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         unsigned i;
 239
 240         if (!fs_usage)
 241                 return -ENOMEM;
 242
 243         pr_buf(&out, "capacity:\t\t\t%llu\n", c->capacity);
 244
 245         pr_buf(&out, "hidden:\t\t\t\t%llu\n",
 246                fs_usage->hidden);
 247         pr_buf(&out, "data:\t\t\t\t%llu\n",
 248                fs_usage->data);
 249         pr_buf(&out, "cached:\t\t\t\t%llu\n",
 250                fs_usage->cached);
 251         pr_buf(&out, "reserved:\t\t\t%llu\n",
 252                fs_usage->reserved);
 253         pr_buf(&out, "nr_inodes:\t\t\t%llu\n",
 254                fs_usage->nr_inodes);
 255         pr_buf(&out, "online reserved:\t\t%llu\n",
 256                fs_usage->online_reserved);
 257
 258         for (i = 0;
 259              i < ARRAY_SIZE(fs_usage->persistent_reserved);
 260              i++) {
 261                 pr_buf(&out, "%u replicas:\n", i + 1);
 262                 pr_buf(&out, "\treserved:\t\t%llu\n",
 263                        fs_usage->persistent_reserved[i]);
 264         }
 265
 266         for (i = 0; i < c->replicas.nr; i++) {
 267                 struct bch_replicas_entry *e =
 268                         cpu_replicas_entry(&c->replicas, i);
 269
 270                 pr_buf(&out, "\t");
 271                 bch2_replicas_entry_to_text(&out, e);
 272                 pr_buf(&out, ":\t%llu\n", fs_usage->replicas[i]);
 273         }
 274
 275         percpu_up_read_preempt_enable(&c->mark_lock);
 276
 277         kfree(fs_usage);
 278
 279         return out.pos - buf;
 280 }
 281
 282 static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 283 {
 284         struct btree_trans trans;
 285         struct btree_iter *iter;
 286         struct bkey_s_c k;
 287         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 288             nr_compressed_extents = 0,
 289             compressed_sectors_compressed = 0,
 290             compressed_sectors_uncompressed = 0;
 291
 292         if (!test_bit(BCH_FS_STARTED, &c->flags))
 293                 return -EPERM;
 294
 295         bch2_trans_init(&trans, c);
 296
 297         for_each_btree_key(&trans, iter, BTREE_ID_EXTENTS, POS_MIN, 0, k)
 298                 if (k.k->type == KEY_TYPE_extent) {
 299                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 300                         const union bch_extent_entry *entry;
 301                         struct extent_ptr_decoded p;
 302
 303                         extent_for_each_ptr_decode(e, p, entry) {
 304                                 if (p.crc.compression_type == BCH_COMPRESSION_NONE) {
 305                                         nr_uncompressed_extents++;
 306                                         uncompressed_sectors += e.k->size;
 307                                 } else {
 308                                         nr_compressed_extents++;
 309                                         compressed_sectors_compressed +=
 310                                                 p.crc.compressed_size;
 311                                         compressed_sectors_uncompressed +=
 312                                                 p.crc.uncompressed_size;
 313                                 }
 314
 315                                 /* only looking at the first ptr */
 316                                 break;
 317                         }
 318                 }
 319         bch2_trans_exit(&trans);
 320
 321         return scnprintf(buf, PAGE_SIZE,
 322                         "uncompressed data:\n"
 323                         "       nr extents:                     %llu\n"
 324                         "       size (bytes):                   %llu\n"
 325                         "compressed data:\n"
 326                         "       nr extents:                     %llu\n"
 327                         "       compressed size (bytes):        %llu\n"
 328                         "       uncompressed size (bytes):      %llu\n",
 329                         nr_uncompressed_extents,
 330                         uncompressed_sectors << 9,
 331                         nr_compressed_extents,
 332                         compressed_sectors_compressed << 9,
 333                         compressed_sectors_uncompressed << 9);
 334 }
 335
 336 static ssize_t bch2_new_stripes(struct bch_fs *c, char *buf)
 337 {
 338         char *out = buf, *end = buf + PAGE_SIZE;
 339         struct ec_stripe_head *h;
 340         struct ec_stripe_new *s;
 341
