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 fs_alloc_debug_to_text(struct printbuf *out, struct bch_fs *c)
 266 {
 267         struct bch_fs_usage_online *fs_usage = bch2_fs_usage_read(c);
 268
 269         if (!fs_usage)
 270                 return -ENOMEM;
 271
 272         bch2_fs_usage_to_text(out, c, fs_usage);
 273
 274         percpu_up_read(&c->mark_lock);
 275
 276         kfree(fs_usage);
 277         return 0;
 278 }
 279
 280 static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c)
 281 {
 282         struct btree_trans trans;
 283         struct btree_iter iter;
 284         struct bkey_s_c k;
 285         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 286             nr_compressed_extents = 0,
 287             compressed_sectors_compressed = 0,
 288             compressed_sectors_uncompressed = 0;
 289         int ret;
 290
 291         if (!test_bit(BCH_FS_STARTED, &c->flags))
 292                 return -EPERM;
 293
 294         bch2_trans_init(&trans, c, 0, 0);
 295
 296         for_each_btree_key(&trans, iter, BTREE_ID_extents, POS_MIN, 0, k, ret)
 297                 if (k.k->type == KEY_TYPE_extent) {
 298                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 299                         const union bch_extent_entry *entry;
 300                         struct extent_ptr_decoded p;
 301
 302                         extent_for_each_ptr_decode(e, p, entry) {
 303                                 if (!crc_is_compressed(p.crc)) {
 304                                         nr_uncompressed_extents++;
 305                                         uncompressed_sectors += e.k->size;
 306                                 } else {
 307                                         nr_compressed_extents++;
 308                                         compressed_sectors_compressed +=
 309                                                 p.crc.compressed_size;
 310                                         compressed_sectors_uncompressed +=
 311                                                 p.crc.uncompressed_size;
 312                                 }
 313
 314                                 /* only looking at the first ptr */
 315                                 break;
 316                         }
 317                 }
 318         bch2_trans_iter_exit(&trans, &iter);
 319
 320         bch2_trans_exit(&trans);
 321         if (ret)
 322                 return ret;
 323
 324         pr_buf(out,
 325                "uncompressed data:\n"
 326                "        nr extents:                     %llu\n"
 327                "        size (bytes):                   %llu\n"
 328                "compressed data:\n"
 329                "        nr extents:                     %llu\n"
 330                "        compressed size (bytes):        %llu\n"
 331                "        uncompressed size (bytes):      %llu\n",
 332                nr_uncompressed_extents,
 333                uncompressed_sectors << 9,
 334                nr_compressed_extents,
 335                compressed_sectors_compressed << 9,
 336                compressed_sectors_uncompressed << 9);
 337         return 0;
 338 }
 339
 340 static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
 341 {
 342         pr_buf(out, "%s: ", bch2_btree_ids[c->gc_gens_btree]);
 343         bch2_bpos_to_text(out, c->gc_gens_pos);
 344         pr_buf(out, "\n");
 345 }
 346
 347 SHOW(bch2_fs)
 348 {
 349         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 350         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 351
 352         sysfs_print(minor,                      c->minor);
 353         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 354
 355         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 356         sysfs_hprint(btree_avg_write_size,      bch2_btree_avg_write_size(c));
 357
 358         sysfs_print(read_realloc_races,
 359                     atomic_long_read(&c->read_realloc_races));
 360         sysfs_print(extent_migrate_done,
 361                     atomic_long_read(&c->extent_migrate_done));
 362         sysfs_print(extent_migrate_raced,
 363                     atomic_long_read(&c->extent_migrate_raced));
 364
