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 "alloc_foreground.h"
  14 #include "sysfs.h"
  15 #include "btree_cache.h"
  16 #include "btree_io.h"
  17 #include "btree_iter.h"
  18 #include "btree_key_cache.h"
  19 #include "btree_update.h"
  20 #include "btree_update_interior.h"
  21 #include "btree_gc.h"
  22 #include "buckets.h"
  23 #include "clock.h"
  24 #include "disk_groups.h"
  25 #include "ec.h"
  26 #include "inode.h"
  27 #include "journal.h"
  28 #include "keylist.h"
  29 #include "move.h"
  30 #include "opts.h"
  31 #include "rebalance.h"
  32 #include "replicas.h"
  33 #include "super-io.h"
  34 #include "tests.h"
  35
  36 #include <linux/blkdev.h>
  37 #include <linux/pretty-printers.h>
  38 #include <linux/sort.h>
  39 #include <linux/sched/clock.h>
  40
  41 #include "util.h"
  42
  43 #define SYSFS_OPS(type)                                                 \
  44 const struct sysfs_ops type ## _sysfs_ops = {                           \
  45         .show   = type ## _show,                                        \
  46         .store  = type ## _store                                        \
  47 }
  48
  49 #define SHOW(fn)                                                        \
  50 static ssize_t fn ## _to_text(struct printbuf *,                        \
  51                               struct kobject *, struct attribute *);    \
  52                                                                         \
  53 static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
  54                            char *buf)                                   \
  55 {                                                                       \
  56         struct printbuf out = PRINTBUF;                                 \
  57         ssize_t ret = fn ## _to_text(&out, kobj, attr);                 \
  58                                                                         \
  59         if (out.pos && out.buf[out.pos - 1] != '\n')                    \
  60                 prt_newline(&out);                                      \
  61                                                                         \
  62         if (!ret && out.allocation_failure)                             \
  63                 ret = -ENOMEM;                                          \
  64                                                                         \
  65         if (!ret) {                                                     \
  66                 ret = min_t(size_t, out.pos, PAGE_SIZE - 1);            \
  67                 memcpy(buf, out.buf, ret);                              \
  68         }                                                               \
  69         printbuf_exit(&out);                                            \
  70         return bch2_err_class(ret);                                     \
  71 }                                                                       \
  72                                                                         \
  73 static ssize_t fn ## _to_text(struct printbuf *out, struct kobject *kobj,\
  74                               struct attribute *attr)
  75
  76 #define STORE(fn)                                                       \
  77 static ssize_t fn ## _store_inner(struct kobject *, struct attribute *,\
  78                             const char *, size_t);                      \
  79                                                                         \
  80 static ssize_t fn ## _store(struct kobject *kobj, struct attribute *attr,\
  81                             const char *buf, size_t size)               \
  82 {                                                                       \
  83         return bch2_err_class(fn##_store_inner(kobj, attr, buf, size)); \
  84 }                                                                       \
  85                                                                         \
  86 static ssize_t fn ## _store_inner(struct kobject *kobj, struct attribute *attr,\
  87                                   const char *buf, size_t size)
  88
  89 #define __sysfs_attribute(_name, _mode)                                 \
  90         static struct attribute sysfs_##_name =                         \
  91                 { .name = #_name, .mode = _mode }
  92
  93 #define write_attribute(n)      __sysfs_attribute(n, 0200)
  94 #define read_attribute(n)       __sysfs_attribute(n, 0444)
  95 #define rw_attribute(n)         __sysfs_attribute(n, 0644)
  96
