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.h"
  12 #include "compress.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 "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                 ssize_t ret = bch2_hprint(buf, val);                    \
  76                 strcat(buf, "\n");                                      \
  77                 return ret + 1;                                         \
  78         }                                                               \
  79 } while (0)
  80
  81 #define var_printf(_var, fmt)   sysfs_printf(_var, fmt, var(_var))
  82 #define var_print(_var)         sysfs_print(_var, var(_var))
  83 #define var_hprint(_var)        sysfs_hprint(_var, var(_var))
  84
  85 #define sysfs_strtoul(file, var)                                        \
  86 do {                                                                    \
  87         if (attr == &sysfs_ ## file)                                    \
  88                 return strtoul_safe(buf, var) ?: (ssize_t) size;        \
  89 } while (0)
  90
  91 #define sysfs_strtoul_clamp(file, var, min, max)                        \
  92 do {                                                                    \
  93         if (attr == &sysfs_ ## file)                                    \
  94                 return strtoul_safe_clamp(buf, var, min, max)           \
  95                         ?: (ssize_t) size;                              \
  96 } while (0)
  97
  98 #define strtoul_or_return(cp)                                           \
  99 ({                                                                      \
 100         unsigned long _v;                                               \
 101         int _r = kstrtoul(cp, 10, &_v);                                 \
 102         if (_r)                                                         \
 103                 return _r;                                              \
 104         _v;                                                             \
 105 })
 106
 107 #define strtoul_restrict_or_return(cp, min, max)                        \
 108 ({                                                                      \
 109         unsigned long __v = 0;                                          \
 110         int _r = strtoul_safe_restrict(cp, __v, min, max);              \
 111         if (_r)                                                         \
 112                 return _r;                                              \
 113         __v;                                                            \
 114 })
 115
 116 #define strtoi_h_or_return(cp)                                          \
 117 ({                                                                      \
 118         u64 _v;                                                         \
 119         int _r = strtoi_h(cp, &_v);                                     \
 120         if (_r)                                                         \
 121                 return _r;                                              \
 122         _v;                                                             \
 123 })
 124
 125 #define sysfs_hatoi(file, var)                                          \
 126 do {                                                                    \
 127         if (attr == &sysfs_ ## file)                                    \
 128                 return strtoi_h(buf, &var) ?: (ssize_t) size;           \
 129 } while (0)
 130
 131 write_attribute(trigger_journal_flush);
 132 write_attribute(trigger_btree_coalesce);
 133 write_attribute(trigger_gc);
 134 write_attribute(prune_cache);
 135 rw_attribute(btree_gc_periodic);
 136
 137 read_attribute(uuid);
 138 read_attribute(minor);
 139 read_attribute(bucket_size);
 140 read_attribute(block_size);
 141 read_attribute(btree_node_size);
 142 read_attribute(first_bucket);
 143 read_attribute(nbuckets);
 144 read_attribute(durability);
 145 read_attribute(iodone);
 146
 147 read_attribute(io_latency_read);
 148 read_attribute(io_latency_write);
 149 read_attribute(io_latency_stats_read);
 150 read_attribute(io_latency_stats_write);
 151 read_attribute(congested);
 152
 153 read_attribute(bucket_quantiles_last_read);
 154 read_attribute(bucket_quantiles_last_write);
 155 read_attribute(bucket_quantiles_fragmentation);
 156 read_attribute(bucket_quantiles_oldest_gen);
 157
 158 read_attribute(reserve_stats);
 159 read_attribute(btree_cache_size);
 160 read_attribute(compression_stats);
 161 read_attribute(journal_debug);