 365         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 366
 367         if (attr == &sysfs_gc_gens_pos) {
 368                 bch2_gc_gens_pos_to_text(&out, c);
 369                 return out.pos - buf;
 370         }
 371
 372         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 373
 374         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 375         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 376         sysfs_hprint(copy_gc_wait,
 377                      max(0LL, c->copygc_wait -
 378                          atomic64_read(&c->io_clock[WRITE].now)) << 9);
 379
 380         if (attr == &sysfs_rebalance_work) {
 381                 bch2_rebalance_work_to_text(&out, c);
 382                 return out.pos - buf;
 383         }
 384
 385         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 386
 387         /* Debugging: */
 388
 389         if (attr == &sysfs_alloc_debug)
 390                 return fs_alloc_debug_to_text(&out, c) ?: out.pos - buf;
 391
 392         if (attr == &sysfs_journal_debug) {
 393                 bch2_journal_debug_to_text(&out, &c->journal);
 394                 return out.pos - buf;
 395         }
 396
 397         if (attr == &sysfs_journal_pins) {
 398                 bch2_journal_pins_to_text(&out, &c->journal);
 399                 return out.pos - buf;
 400         }
 401
 402         if (attr == &sysfs_btree_updates) {
 403                 bch2_btree_updates_to_text(&out, c);
 404                 return out.pos - buf;
 405         }
 406
 407         if (attr == &sysfs_dirty_btree_nodes) {
 408                 bch2_dirty_btree_nodes_to_text(&out, c);
 409                 return out.pos - buf;
 410         }
 411
 412         if (attr == &sysfs_btree_cache) {
 413                 bch2_btree_cache_to_text(&out, c);
 414                 return out.pos - buf;
 415         }
 416
 417         if (attr == &sysfs_btree_key_cache) {
 418                 bch2_btree_key_cache_to_text(&out, &c->btree_key_cache);
 419                 return out.pos - buf;
 420         }
 421
 422         if (attr == &sysfs_btree_transactions) {
 423                 bch2_btree_trans_to_text(&out, c);
 424                 return out.pos - buf;
 425         }
 426
 427         if (attr == &sysfs_stripes_heap) {
 428                 bch2_stripes_heap_to_text(&out, c);
 429                 return out.pos - buf;
 430         }
 431
 432         if (attr == &sysfs_open_buckets) {
 433                 bch2_open_buckets_to_text(&out, c);
 434                 return out.pos - buf;
 435         }
 436
 437         if (attr == &sysfs_compression_stats) {
 438                 bch2_compression_stats_to_text(&out, c);
 439                 return out.pos - buf;
 440         }
 441
 442         if (attr == &sysfs_new_stripes) {
 443                 bch2_new_stripes_to_text(&out, c);
 444                 return out.pos - buf;
 445         }
 446
 447         if (attr == &sysfs_io_timers_read) {
 448                 bch2_io_timers_to_text(&out, &c->io_clock[READ]);
 449                 return out.pos - buf;
 450         }
 451         if (attr == &sysfs_io_timers_write) {
 452                 bch2_io_timers_to_text(&out, &c->io_clock[WRITE]);
 453                 return out.pos - buf;
 454         }
 455
 456         if (attr == &sysfs_data_op_data_progress) {
 457                 data_progress_to_text(&out, c);
 458                 return out.pos - buf;
 459         }
 460
 461         return 0;
 462 }
 463
 464 STORE(bch2_fs)
 465 {
 466         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 467
 468         if (attr == &sysfs_btree_gc_periodic) {
 469                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 470                         ?: (ssize_t) size;
 471
 472                 wake_up_process(c->gc_thread);
 473                 return ret;
 474         }
 475
 476         if (attr == &sysfs_copy_gc_enabled) {
 477                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 478                         ?: (ssize_t) size;
 479
 480                 if (c->copygc_thread)
 481                         wake_up_process(c->copygc_thread);
 482                 return ret;
 483         }
 484
 485         if (attr == &sysfs_rebalance_enabled) {
