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_journal_flush);
 124 write_attribute(trigger_btree_coalesce);
 125 write_attribute(trigger_gc);
 126 write_attribute(prune_cache);
 127
 128 read_attribute(uuid);
 129 read_attribute(minor);
 130 read_attribute(bucket_size);
 131 read_attribute(block_size);
 132 read_attribute(btree_node_size);
 133 read_attribute(first_bucket);
 134 read_attribute(nbuckets);
 135 read_attribute(read_priority_stats);
 136 read_attribute(write_priority_stats);
 137 read_attribute(fragmentation_stats);
 138 read_attribute(oldest_gen_stats);
 139 read_attribute(reserve_stats);
 140 read_attribute(btree_cache_size);
 141 read_attribute(compression_stats);
 142 read_attribute(written);
 143 read_attribute(btree_written);
 144 read_attribute(metadata_written);
 145 read_attribute(journal_debug);
 146 read_attribute(journal_pins);
 147
 148 read_attribute(internal_uuid);
 149
 150 read_attribute(available_buckets);
 151 read_attribute(free_buckets);
 152 read_attribute(dirty_data);
 153 read_attribute(dirty_bytes);
 154 read_attribute(dirty_buckets);
 155 read_attribute(cached_data);
 156 read_attribute(cached_bytes);
 157 read_attribute(cached_buckets);
 158 read_attribute(meta_buckets);
 159 read_attribute(alloc_buckets);
 160 read_attribute(has_data);
 161 read_attribute(has_metadata);
 162 read_attribute(alloc_debug);
 163
 164 read_attribute(read_realloc_races);
 165
 166 rw_attribute(journal_write_delay_ms);
 167 rw_attribute(journal_reclaim_delay_ms);
 168
 169 rw_attribute(discard);
 170 rw_attribute(cache_replacement_policy);
 171
 172 rw_attribute(foreground_write_ratelimit_enabled);
 173 rw_attribute(copy_gc_enabled);
 174 sysfs_pd_controller_attribute(copy_gc);
 175
 176 rw_attribute(tier);
 177 rw_attribute(tiering_enabled);
 178 rw_attribute(tiering_percent);
 179 sysfs_pd_controller_attribute(tiering);
 180
 181 sysfs_pd_controller_attribute(foreground_write);
 182
 183 rw_attribute(pd_controllers_update_seconds);
 184
 185 rw_attribute(foreground_target_percent);
 186
 187 read_attribute(meta_replicas_have);
 188 read_attribute(data_replicas_have);
 189
 190 #define BCH_DEBUG_PARAM(name, description)                              \
 191         rw_attribute(name);
 192
 193         BCH_DEBUG_PARAMS()
 194 #undef BCH_DEBUG_PARAM
 195
 196 #define BCH_OPT(_name, _mode, ...)                                      \
 197         static struct attribute sysfs_opt_##_name = {                   \
 198                 .name = #_name, .mode = _mode,                          \
 199         };
 200
 201         BCH_VISIBLE_OPTS()
 202 #undef BCH_OPT
 203
 204 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 205         sysfs_time_stats_attribute(name, frequency_units, duration_units);
 206         BCH_TIME_STATS()
 207 #undef BCH_TIME_STAT
 208
 209 static struct attribute sysfs_state_rw = {
 210         .name = "state",
 211         .mode = S_IRUGO
 212 };
 213
 214 static size_t bch2_btree_cache_size(struct bch_fs *c)
 215 {
 216         size_t ret = 0;
 217         struct btree *b;
 218
 219         mutex_lock(&c->btree_cache_lock);
 220         list_for_each_entry(b, &c->btree_cache, list)
 221                 ret += btree_bytes(c);
 222
 223         mutex_unlock(&c->btree_cache_lock);
 224         return ret;
 225 }
 226
 227 static ssize_t show_fs_alloc_debug(struct bch_fs *c, char *buf)
 228 {
 229         struct bch_fs_usage stats = bch2_fs_usage_read(c);
 230
 231         return scnprintf(buf, PAGE_SIZE,
 232                          "capacity:\t\t%llu\n"
 233                          "compressed:\n"
 234                          "\tmeta:\t\t%llu\n"
 235                          "\tdirty:\t\t%llu\n"
 236                          "\tcached:\t\t%llu\n"
 237                          "uncompressed:\n"
 238                          "\tmeta:\t\t%llu\n"
 239                          "\tdirty:\t\t%llu\n"
 240                          "\tcached:\t\t%llu\n"
 241                          "persistent reserved sectors:\t%llu\n"
 242                          "online reserved sectors:\t%llu\n",
 243                          c->capacity,
