git.sesse.net Git - bcachefs-tools-debian/blob - libbcachefs/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                 struct printbuf out = _PBUF(buf, PAGE_SIZE);            \
  77                 bch2_hprint(&out, val);                                 \
  78                 pr_buf(&out, "\n");                                     \
  79                 return out.pos - buf;                                   \
  80         }                                                               \
  81 } while (0)
  82
  83 #define var_printf(_var, fmt)   sysfs_printf(_var, fmt, var(_var))
  84 #define var_print(_var)         sysfs_print(_var, var(_var))
  85 #define var_hprint(_var)        sysfs_hprint(_var, var(_var))
  86
  87 #define sysfs_strtoul(file, var)                                        \
  88 do {                                                                    \
  89         if (attr == &sysfs_ ## file)                                    \
  90                 return strtoul_safe(buf, var) ?: (ssize_t) size;        \
  91 } while (0)
  92
  93 #define sysfs_strtoul_clamp(file, var, min, max)                        \
  94 do {                                                                    \
  95         if (attr == &sysfs_ ## file)                                    \
  96                 return strtoul_safe_clamp(buf, var, min, max)           \
  97                         ?: (ssize_t) size;                              \
  98 } while (0)
  99
 100 #define strtoul_or_return(cp)                                           \
 101 ({                                                                      \
 102         unsigned long _v;                                               \
 103         int _r = kstrtoul(cp, 10, &_v);                                 \
 104         if (_r)                                                         \
 105                 return _r;                                              \
 106         _v;                                                             \
 107 })
 108
 109 #define strtoul_restrict_or_return(cp, min, max)                        \
 110 ({                                                                      \
 111         unsigned long __v = 0;                                          \
 112         int _r = strtoul_safe_restrict(cp, __v, min, max);              \
 113         if (_r)                                                         \
 114                 return _r;                                              \
 115         __v;                                                            \
 116 })
 117
 118 #define strtoi_h_or_return(cp)                                          \
 119 ({                                                                      \
 120         u64 _v;                                                         \
 121         int _r = strtoi_h(cp, &_v);                                     \
 122         if (_r)                                                         \
 123                 return _r;                                              \
 124         _v;                                                             \
 125 })
 126
 127 #define sysfs_hatoi(file, var)                                          \
 128 do {                                                                    \
 129         if (attr == &sysfs_ ## file)                                    \
 130                 return strtoi_h(buf, &var) ?: (ssize_t) size;           \
 131 } while (0)
 132
 133 write_attribute(trigger_journal_flush);
 134 write_attribute(trigger_btree_coalesce);
 135 write_attribute(trigger_gc);
 136 write_attribute(trigger_alloc_write);
 137 write_attribute(prune_cache);
 138 rw_attribute(btree_gc_periodic);
 139
 140 read_attribute(uuid);
 141 read_attribute(minor);
 142 read_attribute(bucket_size);
 143 read_attribute(block_size);
 144 read_attribute(btree_node_size);
 145 read_attribute(first_bucket);
 146 read_attribute(nbuckets);
 147 read_attribute(durability);
 148 read_attribute(iodone);
 149
 150 read_attribute(io_latency_read);
 151 read_attribute(io_latency_write);
 152 read_attribute(io_latency_stats_read);
 153 read_attribute(io_latency_stats_write);
 154 read_attribute(congested);
 155
 156 read_attribute(bucket_quantiles_last_read);
 157 read_attribute(bucket_quantiles_last_write);
 158 read_attribute(bucket_quantiles_fragmentation);
 159 read_attribute(bucket_quantiles_oldest_gen);
 160
 161 read_attribute(reserve_stats);
 162 read_attribute(btree_cache_size);
 163 read_attribute(compression_stats);
 164 read_attribute(journal_debug);
 165 read_attribute(journal_pins);
 166 read_attribute(btree_updates);