  97 #define sysfs_printf(file, fmt, ...)                                    \
  98 do {                                                                    \
  99         if (attr == &sysfs_ ## file)                                    \
 100                 prt_printf(out, fmt "\n", __VA_ARGS__);                 \
 101 } while (0)
 102
 103 #define sysfs_print(file, var)                                          \
 104 do {                                                                    \
 105         if (attr == &sysfs_ ## file)                                    \
 106                 snprint(out, var);                                      \
 107 } while (0)
 108
 109 #define sysfs_hprint(file, val)                                         \
 110 do {                                                                    \
 111         if (attr == &sysfs_ ## file)                                    \
 112                 prt_human_readable_s64(out, val);                       \
 113 } while (0)
 114
 115 #define var_printf(_var, fmt)   sysfs_printf(_var, fmt, var(_var))
 116 #define var_print(_var)         sysfs_print(_var, var(_var))
 117 #define var_hprint(_var)        sysfs_hprint(_var, var(_var))
 118
 119 #define sysfs_strtoul(file, var)                                        \
 120 do {                                                                    \
 121         if (attr == &sysfs_ ## file)                                    \
 122                 return strtoul_safe(buf, var) ?: (ssize_t) size;        \
 123 } while (0)
 124
 125 #define sysfs_strtoul_clamp(file, var, min, max)                        \
 126 do {                                                                    \
 127         if (attr == &sysfs_ ## file)                                    \
 128                 return strtoul_safe_clamp(buf, var, min, max)           \
 129                         ?: (ssize_t) size;                              \
 130 } while (0)
 131
 132 #define strtoul_or_return(cp)                                           \
 133 ({                                                                      \
 134         unsigned long _v;                                               \
 135         int _r = kstrtoul(cp, 10, &_v);                                 \
 136         if (_r)                                                         \
 137                 return _r;                                              \
 138         _v;                                                             \
 139 })
 140
 141 #define strtoul_restrict_or_return(cp, min, max)                        \
 142 ({                                                                      \
 143         unsigned long __v = 0;                                          \
 144         int _r = strtoul_safe_restrict(cp, __v, min, max);              \
 145         if (_r)                                                         \
 146                 return _r;                                              \
 147         __v;                                                            \
 148 })
 149
 150 #define strtoi_h_or_return(cp)                                          \
 151 ({                                                                      \
 152         u64 _v;                                                         \
 153         int _r = strtoi_h(cp, &_v);                                     \
 154         if (_r)                                                         \
 155                 return _r;                                              \
 156         _v;                                                             \
 157 })
 158
 159 #define sysfs_hatoi(file, var)                                          \
 160 do {                                                                    \
 161         if (attr == &sysfs_ ## file)                                    \
 162                 return strtoi_h(buf, &var) ?: (ssize_t) size;           \
 163 } while (0)
 164
 165 write_attribute(trigger_gc);
 166 write_attribute(trigger_discards);
 167 write_attribute(trigger_invalidates);
 168 write_attribute(prune_cache);
 169 write_attribute(btree_wakeup);
 170 rw_attribute(btree_gc_periodic);
 171 rw_attribute(gc_gens_pos);
 172
 173 read_attribute(uuid);
 174 read_attribute(minor);
 175 read_attribute(bucket_size);
 176 read_attribute(first_bucket);
 177 read_attribute(nbuckets);
 178 rw_attribute(durability);
 179 read_attribute(iodone);
 180
 181 read_attribute(io_latency_read);
 182 read_attribute(io_latency_write);
 183 read_attribute(io_latency_stats_read);