 244                          stats.s[S_COMPRESSED][S_META],
 245                          stats.s[S_COMPRESSED][S_DIRTY],
 246                          stats.s[S_COMPRESSED][S_CACHED],
 247                          stats.s[S_UNCOMPRESSED][S_META],
 248                          stats.s[S_UNCOMPRESSED][S_DIRTY],
 249                          stats.s[S_UNCOMPRESSED][S_CACHED],
 250                          stats.persistent_reserved,
 251                          stats.online_reserved);
 252 }
 253
 254 static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 255 {
 256         struct btree_iter iter;
 257         struct bkey_s_c k;
 258         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 259             nr_compressed_extents = 0,
 260             compressed_sectors_compressed = 0,
 261             compressed_sectors_uncompressed = 0;
 262
 263         if (!bch2_fs_running(c))
 264                 return -EPERM;
 265
 266         for_each_btree_key(&iter, c, BTREE_ID_EXTENTS, POS_MIN, k)
 267                 if (k.k->type == BCH_EXTENT) {
 268                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 269                         const struct bch_extent_ptr *ptr;
 270                         const union bch_extent_crc *crc;
 271
 272                         extent_for_each_ptr_crc(e, ptr, crc) {
 273                                 if (crc_compression_type(crc) == BCH_COMPRESSION_NONE) {
 274                                         nr_uncompressed_extents++;
 275                                         uncompressed_sectors += e.k->size;
 276                                 } else {
 277                                         nr_compressed_extents++;
 278                                         compressed_sectors_compressed +=
 279                                                 crc_compressed_size(e.k, crc);
 280                                         compressed_sectors_uncompressed +=
 281                                                 crc_uncompressed_size(e.k, crc);
 282                                 }
 283
 284                                 /* only looking at the first ptr */
 285                                 break;
 286                         }
 287                 }
 288         bch2_btree_iter_unlock(&iter);
 289
 290         return snprintf(buf, PAGE_SIZE,
 291                         "uncompressed data:\n"
 292                         "       nr extents:                     %llu\n"
 293                         "       size (bytes):                   %llu\n"
 294                         "compressed data:\n"
 295                         "       nr extents:                     %llu\n"
 296                         "       compressed size (bytes):        %llu\n"
 297                         "       uncompressed size (bytes):      %llu\n",
 298                         nr_uncompressed_extents,
 299                         uncompressed_sectors << 9,
 300                         nr_compressed_extents,
 301                         compressed_sectors_compressed << 9,
 302                         compressed_sectors_uncompressed << 9);
 303 }
 304
 305 SHOW(bch2_fs)
 306 {
 307         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 308
 309         sysfs_print(minor,                      c->minor);
 310         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 311
 312         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 313         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 314
 315         sysfs_print(block_size,                 block_bytes(c));
 316         sysfs_print(btree_node_size,            btree_bytes(c));
 317         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 318
 319         sysfs_print(read_realloc_races,
 320                     atomic_long_read(&c->read_realloc_races));
 321
 322         sysfs_printf(foreground_write_ratelimit_enabled, "%i",
 323                      c->foreground_write_ratelimit_enabled);
 324         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 325         sysfs_pd_controller_show(foreground_write, &c->foreground_write_pd);
 326
