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(block_size);
 144 read_attribute(btree_node_size);
 145 read_attribute(first_bucket);
 146 read_attribute(nbuckets);
 147 read_attribute(durability);
 148 read_attribute(iodone);
 149
 150 read_attribute(io_latency_read);
 151 read_attribute(io_latency_write);
 152 read_attribute(io_latency_stats_read);
 153 read_attribute(io_latency_stats_write);
 154 read_attribute(congested);
 155
 156 read_attribute(btree_avg_write_size);
 157
 158 read_attribute(reserve_stats);
 159 read_attribute(btree_cache_size);
 160 read_attribute(compression_stats);
 161 read_attribute(journal_debug);
 162 read_attribute(journal_pins);
 163 read_attribute(btree_updates);
 164 read_attribute(dirty_btree_nodes);
 165 read_attribute(btree_cache);
 166 read_attribute(btree_key_cache);
 167 read_attribute(btree_transactions);
 168 read_attribute(stripes_heap);
 169 read_attribute(open_buckets);
 170
 171 read_attribute(internal_uuid);
 172
 173 read_attribute(has_data);
 174 read_attribute(alloc_debug);
 175 write_attribute(wake_allocator);
 176
 177 read_attribute(read_realloc_races);
 178 read_attribute(extent_migrate_done);
 179 read_attribute(extent_migrate_raced);
 180
 181 rw_attribute(journal_write_delay_ms);
 182 rw_attribute(journal_reclaim_delay_ms);
 183
 184 rw_attribute(discard);
 185 rw_attribute(cache_replacement_policy);
 186 rw_attribute(label);
 187
 188 rw_attribute(copy_gc_enabled);
 189 read_attribute(copy_gc_wait);
 190
 191 rw_attribute(rebalance_enabled);
 192 sysfs_pd_controller_attribute(rebalance);
 193 read_attribute(rebalance_work);
 194 rw_attribute(promote_whole_extents);
 195
 196 read_attribute(new_stripes);
 197
 198 read_attribute(io_timers_read);
 199 read_attribute(io_timers_write);
 200
 201 read_attribute(data_op_data_progress);
 202
 203 #ifdef CONFIG_BCACHEFS_TESTS
 204 write_attribute(perf_test);
 205 #endif /* CONFIG_BCACHEFS_TESTS */
 206
 207 #define x(_name)                                                \
 208         static struct attribute sysfs_time_stat_##_name =               \
 209                 { .name = #_name, .mode = S_IRUGO };
 210         BCH_TIME_STATS()
 211 #undef x
 212
 213 static struct attribute sysfs_state_rw = {
 214         .name = "state",
 215         .mode = S_IRUGO
 216 };
 217
 218 static size_t bch2_btree_cache_size(struct bch_fs *c)
 219 {
 220         size_t ret = 0;
 221         struct btree *b;
 222
 223         mutex_lock(&c->btree_cache.lock);
 224         list_for_each_entry(b, &c->btree_cache.live, list)
 225                 ret += btree_bytes(c);
 226
 227         mutex_unlock(&c->btree_cache.lock);
 228         return ret;
 229 }
 230
 231 static size_t bch2_btree_avg_write_size(struct bch_fs *c)
 232 {
 233         u64 nr = atomic64_read(&c->btree_writes_nr);
 234         u64 sectors = atomic64_read(&c->btree_writes_sectors);
 235
 236         return nr ? div64_u64(sectors, nr) : 0;
 237 }
 238
 239 static long stats_to_text(struct printbuf *out, struct bch_fs *c,
 240                           struct bch_move_stats *stats)
 241 {
 242         pr_buf(out, "%s: data type %s btree_id %s position: ",
 243                 stats->name,
 244                 bch2_data_types[stats->data_type],
 245                 bch2_btree_ids[stats->btree_id]);
 246         bch2_bpos_to_text(out, stats->pos);
 247         pr_buf(out, "%s", "\n");
 248
 249         return 0;
 250 }
 251
 252 static long data_progress_to_text(struct printbuf *out, struct bch_fs *c)
 253 {
 254         long ret = 0;
 255         struct bch_move_stats *iter;
 256
 257         mutex_lock(&c->data_progress_lock);
 258
 259         if (list_empty(&c->data_progress_list))
 260                 pr_buf(out, "%s", "no progress to report\n");
 261         else
 262                 list_for_each_entry(iter, &c->data_progress_list, list) {
 263                         stats_to_text(out, c, iter);
 264                 }
 265
 266         mutex_unlock(&c->data_progress_lock);
 267         return ret;
 268 }
 269
 270 static int fs_alloc_debug_to_text(struct printbuf *out, struct bch_fs *c)
 271 {
 272         struct bch_fs_usage_online *fs_usage = bch2_fs_usage_read(c);
 273
 274         if (!fs_usage)
 275                 return -ENOMEM;
 276
 277         bch2_fs_usage_to_text(out, c, fs_usage);
