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 #include "bcachefs.h"
   9 #include "alloc.h"
  10 #include "compress.h"
  11 #include "sysfs.h"
  12 #include "btree_cache.h"
  13 #include "btree_iter.h"
  14 #include "btree_update.h"
  15 #include "btree_gc.h"
  16 #include "buckets.h"
  17 #include "inode.h"
  18 #include "journal.h"
  19 #include "keylist.h"
  20 #include "move.h"
  21 #include "opts.h"
  22 #include "super-io.h"
  23 #include "tier.h"
  24
  25 #include <linux/blkdev.h>
  26 #include <linux/sort.h>
  27
  28 #include "util.h"
  29
  30 #define SYSFS_OPS(type)                                                 \
  31 struct sysfs_ops type ## _sysfs_ops = {                                 \
  32         .show   = type ## _show,                                        \
  33         .store  = type ## _store                                        \
  34 }
  35
  36 #define SHOW(fn)                                                        \
  37 static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
  38                            char *buf)                                   \
  39
  40 #define STORE(fn)                                                       \
  41 static ssize_t fn ## _store(struct kobject *kobj, struct attribute *attr,\
  42                             const char *buf, size_t size)               \
  43
  44 #define __sysfs_attribute(_name, _mode)                                 \
  45         static struct attribute sysfs_##_name =                         \
  46                 { .name = #_name, .mode = _mode }
  47
  48 #define write_attribute(n)      __sysfs_attribute(n, S_IWUSR)
  49 #define read_attribute(n)       __sysfs_attribute(n, S_IRUGO)
  50 #define rw_attribute(n)         __sysfs_attribute(n, S_IRUGO|S_IWUSR)
  51
  52 #define sysfs_printf(file, fmt, ...)                                    \
  53 do {                                                                    \
  54         if (attr == &sysfs_ ## file)                                    \
  55                 return snprintf(buf, PAGE_SIZE, fmt "\n", __VA_ARGS__); \
  56 } while (0)
  57
  58 #define sysfs_print(file, var)                                          \
  59 do {                                                                    \
  60         if (attr == &sysfs_ ## file)                                    \
  61                 return snprint(buf, PAGE_SIZE, var);                    \
  62 } while (0)
  63
  64 #define sysfs_hprint(file, val)                                         \
  65 do {                                                                    \
  66         if (attr == &sysfs_ ## file) {                                  \
  67                 ssize_t ret = bch2_hprint(buf, val);                    \
  68                 strcat(buf, "\n");                                      \
  69                 return ret + 1;                                         \
  70         }                                                               \
  71 } while (0)
  72
  73 #define var_printf(_var, fmt)   sysfs_printf(_var, fmt, var(_var))
  74 #define var_print(_var)         sysfs_print(_var, var(_var))
  75 #define var_hprint(_var)        sysfs_hprint(_var, var(_var))
  76
  77 #define sysfs_strtoul(file, var)                                        \
  78 do {                                                                    \
  79         if (attr == &sysfs_ ## file)                                    \
  80                 return strtoul_safe(buf, var) ?: (ssize_t) size;        \
  81 } while (0)
  82
  83 #define sysfs_strtoul_clamp(file, var, min, max)                        \
  84 do {                                                                    \
  85         if (attr == &sysfs_ ## file)                                    \
  86                 return strtoul_safe_clamp(buf, var, min, max)           \
  87                         ?: (ssize_t) size;                              \
  88 } while (0)
  89
  90 #define strtoul_or_return(cp)                                           \
  91 ({                                                                      \
  92         unsigned long _v;                                               \
  93         int _r = kstrtoul(cp, 10, &_v);                                 \
  94         if (_r)                                                         \
  95                 return _r;                                              \
  96         _v;                                                             \
  97 })
  98
  99 #define strtoul_restrict_or_return(cp, min, max)                        \
 100 ({                                                                      \
 101         unsigned long __v = 0;                                          \
 102         int _r = strtoul_safe_restrict(cp, __v, min, max);              \
