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 read_attribute(journal_entry_size_max);
 179
 180 rw_attribute(discard);
 181 rw_attribute(cache_replacement_policy);
 182
 183 rw_attribute(foreground_write_ratelimit_enabled);
 184 rw_attribute(copy_gc_enabled);
 185 sysfs_pd_controller_attribute(copy_gc);
 186
 187 rw_attribute(tier);
 188 rw_attribute(tiering_enabled);
 189 rw_attribute(tiering_percent);
 190 sysfs_pd_controller_attribute(tiering);
 191
 192 sysfs_pd_controller_attribute(foreground_write);
 193
 194 rw_attribute(pd_controllers_update_seconds);
 195
 196 rw_attribute(foreground_target_percent);
 197
 198 read_attribute(meta_replicas_have);
 199 read_attribute(data_replicas_have);
 200
 201 #define BCH_DEBUG_PARAM(name, description)                              \
 202         rw_attribute(name);
 203
 204         BCH_DEBUG_PARAMS()
 205 #undef BCH_DEBUG_PARAM
 206
 207 #define BCH_OPT(_name, _mode, ...)                                      \
 208         static struct attribute sysfs_opt_##_name = {                   \
 209                 .name = #_name, .mode = _mode,                          \
 210         };
 211
 212         BCH_VISIBLE_OPTS()
 213 #undef BCH_OPT
 214
 215 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 216         sysfs_time_stats_attribute(name, frequency_units, duration_units);
 217         BCH_TIME_STATS()
 218 #undef BCH_TIME_STAT
 219
 220 static struct attribute sysfs_state_rw = {
 221         .name = "state",
 222         .mode = S_IRUGO
 223 };
 224
 225 static int bch2_bset_print_stats(struct bch_fs *c, char *buf)
 226 {
 227         struct bset_stats stats;
 228         size_t nodes = 0;
 229         struct btree *b;
 230         struct bucket_table *tbl;
 231         struct rhash_head *pos;
 232         unsigned iter;
 233
 234         memset(&stats, 0, sizeof(stats));
 235
 236         rcu_read_lock();
 237         for_each_cached_btree(b, c, tbl, iter, pos) {
 238                 bch2_btree_keys_stats(b, &stats);
 239                 nodes++;
 240         }
 241         rcu_read_unlock();
 242
 243         return snprintf(buf, PAGE_SIZE,
 244                         "btree nodes:           %zu\n"
 245                         "written sets:          %zu\n"
 246                         "written key bytes:     %zu\n"
 247                         "unwritten sets:                %zu\n"
 248                         "unwritten key bytes:   %zu\n"
 249                         "no table sets:         %zu\n"
 250                         "no table key bytes:    %zu\n"
 251                         "floats:                        %zu\n"
 252                         "failed unpacked:       %zu\n"
 253                         "failed prev:           %zu\n"
 254                         "failed overflow:       %zu\n",
 255                         nodes,
 256                         stats.sets[BSET_RO_AUX_TREE].nr,
 257                         stats.sets[BSET_RO_AUX_TREE].bytes,
 258                         stats.sets[BSET_RW_AUX_TREE].nr,
 259                         stats.sets[BSET_RW_AUX_TREE].bytes,
 260                         stats.sets[BSET_NO_AUX_TREE].nr,
 261                         stats.sets[BSET_NO_AUX_TREE].bytes,
 262                         stats.floats,
 263                         stats.failed_unpacked,
 264                         stats.failed_prev,
 265                         stats.failed_overflow);
 266 }
 267
 268 static unsigned bch2_root_usage(struct bch_fs *c)
 269 {
 270         unsigned bytes = 0;
 271         struct bkey_packed *k;
 272         struct btree *b;
 273         struct btree_node_iter iter;
 274
 275         goto lock_root;
 276
 277         do {
 278                 six_unlock_read(&b->lock);
 279 lock_root:
 280                 b = c->btree_roots[BTREE_ID_EXTENTS].b;
 281                 six_lock_read(&b->lock);
 282         } while (b != c->btree_roots[BTREE_ID_EXTENTS].b);
 283
 284         for_each_btree_node_key(b, k, &iter, btree_node_is_extents(b))
