fs/bcachefs/btree_key_cache.c

   1 // SPDX-License-Identifier: GPL-2.0
   2
   3 #include "bcachefs.h"
   4 #include "btree_cache.h"
   5 #include "btree_iter.h"
   6 #include "btree_key_cache.h"
   7 #include "btree_locking.h"
   8 #include "btree_update.h"
   9 #include "errcode.h"
  10 #include "error.h"
  11 #include "journal.h"
  12 #include "journal_reclaim.h"
  13 #include "trace.h"
  14
  15 #include <linux/sched/mm.h>
  16
  17 static inline bool btree_uses_pcpu_readers(enum btree_id id)
  18 {
  19         return id == BTREE_ID_subvolumes;
  20 }
  21
  22 static struct kmem_cache *bch2_key_cache;
  23
  24 static int bch2_btree_key_cache_cmp_fn(struct rhashtable_compare_arg *arg,
  25                                        const void *obj)
  26 {
  27         const struct bkey_cached *ck = obj;
  28         const struct bkey_cached_key *key = arg->key;
  29
  30         return ck->key.btree_id != key->btree_id ||
  31                 !bpos_eq(ck->key.pos, key->pos);
  32 }
  33
  34 static const struct rhashtable_params bch2_btree_key_cache_params = {
  35         .head_offset    = offsetof(struct bkey_cached, hash),
  36         .key_offset     = offsetof(struct bkey_cached, key),
  37         .key_len        = sizeof(struct bkey_cached_key),
  38         .obj_cmpfn      = bch2_btree_key_cache_cmp_fn,
  39 };
  40
  41 __flatten
  42 inline struct bkey_cached *
  43 bch2_btree_key_cache_find(struct bch_fs *c, enum btree_id btree_id, struct bpos pos)
  44 {
  45         struct bkey_cached_key key = {
  46                 .btree_id       = btree_id,
  47                 .pos            = pos,
  48         };
  49
  50         return rhashtable_lookup_fast(&c->btree_key_cache.table, &key,
  51                                       bch2_btree_key_cache_params);
  52 }
  53
  54 static bool bkey_cached_lock_for_evict(struct bkey_cached *ck)
  55 {
  56         if (!six_trylock_intent(&ck->c.lock))
  57                 return false;
  58
  59         if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
  60                 six_unlock_intent(&ck->c.lock);
  61                 return false;
  62         }
  63
  64         if (!six_trylock_write(&ck->c.lock)) {
  65                 six_unlock_intent(&ck->c.lock);
  66                 return false;
  67         }
  68
  69         return true;
  70 }
  71
  72 static void bkey_cached_evict(struct btree_key_cache *c,
  73                               struct bkey_cached *ck)
  74 {
  75         BUG_ON(rhashtable_remove_fast(&c->table, &ck->hash,
  76                                       bch2_btree_key_cache_params));
  77         memset(&ck->key, ~0, sizeof(ck->key));
  78
  79         atomic_long_dec(&c->nr_keys);
  80 }
  81
  82 static void bkey_cached_free(struct btree_key_cache *bc,
  83                              struct bkey_cached *ck)
  84 {
  85         struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
  86
  87         BUG_ON(test_bit(BKEY_CACHED_DIRTY, &ck->flags));
  88
  89         ck->btree_trans_barrier_seq =
  90                 start_poll_synchronize_srcu(&c->btree_trans_barrier);
  91
  92         if (ck->c.lock.readers) {
  93                 list_move_tail(&ck->list, &bc->freed_pcpu);
  94                 bc->nr_freed_pcpu++;
  95         } else {
  96                 list_move_tail(&ck->list, &bc->freed_nonpcpu);
  97                 bc->nr_freed_nonpcpu++;
  98         }
  99         atomic_long_inc(&bc->nr_freed);
 100
 101         kfree(ck->k);
 102         ck->k           = NULL;
 103         ck->u64s        = 0;
 104
 105         six_unlock_write(&ck->c.lock);
 106         six_unlock_intent(&ck->c.lock);
 107 }
 108
 109 #ifdef __KERNEL__
 110 static void __bkey_cached_move_to_freelist_ordered(struct btree_key_cache *bc,
 111                                                    struct bkey_cached *ck)
 112 {
 113         struct bkey_cached *pos;
 114
 115         bc->nr_freed_nonpcpu++;
 116
 117         list_for_each_entry_reverse(pos, &bc->freed_nonpcpu, list) {
 118                 if (ULONG_CMP_GE(ck->btree_trans_barrier_seq,
 119                                  pos->btree_trans_barrier_seq)) {
 120                         list_move(&ck->list, &pos->list);
 121                         return;
 122                 }
 123         }
 124
 125         list_move(&ck->list, &bc->freed_nonpcpu);
 126 }
 127 #endif
 128
 129 static void bkey_cached_move_to_freelist(struct btree_key_cache *bc,
 130                                          struct bkey_cached *ck)
 131 {
 132         BUG_ON(test_bit(BKEY_CACHED_DIRTY, &ck->flags));
 133
 134         if (!ck->c.lock.readers) {
 135 #ifdef __KERNEL__
 136                 struct btree_key_cache_freelist *f;
 137                 bool freed = false;
 138
 139                 preempt_disable();
 140                 f = this_cpu_ptr(bc->pcpu_freed);
 141
 142                 if (f->nr < ARRAY_SIZE(f->objs)) {
 143                         f->objs[f->nr++] = ck;
 144                         freed = true;
 145                 }
 146                 preempt_enable();
 147
 148                 if (!freed) {
 149                         mutex_lock(&bc->lock);
 150                         preempt_disable();