 162 read_attribute(journal_pins);
 163 read_attribute(btree_updates);
 164 read_attribute(dirty_btree_nodes);
 165
 166 read_attribute(internal_uuid);
 167
 168 read_attribute(has_data);
 169 read_attribute(alloc_debug);
 170 write_attribute(wake_allocator);
 171
 172 read_attribute(read_realloc_races);
 173 read_attribute(extent_migrate_done);
 174 read_attribute(extent_migrate_raced);
 175
 176 rw_attribute(journal_write_delay_ms);
 177 rw_attribute(journal_reclaim_delay_ms);
 178
 179 rw_attribute(discard);
 180 rw_attribute(cache_replacement_policy);
 181 rw_attribute(label);
 182
 183 rw_attribute(copy_gc_enabled);
 184 sysfs_pd_controller_attribute(copy_gc);
 185
 186 rw_attribute(rebalance_enabled);
 187 sysfs_pd_controller_attribute(rebalance);
 188 read_attribute(rebalance_work);
 189 rw_attribute(promote_whole_extents);
 190
 191 rw_attribute(pd_controllers_update_seconds);
 192
 193 read_attribute(meta_replicas_have);
 194 read_attribute(data_replicas_have);
 195
 196 #ifdef CONFIG_BCACHEFS_TESTS
 197 write_attribute(perf_test);
 198 #endif /* CONFIG_BCACHEFS_TESTS */
 199
 200 #define BCH_DEBUG_PARAM(name, description)                              \
 201         rw_attribute(name);
 202
 203         BCH_DEBUG_PARAMS()
 204 #undef BCH_DEBUG_PARAM
 205
 206 #define x(_name)                                                \
 207         static struct attribute sysfs_time_stat_##_name =               \
 208                 { .name = #_name, .mode = S_IRUGO };
 209         BCH_TIME_STATS()
 210 #undef x
 211
 212 static struct attribute sysfs_state_rw = {
 213         .name = "state",
 214         .mode = S_IRUGO
 215 };
 216
 217 static size_t bch2_btree_cache_size(struct bch_fs *c)
 218 {
 219         size_t ret = 0;
 220         struct btree *b;
 221
 222         mutex_lock(&c->btree_cache.lock);
 223         list_for_each_entry(b, &c->btree_cache.live, list)
 224                 ret += btree_bytes(c);
 225
 226         mutex_unlock(&c->btree_cache.lock);
 227         return ret;
 228 }
 229
 230 static ssize_t show_fs_alloc_debug(struct bch_fs *c, char *buf)
 231 {
 232         struct bch_fs_usage stats = bch2_fs_usage_read(c);
 233
 234         return scnprintf(buf, PAGE_SIZE,
 235                          "capacity:\t\t%llu\n"
 236                          "1 replicas:\n"
 237                          "\tmeta:\t\t%llu\n"
 238                          "\tdirty:\t\t%llu\n"
 239                          "\treserved:\t%llu\n"
 240                          "2 replicas:\n"
 241                          "\tmeta:\t\t%llu\n"
 242                          "\tdirty:\t\t%llu\n"
 243                          "\treserved:\t%llu\n"
 244                          "3 replicas:\n"
 245                          "\tmeta:\t\t%llu\n"
 246                          "\tdirty:\t\t%llu\n"
 247                          "\treserved:\t%llu\n"
 248                          "4 replicas:\n"
 249                          "\tmeta:\t\t%llu\n"
 250                          "\tdirty:\t\t%llu\n"
 251                          "\treserved:\t%llu\n"
 252                          "online reserved:\t%llu\n",
 253                          c->capacity,
 254                          stats.s[0].data[S_META],
 255                          stats.s[0].data[S_DIRTY],
 256                          stats.s[0].persistent_reserved,
 257                          stats.s[1].data[S_META],
 258                          stats.s[1].data[S_DIRTY],
 259                          stats.s[1].persistent_reserved,
 260                          stats.s[2].data[S_META],
 261                          stats.s[2].data[S_DIRTY],
 262                          stats.s[2].persistent_reserved,
 263                          stats.s[3].data[S_META],
 264                          stats.s[3].data[S_DIRTY],
 265                          stats.s[3].persistent_reserved,
 266                          stats.online_reserved);
 267 }
 268
 269 static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 270 {
 271         struct btree_iter iter;
 272         struct bkey_s_c k;
 273         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 274             nr_compressed_extents = 0,
 275             compressed_sectors_compressed = 0,
 276             compressed_sectors_uncompressed = 0;
 277
 278         if (!bch2_fs_running(c))
 279                 return -EPERM;
 280
 281         for_each_btree_key(&iter, c, BTREE_ID_EXTENTS, POS_MIN, 0, k)
 282                 if (k.k->type == BCH_EXTENT) {