 285                 bytes += bkey_bytes(k);
 286
 287         six_unlock_read(&b->lock);
 288
 289         return (bytes * 100) / btree_bytes(c);
 290 }
 291
 292 static size_t bch2_btree_cache_size(struct bch_fs *c)
 293 {
 294         size_t ret = 0;
 295         struct btree *b;
 296
 297         mutex_lock(&c->btree_cache_lock);
 298         list_for_each_entry(b, &c->btree_cache, list)
 299                 ret += btree_bytes(c);
 300
 301         mutex_unlock(&c->btree_cache_lock);
 302         return ret;
 303 }
 304
 305 static unsigned bch2_fs_available_percent(struct bch_fs *c)
 306 {
 307         return div64_u64((u64) sectors_available(c) * 100,
 308                          c->capacity ?: 1);
 309 }
 310
 311 #if 0
 312 static unsigned bch2_btree_used(struct bch_fs *c)
 313 {
 314         return div64_u64(c->gc_stats.key_bytes * 100,
 315                          (c->gc_stats.nodes ?: 1) * btree_bytes(c));
 316 }
 317
 318 static unsigned bch2_average_key_size(struct bch_fs *c)
 319 {
 320         return c->gc_stats.nkeys
 321                 ? div64_u64(c->gc_stats.data, c->gc_stats.nkeys)
 322                 : 0;
 323 }
 324 #endif
 325
 326 static ssize_t show_fs_alloc_debug(struct bch_fs *c, char *buf)
 327 {
 328         struct bch_fs_usage stats = bch2_fs_usage_read(c);
 329
 330         return scnprintf(buf, PAGE_SIZE,
 331                          "capacity:\t\t%llu\n"
 332                          "compressed:\n"
 333                          "\tmeta:\t\t%llu\n"
 334                          "\tdirty:\t\t%llu\n"
 335                          "\tcached:\t\t%llu\n"
 336                          "uncompressed:\n"
 337                          "\tmeta:\t\t%llu\n"
 338                          "\tdirty:\t\t%llu\n"
 339                          "\tcached:\t\t%llu\n"
 340                          "persistent reserved sectors:\t%llu\n"
 341                          "online reserved sectors:\t%llu\n",
 342                          c->capacity,
 343                          stats.s[S_COMPRESSED][S_META],
 344                          stats.s[S_COMPRESSED][S_DIRTY],
 345                          stats.s[S_COMPRESSED][S_CACHED],
 346                          stats.s[S_UNCOMPRESSED][S_META],
 347                          stats.s[S_UNCOMPRESSED][S_DIRTY],
 348                          stats.s[S_UNCOMPRESSED][S_CACHED],
 349                          stats.persistent_reserved,
 350                          stats.online_reserved);
 351 }
 352
 353 static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 354 {
 355         struct btree_iter iter;
 356         struct bkey_s_c k;
 357         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 358             nr_compressed_extents = 0,
 359             compressed_sectors_compressed = 0,
 360             compressed_sectors_uncompressed = 0;
 361
 362         for_each_btree_key(&iter, c, BTREE_ID_EXTENTS, POS_MIN, k)
 363                 if (k.k->type == BCH_EXTENT) {
 364                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 365                         const struct bch_extent_ptr *ptr;
 366                         const union bch_extent_crc *crc;
 367
 368                         extent_for_each_ptr_crc(e, ptr, crc) {
 369                                 if (crc_compression_type(crc) == BCH_COMPRESSION_NONE) {
 370                                         nr_uncompressed_extents++;
 371                                         uncompressed_sectors += e.k->size;
 372                                 } else {
 373                                         nr_compressed_extents++;
 374                                         compressed_sectors_compressed +=
 375                                                 crc_compressed_size(e.k, crc);
 376                                         compressed_sectors_uncompressed +=
 377                                                 crc_uncompressed_size(e.k, crc);
 378                                 }