 184 read_attribute(io_latency_stats_write);
 185 read_attribute(congested);
 186
 187 read_attribute(btree_write_stats);
 188
 189 read_attribute(btree_cache_size);
 190 read_attribute(compression_stats);
 191 read_attribute(journal_debug);
 192 read_attribute(btree_updates);
 193 read_attribute(btree_cache);
 194 read_attribute(btree_key_cache);
 195 read_attribute(stripes_heap);
 196 read_attribute(open_buckets);
 197
 198 read_attribute(internal_uuid);
 199
 200 read_attribute(has_data);
 201 read_attribute(alloc_debug);
 202
 203 #define x(t, n, ...) read_attribute(t);
 204 BCH_PERSISTENT_COUNTERS()
 205 #undef x
 206
 207 rw_attribute(discard);
 208 rw_attribute(label);
 209
 210 rw_attribute(copy_gc_enabled);
 211 read_attribute(copy_gc_wait);
 212
 213 rw_attribute(rebalance_enabled);
 214 sysfs_pd_controller_attribute(rebalance);
 215 read_attribute(rebalance_work);
 216 rw_attribute(promote_whole_extents);
 217
 218 read_attribute(new_stripes);
 219
 220 read_attribute(io_timers_read);
 221 read_attribute(io_timers_write);
 222
 223 read_attribute(data_jobs);
 224
 225 #ifdef CONFIG_BCACHEFS_TESTS
 226 write_attribute(perf_test);
 227 #endif /* CONFIG_BCACHEFS_TESTS */
 228
 229 #define x(_name)                                                \
 230         static struct attribute sysfs_time_stat_##_name =               \
 231                 { .name = #_name, .mode = 0444 };
 232         BCH_TIME_STATS()
 233 #undef x
 234
 235 static struct attribute sysfs_state_rw = {
 236         .name = "state",
 237         .mode =  0444,
 238 };
 239
 240 static size_t bch2_btree_cache_size(struct bch_fs *c)
 241 {
 242         size_t ret = 0;
 243         struct btree *b;
 244
 245         mutex_lock(&c->btree_cache.lock);
 246         list_for_each_entry(b, &c->btree_cache.live, list)
 247                 ret += btree_bytes(c);
 248
 249         mutex_unlock(&c->btree_cache.lock);
 250         return ret;
 251 }
 252
 253 static long data_progress_to_text(struct printbuf *out, struct bch_fs *c)
 254 {
 255         long ret = 0;
 256         struct bch_move_stats *stats;
 257
 258         mutex_lock(&c->data_progress_lock);
 259         list_for_each_entry(stats, &c->data_progress_list, list) {
 260                 prt_printf(out, "%s: data type %s btree_id %s position: ",
 261                        stats->name,
 262                        bch2_data_types[stats->data_type],
 263                        bch2_btree_ids[stats->btree_id]);
 264                 bch2_bpos_to_text(out, stats->pos);
 265                 prt_printf(out, "%s", "\n");
 266         }
 267
 268         mutex_unlock(&c->data_progress_lock);
 269         return ret;
 270 }
 271
 272 static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c)
 273 {
 274         struct btree_trans trans;
 275         struct btree_iter iter;
 276         struct bkey_s_c k;
 277         enum btree_id id;
 278         u64 nr_uncompressed_extents = 0,
 279             nr_compressed_extents = 0,
 280             nr_incompressible_extents = 0,
 281             uncompressed_sectors = 0,
 282             incompressible_sectors = 0,
 283             compressed_sectors_compressed = 0,
 284             compressed_sectors_uncompressed = 0;
 285         int ret;
 286
 287         if (!test_bit(BCH_FS_STARTED, &c->flags))
 288                 return -EPERM;
 289
 290         bch2_trans_init(&trans, c, 0, 0);
 291
 292         for (id = 0; id < BTREE_ID_NR; id++) {
 293                 if (!btree_type_has_ptrs(id))
 294                         continue;
 295
 296                 for_each_btree_key(&trans, iter, id, POS_MIN,
 297                                    BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
 298                         struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
 299                         const union bch_extent_entry *entry;
 300                         struct extent_ptr_decoded p;
 301                         bool compressed = false, uncompressed = false, incompressible = false;
 302
 303                         bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
 304                                 switch (p.crc.compression_type) {
 305                                 case BCH_COMPRESSION_TYPE_none:
 306                                         uncompressed = true;
 307                                         uncompressed_sectors += k.k->size;
 308                                         break;
 309                                 case BCH_COMPRESSION_TYPE_incompressible:
