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

   1 /*
   2  * bcache sysfs interfaces
   3  *
   4  * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
   5  * Copyright 2012 Google, Inc.
   6  */
   7
   8 #ifndef NO_BCACHEFS_SYSFS
   9
  10 #include "bcachefs.h"
  11 #include "alloc_background.h"
  12 #include "sysfs.h"
  13 #include "btree_cache.h"
  14 #include "btree_io.h"
  15 #include "btree_iter.h"
  16 #include "btree_update.h"
  17 #include "btree_update_interior.h"
  18 #include "btree_gc.h"
  19 #include "buckets.h"
  20 #include "disk_groups.h"
  21 #include "ec.h"
  22 #include "inode.h"
  23 #include "journal.h"
  24 #include "keylist.h"
  25 #include "move.h"
  26 #include "opts.h"
  27 #include "rebalance.h"
  28 #include "replicas.h"
  29 #include "super-io.h"
  30 #include "tests.h"
  31
  32 #include <linux/blkdev.h>
  33 #include <linux/sort.h>
  34 #include <linux/sched/clock.h>
  35
  36 #include "util.h"
  37
  38 #define SYSFS_OPS(type)                                                 \
  39 struct sysfs_ops type ## _sysfs_ops = {                                 \
  40         .show   = type ## _show,                                        \
  41         .store  = type ## _store                                        \
  42 }
  43
  44 #define SHOW(fn)                                                        \
  45 static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
  46                            char *buf)                                   \
  47
  48 #define STORE(fn)                                                       \
  49 static ssize_t fn ## _store(struct kobject *kobj, struct attribute *attr,\
  50                             const char *buf, size_t size)               \
  51
  52 #define __sysfs_attribute(_name, _mode)                                 \
  53         static struct attribute sysfs_##_name =                         \
  54                 { .name = #_name, .mode = _mode }
  55
  56 #define write_attribute(n)      __sysfs_attribute(n, S_IWUSR)
  57 #define read_attribute(n)       __sysfs_attribute(n, S_IRUGO)
  58 #define rw_attribute(n)         __sysfs_attribute(n, S_IRUGO|S_IWUSR)
  59
  60 #define sysfs_printf(file, fmt, ...)                                    \
  61 do {                                                                    \
  62         if (attr == &sysfs_ ## file)                                    \
  63                 return scnprintf(buf, PAGE_SIZE, fmt "\n", __VA_ARGS__);\
  64 } while (0)
  65
  66 #define sysfs_print(file, var)                                          \
  67 do {                                                                    \
  68         if (attr == &sysfs_ ## file)                                    \
  69                 return snprint(buf, PAGE_SIZE, var);                    \
  70 } while (0)
  71
  72 #define sysfs_hprint(file, val)                                         \
  73 do {                                                                    \
  74         if (attr == &sysfs_ ## file) {                                  \
  75                 struct printbuf out = _PBUF(buf, PAGE_SIZE);            \
  76                 bch2_hprint(&out, val);                                 \
  77                 pr_buf(&out, "\n");                                     \
  78                 return out.pos - buf;                                   \
  79         }                                                               \
  80 } while (0)
  81
  82 #define var_printf(_var, fmt)   sysfs_printf(_var, fmt, var(_var))
  83 #define var_print(_var)         sysfs_print(_var, var(_var))
  84 #define var_hprint(_var)        sysfs_hprint(_var, var(_var))
  85
  86 #define sysfs_strtoul(file, var)                                        \
  87 do {                                                                    \
  88         if (attr == &sysfs_ ## file)                                    \
  89                 return strtoul_safe(buf, var) ?: (ssize_t) size;        \
  90 } while (0)
  91
  92 #define sysfs_strtoul_clamp(file, var, min, max)                        \
  93 do {                                                                    \
  94         if (attr == &sysfs_ ## file)                                    \
  95                 return strtoul_safe_clamp(buf, var, min, max)           \
  96                         ?: (ssize_t) size;                              \
  97 } while (0)
  98
  99 #define strtoul_or_return(cp)                                           \
 100 ({                                                                      \
 101         unsigned long _v;                                               \
 102         int _r = kstrtoul(cp, 10, &_v);                                 \
 103         if (_r)                                                         \
 104                 return _r;                                              \
 105         _v;                                                             \
 106 })
 107
 108 #define strtoul_restrict_or_return(cp, min, max)                        \