 167 read_attribute(dirty_btree_nodes);
 168
 169 read_attribute(internal_uuid);
 170
 171 read_attribute(has_data);
 172 read_attribute(alloc_debug);
 173 write_attribute(wake_allocator);
 174
 175 read_attribute(read_realloc_races);
 176 read_attribute(extent_migrate_done);
 177 read_attribute(extent_migrate_raced);
 178
 179 rw_attribute(journal_write_delay_ms);
 180 rw_attribute(journal_reclaim_delay_ms);
 181
 182 rw_attribute(discard);
 183 rw_attribute(cache_replacement_policy);
 184 rw_attribute(label);
 185
 186 rw_attribute(copy_gc_enabled);
 187 sysfs_pd_controller_attribute(copy_gc);
 188
 189 rw_attribute(rebalance_enabled);
 190 sysfs_pd_controller_attribute(rebalance);
 191 read_attribute(rebalance_work);
 192 rw_attribute(promote_whole_extents);
 193
 194 read_attribute(new_stripes);
 195
 196 rw_attribute(pd_controllers_update_seconds);
 197
 198 read_attribute(meta_replicas_have);
 199 read_attribute(data_replicas_have);
 200
 201 #ifdef CONFIG_BCACHEFS_TESTS
 202 write_attribute(perf_test);
 203 #endif /* CONFIG_BCACHEFS_TESTS */
 204
 205 #define BCH_DEBUG_PARAM(name, description)                              \
 206         rw_attribute(name);
 207
 208         BCH_DEBUG_PARAMS()
 209 #undef BCH_DEBUG_PARAM
 210
 211 #define x(_name)                                                \
 212         static struct attribute sysfs_time_stat_##_name =               \
 213                 { .name = #_name, .mode = S_IRUGO };
 214         BCH_TIME_STATS()
 215 #undef x
 216
 217 static struct attribute sysfs_state_rw = {
 218         .name = "state",
 219         .mode = S_IRUGO
 220 };
 221
 222 static size_t bch2_btree_cache_size(struct bch_fs *c)
 223 {
 224         size_t ret = 0;
 225         struct btree *b;
 226
 227         mutex_lock(&c->btree_cache.lock);
 228         list_for_each_entry(b, &c->btree_cache.live, list)
 229                 ret += btree_bytes(c);
 230
 231         mutex_unlock(&c->btree_cache.lock);
 232         return ret;
 233 }
 234
 235 static ssize_t show_fs_alloc_debug(struct bch_fs *c, char *buf)
 236 {
 237         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 238         struct bch_fs_usage *fs_usage = bch2_fs_usage_read(c);
 239
 240         if (!fs_usage)
 241                 return -ENOMEM;
 242
 243         bch2_fs_usage_to_text(&out, c, fs_usage);
 244
 245         percpu_up_read(&c->mark_lock);
 246
 247         kfree(fs_usage);
 248
 249         return out.pos - buf;
 250 }
 251
 252 static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 253 {
 254         struct btree_trans trans;
 255         struct btree_iter *iter;
 256         struct bkey_s_c k;
 257         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 258             nr_compressed_extents = 0,
 259             compressed_sectors_compressed = 0,
 260             compressed_sectors_uncompressed = 0;
 261         int ret;
 262
 263         if (!test_bit(BCH_FS_STARTED, &c->flags))
 264                 return -EPERM;
 265
 266         bch2_trans_init(&trans, c, 0, 0);
 267
 268         for_each_btree_key(&trans, iter, BTREE_ID_EXTENTS, POS_MIN, 0, k, ret)
 269                 if (k.k->type == KEY_TYPE_extent) {
 270                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 271                         const union bch_extent_entry *entry;
 272                         struct extent_ptr_decoded p;
 273
 274                         extent_for_each_ptr_decode(e, p, entry) {
 275                                 if (p.crc.compression_type == BCH_COMPRESSION_NONE) {
 276                                         nr_uncompressed_extents++;
 277                                         uncompressed_sectors += e.k->size;
 278                                 } else {
 279                                         nr_compressed_extents++;
 280                                         compressed_sectors_compressed +=
 281                                                 p.crc.compressed_size;
 282                                         compressed_sectors_uncompressed +=
 283                                                 p.crc.uncompressed_size;
 284                                 }
 285
 286                                 /* only looking at the first ptr */
 287                                 break;
 288                         }
 289                 }
 290
 291         ret = bch2_trans_exit(&trans) ?: ret;
 292         if (ret)
 293                 return ret;
 294
 295         return scnprintf(buf, PAGE_SIZE,
 296                         "uncompressed data:\n"
 297                         "       nr extents:                     %llu\n"
 298                         "       size (bytes):                   %llu\n"