 151                         f = this_cpu_ptr(bc->pcpu_freed);
 152
 153                         while (f->nr > ARRAY_SIZE(f->objs) / 2) {
 154                                 struct bkey_cached *ck2 = f->objs[--f->nr];
 155
 156                                 __bkey_cached_move_to_freelist_ordered(bc, ck2);
 157                         }
 158                         preempt_enable();
 159
 160                         __bkey_cached_move_to_freelist_ordered(bc, ck);
 161                         mutex_unlock(&bc->lock);
 162                 }
 163 #else
 164                 mutex_lock(&bc->lock);
 165                 list_move_tail(&ck->list, &bc->freed_nonpcpu);
 166                 bc->nr_freed_nonpcpu++;
 167                 mutex_unlock(&bc->lock);
 168 #endif
 169         } else {
 170                 mutex_lock(&bc->lock);
 171                 list_move_tail(&ck->list, &bc->freed_pcpu);
 172                 mutex_unlock(&bc->lock);
 173         }
 174 }
 175
 176 static void bkey_cached_free_fast(struct btree_key_cache *bc,
 177                                   struct bkey_cached *ck)
 178 {
 179         struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
 180
 181         ck->btree_trans_barrier_seq =
 182                 start_poll_synchronize_srcu(&c->btree_trans_barrier);
 183
 184         list_del_init(&ck->list);
 185         atomic_long_inc(&bc->nr_freed);
 186
 187         kfree(ck->k);
 188         ck->k           = NULL;
 189         ck->u64s        = 0;
 190
 191         bkey_cached_move_to_freelist(bc, ck);
 192
 193         six_unlock_write(&ck->c.lock);
 194         six_unlock_intent(&ck->c.lock);
 195 }
 196
 197 static struct bkey_cached *
 198 bkey_cached_alloc(struct btree_trans *trans, struct btree_path *path,
 199                   bool *was_new)
 200 {
 201         struct bch_fs *c = trans->c;
 202         struct btree_key_cache *bc = &c->btree_key_cache;
 203         struct bkey_cached *ck = NULL;
 204         bool pcpu_readers = btree_uses_pcpu_readers(path->btree_id);
 205         int ret;
 206
 207         if (!pcpu_readers) {
 208 #ifdef __KERNEL__
 209                 struct btree_key_cache_freelist *f;
 210
 211                 preempt_disable();
 212                 f = this_cpu_ptr(bc->pcpu_freed);
 213                 if (f->nr)
 214                         ck = f->objs[--f->nr];
 215                 preempt_enable();
 216
 217                 if (!ck) {
 218                         mutex_lock(&bc->lock);
 219                         preempt_disable();
 220                         f = this_cpu_ptr(bc->pcpu_freed);
 221
 222                         while (!list_empty(&bc->freed_nonpcpu) &&
 223                                f->nr < ARRAY_SIZE(f->objs) / 2) {
 224                                 ck = list_last_entry(&bc->freed_nonpcpu, struct bkey_cached, list);
 225                                 list_del_init(&ck->list);
 226                                 bc->nr_freed_nonpcpu--;
 227                                 f->objs[f->nr++] = ck;
 228                         }
 229
 230                         ck = f->nr ? f->objs[--f->nr] : NULL;
 231                         preempt_enable();
 232                         mutex_unlock(&bc->lock);
 233                 }
 234 #else
 235                 mutex_lock(&bc->lock);
 236                 if (!list_empty(&bc->freed_nonpcpu)) {
 237                         ck = list_last_entry(&bc->freed_nonpcpu, struct bkey_cached, list);
 238                         list_del_init(&ck->list);
 239                         bc->nr_freed_nonpcpu--;
 240                 }
 241                 mutex_unlock(&bc->lock);
 242 #endif
 243         } else {
 244                 mutex_lock(&bc->lock);
 245                 if (!list_empty(&bc->freed_pcpu)) {
 246                         ck = list_last_entry(&bc->freed_pcpu, struct bkey_cached, list);
 247                         list_del_init(&ck->list);
 248                 }
 249                 mutex_unlock(&bc->lock);
 250         }
 251
 252         if (ck) {
 253                 ret = btree_node_lock_nopath(trans, &ck->c, SIX_LOCK_intent, _THIS_IP_);
 254                 if (unlikely(ret)) {
 255                         bkey_cached_move_to_freelist(bc, ck);
 256                         return ERR_PTR(ret);
 257                 }
 258
 259                 path->l[0].b = (void *) ck;
 260                 path->l[0].lock_seq = six_lock_seq(&ck->c.lock);
 261                 mark_btree_node_locked(trans, path, 0, BTREE_NODE_INTENT_LOCKED);
 262
 263                 ret = bch2_btree_node_lock_write(trans, path, &ck->c);
 264                 if (unlikely(ret)) {
 265                         btree_node_unlock(trans, path, 0);
 266                         bkey_cached_move_to_freelist(bc, ck);
 267                         return ERR_PTR(ret);
 268                 }
 269
 270                 return ck;
 271         }
 272
 273         ck = allocate_dropping_locks(trans, ret,
 274                         kmem_cache_zalloc(bch2_key_cache, _gfp));
 275         if (ret) {
 276                 kmem_cache_free(bch2_key_cache, ck);
 277                 return ERR_PTR(ret);
 278         }
 279
 280         if (!ck)
 281                 return NULL;
 282
 283         INIT_LIST_HEAD(&ck->list);
