fs/bcachefs/sysfs.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * bcache sysfs interfaces
   4  *
   5  * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
   6  * Copyright 2012 Google, Inc.
   7  */
   8
   9 #ifndef NO_BCACHEFS_SYSFS
  10
  11 #include "bcachefs.h"
  12 #include "alloc_background.h"
  13 #include "sysfs.h"
  14 #include "btree_cache.h"
  15 #include "btree_io.h"
  16 #include "btree_iter.h"
  17 #include "btree_update.h"
  18 #include "btree_update_interior.h"
  19 #include "btree_gc.h"
  20 #include "buckets.h"
  21 #include "disk_groups.h"
  22 #include "ec.h"
  23 #include "inode.h"
  24 #include "journal.h"
  25 #include "keylist.h"
  26 #include "move.h"
  27 #include "opts.h"
  28 #include "rebalance.h"
  29 #include "replicas.h"
  30 #include "super-io.h"
  31 #include "tests.h"
  32
  33 #include <linux/blkdev.h>
  34 #include <linux/sort.h>
  35 #include <linux/sched/clock.h>
  36
  37 #include "util.h"
  38
  39 #define SYSFS_OPS(type)                                                 \
  40 struct sysfs_ops type ## _sysfs_ops = {                                 \
  41         .show   = type ## _show,                                        \
  42         .store  = type ## _store                                        \
  43 }
  44
  45 #define SHOW(fn)                                                        \
  46 static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
  47                            char *buf)                                   \
  48
  49 #define STORE(fn)                                                       \
  50 static ssize_t fn ## _store(struct kobject *kobj, struct attribute *attr,\
  51                             const char *buf, size_t size)               \
  52
  53 #define __sysfs_attribute(_name, _mode)                                 \
  54         static struct attribute sysfs_##_name =                         \
  55                 { .name = #_name, .mode = _mode }
  56
  57 #define write_attribute(n)      __sysfs_attribute(n, S_IWUSR)
  58 #define read_attribute(n)       __sysfs_attribute(n, S_IRUGO)
  59 #define rw_attribute(n)         __sysfs_attribute(n, S_IRUGO|S_IWUSR)
  60
  61 #define sysfs_printf(file, fmt, ...)                                    \
  62 do {                                                                    \
  63         if (attr == &sysfs_ ## file)                                    \
  64                 return scnprintf(buf, PAGE_SIZE, fmt "\n", __VA_ARGS__);\
  65 } while (0)
  66
  67 #define sysfs_print(file, var)                                          \
  68 do {                                                                    \
  69         if (attr == &sysfs_ ## file)                                    \
  70                 return snprint(buf, PAGE_SIZE, var);                    \
  71 } while (0)
  72
  73 #define sysfs_hprint(file, val)                                         \
  74 do {                                                                    \
  75         if (attr == &sysfs_ ## file) {                                  \
  76                 ssize_t ret = bch2_hprint(buf, val);                    \
  77                 strcat(buf, "\n");                                      \
  78                 return ret + 1;                                         \
  79         }                                                               \
  80 } while (0)
  81
  82 #define var_printf(_var, fmt)   sysfs_printf(_var, fmt, var(_var))
  83 #define var_print(_var)         sysfs_print(_var, var(_var))
  84 #define var_hprint(_var)        sysfs_hprint(_var, var(_var))
  85
  86 #define sysfs_strtoul(file, var)                                        \
  87 do {                                                                    \
  88         if (attr == &sysfs_ ## file)                                    \
  89                 return strtoul_safe(buf, var) ?: (ssize_t) size;        \
  90 } while (0)
  91
  92 #define sysfs_strtoul_clamp(file, var, min, max)                        \
  93 do {                                                                    \
  94         if (attr == &sysfs_ ## file)                                    \
  95                 return strtoul_safe_clamp(buf, var, min, max)           \
  96                         ?: (ssize_t) size;                              \
  97 } while (0)
  98
  99 #define strtoul_or_return(cp)                                           \
 100 ({                                                                      \
 101         unsigned long _v;                                               \
 102         int _r = kstrtoul(cp, 10, &_v);                                 \
 103         if (_r)                                                         \
 104                 return _r;                                              \
 105         _v;                                                             \
 106 })
 107
 108 #define strtoul_restrict_or_return(cp, min, max)                        \
