes): %llu\n" + "compressed data:\n" + " nr extents: %llu\n" + " compressed size (bytes): %llu\n" + " uncompressed size (bytes): %llu\n", + nr_uncompressed_extents, + uncompressed_sectors << 9, + nr_compressed_extents, + compressed_sectors_compressed << 9, + compressed_sectors_uncompressed << 9); + return 0; } SHOW(bch2_fs) { struct bch_fs *c = container_of(kobj, struct bch_fs, kobj); + struct printbuf out = _PBUF(buf, PAGE_SIZE); sysfs_print(minor, c->minor); sysfs_printf(internal_uuid, "%pU", c->sb.uuid.b); @@ -320,57 +324,106 @@ 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 */ + sysfs_pd_controller_show(copy_gc, &c->copygc_pd); - sysfs_pd_controller_show(tiering, &c->tiers[1].pd); /* XXX */ + if (attr == &sysfs_rebalance_work) { + bch2_rebalance_work_to_text(&out, c); + return out.pos - buf; + } + + 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: */ if (attr == &sysfs_alloc_debug) - return show_fs_alloc_debug(c, buf); + return fs_alloc_debug_to_text(&out, c) ?: out.pos - buf; + + if (attr == &sysfs_journal_debug) { + bch2_journal_debug_to_text(&out, &c->journal); + return out.pos - buf; + } + + if (attr == &sysfs_journal_pins) { + bch2_journal_pins_to_text(&out, &c->journal); + return out.pos - buf; + } + + if (attr == &sysfs_btree_updates) { + bch2_btree_updates_to_text(&out, c); + return out.pos - buf; + } + + if (attr == &sysfs_dirty_btree_nodes) { + bch2_dirty_btree_nodes_to_text(&out, c); + return out.pos - buf; + } - if (attr == &sysfs_journal_debug) - return bch2_journal_print_debug(&c->journal, buf); + if (attr == &sysfs_btree_cache) { + bch2_btree_cache_to_text(&out, c); + return out.pos - buf; + } - if (attr == &sysfs_journal_pins) - return bch2_journal_print_pins(&c->journal, buf); + if (attr == &sysfs_btree_key_cache) { + bch2_btree_key_cache_to_text(&out, &c->btree_key_cache); + return out.pos - buf; + } - if (attr == &sysfs_compression_stats) - return bch2_compression_stats(c, buf); + if (attr == &sysfs_btree_transactions) { + bch2_btree_trans_to_text(&out, c); + return out.pos - buf; + } -#define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name); - BCH_DEBUG_PARAMS() -#undef BCH_DEBUG_PARAM + if (attr == &sysfs_stripes_heap) { + bch2_stripes_heap_to_text(&out, c); + return out.pos - buf; + } + + if (attr == &sysfs_compression_stats) { + bch2_compression_stats_to_text(&out, c); + return out.pos - buf; + } + + if (attr == &sysfs_new_stripes) { + bch2_new_stripes_to_text(&out, c); + return out.pos - buf; + } + + if (attr == &sysfs_io_timers_read) { + bch2_io_timers_to_text(&out, &c->io_clock[READ]); + return out.pos - buf; + } + if (attr == &sysfs_io_timers_write) { + bch2_io_timers_to_text(&out, &c->io_clock[WRITE]); + return out.pos - buf; + } return 0; } -STORE(__bch2_fs) +STORE(bch2_fs) { struct bch_fs *c = container_of(kobj, struct bch_fs, kobj); 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; @@ -380,73 +433,83 @@ STORE(__bch2_fs) } if (attr == &sysfs_copy_gc_enabled) { - struct bch_dev *ca; - unsigned i; ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled) ?: (ssize_t) size; - for_each_member_device(ca, c, i) - if (ca->moving_gc_read) - wake_up_process(ca->moving_gc_read); + if (c->copygc_thread) + wake_up_process(c->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_pd_controller_store(copy_gc, &c->copygc_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: */ -#define BCH_DEBUG_PARAM(name, description) sysfs_strtoul(name, c->name); - BCH_DEBUG_PARAMS() -#undef BCH_DEBUG_PARAM - - if (!bch2_fs_running(c)) + if (!test_bit(BCH_FS_STARTED, &c->flags)) return -EPERM; /* Debugging: */ if (attr == &sysfs_trigger_journal_flush) - bch2_journal_meta_async(&c->journal, NULL); + bch2_journal_meta(&c->journal); if (attr == &sysfs_trigger_btree_coalesce) bch2_coalesce(c); - if (attr == &sysfs_trigger_gc) - bch2_gc(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, NULL, false, false); + up_read(&c->state_lock); +#else + bch2_gc_gens(c); +#endif + } 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); } - return size; -} - -STORE(bch2_fs) -{ - struct bch_fs *c = container_of(kobj, struct bch_fs, kobj); - - mutex_lock(&c->state_lock); - size = __bch2_fs_store(kobj, attr, buf, size); - mutex_unlock(&c->state_lock); - +#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); @@ -463,10 +526,13 @@ 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 }; @@ -489,25 +555,35 @@ struct attribute *bch2_fs_internal_files[] = { &sysfs_alloc_debug, &sysfs_journal_debug, &sysfs_journal_pins, + &sysfs_btree_updates, + &sysfs_dirty_btree_nodes, + &sysfs_btree_cache, + &sysfs_btree_key_cache, + &sysfs_btree_transactions, + &sysfs_stripes_heap, &sysfs_read_realloc_races, + &sysfs_extent_migrate_done, + &sysfs_extent_migrate_raced, &sysfs_trigger_journal_flush, &sysfs_trigger_btree_coalesce, &sysfs_trigger_gc, &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_internal_uuid, -#define BCH_DEBUG_PARAM(name, description) &sysfs_##name, - BCH_DEBUG_PARAMS() -#undef BCH_DEBUG_PARAM + &sysfs_rebalance_enabled, + &sysfs_rebalance_work, + sysfs_pd_controller_files(rebalance), + sysfs_pd_controller_files(copy_gc), + + &sysfs_new_stripes, + &sysfs_io_timers_read, + &sysfs_io_timers_write, + + &sysfs_internal_uuid, NULL }; @@ -515,135 +591,166 @@ 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); - return bch2_opt_show(&c->opts, attr->name, buf, PAGE_SIZE); + bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST); + pr_buf(&out, "\n"); + + 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; - int 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 */ SHOW(bch2_fs_time_stats) { struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats); + struct printbuf out = _PBUF(buf, PAGE_SIZE); -#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) { \ + bch2_time_stats_to_text(&out, &c->times[BCH_TIME_##name]);\ + return out.pos - buf; \ + } 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) { + 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 int quantiles_to_text(struct printbuf *out, + struct bch_fs *c, struct bch_dev *ca, + 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]) @@ -655,85 +762,104 @@ static ssize_t show_quantiles(struct bch_dev *ca, char *buf, vfree(p); for (i = 0; i < ARRAY_SIZE(q); i++) - ret += scnprintf(buf + ret, PAGE_SIZE - ret, - "%u ", q[i]); - buf[ret - 1] = '\n'; - - return ret; - + pr_buf(out, "%u ", q[i]); + pr_buf(out, "\n"); + return 0; } -static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf) +static void reserve_stats_to_text(struct printbuf *out, struct bch_dev *ca) { 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); - - return ret; + spin_unlock(&ca->fs->freelist_lock); } -static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf) +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); - - 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" - "freelist_wait: %s\n" - "open buckets: %u/%u (reserved %u)\n" - "open_buckets_wait: %s\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, - 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"); + 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]++; + + pr_buf(out, + "\t\t