 109 ({                                                                      \
 110         unsigned long __v = 0;                                          \
 111         int _r = strtoul_safe_restrict(cp, __v, min, max);              \
 112         if (_r)                                                         \
 113                 return _r;                                              \
 114         __v;                                                            \
 115 })
 116
 117 #define strtoi_h_or_return(cp)                                          \
 118 ({                                                                      \
 119         u64 _v;                                                         \
 120         int _r = strtoi_h(cp, &_v);                                     \
 121         if (_r)                                                         \
 122                 return _r;                                              \
 123         _v;                                                             \
 124 })
 125
 126 #define sysfs_hatoi(file, var)                                          \
 127 do {                                                                    \
 128         if (attr == &sysfs_ ## file)                                    \
 129                 return strtoi_h(buf, &var) ?: (ssize_t) size;           \
 130 } while (0)
 131
 132 write_attribute(trigger_journal_flush);
 133 write_attribute(trigger_btree_coalesce);
 134 write_attribute(trigger_gc);
 135 write_attribute(trigger_alloc_write);
 136 write_attribute(prune_cache);
 137 rw_attribute(btree_gc_periodic);
 138
 139 read_attribute(uuid);
 140 read_attribute(minor);
 141 read_attribute(bucket_size);
 142 read_attribute(block_size);
 143 read_attribute(btree_node_size);
 144 read_attribute(first_bucket);
 145 read_attribute(nbuckets);
 146 read_attribute(durability);
 147 read_attribute(iodone);
 148
 149 read_attribute(io_latency_read);
 150 read_attribute(io_latency_write);
 151 read_attribute(io_latency_stats_read);
 152 read_attribute(io_latency_stats_write);
 153 read_attribute(congested);
 154
 155 read_attribute(bucket_quantiles_last_read);
 156 read_attribute(bucket_quantiles_last_write);
 157 read_attribute(bucket_quantiles_fragmentation);
 158 read_attribute(bucket_quantiles_oldest_gen);
 159
 160 read_attribute(reserve_stats);
 161 read_attribute(btree_cache_size);
 162 read_attribute(compression_stats);
 163 read_attribute(journal_debug);
 164 read_attribute(journal_pins);
 165 read_attribute(btree_updates);
 166 read_attribute(dirty_btree_nodes);
 167
 168 read_attribute(internal_uuid);
 169
 170 read_attribute(has_data);
 171 read_attribute(alloc_debug);
 172 write_attribute(wake_allocator);
 173
 174 read_attribute(read_realloc_races);
 175 read_attribute(extent_migrate_done);
 176 read_attribute(extent_migrate_raced);
 177
 178 rw_attribute(journal_write_delay_ms);
 179 rw_attribute(journal_reclaim_delay_ms);
 180
 181 rw_attribute(discard);
 182 rw_attribute(cache_replacement_policy);
 183 rw_attribute(label);
 184
 185 rw_attribute(copy_gc_enabled);
 186 sysfs_pd_controller_attribute(copy_gc);
 187
 188 rw_attribute(rebalance_enabled);
 189 sysfs_pd_controller_attribute(rebalance);
 190 read_attribute(rebalance_work);
 191 rw_attribute(promote_whole_extents);
 192
 193 read_attribute(new_stripes);
 194
 195 rw_attribute(pd_controllers_update_seconds);
 196
 197 read_attribute(meta_replicas_have);
 198 read_attribute(data_replicas_have);
 199
 200 #ifdef CONFIG_BCACHEFS_TESTS
 201 write_attribute(perf_test);
 202 #endif /* CONFIG_BCACHEFS_TESTS */
 203
 204 #define BCH_DEBUG_PARAM(name, description)                              \
 205         rw_attribute(name);
 206
 207         BCH_DEBUG_PARAMS()
 208 #undef BCH_DEBUG_PARAM
 209
 210 #define x(_name)                                                \
 211         static struct attribute sysfs_time_stat_##_name =               \
 212                 { .name = #_name, .mode = S_IRUGO };
 213         BCH_TIME_STATS()
 214 #undef x
 215
 216 static struct attribute sysfs_state_rw = {
 217         .name = "state",
 218         .mode = S_IRUGO
 219 };
 220
 221 static size_t bch2_btree_cache_size(struct bch_fs *c)
 222 {
 223         size_t ret = 0;
 224         struct btree *b;
 225
 226         mutex_lock(&c->btree_cache.lock);
 227         list_for_each_entry(b, &c->btree_cache.live, list)
 228                 ret += btree_bytes(c);
 229
 230         mutex_unlock(&c->btree_cache.lock);
 231         return ret;
 232 }
 233
 234 static ssize_t show_fs_alloc_debug(struct bch_fs *c, char *buf)
 235 {
 236         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 237         struct bch_fs_usage *fs_usage = bch2_fs_usage_read(c);
 238         unsigned i;
 239
 240         if (!fs_usage)
 241                 return -ENOMEM;
 242
 243         pr_buf(&out, "capacity:\t\t\t%llu\n", c->capacity);
 244
 245         pr_buf(&out, "hidden:\t\t\t\t%llu\n",
 246                fs_usage->s.hidden);
 247         pr_buf(&out, "data:\t\t\t\t%llu\n",