 278
 279         percpu_up_read(&c->mark_lock);
 280
 281         kfree(fs_usage);
 282         return 0;
 283 }
 284
 285 static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c)
 286 {
 287         struct btree_trans trans;
 288         struct btree_iter iter;
 289         struct bkey_s_c k;
 290         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 291             nr_compressed_extents = 0,
 292             compressed_sectors_compressed = 0,
 293             compressed_sectors_uncompressed = 0;
 294         int ret;
 295
 296         if (!test_bit(BCH_FS_STARTED, &c->flags))
 297                 return -EPERM;
 298
 299         bch2_trans_init(&trans, c, 0, 0);
 300
 301         for_each_btree_key(&trans, iter, BTREE_ID_extents, POS_MIN, 0, k, ret)
 302                 if (k.k->type == KEY_TYPE_extent) {
 303                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 304                         const union bch_extent_entry *entry;
 305                         struct extent_ptr_decoded p;
 306
 307                         extent_for_each_ptr_decode(e, p, entry) {
 308                                 if (!crc_is_compressed(p.crc)) {
 309                                         nr_uncompressed_extents++;
 310                                         uncompressed_sectors += e.k->size;
 311                                 } else {
 312                                         nr_compressed_extents++;
 313                                         compressed_sectors_compressed +=
 314                                                 p.crc.compressed_size;
 315                                         compressed_sectors_uncompressed +=
 316                                                 p.crc.uncompressed_size;
 317                                 }
 318
 319                                 /* only looking at the first ptr */
 320                                 break;
 321                         }
 322                 }
 323         bch2_trans_iter_exit(&trans, &iter);
 324
 325         bch2_trans_exit(&trans);
 326         if (ret)
 327                 return ret;
 328
 329         pr_buf(out,
 330                "uncompressed data:\n"
 331                "        nr extents:                     %llu\n"
 332                "        size (bytes):                   %llu\n"
 333                "compressed data:\n"
 334                "        nr extents:                     %llu\n"
 335                "        compressed size (bytes):        %llu\n"
 336                "        uncompressed size (bytes):      %llu\n",
 337                nr_uncompressed_extents,
 338                uncompressed_sectors << 9,
 339                nr_compressed_extents,
 340                compressed_sectors_compressed << 9,
 341                compressed_sectors_uncompressed << 9);
 342         return 0;
 343 }
 344
 345 static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
 346 {
 347         pr_buf(out, "%s: ", bch2_btree_ids[c->gc_gens_btree]);
 348         bch2_bpos_to_text(out, c->gc_gens_pos);
 349         pr_buf(out, "\n");
 350 }
 351
 352 SHOW(bch2_fs)
 353 {
 354         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 355         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 356
 357         sysfs_print(minor,                      c->minor);
 358         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 359
 360         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 361         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 362
 363         sysfs_print(block_size,                 block_bytes(c));
 364         sysfs_print(btree_node_size,            btree_bytes(c));
 365         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 366         sysfs_hprint(btree_avg_write_size,      bch2_btree_avg_write_size(c));
 367
 368         sysfs_print(read_realloc_races,
 369                     atomic_long_read(&c->read_realloc_races));
 370         sysfs_print(extent_migrate_done,
 371                     atomic_long_read(&c->extent_migrate_done));
 372         sysfs_print(extent_migrate_raced,
 373                     atomic_long_read(&c->extent_migrate_raced));
 374
 375         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 376
 377         if (attr == &sysfs_gc_gens_pos) {