 299                         "compressed data:\n"
 300                         "       nr extents:                     %llu\n"
 301                         "       compressed size (bytes):        %llu\n"
 302                         "       uncompressed size (bytes):      %llu\n",
 303                         nr_uncompressed_extents,
 304                         uncompressed_sectors << 9,
 305                         nr_compressed_extents,
 306                         compressed_sectors_compressed << 9,
 307                         compressed_sectors_uncompressed << 9);
 308 }
 309
 310 static ssize_t bch2_new_stripes(struct bch_fs *c, char *buf)
 311 {
 312         char *out = buf, *end = buf + PAGE_SIZE;
 313         struct ec_stripe_head *h;
 314         struct ec_stripe_new *s;
 315
 316         mutex_lock(&c->ec_new_stripe_lock);
 317         list_for_each_entry(h, &c->ec_new_stripe_list, list) {
 318                 out += scnprintf(out, end - out,
 319                                  "target %u algo %u redundancy %u:\n",
 320                                  h->target, h->algo, h->redundancy);
 321
 322                 if (h->s)
 323                         out += scnprintf(out, end - out,
 324                                          "\tpending: blocks %u allocated %u\n",
 325                                          h->s->blocks.nr,
 326                                          bitmap_weight(h->s->blocks_allocated,
 327                                                        h->s->blocks.nr));
 328
 329                 mutex_lock(&h->lock);
 330                 list_for_each_entry(s, &h->stripes, list)
 331                         out += scnprintf(out, end - out,
 332                                          "\tin flight: blocks %u allocated %u pin %u\n",
 333                                          s->blocks.nr,
 334                                          bitmap_weight(s->blocks_allocated,
 335                                                        s->blocks.nr),
 336                                          atomic_read(&s->pin));
 337                 mutex_unlock(&h->lock);
 338
 339         }
 340         mutex_unlock(&c->ec_new_stripe_lock);
 341
 342         return out - buf;
 343 }
 344
 345 SHOW(bch2_fs)
 346 {
 347         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 348
 349         sysfs_print(minor,                      c->minor);
 350         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 351
 352         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 353         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 354
 355         sysfs_print(block_size,                 block_bytes(c));
 356         sysfs_print(btree_node_size,            btree_bytes(c));
 357         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 358
 359         sysfs_print(read_realloc_races,
 360                     atomic_long_read(&c->read_realloc_races));
 361         sysfs_print(extent_migrate_done,
 362                     atomic_long_read(&c->extent_migrate_done));
 363         sysfs_print(extent_migrate_raced,
 364                     atomic_long_read(&c->extent_migrate_raced));
 365
 366         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 367
 368         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 369
 370         sysfs_print(pd_controllers_update_seconds,
 371                     c->pd_controllers_update_seconds);
 372
 373         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 374         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 375
 376         if (attr == &sysfs_rebalance_work)
 377                 return bch2_rebalance_work_show(c, buf);
 378
 379         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 380
 381         sysfs_printf(meta_replicas_have, "%i",  bch2_replicas_online(c, true));
 382         sysfs_printf(data_replicas_have, "%i",  bch2_replicas_online(c, false));
 383
 384         /* Debugging: */
 385
 386         if (attr == &sysfs_alloc_debug)
 387                 return show_fs_alloc_debug(c, buf);
 388
 389         if (attr == &sysfs_journal_debug)
 390                 return bch2_journal_print_debug(&c->journal, buf);
 391
 392         if (attr == &sysfs_journal_pins)
 393                 return bch2_journal_print_pins(&c->journal, buf);
 394
 395         if (attr == &sysfs_btree_updates)
 396                 return bch2_btree_updates_print(c, buf);
 397
 398         if (attr == &sysfs_dirty_btree_nodes)
 399                 return bch2_dirty_btree_nodes_print(c, buf);
 400
 401         if (attr == &sysfs_compression_stats)
 402                 return bch2_compression_stats(c, buf);