 310                                         incompressible = true;
 311                                         incompressible_sectors += k.k->size;
 312                                         break;
 313                                 default:
 314                                         compressed_sectors_compressed +=
 315                                                 p.crc.compressed_size;
 316                                         compressed_sectors_uncompressed +=
 317                                                 p.crc.uncompressed_size;
 318                                         compressed = true;
 319                                         break;
 320                                 }
 321                         }
 322
 323                         if (incompressible)
 324                                 nr_incompressible_extents++;
 325                         else if (uncompressed)
 326                                 nr_uncompressed_extents++;
 327                         else if (compressed)
 328                                 nr_compressed_extents++;
 329                 }
 330                 bch2_trans_iter_exit(&trans, &iter);
 331         }
 332
 333         bch2_trans_exit(&trans);
 334
 335         if (ret)
 336                 return ret;
 337
 338         prt_printf(out, "uncompressed:\n");
 339         prt_printf(out, "       nr extents:             %llu\n", nr_uncompressed_extents);
 340         prt_printf(out, "       size:                   ");
 341         prt_human_readable_u64(out, uncompressed_sectors << 9);
 342         prt_printf(out, "\n");
 343
 344         prt_printf(out, "compressed:\n");
 345         prt_printf(out, "       nr extents:             %llu\n", nr_compressed_extents);
 346         prt_printf(out, "       compressed size:        ");
 347         prt_human_readable_u64(out, compressed_sectors_compressed << 9);
 348         prt_printf(out, "\n");
 349         prt_printf(out, "       uncompressed size:      ");
 350         prt_human_readable_u64(out, compressed_sectors_uncompressed << 9);
 351         prt_printf(out, "\n");
 352
 353         prt_printf(out, "incompressible:\n");
 354         prt_printf(out, "       nr extents:             %llu\n", nr_incompressible_extents);
 355         prt_printf(out, "       size:                   ");
 356         prt_human_readable_u64(out, incompressible_sectors << 9);
 357         prt_printf(out, "\n");
 358         return 0;
 359 }
 360
 361 static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
 362 {
 363         prt_printf(out, "%s: ", bch2_btree_ids[c->gc_gens_btree]);
 364         bch2_bpos_to_text(out, c->gc_gens_pos);
 365         prt_printf(out, "\n");
 366 }
 367
 368 static void bch2_btree_wakeup_all(struct bch_fs *c)
 369 {
 370         struct btree_trans *trans;
 371
 372         mutex_lock(&c->btree_trans_lock);
 373         list_for_each_entry(trans, &c->btree_trans_list, list) {
 374                 struct btree_bkey_cached_common *b = READ_ONCE(trans->locking);
 375
 376                 if (b)
 377                         six_lock_wakeup_all(&b->lock);
 378
 379         }
 380         mutex_unlock(&c->btree_trans_lock);
 381 }
 382
 383 SHOW(bch2_fs)
 384 {
 385         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 386
 387         sysfs_print(minor,                      c->minor);
 388         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 389
 390         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 391
 392         if (attr == &sysfs_btree_write_stats)
 393                 bch2_btree_write_stats_to_text(out, c);
 394
 395         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 396
 397         if (attr == &sysfs_gc_gens_pos)
 398                 bch2_gc_gens_pos_to_text(out, c);
 399
 400         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 401
 402         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 403         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 404         sysfs_hprint(copy_gc_wait,
 405                      max(0LL, c->copygc_wait -
 406                          atomic64_read(&c->io_clock[WRITE].now)) << 9);
 407
 408         if (attr == &sysfs_rebalance_work)
 409                 bch2_rebalance_work_to_text(out, c);
 410
 411         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 412
 413         /* Debugging: */
 414
 415         if (attr == &sysfs_journal_debug)
 416                 bch2_journal_debug_to_text(out, &c->journal);
 417
 418         if (attr == &sysfs_btree_updates)
 419                 bch2_btree_updates_to_text(out, c);