 109 ({                                                                      \
 110         unsigned long __v = 0;                                          \
 111         int _r = strtoul_safe_restrict(cp, __v, min, max);              \
 112         if (_r)                                                         \
 113                 return _r;                                              \
 114         __v;                                                            \
 115 })
 116
 117 #define strtoi_h_or_return(cp)                                          \
 118 ({                                                                      \
 119         u64 _v;                                                         \
 120         int _r = strtoi_h(cp, &_v);                                     \
 121         if (_r)                                                         \
 122                 return _r;                                              \
 123         _v;                                                             \
 124 })
 125
 126 #define sysfs_hatoi(file, var)                                          \
 127 do {                                                                    \
 128         if (attr == &sysfs_ ## file)                                    \
 129                 return strtoi_h(buf, &var) ?: (ssize_t) size;           \
 130 } while (0)
 131
 132 write_attribute(trigger_journal_flush);
 133 write_attribute(trigger_btree_coalesce);
 134 write_attribute(trigger_gc);
 135 write_attribute(prune_cache);
 136 rw_attribute(btree_gc_periodic);
 137
 138 read_attribute(uuid);
 139 read_attribute(minor);
 140 read_attribute(bucket_size);
 141 read_attribute(block_size);
 142 read_attribute(btree_node_size);
 143 read_attribute(first_bucket);
 144 read_attribute(nbuckets);
 145 read_attribute(durability);
 146 read_attribute(iodone);
 147
 148 read_attribute(io_latency_read);
 149 read_attribute(io_latency_write);
 150 read_attribute(io_latency_stats_read);
 151 read_attribute(io_latency_stats_write);
 152 read_attribute(congested);
 153
 154 read_attribute(bucket_quantiles_last_read);
 155 read_attribute(bucket_quantiles_last_write);
 156 read_attribute(bucket_quantiles_fragmentation);
 157 read_attribute(bucket_quantiles_oldest_gen);
 158
 159 read_attribute(reserve_stats);
 160 read_attribute(btree_cache_size);
 161 read_attribute(compression_stats);
 162 read_attribute(journal_debug);
 163 read_attribute(journal_pins);
 164 read_attribute(btree_updates);
 165 read_attribute(dirty_btree_nodes);
 166
 167 read_attribute(internal_uuid);
 168
 169 read_attribute(has_data);
 170 read_attribute(alloc_debug);
 171 write_attribute(wake_allocator);
 172
 173 read_attribute(read_realloc_races);
 174 read_attribute(extent_migrate_done);
 175 read_attribute(extent_migrate_raced);
 176
 177 rw_attribute(journal_write_delay_ms);
 178 rw_attribute(journal_reclaim_delay_ms);
 179
 180 rw_attribute(discard);
 181 rw_attribute(cache_replacement_policy);
 182 rw_attribute(label);
 183
 184 rw_attribute(copy_gc_enabled);
 185 sysfs_pd_controller_attribute(copy_gc);
 186
 187 rw_attribute(rebalance_enabled);
 188 sysfs_pd_controller_attribute(rebalance);
 189 read_attribute(rebalance_work);
 190 rw_attribute(promote_whole_extents);
 191
 192 read_attribute(new_stripes);
 193
 194 rw_attribute(pd_controllers_update_seconds);
 195
 196 read_attribute(meta_replicas_have);
 197 read_attribute(data_replicas_have);
 198
 199 #ifdef CONFIG_BCACHEFS_TESTS
 200 write_attribute(perf_test);
 201 #endif /* CONFIG_BCACHEFS_TESTS */
 202
 203 #define BCH_DEBUG_PARAM(name, description)                              \
 204         rw_attribute(name);
 205
 206         BCH_DEBUG_PARAMS()
 207 #undef BCH_DEBUG_PARAM
 208
 209 #define x(_name)                                                \
 210         static struct attribute sysfs_time_stat_##_name =               \
 211                 { .name = #_name, .mode = S_IRUGO };
 212         BCH_TIME_STATS()
 213 #undef x
 214
 215 static struct attribute sysfs_state_rw = {
 216         .name = "state",
 217         .mode = S_IRUGO
 218 };
 219
 220 static size_t bch2_btree_cache_size(struct bch_fs *c)
 221 {
 222         size_t ret = 0;
 223         struct btree *b;
 224
 225         mutex_lock(&c->btree_cache.lock);
 226         list_for_each_entry(b, &c->btree_cache.live, list)
 227                 ret += btree_bytes(c);
 228
 229         mutex_unlock(&c->btree_cache.lock);
 230         return ret;
 231 }
 232
 233 static ssize_t show_fs_alloc_debug(struct bch_fs *c, char *buf)
 234 {
 235         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 236         struct bch_fs_usage stats = bch2_fs_usage_read(c);
 237         unsigned replicas, type;
 238
 239         pr_buf(&out, "capacity:\t\t%llu\n", c->capacity);
 240
 241         for (replicas = 0; replicas < ARRAY_SIZE(stats.replicas); replicas++) {
 242                 pr_buf(&out, "%u replicas:\n", replicas + 1);
 243
 244                 for (type = BCH_DATA_SB; type < BCH_DATA_NR; type++)
