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(bucket_quantiles_last_read);
 159 read_attribute(bucket_quantiles_last_write);
 160 read_attribute(bucket_quantiles_fragmentation);
 161 read_attribute(bucket_quantiles_oldest_gen);
 162
 163 read_attribute(reserve_stats);
 164 read_attribute(btree_cache_size);
 165 read_attribute(compression_stats);
 166 read_attribute(journal_debug);
 167 read_attribute(journal_pins);
 168 read_attribute(btree_updates);
 169 read_attribute(dirty_btree_nodes);
 170 read_attribute(btree_cache);
 171 read_attribute(btree_key_cache);
 172 read_attribute(btree_transactions);
 173 read_attribute(stripes_heap);
 174 read_attribute(open_buckets);
 175
 176 read_attribute(internal_uuid);
 177
 178 read_attribute(has_data);
 179 read_attribute(alloc_debug);
 180 write_attribute(wake_allocator);
 181
 182 read_attribute(read_realloc_races);
 183 read_attribute(extent_migrate_done);
 184 read_attribute(extent_migrate_raced);
 185
 186 rw_attribute(journal_write_delay_ms);
 187 rw_attribute(journal_reclaim_delay_ms);
 188
 189 rw_attribute(discard);
 190 rw_attribute(cache_replacement_policy);
 191 rw_attribute(label);
 192
 193 rw_attribute(copy_gc_enabled);
 194 read_attribute(copy_gc_wait);
 195
 196 rw_attribute(rebalance_enabled);
 197 sysfs_pd_controller_attribute(rebalance);
 198 read_attribute(rebalance_work);
 199 rw_attribute(promote_whole_extents);
 200
 201 read_attribute(new_stripes);
 202
 203 read_attribute(io_timers_read);
 204 read_attribute(io_timers_write);
 205
 206 read_attribute(data_op_data_progress);
 207
 208 #ifdef CONFIG_BCACHEFS_TESTS
 209 write_attribute(perf_test);
 210 #endif /* CONFIG_BCACHEFS_TESTS */
 211
 212 #define x(_name)                                                \
 213         static struct attribute sysfs_time_stat_##_name =               \
 214                 { .name = #_name, .mode = S_IRUGO };
 215         BCH_TIME_STATS()
 216 #undef x
 217
 218 static struct attribute sysfs_state_rw = {
 219         .name = "state",
 220         .mode = S_IRUGO
 221 };
 222
 223 static size_t bch2_btree_cache_size(struct bch_fs *c)
 224 {
 225         size_t ret = 0;
 226         struct btree *b;
 227
 228         mutex_lock(&c->btree_cache.lock);
 229         list_for_each_entry(b, &c->btree_cache.live, list)
 230                 ret += btree_bytes(c);
 231
 232         mutex_unlock(&c->btree_cache.lock);
 233         return ret;
 234 }
 235
 236 static size_t bch2_btree_avg_write_size(struct bch_fs *c)
 237 {
 238         u64 nr = atomic64_read(&c->btree_writes_nr);
 239         u64 sectors = atomic64_read(&c->btree_writes_sectors);
 240
 241         return nr ? div64_u64(sectors, nr) : 0;
 242 }
 243
 244 static long stats_to_text(struct printbuf *out, struct bch_fs *c,
 245                           struct bch_move_stats *stats)
 246 {
 247         pr_buf(out, "%s: data type %s btree_id %s position: ",
 248                 stats->name,
 249                 bch2_data_types[stats->data_type],
 250                 bch2_btree_ids[stats->btree_id]);
 251         bch2_bpos_to_text(out, stats->pos);
 252         pr_buf(out, "%s", "\n");
 253
 254         return 0;
 255 }
 256
 257 static long data_progress_to_text(struct printbuf *out, struct bch_fs *c)
 258 {
 259         long ret = 0;
 260         struct bch_move_stats *iter;
 261
 262         mutex_lock(&c->data_progress_lock);
 263
 264         if (list_empty(&c->data_progress_list))
 265                 pr_buf(out, "%s", "no progress to report\n");
 266         else
 267                 list_for_each_entry(iter, &c->data_progress_list, list) {
 268                         stats_to_text(out, c, iter);
 269                 }
 270
 271         mutex_unlock(&c->data_progress_lock);
 272         return ret;
 273 }
 274
 275 static int fs_alloc_debug_to_text(struct printbuf *out, struct bch_fs *c)
 276 {
 277         struct bch_fs_usage_online *fs_usage = bch2_fs_usage_read(c);
 278
 279         if (!fs_usage)
 280                 return -ENOMEM;
 281
 282         bch2_fs_usage_to_text(out, c, fs_usage);
 283
 284         percpu_up_read(&c->mark_lock);
 285
 286         kfree(fs_usage);