 248                fs_usage->s.data);
 249         pr_buf(&out, "cached:\t\t\t\t%llu\n",
 250                fs_usage->s.cached);
 251         pr_buf(&out, "reserved:\t\t\t%llu\n",
 252                fs_usage->s.reserved);
 253         pr_buf(&out, "nr_inodes:\t\t\t%llu\n",
 254                fs_usage->s.nr_inodes);
 255         pr_buf(&out, "online reserved:\t\t%llu\n",
 256                fs_usage->s.online_reserved);
 257
 258         for (i = 0;
 259              i < ARRAY_SIZE(fs_usage->persistent_reserved);
 260              i++) {
 261                 pr_buf(&out, "%u replicas:\n", i + 1);
 262                 pr_buf(&out, "\treserved:\t\t%llu\n",
 263                        fs_usage->persistent_reserved[i]);
 264         }
 265
 266         for (i = 0; i < c->replicas.nr; i++) {
 267                 struct bch_replicas_entry *e =
 268                         cpu_replicas_entry(&c->replicas, i);
 269
 270                 pr_buf(&out, "\t");
 271                 bch2_replicas_entry_to_text(&out, e);
 272                 pr_buf(&out, ":\t%llu\n", fs_usage->data[i]);
 273         }
 274
 275         percpu_up_read_preempt_enable(&c->mark_lock);
 276
 277         kfree(fs_usage);
 278
 279         return out.pos - buf;
 280 }
 281
 282 static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
 283 {
 284         struct btree_iter iter;
 285         struct bkey_s_c k;
 286         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 287             nr_compressed_extents = 0,
 288             compressed_sectors_compressed = 0,
 289             compressed_sectors_uncompressed = 0;
 290
 291         if (!bch2_fs_running(c))
 292                 return -EPERM;
 293
 294         for_each_btree_key(&iter, c, BTREE_ID_EXTENTS, POS_MIN, 0, k)
 295                 if (k.k->type == KEY_TYPE_extent) {
 296                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 297                         const union bch_extent_entry *entry;
 298                         struct extent_ptr_decoded p;
 299
 300                         extent_for_each_ptr_decode(e, p, entry) {
 301                                 if (p.crc.compression_type == BCH_COMPRESSION_NONE) {
 302                                         nr_uncompressed_extents++;
 303                                         uncompressed_sectors += e.k->size;
 304                                 } else {
 305                                         nr_compressed_extents++;
 306                                         compressed_sectors_compressed +=
 307                                                 p.crc.compressed_size;
 308                                         compressed_sectors_uncompressed +=
 309                                                 p.crc.uncompressed_size;
 310                                 }
 311
 312                                 /* only looking at the first ptr */
 313                                 break;
 314                         }
 315                 }
 316         bch2_btree_iter_unlock(&iter);
 317
 318         return scnprintf(buf, PAGE_SIZE,
 319                         "uncompressed data:\n"
 320                         "       nr extents:                     %llu\n"
 321                         "       size (bytes):                   %llu\n"
 322                         "compressed data:\n"
 323                         "       nr extents:                     %llu\n"
 324                         "       compressed size (bytes):        %llu\n"
 325                         "       uncompressed size (bytes):      %llu\n",
 326                         nr_uncompressed_extents,
 327                         uncompressed_sectors << 9,
 328                         nr_compressed_extents,
 329                         compressed_sectors_compressed << 9,
 330                         compressed_sectors_uncompressed << 9);
 331 }
 332
 333 static ssize_t bch2_new_stripes(struct bch_fs *c, char *buf)
 334 {
 335         char *out = buf, *end = buf + PAGE_SIZE;
 336         struct ec_stripe_head *h;
 337         struct ec_stripe_new *s;
 338
 339         mutex_lock(&c->ec_new_stripe_lock);
 340         list_for_each_entry(h, &c->ec_new_stripe_list, list) {
 341                 out += scnprintf(out, end - out,
 342                                  "target %u algo %u redundancy %u:\n",
 343                                  h->target, h->algo, h->redundancy);
 344
 345                 if (h->s)
 346                         out += scnprintf(out, end - out,
 347                                          "\tpending: blocks %u allocated %u\n",
 348                                          h->s->blocks.nr,
 349                                          bitmap_weight(h->s->blocks_allocated,
 350                                                        h->s->blocks.nr));
 351
 352                 mutex_lock(&h->lock);
 353                 list_for_each_entry(s, &h->stripes, list)
 354                         out += scnprintf(out, end - out,
 355                                          "\tin flight: blocks %u allocated %u pin %u\n",
