git.sesse.net Git - bcachefs-tools-debian/blob - libbcachefs/sysfs.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * bcache sysfs interfaces
   4  *
   5  * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
   6  * Copyright 2012 Google, Inc.
   7  */
   8
   9 #ifndef NO_BCACHEFS_SYSFS
  10
  11 #include "bcachefs.h"
  12 #include "alloc_background.h"
  13 #include "sysfs.h"
  14 #include "btree_cache.h"
  15 #include "btree_io.h"
  16 #include "btree_iter.h"
  17 #include "btree_key_cache.h"
  18 #include "btree_update.h"
  19 #include "btree_update_interior.h"
  20 #include "btree_gc.h"
  21 #include "buckets.h"
  22 #include "clock.h"
  23 #include "disk_groups.h"
  24 #include "ec.h"
  25 #include "inode.h"
  26 #include "journal.h"
  27 #include "keylist.h"
  28 #include "move.h"
  29 #include "opts.h"
  30 #include "rebalance.h"
  31 #include "replicas.h"
  32 #include "super-io.h"
  33 #include "tests.h"
  34
  35 #include <linux/blkdev.h>
  36 #include <linux/sort.h>
  37 #include <linux/sched/clock.h>
  38
  39 #include "util.h"
  40
  41 #define SYSFS_OPS(type)                                                 \
  42 struct sysfs_ops type ## _sysfs_ops = {                                 \
  43         .show   = type ## _show,                                        \
  44         .store  = type ## _store                                        \
  45 }
  46
  47 #define SHOW(fn)                                                        \
  48 static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
  49                            char *buf)                                   \
  50
  51 #define STORE(fn)                                                       \
  52 static ssize_t fn ## _store(struct kobject *kobj, struct attribute *attr,\
  53                             const char *buf, size_t size)               \
  54
  55 #define __sysfs_attribute(_name, _mode)                                 \
  56         static struct attribute sysfs_##_name =                         \
  57                 { .name = #_name, .mode = _mode }
  58
  59 #define write_attribute(n)      __sysfs_attribute(n, S_IWUSR)
  60 #define read_attribute(n)       __sysfs_attribute(n, S_IRUGO)
  61 #define rw_attribute(n)         __sysfs_attribute(n, S_IRUGO|S_IWUSR)
  62
  63 #define sysfs_printf(file, fmt, ...)                                    \
  64 do {                                                                    \
  65         if (attr == &sysfs_ ## file)                                    \
  66                 return scnprintf(buf, PAGE_SIZE, fmt "\n", __VA_ARGS__);\
  67 } while (0)
  68
  69 #define sysfs_print(file, var)                                          \
  70 do {                                                                    \
  71         if (attr == &sysfs_ ## file)                                    \
  72                 return snprint(buf, PAGE_SIZE, var);                    \
  73 } while (0)
  74
  75 #define sysfs_hprint(file, val)                                         \
  76 do {                                                                    \
  77         if (attr == &sysfs_ ## file) {                                  \
  78                 bch2_hprint(&out, val);                                 \
  79                 pr_buf(&out, "\n");                                     \
  80                 return out.pos - buf;                                   \
  81         }                                                               \
  82 } while (0)
  83
  84 #define var_printf(_var, fmt)   sysfs_printf(_var, fmt, var(_var))
  85 #define var_print(_var)         sysfs_print(_var, var(_var))
  86 #define var_hprint(_var)        sysfs_hprint(_var, var(_var))
  87
  88 #define sysfs_strtoul(file, var)                                        \
  89 do {                                                                    \
  90         if (attr == &sysfs_ ## file)                                    \
  91                 return strtoul_safe(buf, var) ?: (ssize_t) size;        \
  92 } while (0)
  93
  94 #define sysfs_strtoul_clamp(file, var, min, max)                        \
  95 do {                                                                    \
  96         if (attr == &sysfs_ ## file)                                    \
  97                 return strtoul_safe_clamp(buf, var, min, max)           \
  98                         ?: (ssize_t) size;                              \