 287         return 0;
 288 }
 289
 290 static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c)
 291 {
 292         struct btree_trans trans;
 293         struct btree_iter iter;
 294         struct bkey_s_c k;
 295         u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
 296             nr_compressed_extents = 0,
 297             compressed_sectors_compressed = 0,
 298             compressed_sectors_uncompressed = 0;
 299         int ret;
 300
 301         if (!test_bit(BCH_FS_STARTED, &c->flags))
 302                 return -EPERM;
 303
 304         bch2_trans_init(&trans, c, 0, 0);
 305
 306         for_each_btree_key(&trans, iter, BTREE_ID_extents, POS_MIN, 0, k, ret)
 307                 if (k.k->type == KEY_TYPE_extent) {
 308                         struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
 309                         const union bch_extent_entry *entry;
 310                         struct extent_ptr_decoded p;
 311
 312                         extent_for_each_ptr_decode(e, p, entry) {
 313                                 if (!crc_is_compressed(p.crc)) {
 314                                         nr_uncompressed_extents++;
 315                                         uncompressed_sectors += e.k->size;
 316                                 } else {
 317                                         nr_compressed_extents++;
 318                                         compressed_sectors_compressed +=
 319                                                 p.crc.compressed_size;
 320                                         compressed_sectors_uncompressed +=
 321                                                 p.crc.uncompressed_size;
 322                                 }
 323
 324                                 /* only looking at the first ptr */
 325                                 break;
 326                         }
 327                 }
 328         bch2_trans_iter_exit(&trans, &iter);
 329
 330         bch2_trans_exit(&trans);
 331         if (ret)
 332                 return ret;
 333
 334         pr_buf(out,
 335                "uncompressed data:\n"
 336                "        nr extents:                     %llu\n"
 337                "        size (bytes):                   %llu\n"
 338                "compressed data:\n"
 339                "        nr extents:                     %llu\n"
 340                "        compressed size (bytes):        %llu\n"
 341                "        uncompressed size (bytes):      %llu\n",
 342                nr_uncompressed_extents,
 343                uncompressed_sectors << 9,
 344                nr_compressed_extents,
 345                compressed_sectors_compressed << 9,
 346                compressed_sectors_uncompressed << 9);
 347         return 0;
 348 }
 349
 350 static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
 351 {
 352         pr_buf(out, "%s: ", bch2_btree_ids[c->gc_gens_btree]);
 353         bch2_bpos_to_text(out, c->gc_gens_pos);
 354         pr_buf(out, "\n");
 355 }
 356
 357 SHOW(bch2_fs)
 358 {
 359         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 360         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 361
 362         sysfs_print(minor,                      c->minor);
 363         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
 364
 365         sysfs_print(journal_write_delay_ms,     c->journal.write_delay_ms);
 366         sysfs_print(journal_reclaim_delay_ms,   c->journal.reclaim_delay_ms);
 367
 368         sysfs_print(block_size,                 block_bytes(c));
 369         sysfs_print(btree_node_size,            btree_bytes(c));
 370         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
 371         sysfs_hprint(btree_avg_write_size,      bch2_btree_avg_write_size(c));
 372
 373         sysfs_print(read_realloc_races,
 374                     atomic_long_read(&c->read_realloc_races));
 375         sysfs_print(extent_migrate_done,
 376                     atomic_long_read(&c->extent_migrate_done));
 377         sysfs_print(extent_migrate_raced,
 378                     atomic_long_read(&c->extent_migrate_raced));
 379