 403
 404         if (attr == &sysfs_new_stripes)
 405                 return bch2_new_stripes(c, buf);
 406
 407 #define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
 408         BCH_DEBUG_PARAMS()
 409 #undef BCH_DEBUG_PARAM
 410
 411         return 0;
 412 }
 413
 414 STORE(__bch2_fs)
 415 {
 416         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 417
 418         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 419         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 420
 421         if (attr == &sysfs_btree_gc_periodic) {
 422                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 423                         ?: (ssize_t) size;
 424
 425                 wake_up_process(c->gc_thread);
 426                 return ret;
 427         }
 428
 429         if (attr == &sysfs_copy_gc_enabled) {
 430                 struct bch_dev *ca;
 431                 unsigned i;
 432                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 433                         ?: (ssize_t) size;
 434
 435                 for_each_member_device(ca, c, i)
 436                         if (ca->copygc_thread)
 437                                 wake_up_process(ca->copygc_thread);
 438                 return ret;
 439         }
 440
 441         if (attr == &sysfs_rebalance_enabled) {
 442                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 443                         ?: (ssize_t) size;
 444
 445                 rebalance_wakeup(c);
 446                 return ret;
 447         }
 448
 449         sysfs_strtoul(pd_controllers_update_seconds,
 450                       c->pd_controllers_update_seconds);
 451         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 452
 453         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 454
 455         /* Debugging: */
 456
 457 #define BCH_DEBUG_PARAM(name, description) sysfs_strtoul(name, c->name);
 458         BCH_DEBUG_PARAMS()
 459 #undef BCH_DEBUG_PARAM
 460
 461         if (!test_bit(BCH_FS_STARTED, &c->flags))
 462                 return -EPERM;
 463
 464         /* Debugging: */
 465
 466         if (attr == &sysfs_trigger_journal_flush)
 467                 bch2_journal_meta_async(&c->journal, NULL);
 468
 469         if (attr == &sysfs_trigger_btree_coalesce)
 470                 bch2_coalesce(c);
 471
 472         if (attr == &sysfs_trigger_gc)
 473                 bch2_gc(c, NULL, false, false);
 474
 475         if (attr == &sysfs_trigger_alloc_write) {
 476                 bool wrote;
 477
 478                 bch2_alloc_write(c, 0, &wrote);
 479         }
 480
 481         if (attr == &sysfs_prune_cache) {
 482                 struct shrink_control sc;
 483
 484                 sc.gfp_mask = GFP_KERNEL;
 485                 sc.nr_to_scan = strtoul_or_return(buf);
 486                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 487         }
 488 #ifdef CONFIG_BCACHEFS_TESTS
 489         if (attr == &sysfs_perf_test) {
 490                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 491                 char *test              = strsep(&p, " \t\n");
 492                 char *nr_str            = strsep(&p, " \t\n");
 493                 char *threads_str       = strsep(&p, " \t\n");
 494                 unsigned threads;
 495                 u64 nr;
 496                 int ret = -EINVAL;
 497
 498                 if (threads_str &&
 499                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 500                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 501                         bch2_btree_perf_test(c, test, nr, threads);
 502                 else
 503                         size = ret;
 504                 kfree(tmp);
 505         }
 506 #endif
 507         return size;
 508 }
 509
 510 STORE(bch2_fs)
 511 {
 512         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 513
 514         mutex_lock(&c->state_lock);
 515         size = __bch2_fs_store(kobj, attr, buf, size);
 516         mutex_unlock(&c->state_lock);
 517
 518         return size;
 519 }
 520 SYSFS_OPS(bch2_fs);
 521
 522 struct attribute *bch2_fs_files[] = {
 523         &sysfs_minor,
 524         &sysfs_block_size,
 525         &sysfs_btree_node_size,
 526         &sysfs_btree_cache_size,
 527
 528         &sysfs_meta_replicas_have,
 529         &sysfs_data_replicas_have,
 530
 531         &sysfs_journal_write_delay_ms,
 532         &sysfs_journal_reclaim_delay_ms,
 533
 534         &sysfs_promote_whole_extents,
 535
 536         &sysfs_compression_stats,
 537
 538 #ifdef CONFIG_BCACHEFS_TESTS
 539         &sysfs_perf_test,
 540 #endif
 541         NULL
 542 };
 543