  99 } while (0)
 100
 101 #define strtoul_or_return(cp)                                           \
 102 ({                                                                      \
 103         unsigned long _v;                                               \
 104         int _r = kstrtoul(cp, 10, &_v);                                 \
 105         if (_r)                                                         \
 106                 return _r;                                              \
 107         _v;                                                             \
 108 })
 109
 110 #define strtoul_restrict_or_return(cp, min, max)                        \
 111 ({                                                                      \
 112         unsigned long __v = 0;                                          \
 113         int _r = strtoul_safe_restrict(cp, __v, min, max);              \
 114         if (_r)                                                         \
 115                 return _r;                                              \
 116         __v;                                                            \
 117 })
 118
 119 #define strtoi_h_or_return(cp)                                          \
 120 ({                                                                      \
 121         u64 _v;                                                         \
 122         int _r = strtoi_h(cp, &_v);                                     \
 123         if (_r)                                                         \
 124                 return _r;                                              \
 125         _v;                                                             \
 126 })
 127
 128 #define sysfs_hatoi(file, var)                                          \
 129 do {                                                                    \
 130         if (attr == &sysfs_ ## file)                                    \
 131                 return strtoi_h(buf, &var) ?: (ssize_t) size;           \
 132 } while (0)
 133
 134 write_attribute(trigger_journal_flush);
 135 write_attribute(trigger_gc);
 136 write_attribute(prune_cache);
 137 rw_attribute(btree_gc_periodic);
 138 rw_attribute(gc_gens_pos);
 139
 140 read_attribute(uuid);
 141 read_attribute(minor);
 142 read_attribute(bucket_size);
 143 read_attribute(first_bucket);
 144 read_attribute(nbuckets);
 145 read_attribute(durability);
 146 read_attribute(iodone);
 147
 148 read_attribute(io_latency_read);
 149 read_attribute(io_latency_write);
 150 read_attribute(io_latency_stats_read);
 151 read_attribute(io_latency_stats_write);
 152 read_attribute(congested);
 153
 154 read_attribute(btree_avg_write_size);
 155
 156 read_attribute(reserve_stats);
 157 read_attribute(btree_cache_size);
 158 read_attribute(compression_stats);
 159 read_attribute(journal_debug);
 160 read_attribute(journal_pins);
 161 read_attribute(btree_updates);
 162 read_attribute(dirty_btree_nodes);
 163 read_attribute(btree_cache);
 164 read_attribute(btree_key_cache);
 165 read_attribute(btree_transactions);
 166 read_attribute(stripes_heap);
 167 read_attribute(open_buckets);
 168
 169 read_attribute(internal_uuid);
 170
 171 read_attribute(has_data);
 172 read_attribute(alloc_debug);
 173 write_attribute(wake_allocator);
 174
 175 read_attribute(read_realloc_races);
 176 read_attribute(extent_migrate_done);
 177 read_attribute(extent_migrate_raced);
 178
 179 rw_attribute(discard);
 180 rw_attribute(cache_replacement_policy);
 181 rw_attribute(label);
 182
 183 rw_attribute(copy_gc_enabled);
 184 read_attribute(copy_gc_wait);
 185
 186 rw_attribute(rebalance_enabled);
 187 sysfs_pd_controller_attribute(rebalance);
 188 read_attribute(rebalance_work);
 189 rw_attribute(promote_whole_extents);
 190
 191 read_attribute(new_stripes);
 192
 193 read_attribute(io_timers_read);
 194 read_attribute(io_timers_write);
 195
 196 read_attribute(data_op_data_progress);
 197
 198 #ifdef CONFIG_BCACHEFS_TESTS
 199 write_attribute(perf_test);
 200 #endif /* CONFIG_BCACHEFS_TESTS */
 201
 202 #define x(_name)                                                \
 203         static struct attribute sysfs_time_stat_##_name =               \
 204                 { .name = #_name, .mode = S_IRUGO };
 205         BCH_TIME_STATS()
 206 #undef x
 207
 208 static struct attribute sysfs_state_rw = {
 209         .name = "state",