 486                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 487                         ?: (ssize_t) size;
 488
 489                 rebalance_wakeup(c);
 490                 return ret;
 491         }
 492
 493         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 494
 495         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 496
 497         /* Debugging: */
 498
 499         if (!test_bit(BCH_FS_STARTED, &c->flags))
 500                 return -EPERM;
 501
 502         /* Debugging: */
 503
 504         if (attr == &sysfs_trigger_journal_flush)
 505                 bch2_journal_meta(&c->journal);
 506
 507         if (attr == &sysfs_trigger_gc) {
 508                 /*
 509                  * Full gc is currently incompatible with btree key cache:
 510                  */
 511 #if 0
 512                 down_read(&c->state_lock);
 513                 bch2_gc(c, false, false);
 514                 up_read(&c->state_lock);
 515 #else
 516                 bch2_gc_gens(c);
 517 #endif
 518         }
 519
 520         if (attr == &sysfs_prune_cache) {
 521                 struct shrink_control sc;
 522
 523                 sc.gfp_mask = GFP_KERNEL;
 524                 sc.nr_to_scan = strtoul_or_return(buf);
 525                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 526         }
 527
 528 #ifdef CONFIG_BCACHEFS_TESTS
 529         if (attr == &sysfs_perf_test) {
 530                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 531                 char *test              = strsep(&p, " \t\n");
 532                 char *nr_str            = strsep(&p, " \t\n");
 533                 char *threads_str       = strsep(&p, " \t\n");
 534                 unsigned threads;
 535                 u64 nr;
 536                 int ret = -EINVAL;
 537
 538                 if (threads_str &&
 539                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 540                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 541                         ret = bch2_btree_perf_test(c, test, nr, threads);
 542                 kfree(tmp);
 543
 544                 if (ret)
 545                         size = ret;
 546         }
 547 #endif
 548         return size;
 549 }
 550 SYSFS_OPS(bch2_fs);
 551
 552 struct attribute *bch2_fs_files[] = {
 553         &sysfs_minor,
 554         &sysfs_btree_cache_size,
 555         &sysfs_btree_avg_write_size,
 556
 557         &sysfs_promote_whole_extents,
 558
 559         &sysfs_compression_stats,
 560
 561 #ifdef CONFIG_BCACHEFS_TESTS
 562         &sysfs_perf_test,
 563 #endif
 564         NULL
 565 };
 566
 567 /* internal dir - just a wrapper */
 568
 569 SHOW(bch2_fs_internal)
 570 {
 571         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 572         return bch2_fs_show(&c->kobj, attr, buf);
 573 }
 574
 575 STORE(bch2_fs_internal)
 576 {
 577         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 578         return bch2_fs_store(&c->kobj, attr, buf, size);
 579 }
 580 SYSFS_OPS(bch2_fs_internal);
 581
 582 struct attribute *bch2_fs_internal_files[] = {
 583         &sysfs_alloc_debug,
 584         &sysfs_journal_debug,
 585         &sysfs_journal_pins,
 586         &sysfs_btree_updates,
 587         &sysfs_dirty_btree_nodes,
 588         &sysfs_btree_cache,
 589         &sysfs_btree_key_cache,
 590         &sysfs_btree_transactions,
 591         &sysfs_stripes_heap,
 592         &sysfs_open_buckets,
 593
 594         &sysfs_read_realloc_races,
 595         &sysfs_extent_migrate_done,
 596         &sysfs_extent_migrate_raced,
 597
 598         &sysfs_trigger_journal_flush,
 599         &sysfs_trigger_gc,
 600         &sysfs_gc_gens_pos,
 601         &sysfs_prune_cache,
 602
 603         &sysfs_copy_gc_enabled,
 604         &sysfs_copy_gc_wait,
 605
 606         &sysfs_rebalance_enabled,
 607         &sysfs_rebalance_work,
 608         sysfs_pd_controller_files(rebalance),
 609
 610         &sysfs_new_stripes,
 611
 612         &sysfs_io_timers_read,
 613         &sysfs_io_timers_write,
 614
 615         &sysfs_data_op_data_progress,
 616
 617         &sysfs_internal_uuid,
 618         NULL
 619 };
 620
 621 /* options */
 622
 623 SHOW(bch2_fs_opts_dir)
 624 {