 327         sysfs_print(pd_controllers_update_seconds,
 328                     c->pd_controllers_update_seconds);
 329         sysfs_print(foreground_target_percent, c->foreground_target_percent);
 330
 331         sysfs_printf(tiering_enabled,           "%i", c->tiering_enabled);
 332         sysfs_print(tiering_percent,            c->tiering_percent);
 333
 334         sysfs_pd_controller_show(tiering,       &c->tiers[1].pd); /* XXX */
 335
 336         sysfs_printf(meta_replicas_have, "%u",  c->sb.meta_replicas_have);
 337         sysfs_printf(data_replicas_have, "%u",  c->sb.data_replicas_have);
 338
 339         /* Debugging: */
 340
 341         if (attr == &sysfs_alloc_debug)
 342                 return show_fs_alloc_debug(c, buf);
 343
 344         if (attr == &sysfs_journal_debug)
 345                 return bch2_journal_print_debug(&c->journal, buf);
 346
 347         if (attr == &sysfs_journal_pins)
 348                 return bch2_journal_print_pins(&c->journal, buf);
 349
 350         if (attr == &sysfs_compression_stats)
 351                 return bch2_compression_stats(c, buf);
 352
 353 #define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
 354         BCH_DEBUG_PARAMS()
 355 #undef BCH_DEBUG_PARAM
 356
 357         return 0;
 358 }
 359
 360 STORE(__bch2_fs)
 361 {
 362         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 363
 364         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 365         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 366
 367         sysfs_strtoul(foreground_write_ratelimit_enabled,
 368                       c->foreground_write_ratelimit_enabled);
 369
 370         if (attr == &sysfs_copy_gc_enabled) {
 371                 struct bch_dev *ca;
 372                 unsigned i;
 373                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 374                         ?: (ssize_t) size;
 375
 376                 for_each_member_device(ca, c, i)
 377                         if (ca->moving_gc_read)
 378                                 wake_up_process(ca->moving_gc_read);
 379                 return ret;
 380         }
 381
 382         if (attr == &sysfs_tiering_enabled) {
 383                 ssize_t ret = strtoul_safe(buf, c->tiering_enabled)
 384                         ?: (ssize_t) size;
 385
 386                 bch2_tiering_start(c); /* issue wakeups */
 387                 return ret;
 388         }
 389
 390         sysfs_pd_controller_store(foreground_write, &c->foreground_write_pd);
 391
 392         sysfs_strtoul(pd_controllers_update_seconds,
 393                       c->pd_controllers_update_seconds);
 394         sysfs_strtoul(foreground_target_percent, c->foreground_target_percent);
 395
 396         sysfs_strtoul(tiering_percent,          c->tiering_percent);
 397         sysfs_pd_controller_store(tiering,      &c->tiers[1].pd); /* XXX */
 398
 399         /* Debugging: */
 400
 401 #define BCH_DEBUG_PARAM(name, description) sysfs_strtoul(name, c->name);
 402         BCH_DEBUG_PARAMS()
 403 #undef BCH_DEBUG_PARAM
 404
 405         if (!bch2_fs_running(c))
 406                 return -EPERM;
 407
 408         /* Debugging: */
 409
 410         if (attr == &sysfs_trigger_journal_flush)
 411                 bch2_journal_meta_async(&c->journal, NULL);
 412
 413         if (attr == &sysfs_trigger_btree_coalesce)
 414                 bch2_coalesce(c);
 415
 416         if (attr == &sysfs_trigger_gc)
 417                 bch2_gc(c);
 418
 419         if (attr == &sysfs_prune_cache) {
 420                 struct shrink_control sc;
 421
 422                 sc.gfp_mask = GFP_KERNEL;
 423                 sc.nr_to_scan = strtoul_or_return(buf);
 424                 c->btree_cache_shrink.scan_objects(&c->btree_cache_shrink, &sc);
 425         }
 426
 427         return size;
 428 }
 429
 430 STORE(bch2_fs)
 431 {
 432         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 433
 434         mutex_lock(&c->state_lock);
 435         size = __bch2_fs_store(kobj, attr, buf, size);
 436         mutex_unlock(&c->state_lock);
 437
 438         return size;
 439 }