 210         .mode = S_IRUGO
 211 };
 212
 213 static size_t bch2_btree_cache_size(struct bch_fs *c)
 214 {
 215         size_t ret = 0;
 216         struct btree *b;
 217
 218         mutex_lock(&c->btree_cache.lock);
 219         list_for_each_entry(b, &c->btree_cache.live, list)
 220                 ret += btree_bytes(c);
 221
 222         mutex_unlock(&c->btree_cache.lock);
 223         return ret;
 224 }
 225
 226 static size_t bch2_btree_avg_write_size(struct bch_fs *c)
 227 {
 228         u64 nr = atomic64_read(&c->btree_writes_nr);
 229         u64 sectors = atomic64_read(&c->btree_writes_sectors);
 230
 231         return nr ? div64_u64(sectors, nr) : 0;
 232 }
 233
 234 static long stats_to_text(struct printbuf *out, struct bch_fs *c,
 235                           struct bch_move_stats *stats)
 236 {
 237         pr_buf(out, "%s: data type %s btree_id %s position: ",
 238                 stats->name,
 239                 bch2_data_types[stats->data_type],
 240                 bch2_btree_ids[stats->btree_id]);
 241         bch2_bpos_to_text(out, stats->pos);
 242         pr_buf(out, "%s", "\n");
 243
 244         return 0;
 245 }
 246
 247 static long data_progress_to_text(struct printbuf *out, struct bch_fs *c)
 248 {
 249         long ret = 0;
 250         struct bch_move_stats *iter;
 251
 252         mutex_lock(&c->data_progress_lock);
 253
 254         if (list_empty(&c->data_progress_list))
 255                 pr_buf(out, "%s", "no progress to report\n");
 256         else
 257                 list_for_each_entry(iter, &c->data_progress_list, list) {
 258                         stats_to_text(out, c, iter);
 259                 }
 260
 261         mutex_unlock(&c->data_progress_lock);
 262         return ret;
 263 }
 264
 265 static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c)
 266 {
 267         struct btree_trans trans;
 268         struct btree_iter iter;
 269         struct bkey_s_c k;
 270         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 271             nr_compressed_extents = 0,
 272             compressed_sectors_compressed = 0,
 273             compressed_sectors_uncompressed = 0;
 274         int ret;
 275
 276         if (!test_bit(BCH_FS_STARTED, &c->flags))
 277                 return -EPERM;
 278
 279         bch2_trans_init(&trans, c, 0, 0);
 280
 281         for_each_btree_key(&trans, iter, BTREE_ID_extents, POS_MIN, 0, k, ret)
 282                 if (k.k->type == KEY_TYPE_extent) {
 283                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 284                         const union bch_extent_entry *entry;
 285                         struct extent_ptr_decoded p;
 286
 287                         extent_for_each_ptr_decode(e, p, entry) {
 288                                 if (!crc_is_compressed(p.crc)) {
 289                                         nr_uncompressed_extents++;
 290                                         uncompressed_sectors += e.k->size;
 291                                 } else {
 292                                         nr_compressed_extents++;
 293                                         compressed_sectors_compressed +=
 294                                                 p.crc.compressed_size;
 295                                         compressed_sectors_uncompressed +=
 296                                                 p.crc.uncompressed_size;
 297                                 }
 298
 299                                 /* only looking at the first ptr */
 300                                 break;
 301                         }
 302                 }
 303         bch2_trans_iter_exit(&trans, &iter);
 304
 305         bch2_trans_exit(&trans);
 306         if (ret)
 307                 return ret;
 308
 309         pr_buf(out,
 310                "uncompressed data:\n"
 311                "        nr extents:                     %llu\n"
 312                "        size (bytes):                   %llu\n"
 313                "compressed data:\n"
 314                "        nr extents:                     %llu\n"
 315                "        compressed size (bytes):        %llu\n"
 316                "        uncompressed size (bytes):      %llu\n",
 317                nr_uncompressed_extents,
 318                uncompressed_sectors << 9,
 319                nr_compressed_extents,
 320                compressed_sectors_compressed << 9,
 321                compressed_sectors_uncompressed << 9);