 356                                          s->blocks.nr,
 357                                          bitmap_weight(s->blocks_allocated,
 358                                                        s->blocks.nr),
 359                                          atomic_read(&s->pin));
 360                 mutex_unlock(&h->lock);
 361
 362         }
 363         mutex_unlock(&c->ec_new_stripe_lock);
 364
 365         return out - buf;
 366 }
 367
 368 SHOW(bch2_fs)
 369 {
 370         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 371
 372         sysfs_print(minor,                      c->minor);
 373         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 374
 375         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 376         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 377
 378         sysfs_print(block_size,                 block_bytes(c));
 379         sysfs_print(btree_node_size,            btree_bytes(c));
 380         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 381
 382         sysfs_print(read_realloc_races,
 383                     atomic_long_read(&c->read_realloc_races));
 384         sysfs_print(extent_migrate_done,
 385                     atomic_long_read(&c->extent_migrate_done));
 386         sysfs_print(extent_migrate_raced,
 387                     atomic_long_read(&c->extent_migrate_raced));
 388
 389         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 390
 391         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 392
 393         sysfs_print(pd_controllers_update_seconds,
 394                     c->pd_controllers_update_seconds);
 395
 396         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 397         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 398
 399         if (attr == &sysfs_rebalance_work)
 400                 return bch2_rebalance_work_show(c, buf);
 401
 402         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 403
 404         sysfs_printf(meta_replicas_have, "%i",  bch2_replicas_online(c, true));
 405         sysfs_printf(data_replicas_have, "%i",  bch2_replicas_online(c, false));
 406
 407         /* Debugging: */
 408
 409         if (attr == &sysfs_alloc_debug)
 410                 return show_fs_alloc_debug(c, buf);
 411
 412         if (attr == &sysfs_journal_debug)
 413                 return bch2_journal_print_debug(&c->journal, buf);
 414
 415         if (attr == &sysfs_journal_pins)
 416                 return bch2_journal_print_pins(&c->journal, buf);
 417
 418         if (attr == &sysfs_btree_updates)
 419                 return bch2_btree_updates_print(c, buf);
 420
 421         if (attr == &sysfs_dirty_btree_nodes)
 422                 return bch2_dirty_btree_nodes_print(c, buf);
 423
 424         if (attr == &sysfs_compression_stats)
 425                 return bch2_compression_stats(c, buf);
 426
 427         if (attr == &sysfs_new_stripes)
 428                 return bch2_new_stripes(c, buf);
 429
 430 #define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
 431         BCH_DEBUG_PARAMS()
 432 #undef BCH_DEBUG_PARAM
 433
 434         return 0;
 435 }
 436
 437 STORE(__bch2_fs)
 438 {
 439         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 440
 441         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 442         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 443
 444         if (attr == &sysfs_btree_gc_periodic) {
 445                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 446                         ?: (ssize_t) size;
 447
 448                 wake_up_process(c->gc_thread);
 449                 return ret;
 450         }
 451
 452         if (attr == &sysfs_copy_gc_enabled) {
 453                 struct bch_dev *ca;
 454                 unsigned i;
 455                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 456                         ?: (ssize_t) size;
 457
 458                 for_each_member_device(ca, c, i)
 459                         if (ca->copygc_thread)
 460                                 wake_up_process(ca->copygc_thread);
 461                 return ret;
 462         }
 463
 464         if (attr == &sysfs_rebalance_enabled) {
 465                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 466                         ?: (ssize_t) size;
 467
 468                 rebalance_wakeup(c);
 469                 return ret;
 470         }
 471
 472         sysfs_strtoul(pd_controllers_update_seconds,
 473                       c->pd_controllers_update_seconds);
 474         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 475
 476         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 477
 478         /* Debugging: */
 479
 480 #define BCH_DEBUG_PARAM(name, description) sysfs_strtoul(name, c->name);
 481         BCH_DEBUG_PARAMS()
 482 #undef BCH_DEBUG_PARAM
 483
 484         if (!bch2_fs_running(c))
 485                 return -EPERM;
 486
 487         /* Debugging: */
 488
 489         if (attr == &sysfs_trigger_journal_flush)
 490                 bch2_journal_meta_async(&c->journal, NULL);
 491