 544 /* internal dir - just a wrapper */
 545
 546 SHOW(bch2_fs_internal)
 547 {
 548         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 549         return bch2_fs_show(&c->kobj, attr, buf);
 550 }
 551
 552 STORE(bch2_fs_internal)
 553 {
 554         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 555         return bch2_fs_store(&c->kobj, attr, buf, size);
 556 }
 557 SYSFS_OPS(bch2_fs_internal);
 558
 559 struct attribute *bch2_fs_internal_files[] = {
 560         &sysfs_alloc_debug,
 561         &sysfs_journal_debug,
 562         &sysfs_journal_pins,
 563         &sysfs_btree_updates,
 564         &sysfs_dirty_btree_nodes,
 565
 566         &sysfs_read_realloc_races,
 567         &sysfs_extent_migrate_done,
 568         &sysfs_extent_migrate_raced,
 569
 570         &sysfs_trigger_journal_flush,
 571         &sysfs_trigger_btree_coalesce,
 572         &sysfs_trigger_gc,
 573         &sysfs_trigger_alloc_write,
 574         &sysfs_prune_cache,
 575
 576         &sysfs_copy_gc_enabled,
 577
 578         &sysfs_rebalance_enabled,
 579         &sysfs_rebalance_work,
 580         sysfs_pd_controller_files(rebalance),
 581
 582         &sysfs_new_stripes,
 583
 584         &sysfs_internal_uuid,
 585
 586 #define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
 587         BCH_DEBUG_PARAMS()
 588 #undef BCH_DEBUG_PARAM
 589
 590         NULL
 591 };
 592
 593 /* options */
 594
 595 SHOW(bch2_fs_opts_dir)
 596 {
 597         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 598         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 599         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 600         int id = opt - bch2_opt_table;
 601         u64 v = bch2_opt_get_by_id(&c->opts, id);
 602
 603         bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
 604         pr_buf(&out, "\n");
 605
 606         return out.pos - buf;
 607 }
 608
 609 STORE(bch2_fs_opts_dir)
 610 {
 611         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 612         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 613         int ret, id = opt - bch2_opt_table;
 614         char *tmp;
 615         u64 v;
 616
 617         tmp = kstrdup(buf, GFP_KERNEL);
 618         if (!tmp)
 619                 return -ENOMEM;
 620
 621         ret = bch2_opt_parse(c, opt, strim(tmp), &v);
 622         kfree(tmp);
 623
 624         if (ret < 0)
 625                 return ret;
 626
 627         ret = bch2_opt_check_may_set(c, id, v);
 628         if (ret < 0)
 629                 return ret;
 630
 631         if (opt->set_sb != SET_NO_SB_OPT) {
 632                 mutex_lock(&c->sb_lock);
 633                 opt->set_sb(c->disk_sb.sb, v);
 634                 bch2_write_super(c);
 635                 mutex_unlock(&c->sb_lock);
 636         }
 637
 638         bch2_opt_set_by_id(&c->opts, id, v);
 639
 640         if ((id == Opt_background_target ||
 641              id == Opt_background_compression) && v) {
 642                 bch2_rebalance_add_work(c, S64_MAX);
 643                 rebalance_wakeup(c);
 644         }
 645
 646         return size;
 647 }
 648 SYSFS_OPS(bch2_fs_opts_dir);
 649
 650 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 651
 652 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 653 {
 654         const struct bch_option *i;
 655         int ret;
 656
 657         for (i = bch2_opt_table;
 658              i < bch2_opt_table + bch2_opts_nr;
 659              i++) {
 660                 if (!(i->mode & (OPT_FORMAT|OPT_MOUNT|OPT_RUNTIME)))
 661                         continue;
 662
 663                 ret = sysfs_create_file(kobj, &i->attr);
 664                 if (ret)
 665                         return ret;
 666         }
 667
 668         return 0;
 669 }
 670
 671 /* time stats */
 672
 673 SHOW(bch2_fs_time_stats)
 674 {
 675         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 676
 677 #define x(name)                                         \
 678         if (attr == &sysfs_time_stat_##name)                            \
 679                 return bch2_time_stats_print(&c->times[BCH_TIME_##name],\
 680                                              buf, PAGE_SIZE);
 681         BCH_TIME_STATS()
 682 #undef x
 683
 684         return 0;
 685 }
 686
 687 STORE(bch2_fs_time_stats)
 688 {
 689         return size;
 690 }
 691 SYSFS_OPS(bch2_fs_time_stats);
 692
 693 struct attribute *bch2_fs_time_stats_files[] = {
 694 #define x(name)                                         \
 695         &sysfs_time_stat_##name,
 696         BCH_TIME_STATS()
 697 #undef x
 698         NULL
 699 };
 700
 701 typedef unsigned (bucket_map_fn)(struct bch_fs *, struct bch_dev *,