 322         return 0;
 323 }
 324
 325 static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
 326 {
 327         pr_buf(out, "%s: ", bch2_btree_ids[c->gc_gens_btree]);
 328         bch2_bpos_to_text(out, c->gc_gens_pos);
 329         pr_buf(out, "\n");
 330 }
 331
 332 SHOW(bch2_fs)
 333 {
 334         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 335         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 336
 337         sysfs_print(minor,                      c->minor);
 338         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 339
 340         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 341         sysfs_hprint(btree_avg_write_size,      bch2_btree_avg_write_size(c));
 342
 343         sysfs_print(read_realloc_races,
 344                     atomic_long_read(&c->read_realloc_races));
 345         sysfs_print(extent_migrate_done,
 346                     atomic_long_read(&c->extent_migrate_done));
 347         sysfs_print(extent_migrate_raced,
 348                     atomic_long_read(&c->extent_migrate_raced));
 349
 350         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 351
 352         if (attr == &sysfs_gc_gens_pos) {
 353                 bch2_gc_gens_pos_to_text(&out, c);
 354                 return out.pos - buf;
 355         }
 356
 357         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 358
 359         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 360         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 361         sysfs_hprint(copy_gc_wait,
 362                      max(0LL, c->copygc_wait -
 363                          atomic64_read(&c->io_clock[WRITE].now)) << 9);
 364
 365         if (attr == &sysfs_rebalance_work) {
 366                 bch2_rebalance_work_to_text(&out, c);
 367                 return out.pos - buf;
 368         }
 369
 370         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 371
 372         /* Debugging: */
 373
 374         if (attr == &sysfs_journal_debug) {
 375                 bch2_journal_debug_to_text(&out, &c->journal);
 376                 return out.pos - buf;
 377         }
 378
 379         if (attr == &sysfs_journal_pins) {
 380                 bch2_journal_pins_to_text(&out, &c->journal);
 381                 return out.pos - buf;
 382         }
 383
 384         if (attr == &sysfs_btree_updates) {
 385                 bch2_btree_updates_to_text(&out, c);
 386                 return out.pos - buf;
 387         }
 388
 389         if (attr == &sysfs_dirty_btree_nodes) {
 390                 bch2_dirty_btree_nodes_to_text(&out, c);
 391                 return out.pos - buf;
 392         }
 393
 394         if (attr == &sysfs_btree_cache) {
 395                 bch2_btree_cache_to_text(&out, c);
 396                 return out.pos - buf;
 397         }
 398
 399         if (attr == &sysfs_btree_key_cache) {
 400                 bch2_btree_key_cache_to_text(&out, &c->btree_key_cache);
 401                 return out.pos - buf;
 402         }
 403
 404         if (attr == &sysfs_btree_transactions) {
 405                 bch2_btree_trans_to_text(&out, c);
 406                 return out.pos - buf;
 407         }
 408
 409         if (attr == &sysfs_stripes_heap) {
 410                 bch2_stripes_heap_to_text(&out, c);
 411                 return out.pos - buf;
 412         }
 413
 414         if (attr == &sysfs_open_buckets) {
 415                 bch2_open_buckets_to_text(&out, c);
 416                 return out.pos - buf;
 417         }
 418
 419         if (attr == &sysfs_compression_stats) {
 420                 bch2_compression_stats_to_text(&out, c);
 421                 return out.pos - buf;
 422         }
 423
 424         if (attr == &sysfs_new_stripes) {
 425                 bch2_new_stripes_to_text(&out, c);
 426                 return out.pos - buf;
 427         }
 428
 429         if (attr == &sysfs_io_timers_read) {
 430                 bch2_io_timers_to_text(&out, &c->io_clock[READ]);
 431                 return out.pos - buf;
 432         }
 433         if (attr == &sysfs_io_timers_write) {
 434                 bch2_io_timers_to_text(&out, &c->io_clock[WRITE]);
 435                 return out.pos - buf;
 436         }
 437
 438         if (attr == &sysfs_data_op_data_progress) {
 439                 data_progress_to_text(&out, c);
 440                 return out.pos - buf;
 441         }
 442
 443         return 0;