 420
 421         if (attr == &sysfs_btree_cache)
 422                 bch2_btree_cache_to_text(out, &c->btree_cache);
 423
 424         if (attr == &sysfs_btree_key_cache)
 425                 bch2_btree_key_cache_to_text(out, &c->btree_key_cache);
 426
 427         if (attr == &sysfs_stripes_heap)
 428                 bch2_stripes_heap_to_text(out, c);
 429
 430         if (attr == &sysfs_open_buckets)
 431                 bch2_open_buckets_to_text(out, c);
 432
 433         if (attr == &sysfs_compression_stats)
 434                 bch2_compression_stats_to_text(out, c);
 435
 436         if (attr == &sysfs_new_stripes)
 437                 bch2_new_stripes_to_text(out, c);
 438
 439         if (attr == &sysfs_io_timers_read)
 440                 bch2_io_timers_to_text(out, &c->io_clock[READ]);
 441
 442         if (attr == &sysfs_io_timers_write)
 443                 bch2_io_timers_to_text(out, &c->io_clock[WRITE]);
 444
 445         if (attr == &sysfs_data_jobs)
 446                 data_progress_to_text(out, c);
 447
 448         return 0;
 449 }
 450
 451 STORE(bch2_fs)
 452 {
 453         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 454
 455         if (attr == &sysfs_btree_gc_periodic) {
 456                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 457                         ?: (ssize_t) size;
 458
 459                 wake_up_process(c->gc_thread);
 460                 return ret;
 461         }
 462
 463         if (attr == &sysfs_copy_gc_enabled) {
 464                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 465                         ?: (ssize_t) size;
 466
 467                 if (c->copygc_thread)
 468                         wake_up_process(c->copygc_thread);
 469                 return ret;
 470         }
 471
 472         if (attr == &sysfs_rebalance_enabled) {
 473                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 474                         ?: (ssize_t) size;
 475
 476                 rebalance_wakeup(c);
 477                 return ret;
 478         }
 479
 480         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 481
 482         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 483
 484         /* Debugging: */
 485
 486         if (!test_bit(BCH_FS_STARTED, &c->flags))
 487                 return -EPERM;
 488
 489         /* Debugging: */
 490
 491         if (!test_bit(BCH_FS_RW, &c->flags))
 492                 return -EROFS;
 493
 494         if (attr == &sysfs_prune_cache) {
 495                 struct shrink_control sc;
 496
 497                 sc.gfp_mask = GFP_KERNEL;
 498                 sc.nr_to_scan = strtoul_or_return(buf);
 499                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 500         }
 501
 502         if (attr == &sysfs_btree_wakeup)
 503                 bch2_btree_wakeup_all(c);
 504
 505         if (attr == &sysfs_trigger_gc) {
 506                 /*
 507                  * Full gc is currently incompatible with btree key cache:
 508                  */
 509 #if 0
 510                 down_read(&c->state_lock);
 511                 bch2_gc(c, false, false);
 512                 up_read(&c->state_lock);
 513 #else
 514                 bch2_gc_gens(c);
 515 #endif
 516         }
 517
 518         if (attr == &sysfs_trigger_discards)
 519                 bch2_do_discards(c);
 520
 521         if (attr == &sysfs_trigger_invalidates)
 522                 bch2_do_invalidates(c);
 523
 524 #ifdef CONFIG_BCACHEFS_TESTS
 525         if (attr == &sysfs_perf_test) {
 526                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 527                 char *test              = strsep(&p, " \t\n");
 528                 char *nr_str            = strsep(&p, " \t\n");
 529                 char *threads_str       = strsep(&p, " \t\n");
 530                 unsigned threads;
 531                 u64 nr;
 532                 int ret = -EINVAL;
 533
 534                 if (threads_str &&
 535                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 536                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 537                         ret = bch2_btree_perf_test(c, test, nr, threads);
 538                 kfree(tmp);
 539
 540                 if (ret)
 541                         size = ret;
 542         }
 543 #endif
 544         return size;
 545 }
 546 SYSFS_OPS(bch2_fs);
 547
 548 struct attribute *bch2_fs_files[] = {
 549         &sysfs_minor,
 550         &sysfs_btree_cache_size,
 551         &sysfs_btree_write_stats,
 552