 702                                  size_t, void *);
 703
 704 static unsigned bucket_last_io_fn(struct bch_fs *c, struct bch_dev *ca,
 705                                   size_t b, void *private)
 706 {
 707         int rw = (private ? 1 : 0);
 708
 709         return bucket_last_io(c, bucket(ca, b), rw);
 710 }
 711
 712 static unsigned bucket_sectors_used_fn(struct bch_fs *c, struct bch_dev *ca,
 713                                        size_t b, void *private)
 714 {
 715         struct bucket *g = bucket(ca, b);
 716         return bucket_sectors_used(g->mark);
 717 }
 718
 719 static unsigned bucket_oldest_gen_fn(struct bch_fs *c, struct bch_dev *ca,
 720                                      size_t b, void *private)
 721 {
 722         return bucket_gc_gen(ca, b);
 723 }
 724
 725 static int unsigned_cmp(const void *_l, const void *_r)
 726 {
 727         const unsigned *l = _l;
 728         const unsigned *r = _r;
 729
 730         return cmp_int(*l, *r);
 731 }
 732
 733 static ssize_t show_quantiles(struct bch_fs *c, struct bch_dev *ca,
 734                               char *buf, bucket_map_fn *fn, void *private)
 735 {
 736         size_t i, n;
 737         /* Compute 31 quantiles */
 738         unsigned q[31], *p;
 739         ssize_t ret = 0;
 740
 741         down_read(&ca->bucket_lock);
 742         n = ca->mi.nbuckets;
 743
 744         p = vzalloc(n * sizeof(unsigned));
 745         if (!p) {
 746                 up_read(&ca->bucket_lock);
 747                 return -ENOMEM;
 748         }
 749
 750         for (i = ca->mi.first_bucket; i < n; i++)
 751                 p[i] = fn(c, ca, i, private);
 752
 753         sort(p, n, sizeof(unsigned), unsigned_cmp, NULL);
 754         up_read(&ca->bucket_lock);
 755
 756         while (n &&
 757                !p[n - 1])
 758                 --n;
 759
 760         for (i = 0; i < ARRAY_SIZE(q); i++)
 761                 q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
 762
 763         vfree(p);
 764
 765         for (i = 0; i < ARRAY_SIZE(q); i++)
 766                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 767                                  "%u ", q[i]);
 768         buf[ret - 1] = '\n';
 769
 770         return ret;
 771 }
 772
 773 static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
 774 {
 775         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 776         enum alloc_reserve i;
 777
 778         spin_lock(&ca->fs->freelist_lock);
 779
 780         pr_buf(&out, "free_inc:\t%zu\t%zu\n",
 781                fifo_used(&ca->free_inc),
 782                ca->free_inc.size);
 783
 784         for (i = 0; i < RESERVE_NR; i++)
 785                 pr_buf(&out, "free[%u]:\t%zu\t%zu\n", i,
 786                        fifo_used(&ca->free[i]),
 787                        ca->free[i].size);
 788
 789         spin_unlock(&ca->fs->freelist_lock);
 790
 791         return out.pos - buf;
 792 }
 793
 794 static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
 795 {
 796         struct bch_fs *c = ca->fs;
 797         struct bch_dev_usage stats = bch2_dev_usage_read(c, ca);
 798         unsigned i, nr[BCH_DATA_NR];
 799
 800         memset(nr, 0, sizeof(nr));
 801
 802         for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 803                 nr[c->open_buckets[i].type]++;
 804
 805         return scnprintf(buf, PAGE_SIZE,
 806                 "free_inc:               %zu/%zu\n"
 807                 "free[RESERVE_BTREE]:    %zu/%zu\n"
 808                 "free[RESERVE_MOVINGGC]: %zu/%zu\n"
 809                 "free[RESERVE_NONE]:     %zu/%zu\n"
 810                 "buckets:\n"
 811                 "    capacity:           %llu\n"
 812                 "    alloc:              %llu\n"
 813                 "    sb:                 %llu\n"
 814                 "    journal:            %llu\n"
 815                 "    meta:               %llu\n"
 816                 "    user:               %llu\n"
 817                 "    cached:             %llu\n"
 818                 "    erasure coded:      %llu\n"
 819                 "    available:          %lli\n"
 820                 "sectors:\n"
 821                 "    sb:                 %llu\n"
 822                 "    journal:            %llu\n"
 823                 "    meta:               %llu\n"
 824                 "    user:               %llu\n"
 825                 "    cached:             %llu\n"
 826                 "    fragmented:         %llu\n"
 827                 "    copygc threshold:   %llu\n"
 828                 "freelist_wait:          %s\n"
 829                 "open buckets:           %u/%u (reserved %u)\n"