 103         if (_r)                                                         \
 104                 return _r;                                              \
 105         __v;                                                            \
 106 })
 107
 108 #define strtoi_h_or_return(cp)                                          \
 109 ({                                                                      \
 110         u64 _v;                                                         \
 111         int _r = strtoi_h(cp, &_v);                                     \
 112         if (_r)                                                         \
 113                 return _r;                                              \
 114         _v;                                                             \
 115 })
 116
 117 #define sysfs_hatoi(file, var)                                          \
 118 do {                                                                    \
 119         if (attr == &sysfs_ ## file)                                    \
 120                 return strtoi_h(buf, &var) ?: (ssize_t) size;           \
 121 } while (0)
 122
 123 write_attribute(trigger_btree_coalesce);
 124 write_attribute(trigger_gc);
 125 write_attribute(prune_cache);
 126
 127 read_attribute(uuid);
 128 read_attribute(minor);
 129 read_attribute(bucket_size);
 130 read_attribute(bucket_size_bytes);
 131 read_attribute(block_size);
 132 read_attribute(block_size_bytes);
 133 read_attribute(btree_node_size);
 134 read_attribute(btree_node_size_bytes);
 135 read_attribute(first_bucket);
 136 read_attribute(nbuckets);
 137 read_attribute(tree_depth);
 138 read_attribute(root_usage_percent);
 139 read_attribute(read_priority_stats);
 140 read_attribute(write_priority_stats);
 141 read_attribute(fragmentation_stats);
 142 read_attribute(oldest_gen_stats);
 143 read_attribute(reserve_stats);
 144 read_attribute(btree_cache_size);
 145 read_attribute(cache_available_percent);
 146 read_attribute(compression_stats);
 147 read_attribute(written);
 148 read_attribute(btree_written);
 149 read_attribute(metadata_written);
 150 read_attribute(journal_debug);
 151 write_attribute(journal_flush);
 152 read_attribute(internal_uuid);
 153
 154 read_attribute(btree_gc_running);
 155
 156 read_attribute(btree_nodes);
 157 read_attribute(btree_used_percent);
 158 read_attribute(average_key_size);
 159 read_attribute(available_buckets);
 160 read_attribute(free_buckets);
 161 read_attribute(dirty_data);
 162 read_attribute(dirty_bytes);
 163 read_attribute(dirty_buckets);
 164 read_attribute(cached_data);
 165 read_attribute(cached_bytes);
 166 read_attribute(cached_buckets);
 167 read_attribute(meta_buckets);
 168 read_attribute(alloc_buckets);
 169 read_attribute(has_data);
 170 read_attribute(has_metadata);
 171 read_attribute(bset_tree_stats);
 172 read_attribute(alloc_debug);
 173
 174 read_attribute(cache_read_races);
 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
 182 rw_attribute(foreground_write_ratelimit_enabled);
 183 rw_attribute(copy_gc_enabled);
 184 sysfs_pd_controller_attribute(copy_gc);
 185
 186 rw_attribute(tier);
 187 rw_attribute(tiering_enabled);
 188 rw_attribute(tiering_percent);
 189 sysfs_pd_controller_attribute(tiering);
 190
 191 sysfs_pd_controller_attribute(foreground_write);
 192
 193 rw_attribute(pd_controllers_update_seconds);
 194
 195 rw_attribute(foreground_target_percent);
 196
 197 read_attribute(meta_replicas_have);
 198 read_attribute(data_replicas_have);
 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 BCH_OPT(_name, _mode, ...)                                      \
 207         static struct attribute sysfs_opt_##_name = {                   \
 208                 .name = #_name, .mode = _mode,                          \
 209         };
 210
 211         BCH_VISIBLE_OPTS()
 212 #undef BCH_OPT
 213
 214 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 215         sysfs_time_stats_attribute(name, frequency_units, duration_units);
 216         BCH_TIME_STATS()
 217 #undef BCH_TIME_STAT
 218
 219 static struct attribute sysfs_state_rw = {
 220         .name = "state",
 221         .mode = S_IRUGO
 222 };
 223
 224 static int bch2_bset_print_stats(struct bch_fs *c, char *buf)
 225 {
 226         struct bset_stats stats;
 227         size_t nodes = 0;
 228         struct btree *b;
 229         struct bucket_table *tbl;
 230         struct rhash_head *pos;
 231         unsigned iter;
 232
 233         memset(&stats, 0, sizeof(stats));
 234
 235         rcu_read_lock();
 236         for_each_cached_btree(b, c, tbl, iter, pos) {
 237                 bch2_btree_keys_stats(b, &stats);
 238                 nodes++;
 239         }
 240         rcu_read_unlock();
 241
 242         return snprintf(buf, PAGE_SIZE,
 243                         "btree nodes:           %zu\n"
 244                         "written sets:          %zu\n"
 245                         "written key bytes:     %zu\n"