 245                         pr_buf(&out, "\t%s:\t\t%llu\n",
 246                                bch2_data_types[type],
 247                                stats.replicas[replicas].data[type]);
 248                 pr_buf(&out, "\terasure coded:\t%llu\n",
 249                        stats.replicas[replicas].ec_data);
 250                 pr_buf(&out, "\treserved:\t%llu\n",
 251                        stats.replicas[replicas].persistent_reserved);
 252         }
 253
 254         pr_buf(&out, "bucket usage\n");
 255
 256         for (type = BCH_DATA_SB; type < BCH_DATA_NR; type++)
 257                 pr_buf(&out, "\t%s:\t\t%llu\n",
 258                        bch2_data_types[type],
 259                        stats.buckets[type]);
 260
 261         pr_buf(&out, "online reserved:\t%llu\n",
 262                stats.online_reserved);
 263
 264         return out.pos - buf;
 265 }
 266
 267 static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 268 {
 269         struct btree_iter iter;
 270         struct bkey_s_c k;
 271         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 272             nr_compressed_extents = 0,
 273             compressed_sectors_compressed = 0,
 274             compressed_sectors_uncompressed = 0;
 275
 276         if (!bch2_fs_running(c))
 277                 return -EPERM;
 278
 279         for_each_btree_key(&iter, c, BTREE_ID_EXTENTS, POS_MIN, 0, k)
 280                 if (k.k->type == KEY_TYPE_extent) {
 281                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 282                         const union bch_extent_entry *entry;
 283                         struct extent_ptr_decoded p;
 284
 285                         extent_for_each_ptr_decode(e, p, entry) {
 286                                 if (p.crc.compression_type == BCH_COMPRESSION_NONE) {
 287                                         nr_uncompressed_extents++;
 288                                         uncompressed_sectors += e.k->size;
 289                                 } else {
 290                                         nr_compressed_extents++;
 291                                         compressed_sectors_compressed +=
 292                                                 p.crc.compressed_size;
 293                                         compressed_sectors_uncompressed +=
 294                                                 p.crc.uncompressed_size;
 295                                 }
 296
 297                                 /* only looking at the first ptr */
 298                                 break;
 299                         }
 300                 }
 301         bch2_btree_iter_unlock(&iter);
 302
 303         return scnprintf(buf, PAGE_SIZE,
 304                         "uncompressed data:\n"
 305                         "       nr extents:                     %llu\n"
 306                         "       size (bytes):                   %llu\n"
 307                         "compressed data:\n"
 308                         "       nr extents:                     %llu\n"
 309                         "       compressed size (bytes):        %llu\n"
 310                         "       uncompressed size (bytes):      %llu\n",
 311                         nr_uncompressed_extents,
 312                         uncompressed_sectors << 9,
 313                         nr_compressed_extents,
 314                         compressed_sectors_compressed << 9,
 315                         compressed_sectors_uncompressed << 9);
 316 }
 317
 318 static ssize_t bch2_new_stripes(struct bch_fs *c, char *buf)
 319 {
 320         char *out = buf, *end = buf + PAGE_SIZE;
 321         struct ec_stripe_head *h;
 322         struct ec_stripe_new *s;
 323
 324         mutex_lock(&c->ec_new_stripe_lock);
 325         list_for_each_entry(h, &c->ec_new_stripe_list, list) {
 326                 out += scnprintf(out, end - out,
 327                                  "target %u algo %u redundancy %u:\n",
 328                                  h->target, h->algo, h->redundancy);
 329
 330                 if (h->s)
 331                         out += scnprintf(out, end - out,
 332                                          "\tpending: blocks %u allocated %u\n",
 333                                          h->s->blocks.nr,
 334                                          bitmap_weight(h->s->blocks_allocated,
 335                                                        h->s->blocks.nr));
 336
 337                 mutex_lock(&h->lock);
 338                 list_for_each_entry(s, &h->stripes, list)
 339                         out += scnprintf(out, end - out,
 340                                          "\tin flight: blocks %u allocated %u pin %u\n",
 341                                          s->blocks.nr,
 342                                          bitmap_weight(s->blocks_allocated,
 343                                                        s->blocks.nr),
 344                                          atomic_read(&s->pin));
 345                 mutex_unlock(&h->lock);
 346
 347         }