 444 }
 445
 446 STORE(bch2_fs)
 447 {
 448         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 449
 450         if (attr == &sysfs_btree_gc_periodic) {
 451                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 452                         ?: (ssize_t) size;
 453
 454                 wake_up_process(c->gc_thread);
 455                 return ret;
 456         }
 457
 458         if (attr == &sysfs_copy_gc_enabled) {
 459                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 460                         ?: (ssize_t) size;
 461
 462                 if (c->copygc_thread)
 463                         wake_up_process(c->copygc_thread);
 464                 return ret;
 465         }
 466
 467         if (attr == &sysfs_rebalance_enabled) {
 468                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 469                         ?: (ssize_t) size;
 470
 471                 rebalance_wakeup(c);
 472                 return ret;
 473         }
 474
 475         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 476
 477         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 478
 479         /* Debugging: */
 480
 481         if (!test_bit(BCH_FS_STARTED, &c->flags))
 482                 return -EPERM;
 483
 484         /* Debugging: */
 485
 486         if (attr == &sysfs_trigger_journal_flush)
 487                 bch2_journal_meta(&c->journal);
 488
 489         if (attr == &sysfs_trigger_gc) {
 490                 /*
 491                  * Full gc is currently incompatible with btree key cache:
 492                  */
 493 #if 0
 494                 down_read(&c->state_lock);
 495                 bch2_gc(c, false, false);
 496                 up_read(&c->state_lock);
 497 #else
 498                 bch2_gc_gens(c);
 499 #endif
 500         }
 501
 502         if (attr == &sysfs_prune_cache) {
 503                 struct shrink_control sc;
 504
 505                 sc.gfp_mask = GFP_KERNEL;
 506                 sc.nr_to_scan = strtoul_or_return(buf);
 507                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 508         }
 509
 510 #ifdef CONFIG_BCACHEFS_TESTS
 511         if (attr == &sysfs_perf_test) {
 512                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 513                 char *test              = strsep(&p, " \t\n");
 514                 char *nr_str            = strsep(&p, " \t\n");
 515                 char *threads_str       = strsep(&p, " \t\n");
 516                 unsigned threads;
 517                 u64 nr;
 518                 int ret = -EINVAL;
 519
 520                 if (threads_str &&
 521                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 522                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 523                         ret = bch2_btree_perf_test(c, test, nr, threads);
 524                 kfree(tmp);
 525
 526                 if (ret)
 527                         size = ret;
 528         }
 529 #endif
 530         return size;
 531 }
 532 SYSFS_OPS(bch2_fs);
 533
 534 struct attribute *bch2_fs_files[] = {
 535         &sysfs_minor,
 536         &sysfs_btree_cache_size,
 537         &sysfs_btree_avg_write_size,
 538
 539         &sysfs_promote_whole_extents,
 540
 541         &sysfs_compression_stats,
 542
 543 #ifdef CONFIG_BCACHEFS_TESTS
 544         &sysfs_perf_test,
 545 #endif
 546         NULL
 547 };
 548
 549 /* internal dir - just a wrapper */
 550
 551 SHOW(bch2_fs_internal)
 552 {
 553         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 554         return bch2_fs_show(&c->kobj, attr, buf);
 555 }
 556
 557 STORE(bch2_fs_internal)
 558 {
 559         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 560         return bch2_fs_store(&c->kobj, attr, buf, size);
 561 }
 562 SYSFS_OPS(bch2_fs_internal);
 563
 564 struct attribute *bch2_fs_internal_files[] = {
 565         &sysfs_journal_debug,
 566         &sysfs_journal_pins,
 567         &sysfs_btree_updates,
 568         &sysfs_dirty_btree_nodes,
 569         &sysfs_btree_cache,
 570         &sysfs_btree_key_cache,
 571         &sysfs_btree_transactions,
 572         &sysfs_new_stripes,
 573         &sysfs_stripes_heap,
 574         &sysfs_open_buckets,
 575         &sysfs_io_timers_read,
 576         &sysfs_io_timers_write,
 577
 578         &sysfs_trigger_journal_flush,
 579         &sysfs_trigger_gc,
 580         &sysfs_prune_cache,
 581