 380         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
 381
 382         if (attr == &sysfs_gc_gens_pos) {
 383                 bch2_gc_gens_pos_to_text(&out, c);
 384                 return out.pos - buf;
 385         }
 386
 387         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
 388
 389         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
 390         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
 391         sysfs_hprint(copy_gc_wait,
 392                      max(0LL, c->copygc_wait -
 393                          atomic64_read(&c->io_clock[WRITE].now)) << 9);
 394
 395         if (attr == &sysfs_rebalance_work) {
 396                 bch2_rebalance_work_to_text(&out, c);
 397                 return out.pos - buf;
 398         }
 399
 400         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
 401
 402         /* Debugging: */
 403
 404         if (attr == &sysfs_alloc_debug)
 405                 return fs_alloc_debug_to_text(&out, c) ?: out.pos - buf;
 406
 407         if (attr == &sysfs_journal_debug) {
 408                 bch2_journal_debug_to_text(&out, &c->journal);
 409                 return out.pos - buf;
 410         }
 411
 412         if (attr == &sysfs_journal_pins) {
 413                 bch2_journal_pins_to_text(&out, &c->journal);
 414                 return out.pos - buf;
 415         }
 416
 417         if (attr == &sysfs_btree_updates) {
 418                 bch2_btree_updates_to_text(&out, c);
 419                 return out.pos - buf;
 420         }
 421
 422         if (attr == &sysfs_dirty_btree_nodes) {
 423                 bch2_dirty_btree_nodes_to_text(&out, c);
 424                 return out.pos - buf;
 425         }
 426
 427         if (attr == &sysfs_btree_cache) {
 428                 bch2_btree_cache_to_text(&out, c);
 429                 return out.pos - buf;
 430         }
 431
 432         if (attr == &sysfs_btree_key_cache) {
 433                 bch2_btree_key_cache_to_text(&out, &c->btree_key_cache);
 434                 return out.pos - buf;
 435         }
 436
 437         if (attr == &sysfs_btree_transactions) {
 438                 bch2_btree_trans_to_text(&out, c);
 439                 return out.pos - buf;
 440         }
 441
 442         if (attr == &sysfs_stripes_heap) {
 443                 bch2_stripes_heap_to_text(&out, c);
 444                 return out.pos - buf;
 445         }
 446
 447         if (attr == &sysfs_open_buckets) {
 448                 bch2_open_buckets_to_text(&out, c);
 449                 return out.pos - buf;
 450         }
 451
 452         if (attr == &sysfs_compression_stats) {
 453                 bch2_compression_stats_to_text(&out, c);
 454                 return out.pos - buf;
 455         }
 456
 457         if (attr == &sysfs_new_stripes) {
 458                 bch2_new_stripes_to_text(&out, c);
 459                 return out.pos - buf;
 460         }
 461
 462         if (attr == &sysfs_io_timers_read) {
 463                 bch2_io_timers_to_text(&out, &c->io_clock[READ]);
 464                 return out.pos - buf;
 465         }
 466         if (attr == &sysfs_io_timers_write) {
 467                 bch2_io_timers_to_text(&out, &c->io_clock[WRITE]);
 468                 return out.pos - buf;
 469         }
 470
 471         if (attr == &sysfs_data_op_data_progress) {
 472                 data_progress_to_text(&out, c);
 473                 return out.pos - buf;
 474         }
 475
 476         return 0;
 477 }
 478
 479 STORE(bch2_fs)
 480 {
 481         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
 482
 483         sysfs_strtoul(journal_write_delay_ms, c->journal.write_delay_ms);
 484         sysfs_strtoul(journal_reclaim_delay_ms, c->journal.reclaim_delay_ms);
 485
 486         if (attr == &sysfs_btree_gc_periodic) {
 487                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
 488                         ?: (ssize_t) size;
 489
 490                 wake_up_process(c->gc_thread);
 491                 return ret;
 492         }
 493
 494         if (attr == &sysfs_copy_gc_enabled) {
 495                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