 348         mutex_unlock(&c->ec_new_stripe_lock);
 349
 350         return out - buf;
 351 }
 352
 353 SHOW(bch2_fs)
 354 {
 355         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 356
 357         sysfs_print(minor,                      c->minor);
 358         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 359
 360         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 361         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 362
 363         sysfs_print(block_size,                 block_bytes(c));
 364         sysfs_print(btree_node_size,            btree_bytes(c));
 365         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 366
 367         sysfs_print(read_realloc_races,
 368                     atomic_long_read(&c->read_realloc_races));
 369         sysfs_print(extent_migrate_done,
 370                     atomic_long_read(&c->extent_migrate_done));
 371         sysfs_print(extent_migrate_raced,
 372                     atomic_long_read(&c->extent_migrate_raced));
 373
 374         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 375
 376         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 377
 378         sysfs_print(pd_controllers_update_seconds,
 379                     c->pd_controllers_update_seconds);
 380
 381         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 382         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 383
 384         if (attr == &sysfs_rebalance_work)
 385                 return bch2_rebalance_work_show(c, buf);
 386
 387         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 388
 389         sysfs_printf(meta_replicas_have, "%i",  bch2_replicas_online(c, true));
 390         sysfs_printf(data_replicas_have, "%i",  bch2_replicas_online(c, false));
 391
 392         /* Debugging: */
 393
 394         if (attr == &sysfs_alloc_debug)
 395                 return show_fs_alloc_debug(c, buf);
 396
 397         if (attr == &sysfs_journal_debug)
 398                 return bch2_journal_print_debug(&c->journal, buf);
 399
 400         if (attr == &sysfs_journal_pins)
 401                 return bch2_journal_print_pins(&c->journal, buf);
 402
 403         if (attr == &sysfs_btree_updates)
 404                 return bch2_btree_updates_print(c, buf);
 405
 406         if (attr == &sysfs_dirty_btree_nodes)
 407                 return bch2_dirty_btree_nodes_print(c, buf);
 408
 409         if (attr == &sysfs_compression_stats)
 410                 return bch2_compression_stats(c, buf);
 411
 412         if (attr == &sysfs_new_stripes)
 413                 return bch2_new_stripes(c, buf);
 414
 415 #define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
 416         BCH_DEBUG_PARAMS()
 417 #undef BCH_DEBUG_PARAM
 418
 419         return 0;
 420 }
 421
 422 STORE(__bch2_fs)
 423 {
 424         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 425
 426         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 427         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 428
 429         if (attr == &sysfs_btree_gc_periodic) {
 430                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 431                         ?: (ssize_t) size;
 432
 433                 wake_up_process(c->gc_thread);
 434                 return ret;
 435         }
 436
 437         if (attr == &sysfs_copy_gc_enabled) {
 438                 struct bch_dev *ca;
 439                 unsigned i;
 440                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 441                         ?: (ssize_t) size;
 442
 443                 for_each_member_device(ca, c, i)
 444                         if (ca->copygc_thread)
 445                                 wake_up_process(ca->copygc_thread);
 446                 return ret;
 447         }
 448
 449         if (attr == &sysfs_rebalance_enabled) {
 450                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 451                         ?: (ssize_t) size;
 452
 453                 rebalance_wakeup(c);
 454                 return ret;
 455         }
 456
 457         sysfs_strtoul(pd_controllers_update_seconds,
 458                       c->pd_controllers_update_seconds);
 459         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 460
 461         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 462
 463         /* Debugging: */
 464
 465 #define BCH_DEBUG_PARAM(name, description) sysfs_strtoul(name, c->name);
 466         BCH_DEBUG_PARAMS()
 467 #undef BCH_DEBUG_PARAM
 468
 469         if (!bch2_fs_running(c))
 470                 return -EPERM;
 471
 472         /* Debugging: */
 473
 474         if (attr == &sysfs_trigger_journal_flush)
 475                 bch2_journal_meta_async(&c->journal, NULL);
 476
 477         if (attr == &sysfs_trigger_btree_coalesce)
 478                 bch2_coalesce(c);
 479
 480         if (attr == &sysfs_trigger_gc)
 481                 bch2_gc(c, NULL, false);
 482
 483         if (attr == &sysfs_prune_cache) {
 484                 struct shrink_control sc;
 485
 486                 sc.gfp_mask = GFP_KERNEL;