 553         &sysfs_promote_whole_extents,
 554
 555         &sysfs_compression_stats,
 556
 557 #ifdef CONFIG_BCACHEFS_TESTS
 558         &sysfs_perf_test,
 559 #endif
 560         NULL
 561 };
 562
 563 /* counters dir */
 564
 565 SHOW(bch2_fs_counters)
 566 {
 567         struct bch_fs *c = container_of(kobj, struct bch_fs, counters_kobj);
 568         u64 counter = 0;
 569         u64 counter_since_mount = 0;
 570
 571         printbuf_tabstop_push(out, 32);
 572
 573         #define x(t, ...) \
 574                 if (attr == &sysfs_##t) {                                       \
 575                         counter             = percpu_u64_get(&c->counters[BCH_COUNTER_##t]);\
 576                         counter_since_mount = counter - c->counters_on_mount[BCH_COUNTER_##t];\
 577                         prt_printf(out, "since mount:");                                \
 578                         prt_tab(out);                                           \
 579                         prt_human_readable_u64(out, counter_since_mount << 9);  \
 580                         prt_newline(out);                                       \
 581                                                                                 \
 582                         prt_printf(out, "since filesystem creation:");          \
 583                         prt_tab(out);                                           \
 584                         prt_human_readable_u64(out, counter << 9);              \
 585                         prt_newline(out);                                       \
 586                 }
 587         BCH_PERSISTENT_COUNTERS()
 588         #undef x
 589         return 0;
 590 }
 591
 592 STORE(bch2_fs_counters) {
 593         return 0;
 594 }
 595
 596 SYSFS_OPS(bch2_fs_counters);
 597
 598 struct attribute *bch2_fs_counters_files[] = {
 599 #define x(t, ...) \
 600         &sysfs_##t,
 601         BCH_PERSISTENT_COUNTERS()
 602 #undef x
 603         NULL
 604 };
 605 /* internal dir - just a wrapper */
 606
 607 SHOW(bch2_fs_internal)
 608 {
 609         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 610
 611         return bch2_fs_to_text(out, &c->kobj, attr);
 612 }
 613
 614 STORE(bch2_fs_internal)
 615 {
 616         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 617
 618         return bch2_fs_store(&c->kobj, attr, buf, size);
 619 }
 620 SYSFS_OPS(bch2_fs_internal);
 621
 622 struct attribute *bch2_fs_internal_files[] = {
 623         &sysfs_journal_debug,
 624         &sysfs_btree_updates,
 625         &sysfs_btree_cache,
 626         &sysfs_btree_key_cache,
 627         &sysfs_new_stripes,
 628         &sysfs_stripes_heap,
 629         &sysfs_open_buckets,
 630         &sysfs_io_timers_read,
 631         &sysfs_io_timers_write,
 632
 633         &sysfs_trigger_gc,
 634         &sysfs_trigger_discards,
 635         &sysfs_trigger_invalidates,
 636         &sysfs_prune_cache,
 637         &sysfs_btree_wakeup,
 638
 639         &sysfs_gc_gens_pos,
 640
 641         &sysfs_copy_gc_enabled,
 642         &sysfs_copy_gc_wait,
 643
 644         &sysfs_rebalance_enabled,
 645         &sysfs_rebalance_work,
 646         sysfs_pd_controller_files(rebalance),
 647
 648         &sysfs_data_jobs,
 649
 650         &sysfs_internal_uuid,
 651         NULL
 652 };
 653
 654 /* options */
 655
 656 SHOW(bch2_fs_opts_dir)
 657 {
 658         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 659         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 660         int id = opt - bch2_opt_table;
 661         u64 v = bch2_opt_get_by_id(&c->opts, id);
 662
 663         bch2_opt_to_text(out, c, c->disk_sb.sb, opt, v, OPT_SHOW_FULL_LIST);
 664         prt_char(out, '\n');
 665
 666         return 0;
 667 }
 668
 669 STORE(bch2_fs_opts_dir)
 670 {
 671         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 672         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 673         int ret, id = opt - bch2_opt_table;
 674         char *tmp;
 675         u64 v;
 676
 677         /*
 678          * We don't need to take c->writes for correctness, but it eliminates an
 679          * unsightly error message in the dmesg log when we're RO:
 680          */
 681         if (unlikely(!percpu_ref_tryget_live(&c->writes)))
 682                 return -EROFS;
 683
 684         tmp = kstrdup(buf, GFP_KERNEL);
 685         if (!tmp) {
 686                 ret = -ENOMEM;
 687                 goto err;
 688         }
 689
 690         ret = bch2_opt_parse(c, opt, strim(tmp), &v, NULL);