 379
 380                                 /* only looking at the first ptr */
 381                                 break;
 382                         }
 383                 }
 384         bch2_btree_iter_unlock(&iter);
 385
 386         return snprintf(buf, PAGE_SIZE,
 387                         "uncompressed data:\n"
 388                         "       nr extents:                     %llu\n"
 389                         "       size (bytes):                   %llu\n"
 390                         "compressed data:\n"
 391                         "       nr extents:                     %llu\n"
 392                         "       compressed size (bytes):        %llu\n"
 393                         "       uncompressed size (bytes):      %llu\n",
 394                         nr_uncompressed_extents,
 395                         uncompressed_sectors << 9,
 396                         nr_compressed_extents,
 397                         compressed_sectors_compressed << 9,
 398                         compressed_sectors_uncompressed << 9);
 399 }
 400
 401 SHOW(bch2_fs)
 402 {
 403         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 404
 405         sysfs_print(minor,                      c->minor);
 406
 407         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 408         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 409         sysfs_hprint(journal_entry_size_max,    c->journal.entry_size_max);
 410
 411         sysfs_hprint(block_size,                block_bytes(c));
 412         sysfs_print(block_size_bytes,           block_bytes(c));
 413         sysfs_hprint(btree_node_size,           c->sb.btree_node_size << 9);
 414         sysfs_print(btree_node_size_bytes,      c->sb.btree_node_size << 9);
 415
 416         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 417         sysfs_print(cache_available_percent,    bch2_fs_available_percent(c));
 418
 419         sysfs_print(btree_gc_running,           c->gc_pos.phase != GC_PHASE_DONE);
 420
 421 #if 0
 422         /* XXX: reimplement */
 423         sysfs_print(btree_used_percent, bch2_btree_used(c));
 424         sysfs_print(btree_nodes,        c->gc_stats.nodes);
 425         sysfs_hprint(average_key_size,  bch2_average_key_size(c));
 426 #endif
 427
 428         sysfs_print(cache_read_races,
 429                     atomic_long_read(&c->cache_read_races));
 430
 431         sysfs_printf(foreground_write_ratelimit_enabled, "%i",
 432                      c->foreground_write_ratelimit_enabled);
 433         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 434         sysfs_pd_controller_show(foreground_write, &c->foreground_write_pd);
 435
 436         sysfs_print(pd_controllers_update_seconds,
 437                     c->pd_controllers_update_seconds);
 438         sysfs_print(foreground_target_percent, c->foreground_target_percent);
 439
 440         sysfs_printf(tiering_enabled,           "%i", c->tiering_enabled);
 441         sysfs_print(tiering_percent,            c->tiering_percent);
 442
 443         sysfs_pd_controller_show(tiering,       &c->tiers[1].pd); /* XXX */
 444
 445         sysfs_printf(meta_replicas_have, "%u",  c->sb.meta_replicas_have);
 446         sysfs_printf(data_replicas_have, "%u",  c->sb.data_replicas_have);
 447
 448         /* Debugging: */
 449
 450         if (attr == &sysfs_journal_debug)
 451                 return bch2_journal_print_debug(&c->journal, buf);
 452
 453 #define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
 454         BCH_DEBUG_PARAMS()
 455 #undef BCH_DEBUG_PARAM
 456
 457         if (!bch2_fs_running(c))
 458                 return -EPERM;
 459
 460         if (attr == &sysfs_bset_tree_stats)
 461                 return bch2_bset_print_stats(c, buf);
 462         if (attr == &sysfs_alloc_debug)
 463                 return show_fs_alloc_debug(c, buf);
 464
 465         sysfs_print(tree_depth, c->btree_roots[BTREE_ID_EXTENTS].b->level);
 466         sysfs_print(root_usage_percent,         bch2_root_usage(c));