 284         bch2_btree_lock_init(&ck->c, pcpu_readers ? SIX_LOCK_INIT_PCPU : 0);
 285
 286         ck->c.cached = true;
 287         BUG_ON(!six_trylock_intent(&ck->c.lock));
 288         BUG_ON(!six_trylock_write(&ck->c.lock));
 289         *was_new = true;
 290         return ck;
 291 }
 292
 293 static struct bkey_cached *
 294 bkey_cached_reuse(struct btree_key_cache *c)
 295 {
 296         struct bucket_table *tbl;
 297         struct rhash_head *pos;
 298         struct bkey_cached *ck;
 299         unsigned i;
 300
 301         mutex_lock(&c->lock);
 302         rcu_read_lock();
 303         tbl = rht_dereference_rcu(c->table.tbl, &c->table);
 304         for (i = 0; i < tbl->size; i++)
 305                 rht_for_each_entry_rcu(ck, pos, tbl, i, hash) {
 306                         if (!test_bit(BKEY_CACHED_DIRTY, &ck->flags) &&
 307                             bkey_cached_lock_for_evict(ck)) {
 308                                 bkey_cached_evict(c, ck);
 309                                 goto out;
 310                         }
 311                 }
 312         ck = NULL;
 313 out:
 314         rcu_read_unlock();
 315         mutex_unlock(&c->lock);
 316         return ck;
 317 }
 318
 319 static struct bkey_cached *
 320 btree_key_cache_create(struct btree_trans *trans, struct btree_path *path)
 321 {
 322         struct bch_fs *c = trans->c;
 323         struct btree_key_cache *bc = &c->btree_key_cache;
 324         struct bkey_cached *ck;
 325         bool was_new = false;
 326
 327         ck = bkey_cached_alloc(trans, path, &was_new);
 328         if (IS_ERR(ck))
 329                 return ck;
 330
 331         if (unlikely(!ck)) {
 332                 ck = bkey_cached_reuse(bc);
 333                 if (unlikely(!ck)) {
 334                         bch_err(c, "error allocating memory for key cache item, btree %s",
 335                                 bch2_btree_id_str(path->btree_id));
 336                         return ERR_PTR(-BCH_ERR_ENOMEM_btree_key_cache_create);
 337                 }
 338
 339                 mark_btree_node_locked(trans, path, 0, BTREE_NODE_INTENT_LOCKED);
 340         }
 341
 342         ck->c.level             = 0;
 343         ck->c.btree_id          = path->btree_id;
 344         ck->key.btree_id        = path->btree_id;
 345         ck->key.pos             = path->pos;
 346         ck->valid               = false;
 347         ck->flags               = 1U << BKEY_CACHED_ACCESSED;
 348
 349         if (unlikely(rhashtable_lookup_insert_fast(&bc->table,
 350                                           &ck->hash,
 351                                           bch2_btree_key_cache_params))) {
 352                 /* We raced with another fill: */
 353
 354                 if (likely(was_new)) {
 355                         six_unlock_write(&ck->c.lock);
 356                         six_unlock_intent(&ck->c.lock);
 357                         kfree(ck);
 358                 } else {
 359                         bkey_cached_free_fast(bc, ck);
 360                 }
 361
 362                 mark_btree_node_locked(trans, path, 0, BTREE_NODE_UNLOCKED);
 363                 return NULL;
 364         }
 365
 366         atomic_long_inc(&bc->nr_keys);
 367
 368         six_unlock_write(&ck->c.lock);
 369
 370         return ck;
 371 }
 372
 373 static int btree_key_cache_fill(struct btree_trans *trans,
 374                                 struct btree_path *ck_path,
 375                                 struct bkey_cached *ck)
 376 {
 377         struct btree_iter iter;
 378         struct bkey_s_c k;
 379         unsigned new_u64s = 0;
 380         struct bkey_i *new_k = NULL;
 381         int ret;
 382
 383         bch2_trans_iter_init(trans, &iter, ck->key.btree_id, ck->key.pos,
 384                              BTREE_ITER_KEY_CACHE_FILL|
 385                              BTREE_ITER_CACHED_NOFILL);
 386         iter.flags &= ~BTREE_ITER_WITH_JOURNAL;
 387         k = bch2_btree_iter_peek_slot(&iter);
 388         ret = bkey_err(k);
 389         if (ret)
 390                 goto err;
 391
 392         if (!bch2_btree_node_relock(trans, ck_path, 0)) {
 393                 trace_and_count(trans->c, trans_restart_relock_key_cache_fill, trans, _THIS_IP_, ck_path);
 394                 ret = btree_trans_restart(trans, BCH_ERR_transaction_restart_key_cache_fill);
 395                 goto err;
 396         }
 397
 398         /*
 399          * bch2_varint_decode can read past the end of the buffer by at
 400          * most 7 bytes (it won't be used):
 401          */
 402         new_u64s = k.k->u64s + 1;
 403
 404         /*
 405          * Allocate some extra space so that the transaction commit path is less
 406          * likely to have to reallocate, since that requires a transaction
 407          * restart:
 408          */
 409         new_u64s = min(256U, (new_u64s * 3) / 2);
 410
 411         if (new_u64s > ck->u64s) {
 412                 new_u64s = roundup_pow_of_two(new_u64s);
 413                 new_k = kmalloc(new_u64s * sizeof(u64), GFP_NOWAIT|__GFP_NOWARN);
 414                 if (!new_k) {
 415                         bch2_trans_unlock(trans);
 416
