git.sesse.net Git - bcachefs-tools-debian/blob - libbcachefs/sysfs.c

   1 /*
   2  * bcache sysfs interfaces
   3  *
   4  * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
   5  * Copyright 2012 Google, Inc.
   6  */
   7
   8 #ifndef NO_BCACHEFS_SYSFS
   9
  10 #include "bcachefs.h"
  11 #include "alloc_background.h"
  12 #include "sysfs.h"
  13 #include "btree_cache.h"
  14 #include "btree_io.h"
  15 #include "btree_iter.h"
  16 #include "btree_update.h"
  17 #include "btree_update_interior.h"
  18 #include "btree_gc.h"
  19 #include "buckets.h"
  20 #include "disk_groups.h"
  21 #include "ec.h"
  22 #include "inode.h"
  23 #include "journal.h"
  24 #include "keylist.h"
  25 #include "move.h"
  26 #include "opts.h"
  27 #include "rebalance.h"
  28 #include "replicas.h"
  29 #include "super-io.h"
  30 #include "tests.h"
  31
  32 #include <linux/blkdev.h>
  33 #include <linux/sort.h>
  34 #include <linux/sched/clock.h>
  35
  36 #include "util.h"
  37
  38 #define SYSFS_OPS(type)                                                 \
  39 struct sysfs_ops type ## _sysfs_ops = {                                 \
  40         .show   = type ## _show,                                        \
  41         .store  = type ## _store                                        \
  42 }
  43
  44 #define SHOW(fn)                                                        \
  45 static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
  46                            char *buf)                                   \
  47
  48 #define STORE(fn)                                                       \
  49 static ssize_t fn ## _store(struct kobject *kobj, struct attribute *attr,\
  50                             const char *buf, size_t size)               \
  51
  52 #define __sysfs_attribute(_name, _mode)                                 \
  53         static struct attribute sysfs_##_name =                         \
  54                 { .name = #_name, .mode = _mode }
  55
  56 #define write_attribute(n)      __sysfs_attribute(n, S_IWUSR)
  57 #define read_attribute(n)       __sysfs_attribute(n, S_IRUGO)
  58 #define rw_attribute(n)         __sysfs_attribute(n, S_IRUGO|S_IWUSR)
  59
  60 #define sysfs_printf(file, fmt, ...)                                    \
  61 do {                                                                    \
  62         if (attr == &sysfs_ ## file)                                    \
  63                 return scnprintf(buf, PAGE_SIZE, fmt "\n", __VA_ARGS__);\
  64 } while (0)
  65
  66 #define sysfs_print(file, var)                                          \
  67 do {                                                                    \
  68         if (attr == &sysfs_ ## file)                                    \
  69                 return snprint(buf, PAGE_SIZE, var);                    \
  70 } while (0)
  71
  72 #define sysfs_hprint(file, val)                                         \
  73 do {                                                                    \
  74         if (attr == &sysfs_ ## file) {                                  \
  75                 struct printbuf out = _PBUF(buf, PAGE_SIZE);            \
  76                 bch2_hprint(&out, val);                                 \
  77                 pr_buf(&out, "\n");                                     \
  78                 return out.pos - buf;                                   \
  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(trigger_alloc_write);
 136 write_attribute(prune_cache);
 137 rw_attribute(btree_gc_periodic);
 138
 139 read_attribute(uuid);
 140 read_attribute(minor);
 141 read_attribute(bucket_size);
 142 read_attribute(block_size);
 143 read_attribute(btree_node_size);
 144 read_attribute(first_bucket);
 145 read_attribute(nbuckets);
 146 read_attribute(durability);
 147 read_attribute(iodone);
 148
 149 read_attribute(io_latency_read);
 150 read_attribute(io_latency_write);
 151 read_attribute(io_latency_stats_read);
 152 read_attribute(io_latency_stats_write);
 153 read_attribute(congested);
 154
 155 read_attribute(bucket_quantiles_last_read);
 156 read_attribute(bucket_quantiles_last_write);
 157 read_attribute(bucket_quantiles_fragmentation);
 158 read_attribute(bucket_quantiles_oldest_gen);
 159
 160 read_attribute(reserve_stats);
 161 read_attribute(btree_cache_size);
 162 read_attribute(compression_stats);
 163 read_attribute(journal_debug);
 164 read_attribute(journal_pins);
 165 read_attribute(btree_updates);
 166 read_attribute(dirty_btree_nodes);
 167
 168 read_attribute(internal_uuid);
 169
 170 read_attribute(has_data);
 171 read_attribute(alloc_debug);