 283                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 284                         const struct bch_extent_ptr *ptr;
 285                         struct bch_extent_crc_unpacked crc;
 286
 287                         extent_for_each_ptr_crc(e, ptr, crc) {
 288                                 if (crc.compression_type == BCH_COMPRESSION_NONE) {
 289                                         nr_uncompressed_extents++;
 290                                         uncompressed_sectors += e.k->size;
 291                                 } else {
 292                                         nr_compressed_extents++;
 293                                         compressed_sectors_compressed +=
 294                                                 crc.compressed_size;
 295                                         compressed_sectors_uncompressed +=
 296                                                 crc.uncompressed_size;
 297                                 }
 298
 299                                 /* only looking at the first ptr */
 300                                 break;
 301                         }
 302                 }
 303         bch2_btree_iter_unlock(&iter);
 304
 305         return scnprintf(buf, PAGE_SIZE,
 306                         "uncompressed data:\n"
 307                         "       nr extents:                     %llu\n"
 308                         "       size (bytes):                   %llu\n"
 309                         "compressed data:\n"
 310                         "       nr extents:                     %llu\n"
 311                         "       compressed size (bytes):        %llu\n"
 312                         "       uncompressed size (bytes):      %llu\n",
 313                         nr_uncompressed_extents,
 314                         uncompressed_sectors << 9,
 315                         nr_compressed_extents,
 316                         compressed_sectors_compressed << 9,
 317                         compressed_sectors_uncompressed << 9);
 318 }
 319
 320 SHOW(bch2_fs)
 321 {
 322         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 323
 324         sysfs_print(minor,                      c->minor);
 325         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 326
 327         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 328         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 329
 330         sysfs_print(block_size,                 block_bytes(c));
 331         sysfs_print(btree_node_size,            btree_bytes(c));
 332         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 333
 334         sysfs_print(read_realloc_races,
 335                     atomic_long_read(&c->read_realloc_races));
 336         sysfs_print(extent_migrate_done,
 337                     atomic_long_read(&c->extent_migrate_done));
 338         sysfs_print(extent_migrate_raced,
 339                     atomic_long_read(&c->extent_migrate_raced));
 340
 341         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 342
 343         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 344
 345         sysfs_print(pd_controllers_update_seconds,
 346                     c->pd_controllers_update_seconds);
 347
 348         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 349         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 350
 351         if (attr == &sysfs_rebalance_work)
 352                 return bch2_rebalance_work_show(c, buf);
 353
 354         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 355
 356         sysfs_printf(meta_replicas_have, "%u",  bch2_replicas_online(c, true));
 357         sysfs_printf(data_replicas_have, "%u",  bch2_replicas_online(c, false));
 358
 359         /* Debugging: */
 360
 361         if (attr == &sysfs_alloc_debug)
 362                 return show_fs_alloc_debug(c, buf);
 363
 364         if (attr == &sysfs_journal_debug)
 365                 return bch2_journal_print_debug(&c->journal, buf);
 366
 367         if (attr == &sysfs_journal_pins)
 368                 return bch2_journal_print_pins(&c->journal, buf);
 369
 370         if (attr == &sysfs_btree_updates)
 371                 return bch2_btree_updates_print(c, buf);
 372
 373         if (attr == &sysfs_dirty_btree_nodes)
 374                 return bch2_dirty_btree_nodes_print(c, buf);
 375
 376         if (attr == &sysfs_compression_stats)
 377                 return bch2_compression_stats(c, buf);
 378
 379 #define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
 380         BCH_DEBUG_PARAMS()
 381 #undef BCH_DEBUG_PARAM
 382
 383         return 0;
 384 }