 378                 bch2_gc_gens_pos_to_text(&out, c);
 379                 return out.pos - buf;
 380         }
 381
 382         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 383
 384         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 385         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 386         sysfs_hprint(copy_gc_wait,
 387                      max(0LL, c->copygc_wait -
 388                          atomic64_read(&c->io_clock[WRITE].now)) << 9);
 389
 390         if (attr == &sysfs_rebalance_work) {
 391                 bch2_rebalance_work_to_text(&out, c);
 392                 return out.pos - buf;
 393         }
 394
 395         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 396
 397         /* Debugging: */
 398
 399         if (attr == &sysfs_alloc_debug)
 400                 return fs_alloc_debug_to_text(&out, c) ?: out.pos - buf;
 401
 402         if (attr == &sysfs_journal_debug) {
 403                 bch2_journal_debug_to_text(&out, &c->journal);
 404                 return out.pos - buf;
 405         }
 406
 407         if (attr == &sysfs_journal_pins) {
 408                 bch2_journal_pins_to_text(&out, &c->journal);
 409                 return out.pos - buf;
 410         }
 411
 412         if (attr == &sysfs_btree_updates) {
 413                 bch2_btree_updates_to_text(&out, c);
 414                 return out.pos - buf;
 415         }
 416
 417         if (attr == &sysfs_dirty_btree_nodes) {
 418                 bch2_dirty_btree_nodes_to_text(&out, c);
 419                 return out.pos - buf;
 420         }
 421
 422         if (attr == &sysfs_btree_cache) {
 423                 bch2_btree_cache_to_text(&out, c);
 424                 return out.pos - buf;
 425         }
 426
 427         if (attr == &sysfs_btree_key_cache) {
 428                 bch2_btree_key_cache_to_text(&out, &c->btree_key_cache);
 429                 return out.pos - buf;
 430         }
 431
 432         if (attr == &sysfs_btree_transactions) {
 433                 bch2_btree_trans_to_text(&out, c);
 434                 return out.pos - buf;
 435         }
 436
 437         if (attr == &sysfs_stripes_heap) {
 438                 bch2_stripes_heap_to_text(&out, c);
 439                 return out.pos - buf;
 440         }
 441
 442         if (attr == &sysfs_open_buckets) {
 443                 bch2_open_buckets_to_text(&out, c);
 444                 return out.pos - buf;
 445         }
 446
 447         if (attr == &sysfs_compression_stats) {
 448                 bch2_compression_stats_to_text(&out, c);
 449                 return out.pos - buf;
 450         }
 451
 452         if (attr == &sysfs_new_stripes) {
 453                 bch2_new_stripes_to_text(&out, c);
 454                 return out.pos - buf;
 455         }
 456
 457         if (attr == &sysfs_io_timers_read) {
 458                 bch2_io_timers_to_text(&out, &c->io_clock[READ]);
 459                 return out.pos - buf;
 460         }
 461         if (attr == &sysfs_io_timers_write) {
 462                 bch2_io_timers_to_text(&out, &c->io_clock[WRITE]);
 463                 return out.pos - buf;
 464         }
 465
 466         if (attr == &sysfs_data_op_data_progress) {
 467                 data_progress_to_text(&out, c);
 468                 return out.pos - buf;
 469         }
 470
 471         return 0;
 472 }
 473
 474 STORE(bch2_fs)
 475 {
 476         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 477
 478         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 479         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 480
 481         if (attr == &sysfs_btree_gc_periodic) {
 482                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 483                         ?: (ssize_t) size;
 484
 485                 wake_up_process(c->gc_thread);
 486                 return ret;
 487         }
 488
 489         if (attr == &sysfs_copy_gc_enabled) {
 490                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 491                         ?: (ssize_t) size;
 492
 493                 if (c->copygc_thread)
 494                         wake_up_process(c->copygc_thread);
 495                 return ret;
 496         }
 497
 498         if (attr == &sysfs_rebalance_enabled) {