 172 write_attribute(wake_allocator);
 173
 174 read_attribute(read_realloc_races);
 175 read_attribute(extent_migrate_done);
 176 read_attribute(extent_migrate_raced);
 177
 178 rw_attribute(journal_write_delay_ms);
 179 rw_attribute(journal_reclaim_delay_ms);
 180
 181 rw_attribute(discard);
 182 rw_attribute(cache_replacement_policy);
 183 rw_attribute(label);
 184
 185 rw_attribute(copy_gc_enabled);
 186 sysfs_pd_controller_attribute(copy_gc);
 187
 188 rw_attribute(rebalance_enabled);
 189 sysfs_pd_controller_attribute(rebalance);
 190 read_attribute(rebalance_work);
 191 rw_attribute(promote_whole_extents);
 192
 193 read_attribute(new_stripes);
 194
 195 rw_attribute(pd_controllers_update_seconds);
 196
 197 read_attribute(meta_replicas_have);
 198 read_attribute(data_replicas_have);
 199
 200 #ifdef CONFIG_BCACHEFS_TESTS
 201 write_attribute(perf_test);
 202 #endif /* CONFIG_BCACHEFS_TESTS */
 203
 204 #define BCH_DEBUG_PARAM(name, description)                              \
 205         rw_attribute(name);
 206
 207         BCH_DEBUG_PARAMS()
 208 #undef BCH_DEBUG_PARAM
 209
 210 #define x(_name)                                                \
 211         static struct attribute sysfs_time_stat_##_name =               \
 212                 { .name = #_name, .mode = S_IRUGO };
 213         BCH_TIME_STATS()
 214 #undef x
 215
 216 static struct attribute sysfs_state_rw = {
 217         .name = "state",
 218         .mode = S_IRUGO
 219 };
 220
 221 static size_t bch2_btree_cache_size(struct bch_fs *c)
 222 {
 223         size_t ret = 0;
 224         struct btree *b;
 225
 226         mutex_lock(&c->btree_cache.lock);
 227         list_for_each_entry(b, &c->btree_cache.live, list)
 228                 ret += btree_bytes(c);
 229
 230         mutex_unlock(&c->btree_cache.lock);
 231         return ret;
 232 }
 233
 234 static ssize_t show_fs_alloc_debug(struct bch_fs *c, char *buf)
 235 {
 236         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 237         struct bch_fs_usage *fs_usage = bch2_fs_usage_read(c);
 238
 239         if (!fs_usage)
 240                 return -ENOMEM;
 241
 242         bch2_fs_usage_to_text(&out, c, fs_usage);
 243
 244         percpu_up_read_preempt_enable(&c->mark_lock);
 245
 246         kfree(fs_usage);
 247
 248         return out.pos - buf;
 249 }
 250
 251 static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 252 {
 253         struct btree_trans trans;
 254         struct btree_iter *iter;
 255         struct bkey_s_c k;
 256         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 257             nr_compressed_extents = 0,
 258             compressed_sectors_compressed = 0,
 259             compressed_sectors_uncompressed = 0;
 260         int ret;
 261
 262         if (!test_bit(BCH_FS_STARTED, &c->flags))
 263                 return -EPERM;
 264
 265         bch2_trans_init(&trans, c, 0, 0);
 266
 267         for_each_btree_key(&trans, iter, BTREE_ID_EXTENTS, POS_MIN, 0, k, ret)
 268                 if (k.k->type == KEY_TYPE_extent) {
 269                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 270                         const union bch_extent_entry *entry;
 271                         struct extent_ptr_decoded p;
 272
 273                         extent_for_each_ptr_decode(e, p, entry) {
 274                                 if (p.crc.compression_type == BCH_COMPRESSION_NONE) {
 275                                         nr_uncompressed_extents++;
 276                                         uncompressed_sectors += e.k->size;
 277                                 } else {
 278                                         nr_compressed_extents++;
 279                                         compressed_sectors_compressed +=
 280                                                 p.crc.compressed_size;
 281                                         compressed_sectors_uncompressed +=
 282                                                 p.crc.uncompressed_size;
 283                                 }
 284
 285                                 /* only looking at the first ptr */
 286                                 break;
 287                         }
 288                 }
 289
 290         ret = bch2_trans_exit(&trans) ?: ret;
 291         if (ret)
 292                 return ret;
 293
 294         return scnprintf(buf, PAGE_SIZE,
 295                         "uncompressed data:\n"
 296                         "       nr extents:                     %llu\n"
 297                         "       size (bytes):                   %llu\n"
 298                         "compressed data:\n"
 299                         "       nr extents:                     %llu\n"
