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

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright 2012 Google, Inc.
   4  *
   5  * Foreground allocator code: allocate buckets from freelist, and allocate in
   6  * sector granularity from writepoints.
   7  *
   8  * bch2_bucket_alloc() allocates a single bucket from a specific device.
   9  *
  10  * bch2_bucket_alloc_set() allocates one or more buckets from different devices
  11  * in a given filesystem.
  12  */
  13
  14 #include "bcachefs.h"
  15 #include "alloc_background.h"
  16 #include "alloc_foreground.h"
  17 #include "backpointers.h"
  18 #include "btree_iter.h"
  19 #include "btree_update.h"
  20 #include "btree_gc.h"
  21 #include "buckets.h"
  22 #include "buckets_waiting_for_journal.h"
  23 #include "clock.h"
  24 #include "debug.h"
  25 #include "disk_groups.h"
  26 #include "ec.h"
  27 #include "error.h"
  28 #include "io_write.h"
  29 #include "journal.h"
  30 #include "movinggc.h"
  31 #include "nocow_locking.h"
  32 #include "trace.h"
  33
  34 #include <linux/math64.h>
  35 #include <linux/rculist.h>
  36 #include <linux/rcupdate.h>
  37
  38 static void bch2_trans_mutex_lock_norelock(struct btree_trans *trans,
  39                                            struct mutex *lock)
  40 {
  41         if (!mutex_trylock(lock)) {
  42                 bch2_trans_unlock(trans);
  43                 mutex_lock(lock);
  44         }
  45 }
  46
  47 const char * const bch2_watermarks[] = {
  48 #define x(t) #t,
  49         BCH_WATERMARKS()
  50 #undef x
  51         NULL
  52 };
  53
  54 /*
  55  * Open buckets represent a bucket that's currently being allocated from.  They
  56  * serve two purposes:
  57  *
  58  *  - They track buckets that have been partially allocated, allowing for
  59  *    sub-bucket sized allocations - they're used by the sector allocator below
  60  *
  61  *  - They provide a reference to the buckets they own that mark and sweep GC
  62  *    can find, until the new allocation has a pointer to it inserted into the
  63  *    btree
  64  *
  65  * When allocating some space with the sector allocator, the allocation comes
  66  * with a reference to an open bucket - the caller is required to put that
  67  * reference _after_ doing the index update that makes its allocation reachable.
  68  */
  69
  70 void bch2_reset_alloc_cursors(struct bch_fs *c)
  71 {
  72         struct bch_dev *ca;
  73         unsigned i;
  74
  75         rcu_read_lock();
  76         for_each_member_device_rcu(ca, c, i, NULL)
  77                 ca->alloc_cursor = 0;
  78         rcu_read_unlock();
  79 }
  80
  81 static void bch2_open_bucket_hash_add(struct bch_fs *c, struct open_bucket *ob)
  82 {
  83         open_bucket_idx_t idx = ob - c->open_buckets;
  84         open_bucket_idx_t *slot = open_bucket_hashslot(c, ob->dev, ob->bucket);
  85
  86         ob->hash = *slot;
  87         *slot = idx;
  88 }
  89
  90 static void bch2_open_bucket_hash_remove(struct bch_fs *c, struct open_bucket *ob)
  91 {
  92         open_bucket_idx_t idx = ob - c->open_buckets;
  93         open_bucket_idx_t *slot = open_bucket_hashslot(c, ob->dev, ob->bucket);
  94
  95         while (*slot != idx) {
  96                 BUG_ON(!*slot);
  97                 slot = &c->open_buckets[*slot].hash;
  98         }
  99
 100         *slot = ob->hash;
 101         ob->hash = 0;
 102 }
 103
 104 void __bch2_open_bucket_put(struct bch_fs *c, struct open_bucket *ob)
 105 {
 106         struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
 107
 108         if (ob->ec) {
 109                 ec_stripe_new_put(c, ob->ec, STRIPE_REF_io);
 110                 return;
 111         }
 112
 113         percpu_down_read(&c->mark_lock);
 114         spin_lock(&ob->lock);
 115
 116         ob->valid = false;
 117         ob->data_type = 0;
 118
 119         spin_unlock(&ob->lock);
 120         percpu_up_read(&c->mark_lock);