 440 SYSFS_OPS(bch2_fs);
 441
 442 struct attribute *bch2_fs_files[] = {
 443         &sysfs_minor,
 444         &sysfs_block_size,
 445         &sysfs_btree_node_size,
 446         &sysfs_btree_cache_size,
 447
 448         &sysfs_meta_replicas_have,
 449         &sysfs_data_replicas_have,
 450
 451         &sysfs_journal_write_delay_ms,
 452         &sysfs_journal_reclaim_delay_ms,
 453
 454         &sysfs_foreground_target_percent,
 455         &sysfs_tiering_percent,
 456
 457         &sysfs_compression_stats,
 458         NULL
 459 };
 460
 461 /* internal dir - just a wrapper */
 462
 463 SHOW(bch2_fs_internal)
 464 {
 465         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 466         return bch2_fs_show(&c->kobj, attr, buf);
 467 }
 468
 469 STORE(bch2_fs_internal)
 470 {
 471         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 472         return bch2_fs_store(&c->kobj, attr, buf, size);
 473 }
 474 SYSFS_OPS(bch2_fs_internal);
 475
 476 struct attribute *bch2_fs_internal_files[] = {
 477         &sysfs_alloc_debug,
 478         &sysfs_journal_debug,
 479         &sysfs_journal_pins,
 480
 481         &sysfs_read_realloc_races,
 482
 483         &sysfs_trigger_journal_flush,
 484         &sysfs_trigger_btree_coalesce,
 485         &sysfs_trigger_gc,
 486         &sysfs_prune_cache,
 487
 488         &sysfs_foreground_write_ratelimit_enabled,
 489         &sysfs_copy_gc_enabled,
 490         &sysfs_tiering_enabled,
 491         sysfs_pd_controller_files(tiering),
 492         sysfs_pd_controller_files(foreground_write),
 493         &sysfs_internal_uuid,
 494
 495 #define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
 496         BCH_DEBUG_PARAMS()
 497 #undef BCH_DEBUG_PARAM
 498
 499         NULL
 500 };
 501
 502 /* options */
 503
 504 SHOW(bch2_fs_opts_dir)
 505 {
 506         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 507
 508         return bch2_opt_show(&c->opts, attr->name, buf, PAGE_SIZE);
 509 }
 510
 511 STORE(bch2_fs_opts_dir)
 512 {
 513         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 514         const struct bch_option *opt;
 515         int id;
 516         u64 v;
 517
 518         id = bch2_parse_sysfs_opt(attr->name, buf, &v);
 519         if (id < 0)
 520                 return id;
 521
 522         opt = &bch2_opt_table[id];
 523
 524         mutex_lock(&c->sb_lock);
 525
 526         if (id == Opt_compression) {
 527                 int ret = bch2_check_set_has_compressed_data(c, v);
 528                 if (ret) {
 529                         mutex_unlock(&c->sb_lock);
 530                         return ret;
 531                 }
 532         }
 533
 534         if (opt->set_sb != SET_NO_SB_OPT) {
 535                 opt->set_sb(c->disk_sb, v);
 536                 bch2_write_super(c);
 537         }
 538
 539         bch2_opt_set(&c->opts, id, v);
 540
 541         mutex_unlock(&c->sb_lock);
 542
 543         return size;
 544 }
 545 SYSFS_OPS(bch2_fs_opts_dir);
 546
 547 struct attribute *bch2_fs_opts_dir_files[] = {
 548 #define BCH_OPT(_name, ...)                                             \
 549         &sysfs_opt_##_name,
 550
 551         BCH_VISIBLE_OPTS()
 552 #undef BCH_OPT
 553
 554         NULL
 555 };
 556
 557 /* time stats */
 558
 559 SHOW(bch2_fs_time_stats)
 560 {
 561         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 562
 563 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 564         sysfs_print_time_stats(&c->name##_time, name,                   \
 565                                frequency_units, duration_units);
 566         BCH_TIME_STATS()
 567 #undef BCH_TIME_STAT
 568
 569         return 0;
 570 }
 571
 572 STORE(bch2_fs_time_stats)
 573 {
 574         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 575
 576 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 577         sysfs_clear_time_stats(&c->name##_time, name);
 578         BCH_TIME_STATS()
 579 #undef BCH_TIME_STAT
 580
 581         return size;
 582 }
 583 SYSFS_OPS(bch2_fs_time_stats);
 584
 585 struct attribute *bch2_fs_time_stats_files[] = {