 492         if (attr == &sysfs_trigger_btree_coalesce)
 493                 bch2_coalesce(c);
 494
 495         if (attr == &sysfs_trigger_gc)
 496                 bch2_gc(c, NULL, false);
 497
 498         if (attr == &sysfs_trigger_alloc_write) {
 499                 bool wrote;
 500
 501                 bch2_alloc_write(c, false, &wrote);
 502         }
 503
 504         if (attr == &sysfs_prune_cache) {
 505                 struct shrink_control sc;
 506
 507                 sc.gfp_mask = GFP_KERNEL;
 508                 sc.nr_to_scan = strtoul_or_return(buf);
 509                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 510         }
 511 #ifdef CONFIG_BCACHEFS_TESTS
 512         if (attr == &sysfs_perf_test) {
 513                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 514                 char *test              = strsep(&p, " \t\n");
 515                 char *nr_str            = strsep(&p, " \t\n");
 516                 char *threads_str       = strsep(&p, " \t\n");
 517                 unsigned threads;
 518                 u64 nr;
 519                 int ret = -EINVAL;
 520
 521                 if (threads_str &&
 522                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 523                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 524                         bch2_btree_perf_test(c, test, nr, threads);
 525                 else
 526                         size = ret;
 527                 kfree(tmp);
 528         }
 529 #endif
 530         return size;
 531 }
 532
 533 STORE(bch2_fs)
 534 {
 535         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 536
 537         mutex_lock(&c->state_lock);
 538         size = __bch2_fs_store(kobj, attr, buf, size);
 539         mutex_unlock(&c->state_lock);
 540
 541         return size;
 542 }
 543 SYSFS_OPS(bch2_fs);
 544
 545 struct attribute *bch2_fs_files[] = {
 546         &sysfs_minor,
 547         &sysfs_block_size,
 548         &sysfs_btree_node_size,
 549         &sysfs_btree_cache_size,
 550
 551         &sysfs_meta_replicas_have,
 552         &sysfs_data_replicas_have,
 553
 554         &sysfs_journal_write_delay_ms,
 555         &sysfs_journal_reclaim_delay_ms,
 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
 589         &sysfs_read_realloc_races,
 590         &sysfs_extent_migrate_done,
 591         &sysfs_extent_migrate_raced,
 592
 593         &sysfs_trigger_journal_flush,
 594         &sysfs_trigger_btree_coalesce,
 595         &sysfs_trigger_gc,
 596         &sysfs_trigger_alloc_write,
 597         &sysfs_prune_cache,
 598
 599         &sysfs_copy_gc_enabled,
 600
 601         &sysfs_rebalance_enabled,
 602         &sysfs_rebalance_work,
 603         sysfs_pd_controller_files(rebalance),
 604
 605         &sysfs_new_stripes,
 606
 607         &sysfs_internal_uuid,
 608
 609 #define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
 610         BCH_DEBUG_PARAMS()
 611 #undef BCH_DEBUG_PARAM
 612
 613         NULL
 614 };
 615
 616 /* options */
 617
 618 SHOW(bch2_fs_opts_dir)
 619 {
 620         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 621         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 622         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 623         int id = opt - bch2_opt_table;
 624         u64 v = bch2_opt_get_by_id(&c->opts, id);
 625
 626         bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
 627         pr_buf(&out, "\n");
 628
 629         return out.pos - buf;
 630 }
 631
 632 STORE(bch2_fs_opts_dir)
 633 {
 634         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 635         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 636         int ret, id = opt - bch2_opt_table;
 637         char *tmp;
 638         u64 v;
 639
 640         tmp = kstrdup(buf, GFP_KERNEL);
 641         if (!tmp)
 642                 return -ENOMEM;
 643
 644         ret = bch2_opt_parse(c, opt, strim(tmp), &v);
 645         kfree(tmp);
 646
 647         if (ret < 0)
 648                 return ret;
 649
 650         ret = bch2_opt_check_may_set(c, id, v);
 651         if (ret < 0)
 652                 return ret;
 653
 654         if (opt->set_sb != SET_NO_SB_OPT) {
 655                 mutex_lock(&c->sb_lock);
 656                 opt->set_sb(c->disk_sb.sb, v);
 657                 bch2_write_super(c);
 658                 mutex_unlock(&c->sb_lock);
 659         }
 660
 661         bch2_opt_set_by_id(&c->opts, id, v);
 662
 663         if ((id == Opt_background_target ||
 664              id == Opt_background_compression) && v) {
 665                 bch2_rebalance_add_work(c, S64_MAX);
 666                 rebalance_wakeup(c);
 667         }
 668
 669         return size;
 670 }
 671 SYSFS_OPS(bch2_fs_opts_dir);
 672
 673 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 674
 675 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 676 {
 677         const struct bch_option *i;
 678         int ret;
 679
 680         for (i = bch2_opt_table;
 681              i < bch2_opt_table + bch2_opts_nr;
 682              i++) {