 342         mutex_lock(&c->ec_new_stripe_lock);
 343         list_for_each_entry(h, &c->ec_new_stripe_list, list) {
 344                 out += scnprintf(out, end - out,
 345                                  "target %u algo %u redundancy %u:\n",
 346                                  h->target, h->algo, h->redundancy);
 347
 348                 if (h->s)
 349                         out += scnprintf(out, end - out,
 350                                          "\tpending: blocks %u allocated %u\n",
 351                                          h->s->blocks.nr,
 352                                          bitmap_weight(h->s->blocks_allocated,
 353                                                        h->s->blocks.nr));
 354
 355                 mutex_lock(&h->lock);
 356                 list_for_each_entry(s, &h->stripes, list)
 357                         out += scnprintf(out, end - out,
 358                                          "\tin flight: blocks %u allocated %u pin %u\n",
 359                                          s->blocks.nr,
 360                                          bitmap_weight(s->blocks_allocated,
 361                                                        s->blocks.nr),
 362                                          atomic_read(&s->pin));
 363                 mutex_unlock(&h->lock);
 364
 365         }
 366         mutex_unlock(&c->ec_new_stripe_lock);
 367
 368         return out - buf;
 369 }
 370
 371 SHOW(bch2_fs)
 372 {
 373         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 374
 375         sysfs_print(minor,                      c->minor);
 376         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 377
 378         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 379         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 380
 381         sysfs_print(block_size,                 block_bytes(c));
 382         sysfs_print(btree_node_size,            btree_bytes(c));
 383         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 384
 385         sysfs_print(read_realloc_races,
 386                     atomic_long_read(&c->read_realloc_races));
 387         sysfs_print(extent_migrate_done,
 388                     atomic_long_read(&c->extent_migrate_done));
 389         sysfs_print(extent_migrate_raced,
 390                     atomic_long_read(&c->extent_migrate_raced));
 391
 392         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 393
 394         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 395
 396         sysfs_print(pd_controllers_update_seconds,
 397                     c->pd_controllers_update_seconds);
 398
 399         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 400         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 401
 402         if (attr == &sysfs_rebalance_work)
 403                 return bch2_rebalance_work_show(c, buf);
 404
 405         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 406
 407         sysfs_printf(meta_replicas_have, "%i",  bch2_replicas_online(c, true));
 408         sysfs_printf(data_replicas_have, "%i",  bch2_replicas_online(c, false));
 409
 410         /* Debugging: */
 411
 412         if (attr == &sysfs_alloc_debug)
 413                 return show_fs_alloc_debug(c, buf);
 414
 415         if (attr == &sysfs_journal_debug)
 416                 return bch2_journal_print_debug(&c->journal, buf);
 417
 418         if (attr == &sysfs_journal_pins)
 419                 return bch2_journal_print_pins(&c->journal, buf);
 420
 421         if (attr == &sysfs_btree_updates)
 422                 return bch2_btree_updates_print(c, buf);
 423
 424         if (attr == &sysfs_dirty_btree_nodes)
 425                 return bch2_dirty_btree_nodes_print(c, buf);
 426
 427         if (attr == &sysfs_compression_stats)
 428                 return bch2_compression_stats(c, buf);
 429
 430         if (attr == &sysfs_new_stripes)
 431                 return bch2_new_stripes(c, buf);
 432
 433 #define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
 434         BCH_DEBUG_PARAMS()
 435 #undef BCH_DEBUG_PARAM
 436
 437         return 0;
 438 }
 439
 440 STORE(__bch2_fs)
 441 {
 442         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 443
 444         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 445         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 446
 447         if (attr == &sysfs_btree_gc_periodic) {
 448                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 449                         ?: (ssize_t) size;
 450