 467
 468         if (attr == &sysfs_compression_stats)
 469                 return bch2_compression_stats(c, buf);
 470
 471         sysfs_printf(internal_uuid, "%pU", c->sb.uuid.b);
 472
 473         return 0;
 474 }
 475
 476 STORE(__bch2_fs)
 477 {
 478         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 479
 480         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 481         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 482
 483         sysfs_strtoul(foreground_write_ratelimit_enabled,
 484                       c->foreground_write_ratelimit_enabled);
 485
 486         if (attr == &sysfs_copy_gc_enabled) {
 487                 struct bch_dev *ca;
 488                 unsigned i;
 489                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 490                         ?: (ssize_t) size;
 491
 492                 for_each_member_device(ca, c, i)
 493                         if (ca->moving_gc_read)
 494                                 wake_up_process(ca->moving_gc_read);
 495                 return ret;
 496         }
 497
 498         if (attr == &sysfs_tiering_enabled) {
 499                 ssize_t ret = strtoul_safe(buf, c->tiering_enabled)
 500                         ?: (ssize_t) size;
 501
 502                 bch2_tiering_start(c); /* issue wakeups */
 503                 return ret;
 504         }
 505
 506         sysfs_pd_controller_store(foreground_write, &c->foreground_write_pd);
 507
 508         sysfs_strtoul(pd_controllers_update_seconds,
 509                       c->pd_controllers_update_seconds);
 510         sysfs_strtoul(foreground_target_percent, c->foreground_target_percent);
 511
 512         sysfs_strtoul(tiering_percent,          c->tiering_percent);
 513         sysfs_pd_controller_store(tiering,      &c->tiers[1].pd); /* XXX */
 514
 515         /* Debugging: */
 516
 517 #define BCH_DEBUG_PARAM(name, description) sysfs_strtoul(name, c->name);
 518         BCH_DEBUG_PARAMS()
 519 #undef BCH_DEBUG_PARAM
 520
 521         if (!bch2_fs_running(c))
 522                 return -EPERM;
 523
 524         if (attr == &sysfs_journal_flush) {
 525                 bch2_journal_meta_async(&c->journal, NULL);
 526
 527                 return size;
 528         }
 529
 530         if (attr == &sysfs_trigger_btree_coalesce)
 531                 bch2_coalesce(c);
 532
 533         /* Debugging: */
 534
 535         if (attr == &sysfs_trigger_gc)
 536                 bch2_gc(c);
 537
 538         if (attr == &sysfs_prune_cache) {
 539                 struct shrink_control sc;
 540
 541                 sc.gfp_mask = GFP_KERNEL;
 542                 sc.nr_to_scan = strtoul_or_return(buf);
 543                 c->btree_cache_shrink.scan_objects(&c->btree_cache_shrink, &sc);
 544         }
 545
 546         return size;
 547 }
 548
 549 STORE(bch2_fs)
 550 {
 551         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 552
 553         mutex_lock(&c->state_lock);
 554         size = __bch2_fs_store(kobj, attr, buf, size);
 555         mutex_unlock(&c->state_lock);
 556
 557         return size;
 558 }
 559 SYSFS_OPS(bch2_fs);
 560
 561 struct attribute *bch2_fs_files[] = {
 562         &sysfs_journal_write_delay_ms,
 563         &sysfs_journal_reclaim_delay_ms,
 564         &sysfs_journal_entry_size_max,
 565
 566         &sysfs_block_size,
 567         &sysfs_block_size_bytes,
 568         &sysfs_btree_node_size,
 569         &sysfs_btree_node_size_bytes,
 570         &sysfs_tree_depth,
 571         &sysfs_root_usage_percent,
 572         &sysfs_btree_cache_size,
 573         &sysfs_cache_available_percent,
 574         &sysfs_compression_stats,
 575
 576         &sysfs_average_key_size,
 577
 578         &sysfs_meta_replicas_have,
 579         &sysfs_data_replicas_have,
 580
 581         &sysfs_foreground_target_percent,
 582         &sysfs_tiering_percent,
 583
 584         &sysfs_journal_flush,
 585         NULL
 586 };
 587