 625         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 626         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 627         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 628         int id = opt - bch2_opt_table;
 629         u64 v = bch2_opt_get_by_id(&c->opts, id);
 630
 631         bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
 632         pr_buf(&out, "\n");
 633
 634         return out.pos - buf;
 635 }
 636
 637 STORE(bch2_fs_opts_dir)
 638 {
 639         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 640         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 641         int ret, id = opt - bch2_opt_table;
 642         char *tmp;
 643         u64 v;
 644
 645         tmp = kstrdup(buf, GFP_KERNEL);
 646         if (!tmp)
 647                 return -ENOMEM;
 648
 649         ret = bch2_opt_parse(c, opt, strim(tmp), &v);
 650         kfree(tmp);
 651
 652         if (ret < 0)
 653                 return ret;
 654
 655         ret = bch2_opt_check_may_set(c, id, v);
 656         if (ret < 0)
 657                 return ret;
 658
 659         if (opt->set_sb != SET_NO_SB_OPT) {
 660                 mutex_lock(&c->sb_lock);
 661                 opt->set_sb(c->disk_sb.sb, v);
 662                 bch2_write_super(c);
 663                 mutex_unlock(&c->sb_lock);
 664         }
 665
 666         bch2_opt_set_by_id(&c->opts, id, v);
 667
 668         if ((id == Opt_background_target ||
 669              id == Opt_background_compression) && v) {
 670                 bch2_rebalance_add_work(c, S64_MAX);
 671                 rebalance_wakeup(c);
 672         }
 673
 674         return size;
 675 }
 676 SYSFS_OPS(bch2_fs_opts_dir);
 677
 678 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 679
 680 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 681 {
 682         const struct bch_option *i;
 683         int ret;
 684
 685         for (i = bch2_opt_table;
 686              i < bch2_opt_table + bch2_opts_nr;
 687              i++) {
 688                 if (!(i->mode & OPT_FS))
 689                         continue;
 690
 691                 ret = sysfs_create_file(kobj, &i->attr);
 692                 if (ret)
 693                         return ret;
 694         }
 695
 696         return 0;
 697 }
 698
 699 /* time stats */
 700
 701 SHOW(bch2_fs_time_stats)
 702 {
 703         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 704         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 705
 706 #define x(name)                                                         \
 707         if (attr == &sysfs_time_stat_##name) {                          \
 708                 bch2_time_stats_to_text(&out, &c->times[BCH_TIME_##name]);\
 709                 return out.pos - buf;                                   \
 710         }
 711         BCH_TIME_STATS()
 712 #undef x
 713
 714         return 0;
 715 }
 716
 717 STORE(bch2_fs_time_stats)
 718 {
 719         return size;
 720 }
 721 SYSFS_OPS(bch2_fs_time_stats);
 722
 723 struct attribute *bch2_fs_time_stats_files[] = {
 724 #define x(name)                                         \
 725         &sysfs_time_stat_##name,
 726         BCH_TIME_STATS()
 727 #undef x
 728         NULL
 729 };
 730
 731 static void reserve_stats_to_text(struct printbuf *out, struct bch_dev *ca)
 732 {
 733         enum alloc_reserve i;
 734
 735         spin_lock(&ca->fs->freelist_lock);
 736
 737         pr_buf(out, "free_inc:\t%zu\t%zu\n",
 738                fifo_used(&ca->free_inc),
 739                ca->free_inc.size);
 740
 741         for (i = 0; i < RESERVE_NR; i++)
 742                 pr_buf(out, "free[%u]:\t%zu\t%zu\n", i,
 743                        fifo_used(&ca->free[i]),
 744                        ca->free[i].size);
 745
 746         spin_unlock(&ca->fs->freelist_lock);
 747 }
 748
 749 static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 750 {
 751         struct bch_fs *c = ca->fs;
 752         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 753         unsigned i, nr[BCH_DATA_NR];
 754
 755         memset(nr, 0, sizeof(nr));
 756
 757         for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 758                 nr[c->open_buckets[i].type]++;
 759
 760         pr_buf(out,