 246                         "unwritten sets:                %zu\n"
 247                         "unwritten key bytes:   %zu\n"
 248                         "no table sets:         %zu\n"
 249                         "no table key bytes:    %zu\n"
 250                         "floats:                        %zu\n"
 251                         "failed unpacked:       %zu\n"
 252                         "failed prev:           %zu\n"
 253                         "failed overflow:       %zu\n",
 254                         nodes,
 255                         stats.sets[BSET_RO_AUX_TREE].nr,
 256                         stats.sets[BSET_RO_AUX_TREE].bytes,
 257                         stats.sets[BSET_RW_AUX_TREE].nr,
 258                         stats.sets[BSET_RW_AUX_TREE].bytes,
 259                         stats.sets[BSET_NO_AUX_TREE].nr,
 260                         stats.sets[BSET_NO_AUX_TREE].bytes,
 261                         stats.floats,
 262                         stats.failed_unpacked,
 263                         stats.failed_prev,
 264                         stats.failed_overflow);
 265 }
 266
 267 static unsigned bch2_root_usage(struct bch_fs *c)
 268 {
 269         unsigned bytes = 0;
 270         struct bkey_packed *k;
 271         struct btree *b;
 272         struct btree_node_iter iter;
 273
 274         goto lock_root;
 275
 276         do {
 277                 six_unlock_read(&b->lock);
 278 lock_root:
 279                 b = c->btree_roots[BTREE_ID_EXTENTS].b;
 280                 six_lock_read(&b->lock);
 281         } while (b != c->btree_roots[BTREE_ID_EXTENTS].b);
 282
 283         for_each_btree_node_key(b, k, &iter, btree_node_is_extents(b))
 284                 bytes += bkey_bytes(k);
 285
 286         six_unlock_read(&b->lock);
 287
 288         return (bytes * 100) / btree_bytes(c);
 289 }
 290
 291 static size_t bch2_btree_cache_size(struct bch_fs *c)
 292 {
 293         size_t ret = 0;
 294         struct btree *b;
 295
 296         mutex_lock(&c->btree_cache_lock);
 297         list_for_each_entry(b, &c->btree_cache, list)
 298                 ret += btree_bytes(c);
 299
 300         mutex_unlock(&c->btree_cache_lock);
 301         return ret;
 302 }
 303
 304 static unsigned bch2_fs_available_percent(struct bch_fs *c)
 305 {
 306         return div64_u64((u64) sectors_available(c) * 100,
 307                          c->capacity ?: 1);
 308 }
 309
 310 #if 0
 311 static unsigned bch2_btree_used(struct bch_fs *c)
 312 {
 313         return div64_u64(c->gc_stats.key_bytes * 100,
 314                          (c->gc_stats.nodes ?: 1) * btree_bytes(c));
 315 }
 316
 317 static unsigned bch2_average_key_size(struct bch_fs *c)
 318 {
 319         return c->gc_stats.nkeys
 320                 ? div64_u64(c->gc_stats.data, c->gc_stats.nkeys)
 321                 : 0;
 322 }
 323 #endif
 324
 325 static ssize_t show_fs_alloc_debug(struct bch_fs *c, char *buf)
 326 {
 327         struct bch_fs_usage stats = bch2_fs_usage_read(c);
 328
 329         return scnprintf(buf, PAGE_SIZE,
 330                          "capacity:\t\t%llu\n"
 331                          "compressed:\n"
 332                          "\tmeta:\t\t%llu\n"
 333                          "\tdirty:\t\t%llu\n"
 334                          "\tcached:\t\t%llu\n"
 335                          "uncompressed:\n"
 336                          "\tmeta:\t\t%llu\n"
 337                          "\tdirty:\t\t%llu\n"
 338                          "\tcached:\t\t%llu\n"
 339                          "persistent reserved sectors:\t%llu\n"
 340                          "online reserved sectors:\t%llu\n",
 341                          c->capacity,
 342                          stats.s[S_COMPRESSED][S_META],
 343                          stats.s[S_COMPRESSED][S_DIRTY],
 344                          stats.s[S_COMPRESSED][S_CACHED],
 345                          stats.s[S_UNCOMPRESSED][S_META],
 346                          stats.s[S_UNCOMPRESSED][S_DIRTY],
 347                          stats.s[S_UNCOMPRESSED][S_CACHED],
 348                          stats.persistent_reserved,
 349                          stats.online_reserved);
 350 }
 351
 352 static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 353 {
 354         struct btree_iter iter;
 355         struct bkey_s_c k;
 356         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 357             nr_compressed_extents = 0,
 358             compressed_sectors_compressed = 0,
 359             compressed_sectors_uncompressed = 0;
 360
 361         for_each_btree_key(&iter, c, BTREE_ID_EXTENTS, POS_MIN, k)
 362                 if (k.k->type == BCH_EXTENT) {
 363                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 364                         const struct bch_extent_ptr *ptr;
 365                         const union bch_extent_crc *crc;
 366