 417                         new_k = kmalloc(new_u64s * sizeof(u64), GFP_KERNEL);
 418                         if (!new_k) {
 419                                 bch_err(trans->c, "error allocating memory for key cache key, btree %s u64s %u",
 420                                         bch2_btree_id_str(ck->key.btree_id), new_u64s);
 421                                 ret = -BCH_ERR_ENOMEM_btree_key_cache_fill;
 422                                 goto err;
 423                         }
 424
 425                         if (!bch2_btree_node_relock(trans, ck_path, 0)) {
 426                                 kfree(new_k);
 427                                 trace_and_count(trans->c, trans_restart_relock_key_cache_fill, trans, _THIS_IP_, ck_path);
 428                                 ret = btree_trans_restart(trans, BCH_ERR_transaction_restart_key_cache_fill);
 429                                 goto err;
 430                         }
 431
 432                         ret = bch2_trans_relock(trans);
 433                         if (ret) {
 434                                 kfree(new_k);
 435                                 goto err;
 436                         }
 437                 }
 438         }
 439
 440         ret = bch2_btree_node_lock_write(trans, ck_path, &ck_path->l[0].b->c);
 441         if (ret) {
 442                 kfree(new_k);
 443                 goto err;
 444         }
 445
 446         if (new_k) {
 447                 kfree(ck->k);
 448                 ck->u64s = new_u64s;
 449                 ck->k = new_k;
 450         }
 451
 452         bkey_reassemble(ck->k, k);
 453         ck->valid = true;
 454         bch2_btree_node_unlock_write(trans, ck_path, ck_path->l[0].b);
 455
 456         /* We're not likely to need this iterator again: */
 457         set_btree_iter_dontneed(&iter);
 458 err:
 459         bch2_trans_iter_exit(trans, &iter);
 460         return ret;
 461 }
 462
 463 static noinline int
 464 bch2_btree_path_traverse_cached_slowpath(struct btree_trans *trans, struct btree_path *path,
 465                                          unsigned flags)
 466 {
 467         struct bch_fs *c = trans->c;
 468         struct bkey_cached *ck;
 469         int ret = 0;
 470
 471         BUG_ON(path->level);
 472
 473         path->l[1].b = NULL;
 474
 475         if (bch2_btree_node_relock_notrace(trans, path, 0)) {
 476                 ck = (void *) path->l[0].b;
 477                 goto fill;
 478         }
 479 retry:
 480         ck = bch2_btree_key_cache_find(c, path->btree_id, path->pos);
 481         if (!ck) {
 482                 ck = btree_key_cache_create(trans, path);
 483                 ret = PTR_ERR_OR_ZERO(ck);
 484                 if (ret)
 485                         goto err;
 486                 if (!ck)
 487                         goto retry;
 488
 489                 mark_btree_node_locked(trans, path, 0, BTREE_NODE_INTENT_LOCKED);
 490                 path->locks_want = 1;
 491         } else {
 492                 enum six_lock_type lock_want = __btree_lock_want(path, 0);
 493
 494                 ret = btree_node_lock(trans, path, (void *) ck, 0,
 495                                       lock_want, _THIS_IP_);
 496                 if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
 497                         goto err;
 498
 499                 BUG_ON(ret);
 500
 501                 if (ck->key.btree_id != path->btree_id ||
 502                     !bpos_eq(ck->key.pos, path->pos)) {
 503                         six_unlock_type(&ck->c.lock, lock_want);
 504                         goto retry;
 505                 }
 506
 507                 mark_btree_node_locked(trans, path, 0,
 508                                        (enum btree_node_locked_type) lock_want);
 509         }
 510
 511         path->l[0].lock_seq     = six_lock_seq(&ck->c.lock);
 512         path->l[0].b            = (void *) ck;
 513 fill:
 514         path->uptodate = BTREE_ITER_UPTODATE;
 515
 516         if (!ck->valid && !(flags & BTREE_ITER_CACHED_NOFILL)) {
 517                 /*
 518                  * Using the underscore version because we haven't set
 519                  * path->uptodate yet:
 520                  */
 521                 if (!path->locks_want &&
 522                     !__bch2_btree_path_upgrade(trans, path, 1, NULL)) {
 523                         trace_and_count(trans->c, trans_restart_key_cache_upgrade, trans, _THIS_IP_);
 524                         ret = btree_trans_restart(trans, BCH_ERR_transaction_restart_key_cache_upgrade);
 525                         goto err;
 526                 }
 527
 528                 ret = btree_key_cache_fill(trans, path, ck);
 529                 if (ret)
 530                         goto err;
 531
 532                 ret = bch2_btree_path_relock(trans, path, _THIS_IP_);
 533                 if (ret)
 534                         goto err;
 535
 536                 path->uptodate = BTREE_ITER_UPTODATE;
 537         }
 538
 539         if (!test_bit(BKEY_CACHED_ACCESSED, &ck->flags))
 540                 set_bit(BKEY_CACHED_ACCESSED, &ck->flags);
 541
 542         BUG_ON(btree_node_locked_type(path, 0) != btree_lock_want(path, 0));
 543         BUG_ON(path->uptodate);
 544
 545         return ret;