 582         &sysfs_read_realloc_races,
 583         &sysfs_extent_migrate_done,
 584         &sysfs_extent_migrate_raced,
 585
 586         &sysfs_gc_gens_pos,
 587
 588         &sysfs_copy_gc_enabled,
 589         &sysfs_copy_gc_wait,
 590
 591         &sysfs_rebalance_enabled,
 592         &sysfs_rebalance_work,
 593         sysfs_pd_controller_files(rebalance),
 594
 595         &sysfs_data_op_data_progress,
 596
 597         &sysfs_internal_uuid,
 598         NULL
 599 };
 600
 601 /* options */
 602
 603 SHOW(bch2_fs_opts_dir)
 604 {
 605         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 606         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 607         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 608         int id = opt - bch2_opt_table;
 609         u64 v = bch2_opt_get_by_id(&c->opts, id);
 610
 611         bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
 612         pr_buf(&out, "\n");
 613
 614         return out.pos - buf;
 615 }
 616
 617 STORE(bch2_fs_opts_dir)
 618 {
 619         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 620         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 621         int ret, id = opt - bch2_opt_table;
 622         char *tmp;
 623         u64 v;
 624
 625         tmp = kstrdup(buf, GFP_KERNEL);
 626         if (!tmp)
 627                 return -ENOMEM;
 628
 629         ret = bch2_opt_parse(c, opt, strim(tmp), &v);
 630         kfree(tmp);
 631
 632         if (ret < 0)
 633                 return ret;
 634
 635         ret = bch2_opt_check_may_set(c, id, v);
 636         if (ret < 0)
 637                 return ret;
 638
 639         if (opt->set_sb != SET_NO_SB_OPT) {
 640                 mutex_lock(&c->sb_lock);
 641                 opt->set_sb(c->disk_sb.sb, v);
 642                 bch2_write_super(c);
 643                 mutex_unlock(&c->sb_lock);
 644         }
 645
 646         bch2_opt_set_by_id(&c->opts, id, v);
 647
 648         if ((id == Opt_background_target ||
 649              id == Opt_background_compression) && v) {
 650                 bch2_rebalance_add_work(c, S64_MAX);
 651                 rebalance_wakeup(c);
 652         }
 653
 654         return size;
 655 }
 656 SYSFS_OPS(bch2_fs_opts_dir);
 657
 658 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 659
 660 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 661 {
 662         const struct bch_option *i;
 663         int ret;
 664
 665         for (i = bch2_opt_table;
 666              i < bch2_opt_table + bch2_opts_nr;
 667              i++) {
 668                 if (!(i->mode & OPT_FS))
 669                         continue;
 670
 671                 ret = sysfs_create_file(kobj, &i->attr);
 672                 if (ret)
 673                         return ret;
 674         }
 675
 676         return 0;
 677 }
 678
 679 /* time stats */
 680
 681 SHOW(bch2_fs_time_stats)
 682 {
 683         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 684         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 685
 686 #define x(name)                                                         \
 687         if (attr == &sysfs_time_stat_##name) {                          \
 688                 bch2_time_stats_to_text(&out, &c->times[BCH_TIME_##name]);\
 689                 return out.pos - buf;                                   \
 690         }
 691         BCH_TIME_STATS()
 692 #undef x
 693
 694         return 0;
 695 }
 696
 697 STORE(bch2_fs_time_stats)
 698 {
 699         return size;
 700 }
 701 SYSFS_OPS(bch2_fs_time_stats);
 702
 703 struct attribute *bch2_fs_time_stats_files[] = {
 704 #define x(name)                                         \
 705         &sysfs_time_stat_##name,
 706         BCH_TIME_STATS()
 707 #undef x
 708         NULL
 709 };
 710
 711 static void reserve_stats_to_text(struct printbuf *out, struct bch_dev *ca)
 712 {
 713         enum alloc_reserve i;
 714
 715         spin_lock(&ca->fs->freelist_lock);
 716
 717         pr_buf(out, "free_inc:\t%zu\t%zu\n",
 718                fifo_used(&ca->free_inc),
 719                ca->free_inc.size);
 720
 721         for (i = 0; i < RESERVE_NR; i++)
 722                 pr_buf(out, "free[%u]:\t%zu\t%zu\n", i,
 723                        fifo_used(&ca->free[i]),
 724                        ca->free[i].size);
 725
 726         spin_unlock(&ca->fs->freelist_lock);
 727 }
 728
 729 static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 730 {