 496                         ?: (ssize_t) size;
 497
 498                 if (c->copygc_thread)
 499                         wake_up_process(c->copygc_thread);
 500                 return ret;
 501         }
 502
 503         if (attr == &sysfs_rebalance_enabled) {
 504                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
 505                         ?: (ssize_t) size;
 506
 507                 rebalance_wakeup(c);
 508                 return ret;
 509         }
 510
 511         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
 512
 513         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
 514
 515         /* Debugging: */
 516
 517         if (!test_bit(BCH_FS_STARTED, &c->flags))
 518                 return -EPERM;
 519
 520         /* Debugging: */
 521
 522         if (attr == &sysfs_trigger_journal_flush)
 523                 bch2_journal_meta(&c->journal);
 524
 525         if (attr == &sysfs_trigger_gc) {
 526                 /*
 527                  * Full gc is currently incompatible with btree key cache:
 528                  */
 529 #if 0
 530                 down_read(&c->state_lock);
 531                 bch2_gc(c, false, false);
 532                 up_read(&c->state_lock);
 533 #else
 534                 bch2_gc_gens(c);
 535 #endif
 536         }
 537
 538         if (attr == &sysfs_prune_cache) {
 539                 struct shrink_control sc;
 540
 541                 sc.gfp_mask = GFP_KERNEL;
 542                 sc.nr_to_scan = strtoul_or_return(buf);
 543                 c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
 544         }
 545
 546 #ifdef CONFIG_BCACHEFS_TESTS
 547         if (attr == &sysfs_perf_test) {
 548                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
 549                 char *test              = strsep(&p, " \t\n");
 550                 char *nr_str            = strsep(&p, " \t\n");
 551                 char *threads_str       = strsep(&p, " \t\n");
 552                 unsigned threads;
 553                 u64 nr;
 554                 int ret = -EINVAL;
 555
 556                 if (threads_str &&
 557                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
 558                     !(ret = bch2_strtoull_h(nr_str, &nr)))
 559                         ret = bch2_btree_perf_test(c, test, nr, threads);
 560                 kfree(tmp);
 561
 562                 if (ret)
 563                         size = ret;
 564         }
 565 #endif
 566         return size;
 567 }
 568 SYSFS_OPS(bch2_fs);
 569
 570 struct attribute *bch2_fs_files[] = {
 571         &sysfs_minor,
 572         &sysfs_block_size,
 573         &sysfs_btree_node_size,
 574         &sysfs_btree_cache_size,
 575         &sysfs_btree_avg_write_size,
 576
 577         &sysfs_journal_write_delay_ms,
 578         &sysfs_journal_reclaim_delay_ms,
 579
 580         &sysfs_promote_whole_extents,
 581
 582         &sysfs_compression_stats,
 583
 584 #ifdef CONFIG_BCACHEFS_TESTS
 585         &sysfs_perf_test,
 586 #endif
 587         NULL
 588 };
 589
 590 /* internal dir - just a wrapper */
 591
 592 SHOW(bch2_fs_internal)
 593 {
 594         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 595         return bch2_fs_show(&c->kobj, attr, buf);
 596 }
 597
 598 STORE(bch2_fs_internal)
 599 {
 600         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
 601         return bch2_fs_store(&c->kobj, attr, buf, size);
 602 }
 603 SYSFS_OPS(bch2_fs_internal);
 604
 605 struct attribute *bch2_fs_internal_files[] = {
 606         &sysfs_alloc_debug,
 607         &sysfs_journal_debug,
 608         &sysfs_journal_pins,
 609         &sysfs_btree_updates,
 610         &sysfs_dirty_btree_nodes,
 611         &sysfs_btree_cache,
 612         &sysfs_btree_key_cache,
 613         &sysfs_btree_transactions,
 614         &sysfs_stripes_heap,
 615         &sysfs_open_buckets,
 616
 617         &sysfs_read_realloc_races,
 618         &sysfs_extent_migrate_done,
 619         &sysfs_extent_migrate_raced,
 620
 621         &sysfs_trigger_journal_flush,
 622         &sysfs_trigger_gc,
 623         &sysfs_gc_gens_pos,
 624         &sysfs_prune_cache,