 121
 122         spin_lock(&c->freelist_lock);
 123         bch2_open_bucket_hash_remove(c, ob);
 124
 125         ob->freelist = c->open_buckets_freelist;
 126         c->open_buckets_freelist = ob - c->open_buckets;
 127
 128         c->open_buckets_nr_free++;
 129         ca->nr_open_buckets--;
 130         spin_unlock(&c->freelist_lock);
 131
 132         closure_wake_up(&c->open_buckets_wait);
 133 }
 134
 135 void bch2_open_bucket_write_error(struct bch_fs *c,
 136                                   struct open_buckets *obs,
 137                                   unsigned dev)
 138 {
 139         struct open_bucket *ob;
 140         unsigned i;
 141
 142         open_bucket_for_each(c, obs, ob, i)
 143                 if (ob->dev == dev && ob->ec)
 144                         bch2_ec_bucket_cancel(c, ob);
 145 }
 146
 147 static struct open_bucket *bch2_open_bucket_alloc(struct bch_fs *c)
 148 {
 149         struct open_bucket *ob;
 150
 151         BUG_ON(!c->open_buckets_freelist || !c->open_buckets_nr_free);
 152
 153         ob = c->open_buckets + c->open_buckets_freelist;
 154         c->open_buckets_freelist = ob->freelist;
 155         atomic_set(&ob->pin, 1);
 156         ob->data_type = 0;
 157
 158         c->open_buckets_nr_free--;
 159         return ob;
 160 }
 161
 162 static void open_bucket_free_unused(struct bch_fs *c, struct open_bucket *ob)
 163 {
 164         BUG_ON(c->open_buckets_partial_nr >=
 165                ARRAY_SIZE(c->open_buckets_partial));
 166
 167         spin_lock(&c->freelist_lock);
 168         ob->on_partial_list = true;
 169         c->open_buckets_partial[c->open_buckets_partial_nr++] =
 170                 ob - c->open_buckets;
 171         spin_unlock(&c->freelist_lock);
 172
 173         closure_wake_up(&c->open_buckets_wait);
 174         closure_wake_up(&c->freelist_wait);
 175 }
 176
 177 /* _only_ for allocating the journal on a new device: */
 178 long bch2_bucket_alloc_new_fs(struct bch_dev *ca)
 179 {
 180         while (ca->new_fs_bucket_idx < ca->mi.nbuckets) {
 181                 u64 b = ca->new_fs_bucket_idx++;
 182
 183                 if (!is_superblock_bucket(ca, b) &&
 184                     (!ca->buckets_nouse || !test_bit(b, ca->buckets_nouse)))
 185                         return b;
 186         }
 187
 188         return -1;
 189 }
 190
 191 static inline unsigned open_buckets_reserved(enum bch_watermark watermark)
 192 {
 193         switch (watermark) {
 194         case BCH_WATERMARK_reclaim:
 195                 return 0;
 196         case BCH_WATERMARK_btree:
 197         case BCH_WATERMARK_btree_copygc:
 198                 return OPEN_BUCKETS_COUNT / 4;
 199         case BCH_WATERMARK_copygc:
 200                 return OPEN_BUCKETS_COUNT / 3;
 201         default:
 202                 return OPEN_BUCKETS_COUNT / 2;
 203         }
 204 }
 205
 206 static struct open_bucket *__try_alloc_bucket(struct bch_fs *c, struct bch_dev *ca,
 207                                               u64 bucket,
 208                                               enum bch_watermark watermark,
 209                                               const struct bch_alloc_v4 *a,
 210                                               struct bucket_alloc_state *s,
 211                                               struct closure *cl)
 212 {
 213         struct open_bucket *ob;
 214
 215         if (unlikely(ca->buckets_nouse && test_bit(bucket, ca->buckets_nouse))) {
 216                 s->skipped_nouse++;
 217                 return NULL;
 218         }
 219
 220         if (bch2_bucket_is_open(c, ca->dev_idx, bucket)) {
 221                 s->skipped_open++;
 222                 return NULL;
 223         }
 224
 225         if (bch2_bucket_needs_journal_commit(&c->buckets_waiting_for_journal,
 226                         c->journal.flushed_seq_ondisk, ca->dev_idx, bucket)) {