 588 /* internal dir - just a wrapper */
 589
 590 SHOW(bch2_fs_internal)
 591 {
 592         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 593         return bch2_fs_show(&c->kobj, attr, buf);
 594 }
 595
 596 STORE(bch2_fs_internal)
 597 {
 598         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 599         return bch2_fs_store(&c->kobj, attr, buf, size);
 600 }
 601 SYSFS_OPS(bch2_fs_internal);
 602
 603 struct attribute *bch2_fs_internal_files[] = {
 604         &sysfs_journal_debug,
 605
 606         &sysfs_alloc_debug,
 607
 608         &sysfs_btree_gc_running,
 609
 610         &sysfs_btree_nodes,
 611         &sysfs_btree_used_percent,
 612
 613         &sysfs_bset_tree_stats,
 614         &sysfs_cache_read_races,
 615
 616         &sysfs_trigger_btree_coalesce,
 617         &sysfs_trigger_gc,
 618         &sysfs_prune_cache,
 619         &sysfs_foreground_write_ratelimit_enabled,
 620         &sysfs_copy_gc_enabled,
 621         &sysfs_tiering_enabled,
 622         sysfs_pd_controller_files(tiering),
 623         sysfs_pd_controller_files(foreground_write),
 624         &sysfs_internal_uuid,
 625
 626 #define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
 627         BCH_DEBUG_PARAMS()
 628 #undef BCH_DEBUG_PARAM
 629
 630         NULL
 631 };
 632
 633 /* options */
 634
 635 SHOW(bch2_fs_opts_dir)
 636 {
 637         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 638
 639         return bch2_opt_show(&c->opts, attr->name, buf, PAGE_SIZE);
 640 }
 641
 642 STORE(bch2_fs_opts_dir)
 643 {
 644         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 645         const struct bch_option *opt;
 646         enum bch_opt_id id;
 647         u64 v;
 648
 649         id = bch2_parse_sysfs_opt(attr->name, buf, &v);
 650         if (id < 0)
 651                 return id;
 652
 653         opt = &bch2_opt_table[id];
 654
 655         mutex_lock(&c->sb_lock);
 656
 657         if (id == Opt_compression) {
 658                 int ret = bch2_check_set_has_compressed_data(c, v);
 659                 if (ret) {
 660                         mutex_unlock(&c->sb_lock);
 661                         return ret;
 662                 }
 663         }
 664
 665         if (opt->set_sb != SET_NO_SB_OPT) {
 666                 opt->set_sb(c->disk_sb, v);
 667                 bch2_write_super(c);
 668         }
 669
 670         bch2_opt_set(&c->opts, id, v);
 671
 672         mutex_unlock(&c->sb_lock);
 673
 674         return size;
 675 }
 676 SYSFS_OPS(bch2_fs_opts_dir);
 677
 678 struct attribute *bch2_fs_opts_dir_files[] = {
 679 #define BCH_OPT(_name, ...)                                             \
 680         &sysfs_opt_##_name,
 681
 682         BCH_VISIBLE_OPTS()
 683 #undef BCH_OPT
 684
 685         NULL
 686 };
 687
 688 /* time stats */
 689
 690 SHOW(bch2_fs_time_stats)
 691 {
 692         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 693
 694 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 695         sysfs_print_time_stats(&c->name##_time, name,                   \
 696                                frequency_units, duration_units);
 697         BCH_TIME_STATS()
 698 #undef BCH_TIME_STAT
 699
 700         return 0;
 701 }
 702
 703 STORE(bch2_fs_time_stats)
 704 {
 705         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 706
 707 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 708         sysfs_clear_time_stats(&c->name##_time, name);
 709         BCH_TIME_STATS()
 710 #undef BCH_TIME_STAT
 711
 712         return size;
 713 }
 714 SYSFS_OPS(bch2_fs_time_stats);
 715
 716 struct attribute *bch2_fs_time_stats_files[] = {
 717 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 718         sysfs_time_stats_attribute_list(name, frequency_units, duration_units)
 719         BCH_TIME_STATS()
 720 #undef BCH_TIME_STAT
 721
 722         NULL
 723 };
 724
 725 typedef unsigned (bucket_map_fn)(struct bch_dev *, struct bucket *, void *);
 726