 487                 sc.nr_to_scan = strtoul_or_return(buf);
 488                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 489         }
 490 #ifdef CONFIG_BCACHEFS_TESTS
 491         if (attr == &sysfs_perf_test) {
 492                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 493                 char *test              = strsep(&p, " \t\n");
 494                 char *nr_str            = strsep(&p, " \t\n");
 495                 char *threads_str       = strsep(&p, " \t\n");
 496                 unsigned threads;
 497                 u64 nr;
 498                 int ret = -EINVAL;
 499
 500                 if (threads_str &&
 501                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 502                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 503                         bch2_btree_perf_test(c, test, nr, threads);
 504                 else
 505                         size = ret;
 506                 kfree(tmp);
 507         }
 508 #endif
 509         return size;
 510 }
 511
 512 STORE(bch2_fs)
 513 {
 514         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 515
 516         mutex_lock(&c->state_lock);
 517         size = __bch2_fs_store(kobj, attr, buf, size);
 518         mutex_unlock(&c->state_lock);
 519
 520         return size;
 521 }
 522 SYSFS_OPS(bch2_fs);
 523
 524 struct attribute *bch2_fs_files[] = {
 525         &sysfs_minor,
 526         &sysfs_block_size,
 527         &sysfs_btree_node_size,
 528         &sysfs_btree_cache_size,
 529
 530         &sysfs_meta_replicas_have,
 531         &sysfs_data_replicas_have,
 532
 533         &sysfs_journal_write_delay_ms,
 534         &sysfs_journal_reclaim_delay_ms,
 535
 536         &sysfs_promote_whole_extents,
 537
 538         &sysfs_compression_stats,
 539
 540 #ifdef CONFIG_BCACHEFS_TESTS
 541         &sysfs_perf_test,
 542 #endif
 543         NULL
 544 };
 545
 546 /* internal dir - just a wrapper */
 547
 548 SHOW(bch2_fs_internal)
 549 {
 550         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 551         return bch2_fs_show(&c->kobj, attr, buf);
 552 }
 553
 554 STORE(bch2_fs_internal)
 555 {
 556         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 557         return bch2_fs_store(&c->kobj, attr, buf, size);
 558 }
 559 SYSFS_OPS(bch2_fs_internal);
 560
 561 struct attribute *bch2_fs_internal_files[] = {
 562         &sysfs_alloc_debug,
 563         &sysfs_journal_debug,
 564         &sysfs_journal_pins,
 565         &sysfs_btree_updates,
 566         &sysfs_dirty_btree_nodes,
 567
 568         &sysfs_read_realloc_races,
 569         &sysfs_extent_migrate_done,
 570         &sysfs_extent_migrate_raced,
 571
 572         &sysfs_trigger_journal_flush,
 573         &sysfs_trigger_btree_coalesce,
 574         &sysfs_trigger_gc,
 575         &sysfs_prune_cache,
 576
 577         &sysfs_copy_gc_enabled,
 578
 579         &sysfs_rebalance_enabled,
 580         &sysfs_rebalance_work,
 581         sysfs_pd_controller_files(rebalance),
 582
 583         &sysfs_new_stripes,
 584
 585         &sysfs_internal_uuid,
 586
 587 #define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
 588         BCH_DEBUG_PARAMS()
 589 #undef BCH_DEBUG_PARAM
 590
 591         NULL
 592 };
 593
 594 /* options */
 595
 596 SHOW(bch2_fs_opts_dir)
 597 {
 598         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 599         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 600         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 601         int id = opt - bch2_opt_table;
 602         u64 v = bch2_opt_get_by_id(&c->opts, id);
 603
 604         bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
 605         pr_buf(&out, "\n");
 606
 607         return out.pos - buf;
 608 }
 609
 610 STORE(bch2_fs_opts_dir)
 611 {
 612         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 613         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 614         int ret, id = opt - bch2_opt_table;
 615         char *tmp;
 616         u64 v;
 617
 618         tmp = kstrdup(buf, GFP_KERNEL);
 619         if (!tmp)
 620                 return -ENOMEM;
 621
 622         ret = bch2_opt_parse(c, opt, strim(tmp), &v);
 623         kfree(tmp);
 624
 625         if (ret < 0)
 626                 return ret;
 627
 628         ret = bch2_opt_check_may_set(c, id, v);
 629         if (ret < 0)
 630                 return ret;
 631
 632         if (opt->set_sb != SET_NO_SB_OPT) {
 633                 mutex_lock(&c->sb_lock);
 634                 opt->set_sb(c->disk_sb.sb, v);
 635                 bch2_write_super(c);
 636                 mutex_unlock(&c->sb_lock);
 637         }
 638
 639         bch2_opt_set_by_id(&c->opts, id, v);
 640
 641         if ((id == Opt_background_target ||
 642              id == Opt_background_compression) && v) {
 643                 bch2_rebalance_add_work(c, S64_MAX);
 644                 rebalance_wakeup(c);
 645         }
 646
 647         return size;
 648 }
 649 SYSFS_OPS(bch2_fs_opts_dir);
 650