 546 err:
 547         path->uptodate = BTREE_ITER_NEED_TRAVERSE;
 548         if (!bch2_err_matches(ret, BCH_ERR_transaction_restart)) {
 549                 btree_node_unlock(trans, path, 0);
 550                 path->l[0].b = ERR_PTR(ret);
 551         }
 552         return ret;
 553 }
 554
 555 int bch2_btree_path_traverse_cached(struct btree_trans *trans, struct btree_path *path,
 556                                     unsigned flags)
 557 {
 558         struct bch_fs *c = trans->c;
 559         struct bkey_cached *ck;
 560         int ret = 0;
 561
 562         EBUG_ON(path->level);
 563
 564         path->l[1].b = NULL;
 565
 566         if (bch2_btree_node_relock_notrace(trans, path, 0)) {
 567                 ck = (void *) path->l[0].b;
 568                 goto fill;
 569         }
 570 retry:
 571         ck = bch2_btree_key_cache_find(c, path->btree_id, path->pos);
 572         if (!ck) {
 573                 return bch2_btree_path_traverse_cached_slowpath(trans, path, flags);
 574         } else {
 575                 enum six_lock_type lock_want = __btree_lock_want(path, 0);
 576
 577                 ret = btree_node_lock(trans, path, (void *) ck, 0,
 578                                       lock_want, _THIS_IP_);
 579                 EBUG_ON(ret && !bch2_err_matches(ret, BCH_ERR_transaction_restart));
 580
 581                 if (ret)
 582                         return ret;
 583
 584                 if (ck->key.btree_id != path->btree_id ||
 585                     !bpos_eq(ck->key.pos, path->pos)) {
 586                         six_unlock_type(&ck->c.lock, lock_want);
 587                         goto retry;
 588                 }
 589
 590                 mark_btree_node_locked(trans, path, 0,
 591                                        (enum btree_node_locked_type) lock_want);
 592         }
 593
 594         path->l[0].lock_seq     = six_lock_seq(&ck->c.lock);
 595         path->l[0].b            = (void *) ck;
 596 fill:
 597         if (!ck->valid)
 598                 return bch2_btree_path_traverse_cached_slowpath(trans, path, flags);
 599
 600         if (!test_bit(BKEY_CACHED_ACCESSED, &ck->flags))
 601                 set_bit(BKEY_CACHED_ACCESSED, &ck->flags);
 602
 603         path->uptodate = BTREE_ITER_UPTODATE;
 604         EBUG_ON(!ck->valid);
 605         EBUG_ON(btree_node_locked_type(path, 0) != btree_lock_want(path, 0));
 606
 607         return ret;
 608 }
 609
 610 static int btree_key_cache_flush_pos(struct btree_trans *trans,
 611                                      struct bkey_cached_key key,
 612                                      u64 journal_seq,
 613                                      unsigned commit_flags,
 614                                      bool evict)
 615 {
 616         struct bch_fs *c = trans->c;
 617         struct journal *j = &c->journal;
 618         struct btree_iter c_iter, b_iter;
 619         struct bkey_cached *ck = NULL;
 620         int ret;
 621
 622         bch2_trans_iter_init(trans, &b_iter, key.btree_id, key.pos,
 623                              BTREE_ITER_SLOTS|
 624                              BTREE_ITER_INTENT|
 625                              BTREE_ITER_ALL_SNAPSHOTS);
 626         bch2_trans_iter_init(trans, &c_iter, key.btree_id, key.pos,
 627                              BTREE_ITER_CACHED|
 628                              BTREE_ITER_INTENT);
 629         b_iter.flags &= ~BTREE_ITER_WITH_KEY_CACHE;
 630
 631         ret = bch2_btree_iter_traverse(&c_iter);
 632         if (ret)
 633                 goto out;
 634
 635         ck = (void *) btree_iter_path(trans, &c_iter)->l[0].b;
 636         if (!ck)
 637                 goto out;
 638
 639         if (!test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
 640                 if (evict)
 641                         goto evict;
 642                 goto out;
 643         }
 644
 645         BUG_ON(!ck->valid);
 646
 647         if (journal_seq && ck->journal.seq != journal_seq)
 648                 goto out;
 649
 650         trans->journal_res.seq = ck->journal.seq;
 651
 652         /*
 653          * If we're at the end of the journal, we really want to free up space
 654          * in the journal right away - we don't want to pin that old journal
 655          * sequence number with a new btree node write, we want to re-journal
 656          * the update
 657          */
 658         if (ck->journal.seq == journal_last_seq(j))
 659                 commit_flags |= BCH_WATERMARK_reclaim;
 660
 661         if (ck->journal.seq != journal_last_seq(j) ||
 662             j->watermark == BCH_WATERMARK_stripe)
 663                 commit_flags |= BCH_TRANS_COMMIT_no_journal_res;
 664
 665         ret   = bch2_btree_iter_traverse(&b_iter) ?:
 666                 bch2_trans_update(trans, &b_iter, ck->k,
 667                                   BTREE_UPDATE_KEY_CACHE_RECLAIM|
 668                                   BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|
 669                                   BTREE_TRIGGER_NORUN) ?:
 670                 bch2_trans_commit(trans, NULL, NULL,
 671                                   BCH_TRANS_COMMIT_no_check_rw|
 672                                   BCH_TRANS_COMMIT_no_enospc|