 300                         "       compressed size (bytes):        %llu\n"
 301                         "       uncompressed size (bytes):      %llu\n",
 302                         nr_uncompressed_extents,
 303                         uncompressed_sectors << 9,
 304                         nr_compressed_extents,
 305                         compressed_sectors_compressed << 9,
 306                         compressed_sectors_uncompressed << 9);
 307 }
 308
 309 static ssize_t bch2_new_stripes(struct bch_fs *c, char *buf)
 310 {
 311         char *out = buf, *end = buf + PAGE_SIZE;
 312         struct ec_stripe_head *h;
 313         struct ec_stripe_new *s;
 314
 315         mutex_lock(&c->ec_new_stripe_lock);
 316         list_for_each_entry(h, &c->ec_new_stripe_list, list) {
 317                 out += scnprintf(out, end - out,
 318                                  "target %u algo %u redundancy %u:\n",
 319                                  h->target, h->algo, h->redundancy);
 320
 321                 if (h->s)
 322                         out += scnprintf(out, end - out,
 323                                          "\tpending: blocks %u allocated %u\n",
 324                                          h->s->blocks.nr,
 325                                          bitmap_weight(h->s->blocks_allocated,
 326                                                        h->s->blocks.nr));
 327
 328                 mutex_lock(&h->lock);
 329                 list_for_each_entry(s, &h->stripes, list)
 330                         out += scnprintf(out, end - out,
 331                                          "\tin flight: blocks %u allocated %u pin %u\n",
 332                                          s->blocks.nr,
 333                                          bitmap_weight(s->blocks_allocated,
 334                                                        s->blocks.nr),
 335                                          atomic_read(&s->pin));
 336                 mutex_unlock(&h->lock);
 337
 338         }
 339         mutex_unlock(&c->ec_new_stripe_lock);
 340
 341         return out - buf;
 342 }
 343
 344 SHOW(bch2_fs)
 345 {
 346         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 347
 348         sysfs_print(minor,                      c->minor);
 349         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 350
 351         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 352         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 353
 354         sysfs_print(block_size,                 block_bytes(c));
 355         sysfs_print(btree_node_size,            btree_bytes(c));
 356         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 357
 358         sysfs_print(read_realloc_races,
 359                     atomic_long_read(&c->read_realloc_races));
 360         sysfs_print(extent_migrate_done,
 361                     atomic_long_read(&c->extent_migrate_done));
 362         sysfs_print(extent_migrate_raced,
 363                     atomic_long_read(&c->extent_migrate_raced));
 364
 365         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 366
 367         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 368
 369         sysfs_print(pd_controllers_update_seconds,
 370                     c->pd_controllers_update_seconds);
 371
 372         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 373         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 374
 375         if (attr == &sysfs_rebalance_work)
 376                 return bch2_rebalance_work_show(c, buf);
 377
 378         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 379
 380         sysfs_printf(meta_replicas_have, "%i",  bch2_replicas_online(c, true));
 381         sysfs_printf(data_replicas_have, "%i",  bch2_replicas_online(c, false));
 382
 383         /* Debugging: */
 384
 385         if (attr == &sysfs_alloc_debug)
 386                 return show_fs_alloc_debug(c, buf);
 387
 388         if (attr == &sysfs_journal_debug)
 389                 return bch2_journal_print_debug(&c->journal, buf);
 390
 391         if (attr == &sysfs_journal_pins)
 392                 return bch2_journal_print_pins(&c->journal, buf);
 393
 394         if (attr == &sysfs_btree_updates)
 395                 return bch2_btree_updates_print(c, buf);
 396
 397         if (attr == &sysfs_dirty_btree_nodes)
 398                 return bch2_dirty_btree_nodes_print(c, buf);
 399
 400         if (attr == &sysfs_compression_stats)
 401                 return bch2_compression_stats(c, buf);
 402
 403         if (attr == &sysfs_new_stripes)
 404                 return bch2_new_stripes(c, buf);