 727 static unsigned bucket_priority_fn(struct bch_dev *ca, struct bucket *g,
 728                                    void *private)
 729 {
 730         int rw = (private ? 1 : 0);
 731
 732         return ca->fs->prio_clock[rw].hand - g->prio[rw];
 733 }
 734
 735 static unsigned bucket_sectors_used_fn(struct bch_dev *ca, struct bucket *g,
 736                                        void *private)
 737 {
 738         return bucket_sectors_used(g);
 739 }
 740
 741 static unsigned bucket_oldest_gen_fn(struct bch_dev *ca, struct bucket *g,
 742                                      void *private)
 743 {
 744         return bucket_gc_gen(ca, g);
 745 }
 746
 747 static ssize_t show_quantiles(struct bch_dev *ca, char *buf,
 748                               bucket_map_fn *fn, void *private)
 749 {
 750         int cmp(const void *l, const void *r)
 751         {       return *((unsigned *) r) - *((unsigned *) l); }
 752
 753         size_t n = ca->mi.nbuckets, i;
 754         /* Compute 31 quantiles */
 755         unsigned q[31], *p;
 756         ssize_t ret = 0;
 757
 758         p = vzalloc(ca->mi.nbuckets * sizeof(unsigned));
 759         if (!p)
 760                 return -ENOMEM;
 761
 762         for (i = ca->mi.first_bucket; i < n; i++)
 763                 p[i] = fn(ca, &ca->buckets[i], private);
 764
 765         sort(p, n, sizeof(unsigned), cmp, NULL);
 766
 767         while (n &&
 768                !p[n - 1])
 769                 --n;
 770
 771         for (i = 0; i < ARRAY_SIZE(q); i++)
 772                 q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
 773
 774         vfree(p);
 775
 776         for (i = 0; i < ARRAY_SIZE(q); i++)
 777                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 778                                  "%u ", q[i]);
 779         buf[ret - 1] = '\n';
 780
 781         return ret;
 782
 783 }
 784
 785 static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
 786 {
 787         enum alloc_reserve i;
 788         ssize_t ret;
 789
 790         spin_lock(&ca->freelist_lock);
 791
 792         ret = scnprintf(buf, PAGE_SIZE,
 793                         "free_inc:\t%zu\t%zu\n",
 794                         fifo_used(&ca->free_inc),
 795                         ca->free_inc.size);
 796
 797         for (i = 0; i < RESERVE_NR; i++)
 798                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 799                                  "free[%u]:\t%zu\t%zu\n", i,
 800                                  fifo_used(&ca->free[i]),
 801                                  ca->free[i].size);
 802
 803         spin_unlock(&ca->freelist_lock);
 804
 805         return ret;
 806 }
 807
 808 static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
 809 {
 810         struct bch_fs *c = ca->fs;
 811         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 812
 813         return scnprintf(buf, PAGE_SIZE,
 814                 "free_inc:               %zu/%zu\n"
 815                 "free[RESERVE_PRIO]:     %zu/%zu\n"
 816                 "free[RESERVE_BTREE]:    %zu/%zu\n"
 817                 "free[RESERVE_MOVINGGC]: %zu/%zu\n"
 818                 "free[RESERVE_NONE]:     %zu/%zu\n"
 819                 "alloc:                  %llu/%llu\n"
 820                 "meta:                   %llu/%llu\n"
 821                 "dirty:                  %llu/%llu\n"
 822                 "available:              %llu/%llu\n"
 823                 "freelist_wait:          %s\n"
 824                 "open buckets:           %u/%u (reserved %u)\n"
 825                 "open_buckets_wait:      %s\n",
 826                 fifo_used(&ca->free_inc),               ca->free_inc.size,
 827                 fifo_used(&ca->free[RESERVE_PRIO]),     ca->free[RESERVE_PRIO].size,
 828                 fifo_used(&ca->free[RESERVE_BTREE]),    ca->free[RESERVE_BTREE].size,
 829                 fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size,
 830                 fifo_used(&ca->free[RESERVE_NONE]),     ca->free[RESERVE_NONE].size,