 731         struct bch_fs *c = ca->fs;
 732         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 733         unsigned i, nr[BCH_DATA_NR];
 734
 735         memset(nr, 0, sizeof(nr));
 736
 737         for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 738                 nr[c->open_buckets[i].type]++;
 739
 740         pr_buf(out,
 741                "\t\t buckets\t sectors      fragmented\n"
 742                "capacity%16llu\n",
 743                ca->mi.nbuckets - ca->mi.first_bucket);
 744
 745         for (i = 1; i < BCH_DATA_NR; i++)
 746                 pr_buf(out, "%-8s%16llu%16llu%16llu\n",
 747                        bch2_data_types[i], stats.d[i].buckets,
 748                        stats.d[i].sectors, stats.d[i].fragmented);
 749
 750         pr_buf(out,
 751                "ec\t%16llu\n"
 752                "available%15llu\n"
 753                "\n"
 754                "free_inc\t\t%zu/%zu\n"
 755                "free[RESERVE_MOVINGGC]\t%zu/%zu\n"
 756                "free[RESERVE_NONE]\t%zu/%zu\n"
 757                "freelist_wait\t\t%s\n"
 758                "open buckets allocated\t%u\n"
 759                "open buckets this dev\t%u\n"
 760                "open buckets total\t%u\n"
 761                "open_buckets_wait\t%s\n"
 762                "open_buckets_btree\t%u\n"
 763                "open_buckets_user\t%u\n"
 764                "btree reserve cache\t%u\n"
 765                "thread state:\t\t%s\n",
 766                stats.buckets_ec,
 767                __dev_buckets_available(ca, stats),
 768                fifo_used(&ca->free_inc),                ca->free_inc.size,
 769                fifo_used(&ca->free[RESERVE_MOVINGGC]),  ca->free[RESERVE_MOVINGGC].size,
 770                fifo_used(&ca->free[RESERVE_NONE]),      ca->free[RESERVE_NONE].size,
 771                c->freelist_wait.list.first              ? "waiting" : "empty",
 772                OPEN_BUCKETS_COUNT - c->open_buckets_nr_free,
 773                ca->nr_open_buckets,
 774                OPEN_BUCKETS_COUNT,
 775                c->open_buckets_wait.list.first          ? "waiting" : "empty",
 776                nr[BCH_DATA_btree],
 777                nr[BCH_DATA_user],
 778                c->btree_reserve_cache_nr,
 779                bch2_allocator_states[ca->allocator_state]);
 780 }
 781
 782 static const char * const bch2_rw[] = {
 783         "read",
 784         "write",
 785         NULL
 786 };
 787
 788 static void dev_iodone_to_text(struct printbuf *out, struct bch_dev *ca)
 789 {
 790         int rw, i;
 791
 792         for (rw = 0; rw < 2; rw++) {
 793                 pr_buf(out, "%s:\n", bch2_rw[rw]);
 794
 795                 for (i = 1; i < BCH_DATA_NR; i++)
 796                         pr_buf(out, "%-12s:%12llu\n",
 797                                bch2_data_types[i],
 798                                percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
 799         }
 800 }
 801
 802 SHOW(bch2_dev)
 803 {
 804         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 805         struct bch_fs *c = ca->fs;
 806         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 807
 808         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 809
 810         sysfs_print(bucket_size,        bucket_bytes(ca));
 811         sysfs_print(first_bucket,       ca->mi.first_bucket);
 812         sysfs_print(nbuckets,           ca->mi.nbuckets);
 813         sysfs_print(durability,         ca->mi.durability);
 814         sysfs_print(discard,            ca->mi.discard);
 815
 816         if (attr == &sysfs_label) {
 817                 if (ca->mi.group) {
 818                         mutex_lock(&c->sb_lock);
 819                         bch2_disk_path_to_text(&out, &c->disk_sb,
 820                                                ca->mi.group - 1);
 821                         mutex_unlock(&c->sb_lock);
 822                 }
 823
 824                 pr_buf(&out, "\n");
 825                 return out.pos - buf;
 826         }
 827
 828         if (attr == &sysfs_has_data) {
 829                 bch2_flags_to_text(&out, bch2_data_types,
 830                                    bch2_dev_has_data(c, ca));
 831                 pr_buf(&out, "\n");
 832                 return out.pos - buf;
 833         }
 834
 835         if (attr == &sysfs_cache_replacement_policy) {
 836                 bch2_string_opt_to_text(&out,