 405
 406 #define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
 407         BCH_DEBUG_PARAMS()
 408 #undef BCH_DEBUG_PARAM
 409
 410         return 0;
 411 }
 412
 413 STORE(__bch2_fs)
 414 {
 415         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 416
 417         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 418         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 419
 420         if (attr == &sysfs_btree_gc_periodic) {
 421                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 422                         ?: (ssize_t) size;
 423
 424                 wake_up_process(c->gc_thread);
 425                 return ret;
 426         }
 427
 428         if (attr == &sysfs_copy_gc_enabled) {
 429                 struct bch_dev *ca;
 430                 unsigned i;
 431                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 432                         ?: (ssize_t) size;
 433
 434                 for_each_member_device(ca, c, i)
 435                         if (ca->copygc_thread)
 436                                 wake_up_process(ca->copygc_thread);
 437                 return ret;
 438         }
 439
 440         if (attr == &sysfs_rebalance_enabled) {
 441                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 442                         ?: (ssize_t) size;
 443
 444                 rebalance_wakeup(c);
 445                 return ret;
 446         }
 447
 448         sysfs_strtoul(pd_controllers_update_seconds,
 449                       c->pd_controllers_update_seconds);
 450         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 451
 452         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 453
 454         /* Debugging: */
 455
 456 #define BCH_DEBUG_PARAM(name, description) sysfs_strtoul(name, c->name);
 457         BCH_DEBUG_PARAMS()
 458 #undef BCH_DEBUG_PARAM
 459
 460         if (!test_bit(BCH_FS_STARTED, &c->flags))
 461                 return -EPERM;
 462
 463         /* Debugging: */
 464
 465         if (attr == &sysfs_trigger_journal_flush)
 466                 bch2_journal_meta_async(&c->journal, NULL);
 467
 468         if (attr == &sysfs_trigger_btree_coalesce)
 469                 bch2_coalesce(c);
 470
 471         if (attr == &sysfs_trigger_gc)
 472                 bch2_gc(c, NULL, false, false);
 473
 474         if (attr == &sysfs_trigger_alloc_write) {
 475                 bool wrote;
 476
 477                 bch2_alloc_write(c, 0, &wrote);
 478         }
 479
 480         if (attr == &sysfs_prune_cache) {
 481                 struct shrink_control sc;
 482
 483                 sc.gfp_mask = GFP_KERNEL;
 484                 sc.nr_to_scan = strtoul_or_return(buf);
 485                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 486         }
 487 #ifdef CONFIG_BCACHEFS_TESTS
 488         if (attr == &sysfs_perf_test) {
 489                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 490                 char *test              = strsep(&p, " \t\n");
 491                 char *nr_str            = strsep(&p, " \t\n");
 492                 char *threads_str       = strsep(&p, " \t\n");
 493                 unsigned threads;
 494                 u64 nr;
 495                 int ret = -EINVAL;
 496
 497                 if (threads_str &&
 498                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 499                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 500                         bch2_btree_perf_test(c, test, nr, threads);
 501                 else
 502                         size = ret;
 503                 kfree(tmp);
 504         }
 505 #endif
 506         return size;
 507 }
 508
 509 STORE(bch2_fs)
 510 {
 511         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 512
 513         mutex_lock(&c->state_lock);
 514         size = __bch2_fs_store(kobj, attr, buf, size);
 515         mutex_unlock(&c->state_lock);
 516
 517         return size;
 518 }
 519 SYSFS_OPS(bch2_fs);
 520
 521 struct attribute *bch2_fs_files[] = {
 522         &sysfs_minor,
 523         &sysfs_block_size,
 524         &sysfs_btree_node_size,
 525         &sysfs_btree_cache_size,
 526
 527         &sysfs_meta_replicas_have,
 528         &sysfs_data_replicas_have,
 529
 530         &sysfs_journal_write_delay_ms,
 531         &sysfs_journal_reclaim_delay_ms,
 532
 533         &sysfs_promote_whole_extents,
 534
 535         &sysfs_compression_stats,
 536
 537 #ifdef CONFIG_BCACHEFS_TESTS
 538         &sysfs_perf_test,
 539 #endif
 540         NULL
 541 };
 542
 543 /* internal dir - just a wrapper */
 544
 545 SHOW(bch2_fs_internal)
 546 {