 673                                   commit_flags);
 674
 675         bch2_fs_fatal_err_on(ret &&
 676                              !bch2_err_matches(ret, BCH_ERR_transaction_restart) &&
 677                              !bch2_err_matches(ret, BCH_ERR_journal_reclaim_would_deadlock) &&
 678                              !bch2_journal_error(j), c,
 679                              "flushing key cache: %s", bch2_err_str(ret));
 680         if (ret)
 681                 goto out;
 682
 683         bch2_journal_pin_drop(j, &ck->journal);
 684
 685         struct btree_path *path = btree_iter_path(trans, &c_iter);
 686         BUG_ON(!btree_node_locked(path, 0));
 687
 688         if (!evict) {
 689                 if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
 690                         clear_bit(BKEY_CACHED_DIRTY, &ck->flags);
 691                         atomic_long_dec(&c->btree_key_cache.nr_dirty);
 692                 }
 693         } else {
 694                 struct btree_path *path2;
 695                 unsigned i;
 696 evict:
 697                 trans_for_each_path(trans, path2, i)
 698                         if (path2 != path)
 699                                 __bch2_btree_path_unlock(trans, path2);
 700
 701                 bch2_btree_node_lock_write_nofail(trans, path, &ck->c);
 702
 703                 if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
 704                         clear_bit(BKEY_CACHED_DIRTY, &ck->flags);
 705                         atomic_long_dec(&c->btree_key_cache.nr_dirty);
 706                 }
 707
 708                 mark_btree_node_locked_noreset(path, 0, BTREE_NODE_UNLOCKED);
 709                 bkey_cached_evict(&c->btree_key_cache, ck);
 710                 bkey_cached_free_fast(&c->btree_key_cache, ck);
 711         }
 712 out:
 713         bch2_trans_iter_exit(trans, &b_iter);
 714         bch2_trans_iter_exit(trans, &c_iter);
 715         return ret;
 716 }
 717
 718 int bch2_btree_key_cache_journal_flush(struct journal *j,
 719                                 struct journal_entry_pin *pin, u64 seq)
 720 {
 721         struct bch_fs *c = container_of(j, struct bch_fs, journal);
 722         struct bkey_cached *ck =
 723                 container_of(pin, struct bkey_cached, journal);
 724         struct bkey_cached_key key;
 725         struct btree_trans *trans = bch2_trans_get(c);
 726         int srcu_idx = srcu_read_lock(&c->btree_trans_barrier);
 727         int ret = 0;
 728
 729         btree_node_lock_nopath_nofail(trans, &ck->c, SIX_LOCK_read);
 730         key = ck->key;
 731
 732         if (ck->journal.seq != seq ||
 733             !test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
 734                 six_unlock_read(&ck->c.lock);
 735                 goto unlock;
 736         }
 737
 738         if (ck->seq != seq) {
 739                 bch2_journal_pin_update(&c->journal, ck->seq, &ck->journal,
 740                                         bch2_btree_key_cache_journal_flush);
 741                 six_unlock_read(&ck->c.lock);
 742                 goto unlock;
 743         }
 744         six_unlock_read(&ck->c.lock);
 745
 746         ret = lockrestart_do(trans,
 747                 btree_key_cache_flush_pos(trans, key, seq,
 748                                 BCH_TRANS_COMMIT_journal_reclaim, false));
 749 unlock:
 750         srcu_read_unlock(&c->btree_trans_barrier, srcu_idx);
 751
 752         bch2_trans_put(trans);
 753         return ret;
 754 }
 755
 756 bool bch2_btree_insert_key_cached(struct btree_trans *trans,
 757                                   unsigned flags,
 758                                   struct btree_insert_entry *insert_entry)
 759 {
 760         struct bch_fs *c = trans->c;
 761         struct bkey_cached *ck = (void *) (trans->paths + insert_entry->path)->l[0].b;
 762         struct bkey_i *insert = insert_entry->k;
 763         bool kick_reclaim = false;
 764
 765         BUG_ON(insert->k.u64s > ck->u64s);
 766
 767         bkey_copy(ck->k, insert);
 768         ck->valid = true;
 769
 770         if (!test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
 771                 EBUG_ON(test_bit(BCH_FS_clean_shutdown, &c->flags));
 772                 set_bit(BKEY_CACHED_DIRTY, &ck->flags);
 773                 atomic_long_inc(&c->btree_key_cache.nr_dirty);
 774
 775                 if (bch2_nr_btree_keys_need_flush(c))
 776                         kick_reclaim = true;
 777         }
 778
 779         /*
 780          * To minimize lock contention, we only add the journal pin here and
 781          * defer pin updates to the flush callback via ->seq. Be careful not to
 782          * update ->seq on nojournal commits because we don't want to update the
 783          * pin to a seq that doesn't include journal updates on disk. Otherwise
 784          * we risk losing the update after a crash.
 785          *
 786          * The only exception is if the pin is not active in the first place. We
 787          * have to add the pin because journal reclaim drives key cache
 788          * flushing. The flush callback will not proceed unless ->seq matches
 789          * the latest pin, so make sure it starts with a consistent value.