 651 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 652
 653 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 654 {
 655         const struct bch_option *i;
 656         int ret;
 657
 658         for (i = bch2_opt_table;
 659              i < bch2_opt_table + bch2_opts_nr;
 660              i++) {
 661                 if (i->mode == OPT_INTERNAL)
 662                         continue;
 663
 664                 ret = sysfs_create_file(kobj, &i->attr);
 665                 if (ret)
 666                         return ret;
 667         }
 668
 669         return 0;
 670 }
 671
 672 /* time stats */
 673
 674 SHOW(bch2_fs_time_stats)
 675 {
 676         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 677
 678 #define x(name)                                         \
 679         if (attr == &sysfs_time_stat_##name)                            \
 680                 return bch2_time_stats_print(&c->times[BCH_TIME_##name],\
 681                                              buf, PAGE_SIZE);
 682         BCH_TIME_STATS()
 683 #undef x
 684
 685         return 0;
 686 }
 687
 688 STORE(bch2_fs_time_stats)
 689 {
 690         return size;
 691 }
 692 SYSFS_OPS(bch2_fs_time_stats);
 693
 694 struct attribute *bch2_fs_time_stats_files[] = {
 695 #define x(name)                                         \
 696         &sysfs_time_stat_##name,
 697         BCH_TIME_STATS()
 698 #undef x
 699         NULL
 700 };
 701
 702 typedef unsigned (bucket_map_fn)(struct bch_fs *, struct bch_dev *,
 703                                  size_t, void *);
 704
 705 static unsigned bucket_last_io_fn(struct bch_fs *c, struct bch_dev *ca,
 706                                   size_t b, void *private)
 707 {
 708         int rw = (private ? 1 : 0);
 709
 710         return bucket_last_io(c, bucket(ca, b), rw);
 711 }
 712
 713 static unsigned bucket_sectors_used_fn(struct bch_fs *c, struct bch_dev *ca,
 714                                        size_t b, void *private)
 715 {
 716         struct bucket *g = bucket(ca, b);
 717         return bucket_sectors_used(g->mark);
 718 }
 719
 720 static unsigned bucket_oldest_gen_fn(struct bch_fs *c, struct bch_dev *ca,
 721                                      size_t b, void *private)
 722 {
 723         return bucket_gc_gen(ca, b);
 724 }
 725
 726 static int unsigned_cmp(const void *_l, const void *_r)
 727 {
 728         unsigned l = *((unsigned *) _l);
 729         unsigned r = *((unsigned *) _r);
 730
 731         return (l > r) - (l < r);
 732 }
 733
 734 static ssize_t show_quantiles(struct bch_fs *c, struct bch_dev *ca,
 735                               char *buf, bucket_map_fn *fn, void *private)
 736 {
 737         size_t i, n;
 738         /* Compute 31 quantiles */
 739         unsigned q[31], *p;
 740         ssize_t ret = 0;
 741
 742         down_read(&ca->bucket_lock);
 743         n = ca->mi.nbuckets;
 744
 745         p = vzalloc(n * sizeof(unsigned));
 746         if (!p) {
 747                 up_read(&ca->bucket_lock);
 748                 return -ENOMEM;
 749         }
 750
 751         for (i = ca->mi.first_bucket; i < n; i++)
 752                 p[i] = fn(c, ca, i, private);
 753
 754         sort(p, n, sizeof(unsigned), unsigned_cmp, NULL);
 755         up_read(&ca->bucket_lock);
 756
 757         while (n &&
 758                !p[n - 1])
 759                 --n;
 760
 761         for (i = 0; i < ARRAY_SIZE(q); i++)
 762                 q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
 763
 764         vfree(p);
 765
 766         for (i = 0; i < ARRAY_SIZE(q); i++)
 767                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 768                                  "%u ", q[i]);
 769         buf[ret - 1] = '\n';
 770
 771         return ret;
 772 }
 773
 774 static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
 775 {
 776         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 777         enum alloc_reserve i;
 778
 779         spin_lock(&ca->freelist_lock);
 780
 781         pr_buf(&out, "free_inc:\t%zu\t%zu\n",
 782                fifo_used(&ca->free_inc),
 783                ca->free_inc.size);
 784
 785         for (i = 0; i < RESERVE_NR; i++)
 786                 pr_buf(&out, "free[%u]:\t%zu\t%zu\n", i,
 787                        fifo_used(&ca->free[i]),
 788                        ca->free[i].size);
 789
 790         spin_unlock(&ca->freelist_lock);
 791
 792         return out.pos - buf;
 793 }
 794
 795 static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
 796 {
 797         struct bch_fs *c = ca->fs;
 798         struct bch_dev_usage stats = bch2_dev_usage_read(c, ca);
 799
 800         return scnprintf(buf, PAGE_SIZE,
 801                 "free_inc:               %zu/%zu\n"
 802                 "free[RESERVE_BTREE]:    %zu/%zu\n"
 803                 "free[RESERVE_MOVINGGC]: %zu/%zu\n"
 804                 "free[RESERVE_NONE]:     %zu/%zu\n"
 805                 "buckets:\n"
 806                 "    capacity:           %llu\n"
 807                 "    alloc:              %llu\n"