 547         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 548         return bch2_fs_show(&c->kobj, attr, buf);
 549 }
 550
 551 STORE(bch2_fs_internal)
 552 {
 553         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 554         return bch2_fs_store(&c->kobj, attr, buf, size);
 555 }
 556 SYSFS_OPS(bch2_fs_internal);
 557
 558 struct attribute *bch2_fs_internal_files[] = {
 559         &sysfs_alloc_debug,
 560         &sysfs_journal_debug,
 561         &sysfs_journal_pins,
 562         &sysfs_btree_updates,
 563         &sysfs_dirty_btree_nodes,
 564
 565         &sysfs_read_realloc_races,
 566         &sysfs_extent_migrate_done,
 567         &sysfs_extent_migrate_raced,
 568
 569         &sysfs_trigger_journal_flush,
 570         &sysfs_trigger_btree_coalesce,
 571         &sysfs_trigger_gc,
 572         &sysfs_trigger_alloc_write,
 573         &sysfs_prune_cache,
 574
 575         &sysfs_copy_gc_enabled,
 576
 577         &sysfs_rebalance_enabled,
 578         &sysfs_rebalance_work,
 579         sysfs_pd_controller_files(rebalance),
 580
 581         &sysfs_new_stripes,
 582
 583         &sysfs_internal_uuid,
 584
 585 #define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
 586         BCH_DEBUG_PARAMS()
 587 #undef BCH_DEBUG_PARAM
 588
 589         NULL
 590 };
 591
 592 /* options */
 593
 594 SHOW(bch2_fs_opts_dir)
 595 {
 596         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 597         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 598         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 599         int id = opt - bch2_opt_table;
 600         u64 v = bch2_opt_get_by_id(&c->opts, id);
 601
 602         bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
 603         pr_buf(&out, "\n");
 604
 605         return out.pos - buf;
 606 }
 607
 608 STORE(bch2_fs_opts_dir)
 609 {
 610         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 611         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 612         int ret, id = opt - bch2_opt_table;
 613         char *tmp;
 614         u64 v;
 615
 616         tmp = kstrdup(buf, GFP_KERNEL);
 617         if (!tmp)
 618                 return -ENOMEM;
 619
 620         ret = bch2_opt_parse(c, opt, strim(tmp), &v);
 621         kfree(tmp);
 622
 623         if (ret < 0)
 624                 return ret;
 625
 626         ret = bch2_opt_check_may_set(c, id, v);
 627         if (ret < 0)
 628                 return ret;
 629
 630         if (opt->set_sb != SET_NO_SB_OPT) {
 631                 mutex_lock(&c->sb_lock);
 632                 opt->set_sb(c->disk_sb.sb, v);
 633                 bch2_write_super(c);
 634                 mutex_unlock(&c->sb_lock);
 635         }
 636
 637         bch2_opt_set_by_id(&c->opts, id, v);
 638
 639         if ((id == Opt_background_target ||
 640              id == Opt_background_compression) && v) {
 641                 bch2_rebalance_add_work(c, S64_MAX);
 642                 rebalance_wakeup(c);
 643         }
 644
 645         return size;
 646 }
 647 SYSFS_OPS(bch2_fs_opts_dir);
 648
 649 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 650
 651 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 652 {
 653         const struct bch_option *i;
 654         int ret;
 655
 656         for (i = bch2_opt_table;
 657              i < bch2_opt_table + bch2_opts_nr;
 658              i++) {
 659                 if (!(i->mode & (OPT_FORMAT|OPT_MOUNT|OPT_RUNTIME)))
 660                         continue;
 661
 662                 ret = sysfs_create_file(kobj, &i->attr);
 663                 if (ret)
 664                         return ret;
 665         }
 666
 667         return 0;
 668 }
 669
 670 /* time stats */
 671
 672 SHOW(bch2_fs_time_stats)
 673 {
 674         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 675
 676 #define x(name)                                         \
 677         if (attr == &sysfs_time_stat_##name)                            \
 678                 return bch2_time_stats_print(&c->times[BCH_TIME_##name],\
 679                                              buf, PAGE_SIZE);
 680         BCH_TIME_STATS()
 681 #undef x
 682
 683         return 0;
 684 }
 685
 686 STORE(bch2_fs_time_stats)
 687 {
 688         return size;
 689 }
 690 SYSFS_OPS(bch2_fs_time_stats);
 691
 692 struct attribute *bch2_fs_time_stats_files[] = {
 693 #define x(name)                                         \
 694         &sysfs_time_stat_##name,
 695         BCH_TIME_STATS()
 696 #undef x
 697         NULL
 698 };
 699
 700 typedef unsigned (bucket_map_fn)(struct bch_fs *, struct bch_dev *,
 701                                  size_t, void *);