 586 #define BCH_TIME_STAT(name, frequency_units, duration_units)            \
 587         sysfs_time_stats_attribute_list(name, frequency_units, duration_units)
 588         BCH_TIME_STATS()
 589 #undef BCH_TIME_STAT
 590
 591         NULL
 592 };
 593
 594 typedef unsigned (bucket_map_fn)(struct bch_dev *, struct bucket *, void *);
 595
 596 static unsigned bucket_priority_fn(struct bch_dev *ca, struct bucket *g,
 597                                    void *private)
 598 {
 599         int rw = (private ? 1 : 0);
 600
 601         return ca->fs->prio_clock[rw].hand - g->prio[rw];
 602 }
 603
 604 static unsigned bucket_sectors_used_fn(struct bch_dev *ca, struct bucket *g,
 605                                        void *private)
 606 {
 607         return bucket_sectors_used(g);
 608 }
 609
 610 static unsigned bucket_oldest_gen_fn(struct bch_dev *ca, struct bucket *g,
 611                                      void *private)
 612 {
 613         return bucket_gc_gen(ca, g);
 614 }
 615
 616 static ssize_t show_quantiles(struct bch_dev *ca, char *buf,
 617                               bucket_map_fn *fn, void *private)
 618 {
 619         int cmp(const void *l, const void *r)
 620         {       return *((unsigned *) r) - *((unsigned *) l); }
 621
 622         size_t n = ca->mi.nbuckets, i;
 623         /* Compute 31 quantiles */
 624         unsigned q[31], *p;
 625         ssize_t ret = 0;
 626
 627         p = vzalloc(ca->mi.nbuckets * sizeof(unsigned));
 628         if (!p)
 629                 return -ENOMEM;
 630
 631         for (i = ca->mi.first_bucket; i < n; i++)
 632                 p[i] = fn(ca, &ca->buckets[i], private);
 633
 634         sort(p, n, sizeof(unsigned), cmp, NULL);
 635
 636         while (n &&
 637                !p[n - 1])
 638                 --n;
 639
 640         for (i = 0; i < ARRAY_SIZE(q); i++)
 641                 q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
 642
 643         vfree(p);
 644
 645         for (i = 0; i < ARRAY_SIZE(q); i++)
 646                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 647                                  "%u ", q[i]);
 648         buf[ret - 1] = '\n';
 649
 650         return ret;
 651
 652 }
 653
 654 static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
 655 {
 656         enum alloc_reserve i;
 657         ssize_t ret;
 658
 659         spin_lock(&ca->freelist_lock);
 660
 661         ret = scnprintf(buf, PAGE_SIZE,
 662                         "free_inc:\t%zu\t%zu\n",
 663                         fifo_used(&ca->free_inc),
 664                         ca->free_inc.size);
 665
 666         for (i = 0; i < RESERVE_NR; i++)
 667                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 668                                  "free[%u]:\t%zu\t%zu\n", i,
 669                                  fifo_used(&ca->free[i]),
 670                                  ca->free[i].size);
 671
 672         spin_unlock(&ca->freelist_lock);
 673
 674         return ret;
 675 }
 676
 677 static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
 678 {
 679         struct bch_fs *c = ca->fs;
 680         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 681
 682         return scnprintf(buf, PAGE_SIZE,
 683                 "free_inc:               %zu/%zu\n"
 684                 "free[RESERVE_PRIO]:     %zu/%zu\n"
 685                 "free[RESERVE_BTREE]:    %zu/%zu\n"
 686                 "free[RESERVE_MOVINGGC]: %zu/%zu\n"
 687                 "free[RESERVE_NONE]:     %zu/%zu\n"
 688                 "alloc:                  %llu/%llu\n"
 689                 "meta:                   %llu/%llu\n"
 690                 "dirty:                  %llu/%llu\n"
 691                 "available:              %llu/%llu\n"
 692                 "freelist_wait:          %s\n"
 693                 "open buckets:           %u/%u (reserved %u)\n"
 694                 "open_buckets_wait:      %s\n",
 695                 fifo_used(&ca->free_inc),               ca->free_inc.size,
 696                 fifo_used(&ca->free[RESERVE_PRIO]),     ca->free[RESERVE_PRIO].size,