 761                "\t\t buckets\t sectors      fragmented\n"
 762                "capacity%16llu\n",
 763                ca->mi.nbuckets - ca->mi.first_bucket);
 764
 765         for (i = 1; i < BCH_DATA_NR; i++)
 766                 pr_buf(out, "%-8s%16llu%16llu%16llu\n",
 767                        bch2_data_types[i], stats.d[i].buckets,
 768                        stats.d[i].sectors, stats.d[i].fragmented);
 769
 770         pr_buf(out,
 771                "ec\t%16llu\n"
 772                "available%15llu\n"
 773                "\n"
 774                "free_inc\t\t%zu/%zu\n"
 775                "free[RESERVE_MOVINGGC]\t%zu/%zu\n"
 776                "free[RESERVE_NONE]\t%zu/%zu\n"
 777                "freelist_wait\t\t%s\n"
 778                "open buckets allocated\t%u\n"
 779                "open buckets this dev\t%u\n"
 780                "open buckets total\t%u\n"
 781                "open_buckets_wait\t%s\n"
 782                "open_buckets_btree\t%u\n"
 783                "open_buckets_user\t%u\n"
 784                "btree reserve cache\t%u\n"
 785                "thread state:\t\t%s\n",
 786                stats.buckets_ec,
 787                __dev_buckets_available(ca, stats),
 788                fifo_used(&ca->free_inc),                ca->free_inc.size,
 789                fifo_used(&ca->free[RESERVE_MOVINGGC]),  ca->free[RESERVE_MOVINGGC].size,
 790                fifo_used(&ca->free[RESERVE_NONE]),      ca->free[RESERVE_NONE].size,
 791                c->freelist_wait.list.first              ? "waiting" : "empty",
 792                OPEN_BUCKETS_COUNT - c->open_buckets_nr_free,
 793                ca->nr_open_buckets,
 794                OPEN_BUCKETS_COUNT,
 795                c->open_buckets_wait.list.first          ? "waiting" : "empty",
 796                nr[BCH_DATA_btree],
 797                nr[BCH_DATA_user],
 798                c->btree_reserve_cache_nr,
 799                bch2_allocator_states[ca->allocator_state]);
 800 }
 801
 802 static const char * const bch2_rw[] = {
 803         "read",
 804         "write",
 805         NULL
 806 };
 807
 808 static void dev_iodone_to_text(struct printbuf *out, struct bch_dev *ca)
 809 {
 810         int rw, i;
 811
 812         for (rw = 0; rw < 2; rw++) {
 813                 pr_buf(out, "%s:\n", bch2_rw[rw]);
 814
 815                 for (i = 1; i < BCH_DATA_NR; i++)
 816                         pr_buf(out, "%-12s:%12llu\n",
 817                                bch2_data_types[i],
 818                                percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
 819         }
 820 }
 821
 822 SHOW(bch2_dev)
 823 {
 824         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 825         struct bch_fs *c = ca->fs;
 826         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 827
 828         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 829
 830         sysfs_print(bucket_size,        bucket_bytes(ca));
 831         sysfs_print(first_bucket,       ca->mi.first_bucket);
 832         sysfs_print(nbuckets,           ca->mi.nbuckets);
 833         sysfs_print(durability,         ca->mi.durability);
 834         sysfs_print(discard,            ca->mi.discard);
 835
 836         if (attr == &sysfs_label) {
 837                 if (ca->mi.group) {
 838                         mutex_lock(&c->sb_lock);
 839                         bch2_disk_path_to_text(&out, &c->disk_sb,
 840                                                ca->mi.group - 1);
 841                         mutex_unlock(&c->sb_lock);
 842                 }
 843
 844                 pr_buf(&out, "\n");
 845                 return out.pos - buf;
 846         }
 847
 848         if (attr == &sysfs_has_data) {
 849                 bch2_flags_to_text(&out, bch2_data_types,
 850                                    bch2_dev_has_data(c, ca));
 851                 pr_buf(&out, "\n");
 852                 return out.pos - buf;
 853         }
 854
 855         if (attr == &sysfs_cache_replacement_policy) {
 856                 bch2_string_opt_to_text(&out,
 857                                         bch2_cache_replacement_policies,
 858                                         ca->mi.replacement);
 859                 pr_buf(&out, "\n");
 860                 return out.pos - buf;