 499                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 500                         ?: (ssize_t) size;
 501
 502                 rebalance_wakeup(c);
 503                 return ret;
 504         }
 505
 506         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 507
 508         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 509
 510         /* Debugging: */
 511
 512         if (!test_bit(BCH_FS_STARTED, &c->flags))
 513                 return -EPERM;
 514
 515         /* Debugging: */
 516
 517         if (attr == &sysfs_trigger_journal_flush)
 518                 bch2_journal_meta(&c->journal);
 519
 520         if (attr == &sysfs_trigger_gc) {
 521                 /*
 522                  * Full gc is currently incompatible with btree key cache:
 523                  */
 524 #if 0
 525                 down_read(&c->state_lock);
 526                 bch2_gc(c, false, false);
 527                 up_read(&c->state_lock);
 528 #else
 529                 bch2_gc_gens(c);
 530 #endif
 531         }
 532
 533         if (attr == &sysfs_prune_cache) {
 534                 struct shrink_control sc;
 535
 536                 sc.gfp_mask = GFP_KERNEL;
 537                 sc.nr_to_scan = strtoul_or_return(buf);
 538                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 539         }
 540
 541 #ifdef CONFIG_BCACHEFS_TESTS
 542         if (attr == &sysfs_perf_test) {
 543                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 544                 char *test              = strsep(&p, " \t\n");
 545                 char *nr_str            = strsep(&p, " \t\n");
 546                 char *threads_str       = strsep(&p, " \t\n");
 547                 unsigned threads;
 548                 u64 nr;
 549                 int ret = -EINVAL;
 550
 551                 if (threads_str &&
 552                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 553                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 554                         ret = bch2_btree_perf_test(c, test, nr, threads);
 555                 kfree(tmp);
 556
 557                 if (ret)
 558                         size = ret;
 559         }
 560 #endif
 561         return size;
 562 }
 563 SYSFS_OPS(bch2_fs);
 564
 565 struct attribute *bch2_fs_files[] = {
 566         &sysfs_minor,
 567         &sysfs_block_size,
 568         &sysfs_btree_node_size,
 569         &sysfs_btree_cache_size,
 570         &sysfs_btree_avg_write_size,
 571
 572         &sysfs_journal_write_delay_ms,
 573         &sysfs_journal_reclaim_delay_ms,
 574
 575         &sysfs_promote_whole_extents,
 576
 577         &sysfs_compression_stats,
 578
 579 #ifdef CONFIG_BCACHEFS_TESTS
 580         &sysfs_perf_test,
 581 #endif
 582         NULL
 583 };
 584
 585 /* internal dir - just a wrapper */
 586
 587 SHOW(bch2_fs_internal)
 588 {
 589         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 590         return bch2_fs_show(&c->kobj, attr, buf);
 591 }
 592
 593 STORE(bch2_fs_internal)
 594 {
 595         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 596         return bch2_fs_store(&c->kobj, attr, buf, size);
 597 }
 598 SYSFS_OPS(bch2_fs_internal);
 599
 600 struct attribute *bch2_fs_internal_files[] = {
 601         &sysfs_alloc_debug,
 602         &sysfs_journal_debug,
 603         &sysfs_journal_pins,
 604         &sysfs_btree_updates,
 605         &sysfs_dirty_btree_nodes,
 606         &sysfs_btree_cache,
 607         &sysfs_btree_key_cache,
 608         &sysfs_btree_transactions,
 609         &sysfs_stripes_heap,
 610         &sysfs_open_buckets,
 611
 612         &sysfs_read_realloc_races,
 613         &sysfs_extent_migrate_done,
 614         &sysfs_extent_migrate_raced,
 615
 616         &sysfs_trigger_journal_flush,
 617         &sysfs_trigger_gc,
 618         &sysfs_gc_gens_pos,
 619         &sysfs_prune_cache,
 620
 621         &sysfs_copy_gc_enabled,
 622         &sysfs_copy_gc_wait,
 623
 624         &sysfs_rebalance_enabled,
 625         &sysfs_rebalance_work,
 626         sysfs_pd_controller_files(rebalance),
 627
 628         &sysfs_new_stripes,
 629
 630         &sysfs_io_timers_read,
 631         &sysfs_io_timers_write,
 632
 633         &sysfs_data_op_data_progress,
 634
 635         &sysfs_internal_uuid,
 636         NULL
 637 };
 638
 639 /* options */
 640