 625
 626         &sysfs_copy_gc_enabled,
 627         &sysfs_copy_gc_wait,
 628
 629         &sysfs_rebalance_enabled,
 630         &sysfs_rebalance_work,
 631         sysfs_pd_controller_files(rebalance),
 632
 633         &sysfs_new_stripes,
 634
 635         &sysfs_io_timers_read,
 636         &sysfs_io_timers_write,
 637
 638         &sysfs_data_op_data_progress,
 639
 640         &sysfs_internal_uuid,
 641         NULL
 642 };
 643
 644 /* options */
 645
 646 SHOW(bch2_fs_opts_dir)
 647 {
 648         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 649         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 650         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 651         int id = opt - bch2_opt_table;
 652         u64 v = bch2_opt_get_by_id(&c->opts, id);
 653
 654         bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
 655         pr_buf(&out, "\n");
 656
 657         return out.pos - buf;
 658 }
 659
 660 STORE(bch2_fs_opts_dir)
 661 {
 662         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
 663         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
 664         int ret, id = opt - bch2_opt_table;
 665         char *tmp;
 666         u64 v;
 667
 668         tmp = kstrdup(buf, GFP_KERNEL);
 669         if (!tmp)
 670                 return -ENOMEM;
 671
 672         ret = bch2_opt_parse(c, opt, strim(tmp), &v);
 673         kfree(tmp);
 674
 675         if (ret < 0)
 676                 return ret;
 677
 678         ret = bch2_opt_check_may_set(c, id, v);
 679         if (ret < 0)
 680                 return ret;
 681
 682         if (opt->set_sb != SET_NO_SB_OPT) {
 683                 mutex_lock(&c->sb_lock);
 684                 opt->set_sb(c->disk_sb.sb, v);
 685                 bch2_write_super(c);
 686                 mutex_unlock(&c->sb_lock);
 687         }
 688
 689         bch2_opt_set_by_id(&c->opts, id, v);
 690
 691         if ((id == Opt_background_target ||
 692              id == Opt_background_compression) && v) {
 693                 bch2_rebalance_add_work(c, S64_MAX);
 694                 rebalance_wakeup(c);
 695         }
 696
 697         return size;
 698 }
 699 SYSFS_OPS(bch2_fs_opts_dir);
 700
 701 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
 702
 703 int bch2_opts_create_sysfs_files(struct kobject *kobj)
 704 {
 705         const struct bch_option *i;
 706         int ret;
 707
 708         for (i = bch2_opt_table;
 709              i < bch2_opt_table + bch2_opts_nr;
 710              i++) {
 711                 if (!(i->mode & (OPT_FORMAT|OPT_MOUNT|OPT_RUNTIME)))
 712                         continue;
 713
 714                 ret = sysfs_create_file(kobj, &i->attr);
 715                 if (ret)
 716                         return ret;
 717         }
 718
 719         return 0;
 720 }
 721
 722 /* time stats */
 723
 724 SHOW(bch2_fs_time_stats)
 725 {
 726         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
 727         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 728
 729 #define x(name)                                                         \
 730         if (attr == &sysfs_time_stat_##name) {                          \
 731                 bch2_time_stats_to_text(&out, &c->times[BCH_TIME_##name]);\
 732                 return out.pos - buf;                                   \
 733         }
 734         BCH_TIME_STATS()
 735 #undef x
 736
 737         return 0;
 738 }
 739
 740 STORE(bch2_fs_time_stats)
 741 {
 742         return size;
 743 }
 744 SYSFS_OPS(bch2_fs_time_stats);
 745
 746 struct attribute *bch2_fs_time_stats_files[] = {
 747 #define x(name)                                         \
 748         &sysfs_time_stat_##name,
 749         BCH_TIME_STATS()
 750 #undef x
 751         NULL
 752 };
 753
 754 typedef unsigned (bucket_map_fn)(struct bch_fs *, struct bch_dev *,
 755                                  size_t, void *);
 756
 757 static unsigned bucket_last_io_fn(struct bch_fs *c, struct bch_dev *ca,
 758                                   size_t b, void *private)
 759 {
 760         int rw = (private ? 1 : 0);
 761