 790          */
 791         if (!(insert_entry->flags & BTREE_UPDATE_NOJOURNAL) ||
 792             !journal_pin_active(&ck->journal)) {
 793                 ck->seq = trans->journal_res.seq;
 794         }
 795         bch2_journal_pin_add(&c->journal, trans->journal_res.seq,
 796                              &ck->journal, bch2_btree_key_cache_journal_flush);
 797
 798         if (kick_reclaim)
 799                 journal_reclaim_kick(&c->journal);
 800         return true;
 801 }
 802
 803 void bch2_btree_key_cache_drop(struct btree_trans *trans,
 804                                struct btree_path *path)
 805 {
 806         struct bch_fs *c = trans->c;
 807         struct bkey_cached *ck = (void *) path->l[0].b;
 808
 809         BUG_ON(!ck->valid);
 810
 811         /*
 812          * We just did an update to the btree, bypassing the key cache: the key
 813          * cache key is now stale and must be dropped, even if dirty:
 814          */
 815         if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
 816                 clear_bit(BKEY_CACHED_DIRTY, &ck->flags);
 817                 atomic_long_dec(&c->btree_key_cache.nr_dirty);
 818                 bch2_journal_pin_drop(&c->journal, &ck->journal);
 819         }
 820
 821         ck->valid = false;
 822 }
 823
 824 static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
 825                                            struct shrink_control *sc)
 826 {
 827         struct bch_fs *c = shrink->private_data;
 828         struct btree_key_cache *bc = &c->btree_key_cache;
 829         struct bucket_table *tbl;
 830         struct bkey_cached *ck, *t;
 831         size_t scanned = 0, freed = 0, nr = sc->nr_to_scan;
 832         unsigned start, flags;
 833         int srcu_idx;
 834
 835         mutex_lock(&bc->lock);
 836         srcu_idx = srcu_read_lock(&c->btree_trans_barrier);
 837         flags = memalloc_nofs_save();
 838
 839         /*
 840          * Newest freed entries are at the end of the list - once we hit one
 841          * that's too new to be freed, we can bail out:
 842          */
 843         scanned += bc->nr_freed_nonpcpu;
 844
 845         list_for_each_entry_safe(ck, t, &bc->freed_nonpcpu, list) {
 846                 if (!poll_state_synchronize_srcu(&c->btree_trans_barrier,
 847                                                  ck->btree_trans_barrier_seq))
 848                         break;
 849
 850                 list_del(&ck->list);
 851                 six_lock_exit(&ck->c.lock);
 852                 kmem_cache_free(bch2_key_cache, ck);
 853                 atomic_long_dec(&bc->nr_freed);
 854                 freed++;
 855                 bc->nr_freed_nonpcpu--;
 856         }
 857
 858         if (scanned >= nr)
 859                 goto out;
 860
 861         scanned += bc->nr_freed_pcpu;
 862
 863         list_for_each_entry_safe(ck, t, &bc->freed_pcpu, list) {
 864                 if (!poll_state_synchronize_srcu(&c->btree_trans_barrier,
 865                                                  ck->btree_trans_barrier_seq))
 866                         break;
 867
 868                 list_del(&ck->list);
 869                 six_lock_exit(&ck->c.lock);
 870                 kmem_cache_free(bch2_key_cache, ck);
 871                 atomic_long_dec(&bc->nr_freed);
 872                 freed++;
 873                 bc->nr_freed_pcpu--;
 874         }
 875
 876         if (scanned >= nr)
 877                 goto out;
 878
 879         rcu_read_lock();
 880         tbl = rht_dereference_rcu(bc->table.tbl, &bc->table);
 881         if (bc->shrink_iter >= tbl->size)
 882                 bc->shrink_iter = 0;
 883         start = bc->shrink_iter;
 884
 885         do {
 886                 struct rhash_head *pos, *next;
 887
 888                 pos = rht_ptr_rcu(rht_bucket(tbl, bc->shrink_iter));
 889
 890                 while (!rht_is_a_nulls(pos)) {
 891                         next = rht_dereference_bucket_rcu(pos->next, tbl, bc->shrink_iter);
 892                         ck = container_of(pos, struct bkey_cached, hash);
 893
 894                         if (test_bit(BKEY_CACHED_DIRTY, &ck->flags))
 895                                 goto next;
 896
 897                         if (test_bit(BKEY_CACHED_ACCESSED, &ck->flags))
 898                                 clear_bit(BKEY_CACHED_ACCESSED, &ck->flags);
 899                         else if (bkey_cached_lock_for_evict(ck)) {
 900                                 bkey_cached_evict(bc, ck);
 901                                 bkey_cached_free(bc, ck);
 902                         }
 903
 904                         scanned++;
 905                         if (scanned >= nr)
 906                                 break;
 907 next:
 908                         pos = next;
 909                 }
 910
 911                 bc->shrink_iter++;
 912                 if (bc->shrink_iter >= tbl->size)
 913                         bc->shrink_iter = 0;
 914         } while (scanned < nr && bc->shrink_iter != start);
 915
 916         rcu_read_unlock();
 917 out:
 918         memalloc_nofs_restore(flags);
 919         srcu_read_unlock(&c->btree_trans_barrier, srcu_idx);
 920         mutex_unlock(&bc->lock);
 921
 922         return freed;
 923 }
 924
 925 static unsigned long bch2_btree_key_cache_count(struct shrinker *shrink,