 641 SHOW(bch2_fs_opts_dir)
 642 {
 643         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 644         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 645         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 646         int id = opt - bch2_opt_table;
 647         u64 v = bch2_opt_get_by_id(&c->opts, id);
 648
 649         bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
 650         pr_buf(&out, "\n");
 651
 652         return out.pos - buf;
 653 }
 654
 655 STORE(bch2_fs_opts_dir)
 656 {
 657         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 658         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 659         int ret, id = opt - bch2_opt_table;
 660         char *tmp;
 661         u64 v;
 662
 663         tmp = kstrdup(buf, GFP_KERNEL);
 664         if (!tmp)
 665                 return -ENOMEM;
 666
 667         ret = bch2_opt_parse(c, opt, strim(tmp), &v);
 668         kfree(tmp);
 669
 670         if (ret < 0)
 671                 return ret;
 672
 673         ret = bch2_opt_check_may_set(c, id, v);
 674         if (ret < 0)
 675                 return ret;
 676
 677         if (opt->set_sb != SET_NO_SB_OPT) {
 678                 mutex_lock(&c->sb_lock);
 679                 opt->set_sb(c->disk_sb.sb, v);
 680                 bch2_write_super(c);
 681                 mutex_unlock(&c->sb_lock);
 682         }
 683
 684         bch2_opt_set_by_id(&c->opts, id, v);
 685
 686         if ((id == Opt_background_target ||
 687              id == Opt_background_compression) && v) {
 688                 bch2_rebalance_add_work(c, S64_MAX);
 689                 rebalance_wakeup(c);
 690         }
 691
 692         return size;
 693 }
 694 SYSFS_OPS(bch2_fs_opts_dir);
 695
 696 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 697
 698 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 699 {
 700         const struct bch_option *i;
 701         int ret;
 702
 703         for (i = bch2_opt_table;
 704              i < bch2_opt_table + bch2_opts_nr;
 705              i++) {
 706                 if (!(i->mode & (OPT_FORMAT|OPT_MOUNT|OPT_RUNTIME)))
 707                         continue;
 708
 709                 ret = sysfs_create_file(kobj, &i->attr);
 710                 if (ret)
 711                         return ret;
 712         }
 713
 714         return 0;
 715 }
 716
 717 /* time stats */
 718
 719 SHOW(bch2_fs_time_stats)
 720 {
 721         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 722         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 723
 724 #define x(name)                                                         \
 725         if (attr == &sysfs_time_stat_##name) {                          \
 726                 bch2_time_stats_to_text(&out, &c->times[BCH_TIME_##name]);\
 727                 return out.pos - buf;                                   \
 728         }
 729         BCH_TIME_STATS()
 730 #undef x
 731
 732         return 0;
 733 }
 734
 735 STORE(bch2_fs_time_stats)
 736 {
 737         return size;
 738 }
 739 SYSFS_OPS(bch2_fs_time_stats);
 740
 741 struct attribute *bch2_fs_time_stats_files[] = {
 742 #define x(name)                                         \
 743         &sysfs_time_stat_##name,
 744         BCH_TIME_STATS()
 745 #undef x
 746         NULL
 747 };
 748
 749 static void reserve_stats_to_text(struct printbuf *out, struct bch_dev *ca)
 750 {
 751         enum alloc_reserve i;
 752
 753         spin_lock(&ca->fs->freelist_lock);
 754
 755         pr_buf(out, "free_inc:\t%zu\t%zu\n",
 756                fifo_used(&ca->free_inc),
 757                ca->free_inc.size);
 758
 759         for (i = 0; i < RESERVE_NR; i++)
 760                 pr_buf(out, "free[%u]:\t%zu\t%zu\n", i,
 761                        fifo_used(&ca->free[i]),
 762                        ca->free[i].size);
 763
 764         spin_unlock(&ca->fs->freelist_lock);
 765 }
 766
 767 static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 768 {
 769         struct bch_fs *c = ca->fs;
 770         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 771         unsigned i, nr[BCH_DATA_NR];
 772
 773         memset(nr, 0, sizeof(nr));
 774
 775         for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 776                 nr[c->open_buckets[i].type]++;
 777
 778         pr_buf(out,