 691         kfree(tmp);
 692
 693         if (ret < 0)
 694                 goto err;
 695
 696         ret = bch2_opt_check_may_set(c, id, v);
 697         if (ret < 0)
 698                 goto err;
 699
 700         bch2_opt_set_sb(c, opt, v);
 701         bch2_opt_set_by_id(&c->opts, id, v);
 702
 703         if ((id == Opt_background_target ||
 704              id == Opt_background_compression) && v) {
 705                 bch2_rebalance_add_work(c, S64_MAX);
 706                 rebalance_wakeup(c);
 707         }
 708
 709         ret = size;
 710 err:
 711         percpu_ref_put(&c->writes);
 712         return ret;
 713 }
 714 SYSFS_OPS(bch2_fs_opts_dir);
 715
 716 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 717
 718 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 719 {
 720         const struct bch_option *i;
 721         int ret;
 722
 723         for (i = bch2_opt_table;
 724              i < bch2_opt_table + bch2_opts_nr;
 725              i++) {
 726                 if (!(i->flags & OPT_FS))
 727                         continue;
 728
 729                 ret = sysfs_create_file(kobj, &i->attr);
 730                 if (ret)
 731                         return ret;
 732         }
 733
 734         return 0;
 735 }
 736
 737 /* time stats */
 738
 739 SHOW(bch2_fs_time_stats)
 740 {
 741         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 742
 743 #define x(name)                                                         \
 744         if (attr == &sysfs_time_stat_##name)                            \
 745                 bch2_time_stats_to_text(out, &c->times[BCH_TIME_##name]);
 746         BCH_TIME_STATS()
 747 #undef x
 748
 749         return 0;
 750 }
 751
 752 STORE(bch2_fs_time_stats)
 753 {
 754         return size;
 755 }
 756 SYSFS_OPS(bch2_fs_time_stats);
 757
 758 struct attribute *bch2_fs_time_stats_files[] = {
 759 #define x(name)                                         \
 760         &sysfs_time_stat_##name,
 761         BCH_TIME_STATS()
 762 #undef x
 763         NULL
 764 };
 765
 766 static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 767 {
 768         struct bch_fs *c = ca->fs;
 769         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 770         unsigned i, nr[BCH_DATA_NR];
 771
 772         memset(nr, 0, sizeof(nr));
 773
 774         for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 775                 nr[c->open_buckets[i].data_type]++;
 776
 777         prt_printf(out,
 778                "\t\t\t buckets\t sectors      fragmented\n"
 779                "capacity\t%16llu\n",
 780                ca->mi.nbuckets - ca->mi.first_bucket);
 781
 782         for (i = 0; i < BCH_DATA_NR; i++)
 783                 prt_printf(out, "%-16s%16llu%16llu%16llu\n",
 784                        bch2_data_types[i], stats.d[i].buckets,
 785                        stats.d[i].sectors, stats.d[i].fragmented);
 786
 787         prt_printf(out,
 788                "ec\t\t%16llu\n"
 789                "\n"
 790                "freelist_wait\t\t%s\n"
 791                "open buckets allocated\t%u\n"
 792                "open buckets this dev\t%u\n"
 793                "open buckets total\t%u\n"
 794                "open_buckets_wait\t%s\n"
 795                "open_buckets_btree\t%u\n"
 796                "open_buckets_user\t%u\n"
 797                "buckets_to_invalidate\t%llu\n"
 798                "btree reserve cache\t%u\n",
 799                stats.buckets_ec,
 800                c->freelist_wait.list.first              ? "waiting" : "empty",
 801                OPEN_BUCKETS_COUNT - c->open_buckets_nr_free,
 802                ca->nr_open_buckets,
 803                OPEN_BUCKETS_COUNT,
 804                c->open_buckets_wait.list.first          ? "waiting" : "empty",
 805                nr[BCH_DATA_btree],
 806                nr[BCH_DATA_user],
 807                should_invalidate_buckets(ca, stats),
 808                c->btree_reserve_cache_nr);
 809 }
 810
 811 static const char * const bch2_rw[] = {
 812         "read",
 813         "write",
 814         NULL
 815 };
 816
 817 static void dev_iodone_to_text(struct printbuf *out, struct bch_dev *ca)
 818 {
 819         int rw, i;
 820
 821         for (rw = 0; rw < 2; rw++) {
 822                 prt_printf(out, "%s:\n", bch2_rw[rw]);
 823
 824                 for (i = 1; i < BCH_DATA_NR; i++)
 825                         prt_printf(out, "%-12s:%12llu\n",
 826                                bch2_data_types[i],
 827                                percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