 702
 703 static unsigned bucket_last_io_fn(struct bch_fs *c, struct bch_dev *ca,
 704                                   size_t b, void *private)
 705 {
 706         int rw = (private ? 1 : 0);
 707
 708         return bucket_last_io(c, bucket(ca, b), rw);
 709 }
 710
 711 static unsigned bucket_sectors_used_fn(struct bch_fs *c, struct bch_dev *ca,
 712                                        size_t b, void *private)
 713 {
 714         struct bucket *g = bucket(ca, b);
 715         return bucket_sectors_used(g->mark);
 716 }
 717
 718 static unsigned bucket_oldest_gen_fn(struct bch_fs *c, struct bch_dev *ca,
 719                                      size_t b, void *private)
 720 {
 721         return bucket_gc_gen(ca, b);
 722 }
 723
 724 static int unsigned_cmp(const void *_l, const void *_r)
 725 {
 726         const unsigned *l = _l;
 727         const unsigned *r = _r;
 728
 729         return cmp_int(*l, *r);
 730 }
 731
 732 static ssize_t show_quantiles(struct bch_fs *c, struct bch_dev *ca,
 733                               char *buf, bucket_map_fn *fn, void *private)
 734 {
 735         size_t i, n;
 736         /* Compute 31 quantiles */
 737         unsigned q[31], *p;
 738         ssize_t ret = 0;
 739
 740         down_read(&ca->bucket_lock);
 741         n = ca->mi.nbuckets;
 742
 743         p = vzalloc(n * sizeof(unsigned));
 744         if (!p) {
 745                 up_read(&ca->bucket_lock);
 746                 return -ENOMEM;
 747         }
 748
 749         for (i = ca->mi.first_bucket; i < n; i++)
 750                 p[i] = fn(c, ca, i, private);
 751
 752         sort(p, n, sizeof(unsigned), unsigned_cmp, NULL);
 753         up_read(&ca->bucket_lock);
 754
 755         while (n &&
 756                !p[n - 1])
 757                 --n;
 758
 759         for (i = 0; i < ARRAY_SIZE(q); i++)
 760                 q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
 761
 762         vfree(p);
 763
 764         for (i = 0; i < ARRAY_SIZE(q); i++)
 765                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 766                                  "%u ", q[i]);
 767         buf[ret - 1] = '\n';
 768
 769         return ret;
 770 }
 771
 772 static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
 773 {
 774         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 775         enum alloc_reserve i;
 776
 777         spin_lock(&ca->freelist_lock);
 778
 779         pr_buf(&out, "free_inc:\t%zu\t%zu\n",
 780                fifo_used(&ca->free_inc),
 781                ca->free_inc.size);
 782
 783         for (i = 0; i < RESERVE_NR; i++)
 784                 pr_buf(&out, "free[%u]:\t%zu\t%zu\n", i,
 785                        fifo_used(&ca->free[i]),
 786                        ca->free[i].size);
 787
 788         spin_unlock(&ca->freelist_lock);
 789
 790         return out.pos - buf;
 791 }
 792
 793 static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
 794 {
 795         struct bch_fs *c = ca->fs;
 796         struct bch_dev_usage stats = bch2_dev_usage_read(c, ca);
 797         unsigned i, nr[BCH_DATA_NR];
 798
 799         memset(nr, 0, sizeof(nr));
 800
 801         for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 802                 nr[c->open_buckets[i].type]++;
 803
 804         return scnprintf(buf, PAGE_SIZE,
 805                 "free_inc:               %zu/%zu\n"
 806                 "free[RESERVE_BTREE]:    %zu/%zu\n"
 807                 "free[RESERVE_MOVINGGC]: %zu/%zu\n"
 808                 "free[RESERVE_NONE]:     %zu/%zu\n"
 809                 "buckets:\n"
 810                 "    capacity:           %llu\n"
 811                 "    alloc:              %llu\n"
 812                 "    sb:                 %llu\n"
 813                 "    journal:            %llu\n"
 814                 "    meta:               %llu\n"
 815                 "    user:               %llu\n"
 816                 "    cached:             %llu\n"
 817                 "    erasure coded:      %llu\n"
 818                 "    available:          %lli\n"
 819                 "sectors:\n"
 820                 "    sb:                 %llu\n"
 821                 "    journal:            %llu\n"
 822                 "    meta:               %llu\n"
 823                 "    user:               %llu\n"
 824                 "    cached:             %llu\n"
 825                 "    fragmented:         %llu\n"
 826                 "    copygc threshold:   %llu\n"
 827                 "freelist_wait:          %s\n"
 828                 "open buckets:           %u/%u (reserved %u)\n"
 829                 "open_buckets_wait:      %s\n"