 451                 wake_up_process(c->gc_thread);
 452                 return ret;
 453         }
 454
 455         if (attr == &sysfs_copy_gc_enabled) {
 456                 struct bch_dev *ca;
 457                 unsigned i;
 458                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 459                         ?: (ssize_t) size;
 460
 461                 for_each_member_device(ca, c, i)
 462                         if (ca->copygc_thread)
 463                                 wake_up_process(ca->copygc_thread);
 464                 return ret;
 465         }
 466
 467         if (attr == &sysfs_rebalance_enabled) {
 468                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 469                         ?: (ssize_t) size;
 470
 471                 rebalance_wakeup(c);
 472                 return ret;
 473         }
 474
 475         sysfs_strtoul(pd_controllers_update_seconds,
 476                       c->pd_controllers_update_seconds);
 477         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 478
 479         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 480
 481         /* Debugging: */
 482
 483 #define BCH_DEBUG_PARAM(name, description) sysfs_strtoul(name, c->name);
 484         BCH_DEBUG_PARAMS()
 485 #undef BCH_DEBUG_PARAM
 486
 487         if (!test_bit(BCH_FS_STARTED, &c->flags))
 488                 return -EPERM;
 489
 490         /* Debugging: */
 491
 492         if (attr == &sysfs_trigger_journal_flush)
 493                 bch2_journal_meta_async(&c->journal, NULL);
 494
 495         if (attr == &sysfs_trigger_btree_coalesce)
 496                 bch2_coalesce(c);
 497
 498         if (attr == &sysfs_trigger_gc)
 499                 bch2_gc(c, NULL, false, false);
 500
 501         if (attr == &sysfs_trigger_alloc_write) {
 502                 bool wrote;
 503
 504                 bch2_alloc_write(c, false, &wrote);
 505         }
 506
 507         if (attr == &sysfs_prune_cache) {
 508                 struct shrink_control sc;
 509
 510                 sc.gfp_mask = GFP_KERNEL;
 511                 sc.nr_to_scan = strtoul_or_return(buf);
 512                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 513         }
 514 #ifdef CONFIG_BCACHEFS_TESTS
 515         if (attr == &sysfs_perf_test) {
 516                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 517                 char *test              = strsep(&p, " \t\n");
 518                 char *nr_str            = strsep(&p, " \t\n");
 519                 char *threads_str       = strsep(&p, " \t\n");
 520                 unsigned threads;
 521                 u64 nr;
 522                 int ret = -EINVAL;
 523
 524                 if (threads_str &&
 525                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 526                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 527                         bch2_btree_perf_test(c, test, nr, threads);
 528                 else
 529                         size = ret;
 530                 kfree(tmp);
 531         }
 532 #endif
 533         return size;
 534 }
 535
 536 STORE(bch2_fs)
 537 {
 538         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 539
 540         mutex_lock(&c->state_lock);
 541         size = __bch2_fs_store(kobj, attr, buf, size);
 542         mutex_unlock(&c->state_lock);
 543
 544         return size;
 545 }
 546 SYSFS_OPS(bch2_fs);
 547
 548 struct attribute *bch2_fs_files[] = {
 549         &sysfs_minor,
 550         &sysfs_block_size,
 551         &sysfs_btree_node_size,
 552         &sysfs_btree_cache_size,
 553
 554         &sysfs_meta_replicas_have,
 555         &sysfs_data_replicas_have,
 556
 557         &sysfs_journal_write_delay_ms,
 558         &sysfs_journal_reclaim_delay_ms,
 559
 560         &sysfs_promote_whole_extents,
 561
 562         &sysfs_compression_stats,
 563
 564 #ifdef CONFIG_BCACHEFS_TESTS
 565         &sysfs_perf_test,
 566 #endif
 567         NULL
 568 };
 569
 570 /* internal dir - just a wrapper */
 571
 572 SHOW(bch2_fs_internal)
 573 {
 574         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 575         return bch2_fs_show(&c->kobj, attr, buf);
 576 }
 577
 578 STORE(bch2_fs_internal)
 579 {
 580         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 581         return bch2_fs_store(&c->kobj, attr, buf, size);
 582 }
 583 SYSFS_OPS(bch2_fs_internal);
 584
 585 struct attribute *bch2_fs_internal_files[] = {
 586         &sysfs_alloc_debug,
 587         &sysfs_journal_debug,
 588         &sysfs_journal_pins,