 831                 stats.buckets_alloc,                    ca->mi.nbuckets - ca->mi.first_bucket,
 832                 stats.buckets_meta,                     ca->mi.nbuckets - ca->mi.first_bucket,
 833                 stats.buckets_dirty,                    ca->mi.nbuckets - ca->mi.first_bucket,
 834                 __dev_buckets_available(ca, stats),     ca->mi.nbuckets - ca->mi.first_bucket,
 835                 c->freelist_wait.list.first             ? "waiting" : "empty",
 836                 c->open_buckets_nr_free, OPEN_BUCKETS_COUNT, BTREE_NODE_RESERVE,
 837                 c->open_buckets_wait.list.first         ? "waiting" : "empty");
 838 }
 839
 840 static u64 sectors_written(struct bch_dev *ca)
 841 {
 842         u64 ret = 0;
 843         int cpu;
 844
 845         for_each_possible_cpu(cpu)
 846                 ret += *per_cpu_ptr(ca->sectors_written, cpu);
 847
 848         return ret;
 849 }
 850
 851 SHOW(bch2_dev)
 852 {
 853         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 854         struct bch_fs *c = ca->fs;
 855         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 856
 857         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 858
 859         sysfs_hprint(bucket_size,       bucket_bytes(ca));
 860         sysfs_print(bucket_size_bytes,  bucket_bytes(ca));
 861         sysfs_hprint(block_size,        block_bytes(c));
 862         sysfs_print(block_size_bytes,   block_bytes(c));
 863         sysfs_print(first_bucket,       ca->mi.first_bucket);
 864         sysfs_print(nbuckets,           ca->mi.nbuckets);
 865         sysfs_print(discard,            ca->mi.discard);
 866         sysfs_hprint(written, sectors_written(ca) << 9);
 867         sysfs_hprint(btree_written,
 868                      atomic64_read(&ca->btree_sectors_written) << 9);
 869         sysfs_hprint(metadata_written,
 870                      (atomic64_read(&ca->meta_sectors_written) +
 871                       atomic64_read(&ca->btree_sectors_written)) << 9);
 872
 873         sysfs_hprint(dirty_data,        stats.sectors[S_DIRTY] << 9);
 874         sysfs_print(dirty_bytes,        stats.sectors[S_DIRTY] << 9);
 875         sysfs_print(dirty_buckets,      stats.buckets_dirty);
 876         sysfs_hprint(cached_data,       stats.sectors[S_CACHED] << 9);
 877         sysfs_print(cached_bytes,       stats.sectors[S_CACHED] << 9);
 878         sysfs_print(cached_buckets,     stats.buckets_cached);
 879         sysfs_print(meta_buckets,       stats.buckets_meta);
 880         sysfs_print(alloc_buckets,      stats.buckets_alloc);
 881         sysfs_print(available_buckets,  dev_buckets_available(ca));
 882         sysfs_print(free_buckets,       dev_buckets_free(ca));
 883         sysfs_print(has_data,           ca->mi.has_data);
 884         sysfs_print(has_metadata,       ca->mi.has_metadata);
 885
 886         sysfs_pd_controller_show(copy_gc, &ca->moving_gc_pd);
 887
 888         if (attr == &sysfs_cache_replacement_policy)
 889                 return bch2_snprint_string_list(buf, PAGE_SIZE,
 890                                                 bch2_cache_replacement_policies,
 891                                                 ca->mi.replacement);
 892
 893         sysfs_print(tier,               ca->mi.tier);
 894
 895         if (attr == &sysfs_state_rw)
 896                 return bch2_snprint_string_list(buf, PAGE_SIZE,
 897                                                 bch2_dev_state,
 898                                                 ca->mi.state);
 899
 900         if (attr == &sysfs_read_priority_stats)
 901                 return show_quantiles(ca, buf, bucket_priority_fn, (void *) 0);
 902         if (attr == &sysfs_write_priority_stats)
 903                 return show_quantiles(ca, buf, bucket_priority_fn, (void *) 1);
 904         if (attr == &sysfs_fragmentation_stats)
 905                 return show_quantiles(ca, buf, bucket_sectors_used_fn, NULL);
 906         if (attr == &sysfs_oldest_gen_stats)