 830                 "open_buckets_btree:     %u\n"
 831                 "open_buckets_user:      %u\n"
 832                 "btree reserve cache:    %u\n",
 833                 fifo_used(&ca->free_inc),               ca->free_inc.size,
 834                 fifo_used(&ca->free[RESERVE_BTREE]),    ca->free[RESERVE_BTREE].size,
 835                 fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size,
 836                 fifo_used(&ca->free[RESERVE_NONE]),     ca->free[RESERVE_NONE].size,
 837                 ca->mi.nbuckets - ca->mi.first_bucket,
 838                 stats.buckets_alloc,
 839                 stats.buckets[BCH_DATA_SB],
 840                 stats.buckets[BCH_DATA_JOURNAL],
 841                 stats.buckets[BCH_DATA_BTREE],
 842                 stats.buckets[BCH_DATA_USER],
 843                 stats.buckets[BCH_DATA_CACHED],
 844                 stats.buckets_ec,
 845                 ca->mi.nbuckets - ca->mi.first_bucket - stats.buckets_unavailable,
 846                 stats.sectors[BCH_DATA_SB],
 847                 stats.sectors[BCH_DATA_JOURNAL],
 848                 stats.sectors[BCH_DATA_BTREE],
 849                 stats.sectors[BCH_DATA_USER],
 850                 stats.sectors[BCH_DATA_CACHED],
 851                 stats.sectors_fragmented,
 852                 ca->copygc_threshold,
 853                 c->freelist_wait.list.first             ? "waiting" : "empty",
 854                 c->open_buckets_nr_free, OPEN_BUCKETS_COUNT,
 855                 BTREE_NODE_OPEN_BUCKET_RESERVE,
 856                 c->open_buckets_wait.list.first         ? "waiting" : "empty",
 857                 nr[BCH_DATA_BTREE],
 858                 nr[BCH_DATA_USER],
 859                 c->btree_reserve_cache_nr);
 860 }
 861
 862 static const char * const bch2_rw[] = {
 863         "read",
 864         "write",
 865         NULL
 866 };
 867
 868 static ssize_t show_dev_iodone(struct bch_dev *ca, char *buf)
 869 {
 870         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 871         int rw, i;
 872
 873         for (rw = 0; rw < 2; rw++) {
 874                 pr_buf(&out, "%s:\n", bch2_rw[rw]);
 875
 876                 for (i = 1; i < BCH_DATA_NR; i++)
 877                         pr_buf(&out, "%-12s:%12llu\n",
 878                                bch2_data_types[i],
 879                                percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
 880         }
 881
 882         return out.pos - buf;
 883 }
 884
 885 SHOW(bch2_dev)
 886 {
 887         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 888         struct bch_fs *c = ca->fs;
 889         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 890
 891         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 892
 893         sysfs_print(bucket_size,        bucket_bytes(ca));
 894         sysfs_print(block_size,         block_bytes(c));
 895         sysfs_print(first_bucket,       ca->mi.first_bucket);
 896         sysfs_print(nbuckets,           ca->mi.nbuckets);
 897         sysfs_print(durability,         ca->mi.durability);
 898         sysfs_print(discard,            ca->mi.discard);
 899
 900         if (attr == &sysfs_label) {
 901                 if (ca->mi.group) {
 902                         mutex_lock(&c->sb_lock);
 903                         bch2_disk_path_to_text(&out, &c->disk_sb,
 904                                                ca->mi.group - 1);
 905                         mutex_unlock(&c->sb_lock);
 906                 } else {
 907                         pr_buf(&out, "none");
 908                 }
 909
 910                 pr_buf(&out, "\n");
 911                 return out.pos - buf;
 912         }
 913
 914         if (attr == &sysfs_has_data) {
 915                 bch2_flags_to_text(&out, bch2_data_types,
 916                                    bch2_dev_has_data(c, ca));
 917                 pr_buf(&out, "\n");
 918                 return out.pos - buf;
 919         }
 920
 921         sysfs_pd_controller_show(copy_gc, &ca->copygc_pd);
 922
 923         if (attr == &sysfs_cache_replacement_policy) {
 924                 bch2_string_opt_to_text(&out,
 925                                         bch2_cache_replacement_policies,
 926                                         ca->mi.replacement);
 927                 pr_buf(&out, "\n");
 928                 return out.pos - buf;
 929         }
 930
 931         if (attr == &sysfs_state_rw) {
 932                 bch2_string_opt_to_text(&out, bch2_dev_state,
 933                                         ca->mi.state);
 934                 pr_buf(&out, "\n");
 935                 return out.pos - buf;
 936         }