 589         &sysfs_btree_updates,
 590         &sysfs_dirty_btree_nodes,
 591
 592         &sysfs_read_realloc_races,
 593         &sysfs_extent_migrate_done,
 594         &sysfs_extent_migrate_raced,
 595
 596         &sysfs_trigger_journal_flush,
 597         &sysfs_trigger_btree_coalesce,
 598         &sysfs_trigger_gc,
 599         &sysfs_trigger_alloc_write,
 600         &sysfs_prune_cache,
 601
 602         &sysfs_copy_gc_enabled,
 603
 604         &sysfs_rebalance_enabled,
 605         &sysfs_rebalance_work,
 606         sysfs_pd_controller_files(rebalance),
 607
 608         &sysfs_new_stripes,
 609
 610         &sysfs_internal_uuid,
 611
 612 #define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
 613         BCH_DEBUG_PARAMS()
 614 #undef BCH_DEBUG_PARAM
 615
 616         NULL
 617 };
 618
 619 /* options */
 620
 621 SHOW(bch2_fs_opts_dir)
 622 {
 623         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 624         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 625         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 626         int id = opt - bch2_opt_table;
 627         u64 v = bch2_opt_get_by_id(&c->opts, id);
 628
 629         bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
 630         pr_buf(&out, "\n");
 631
 632         return out.pos - buf;
 633 }
 634
 635 STORE(bch2_fs_opts_dir)
 636 {
 637         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 638         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 639         int ret, id = opt - bch2_opt_table;
 640         char *tmp;
 641         u64 v;
 642
 643         tmp = kstrdup(buf, GFP_KERNEL);
 644         if (!tmp)
 645                 return -ENOMEM;
 646
 647         ret = bch2_opt_parse(c, opt, strim(tmp), &v);
 648         kfree(tmp);
 649
 650         if (ret < 0)
 651                 return ret;
 652
 653         ret = bch2_opt_check_may_set(c, id, v);
 654         if (ret < 0)
 655                 return ret;
 656
 657         if (opt->set_sb != SET_NO_SB_OPT) {
 658                 mutex_lock(&c->sb_lock);
 659                 opt->set_sb(c->disk_sb.sb, v);
 660                 bch2_write_super(c);
 661                 mutex_unlock(&c->sb_lock);
 662         }
 663
 664         bch2_opt_set_by_id(&c->opts, id, v);
 665
 666         if ((id == Opt_background_target ||
 667              id == Opt_background_compression) && v) {
 668                 bch2_rebalance_add_work(c, S64_MAX);
 669                 rebalance_wakeup(c);
 670         }
 671
 672         return size;
 673 }
 674 SYSFS_OPS(bch2_fs_opts_dir);
 675
 676 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 677
 678 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 679 {
 680         const struct bch_option *i;
 681         int ret;
 682
 683         for (i = bch2_opt_table;
 684              i < bch2_opt_table + bch2_opts_nr;
 685              i++) {
 686                 if (!(i->mode & (OPT_FORMAT|OPT_MOUNT|OPT_RUNTIME)))
 687                         continue;
 688
 689                 ret = sysfs_create_file(kobj, &i->attr);
 690                 if (ret)
 691                         return ret;
 692         }
 693
 694         return 0;
 695 }
 696
 697 /* time stats */
 698
 699 SHOW(bch2_fs_time_stats)
 700 {
 701         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 702
 703 #define x(name)                                         \
 704         if (attr == &sysfs_time_stat_##name)                            \
 705                 return bch2_time_stats_print(&c->times[BCH_TIME_##name],\
 706                                              buf, PAGE_SIZE);
 707         BCH_TIME_STATS()
 708 #undef x
 709
 710         return 0;
 711 }
 712
 713 STORE(bch2_fs_time_stats)
 714 {
 715         return size;
 716 }
 717 SYSFS_OPS(bch2_fs_time_stats);
 718
 719 struct attribute *bch2_fs_time_stats_files[] = {
 720 #define x(name)                                         \
 721         &sysfs_time_stat_##name,
 722         BCH_TIME_STATS()
 723 #undef x
 724         NULL
 725 };
 726
 727 typedef unsigned (bucket_map_fn)(struct bch_fs *, struct bch_dev *,
 728                                  size_t, void *);
 729
 730 static unsigned bucket_last_io_fn(struct bch_fs *c, struct bch_dev *ca,
 731                                   size_t b, void *private)
 732 {
 733         int rw = (private ? 1 : 0);
 734
 735         return bucket_last_io(c, bucket(ca, b), rw);
 736 }
 737
 738 static unsigned bucket_sectors_used_fn(struct bch_fs *c, struct bch_dev *ca,