 227                 s->skipped_need_journal_commit++;
 228                 return NULL;
 229         }
 230
 231         if (bch2_bucket_nocow_is_locked(&c->nocow_locks, POS(ca->dev_idx, bucket))) {
 232                 s->skipped_nocow++;
 233                 return NULL;
 234         }
 235
 236         spin_lock(&c->freelist_lock);
 237
 238         if (unlikely(c->open_buckets_nr_free <= open_buckets_reserved(watermark))) {
 239                 if (cl)
 240                         closure_wait(&c->open_buckets_wait, cl);
 241
 242                 track_event_change(&c->times[BCH_TIME_blocked_allocate_open_bucket],
 243                                    &c->blocked_allocate_open_bucket, true);
 244                 spin_unlock(&c->freelist_lock);
 245                 return ERR_PTR(-BCH_ERR_open_buckets_empty);
 246         }
 247
 248         /* Recheck under lock: */
 249         if (bch2_bucket_is_open(c, ca->dev_idx, bucket)) {
 250                 spin_unlock(&c->freelist_lock);
 251                 s->skipped_open++;
 252                 return NULL;
 253         }
 254
 255         ob = bch2_open_bucket_alloc(c);
 256
 257         spin_lock(&ob->lock);
 258
 259         ob->valid       = true;
 260         ob->sectors_free = ca->mi.bucket_size;
 261         ob->dev         = ca->dev_idx;
 262         ob->gen         = a->gen;
 263         ob->bucket      = bucket;
 264         spin_unlock(&ob->lock);
 265
 266         ca->nr_open_buckets++;
 267         bch2_open_bucket_hash_add(c, ob);
 268
 269         track_event_change(&c->times[BCH_TIME_blocked_allocate_open_bucket],
 270                            &c->blocked_allocate_open_bucket, false);
 271
 272         track_event_change(&c->times[BCH_TIME_blocked_allocate],
 273                            &c->blocked_allocate, false);
 274
 275         spin_unlock(&c->freelist_lock);
 276         return ob;
 277 }
 278
 279 static struct open_bucket *try_alloc_bucket(struct btree_trans *trans, struct bch_dev *ca,
 280                                             enum bch_watermark watermark, u64 free_entry,
 281                                             struct bucket_alloc_state *s,
 282                                             struct bkey_s_c freespace_k,
 283                                             struct closure *cl)
 284 {
 285         struct bch_fs *c = trans->c;
 286         struct btree_iter iter = { NULL };
 287         struct bkey_s_c k;
 288         struct open_bucket *ob;
 289         struct bch_alloc_v4 a_convert;
 290         const struct bch_alloc_v4 *a;
 291         u64 b = free_entry & ~(~0ULL << 56);
 292         unsigned genbits = free_entry >> 56;
 293         struct printbuf buf = PRINTBUF;
 294         int ret;
 295
 296         if (b < ca->mi.first_bucket || b >= ca->mi.nbuckets) {
 297                 prt_printf(&buf, "freespace btree has bucket outside allowed range %u-%llu\n"
 298                        "  freespace key ",
 299                         ca->mi.first_bucket, ca->mi.nbuckets);
 300                 bch2_bkey_val_to_text(&buf, c, freespace_k);
 301                 bch2_trans_inconsistent(trans, "%s", buf.buf);
 302                 ob = ERR_PTR(-EIO);
 303                 goto err;
 304         }
 305
 306         k = bch2_bkey_get_iter(trans, &iter,
 307                                BTREE_ID_alloc, POS(ca->dev_idx, b),
 308                                BTREE_ITER_CACHED);
 309         ret = bkey_err(k);
 310         if (ret) {
 311                 ob = ERR_PTR(ret);
 312                 goto err;
 313         }
 314
 315         a = bch2_alloc_to_v4(k, &a_convert);
 316
 317         if (a->data_type != BCH_DATA_free) {
 318                 if (c->curr_recovery_pass <= BCH_RECOVERY_PASS_check_alloc_info) {
 319                         ob = NULL;
 320                         goto err;
 321                 }
 322
 323                 prt_printf(&buf, "non free bucket in freespace btree\n"