 683                 if (!(i->mode & (OPT_FORMAT|OPT_MOUNT|OPT_RUNTIME)))
 684                         continue;
 685
 686                 ret = sysfs_create_file(kobj, &i->attr);
 687                 if (ret)
 688                         return ret;
 689         }
 690
 691         return 0;
 692 }
 693
 694 /* time stats */
 695
 696 SHOW(bch2_fs_time_stats)
 697 {
 698         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 699
 700 #define x(name)                                         \
 701         if (attr == &sysfs_time_stat_##name)                            \
 702                 return bch2_time_stats_print(&c->times[BCH_TIME_##name],\
 703                                              buf, PAGE_SIZE);
 704         BCH_TIME_STATS()
 705 #undef x
 706
 707         return 0;
 708 }
 709
 710 STORE(bch2_fs_time_stats)
 711 {
 712         return size;
 713 }
 714 SYSFS_OPS(bch2_fs_time_stats);
 715
 716 struct attribute *bch2_fs_time_stats_files[] = {
 717 #define x(name)                                         \
 718         &sysfs_time_stat_##name,
 719         BCH_TIME_STATS()
 720 #undef x
 721         NULL
 722 };
 723
 724 typedef unsigned (bucket_map_fn)(struct bch_fs *, struct bch_dev *,
 725                                  size_t, void *);
 726
 727 static unsigned bucket_last_io_fn(struct bch_fs *c, struct bch_dev *ca,
 728                                   size_t b, void *private)
 729 {
 730         int rw = (private ? 1 : 0);
 731
 732         return bucket_last_io(c, bucket(ca, b), rw);
 733 }
 734
 735 static unsigned bucket_sectors_used_fn(struct bch_fs *c, struct bch_dev *ca,
 736                                        size_t b, void *private)
 737 {
 738         struct bucket *g = bucket(ca, b);
 739         return bucket_sectors_used(g->mark);
 740 }
 741
 742 static unsigned bucket_oldest_gen_fn(struct bch_fs *c, struct bch_dev *ca,
 743                                      size_t b, void *private)
 744 {
 745         return bucket_gc_gen(ca, b);
 746 }
 747
 748 static int unsigned_cmp(const void *_l, const void *_r)
 749 {
 750         unsigned l = *((unsigned *) _l);
 751         unsigned r = *((unsigned *) _r);
 752
 753         return (l > r) - (l < r);
 754 }
 755
 756 static ssize_t show_quantiles(struct bch_fs *c, struct bch_dev *ca,
 757                               char *buf, bucket_map_fn *fn, void *private)
 758 {
 759         size_t i, n;
 760         /* Compute 31 quantiles */
 761         unsigned q[31], *p;
 762         ssize_t ret = 0;
 763
 764         down_read(&ca->bucket_lock);
 765         n = ca->mi.nbuckets;
 766
 767         p = vzalloc(n * sizeof(unsigned));
 768         if (!p) {
 769                 up_read(&ca->bucket_lock);
 770                 return -ENOMEM;
 771         }
 772
 773         for (i = ca->mi.first_bucket; i < n; i++)
 774                 p[i] = fn(c, ca, i, private);
 775
 776         sort(p, n, sizeof(unsigned), unsigned_cmp, NULL);
 777         up_read(&ca->bucket_lock);
 778
 779         while (n &&
 780                !p[n - 1])
 781                 --n;
 782
 783         for (i = 0; i < ARRAY_SIZE(q); i++)
 784                 q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
 785
 786         vfree(p);
 787
 788         for (i = 0; i < ARRAY_SIZE(q); i++)
 789                 ret += scnprintf(buf + ret, PAGE_SIZE - ret,
 790                                  "%u ", q[i]);
 791         buf[ret - 1] = '\n';
 792
 793         return ret;
 794 }
 795
 796 static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
 797 {
 798         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 799         enum alloc_reserve i;
 800
 801         spin_lock(&ca->freelist_lock);
 802
 803         pr_buf(&out, "free_inc:\t%zu\t%zu\n",
 804                fifo_used(&ca->free_inc),
 805                ca->free_inc.size);
 806
 807         for (i = 0; i < RESERVE_NR; i++)
 808                 pr_buf(&out, "free[%u]:\t%zu\t%zu\n", i,
 809                        fifo_used(&ca->free[i]),
 810                        ca->free[i].size);
 811
 812         spin_unlock(&ca->freelist_lock);
 813
 814         return out.pos - buf;
 815 }
 816
 817 static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
 818 {
 819         struct bch_fs *c = ca->fs;
 820         struct bch_dev_usage stats = bch2_dev_usage_read(c, ca);
 821         unsigned i, nr[BCH_DATA_NR];
 822
 823         memset(nr, 0, sizeof(nr));
 824
 825         for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 826                 nr[c->open_buckets[i].type]++;
 827
 828         return scnprintf(buf, PAGE_SIZE,
 829                 "free_inc:               %zu/%zu\n"
 830                 "free[RESERVE_BTREE]:    %zu/%zu\n"
 831                 "free[RESERVE_MOVINGGC]: %zu/%zu\n"
 832                 "free[RESERVE_NONE]:     %zu/%zu\n"
 833                 "buckets:\n"
 834                 "    capacity:           %llu\n"
 835                 "    alloc:              %llu\n"
 836                 "    sb:                 %llu\n"
 837                 "    journal:            %llu\n"