 739                                        size_t b, void *private)
 740 {
 741         struct bucket *g = bucket(ca, b);
 742         return bucket_sectors_used(g->mark);
 743 }
 744
 745 static unsigned bucket_oldest_gen_fn(struct bch_fs *c, struct bch_dev *ca,
 746                                      size_t b, void *private)
 747 {
 748         return bucket_gc_gen(ca, b);
 749 }
 750
 751 static int unsigned_cmp(const void *_l, const void *_r)
 752 {
 753         unsigned l = *((unsigned *) _l);
 754         unsigned r = *((unsigned *) _r);
 755
 756         return (l > r) - (l < r);
 757 }
 758
 759 static ssize_t show_quantiles(struct bch_fs *c, struct bch_dev *ca,
 760                               char *buf, bucket_map_fn *fn, void *private)
 761 {
 762         size_t i, n;
 763         /* Compute 31 quantiles */
 764         unsigned q[31], *p;
 765         ssize_t ret = 0;
 766
 767         down_read(&ca->bucket_lock);
 768         n = ca->mi.nbuckets;
 769
 770         p = vzalloc(n * sizeof(unsigned));
 771         if (!p) {
 772                 up_read(&ca->bucket_lock);
 773                 return -ENOMEM;
 774         }
 775
 776         for (i = ca->mi.first_bucket; i < n; i++)
 777                 p[i] = fn(c, ca, i, private);
 778
 779         sort(p, n, sizeof(unsigned), unsigned_cmp, NULL);
 780         up_read(&ca->bucket_lock);
 781
 782         while (n &&
 783                !p[n - 1])
 784                 --n;
 785
 786         for (i = 0; i < ARRAY_SIZE(q); i++)
 787                 q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
 788
 789         vfree(p);
 790
 791         for (i = 0; i < ARRAY_SIZE(q); i++)
 792                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 793                                  "%u ", q[i]);
 794         buf[ret - 1] = '\n';
 795
 796         return ret;
 797 }
 798
 799 static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
 800 {
 801         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 802         enum alloc_reserve i;
 803
 804         spin_lock(&ca->freelist_lock);
 805
 806         pr_buf(&out, "free_inc:\t%zu\t%zu\n",
 807                fifo_used(&ca->free_inc),
 808                ca->free_inc.size);
 809
 810         for (i = 0; i < RESERVE_NR; i++)
 811                 pr_buf(&out, "free[%u]:\t%zu\t%zu\n", i,
 812                        fifo_used(&ca->free[i]),
 813                        ca->free[i].size);
 814
 815         spin_unlock(&ca->freelist_lock);
 816
 817         return out.pos - buf;
 818 }
 819
 820 static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
 821 {
 822         struct bch_fs *c = ca->fs;
 823         struct bch_dev_usage stats = bch2_dev_usage_read(c, ca);
 824         unsigned i, nr[BCH_DATA_NR];
 825
 826         memset(nr, 0, sizeof(nr));
 827
 828         for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 829                 nr[c->open_buckets[i].type]++;
 830
 831         return scnprintf(buf, PAGE_SIZE,
 832                 "free_inc:               %zu/%zu\n"
 833                 "free[RESERVE_BTREE]:    %zu/%zu\n"
 834                 "free[RESERVE_MOVINGGC]: %zu/%zu\n"
 835                 "free[RESERVE_NONE]:     %zu/%zu\n"
 836                 "buckets:\n"
 837                 "    capacity:           %llu\n"
 838                 "    alloc:              %llu\n"
 839                 "    sb:                 %llu\n"
 840                 "    journal:            %llu\n"
 841                 "    meta:               %llu\n"
 842                 "    user:               %llu\n"
 843                 "    cached:             %llu\n"
 844                 "    erasure coded:      %llu\n"
 845                 "    available:          %lli\n"
 846                 "sectors:\n"
 847                 "    sb:                 %llu\n"
 848                 "    journal:            %llu\n"
 849                 "    meta:               %llu\n"
 850                 "    user:               %llu\n"
 851                 "    cached:             %llu\n"
 852                 "    fragmented:         %llu\n"
 853                 "    copygc threshold:   %llu\n"
 854                 "freelist_wait:          %s\n"
 855                 "open buckets:           %u/%u (reserved %u)\n"
 856                 "open_buckets_wait:      %s\n"
 857                 "open_buckets_btree:     %u\n"
 858                 "open_buckets_user:      %u\n"
 859                 "btree reserve cache:    %u\n",
 860                 fifo_used(&ca->free_inc),               ca->free_inc.size,