 828         }
 829 }
 830
 831 SHOW(bch2_dev)
 832 {
 833         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 834         struct bch_fs *c = ca->fs;
 835
 836         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 837
 838         sysfs_print(bucket_size,        bucket_bytes(ca));
 839         sysfs_print(first_bucket,       ca->mi.first_bucket);
 840         sysfs_print(nbuckets,           ca->mi.nbuckets);
 841         sysfs_print(durability,         ca->mi.durability);
 842         sysfs_print(discard,            ca->mi.discard);
 843
 844         if (attr == &sysfs_label) {
 845                 if (ca->mi.group) {
 846                         mutex_lock(&c->sb_lock);
 847                         bch2_disk_path_to_text(out, c->disk_sb.sb,
 848                                                ca->mi.group - 1);
 849                         mutex_unlock(&c->sb_lock);
 850                 }
 851
 852                 prt_char(out, '\n');
 853         }
 854
 855         if (attr == &sysfs_has_data) {
 856                 prt_bitflags(out, bch2_data_types, bch2_dev_has_data(c, ca));
 857                 prt_char(out, '\n');
 858         }
 859
 860         if (attr == &sysfs_state_rw) {
 861                 prt_string_option(out, bch2_member_states, ca->mi.state);
 862                 prt_char(out, '\n');
 863         }
 864
 865         if (attr == &sysfs_iodone)
 866                 dev_iodone_to_text(out, ca);
 867
 868         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 869         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 870
 871         if (attr == &sysfs_io_latency_stats_read)
 872                 bch2_time_stats_to_text(out, &ca->io_latency[READ]);
 873
 874         if (attr == &sysfs_io_latency_stats_write)
 875                 bch2_time_stats_to_text(out, &ca->io_latency[WRITE]);
 876
 877         sysfs_printf(congested,                 "%u%%",
 878                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 879                      * 100 / CONGESTED_MAX);
 880
 881         if (attr == &sysfs_alloc_debug)
 882                 dev_alloc_debug_to_text(out, ca);
 883
 884         return 0;
 885 }
 886
 887 STORE(bch2_dev)
 888 {
 889         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 890         struct bch_fs *c = ca->fs;
 891         struct bch_member *mi;
 892
 893         if (attr == &sysfs_discard) {
 894                 bool v = strtoul_or_return(buf);
 895
 896                 mutex_lock(&c->sb_lock);
 897                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 898
 899                 if (v != BCH_MEMBER_DISCARD(mi)) {
 900                         SET_BCH_MEMBER_DISCARD(mi, v);
 901                         bch2_write_super(c);
 902                 }
 903                 mutex_unlock(&c->sb_lock);
 904         }
 905
 906         if (attr == &sysfs_durability) {
 907                 u64 v = strtoul_or_return(buf);
 908
 909                 mutex_lock(&c->sb_lock);
 910                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
 911
 912                 if (v != BCH_MEMBER_DURABILITY(mi)) {
 913                         SET_BCH_MEMBER_DURABILITY(mi, v + 1);
 914                         bch2_write_super(c);
 915                 }
 916                 mutex_unlock(&c->sb_lock);
 917         }
 918
 919         if (attr == &sysfs_label) {
 920                 char *tmp;
 921                 int ret;
 922
 923                 tmp = kstrdup(buf, GFP_KERNEL);
 924                 if (!tmp)
 925                         return -ENOMEM;
 926
 927                 ret = bch2_dev_group_set(c, ca, strim(tmp));
 928                 kfree(tmp);
 929                 if (ret)
 930                         return ret;
 931         }
 932
 933         return size;
 934 }
 935 SYSFS_OPS(bch2_dev);
 936
 937 struct attribute *bch2_dev_files[] = {
 938         &sysfs_uuid,
 939         &sysfs_bucket_size,
 940         &sysfs_first_bucket,
 941         &sysfs_nbuckets,
 942         &sysfs_durability,
 943
 944         /* settings: */
 945         &sysfs_discard,
 946         &sysfs_state_rw,
 947         &sysfs_label,
 948
 949         &sysfs_has_data,
 950         &sysfs_iodone,
 951
 952         &sysfs_io_latency_read,
 953         &sysfs_io_latency_write,
 954         &sysfs_io_latency_stats_read,
 955         &sysfs_io_latency_stats_write,
 956         &sysfs_congested,
 957
 958         /* debug: */
 959         &sysfs_alloc_debug,
 960         NULL
 961 };
 962
 963 #endif  /* _BCACHEFS_SYSFS_H_ */