 697                 fifo_used(&ca->free[RESERVE_BTREE]),    ca->free[RESERVE_BTREE].size,
 698                 fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size,
 699                 fifo_used(&ca->free[RESERVE_NONE]),     ca->free[RESERVE_NONE].size,
 700                 stats.buckets_alloc,                    ca->mi.nbuckets - ca->mi.first_bucket,
 701                 stats.buckets_meta,                     ca->mi.nbuckets - ca->mi.first_bucket,
 702                 stats.buckets_dirty,                    ca->mi.nbuckets - ca->mi.first_bucket,
 703                 __dev_buckets_available(ca, stats),     ca->mi.nbuckets - ca->mi.first_bucket,
 704                 c->freelist_wait.list.first             ? "waiting" : "empty",
 705                 c->open_buckets_nr_free, OPEN_BUCKETS_COUNT, BTREE_NODE_RESERVE,
 706                 c->open_buckets_wait.list.first         ? "waiting" : "empty");
 707 }
 708
 709 static u64 sectors_written(struct bch_dev *ca)
 710 {
 711         u64 ret = 0;
 712         int cpu;
 713
 714         for_each_possible_cpu(cpu)
 715                 ret += *per_cpu_ptr(ca->sectors_written, cpu);
 716
 717         return ret;
 718 }
 719
 720 SHOW(bch2_dev)
 721 {
 722         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 723         struct bch_fs *c = ca->fs;
 724         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 725
 726         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 727
 728         sysfs_print(bucket_size,        bucket_bytes(ca));
 729         sysfs_print(block_size,         block_bytes(c));
 730         sysfs_print(first_bucket,       ca->mi.first_bucket);
 731         sysfs_print(nbuckets,           ca->mi.nbuckets);
 732         sysfs_print(discard,            ca->mi.discard);
 733         sysfs_hprint(written, sectors_written(ca) << 9);
 734         sysfs_hprint(btree_written,
 735                      atomic64_read(&ca->btree_sectors_written) << 9);
 736         sysfs_hprint(metadata_written,
 737                      (atomic64_read(&ca->meta_sectors_written) +
 738                       atomic64_read(&ca->btree_sectors_written)) << 9);
 739
 740         sysfs_hprint(dirty_data,        stats.sectors[S_DIRTY] << 9);
 741         sysfs_print(dirty_bytes,        stats.sectors[S_DIRTY] << 9);
 742         sysfs_print(dirty_buckets,      stats.buckets_dirty);
 743         sysfs_hprint(cached_data,       stats.sectors[S_CACHED] << 9);
 744         sysfs_print(cached_bytes,       stats.sectors[S_CACHED] << 9);
 745         sysfs_print(cached_buckets,     stats.buckets_cached);
 746         sysfs_print(meta_buckets,       stats.buckets_meta);
 747         sysfs_print(alloc_buckets,      stats.buckets_alloc);
 748         sysfs_print(available_buckets,  dev_buckets_available(ca));
 749         sysfs_print(free_buckets,       dev_buckets_free(ca));
 750         sysfs_print(has_data,           ca->mi.has_data);
 751         sysfs_print(has_metadata,       ca->mi.has_metadata);
 752
 753         sysfs_pd_controller_show(copy_gc, &ca->moving_gc_pd);
 754
 755         if (attr == &sysfs_cache_replacement_policy)
 756                 return bch2_snprint_string_list(buf, PAGE_SIZE,
 757                                                 bch2_cache_replacement_policies,
 758                                                 ca->mi.replacement);
 759
 760         sysfs_print(tier,               ca->mi.tier);
 761
 762         if (attr == &sysfs_state_rw)
 763                 return bch2_snprint_string_list(buf, PAGE_SIZE,
 764                                                 bch2_dev_state,
 765                                                 ca->mi.state);
 766
 767         if (attr == &sysfs_read_priority_stats)
 768                 return show_quantiles(ca, buf, bucket_priority_fn, (void *) 0);
 769         if (attr == &sysfs_write_priority_stats)
 770                 return show_quantiles(ca, buf, bucket_priority_fn, (void *) 1);
 771         if (attr == &sysfs_fragmentation_stats)
 772                 return show_quantiles(ca, buf, bucket_sectors_used_fn, NULL);
 773         if (attr == &sysfs_oldest_gen_stats)