 762         return atomic64_read(&c->io_clock[rw].now) - bucket(ca, b)->io_time[rw];
 763 }
 764
 765 static unsigned bucket_sectors_used_fn(struct bch_fs *c, struct bch_dev *ca,
 766                                        size_t b, void *private)
 767 {
 768         struct bucket *g = bucket(ca, b);
 769         return bucket_sectors_used(g->mark);
 770 }
 771
 772 static unsigned bucket_oldest_gen_fn(struct bch_fs *c, struct bch_dev *ca,
 773                                      size_t b, void *private)
 774 {
 775         return bucket_gc_gen(bucket(ca, b));
 776 }
 777
 778 static int unsigned_cmp(const void *_l, const void *_r)
 779 {
 780         const unsigned *l = _l;
 781         const unsigned *r = _r;
 782
 783         return cmp_int(*l, *r);
 784 }
 785
 786 static int quantiles_to_text(struct printbuf *out,
 787                              struct bch_fs *c, struct bch_dev *ca,
 788                              bucket_map_fn *fn, void *private)
 789 {
 790         size_t i, n;
 791         /* Compute 31 quantiles */
 792         unsigned q[31], *p;
 793
 794         down_read(&ca->bucket_lock);
 795         n = ca->mi.nbuckets;
 796
 797         p = vzalloc(n * sizeof(unsigned));
 798         if (!p) {
 799                 up_read(&ca->bucket_lock);
 800                 return -ENOMEM;
 801         }
 802
 803         for (i = ca->mi.first_bucket; i < n; i++)
 804                 p[i] = fn(c, ca, i, private);
 805
 806         sort(p, n, sizeof(unsigned), unsigned_cmp, NULL);
 807         up_read(&ca->bucket_lock);
 808
 809         while (n &&
 810                !p[n - 1])
 811                 --n;
 812
 813         for (i = 0; i < ARRAY_SIZE(q); i++)
 814                 q[i] = p[n * (i + 1) / (ARRAY_SIZE(q) + 1)];
 815
 816         vfree(p);
 817
 818         for (i = 0; i < ARRAY_SIZE(q); i++)
 819                 pr_buf(out, "%u ", q[i]);
 820         pr_buf(out, "\n");
 821         return 0;
 822 }
 823
 824 static void reserve_stats_to_text(struct printbuf *out, struct bch_dev *ca)
 825 {
 826         enum alloc_reserve i;
 827
 828         spin_lock(&ca->fs->freelist_lock);
 829
 830         pr_buf(out, "free_inc:\t%zu\t%zu\n",
 831                fifo_used(&ca->free_inc),
 832                ca->free_inc.size);
 833
 834         for (i = 0; i < RESERVE_NR; i++)
 835                 pr_buf(out, "free[%u]:\t%zu\t%zu\n", i,
 836                        fifo_used(&ca->free[i]),
 837                        ca->free[i].size);
 838
 839         spin_unlock(&ca->fs->freelist_lock);
 840 }
 841
 842 static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
 843 {
 844         struct bch_fs *c = ca->fs;
 845         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
 846         unsigned i, nr[BCH_DATA_NR];
 847
 848         memset(nr, 0, sizeof(nr));
 849
 850         for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
 851                 nr[c->open_buckets[i].type]++;
 852
 853         pr_buf(out,
 854                "\t\t buckets\t sectors      fragmented\n"
 855                "capacity%16llu\n",
 856                ca->mi.nbuckets - ca->mi.first_bucket);
 857
 858         for (i = 1; i < BCH_DATA_NR; i++)
 859                 pr_buf(out, "%-8s%16llu%16llu%16llu\n",
 860                        bch2_data_types[i], stats.d[i].buckets,
 861                        stats.d[i].sectors, stats.d[i].fragmented);
 862
 863         pr_buf(out,
 864                "ec\t%16llu\n"
 865                "available%15llu\n"
 866                "\n"
 867                "free_inc\t\t%zu/%zu\n"
 868                "free[RESERVE_MOVINGGC]\t%zu/%zu\n"
 869                "free[RESERVE_NONE]\t%zu/%zu\n"
 870                "freelist_wait\t\t%s\n"
 871                "open buckets allocated\t%u\n"
 872                "open buckets this dev\t%u\n"
 873                "open buckets total\t%u\n"
 874                "open_buckets_wait\t%s\n"
 875                "open_buckets_btree\t%u\n"
 876                "open_buckets_user\t%u\n"
 877                "btree reserve cache\t%u\n"
 878                "thread state:\t\t%s\n",
 879                stats.buckets_ec,
 880                __dev_buckets_available(ca, stats),