 367                         extent_for_each_ptr_crc(e, ptr, crc) {
 368                                 if (crc_compression_type(crc) == BCH_COMPRESSION_NONE) {
 369                                         nr_uncompressed_extents++;
 370                                         uncompressed_sectors += e.k->size;
 371                                 } else {
 372                                         nr_compressed_extents++;
 373                                         compressed_sectors_compressed +=
 374                                                 crc_compressed_size(e.k, crc);
 375                                         compressed_sectors_uncompressed +=
 376                                                 crc_uncompressed_size(e.k, crc);
 377                                 }
 378
 379                                 /* only looking at the first ptr */
 380                                 break;
 381                         }
 382                 }
 383         bch2_btree_iter_unlock(&iter);
 384
 385         return snprintf(buf, PAGE_SIZE,
 386                         "uncompressed data:\n"
 387                         "       nr extents:                     %llu\n"
 388                         "       size (bytes):                   %llu\n"
 389                         "compressed data:\n"
 390                         "       nr extents:                     %llu\n"
 391                         "       compressed size (bytes):        %llu\n"
 392                         "       uncompressed size (bytes):      %llu\n",
 393                         nr_uncompressed_extents,
 394                         uncompressed_sectors << 9,
 395                         nr_compressed_extents,
 396                         compressed_sectors_compressed << 9,
 397                         compressed_sectors_uncompressed << 9);
 398 }
 399
 400 SHOW(bch2_fs)
 401 {
 402         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 403
 404         sysfs_print(minor,                      c->minor);
 405
 406         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 407         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 408
 409         sysfs_hprint(block_size,                block_bytes(c));
 410         sysfs_print(block_size_bytes,           block_bytes(c));
 411         sysfs_hprint(btree_node_size,           c->sb.btree_node_size << 9);
 412         sysfs_print(btree_node_size_bytes,      c->sb.btree_node_size << 9);
 413
 414         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 415         sysfs_print(cache_available_percent,    bch2_fs_available_percent(c));
 416
 417         sysfs_print(btree_gc_running,           c->gc_pos.phase != GC_PHASE_DONE);
 418
 419 #if 0
 420         /* XXX: reimplement */
 421         sysfs_print(btree_used_percent, bch2_btree_used(c));
 422         sysfs_print(btree_nodes,        c->gc_stats.nodes);
 423         sysfs_hprint(average_key_size,  bch2_average_key_size(c));
 424 #endif
 425
 426         sysfs_print(cache_read_races,
 427                     atomic_long_read(&c->cache_read_races));
 428
 429         sysfs_printf(foreground_write_ratelimit_enabled, "%i",
 430                      c->foreground_write_ratelimit_enabled);
 431         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 432         sysfs_pd_controller_show(foreground_write, &c->foreground_write_pd);
 433
 434         sysfs_print(pd_controllers_update_seconds,
 435                     c->pd_controllers_update_seconds);
 436         sysfs_print(foreground_target_percent, c->foreground_target_percent);
 437
 438         sysfs_printf(tiering_enabled,           "%i", c->tiering_enabled);
 439         sysfs_print(tiering_percent,            c->tiering_percent);
 440
 441         sysfs_pd_controller_show(tiering,       &c->tiers[1].pd); /* XXX */
 442
 443         sysfs_printf(meta_replicas_have, "%u",  c->sb.meta_replicas_have);
 444         sysfs_printf(data_replicas_have, "%u",  c->sb.data_replicas_have);
 445
 446         /* Debugging: */
 447
 448         if (attr == &sysfs_journal_debug)
 449                 return bch2_journal_print_debug(&c->journal, buf);
 450
 451 #define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
 452         BCH_DEBUG_PARAMS()
 453 #undef BCH_DEBUG_PARAM
 454
 455         if (!bch2_fs_running(c))
 456                 return -EPERM;
 457
 458         if (attr == &sysfs_bset_tree_stats)
 459                 return bch2_bset_print_stats(c, buf);
 460         if (attr == &sysfs_alloc_debug)
 461                 return show_fs_alloc_debug(c, buf);
 462
 463         sysfs_print(tree_depth, c->btree_roots[BTREE_ID_EXTENTS].b->level);
 464         sysfs_print(root_usage_percent,         bch2_root_usage(c));
 465
 466         if (attr == &sysfs_compression_stats)
 467                 return bch2_compression_stats(c, buf);
 468
 469         sysfs_printf(internal_uuid, "%pU", c->sb.uuid.b);
 470
 471         return 0;