 937
 938         if (attr == &sysfs_iodone)
 939                 return show_dev_iodone(ca, buf);
 940
 941         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 942         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 943
 944         if (attr == &sysfs_io_latency_stats_read)
 945                 return bch2_time_stats_print(&ca->io_latency[READ], buf, PAGE_SIZE);
 946         if (attr == &sysfs_io_latency_stats_write)
 947                 return bch2_time_stats_print(&ca->io_latency[WRITE], buf, PAGE_SIZE);
 948
 949         sysfs_printf(congested,                 "%u%%",
 950                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 951                      * 100 / CONGESTED_MAX);
 952
 953         if (attr == &sysfs_bucket_quantiles_last_read)
 954                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 0);
 955         if (attr == &sysfs_bucket_quantiles_last_write)
 956                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 1);
 957         if (attr == &sysfs_bucket_quantiles_fragmentation)
 958                 return show_quantiles(c, ca, buf, bucket_sectors_used_fn, NULL);
 959         if (attr == &sysfs_bucket_quantiles_oldest_gen)
 960                 return show_quantiles(c, ca, buf, bucket_oldest_gen_fn, NULL);
 961
 962         if (attr == &sysfs_reserve_stats)
 963                 return show_reserve_stats(ca, buf);
 964         if (attr == &sysfs_alloc_debug)
 965                 return show_dev_alloc_debug(ca, buf);
 966
 967         return 0;
 968 }
 969
 970 STORE(bch2_dev)
 971 {
 972         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 973         struct bch_fs *c = ca->fs;
 974         struct bch_member *mi;
 975
 976         sysfs_pd_controller_store(copy_gc, &ca->copygc_pd);
 977
 978         if (attr == &sysfs_discard) {
 979                 bool v = strtoul_or_return(buf);
 980
 981                 mutex_lock(&c->sb_lock);
 982                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 983
 984                 if (v != BCH_MEMBER_DISCARD(mi)) {
 985                         SET_BCH_MEMBER_DISCARD(mi, v);
 986                         bch2_write_super(c);
 987                 }
 988                 mutex_unlock(&c->sb_lock);
 989         }
 990
 991         if (attr == &sysfs_cache_replacement_policy) {
 992                 ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
 993
 994                 if (v < 0)
 995                         return v;
 996
 997                 mutex_lock(&c->sb_lock);
 998                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 999
1000                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
1001                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
1002                         bch2_write_super(c);
1003                 }
1004                 mutex_unlock(&c->sb_lock);
1005         }
1006
1007         if (attr == &sysfs_label) {
1008                 char *tmp;
1009                 int ret;
1010
1011                 tmp = kstrdup(buf, GFP_KERNEL);
1012                 if (!tmp)
1013                         return -ENOMEM;
1014
1015                 ret = bch2_dev_group_set(c, ca, strim(tmp));
1016                 kfree(tmp);
1017                 if (ret)
1018                         return ret;
1019         }
1020
1021         if (attr == &sysfs_wake_allocator)
1022                 bch2_wake_allocator(ca);
1023
1024         return size;
1025 }
1026 SYSFS_OPS(bch2_dev);
1027
1028 struct attribute *bch2_dev_files[] = {
1029         &sysfs_uuid,
1030         &sysfs_bucket_size,
1031         &sysfs_block_size,
1032         &sysfs_first_bucket,
1033         &sysfs_nbuckets,
1034         &sysfs_durability,
1035
1036         /* settings: */
1037         &sysfs_discard,
1038         &sysfs_cache_replacement_policy,
1039         &sysfs_state_rw,
1040         &sysfs_label,
1041
1042         &sysfs_has_data,
1043         &sysfs_iodone,
1044
1045         &sysfs_io_latency_read,
1046         &sysfs_io_latency_write,
1047         &sysfs_io_latency_stats_read,
1048         &sysfs_io_latency_stats_write,
1049         &sysfs_congested,
1050
1051         /* alloc info - other stats: */
1052         &sysfs_bucket_quantiles_last_read,
1053         &sysfs_bucket_quantiles_last_write,
1054         &sysfs_bucket_quantiles_fragmentation,
1055         &sysfs_bucket_quantiles_oldest_gen,
1056
1057         &sysfs_reserve_stats,
1058
1059         /* debug: */
1060         &sysfs_alloc_debug,
1061         &sysfs_wake_allocator,
1062
1063         sysfs_pd_controller_files(copy_gc),
1064         NULL
1065 };
1066
1067 #endif  /* _BCACHEFS_SYSFS_H_ */