 881                fifo_used(&ca->free_inc),                ca->free_inc.size,
 882                fifo_used(&ca->free[RESERVE_MOVINGGC]),  ca->free[RESERVE_MOVINGGC].size,
 883                fifo_used(&ca->free[RESERVE_NONE]),      ca->free[RESERVE_NONE].size,
 884                c->freelist_wait.list.first              ? "waiting" : "empty",
 885                OPEN_BUCKETS_COUNT - c->open_buckets_nr_free,
 886                ca->nr_open_buckets,
 887                OPEN_BUCKETS_COUNT,
 888                c->open_buckets_wait.list.first          ? "waiting" : "empty",
 889                nr[BCH_DATA_btree],
 890                nr[BCH_DATA_user],
 891                c->btree_reserve_cache_nr,
 892                bch2_allocator_states[ca->allocator_state]);
 893 }
 894
 895 static const char * const bch2_rw[] = {
 896         "read",
 897         "write",
 898         NULL
 899 };
 900
 901 static void dev_iodone_to_text(struct printbuf *out, struct bch_dev *ca)
 902 {
 903         int rw, i;
 904
 905         for (rw = 0; rw < 2; rw++) {
 906                 pr_buf(out, "%s:\n", bch2_rw[rw]);
 907
 908                 for (i = 1; i < BCH_DATA_NR; i++)
 909                         pr_buf(out, "%-12s:%12llu\n",
 910                                bch2_data_types[i],
 911                                percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
 912         }
 913 }
 914
 915 SHOW(bch2_dev)
 916 {
 917         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
 918         struct bch_fs *c = ca->fs;
 919         struct printbuf out = _PBUF(buf, PAGE_SIZE);
 920
 921         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
 922
 923         sysfs_print(bucket_size,        bucket_bytes(ca));
 924         sysfs_print(block_size,         block_bytes(c));
 925         sysfs_print(first_bucket,       ca->mi.first_bucket);
 926         sysfs_print(nbuckets,           ca->mi.nbuckets);
 927         sysfs_print(durability,         ca->mi.durability);
 928         sysfs_print(discard,            ca->mi.discard);
 929
 930         if (attr == &sysfs_label) {
 931                 if (ca->mi.group) {
 932                         mutex_lock(&c->sb_lock);
 933                         bch2_disk_path_to_text(&out, &c->disk_sb,
 934                                                ca->mi.group - 1);
 935                         mutex_unlock(&c->sb_lock);
 936                 }
 937
 938                 pr_buf(&out, "\n");
 939                 return out.pos - buf;
 940         }
 941
 942         if (attr == &sysfs_has_data) {
 943                 bch2_flags_to_text(&out, bch2_data_types,
 944                                    bch2_dev_has_data(c, ca));
 945                 pr_buf(&out, "\n");
 946                 return out.pos - buf;
 947         }
 948
 949         if (attr == &sysfs_cache_replacement_policy) {
 950                 bch2_string_opt_to_text(&out,
 951                                         bch2_cache_replacement_policies,
 952                                         ca->mi.replacement);
 953                 pr_buf(&out, "\n");
 954                 return out.pos - buf;
 955         }
 956
 957         if (attr == &sysfs_state_rw) {
 958                 bch2_string_opt_to_text(&out, bch2_member_states,
 959                                         ca->mi.state);
 960                 pr_buf(&out, "\n");
 961                 return out.pos - buf;
 962         }
 963
 964         if (attr == &sysfs_iodone) {
 965                 dev_iodone_to_text(&out, ca);
 966                 return out.pos - buf;
 967         }
 968
 969         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
 970         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
 971
 972         if (attr == &sysfs_io_latency_stats_read) {
 973                 bch2_time_stats_to_text(&out, &ca->io_latency[READ]);
 974                 return out.pos - buf;
 975         }
 976         if (attr == &sysfs_io_latency_stats_write) {
 977                 bch2_time_stats_to_text(&out, &ca->io_latency[WRITE]);
 978                 return out.pos - buf;
 979         }
 980
 981         sysfs_printf(congested,                 "%u%%",