 472 }
 473
 474 STORE(__bch2_fs)
 475 {
 476         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 477
 478         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 479         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 480
 481         sysfs_strtoul(foreground_write_ratelimit_enabled,
 482                       c->foreground_write_ratelimit_enabled);
 483
 484         if (attr == &sysfs_copy_gc_enabled) {
 485                 struct bch_dev *ca;
 486                 unsigned i;
 487                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 488                         ?: (ssize_t) size;
 489
 490                 for_each_member_device(ca, c, i)
 491                         if (ca->moving_gc_read)
 492                                 wake_up_process(ca->moving_gc_read);
 493                 return ret;
 494         }
 495
 496         if (attr == &sysfs_tiering_enabled) {
 497                 ssize_t ret = strtoul_safe(buf, c->tiering_enabled)
 498                         ?: (ssize_t) size;
 499
 500                 bch2_tiering_start(c); /* issue wakeups */
 501                 return ret;
 502         }
 503
 504         sysfs_pd_controller_store(foreground_write, &c->foreground_write_pd);
 505
 506         sysfs_strtoul(pd_controllers_update_seconds,
 507                       c->pd_controllers_update_seconds);
 508         sysfs_strtoul(foreground_target_percent, c->foreground_target_percent);
 509
 510         sysfs_strtoul(tiering_percent,          c->tiering_percent);
 511         sysfs_pd_controller_store(tiering,      &c->tiers[1].pd); /* XXX */
 512
 513         /* Debugging: */
 514
 515 #define BCH_DEBUG_PARAM(name, description) sysfs_strtoul(name, c->name);
 516         BCH_DEBUG_PARAMS()
 517 #undef BCH_DEBUG_PARAM
 518
 519         if (!bch2_fs_running(c))
 520                 return -EPERM;
 521
 522         if (attr == &sysfs_journal_flush) {
 523                 bch2_journal_meta_async(&c->journal, NULL);
 524
 525                 return size;
 526         }
 527
 528         if (attr == &sysfs_trigger_btree_coalesce)
 529                 bch2_coalesce(c);
 530
 531         /* Debugging: */
 532
 533         if (attr == &sysfs_trigger_gc)
 534                 bch2_gc(c);
 535
 536         if (attr == &sysfs_prune_cache) {
 537                 struct shrink_control sc;
 538
 539                 sc.gfp_mask = GFP_KERNEL;
 540                 sc.nr_to_scan = strtoul_or_return(buf);
 541                 c->btree_cache_shrink.scan_objects(&c->btree_cache_shrink, &sc);
 542         }
 543
 544         return size;
 545 }
 546
 547 STORE(bch2_fs)
 548 {
 549         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 550
 551         mutex_lock(&c->state_lock);
 552         size = __bch2_fs_store(kobj, attr, buf, size);
 553         mutex_unlock(&c->state_lock);
 554
 555         return size;
 556 }
 557 SYSFS_OPS(bch2_fs);
 558
 559 struct attribute *bch2_fs_files[] = {
 560         &sysfs_journal_write_delay_ms,
 561         &sysfs_journal_reclaim_delay_ms,
 562
 563         &sysfs_block_size,
 564         &sysfs_block_size_bytes,
 565         &sysfs_btree_node_size,
 566         &sysfs_btree_node_size_bytes,
 567         &sysfs_tree_depth,
 568         &sysfs_root_usage_percent,
 569         &sysfs_btree_cache_size,
 570         &sysfs_cache_available_percent,
 571         &sysfs_compression_stats,
 572
 573         &sysfs_average_key_size,
 574
 575         &sysfs_meta_replicas_have,
 576         &sysfs_data_replicas_have,
 577
 578         &sysfs_foreground_target_percent,
 579         &sysfs_tiering_percent,
 580
 581         &sysfs_journal_flush,
 582         NULL
 583 };
 584
 585 /* internal dir - just a wrapper */
 586
 587 SHOW(bch2_fs_internal)
 588 {
 589         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 590         return bch2_fs_show(&c->kobj, attr, buf);
 591 }
 592
 593 STORE(bch2_fs_internal)
 594 {
 595         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 596         return bch2_fs_store(&c->kobj, attr, buf, size);
 597 }
 598 SYSFS_OPS(bch2_fs_internal);
 599
 600 struct attribute *bch2_fs_internal_files[] = {
 601         &sysfs_journal_debug,
 602
 603         &sysfs_alloc_debug,
 604
 605         &sysfs_btree_gc_running,
 606
 607         &sysfs_btree_nodes,
 608         &sysfs_btree_used_percent,
 609
 610         &sysfs_bset_tree_stats,
 611         &sysfs_cache_read_races,
 612
 613         &sysfs_trigger_btree_coalesce,
 614         &sysfs_trigger_gc,
 615         &sysfs_prune_cache,
 616         &sysfs_foreground_write_ratelimit_enabled,
 617         &sysfs_copy_gc_enabled,
 618         &sysfs_tiering_enabled,
 619         sysfs_pd_controller_files(tiering),
 620         sysfs_pd_controller_files(foreground_write),
 621         &sysfs_internal_uuid,
 622
 623 #define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