 907                 return show_quantiles(ca, buf, bucket_oldest_gen_fn, NULL);
 908         if (attr == &sysfs_reserve_stats)
 909                 return show_reserve_stats(ca, buf);
 910         if (attr == &sysfs_alloc_debug)
 911                 return show_dev_alloc_debug(ca, buf);
 912
 913         return 0;
 914 }
 915
 916 STORE(bch2_dev)
 917 {
 918         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 919         struct bch_fs *c = ca->fs;
 920         struct bch_member *mi;
 921
 922         sysfs_pd_controller_store(copy_gc, &ca->moving_gc_pd);
 923
 924         if (attr == &sysfs_discard) {
 925                 bool v = strtoul_or_return(buf);
 926
 927                 mutex_lock(&c->sb_lock);
 928                 mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
 929
 930                 if (v != BCH_MEMBER_DISCARD(mi)) {
 931                         SET_BCH_MEMBER_DISCARD(mi, v);
 932                         bch2_write_super(c);
 933                 }
 934                 mutex_unlock(&c->sb_lock);
 935         }
 936
 937         if (attr == &sysfs_cache_replacement_policy) {
 938                 ssize_t v = bch2_read_string_list(buf, bch2_cache_replacement_policies);
 939
 940                 if (v < 0)
 941                         return v;
 942
 943                 mutex_lock(&c->sb_lock);
 944                 mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
 945
 946                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
 947                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
 948                         bch2_write_super(c);
 949                 }
 950                 mutex_unlock(&c->sb_lock);
 951         }
 952
 953         if (attr == &sysfs_tier) {
 954                 unsigned prev_tier;
 955                 unsigned v = strtoul_restrict_or_return(buf,
 956                                         0, BCH_TIER_MAX - 1);
 957
 958                 mutex_lock(&c->sb_lock);
 959                 prev_tier = ca->mi.tier;
 960
 961                 if (v == ca->mi.tier) {
 962                         mutex_unlock(&c->sb_lock);
 963                         return size;
 964                 }
 965
 966                 mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
 967                 SET_BCH_MEMBER_TIER(mi, v);
 968                 bch2_write_super(c);
 969
 970                 bch2_dev_group_remove(&c->tiers[prev_tier].devs, ca);
 971                 bch2_dev_group_add(&c->tiers[ca->mi.tier].devs, ca);
 972                 mutex_unlock(&c->sb_lock);
 973
 974                 bch2_recalc_capacity(c);
 975                 bch2_tiering_start(c);
 976         }
 977
 978         return size;
 979 }
 980 SYSFS_OPS(bch2_dev);
 981
 982 struct attribute *bch2_dev_files[] = {
 983         &sysfs_uuid,
 984         &sysfs_bucket_size,
 985         &sysfs_bucket_size_bytes,
 986         &sysfs_block_size,
 987         &sysfs_block_size_bytes,
 988         &sysfs_first_bucket,
 989         &sysfs_nbuckets,
 990         &sysfs_read_priority_stats,
 991         &sysfs_write_priority_stats,
 992         &sysfs_fragmentation_stats,
 993         &sysfs_oldest_gen_stats,
 994         &sysfs_reserve_stats,
 995         &sysfs_available_buckets,
 996         &sysfs_free_buckets,
 997         &sysfs_dirty_data,
 998         &sysfs_dirty_bytes,
 999         &sysfs_dirty_buckets,
1000         &sysfs_cached_data,
1001         &sysfs_cached_bytes,
1002         &sysfs_cached_buckets,
1003         &sysfs_meta_buckets,
1004         &sysfs_alloc_buckets,
1005         &sysfs_has_data,
1006         &sysfs_has_metadata,
1007         &sysfs_discard,
1008         &sysfs_written,
1009         &sysfs_btree_written,
1010         &sysfs_metadata_written,
1011         &sysfs_cache_replacement_policy,
1012         &sysfs_tier,
1013         &sysfs_state_rw,
1014         &sysfs_alloc_debug,
1015
1016         sysfs_pd_controller_files(copy_gc),
1017         NULL
1018 };