 385
 386 STORE(__bch2_fs)
 387 {
 388         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 389
 390         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 391         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 392
 393         if (attr == &sysfs_btree_gc_periodic) {
 394                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 395                         ?: (ssize_t) size;
 396
 397                 wake_up_process(c->gc_thread);
 398                 return ret;
 399         }
 400
 401         if (attr == &sysfs_copy_gc_enabled) {
 402                 struct bch_dev *ca;
 403                 unsigned i;
 404                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 405                         ?: (ssize_t) size;
 406
 407                 for_each_member_device(ca, c, i)
 408                         if (ca->copygc_thread)
 409                                 wake_up_process(ca->copygc_thread);
 410                 return ret;
 411         }
 412
 413         if (attr == &sysfs_rebalance_enabled) {
 414                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 415                         ?: (ssize_t) size;
 416
 417                 rebalance_wakeup(c);
 418                 return ret;
 419         }
 420
 421         sysfs_strtoul(pd_controllers_update_seconds,
 422                       c->pd_controllers_update_seconds);
 423         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 424
 425         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 426
 427         /* Debugging: */
 428
 429 #define BCH_DEBUG_PARAM(name, description) sysfs_strtoul(name, c->name);
 430         BCH_DEBUG_PARAMS()
 431 #undef BCH_DEBUG_PARAM
 432
 433         if (!bch2_fs_running(c))
 434                 return -EPERM;
 435
 436         /* Debugging: */
 437
 438         if (attr == &sysfs_trigger_journal_flush)
 439                 bch2_journal_meta_async(&c->journal, NULL);
 440
 441         if (attr == &sysfs_trigger_btree_coalesce)
 442                 bch2_coalesce(c);
 443
 444         if (attr == &sysfs_trigger_gc)
 445                 bch2_gc(c);
 446
 447         if (attr == &sysfs_prune_cache) {
 448                 struct shrink_control sc;
 449
 450                 sc.gfp_mask = GFP_KERNEL;
 451                 sc.nr_to_scan = strtoul_or_return(buf);
 452                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 453         }
 454 #ifdef CONFIG_BCACHEFS_TESTS
 455         if (attr == &sysfs_perf_test) {
 456                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 457                 char *test              = strsep(&p, " \t\n");
 458                 char *nr_str            = strsep(&p, " \t\n");
 459                 char *threads_str       = strsep(&p, " \t\n");
 460                 unsigned threads;
 461                 u64 nr;
 462                 int ret = -EINVAL;
 463
 464                 if (threads_str &&
 465                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 466                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 467                         bch2_btree_perf_test(c, test, nr, threads);
 468                 else
 469                         size = ret;
 470                 kfree(tmp);
 471         }
 472 #endif
 473         return size;
 474 }
 475
 476 STORE(bch2_fs)
 477 {
 478         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 479
 480         mutex_lock(&c->state_lock);
 481         size = __bch2_fs_store(kobj, attr, buf, size);
 482         mutex_unlock(&c->state_lock);
 483
 484         return size;
 485 }
 486 SYSFS_OPS(bch2_fs);
 487
 488 struct attribute *bch2_fs_files[] = {
 489         &sysfs_minor,
 490         &sysfs_block_size,
 491         &sysfs_btree_node_size,
 492         &sysfs_btree_cache_size,
 493
 494         &sysfs_meta_replicas_have,
 495         &sysfs_data_replicas_have,
 496
 497         &sysfs_journal_write_delay_ms,
 498         &sysfs_journal_reclaim_delay_ms,
 499
 500         &sysfs_promote_whole_extents,
 501
 502         &sysfs_compression_stats,
 503
 504 #ifdef CONFIG_BCACHEFS_TESTS
 505         &sysfs_perf_test,
 506 #endif
 507         NULL
 508 };
 509
 510 /* internal dir - just a wrapper */
 511
 512 SHOW(bch2_fs_internal)
 513 {
 514         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 515         return bch2_fs_show(&c->kobj, attr, buf);
 516 }
 517
 518 STORE(bch2_fs_internal)
 519 {
 520         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 521         return bch2_fs_store(&c->kobj, attr, buf, size);