 830                 "open_buckets_wait:      %s\n"
 831                 "open_buckets_btree:     %u\n"
 832                 "open_buckets_user:      %u\n"
 833                 "btree reserve cache:    %u\n",
 834                 fifo_used(&ca->free_inc),               ca->free_inc.size,
 835                 fifo_used(&ca->free[RESERVE_BTREE]),    ca->free[RESERVE_BTREE].size,
 836                 fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size,
 837                 fifo_used(&ca->free[RESERVE_NONE]),     ca->free[RESERVE_NONE].size,
 838                 ca->mi.nbuckets - ca->mi.first_bucket,
 839                 stats.buckets_alloc,
 840                 stats.buckets[BCH_DATA_SB],
 841                 stats.buckets[BCH_DATA_JOURNAL],
 842                 stats.buckets[BCH_DATA_BTREE],
 843                 stats.buckets[BCH_DATA_USER],
 844                 stats.buckets[BCH_DATA_CACHED],
 845                 stats.buckets_ec,
 846                 ca->mi.nbuckets - ca->mi.first_bucket - stats.buckets_unavailable,
 847                 stats.sectors[BCH_DATA_SB],
 848                 stats.sectors[BCH_DATA_JOURNAL],
 849                 stats.sectors[BCH_DATA_BTREE],
 850                 stats.sectors[BCH_DATA_USER],
 851                 stats.sectors[BCH_DATA_CACHED],
 852                 stats.sectors_fragmented,
 853                 ca->copygc_threshold,
 854                 c->freelist_wait.list.first             ? "waiting" : "empty",
 855                 c->open_buckets_nr_free, OPEN_BUCKETS_COUNT,
 856                 BTREE_NODE_OPEN_BUCKET_RESERVE,
 857                 c->open_buckets_wait.list.first         ? "waiting" : "empty",
 858                 nr[BCH_DATA_BTREE],
 859                 nr[BCH_DATA_USER],
 860                 c->btree_reserve_cache_nr);
 861 }
 862
 863 static const char * const bch2_rw[] = {
 864         "read",
 865         "write",
 866         NULL
 867 };
 868
 869 static ssize_t show_dev_iodone(struct bch_dev *ca, char *buf)
 870 {
 871         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 872         int rw, i;
 873
 874         for (rw = 0; rw < 2; rw++) {
 875                 pr_buf(&out, "%s:\n", bch2_rw[rw]);
 876
 877                 for (i = 1; i < BCH_DATA_NR; i++)
 878                         pr_buf(&out, "%-12s:%12llu\n",
 879                                bch2_data_types[i],
 880                                percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
 881         }
 882
 883         return out.pos - buf;
 884 }
 885
 886 SHOW(bch2_dev)
 887 {
 888         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 889         struct bch_fs *c = ca->fs;
 890         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 891
 892         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 893
 894         sysfs_print(bucket_size,        bucket_bytes(ca));
 895         sysfs_print(block_size,         block_bytes(c));
 896         sysfs_print(first_bucket,       ca->mi.first_bucket);
 897         sysfs_print(nbuckets,           ca->mi.nbuckets);
 898         sysfs_print(durability,         ca->mi.durability);
 899         sysfs_print(discard,            ca->mi.discard);
 900
 901         if (attr == &sysfs_label) {
 902                 if (ca->mi.group) {
 903                         mutex_lock(&c->sb_lock);
 904                         bch2_disk_path_to_text(&out, &c->disk_sb,
 905                                                ca->mi.group - 1);
 906                         mutex_unlock(&c->sb_lock);
 907                 } else {
 908                         pr_buf(&out, "none");
 909                 }
 910
 911                 pr_buf(&out, "\n");
 912                 return out.pos - buf;
 913         }
 914
 915         if (attr == &sysfs_has_data) {
 916                 bch2_flags_to_text(&out, bch2_data_types,
 917                                    bch2_dev_has_data(c, ca));
 918                 pr_buf(&out, "\n");
 919                 return out.pos - buf;
 920         }
 921
 922         sysfs_pd_controller_show(copy_gc, &ca->copygc_pd);
 923
 924         if (attr == &sysfs_cache_replacement_policy) {
 925                 bch2_string_opt_to_text(&out,
 926                                         bch2_cache_replacement_policies,
 927                                         ca->mi.replacement);
 928                 pr_buf(&out, "\n");
 929                 return out.pos - buf;
 930         }
 931
 932         if (attr == &sysfs_state_rw) {
 933                 bch2_string_opt_to_text(&out, bch2_dev_state,
 934                                         ca->mi.state);
 935                 pr_buf(&out, "\n");
 936                 return out.pos - buf;