 774                 return show_quantiles(ca, buf, bucket_oldest_gen_fn, NULL);
 775         if (attr == &sysfs_reserve_stats)
 776                 return show_reserve_stats(ca, buf);
 777         if (attr == &sysfs_alloc_debug)
 778                 return show_dev_alloc_debug(ca, buf);
 779
 780         return 0;
 781 }
 782
 783 STORE(bch2_dev)
 784 {
 785         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 786         struct bch_fs *c = ca->fs;
 787         struct bch_member *mi;
 788
 789         sysfs_pd_controller_store(copy_gc, &ca->moving_gc_pd);
 790
 791         if (attr == &sysfs_discard) {
 792                 bool v = strtoul_or_return(buf);
 793
 794                 mutex_lock(&c->sb_lock);
 795                 mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
 796
 797                 if (v != BCH_MEMBER_DISCARD(mi)) {
 798                         SET_BCH_MEMBER_DISCARD(mi, v);
 799                         bch2_write_super(c);
 800                 }
 801                 mutex_unlock(&c->sb_lock);
 802         }
 803
 804         if (attr == &sysfs_cache_replacement_policy) {
 805                 ssize_t v = bch2_read_string_list(buf, bch2_cache_replacement_policies);
 806
 807                 if (v < 0)
 808                         return v;
 809
 810                 mutex_lock(&c->sb_lock);
 811                 mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
 812
 813                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
 814                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
 815                         bch2_write_super(c);
 816                 }
 817                 mutex_unlock(&c->sb_lock);
 818         }
 819
 820         if (attr == &sysfs_tier) {
 821                 unsigned prev_tier;
 822                 unsigned v = strtoul_restrict_or_return(buf,
 823                                         0, BCH_TIER_MAX - 1);
 824
 825                 mutex_lock(&c->sb_lock);
 826                 prev_tier = ca->mi.tier;
 827
 828                 if (v == ca->mi.tier) {
 829                         mutex_unlock(&c->sb_lock);
 830                         return size;
 831                 }
 832
 833                 mi = &bch2_sb_get_members(c->disk_sb)->members[ca->dev_idx];
 834                 SET_BCH_MEMBER_TIER(mi, v);
 835                 bch2_write_super(c);
 836
 837                 bch2_dev_group_remove(&c->tiers[prev_tier].devs, ca);
 838                 bch2_dev_group_add(&c->tiers[ca->mi.tier].devs, ca);
 839                 mutex_unlock(&c->sb_lock);
 840
 841                 bch2_recalc_capacity(c);
 842                 bch2_tiering_start(c);
 843         }
 844
 845         return size;
 846 }
 847 SYSFS_OPS(bch2_dev);
 848
 849 struct attribute *bch2_dev_files[] = {
 850         &sysfs_uuid,
 851         &sysfs_bucket_size,
 852         &sysfs_block_size,
 853         &sysfs_first_bucket,
 854         &sysfs_nbuckets,
 855
 856         /* settings: */
 857         &sysfs_discard,
 858         &sysfs_cache_replacement_policy,
 859         &sysfs_tier,
 860         &sysfs_state_rw,
 861
 862         &sysfs_has_data,
 863         &sysfs_has_metadata,
 864
 865         /* io stats: */
 866         &sysfs_written,
 867         &sysfs_btree_written,
 868         &sysfs_metadata_written,
 869
 870         /* alloc info - data: */
 871         &sysfs_dirty_data,
 872         &sysfs_dirty_bytes,
 873         &sysfs_cached_data,
 874         &sysfs_cached_bytes,
 875
 876         /* alloc info - buckets: */
 877         &sysfs_available_buckets,
 878         &sysfs_free_buckets,
 879         &sysfs_dirty_buckets,
 880         &sysfs_cached_buckets,
 881         &sysfs_meta_buckets,
 882         &sysfs_alloc_buckets,
 883
 884         /* alloc info - other stats: */
 885         &sysfs_read_priority_stats,
 886         &sysfs_write_priority_stats,
 887         &sysfs_fragmentation_stats,
 888         &sysfs_oldest_gen_stats,
 889         &sysfs_reserve_stats,
 890
 891         /* debug: */
 892         &sysfs_alloc_debug,
 893
 894         sysfs_pd_controller_files(copy_gc),
 895         NULL
 896 };