 324                        "  freespace key ");
 325                 bch2_bkey_val_to_text(&buf, c, freespace_k);
 326                 prt_printf(&buf, "\n  ");
 327                 bch2_bkey_val_to_text(&buf, c, k);
 328                 bch2_trans_inconsistent(trans, "%s", buf.buf);
 329                 ob = ERR_PTR(-EIO);
 330                 goto err;
 331         }
 332
 333         if (genbits != (alloc_freespace_genbits(*a) >> 56) &&
 334             c->curr_recovery_pass > BCH_RECOVERY_PASS_check_alloc_info) {
 335                 prt_printf(&buf, "bucket in freespace btree with wrong genbits (got %u should be %llu)\n"
 336                        "  freespace key ",
 337                        genbits, alloc_freespace_genbits(*a) >> 56);
 338                 bch2_bkey_val_to_text(&buf, c, freespace_k);
 339                 prt_printf(&buf, "\n  ");
 340                 bch2_bkey_val_to_text(&buf, c, k);
 341                 bch2_trans_inconsistent(trans, "%s", buf.buf);
 342                 ob = ERR_PTR(-EIO);
 343                 goto err;
 344         }
 345
 346         if (c->curr_recovery_pass <= BCH_RECOVERY_PASS_check_extents_to_backpointers) {
 347                 struct bch_backpointer bp;
 348                 struct bpos bp_pos = POS_MIN;
 349
 350                 ret = bch2_get_next_backpointer(trans, POS(ca->dev_idx, b), -1,
 351                                                 &bp_pos, &bp,
 352                                                 BTREE_ITER_NOPRESERVE);
 353                 if (ret) {
 354                         ob = ERR_PTR(ret);
 355                         goto err;
 356                 }
 357
 358                 if (!bkey_eq(bp_pos, POS_MAX)) {
 359                         /*
 360                          * Bucket may have data in it - we don't call
 361                          * bc2h_trans_inconnsistent() because fsck hasn't
 362                          * finished yet
 363                          */
 364                         ob = NULL;
 365                         goto err;
 366                 }
 367         }
 368
 369         ob = __try_alloc_bucket(c, ca, b, watermark, a, s, cl);
 370         if (!ob)
 371                 set_btree_iter_dontneed(&iter);
 372 err:
 373         if (iter.path)
 374                 set_btree_iter_dontneed(&iter);
 375         bch2_trans_iter_exit(trans, &iter);
 376         printbuf_exit(&buf);
 377         return ob;
 378 }
 379
 380 /*
 381  * This path is for before the freespace btree is initialized:
 382  *
 383  * If ca->new_fs_bucket_idx is nonzero, we haven't yet marked superblock &
 384  * journal buckets - journal buckets will be < ca->new_fs_bucket_idx
 385  */
 386 static noinline struct open_bucket *
 387 bch2_bucket_alloc_early(struct btree_trans *trans,
 388                         struct bch_dev *ca,
 389                         enum bch_watermark watermark,
 390                         struct bucket_alloc_state *s,
 391                         struct closure *cl)
 392 {
 393         struct btree_iter iter, citer;
 394         struct bkey_s_c k, ck;
 395         struct open_bucket *ob = NULL;
 396         u64 first_bucket = max_t(u64, ca->mi.first_bucket, ca->new_fs_bucket_idx);
 397         u64 alloc_start = max(first_bucket, READ_ONCE(ca->alloc_cursor));
 398         u64 alloc_cursor = alloc_start;
 399         int ret;
 400
 401         /*
 402          * Scan with an uncached iterator to avoid polluting the key cache. An
 403          * uncached iter will return a cached key if one exists, but if not
 404          * there is no other underlying protection for the associated key cache
 405          * slot. To avoid racing bucket allocations, look up the cached key slot
 406          * of any likely allocation candidate before attempting to proceed with
 407          * the allocation. This provides proper exclusion on the associated
 408          * bucket.