 861                 fifo_used(&ca->free[RESERVE_BTREE]),    ca->free[RESERVE_BTREE].size,
 862                 fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size,
 863                 fifo_used(&ca->free[RESERVE_NONE]),     ca->free[RESERVE_NONE].size,
 864                 ca->mi.nbuckets - ca->mi.first_bucket,
 865                 stats.buckets_alloc,
 866                 stats.buckets[BCH_DATA_SB],
 867                 stats.buckets[BCH_DATA_JOURNAL],
 868                 stats.buckets[BCH_DATA_BTREE],
 869                 stats.buckets[BCH_DATA_USER],
 870                 stats.buckets[BCH_DATA_CACHED],
 871                 stats.buckets_ec,
 872                 ca->mi.nbuckets - ca->mi.first_bucket - stats.buckets_unavailable,
 873                 stats.sectors[BCH_DATA_SB],
 874                 stats.sectors[BCH_DATA_JOURNAL],
 875                 stats.sectors[BCH_DATA_BTREE],
 876                 stats.sectors[BCH_DATA_USER],
 877                 stats.sectors[BCH_DATA_CACHED],
 878                 stats.sectors_fragmented,
 879                 ca->copygc_threshold,
 880                 c->freelist_wait.list.first             ? "waiting" : "empty",
 881                 c->open_buckets_nr_free, OPEN_BUCKETS_COUNT,
 882                 BTREE_NODE_OPEN_BUCKET_RESERVE,
 883                 c->open_buckets_wait.list.first         ? "waiting" : "empty",
 884                 nr[BCH_DATA_BTREE],
 885                 nr[BCH_DATA_USER],
 886                 c->btree_reserve_cache_nr);
 887 }
 888
 889 static const char * const bch2_rw[] = {
 890         "read",
 891         "write",
 892         NULL
 893 };
 894
 895 static ssize_t show_dev_iodone(struct bch_dev *ca, char *buf)
 896 {
 897         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 898         int rw, i;
 899
 900         for (rw = 0; rw < 2; rw++) {
 901                 pr_buf(&out, "%s:\n", bch2_rw[rw]);
 902
 903                 for (i = 1; i < BCH_DATA_NR; i++)
 904                         pr_buf(&out, "%-12s:%12llu\n",
 905                                bch2_data_types[i],
 906                                percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
 907         }
 908
 909         return out.pos - buf;
 910 }
 911
 912 SHOW(bch2_dev)
 913 {
 914         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 915         struct bch_fs *c = ca->fs;
 916         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 917
 918         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 919
 920         sysfs_print(bucket_size,        bucket_bytes(ca));
 921         sysfs_print(block_size,         block_bytes(c));
 922         sysfs_print(first_bucket,       ca->mi.first_bucket);
 923         sysfs_print(nbuckets,           ca->mi.nbuckets);
 924         sysfs_print(durability,         ca->mi.durability);
 925         sysfs_print(discard,            ca->mi.discard);
 926
 927         if (attr == &sysfs_label) {
 928                 if (ca->mi.group) {
 929                         mutex_lock(&c->sb_lock);
 930                         bch2_disk_path_to_text(&out, &c->disk_sb,
 931                                                ca->mi.group - 1);
 932                         mutex_unlock(&c->sb_lock);
 933                 } else {
 934                         pr_buf(&out, "none");
 935                 }
 936
 937                 pr_buf(&out, "\n");
 938                 return out.pos - buf;
 939         }
 940
 941         if (attr == &sysfs_has_data) {
 942                 bch2_flags_to_text(&out, bch2_data_types,
 943                                    bch2_dev_has_data(c, ca));
 944                 pr_buf(&out, "\n");
 945                 return out.pos - buf;
 946         }
 947
 948         sysfs_pd_controller_show(copy_gc, &ca->copygc_pd);
 949
 950         if (attr == &sysfs_cache_replacement_policy) {
 951                 bch2_string_opt_to_text(&out,
 952                                         bch2_cache_replacement_policies,
 953                                         ca->mi.replacement);
 954                 pr_buf(&out, "\n");
 955                 return out.pos - buf;
 956         }
 957
 958         if (attr == &sysfs_state_rw) {
 959                 bch2_string_opt_to_text(&out, bch2_dev_state,
 960                                         ca->mi.state);
 961                 pr_buf(&out, "\n");
 962                 return out.pos - buf;
 963         }
 964
 965         if (attr == &sysfs_iodone)
 966                 return show_dev_iodone(ca, buf);
 967