 624         BCH_DEBUG_PARAMS()
 625 #undef BCH_DEBUG_PARAM
 626
 627         NULL
 628 };
 629
 630 /* options */
 631
 632 SHOW(bch2_fs_opts_dir)
 633 {
 634         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 635
 636         return bch2_opt_show(&c->opts, attr->name, buf, PAGE_SIZE);
 637 }
 638
 639 STORE(bch2_fs_opts_dir)
 640 {
 641         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 642         const struct bch_option *opt;
 643         enum bch_opt_id id;
 644         u64 v;
 645
 646         id = bch2_parse_sysfs_opt(attr->name, buf, &v);
 647         if (id < 0)
 648                 return id;
 649
 650         opt = &bch2_opt_table[id];
 651
 652         mutex_lock(&c->sb_lock);
 653
 654         if (id == Opt_compression) {
 655                 int ret = bch2_check_set_has_compressed_data(c, v);
 656                 if (ret) {
 657                         mutex_unlock(&c->sb_lock);
 658                         return ret;
 659                 }
 660         }
 661
 662         if (opt->set_sb != SET_NO_SB_OPT) {
 663                 opt->set_sb(c->disk_sb, v);
 664                 bch2_write_super(c);
 665         }
 666
 667         bch2_opt_set(&c->opts, id, v);
 668
 669         mutex_unlock(&c->sb_lock);
 670
 671         return size;
 672 }
 673 SYSFS_OPS(bch2_fs_opts_dir);
 674
 675 struct attribute *bch2_fs_opts_dir_files[] = {
 676 #define BCH_OPT(_name, ...)                                             \
 677         &sysfs_opt_##_name,
 678
 679         BCH_VISIBLE_OPTS()
 680 #undef BCH_OPT
 681
 682         NULL
 683 };
 684
 685 /* time stats */
 686
 687 SHOW(bch2_fs_time_stats)
 688 {
 689         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 690
 691 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 692         sysfs_print_time_stats(&c->name##_time, name,                   \
 693                                frequency_units, duration_units);
 694         BCH_TIME_STATS()
 695 #undef BCH_TIME_STAT
 696
 697         return 0;
 698 }
 699
 700 STORE(bch2_fs_time_stats)
 701 {
 702         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 703
 704 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 705         sysfs_clear_time_stats(&c->name##_time, name);
 706         BCH_TIME_STATS()
 707 #undef BCH_TIME_STAT
 708
 709         return size;
 710 }
 711 SYSFS_OPS(bch2_fs_time_stats);
 712
 713 struct attribute *bch2_fs_time_stats_files[] = {
 714 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 715         sysfs_time_stats_attribute_list(name, frequency_units, duration_units)
 716         BCH_TIME_STATS()
 717 #undef BCH_TIME_STAT
 718
 719         NULL
 720 };
 721
 722 typedef unsigned (bucket_map_fn)(struct bch_dev *, struct bucket *, void *);
 723
 724 static unsigned bucket_priority_fn(struct bch_dev *ca, struct bucket *g,
 725                                    void *private)
 726 {
 727         int rw = (private ? 1 : 0);
 728
 729         return ca->fs->prio_clock[rw].hand - g->prio[rw];
 730 }
 731
 732 static unsigned bucket_sectors_used_fn(struct bch_dev *ca, struct bucket *g,
 733                                        void *private)
 734 {
 735         return bucket_sectors_used(g);
 736 }
 737
 738 static unsigned bucket_oldest_gen_fn(struct bch_dev *ca, struct bucket *g,
 739                                      void *private)
 740 {
 741         return bucket_gc_gen(ca, g);
 742 }
 743
 744 static ssize_t show_quantiles(struct bch_dev *ca, char *buf,
 745                               bucket_map_fn *fn, void *private)
 746 {
 747         int cmp(const void *l, const void *r)
 748         {       return *((unsigned *) r) - *((unsigned *) l); }
 749
 750         size_t n = ca->mi.nbuckets, i;
 751         /* Compute 31 quantiles */
 752         unsigned q[31], *p;
 753         ssize_t ret = 0;
 754
 755         p = vzalloc(ca->mi.nbuckets * sizeof(unsigned));
 756         if (!p)
 757                 return -ENOMEM;
 758
 759         for (i = ca->mi.first_bucket; i < n; i++)
 760                 p[i] = fn(ca, &ca->buckets[i], private);
 761
 762         sort(p, n, sizeof(unsigned), cmp, NULL);
 763
 764         while (n &&
 765                !p[n - 1])
 766                 --n;
 767
 768         for (i = 0; i < ARRAY_SIZE(q); i++)
 769                 q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
 770
 771         vfree(p);
 772
 773         for (i = 0; i < ARRAY_SIZE(q); i++)
 774                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 775                                  "%u ", q[i]);
 776         buf[ret - 1] = '\n';
 777
 778         return ret;
 779
 780 }
 781
 782 static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
 783 {
 784         enum alloc_reserve i;
 785         ssize_t ret;
 786
 787         spin_lock(&ca->freelist_lock);
 788
 789         ret = scnprintf(buf, PAGE_SIZE,
 790                         "free_inc:\t%zu\t%zu\n",