 926                                             struct shrink_control *sc)
 927 {
 928         struct bch_fs *c = shrink->private_data;
 929         struct btree_key_cache *bc = &c->btree_key_cache;
 930         long nr = atomic_long_read(&bc->nr_keys) -
 931                 atomic_long_read(&bc->nr_dirty);
 932
 933         return max(0L, nr);
 934 }
 935
 936 void bch2_fs_btree_key_cache_exit(struct btree_key_cache *bc)
 937 {
 938         struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
 939         struct bucket_table *tbl;
 940         struct bkey_cached *ck, *n;
 941         struct rhash_head *pos;
 942         LIST_HEAD(items);
 943         unsigned i;
 944 #ifdef __KERNEL__
 945         int cpu;
 946 #endif
 947
 948         shrinker_free(bc->shrink);
 949
 950         mutex_lock(&bc->lock);
 951
 952         /*
 953          * The loop is needed to guard against racing with rehash:
 954          */
 955         while (atomic_long_read(&bc->nr_keys)) {
 956                 rcu_read_lock();
 957                 tbl = rht_dereference_rcu(bc->table.tbl, &bc->table);
 958                 if (tbl)
 959                         for (i = 0; i < tbl->size; i++)
 960                                 rht_for_each_entry_rcu(ck, pos, tbl, i, hash) {
 961                                         bkey_cached_evict(bc, ck);
 962                                         list_add(&ck->list, &items);
 963                                 }
 964                 rcu_read_unlock();
 965         }
 966
 967 #ifdef __KERNEL__
 968         for_each_possible_cpu(cpu) {
 969                 struct btree_key_cache_freelist *f =
 970                         per_cpu_ptr(bc->pcpu_freed, cpu);
 971
 972                 for (i = 0; i < f->nr; i++) {
 973                         ck = f->objs[i];
 974                         list_add(&ck->list, &items);
 975                 }
 976         }
 977 #endif
 978
 979         BUG_ON(list_count_nodes(&bc->freed_pcpu) != bc->nr_freed_pcpu);
 980         BUG_ON(list_count_nodes(&bc->freed_nonpcpu) != bc->nr_freed_nonpcpu);
 981
 982         list_splice(&bc->freed_pcpu,    &items);
 983         list_splice(&bc->freed_nonpcpu, &items);
 984
 985         mutex_unlock(&bc->lock);
 986
 987         list_for_each_entry_safe(ck, n, &items, list) {
 988                 cond_resched();
 989
 990                 list_del(&ck->list);
 991                 kfree(ck->k);
 992                 six_lock_exit(&ck->c.lock);
 993                 kmem_cache_free(bch2_key_cache, ck);
 994         }
 995
 996         if (atomic_long_read(&bc->nr_dirty) &&
 997             !bch2_journal_error(&c->journal) &&
 998             test_bit(BCH_FS_was_rw, &c->flags))
 999                 panic("btree key cache shutdown error: nr_dirty nonzero (%li)\n",
1000                       atomic_long_read(&bc->nr_dirty));
1001
1002         if (atomic_long_read(&bc->nr_keys))
1003                 panic("btree key cache shutdown error: nr_keys nonzero (%li)\n",
1004                       atomic_long_read(&bc->nr_keys));
1005
1006         if (bc->table_init_done)
1007                 rhashtable_destroy(&bc->table);
1008
1009         free_percpu(bc->pcpu_freed);
1010 }
1011
1012 void bch2_fs_btree_key_cache_init_early(struct btree_key_cache *c)
1013 {
1014         mutex_init(&c->lock);
1015         INIT_LIST_HEAD(&c->freed_pcpu);
1016         INIT_LIST_HEAD(&c->freed_nonpcpu);
1017 }
1018
1019 int bch2_fs_btree_key_cache_init(struct btree_key_cache *bc)
1020 {
1021         struct bch_fs *c = container_of(bc, struct bch_fs, btree_key_cache);
1022         struct shrinker *shrink;
1023
1024 #ifdef __KERNEL__
1025         bc->pcpu_freed = alloc_percpu(struct btree_key_cache_freelist);
1026         if (!bc->pcpu_freed)
1027                 return -BCH_ERR_ENOMEM_fs_btree_cache_init;
1028 #endif
1029
1030         if (rhashtable_init(&bc->table, &bch2_btree_key_cache_params))
1031                 return -BCH_ERR_ENOMEM_fs_btree_cache_init;
1032
1033         bc->table_init_done = true;
1034
1035         shrink = shrinker_alloc(0, "%s-btree_key_cache", c->name);
1036         if (!shrink)
1037                 return -BCH_ERR_ENOMEM_fs_btree_cache_init;
1038         bc->shrink = shrink;
1039         shrink->seeks           = 0;
1040         shrink->count_objects   = bch2_btree_key_cache_count;
1041         shrink->scan_objects    = bch2_btree_key_cache_scan;
1042         shrink->private_data    = c;
1043         shrinker_register(shrink);
1044         return 0;
1045 }
1046
1047 void bch2_btree_key_cache_to_text(struct printbuf *out, struct btree_key_cache *c)
1048 {
1049         prt_printf(out, "nr_freed:\t%lu",       atomic_long_read(&c->nr_freed));
1050         prt_newline(out);
1051         prt_printf(out, "nr_keys:\t%lu",        atomic_long_read(&c->nr_keys));
1052         prt_newline(out);
1053         prt_printf(out, "nr_dirty:\t%lu",       atomic_long_read(&c->nr_dirty));
1054         prt_newline(out);
1055 }
1056
1057 void bch2_btree_key_cache_exit(void)
1058 {
1059         kmem_cache_destroy(bch2_key_cache);
1060 }
1061
1062 int __init bch2_btree_key_cache_init(void)
1063 {
1064         bch2_key_cache = KMEM_CACHE(bkey_cached, SLAB_RECLAIM_ACCOUNT);
1065         if (!bch2_key_cache)
1066                 return -ENOMEM;
1067
1068         return 0;
1069 }