buckets\t sectors fragmented\n" + "capacity%16llu\n", + ca->mi.nbuckets - ca->mi.first_bucket); + + for (i = 1; i < BCH_DATA_NR; i++) + pr_buf(out, "%-8s%16llu%16llu%16llu\n", + bch2_data_types[i], stats.d[i].buckets, + stats.d[i].sectors, stats.d[i].fragmented); + + pr_buf(out, + "ec\t%16llu\n" + "available%15llu\n" + "alloc\t%16llu\n" + "\n" + "free_inc\t\t%zu/%zu\n" + "free[RESERVE_MOVINGGC]\t%zu/%zu\n" + "free[RESERVE_NONE]\t%zu/%zu\n" + "freelist_wait\t\t%s\n" + "open buckets\t\t%u/%u (reserved %u)\n" + "open_buckets_wait\t%s\n" + "open_buckets_btree\t%u\n" + "open_buckets_user\t%u\n" + "btree reserve cache\t%u\n", + stats.buckets_ec, + __dev_buckets_available(ca, stats), + stats.buckets_alloc, + fifo_used(&ca->free_inc), ca->free_inc.size, + fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size, + fifo_used(&ca->free[RESERVE_NONE]), ca->free[RESERVE_NONE].size, + c->freelist_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 void dev_iodone_to_text(struct printbuf *out, struct bch_dev *ca) { - u64 ret = 0; - int cpu; + 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); + } } 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); @@ -741,53 +867,81 @@ 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_reserve_stats) - return show_reserve_stats(ca, buf); - if (attr == &sysfs_alloc_debug) - return show_dev_alloc_debug(ca, buf); + + 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; + } + + 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) { + dev_iodone_to_text(&out, ca); + return out.pos - 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) { + bch2_time_stats_to_text(&out, &ca->io_latency[READ]); + return out.pos - buf; + } + if (attr == &sysfs_io_latency_stats_write) { + bch2_time_stats_to_text(&out, &ca->io_latency[WRITE]); + return out.pos - buf; + } + + sysfs_printf(congested, "%u%%", + clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX) + * 100 / CONGESTED_MAX); + + if (attr == &sysfs_bucket_quantiles_last_read) + return quantiles_to_text(&out, c, ca, bucket_last_io_fn, (void *) 0) ?: out.pos - buf; + if (attr == &sysfs_bucket_quantiles_last_write) + return quantiles_to_text(&out, c, ca, bucket_last_io_fn, (void *) 1) ?: out.pos - buf; + if (attr == &sysfs_bucket_quantiles_fragmentation) + return quantiles_to_text(&out, c, ca, bucket_sectors_used_fn, NULL) ?: out.pos - buf; + if (attr == &sysfs_bucket_quantiles_oldest_gen) + return quantiles_to_text(&out, c, ca, bucket_oldest_gen_fn, NULL) ?: out.pos - buf; + + if (attr == &sysfs_reserve_stats) { + reserve_stats_to_text(&out, ca); + return out.pos - buf; + } + if (attr == &sysfs_alloc_debug) { + dev_alloc_debug_to_text(&out, ca); + return out.pos - buf; + } return 0; } @@ -798,13 +952,11 @@ STORE(bch2_dev) struct bch_fs *c = ca->fs; struct bch_member *mi; - sysfs_pd_controller_store(copy_gc, &ca->moving_gc_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); @@ -814,13 +966,13 @@ 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); @@ -829,31 +981,23 @@ 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); + if (attr == &sysfs_label) { + char *tmp; + int ret; - mutex_lock(&c->sb_lock); - prev_tier = ca->mi.tier; + tmp = kstrdup(buf, GFP_KERNEL); + if (!tmp) + return -ENOMEM; - 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); - - 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); - - 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); @@ -864,45 +1008,35 @@ 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_pd_controller_files(copy_gc), + &sysfs_wake_allocator, NULL }; + +#endif /* _BCACHEFS_SYSFS_H_ */