 791                         fifo_used(&ca->free_inc),
 792                         ca->free_inc.size);
 793
 794         for (i = 0; i < RESERVE_NR; i++)
 795                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 796                                  "free[%u]:\t%zu\t%zu\n", i,
 797                                  fifo_used(&ca->free[i]),
 798                                  ca->free[i].size);
 799
 800         spin_unlock(&ca->freelist_lock);
 801
 802         return ret;
 803 }
 804
 805 static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
 806 {
 807         struct bch_fs *c = ca->fs;
 808         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 809
 810         return scnprintf(buf, PAGE_SIZE,
 811                 "free_inc:               %zu/%zu\n"
 812                 "free[RESERVE_PRIO]:     %zu/%zu\n"
 813                 "free[RESERVE_BTREE]:    %zu/%zu\n"
 814                 "free[RESERVE_MOVINGGC]: %zu/%zu\n"
 815                 "free[RESERVE_NONE]:     %zu/%zu\n"
 816                 "alloc:                  %llu/%llu\n"
 817                 "meta:                   %llu/%llu\n"
 818                 "dirty:                  %llu/%llu\n"
 819                 "available:              %llu/%llu\n"
 820                 "freelist_wait:          %s\n"
 821                 "open buckets:           %u/%u (reserved %u)\n"
 822                 "open_buckets_wait:      %s\n",
 823                 fifo_used(&ca->free_inc),               ca->free_inc.size,
 824                 fifo_used(&ca->free[RESERVE_PRIO]),     ca->free[RESERVE_PRIO].size,
 825                 fifo_used(&ca->free[RESERVE_BTREE]),    ca->free[RESERVE_BTREE].size,
 826                 fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size,
 827                 fifo_used(&ca->free[RESERVE_NONE]),     ca->free[RESERVE_NONE].size,
 828                 stats.buckets_alloc,                    ca->mi.nbuckets - ca->mi.first_bucket,
 829                 stats.buckets_meta,                     ca->mi.nbuckets - ca->mi.first_bucket,
 830                 stats.buckets_dirty,                    ca->mi.nbuckets - ca->mi.first_bucket,
 831                 __dev_buckets_available(ca, stats),     ca->mi.nbuckets - ca->mi.first_bucket,
 832                 c->freelist_wait.list.first             ? "waiting" : "empty",
 833                 c->open_buckets_nr_free, OPEN_BUCKETS_COUNT, BTREE_NODE_RESERVE,
 834                 c->open_buckets_wait.list.first         ? "waiting" : "empty");
 835 }
 836
 837 static u64 sectors_written(struct bch_dev *ca)
 838 {
 839         u64 ret = 0;
 840         int cpu;
 841
 842         for_each_possible_cpu(cpu)
 843                 ret += *per_cpu_ptr(ca->sectors_written, cpu);
 844
 845         return ret;
 846 }
 847
 848 SHOW(bch2_dev)
 849 {
 850         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 851         struct bch_fs *c = ca->fs;
 852         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 853
 854         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 855
 856         sysfs_hprint(bucket_size,       bucket_bytes(ca));
 857         sysfs_print(bucket_size_bytes,  bucket_bytes(ca));
 858         sysfs_hprint(block_size,        block_bytes(c));
 859         sysfs_print(block_size_bytes,   block_bytes(c));
 860         sysfs_print(first_bucket,       ca->mi.first_bucket);
 861         sysfs_print(nbuckets,           ca->mi.nbuckets);
 862         sysfs_print(discard,            ca->mi.discard);
 863         sysfs_hprint(written, sectors_written(ca) << 9);
 864         sysfs_hprint(btree_written,
 865                      atomic64_read(&ca->btree_sectors_written) << 9);
 866         sysfs_hprint(metadata_written,
 867                      (atomic64_read(&ca->meta_sectors_written) +
 868                       atomic64_read(&ca->btree_sectors_written)) << 9);
 869
 870         sysfs_hprint(dirty_data,        stats.sectors[S_DIRTY] << 9);
 871         sysfs_print(dirty_bytes,        stats.sectors[S_DIRTY] << 9);
 872         sysfs_print(dirty_buckets,      stats.buckets_dirty);
 873         sysfs_hprint(cached_data,       stats.sectors[S_CACHED] << 9);
 874         sysfs_print(cached_bytes,       stats.sectors[S_CACHED] << 9);
 875         sysfs_print(cached_buckets,     stats.buckets_cached);
 876         sysfs_print(meta_buckets,       stats.buckets_meta);
 877         sysfs_print(alloc_buckets,      stats.buckets_alloc);
 878         sysfs_print(available_buckets,  dev_buckets_available(ca));
 879         sysfs_print(free_buckets,       dev_buckets_free(ca));
 880         sysfs_print(has_data,           ca->mi.has_data);
 881         sysfs_print(has_metadata,       ca->mi.has_metadata);
 882
 883         sysfs_pd_controller_show(copy_gc, &ca->moving_gc_pd);
 884
 885         if (attr == &sysfs_cache_replacement_policy)
 886                 return bch2_snprint_string_list(buf, PAGE_SIZE,