 937         }
 938
 939         if (attr == &sysfs_iodone)
 940                 return show_dev_iodone(ca, buf);
 941
 942         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 943         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 944
 945         if (attr == &sysfs_io_latency_stats_read)
 946                 return bch2_time_stats_print(&ca->io_latency[READ], buf, PAGE_SIZE);
 947         if (attr == &sysfs_io_latency_stats_write)
 948                 return bch2_time_stats_print(&ca->io_latency[WRITE], buf, PAGE_SIZE);
 949
 950         sysfs_printf(congested,                 "%u%%",
 951                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 952                      * 100 / CONGESTED_MAX);
 953
 954         if (attr == &sysfs_bucket_quantiles_last_read)
 955                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 0);
 956         if (attr == &sysfs_bucket_quantiles_last_write)
 957                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 1);
 958         if (attr == &sysfs_bucket_quantiles_fragmentation)
 959                 return show_quantiles(c, ca, buf, bucket_sectors_used_fn, NULL);
 960         if (attr == &sysfs_bucket_quantiles_oldest_gen)
 961                 return show_quantiles(c, ca, buf, bucket_oldest_gen_fn, NULL);
 962
 963         if (attr == &sysfs_reserve_stats)
 964                 return show_reserve_stats(ca, buf);
 965         if (attr == &sysfs_alloc_debug)
 966                 return show_dev_alloc_debug(ca, buf);
 967
 968         return 0;
 969 }
 970
 971 STORE(bch2_dev)
 972 {
 973         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 974         struct bch_fs *c = ca->fs;
 975         struct bch_member *mi;
 976
 977         sysfs_pd_controller_store(copy_gc, &ca->copygc_pd);
 978
 979         if (attr == &sysfs_discard) {
 980                 bool v = strtoul_or_return(buf);
 981
 982                 mutex_lock(&c->sb_lock);
 983                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 984
 985                 if (v != BCH_MEMBER_DISCARD(mi)) {
 986                         SET_BCH_MEMBER_DISCARD(mi, v);
 987                         bch2_write_super(c);
 988                 }
 989                 mutex_unlock(&c->sb_lock);
 990         }
 991
 992         if (attr == &sysfs_cache_replacement_policy) {
 993                 ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
 994
 995                 if (v < 0)
 996                         return v;
 997
 998                 mutex_lock(&c->sb_lock);
 999                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
1000
1001                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
1002                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
1003                         bch2_write_super(c);
1004                 }
1005                 mutex_unlock(&c->sb_lock);
1006         }
1007
1008         if (attr == &sysfs_label) {
1009                 char *tmp;
1010                 int ret;
1011
1012                 tmp = kstrdup(buf, GFP_KERNEL);
1013                 if (!tmp)
1014                         return -ENOMEM;
1015
1016                 ret = bch2_dev_group_set(c, ca, strim(tmp));
1017                 kfree(tmp);
1018                 if (ret)
1019                         return ret;
1020         }
1021
1022         if (attr == &sysfs_wake_allocator)
1023                 bch2_wake_allocator(ca);
1024
1025         return size;
1026 }
1027 SYSFS_OPS(bch2_dev);
1028
1029 struct attribute *bch2_dev_files[] = {
1030         &sysfs_uuid,
1031         &sysfs_bucket_size,
1032         &sysfs_block_size,
1033         &sysfs_first_bucket,
1034         &sysfs_nbuckets,
1035         &sysfs_durability,
1036
1037         /* settings: */
1038         &sysfs_discard,
1039         &sysfs_cache_replacement_policy,
1040         &sysfs_state_rw,
1041         &sysfs_label,
1042
1043         &sysfs_has_data,
1044         &sysfs_iodone,
1045
1046         &sysfs_io_latency_read,
1047         &sysfs_io_latency_write,
1048         &sysfs_io_latency_stats_read,
1049         &sysfs_io_latency_stats_write,
1050         &sysfs_congested,
1051
1052         /* alloc info - other stats: */
1053         &sysfs_bucket_quantiles_last_read,
1054         &sysfs_bucket_quantiles_last_write,
1055         &sysfs_bucket_quantiles_fragmentation,
1056         &sysfs_bucket_quantiles_oldest_gen,
1057
1058         &sysfs_reserve_stats,
1059
1060         /* debug: */
1061         &sysfs_alloc_debug,
1062         &sysfs_wake_allocator,
1063
1064         sysfs_pd_controller_files(copy_gc),
1065         NULL
1066 };
1067
1068 #endif  /* _BCACHEFS_SYSFS_H_ */