 409          */
 410 again:
 411         for_each_btree_key_norestart(trans, iter, BTREE_ID_alloc, POS(ca->dev_idx, alloc_cursor),
 412                            BTREE_ITER_SLOTS, k, ret) {
 413                 struct bch_alloc_v4 a_convert;
 414                 const struct bch_alloc_v4 *a;
 415
 416                 if (bkey_ge(k.k->p, POS(ca->dev_idx, ca->mi.nbuckets)))
 417                         break;
 418
 419                 if (ca->new_fs_bucket_idx &&
 420                     is_superblock_bucket(ca, k.k->p.offset))
 421                         continue;
 422
 423                 a = bch2_alloc_to_v4(k, &a_convert);
 424                 if (a->data_type != BCH_DATA_free)
 425                         continue;
 426
 427                 /* now check the cached key to serialize concurrent allocs of the bucket */
 428                 ck = bch2_bkey_get_iter(trans, &citer, BTREE_ID_alloc, k.k->p, BTREE_ITER_CACHED);
 429                 ret = bkey_err(ck);
 430                 if (ret)
 431                         break;
 432
 433                 a = bch2_alloc_to_v4(ck, &a_convert);
 434                 if (a->data_type != BCH_DATA_free)
 435                         goto next;
 436
 437                 s->buckets_seen++;
 438
 439                 ob = __try_alloc_bucket(trans->c, ca, k.k->p.offset, watermark, a, s, cl);
 440 next:
 441                 set_btree_iter_dontneed(&citer);
 442                 bch2_trans_iter_exit(trans, &citer);
 443                 if (ob)
 444                         break;
 445         }
 446         bch2_trans_iter_exit(trans, &iter);
 447
 448         alloc_cursor = iter.pos.offset;
 449         ca->alloc_cursor = alloc_cursor;
 450
 451         if (!ob && ret)
 452                 ob = ERR_PTR(ret);
 453
 454         if (!ob && alloc_start > first_bucket) {
 455                 alloc_cursor = alloc_start = first_bucket;
 456                 goto again;
 457         }
 458
 459         return ob;
 460 }
 461
 462 static struct open_bucket *bch2_bucket_alloc_freelist(struct btree_trans *trans,
 463                                                    struct bch_dev *ca,
 464                                                    enum bch_watermark watermark,
 465                                                    struct bucket_alloc_state *s,
 466                                                    struct closure *cl)
 467 {
 468         struct btree_iter iter;
 469         struct bkey_s_c k;
 470         struct open_bucket *ob = NULL;
 471         u64 alloc_start = max_t(u64, ca->mi.first_bucket, READ_ONCE(ca->alloc_cursor));
 472         u64 alloc_cursor = alloc_start;
 473         int ret;
 474
 475         BUG_ON(ca->new_fs_bucket_idx);
 476 again:
 477         for_each_btree_key_norestart(trans, iter, BTREE_ID_freespace,
 478                                      POS(ca->dev_idx, alloc_cursor), 0, k, ret) {
 479                 if (k.k->p.inode != ca->dev_idx)
 480                         break;
 481
 482                 for (alloc_cursor = max(alloc_cursor, bkey_start_offset(k.k));
 483                      alloc_cursor < k.k->p.offset;
 484                      alloc_cursor++) {
 485                         ret = btree_trans_too_many_iters(trans);
 486                         if (ret) {
 487                                 ob = ERR_PTR(ret);
 488                                 break;
 489                         }
 490
 491                         s->buckets_seen++;
 492
 493                         ob = try_alloc_bucket(trans, ca, watermark,
 494                                               alloc_cursor, s, k, cl);