 808                 "    sb:                 %llu\n"
 809                 "    journal:            %llu\n"
 810                 "    meta:               %llu\n"
 811                 "    user:               %llu\n"
 812                 "    cached:             %llu\n"
 813                 "    erasure coded:      %llu\n"
 814                 "    available:          %lli\n"
 815                 "sectors:\n"
 816                 "    sb:                 %llu\n"
 817                 "    journal:            %llu\n"
 818                 "    meta:               %llu\n"
 819                 "    user:               %llu\n"
 820                 "    cached:             %llu\n"
 821                 "    fragmented:         %llu\n"
 822                 "    copygc threshold:   %llu\n"
 823                 "freelist_wait:          %s\n"
 824                 "open buckets:           %u/%u (reserved %u)\n"
 825                 "open_buckets_wait:      %s\n",
 826                 fifo_used(&ca->free_inc),               ca->free_inc.size,
 827                 fifo_used(&ca->free[RESERVE_BTREE]),    ca->free[RESERVE_BTREE].size,
 828                 fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size,
 829                 fifo_used(&ca->free[RESERVE_NONE]),     ca->free[RESERVE_NONE].size,
 830                 ca->mi.nbuckets - ca->mi.first_bucket,
 831                 stats.buckets_alloc,
 832                 stats.buckets[BCH_DATA_SB],
 833                 stats.buckets[BCH_DATA_JOURNAL],
 834                 stats.buckets[BCH_DATA_BTREE],
 835                 stats.buckets[BCH_DATA_USER],
 836                 stats.buckets[BCH_DATA_CACHED],
 837                 stats.buckets_ec,
 838                 ca->mi.nbuckets - ca->mi.first_bucket - stats.buckets_unavailable,
 839                 stats.sectors[BCH_DATA_SB],
 840                 stats.sectors[BCH_DATA_JOURNAL],
 841                 stats.sectors[BCH_DATA_BTREE],
 842                 stats.sectors[BCH_DATA_USER],
 843                 stats.sectors[BCH_DATA_CACHED],
 844                 stats.sectors_fragmented,
 845                 ca->copygc_threshold,
 846                 c->freelist_wait.list.first             ? "waiting" : "empty",
 847                 c->open_buckets_nr_free, OPEN_BUCKETS_COUNT, BTREE_NODE_RESERVE,
 848                 c->open_buckets_wait.list.first         ? "waiting" : "empty");
 849 }
 850
 851 static const char * const bch2_rw[] = {
 852         "read",
 853         "write",
 854         NULL
 855 };
 856
 857 static ssize_t show_dev_iodone(struct bch_dev *ca, char *buf)
 858 {
 859         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 860         int rw, i, cpu;
 861
 862         for (rw = 0; rw < 2; rw++) {
 863                 pr_buf(&out, "%s:\n", bch2_rw[rw]);
 864
 865                 for (i = 1; i < BCH_DATA_NR; i++) {
 866                         u64 n = 0;
 867
 868                         for_each_possible_cpu(cpu)
 869                                 n += per_cpu_ptr(ca->io_done, cpu)->sectors[rw][i];
 870
 871                         pr_buf(&out, "%-12s:%12llu\n",
 872                                bch2_data_types[i], n << 9);
 873                 }
 874         }
 875
 876         return out.pos - buf;
 877 }
 878
 879 SHOW(bch2_dev)
 880 {
 881         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 882         struct bch_fs *c = ca->fs;
 883         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 884
 885         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 886
 887         sysfs_print(bucket_size,        bucket_bytes(ca));
 888         sysfs_print(block_size,         block_bytes(c));
 889         sysfs_print(first_bucket,       ca->mi.first_bucket);
 890         sysfs_print(nbuckets,           ca->mi.nbuckets);
 891         sysfs_print(durability,         ca->mi.durability);
 892         sysfs_print(discard,            ca->mi.discard);
 893
 894         if (attr == &sysfs_label) {
 895                 if (ca->mi.group) {
 896                         mutex_lock(&c->sb_lock);
 897                         bch2_disk_path_to_text(&out, &c->disk_sb,
 898                                                ca->mi.group - 1);
 899                         mutex_unlock(&c->sb_lock);
 900                 } else {
 901                         pr_buf(&out, "none");
 902                 }
 903
 904                 pr_buf(&out, "\n");
 905                 return out.pos - buf;
 906         }
 907
 908         if (attr == &sysfs_has_data) {
 909                 bch2_flags_to_text(&out, bch2_data_types,
 910                                    bch2_dev_has_data(c, ca));
 911                 pr_buf(&out, "\n");
 912                 return out.pos - buf;
 913         }
 914
 915         sysfs_pd_controller_show(copy_gc, &ca->copygc_pd);
 916
 917         if (attr == &sysfs_cache_replacement_policy) {
 918                 bch2_string_opt_to_text(&out,
 919                                         bch2_cache_replacement_policies,
 920                                         ca->mi.replacement);