 522 }
 523 SYSFS_OPS(bch2_fs_internal);
 524
 525 struct attribute *bch2_fs_internal_files[] = {
 526         &sysfs_alloc_debug,
 527         &sysfs_journal_debug,
 528         &sysfs_journal_pins,
 529         &sysfs_btree_updates,
 530         &sysfs_dirty_btree_nodes,
 531
 532         &sysfs_read_realloc_races,
 533         &sysfs_extent_migrate_done,
 534         &sysfs_extent_migrate_raced,
 535
 536         &sysfs_trigger_journal_flush,
 537         &sysfs_trigger_btree_coalesce,
 538         &sysfs_trigger_gc,
 539         &sysfs_prune_cache,
 540
 541         &sysfs_copy_gc_enabled,
 542
 543         &sysfs_rebalance_enabled,
 544         &sysfs_rebalance_work,
 545         sysfs_pd_controller_files(rebalance),
 546
 547         &sysfs_internal_uuid,
 548
 549 #define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
 550         BCH_DEBUG_PARAMS()
 551 #undef BCH_DEBUG_PARAM
 552
 553         NULL
 554 };
 555
 556 /* options */
 557
 558 SHOW(bch2_fs_opts_dir)
 559 {
 560         char *out = buf, *end = buf + PAGE_SIZE;
 561         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 562         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 563         int id = opt - bch2_opt_table;
 564         u64 v = bch2_opt_get_by_id(&c->opts, id);
 565
 566         out += bch2_opt_to_text(c, out, end - out, opt, v, OPT_SHOW_FULL_LIST);
 567         out += scnprintf(out, end - out, "\n");
 568
 569         return out - buf;
 570 }
 571
 572 STORE(bch2_fs_opts_dir)
 573 {
 574         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 575         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 576         int ret, id = opt - bch2_opt_table;
 577         char *tmp;
 578         u64 v;
 579
 580         tmp = kstrdup(buf, GFP_KERNEL);
 581         if (!tmp)
 582                 return -ENOMEM;
 583
 584         ret = bch2_opt_parse(c, opt, strim(tmp), &v);
 585         kfree(tmp);
 586
 587         if (ret < 0)
 588                 return ret;
 589
 590         if (id == Opt_compression ||
 591             id == Opt_background_compression) {
 592                 int ret = bch2_check_set_has_compressed_data(c, v);
 593                 if (ret) {
 594                         mutex_unlock(&c->sb_lock);
 595                         return ret;
 596                 }
 597         }
 598
 599         if (opt->set_sb != SET_NO_SB_OPT) {
 600                 mutex_lock(&c->sb_lock);
 601                 opt->set_sb(c->disk_sb.sb, v);
 602                 bch2_write_super(c);
 603                 mutex_unlock(&c->sb_lock);
 604         }
 605
 606         bch2_opt_set_by_id(&c->opts, id, v);
 607
 608         if ((id == Opt_background_target ||
 609              id == Opt_background_compression) && v) {
 610                 bch2_rebalance_add_work(c, S64_MAX);
 611                 rebalance_wakeup(c);
 612         }
 613
 614         return size;
 615 }
 616 SYSFS_OPS(bch2_fs_opts_dir);
 617
 618 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 619
 620 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 621 {
 622         const struct bch_option *i;
 623         int ret;
 624
 625         for (i = bch2_opt_table;
 626              i < bch2_opt_table + bch2_opts_nr;
 627              i++) {
 628                 if (i->mode == OPT_INTERNAL)
 629                         continue;
 630
 631                 ret = sysfs_create_file(kobj, &i->attr);
 632                 if (ret)
 633                         return ret;
 634         }
 635
 636         return 0;
 637 }
 638
 639 /* time stats */
 640
 641 SHOW(bch2_fs_time_stats)
 642 {
 643         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 644
 645 #define x(name)                                         \
 646         if (attr == &sysfs_time_stat_##name)                            \
 647                 return bch2_time_stats_print(&c->times[BCH_TIME_##name],\
 648                                              buf, PAGE_SIZE);
 649         BCH_TIME_STATS()
 650 #undef x
 651
 652         return 0;
 653 }
 654
 655 STORE(bch2_fs_time_stats)
 656 {
 657         return size;
 658 }
 659 SYSFS_OPS(bch2_fs_time_stats);
 660
 661 struct attribute *bch2_fs_time_stats_files[] = {
 662 #define x(name)                                         \
 663         &sysfs_time_stat_##name,
 664         BCH_TIME_STATS()
 665 #undef x
 666         NULL
 667 };
 668
 669 typedef unsigned (bucket_map_fn)(struct bch_fs *, struct bch_dev *,
 670                                  size_t, void *);
 671