 861         }
 862
 863         if (attr == &sysfs_state_rw) {
 864                 bch2_string_opt_to_text(&out, bch2_member_states,
 865                                         ca->mi.state);
 866                 pr_buf(&out, "\n");
 867                 return out.pos - buf;
 868         }
 869
 870         if (attr == &sysfs_iodone) {
 871                 dev_iodone_to_text(&out, ca);
 872                 return out.pos - buf;
 873         }
 874
 875         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 876         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 877
 878         if (attr == &sysfs_io_latency_stats_read) {
 879                 bch2_time_stats_to_text(&out, &ca->io_latency[READ]);
 880                 return out.pos - buf;
 881         }
 882         if (attr == &sysfs_io_latency_stats_write) {
 883                 bch2_time_stats_to_text(&out, &ca->io_latency[WRITE]);
 884                 return out.pos - buf;
 885         }
 886
 887         sysfs_printf(congested,                 "%u%%",
 888                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 889                      * 100 / CONGESTED_MAX);
 890
 891         if (attr == &sysfs_reserve_stats) {
 892                 reserve_stats_to_text(&out, ca);
 893                 return out.pos - buf;
 894         }
 895         if (attr == &sysfs_alloc_debug) {
 896                 dev_alloc_debug_to_text(&out, ca);
 897                 return out.pos - buf;
 898         }
 899
 900         return 0;
 901 }
 902
 903 STORE(bch2_dev)
 904 {
 905         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 906         struct bch_fs *c = ca->fs;
 907         struct bch_member *mi;
 908
 909         if (attr == &sysfs_discard) {
 910                 bool v = strtoul_or_return(buf);
 911
 912                 mutex_lock(&c->sb_lock);
 913                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 914
 915                 if (v != BCH_MEMBER_DISCARD(mi)) {
 916                         SET_BCH_MEMBER_DISCARD(mi, v);
 917                         bch2_write_super(c);
 918                 }
 919                 mutex_unlock(&c->sb_lock);
 920         }
 921
 922         if (attr == &sysfs_cache_replacement_policy) {
 923                 ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
 924
 925                 if (v < 0)
 926                         return v;
 927
 928                 mutex_lock(&c->sb_lock);
 929                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 930
 931                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
 932                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
 933                         bch2_write_super(c);
 934                 }
 935                 mutex_unlock(&c->sb_lock);
 936         }
 937
 938         if (attr == &sysfs_label) {
 939                 char *tmp;
 940                 int ret;
 941
 942                 tmp = kstrdup(buf, GFP_KERNEL);
 943                 if (!tmp)
 944                         return -ENOMEM;
 945
 946                 ret = bch2_dev_group_set(c, ca, strim(tmp));
 947                 kfree(tmp);
 948                 if (ret)
 949                         return ret;
 950         }
 951
 952         if (attr == &sysfs_wake_allocator)
 953                 bch2_wake_allocator(ca);
 954
 955         return size;
 956 }
 957 SYSFS_OPS(bch2_dev);
 958
 959 struct attribute *bch2_dev_files[] = {
 960         &sysfs_uuid,
 961         &sysfs_bucket_size,
 962         &sysfs_first_bucket,
 963         &sysfs_nbuckets,
 964         &sysfs_durability,
 965
 966         /* settings: */
 967         &sysfs_discard,
 968         &sysfs_cache_replacement_policy,
 969         &sysfs_state_rw,
 970         &sysfs_label,
 971
 972         &sysfs_has_data,
 973         &sysfs_iodone,
 974
 975         &sysfs_io_latency_read,
 976         &sysfs_io_latency_write,
 977         &sysfs_io_latency_stats_read,
 978         &sysfs_io_latency_stats_write,
 979         &sysfs_congested,
 980
 981         &sysfs_reserve_stats,
 982
 983         /* debug: */
 984         &sysfs_alloc_debug,
 985         &sysfs_wake_allocator,
 986         NULL
 987 };
 988
 989 #endif  /* _BCACHEFS_SYSFS_H_ */