 838                 "    meta:               %llu\n"
 839                 "    user:               %llu\n"
 840                 "    cached:             %llu\n"
 841                 "    erasure coded:      %llu\n"
 842                 "    available:          %lli\n"
 843                 "sectors:\n"
 844                 "    sb:                 %llu\n"
 845                 "    journal:            %llu\n"
 846                 "    meta:               %llu\n"
 847                 "    user:               %llu\n"
 848                 "    cached:             %llu\n"
 849                 "    fragmented:         %llu\n"
 850                 "    copygc threshold:   %llu\n"
 851                 "freelist_wait:          %s\n"
 852                 "open buckets:           %u/%u (reserved %u)\n"
 853                 "open_buckets_wait:      %s\n"
 854                 "open_buckets_btree:     %u\n"
 855                 "open_buckets_user:      %u\n"
 856                 "btree reserve cache:    %u\n",
 857                 fifo_used(&ca->free_inc),               ca->free_inc.size,
 858                 fifo_used(&ca->free[RESERVE_BTREE]),    ca->free[RESERVE_BTREE].size,
 859                 fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size,
 860                 fifo_used(&ca->free[RESERVE_NONE]),     ca->free[RESERVE_NONE].size,
 861                 ca->mi.nbuckets - ca->mi.first_bucket,
 862                 stats.buckets_alloc,
 863                 stats.buckets[BCH_DATA_SB],
 864                 stats.buckets[BCH_DATA_JOURNAL],
 865                 stats.buckets[BCH_DATA_BTREE],
 866                 stats.buckets[BCH_DATA_USER],
 867                 stats.buckets[BCH_DATA_CACHED],
 868                 stats.buckets_ec,
 869                 ca->mi.nbuckets - ca->mi.first_bucket - stats.buckets_unavailable,
 870                 stats.sectors[BCH_DATA_SB],
 871                 stats.sectors[BCH_DATA_JOURNAL],
 872                 stats.sectors[BCH_DATA_BTREE],
 873                 stats.sectors[BCH_DATA_USER],
 874                 stats.sectors[BCH_DATA_CACHED],
 875                 stats.sectors_fragmented,
 876                 ca->copygc_threshold,
 877                 c->freelist_wait.list.first             ? "waiting" : "empty",
 878                 c->open_buckets_nr_free, OPEN_BUCKETS_COUNT,
 879                 BTREE_NODE_OPEN_BUCKET_RESERVE,
 880                 c->open_buckets_wait.list.first         ? "waiting" : "empty",
 881                 nr[BCH_DATA_BTREE],
 882                 nr[BCH_DATA_USER],
 883                 c->btree_reserve_cache_nr);
 884 }
 885
 886 static const char * const bch2_rw[] = {
 887         "read",
 888         "write",
 889         NULL
 890 };
 891
 892 static ssize_t show_dev_iodone(struct bch_dev *ca, char *buf)
 893 {
 894         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 895         int rw, i;
 896
 897         for (rw = 0; rw < 2; rw++) {
 898                 pr_buf(&out, "%s:\n", bch2_rw[rw]);
 899
 900                 for (i = 1; i < BCH_DATA_NR; i++)
 901                         pr_buf(&out, "%-12s:%12llu\n",
 902                                bch2_data_types[i],
 903                                percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
 904         }
 905
 906         return out.pos - buf;
 907 }
 908
 909 SHOW(bch2_dev)
 910 {
 911         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 912         struct bch_fs *c = ca->fs;
 913         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 914
 915         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 916
 917         sysfs_print(bucket_size,        bucket_bytes(ca));
 918         sysfs_print(block_size,         block_bytes(c));
 919         sysfs_print(first_bucket,       ca->mi.first_bucket);
 920         sysfs_print(nbuckets,           ca->mi.nbuckets);
 921         sysfs_print(durability,         ca->mi.durability);
 922         sysfs_print(discard,            ca->mi.discard);
 923
 924         if (attr == &sysfs_label) {
 925                 if (ca->mi.group) {
 926                         mutex_lock(&c->sb_lock);
 927                         bch2_disk_path_to_text(&out, &c->disk_sb,
 928                                                ca->mi.group - 1);
 929                         mutex_unlock(&c->sb_lock);
 930                 } else {
 931                         pr_buf(&out, "none");
 932                 }
 933
 934                 pr_buf(&out, "\n");
 935                 return out.pos - buf;
 936         }
 937
 938         if (attr == &sysfs_has_data) {
 939                 bch2_flags_to_text(&out, bch2_data_types,
 940                                    bch2_dev_has_data(c, ca));
 941                 pr_buf(&out, "\n");
 942                 return out.pos - buf;
 943         }
 944
 945         sysfs_pd_controller_show(copy_gc, &ca->copygc_pd);
 946
 947         if (attr == &sysfs_cache_replacement_policy) {
 948                 bch2_string_opt_to_text(&out,
 949                                         bch2_cache_replacement_policies,