 672 static unsigned bucket_last_io_fn(struct bch_fs *c, struct bch_dev *ca,
 673                                   size_t b, void *private)
 674 {
 675         int rw = (private ? 1 : 0);
 676
 677         return bucket_last_io(c, bucket(ca, b), rw);
 678 }
 679
 680 static unsigned bucket_sectors_used_fn(struct bch_fs *c, struct bch_dev *ca,
 681                                        size_t b, void *private)
 682 {
 683         struct bucket *g = bucket(ca, b);
 684         return bucket_sectors_used(g->mark);
 685 }
 686
 687 static unsigned bucket_oldest_gen_fn(struct bch_fs *c, struct bch_dev *ca,
 688                                      size_t b, void *private)
 689 {
 690         return bucket_gc_gen(ca, b);
 691 }
 692
 693 static int unsigned_cmp(const void *_l, const void *_r)
 694 {
 695         unsigned l = *((unsigned *) _l);
 696         unsigned r = *((unsigned *) _r);
 697
 698         return (l > r) - (l < r);
 699 }
 700
 701 static ssize_t show_quantiles(struct bch_fs *c, struct bch_dev *ca,
 702                               char *buf, bucket_map_fn *fn, void *private)
 703 {
 704         size_t i, n;
 705         /* Compute 31 quantiles */
 706         unsigned q[31], *p;
 707         ssize_t ret = 0;
 708
 709         down_read(&ca->bucket_lock);
 710         n = ca->mi.nbuckets;
 711
 712         p = vzalloc(n * sizeof(unsigned));
 713         if (!p) {
 714                 up_read(&ca->bucket_lock);
 715                 return -ENOMEM;
 716         }
 717
 718         for (i = ca->mi.first_bucket; i < n; i++)
 719                 p[i] = fn(c, ca, i, private);
 720
 721         sort(p, n, sizeof(unsigned), unsigned_cmp, NULL);
 722         up_read(&ca->bucket_lock);
 723
 724         while (n &&
 725                !p[n - 1])
 726                 --n;
 727
 728         for (i = 0; i < ARRAY_SIZE(q); i++)
 729                 q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
 730
 731         vfree(p);
 732
 733         for (i = 0; i < ARRAY_SIZE(q); i++)
 734                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 735                                  "%u ", q[i]);
 736         buf[ret - 1] = '\n';
 737
 738         return ret;
 739 }
 740
 741 static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
 742 {
 743         enum alloc_reserve i;
 744         ssize_t ret;
 745
 746         spin_lock(&ca->freelist_lock);
 747
 748         ret = scnprintf(buf, PAGE_SIZE,
 749                         "free_inc:\t%zu\t%zu\n",
 750                         fifo_used(&ca->free_inc),
 751                         ca->free_inc.size);
 752
 753         for (i = 0; i < RESERVE_NR; i++)
 754                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 755                                  "free[%u]:\t%zu\t%zu\n", i,
 756                                  fifo_used(&ca->free[i]),
 757                                  ca->free[i].size);
 758
 759         spin_unlock(&ca->freelist_lock);
 760
 761         return ret;
 762 }
 763
 764 static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
 765 {
 766         struct bch_fs *c = ca->fs;
 767         struct bch_dev_usage stats = bch2_dev_usage_read(c, ca);
 768
 769         return scnprintf(buf, PAGE_SIZE,
 770                 "free_inc:               %zu/%zu\n"
 771                 "free[RESERVE_BTREE]:    %zu/%zu\n"
 772                 "free[RESERVE_MOVINGGC]: %zu/%zu\n"
 773                 "free[RESERVE_NONE]:     %zu/%zu\n"
 774                 "buckets:\n"
 775                 "    capacity:           %llu\n"
 776                 "    alloc:              %llu\n"
 777                 "    sb:                 %llu\n"
 778                 "    journal:            %llu\n"
 779                 "    meta:               %llu\n"
 780                 "    user:               %llu\n"
 781                 "    cached:             %llu\n"
 782                 "    available:          %llu\n"
 783                 "sectors:\n"
 784                 "    sb:                 %llu\n"
 785                 "    journal:            %llu\n"
 786                 "    meta:               %llu\n"
 787                 "    user:               %llu\n"
 788                 "    cached:             %llu\n"
 789                 "freelist_wait:          %s\n"
 790                 "open buckets:           %u/%u (reserved %u)\n"
 791                 "open_buckets_wait:      %s\n",
 792                 fifo_used(&ca->free_inc),               ca->free_inc.size,
 793                 fifo_used(&ca->free[RESERVE_BTREE]),    ca->free[RESERVE_BTREE].size,