 982                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
 983                      * 100 / CONGESTED_MAX);
 984
 985         if (attr == &sysfs_bucket_quantiles_last_read)
 986                 return quantiles_to_text(&out, c, ca, bucket_last_io_fn, (void *) 0) ?: out.pos - buf;
 987         if (attr == &sysfs_bucket_quantiles_last_write)
 988                 return quantiles_to_text(&out, c, ca, bucket_last_io_fn, (void *) 1) ?: out.pos - buf;
 989         if (attr == &sysfs_bucket_quantiles_fragmentation)
 990                 return quantiles_to_text(&out, c, ca, bucket_sectors_used_fn, NULL)  ?: out.pos - buf;
 991         if (attr == &sysfs_bucket_quantiles_oldest_gen)
 992                 return quantiles_to_text(&out, c, ca, bucket_oldest_gen_fn, NULL)    ?: out.pos - buf;
 993
 994         if (attr == &sysfs_reserve_stats) {
 995                 reserve_stats_to_text(&out, ca);
 996                 return out.pos - buf;
 997         }
 998         if (attr == &sysfs_alloc_debug) {
 999                 dev_alloc_debug_to_text(&out, ca);
1000                 return out.pos - buf;
1001         }
1002
1003         return 0;
1004 }
1005
1006 STORE(bch2_dev)
1007 {
1008         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
1009         struct bch_fs *c = ca->fs;
1010         struct bch_member *mi;
1011
1012         if (attr == &sysfs_discard) {
1013                 bool v = strtoul_or_return(buf);
1014
1015                 mutex_lock(&c->sb_lock);
1016                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
1017
1018                 if (v != BCH_MEMBER_DISCARD(mi)) {
1019                         SET_BCH_MEMBER_DISCARD(mi, v);
1020                         bch2_write_super(c);
1021                 }
1022                 mutex_unlock(&c->sb_lock);
1023         }
1024
1025         if (attr == &sysfs_cache_replacement_policy) {
1026                 ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
1027
1028                 if (v < 0)
1029                         return v;
1030
1031                 mutex_lock(&c->sb_lock);
1032                 mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
1033
1034                 if ((unsigned) v != BCH_MEMBER_REPLACEMENT(mi)) {
1035                         SET_BCH_MEMBER_REPLACEMENT(mi, v);
1036                         bch2_write_super(c);
1037                 }
1038                 mutex_unlock(&c->sb_lock);
1039         }
1040
1041         if (attr == &sysfs_label) {
1042                 char *tmp;
1043                 int ret;
1044
1045                 tmp = kstrdup(buf, GFP_KERNEL);
1046                 if (!tmp)
1047                         return -ENOMEM;
1048
1049                 ret = bch2_dev_group_set(c, ca, strim(tmp));
1050                 kfree(tmp);
1051                 if (ret)
1052                         return ret;
1053         }
1054
1055         if (attr == &sysfs_wake_allocator)
1056                 bch2_wake_allocator(ca);
1057
1058         return size;
1059 }
1060 SYSFS_OPS(bch2_dev);
1061
1062 struct attribute *bch2_dev_files[] = {
1063         &sysfs_uuid,
1064         &sysfs_bucket_size,
1065         &sysfs_block_size,
1066         &sysfs_first_bucket,
1067         &sysfs_nbuckets,
1068         &sysfs_durability,
1069
1070         /* settings: */
1071         &sysfs_discard,
1072         &sysfs_cache_replacement_policy,
1073         &sysfs_state_rw,
1074         &sysfs_label,
1075
1076         &sysfs_has_data,
1077         &sysfs_iodone,
1078
1079         &sysfs_io_latency_read,
1080         &sysfs_io_latency_write,
1081         &sysfs_io_latency_stats_read,
1082         &sysfs_io_latency_stats_write,
1083         &sysfs_congested,
1084
1085         /* alloc info - other stats: */
1086         &sysfs_bucket_quantiles_last_read,
1087         &sysfs_bucket_quantiles_last_write,
1088         &sysfs_bucket_quantiles_fragmentation,
1089         &sysfs_bucket_quantiles_oldest_gen,
1090
1091         &sysfs_reserve_stats,
1092
1093         /* debug: */
1094         &sysfs_alloc_debug,
1095         &sysfs_wake_allocator,
1096         NULL
1097 };
1098
1099 #endif  /* _BCACHEFS_SYSFS_H_ */