 950                                         ca->mi.replacement);
 951                 pr_buf(&out, "\n");
 952                 return out.pos - buf;
 953         }
 954
 955         if (attr == &sysfs_state_rw) {
 956                 bch2_string_opt_to_text(&out, bch2_dev_state,
 957                                         ca->mi.state);
 958                 pr_buf(&out, "\n");
 959                 return out.pos - buf;
 960         }
 961
 962         if (attr == &sysfs_iodone)
 963                 return show_dev_iodone(ca, buf);
 964
 965         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 966         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 967
 968         if (attr == &sysfs_io_latency_stats_read)
 969                 return bch2_time_stats_print(&ca->io_latency[READ], buf, PAGE_SIZE);
 970         if (attr == &sysfs_io_latency_stats_write)
 971                 return bch2_time_stats_print(&ca->io_latency[WRITE], buf, PAGE_SIZE);
 972
 973         sysfs_printf(congested,                 "%u%%",
 974                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 975                      * 100 / CONGESTED_MAX);
 976
 977         if (attr == &sysfs_bucket_quantiles_last_read)
 978                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 0);
 979         if (attr == &sysfs_bucket_quantiles_last_write)
 980                 return show_quantiles(c, ca, buf, bucket_last_io_fn, (void *) 1);
 981         if (attr == &sysfs_bucket_quantiles_fragmentation)
 982                 return show_quantiles(c, ca, buf, bucket_sectors_used_fn, NULL);
 983         if (attr == &sysfs_bucket_quantiles_oldest_gen)
 984                 return show_quantiles(c, ca, buf, bucket_oldest_gen_fn, NULL);
 985
 986         if (attr == &sysfs_reserve_stats)
 987                 return show_reserve_stats(ca, buf);
 988         if (attr == &sysfs_alloc_debug)
 989                 return show_dev_alloc_debug(ca, buf);
 990
 991         return 0;
 992 }
 993
 994 STORE(bch2_dev)
 995 {
 996         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 997         struct bch_fs *c = ca->fs;
 998         struct bch_member *mi;
 999
1000         sysfs_pd_controller_store(copy_gc, &ca->copygc_pd);
1001
1002         if (attr == &sysfs_discard) {
1003                 bool v = strtoul_or_return(buf);
1004
1005                 mutex_lock(&c->sb_lock);
1006                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
1007
1008                 if (v != BCH_MEMBER_DISCARD(mi)) {
1009                         SET_BCH_MEMBER_DISCARD(mi, v);
1010                         bch2_write_super(c);
1011                 }
1012                 mutex_unlock(&c->sb_lock);
1013         }
1014
1015         if (attr == &sysfs_cache_replacement_policy) {
1016                 ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
1017
1018                 if (v < 0)
1019                         return v;
1020
1021                 mutex_lock(&c->sb_lock);
1022                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
1023
1024                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
1025                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
1026                         bch2_write_super(c);
1027                 }
1028                 mutex_unlock(&c->sb_lock);
1029         }
1030
1031         if (attr == &sysfs_label) {
1032                 char *tmp;
1033                 int ret;
1034
1035                 tmp = kstrdup(buf, GFP_KERNEL);
1036                 if (!tmp)
1037                         return -ENOMEM;
1038
1039                 ret = bch2_dev_group_set(c, ca, strim(tmp));
1040                 kfree(tmp);
1041                 if (ret)
1042                         return ret;
1043         }
1044
1045         if (attr == &sysfs_wake_allocator)
1046                 bch2_wake_allocator(ca);
1047
1048         return size;
1049 }
1050 SYSFS_OPS(bch2_dev);
1051
1052 struct attribute *bch2_dev_files[] = {
1053         &sysfs_uuid,
1054         &sysfs_bucket_size,
1055         &sysfs_block_size,
1056         &sysfs_first_bucket,
1057         &sysfs_nbuckets,
1058         &sysfs_durability,
1059
1060         /* settings: */
1061         &sysfs_discard,
1062         &sysfs_cache_replacement_policy,
1063         &sysfs_state_rw,
1064         &sysfs_label,
1065
1066         &sysfs_has_data,
1067         &sysfs_iodone,
1068
1069         &sysfs_io_latency_read,
1070         &sysfs_io_latency_write,
1071         &sysfs_io_latency_stats_read,
1072         &sysfs_io_latency_stats_write,
1073         &sysfs_congested,
1074
1075         /* alloc info - other stats: */
1076         &sysfs_bucket_quantiles_last_read,
1077         &sysfs_bucket_quantiles_last_write,
1078         &sysfs_bucket_quantiles_fragmentation,
1079         &sysfs_bucket_quantiles_oldest_gen,
1080
1081         &sysfs_reserve_stats,
1082
1083         /* debug: */
1084         &sysfs_alloc_debug,
1085         &sysfs_wake_allocator,
1086
1087         sysfs_pd_controller_files(copy_gc),
1088         NULL
1089 };
1090
1091 #endif  /* _BCACHEFS_SYSFS_H_ */