 794                 fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size,
 795                 fifo_used(&ca->free[RESERVE_NONE]),     ca->free[RESERVE_NONE].size,
 796                 ca->mi.nbuckets - ca->mi.first_bucket,
 797                 stats.buckets_alloc,
 798                 stats.buckets[BCH_DATA_SB],
 799                 stats.buckets[BCH_DATA_JOURNAL],
 800                 stats.buckets[BCH_DATA_BTREE],
 801                 stats.buckets[BCH_DATA_USER],
 802                 stats.buckets[BCH_DATA_CACHED],
 803                 __dev_buckets_available(ca, stats),
 804                 stats.sectors[BCH_DATA_SB],
 805                 stats.sectors[BCH_DATA_JOURNAL],
 806                 stats.sectors[BCH_DATA_BTREE],
 807                 stats.sectors[BCH_DATA_USER],
 808                 stats.sectors[BCH_DATA_CACHED],
 809                 c->freelist_wait.list.first             ? "waiting" : "empty",
 810                 c->open_buckets_nr_free, OPEN_BUCKETS_COUNT, BTREE_NODE_RESERVE,
 811                 c->open_buckets_wait.list.first         ? "waiting" : "empty");
 812 }
 813
 814 static const char * const bch2_rw[] = {
 815         "read",
 816         "write",
 817         NULL
 818 };
 819
 820 static ssize_t show_dev_iodone(struct bch_dev *ca, char *buf)
 821 {
 822         char *out = buf, *end = buf + PAGE_SIZE;
 823         int rw, i, cpu;
 824
 825         for (rw = 0; rw < 2; rw++) {
 826                 out += scnprintf(out, end - out, "%s:\n", bch2_rw[rw]);
 827
 828                 for (i = 1; i < BCH_DATA_NR; i++) {
 829                         u64 n = 0;
 830
 831                         for_each_possible_cpu(cpu)
 832                                 n += per_cpu_ptr(ca->io_done, cpu)->sectors[rw][i];
 833
 834                         out += scnprintf(out, end - out, "%-12s:%12llu\n",
 835                                          bch2_data_types[i], n << 9);
 836                 }
 837         }
 838
 839         return out - buf;
 840 }
 841
 842 SHOW(bch2_dev)
 843 {
 844         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 845         struct bch_fs *c = ca->fs;
 846         char *out = buf, *end = buf + PAGE_SIZE;
 847
 848         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 849
 850         sysfs_print(bucket_size,        bucket_bytes(ca));
 851         sysfs_print(block_size,         block_bytes(c));
 852         sysfs_print(first_bucket,       ca->mi.first_bucket);
 853         sysfs_print(nbuckets,           ca->mi.nbuckets);
 854         sysfs_print(durability,         ca->mi.durability);
 855         sysfs_print(discard,            ca->mi.discard);
 856
 857         if (attr == &sysfs_label) {
 858                 if (ca->mi.group) {
 859                         mutex_lock(&c->sb_lock);
 860                         out += bch2_disk_path_print(&c->disk_sb, out, end - out,
 861                                                     ca->mi.group - 1);
 862                         mutex_unlock(&c->sb_lock);
 863                 } else {
 864                         out += scnprintf(out, end - out, "none");
 865                 }
 866
 867                 out += scnprintf(out, end - out, "\n");
 868                 return out - buf;
 869         }
 870
 871         if (attr == &sysfs_has_data) {
 872                 out += bch2_scnprint_flag_list(out, end - out,
 873                                                bch2_data_types,
 874                                                bch2_dev_has_data(c, ca));
 875                 out += scnprintf(out, end - out, "\n");
 876                 return out - buf;
 877         }
 878
 879         sysfs_pd_controller_show(copy_gc, &ca->copygc_pd);
 880
 881         if (attr == &sysfs_cache_replacement_policy) {
 882                 out += bch2_scnprint_string_list(out, end - out,
 883                                                  bch2_cache_replacement_policies,
 884                                                  ca->mi.replacement);
 885                 out += scnprintf(out, end - out, "\n");
 886                 return out - buf;
 887         }
 888
 889         if (attr == &sysfs_state_rw) {
 890                 out += bch2_scnprint_string_list(out, end - out,
 891                                                  bch2_dev_state,
 892                                                  ca->mi.state);
 893                 out += scnprintf(out, end - out, "\n");
 894                 return out - buf;
 895         }
 896
 897         if (attr == &sysfs_iodone)
 898                 return show_dev_iodone(ca, buf);