 779                "\t\t buckets\t sectors      fragmented\n"
 780                "capacity%16llu\n",
 781                ca->mi.nbuckets - ca->mi.first_bucket);
 782
 783         for (i = 1; i < BCH_DATA_NR; i++)
 784                 pr_buf(out, "%-8s%16llu%16llu%16llu\n",
 785                        bch2_data_types[i], stats.d[i].buckets,
 786                        stats.d[i].sectors, stats.d[i].fragmented);
 787
 788         pr_buf(out,
 789                "ec\t%16llu\n"
 790                "available%15llu\n"
 791                "\n"
 792                "free_inc\t\t%zu/%zu\n"
 793                "free[RESERVE_MOVINGGC]\t%zu/%zu\n"
 794                "free[RESERVE_NONE]\t%zu/%zu\n"
 795                "freelist_wait\t\t%s\n"
 796                "open buckets allocated\t%u\n"
 797                "open buckets this dev\t%u\n"
 798                "open buckets total\t%u\n"
 799                "open_buckets_wait\t%s\n"
 800                "open_buckets_btree\t%u\n"
 801                "open_buckets_user\t%u\n"
 802                "btree reserve cache\t%u\n"
 803                "thread state:\t\t%s\n",
 804                stats.buckets_ec,
 805                __dev_buckets_available(ca, stats),
 806                fifo_used(&ca->free_inc),                ca->free_inc.size,
 807                fifo_used(&ca->free[RESERVE_MOVINGGC]),  ca->free[RESERVE_MOVINGGC].size,
 808                fifo_used(&ca->free[RESERVE_NONE]),      ca->free[RESERVE_NONE].size,
 809                c->freelist_wait.list.first              ? "waiting" : "empty",
 810                OPEN_BUCKETS_COUNT - c->open_buckets_nr_free,
 811                ca->nr_open_buckets,
 812                OPEN_BUCKETS_COUNT,
 813                c->open_buckets_wait.list.first          ? "waiting" : "empty",
 814                nr[BCH_DATA_btree],
 815                nr[BCH_DATA_user],
 816                c->btree_reserve_cache_nr,
 817                bch2_allocator_states[ca->allocator_state]);
 818 }
 819
 820 static const char * const bch2_rw[] = {
 821         "read",
 822         "write",
 823         NULL
 824 };
 825
 826 static void dev_iodone_to_text(struct printbuf *out, struct bch_dev *ca)
 827 {
 828         int rw, i;
 829
 830         for (rw = 0; rw < 2; rw++) {
 831                 pr_buf(out, "%s:\n", bch2_rw[rw]);
 832
 833                 for (i = 1; i < BCH_DATA_NR; i++)
 834                         pr_buf(out, "%-12s:%12llu\n",
 835                                bch2_data_types[i],
 836                                percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
 837         }
 838 }
 839
 840 SHOW(bch2_dev)
 841 {
 842         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 843         struct bch_fs *c = ca->fs;
 844         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 845
 846         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 847
 848         sysfs_print(bucket_size,        bucket_bytes(ca));
 849         sysfs_print(block_size,         block_bytes(c));
 850         sysfs_print(first_bucket,       ca->mi.first_bucket);
 851         sysfs_print(nbuckets,           ca->mi.nbuckets);
 852         sysfs_print(durability,         ca->mi.durability);
 853         sysfs_print(discard,            ca->mi.discard);
 854
 855         if (attr == &sysfs_label) {
 856                 if (ca->mi.group) {
 857                         mutex_lock(&c->sb_lock);
 858                         bch2_disk_path_to_text(&out, &c->disk_sb,
 859                                                ca->mi.group - 1);
 860                         mutex_unlock(&c->sb_lock);
 861                 }
 862
 863                 pr_buf(&out, "\n");
 864                 return out.pos - buf;
 865         }
 866
 867         if (attr == &sysfs_has_data) {
 868                 bch2_flags_to_text(&out, bch2_data_types,
 869                                    bch2_dev_has_data(c, ca));
 870                 pr_buf(&out, "\n");
 871                 return out.pos - buf;
 872         }
 873
 874         if (attr == &sysfs_cache_replacement_policy) {
 875                 bch2_string_opt_to_text(&out,
 876                                         bch2_cache_replacement_policies,
 877                                         ca->mi.replacement);
 878                 pr_buf(&out, "\n");