 887                                                 bch2_cache_replacement_policies,
 888                                                 ca->mi.replacement);
 889
 890         sysfs_print(tier,               ca->mi.tier);
 891
 892         if (attr == &sysfs_state_rw)
 893                 return bch2_snprint_string_list(buf, PAGE_SIZE,
 894                                                 bch2_dev_state,
 895                                                 ca->mi.state);
 896
 897         if (attr == &sysfs_read_priority_stats)
 898                 return show_quantiles(ca, buf, bucket_priority_fn, (void *) 0);
 899         if (attr == &sysfs_write_priority_stats)
 900                 return show_quantiles(ca, buf, bucket_priority_fn, (void *) 1);
 901         if (attr == &sysfs_fragmentation_stats)
 902                 return show_quantiles(ca, buf, bucket_sectors_used_fn, NULL);
 903         if (attr == &sysfs_oldest_gen_stats)
 904                 return show_quantiles(ca, buf, bucket_oldest_gen_fn, NULL);
 905         if (attr == &sysfs_reserve_stats)
 906                 return show_reserve_stats(ca, buf);
 907         if (attr == &sysfs_alloc_debug)
 908                 return show_dev_alloc_debug(ca, buf);
 909
 910         return 0;
 911 }
 912
 913 STORE(bch2_dev)
 914 {
 915         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 916         struct bch_fs *c = ca->fs;
 917         struct bch_member *mi;
 918
 919         sysfs_pd_controller_store(copy_gc, &ca->moving_gc_pd);
 920
 921         if (attr == &sysfs_discard) {
 922                 bool v = strtoul_or_return(buf);
 923
 924                 mutex_lock(&c->sb_lock);
 925                 mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
 926
 927                 if (v != BCH_MEMBER_DISCARD(mi)) {
 928                         SET_BCH_MEMBER_DISCARD(mi, v);
 929                         bch2_write_super(c);
 930                 }
 931                 mutex_unlock(&c->sb_lock);
 932         }
 933
 934         if (attr == &sysfs_cache_replacement_policy) {
 935                 ssize_t v = bch2_read_string_list(buf, bch2_cache_replacement_policies);
 936
 937                 if (v < 0)
 938                         return v;
 939
 940                 mutex_lock(&c->sb_lock);
 941                 mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
 942
 943                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
 944                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
 945                         bch2_write_super(c);
 946                 }
 947                 mutex_unlock(&c->sb_lock);
 948         }
 949
 950         if (attr == &sysfs_tier) {
 951                 unsigned prev_tier;
 952                 unsigned v = strtoul_restrict_or_return(buf,
 953                                         0, BCH_TIER_MAX - 1);
 954
 955                 mutex_lock(&c->sb_lock);
 956                 prev_tier = ca->mi.tier;
 957
 958                 if (v == ca->mi.tier) {
 959                         mutex_unlock(&c->sb_lock);
 960                         return size;
 961                 }
 962
 963                 mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
 964                 SET_BCH_MEMBER_TIER(mi, v);
 965                 bch2_write_super(c);
 966
 967                 bch2_dev_group_remove(&c->tiers[prev_tier].devs, ca);
 968                 bch2_dev_group_add(&c->tiers[ca->mi.tier].devs, ca);
 969                 mutex_unlock(&c->sb_lock);
 970
 971                 bch2_recalc_capacity(c);
 972                 bch2_tiering_start(c);
 973         }
 974
 975         return size;
 976 }
 977 SYSFS_OPS(bch2_dev);
 978
 979 struct attribute *bch2_dev_files[] = {
 980         &sysfs_uuid,
 981         &sysfs_bucket_size,
 982         &sysfs_bucket_size_bytes,
 983         &sysfs_block_size,
 984         &sysfs_block_size_bytes,
 985         &sysfs_first_bucket,
 986         &sysfs_nbuckets,
 987         &sysfs_read_priority_stats,
 988         &sysfs_write_priority_stats,
 989         &sysfs_fragmentation_stats,
 990         &sysfs_oldest_gen_stats,
 991         &sysfs_reserve_stats,
 992         &sysfs_available_buckets,
 993         &sysfs_free_buckets,
 994         &sysfs_dirty_data,
 995         &sysfs_dirty_bytes,
 996         &sysfs_dirty_buckets,
 997         &sysfs_cached_data,
 998         &sysfs_cached_bytes,
 999         &sysfs_cached_buckets,
1000         &sysfs_meta_buckets,
1001         &sysfs_alloc_buckets,
1002         &sysfs_has_data,
1003         &sysfs_has_metadata,
1004         &sysfs_discard,
1005         &sysfs_written,
1006         &sysfs_btree_written,
1007         &sysfs_metadata_written,
1008         &sysfs_cache_replacement_policy,
1009         &sysfs_tier,
1010         &sysfs_state_rw,
1011         &sysfs_alloc_debug,
1012
1013         sysfs_pd_controller_files(copy_gc),
1014         NULL
1015 };