 921                 pr_buf(&out, "\n");
 922                 return out.pos - buf;
 923         }
 924
 925         if (attr == &sysfs_state_rw) {
 926                 bch2_string_opt_to_text(&out, bch2_dev_state,
 927                                         ca->mi.state);
 928                 pr_buf(&out, "\n");
 929                 return out.pos - buf;
 930         }
 931
 932         if (attr == &sysfs_iodone)
 933                 return show_dev_iodone(ca, buf);
 934
 935         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 936         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 937
 938         if (attr == &sysfs_io_latency_stats_read)
 939                 return bch2_time_stats_print(&ca->io_latency[READ], buf, PAGE_SIZE);
 940         if (attr == &sysfs_io_latency_stats_write)
 941                 return bch2_time_stats_print(&ca->io_latency[WRITE], buf, PAGE_SIZE);
 942
 943         sysfs_printf(congested,                 "%u%%",
 944                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 945                      * 100 / CONGESTED_MAX);
 946
 947         if (attr == &sysfs_bucket_quantiles_last_read)
 948                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 0);
 949         if (attr == &sysfs_bucket_quantiles_last_write)
 950                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 1);
 951         if (attr == &sysfs_bucket_quantiles_fragmentation)
 952                 return show_quantiles(c, ca, buf, bucket_sectors_used_fn, NULL);
 953         if (attr == &sysfs_bucket_quantiles_oldest_gen)
 954                 return show_quantiles(c, ca, buf, bucket_oldest_gen_fn, NULL);
 955
 956         if (attr == &sysfs_reserve_stats)
 957                 return show_reserve_stats(ca, buf);
 958         if (attr == &sysfs_alloc_debug)
 959                 return show_dev_alloc_debug(ca, buf);
 960
 961         return 0;
 962 }
 963
 964 STORE(bch2_dev)
 965 {
 966         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 967         struct bch_fs *c = ca->fs;
 968         struct bch_member *mi;
 969
 970         sysfs_pd_controller_store(copy_gc, &ca->copygc_pd);
 971
 972         if (attr == &sysfs_discard) {
 973                 bool v = strtoul_or_return(buf);
 974
 975                 mutex_lock(&c->sb_lock);
 976                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 977
 978                 if (v != BCH_MEMBER_DISCARD(mi)) {
 979                         SET_BCH_MEMBER_DISCARD(mi, v);
 980                         bch2_write_super(c);
 981                 }
 982                 mutex_unlock(&c->sb_lock);
 983         }
 984
 985         if (attr == &sysfs_cache_replacement_policy) {
 986                 ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
 987
 988                 if (v < 0)
 989                         return v;
 990
 991                 mutex_lock(&c->sb_lock);
 992                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 993
 994                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
 995                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
 996                         bch2_write_super(c);
 997                 }
 998                 mutex_unlock(&c->sb_lock);
 999         }
1000
1001         if (attr == &sysfs_label) {
1002                 char *tmp;
1003                 int ret;
1004
1005                 tmp = kstrdup(buf, GFP_KERNEL);
1006                 if (!tmp)
1007                         return -ENOMEM;
1008
1009                 ret = bch2_dev_group_set(c, ca, strim(tmp));
1010                 kfree(tmp);
1011                 if (ret)
1012                         return ret;
1013         }
1014
1015         if (attr == &sysfs_wake_allocator)
1016                 bch2_wake_allocator(ca);
1017
1018         return size;
1019 }
1020 SYSFS_OPS(bch2_dev);
1021
1022 struct attribute *bch2_dev_files[] = {
1023         &sysfs_uuid,
1024         &sysfs_bucket_size,
1025         &sysfs_block_size,
1026         &sysfs_first_bucket,
1027         &sysfs_nbuckets,
1028         &sysfs_durability,
1029
1030         /* settings: */
1031         &sysfs_discard,
1032         &sysfs_cache_replacement_policy,
1033         &sysfs_state_rw,
1034         &sysfs_label,
1035
1036         &sysfs_has_data,
1037         &sysfs_iodone,
1038
1039         &sysfs_io_latency_read,
1040         &sysfs_io_latency_write,
1041         &sysfs_io_latency_stats_read,
1042         &sysfs_io_latency_stats_write,
1043         &sysfs_congested,
1044
1045         /* alloc info - other stats: */
1046         &sysfs_bucket_quantiles_last_read,
1047         &sysfs_bucket_quantiles_last_write,
1048         &sysfs_bucket_quantiles_fragmentation,
1049         &sysfs_bucket_quantiles_oldest_gen,
1050
1051         &sysfs_reserve_stats,
1052
1053         /* debug: */
1054         &sysfs_alloc_debug,
1055         &sysfs_wake_allocator,
1056
1057         sysfs_pd_controller_files(copy_gc),
1058         NULL
1059 };
1060
1061 #endif  /* _BCACHEFS_SYSFS_H_ */