 899
 900         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 901         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 902
 903         if (attr == &sysfs_io_latency_stats_read)
 904                 return bch2_time_stats_print(&ca->io_latency[READ], buf, PAGE_SIZE);
 905         if (attr == &sysfs_io_latency_stats_write)
 906                 return bch2_time_stats_print(&ca->io_latency[WRITE], buf, PAGE_SIZE);
 907
 908         sysfs_printf(congested,                 "%u%%",
 909                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 910                      * 100 / CONGESTED_MAX);
 911
 912         if (attr == &sysfs_bucket_quantiles_last_read)
 913                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 0);
 914         if (attr == &sysfs_bucket_quantiles_last_write)
 915                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 1);
 916         if (attr == &sysfs_bucket_quantiles_fragmentation)
 917                 return show_quantiles(c, ca, buf, bucket_sectors_used_fn, NULL);
 918         if (attr == &sysfs_bucket_quantiles_oldest_gen)
 919                 return show_quantiles(c, ca, buf, bucket_oldest_gen_fn, NULL);
 920
 921         if (attr == &sysfs_reserve_stats)
 922                 return show_reserve_stats(ca, buf);
 923         if (attr == &sysfs_alloc_debug)
 924                 return show_dev_alloc_debug(ca, buf);
 925
 926         return 0;
 927 }
 928
 929 STORE(bch2_dev)
 930 {
 931         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 932         struct bch_fs *c = ca->fs;
 933         struct bch_member *mi;
 934
 935         sysfs_pd_controller_store(copy_gc, &ca->copygc_pd);
 936
 937         if (attr == &sysfs_discard) {
 938                 bool v = strtoul_or_return(buf);
 939
 940                 mutex_lock(&c->sb_lock);
 941                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 942
 943                 if (v != BCH_MEMBER_DISCARD(mi)) {
 944                         SET_BCH_MEMBER_DISCARD(mi, v);
 945                         bch2_write_super(c);
 946                 }
 947                 mutex_unlock(&c->sb_lock);
 948         }
 949
 950         if (attr == &sysfs_cache_replacement_policy) {
 951                 ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
 952
 953                 if (v < 0)
 954                         return v;
 955
 956                 mutex_lock(&c->sb_lock);
 957                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 958
 959                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
 960                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
 961                         bch2_write_super(c);
 962                 }
 963                 mutex_unlock(&c->sb_lock);
 964         }
 965
 966         if (attr == &sysfs_label) {
 967                 char *tmp;
 968                 int ret;
 969
 970                 tmp = kstrdup(buf, GFP_KERNEL);
 971                 if (!tmp)
 972                         return -ENOMEM;
 973
 974                 ret = bch2_dev_group_set(c, ca, strim(tmp));
 975                 kfree(tmp);
 976                 if (ret)
 977                         return ret;
 978         }
 979
 980         if (attr == &sysfs_wake_allocator)
 981                 bch2_wake_allocator(ca);
 982
 983         return size;
 984 }
 985 SYSFS_OPS(bch2_dev);
 986
 987 struct attribute *bch2_dev_files[] = {
 988         &sysfs_uuid,
 989         &sysfs_bucket_size,
 990         &sysfs_block_size,
 991         &sysfs_first_bucket,
 992         &sysfs_nbuckets,
 993         &sysfs_durability,
 994
 995         /* settings: */
 996         &sysfs_discard,
 997         &sysfs_cache_replacement_policy,
 998         &sysfs_state_rw,
 999         &sysfs_label,
1000
1001         &sysfs_has_data,
1002         &sysfs_iodone,
1003
1004         &sysfs_io_latency_read,
1005         &sysfs_io_latency_write,
1006         &sysfs_io_latency_stats_read,
1007         &sysfs_io_latency_stats_write,
1008         &sysfs_congested,
1009
1010         /* alloc info - other stats: */
1011         &sysfs_bucket_quantiles_last_read,
1012         &sysfs_bucket_quantiles_last_write,
1013         &sysfs_bucket_quantiles_fragmentation,
1014         &sysfs_bucket_quantiles_oldest_gen,
1015
1016         &sysfs_reserve_stats,
1017
1018         /* debug: */
1019         &sysfs_alloc_debug,
1020         &sysfs_wake_allocator,
1021
1022         sysfs_pd_controller_files(copy_gc),
1023         NULL
1024 };
1025
1026 #endif  /* _BCACHEFS_SYSFS_H_ */