 837                                         bch2_cache_replacement_policies,
 838                                         ca->mi.replacement);
 839                 pr_buf(&out, "\n");
 840                 return out.pos - buf;
 841         }
 842
 843         if (attr == &sysfs_state_rw) {
 844                 bch2_string_opt_to_text(&out, bch2_member_states,
 845                                         ca->mi.state);
 846                 pr_buf(&out, "\n");
 847                 return out.pos - buf;
 848         }
 849
 850         if (attr == &sysfs_iodone) {
 851                 dev_iodone_to_text(&out, ca);
 852                 return out.pos - buf;
 853         }
 854
 855         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 856         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 857
 858         if (attr == &sysfs_io_latency_stats_read) {
 859                 bch2_time_stats_to_text(&out, &ca->io_latency[READ]);
 860                 return out.pos - buf;
 861         }
 862         if (attr == &sysfs_io_latency_stats_write) {
 863                 bch2_time_stats_to_text(&out, &ca->io_latency[WRITE]);
 864                 return out.pos - buf;
 865         }
 866
 867         sysfs_printf(congested,                 "%u%%",
 868                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 869                      * 100 / CONGESTED_MAX);
 870
 871         if (attr == &sysfs_reserve_stats) {
 872                 reserve_stats_to_text(&out, ca);
 873                 return out.pos - buf;
 874         }
 875         if (attr == &sysfs_alloc_debug) {
 876                 dev_alloc_debug_to_text(&out, ca);
 877                 return out.pos - buf;
 878         }
 879
 880         return 0;
 881 }
 882
 883 STORE(bch2_dev)
 884 {
 885         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 886         struct bch_fs *c = ca->fs;
 887         struct bch_member *mi;
 888
 889         if (attr == &sysfs_discard) {
 890                 bool v = strtoul_or_return(buf);
 891
 892                 mutex_lock(&c->sb_lock);
 893                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 894
 895                 if (v != BCH_MEMBER_DISCARD(mi)) {
 896                         SET_BCH_MEMBER_DISCARD(mi, v);
 897                         bch2_write_super(c);
 898                 }
 899                 mutex_unlock(&c->sb_lock);
 900         }
 901
 902         if (attr == &sysfs_cache_replacement_policy) {
 903                 ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
 904
 905                 if (v < 0)
 906                         return v;
 907
 908                 mutex_lock(&c->sb_lock);
 909                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 910
 911                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
 912                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
 913                         bch2_write_super(c);
 914                 }
 915                 mutex_unlock(&c->sb_lock);
 916         }
 917
 918         if (attr == &sysfs_label) {
 919                 char *tmp;
 920                 int ret;
 921
 922                 tmp = kstrdup(buf, GFP_KERNEL);
 923                 if (!tmp)
 924                         return -ENOMEM;
 925
 926                 ret = bch2_dev_group_set(c, ca, strim(tmp));
 927                 kfree(tmp);
 928                 if (ret)
 929                         return ret;
 930         }
 931
 932         if (attr == &sysfs_wake_allocator)
 933                 bch2_wake_allocator(ca);
 934
 935         return size;
 936 }
 937 SYSFS_OPS(bch2_dev);
 938
 939 struct attribute *bch2_dev_files[] = {
 940         &sysfs_uuid,
 941         &sysfs_bucket_size,
 942         &sysfs_first_bucket,
 943         &sysfs_nbuckets,
 944         &sysfs_durability,
 945
 946         /* settings: */
 947         &sysfs_discard,
 948         &sysfs_cache_replacement_policy,
 949         &sysfs_state_rw,
 950         &sysfs_label,
 951
 952         &sysfs_has_data,
 953         &sysfs_iodone,
 954
 955         &sysfs_io_latency_read,
 956         &sysfs_io_latency_write,
 957         &sysfs_io_latency_stats_read,
 958         &sysfs_io_latency_stats_write,
 959         &sysfs_congested,
 960
 961         &sysfs_reserve_stats,
 962
 963         /* debug: */
 964         &sysfs_alloc_debug,
 965         &sysfs_wake_allocator,
 966         NULL
 967 };
 968
 969 #endif  /* _BCACHEFS_SYSFS_H_ */