 968         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 969         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 970
 971         if (attr == &sysfs_io_latency_stats_read)
 972                 return bch2_time_stats_print(&ca->io_latency[READ], buf, PAGE_SIZE);
 973         if (attr == &sysfs_io_latency_stats_write)
 974                 return bch2_time_stats_print(&ca->io_latency[WRITE], buf, PAGE_SIZE);
 975
 976         sysfs_printf(congested,                 "%u%%",
 977                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 978                      * 100 / CONGESTED_MAX);
 979
 980         if (attr == &sysfs_bucket_quantiles_last_read)
 981                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 0);
 982         if (attr == &sysfs_bucket_quantiles_last_write)
 983                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 1);
 984         if (attr == &sysfs_bucket_quantiles_fragmentation)
 985                 return show_quantiles(c, ca, buf, bucket_sectors_used_fn, NULL);
 986         if (attr == &sysfs_bucket_quantiles_oldest_gen)
 987                 return show_quantiles(c, ca, buf, bucket_oldest_gen_fn, NULL);
 988
 989         if (attr == &sysfs_reserve_stats)
 990                 return show_reserve_stats(ca, buf);
 991         if (attr == &sysfs_alloc_debug)
 992                 return show_dev_alloc_debug(ca, buf);
 993
 994         return 0;
 995 }
 996
 997 STORE(bch2_dev)
 998 {
 999         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
1000         struct bch_fs *c = ca->fs;
1001         struct bch_member *mi;
1002
1003         sysfs_pd_controller_store(copy_gc, &ca->copygc_pd);
1004
1005         if (attr == &sysfs_discard) {
1006                 bool v = strtoul_or_return(buf);
1007
1008                 mutex_lock(&c->sb_lock);
1009                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
1010
1011                 if (v != BCH_MEMBER_DISCARD(mi)) {
1012                         SET_BCH_MEMBER_DISCARD(mi, v);
1013                         bch2_write_super(c);
1014                 }
1015                 mutex_unlock(&c->sb_lock);
1016         }
1017
1018         if (attr == &sysfs_cache_replacement_policy) {
1019                 ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
1020
1021                 if (v < 0)
1022                         return v;
1023
1024                 mutex_lock(&c->sb_lock);
1025                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
1026
1027                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
1028                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
1029                         bch2_write_super(c);
1030                 }
1031                 mutex_unlock(&c->sb_lock);
1032         }
1033
1034         if (attr == &sysfs_label) {
1035                 char *tmp;
1036                 int ret;
1037
1038                 tmp = kstrdup(buf, GFP_KERNEL);
1039                 if (!tmp)
1040                         return -ENOMEM;
1041
1042                 ret = bch2_dev_group_set(c, ca, strim(tmp));
1043                 kfree(tmp);
1044                 if (ret)
1045                         return ret;
1046         }
1047
1048         if (attr == &sysfs_wake_allocator)
1049                 bch2_wake_allocator(ca);
1050
1051         return size;
1052 }
1053 SYSFS_OPS(bch2_dev);
1054
1055 struct attribute *bch2_dev_files[] = {
1056         &sysfs_uuid,
1057         &sysfs_bucket_size,
1058         &sysfs_block_size,
1059         &sysfs_first_bucket,
1060         &sysfs_nbuckets,
1061         &sysfs_durability,
1062
1063         /* settings: */
1064         &sysfs_discard,
1065         &sysfs_cache_replacement_policy,
1066         &sysfs_state_rw,
1067         &sysfs_label,
1068
1069         &sysfs_has_data,
1070         &sysfs_iodone,
1071
1072         &sysfs_io_latency_read,
1073         &sysfs_io_latency_write,
1074         &sysfs_io_latency_stats_read,
1075         &sysfs_io_latency_stats_write,
1076         &sysfs_congested,
1077
1078         /* alloc info - other stats: */
1079         &sysfs_bucket_quantiles_last_read,
1080         &sysfs_bucket_quantiles_last_write,
1081         &sysfs_bucket_quantiles_fragmentation,
1082         &sysfs_bucket_quantiles_oldest_gen,
1083
1084         &sysfs_reserve_stats,
1085
1086         /* debug: */
1087         &sysfs_alloc_debug,
1088         &sysfs_wake_allocator,
1089
1090         sysfs_pd_controller_files(copy_gc),
1091         NULL
1092 };
1093
1094 #endif  /* _BCACHEFS_SYSFS_H_ */