 879                 return out.pos - buf;
 880         }
 881
 882         if (attr == &sysfs_state_rw) {
 883                 bch2_string_opt_to_text(&out, bch2_member_states,
 884                                         ca->mi.state);
 885                 pr_buf(&out, "\n");
 886                 return out.pos - buf;
 887         }
 888
 889         if (attr == &sysfs_iodone) {
 890                 dev_iodone_to_text(&out, ca);
 891                 return out.pos - buf;
 892         }
 893
 894         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 895         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 896
 897         if (attr == &sysfs_io_latency_stats_read) {
 898                 bch2_time_stats_to_text(&out, &ca->io_latency[READ]);
 899                 return out.pos - buf;
 900         }
 901         if (attr == &sysfs_io_latency_stats_write) {
 902                 bch2_time_stats_to_text(&out, &ca->io_latency[WRITE]);
 903                 return out.pos - buf;
 904         }
 905
 906         sysfs_printf(congested,                 "%u%%",
 907                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 908                      * 100 / CONGESTED_MAX);
 909
 910         if (attr == &sysfs_reserve_stats) {
 911                 reserve_stats_to_text(&out, ca);
 912                 return out.pos - buf;
 913         }
 914         if (attr == &sysfs_alloc_debug) {
 915                 dev_alloc_debug_to_text(&out, ca);
 916                 return out.pos - buf;
 917         }
 918
 919         return 0;
 920 }
 921
 922 STORE(bch2_dev)
 923 {
 924         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 925         struct bch_fs *c = ca->fs;
 926         struct bch_member *mi;
 927
 928         if (attr == &sysfs_discard) {
 929                 bool v = strtoul_or_return(buf);
 930
 931                 mutex_lock(&c->sb_lock);
 932                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 933
 934                 if (v != BCH_MEMBER_DISCARD(mi)) {
 935                         SET_BCH_MEMBER_DISCARD(mi, v);
 936                         bch2_write_super(c);
 937                 }
 938                 mutex_unlock(&c->sb_lock);
 939         }
 940
 941         if (attr == &sysfs_cache_replacement_policy) {
 942                 ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
 943
 944                 if (v < 0)
 945                         return v;
 946
 947                 mutex_lock(&c->sb_lock);
 948                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 949
 950                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
 951                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
 952                         bch2_write_super(c);
 953                 }
 954                 mutex_unlock(&c->sb_lock);
 955         }
 956
 957         if (attr == &sysfs_label) {
 958                 char *tmp;
 959                 int ret;
 960
 961                 tmp = kstrdup(buf, GFP_KERNEL);
 962                 if (!tmp)
 963                         return -ENOMEM;
 964
 965                 ret = bch2_dev_group_set(c, ca, strim(tmp));
 966                 kfree(tmp);
 967                 if (ret)
 968                         return ret;
 969         }
 970
 971         if (attr == &sysfs_wake_allocator)
 972                 bch2_wake_allocator(ca);
 973
 974         return size;
 975 }
 976 SYSFS_OPS(bch2_dev);
 977
 978 struct attribute *bch2_dev_files[] = {
 979         &sysfs_uuid,
 980         &sysfs_bucket_size,
 981         &sysfs_block_size,
 982         &sysfs_first_bucket,
 983         &sysfs_nbuckets,
 984         &sysfs_durability,
 985
 986         /* settings: */
 987         &sysfs_discard,
 988         &sysfs_cache_replacement_policy,
 989         &sysfs_state_rw,
 990         &sysfs_label,
 991
 992         &sysfs_has_data,
 993         &sysfs_iodone,
 994
 995         &sysfs_io_latency_read,
 996         &sysfs_io_latency_write,
 997         &sysfs_io_latency_stats_read,
 998         &sysfs_io_latency_stats_write,
 999         &sysfs_congested,
1000
1001         &sysfs_reserve_stats,
1002
1003         /* debug: */
1004         &sysfs_alloc_debug,
1005         &sysfs_wake_allocator,
1006         NULL
1007 };
1008
1009 #endif  /* _BCACHEFS_SYSFS_H_ */