 495                         if (ob) {
 496                                 set_btree_iter_dontneed(&iter);
 497                                 break;
 498                         }
 499                 }
 500
 501                 if (ob || ret)
 502                         break;
 503         }
 504         bch2_trans_iter_exit(trans, &iter);
 505
 506         ca->alloc_cursor = alloc_cursor;
 507
 508         if (!ob && ret)
 509                 ob = ERR_PTR(ret);
 510
 511         if (!ob && alloc_start > ca->mi.first_bucket) {
 512                 alloc_cursor = alloc_start = ca->mi.first_bucket;
 513                 goto again;
 514         }
 515
 516         return ob;
 517 }
 518
 519 /**
 520  * bch2_bucket_alloc_trans - allocate a single bucket from a specific device
 521  * @trans:      transaction object
 522  * @ca:         device to allocate from
 523  * @watermark:  how important is this allocation?
 524  * @cl:         if not NULL, closure to be used to wait if buckets not available
 525  * @usage:      for secondarily also returning the current device usage
 526  *
 527  * Returns:     an open_bucket on success, or an ERR_PTR() on failure.
 528  */
 529 static struct open_bucket *bch2_bucket_alloc_trans(struct btree_trans *trans,
 530                                       struct bch_dev *ca,
 531                                       enum bch_watermark watermark,
 532                                       struct closure *cl,
 533                                       struct bch_dev_usage *usage)
 534 {
 535         struct bch_fs *c = trans->c;
 536         struct open_bucket *ob = NULL;
 537         bool freespace = READ_ONCE(ca->mi.freespace_initialized);
 538         u64 avail;
 539         struct bucket_alloc_state s = { 0 };
 540         bool waiting = false;
 541 again:
 542         bch2_dev_usage_read_fast(ca, usage);
 543         avail = dev_buckets_free(ca, *usage, watermark);
 544
 545         if (usage->d[BCH_DATA_need_discard].buckets > avail)
 546                 bch2_do_discards(c);
 547
 548         if (usage->d[BCH_DATA_need_gc_gens].buckets > avail)
 549                 bch2_do_gc_gens(c);
 550
 551         if (should_invalidate_buckets(ca, *usage))
 552                 bch2_do_invalidates(c);
 553
 554         if (!avail) {
 555                 if (cl && !waiting) {
 556                         closure_wait(&c->freelist_wait, cl);
 557                         waiting = true;
 558                         goto again;
 559                 }
 560
 561                 track_event_change(&c->times[BCH_TIME_blocked_allocate],
 562                                    &c->blocked_allocate, true);
 563
 564                 ob = ERR_PTR(-BCH_ERR_freelist_empty);
 565                 goto err;
 566         }
 567
 568         if (waiting)
 569                 closure_wake_up(&c->freelist_wait);
 570 alloc:
 571         ob = likely(freespace)
 572                 ? bch2_bucket_alloc_freelist(trans, ca, watermark, &s, cl)
 573                 : bch2_bucket_alloc_early(trans, ca, watermark, &s, cl);
 574
 575         if (s.skipped_need_journal_commit * 2 > avail)
 576                 bch2_journal_flush_async(&c->journal, NULL);
 577
 578         if (!ob && freespace && c->curr_recovery_pass <= BCH_RECOVERY_PASS_check_alloc_info) {
 579                 freespace = false;
 580                 goto alloc;
 581         }
 582 err:
 583         if (!ob)
 584                 ob = ERR_PTR(-BCH_ERR_no_buckets_found);
 585
 586         if (!IS_ERR(ob))
 587                 trace_and_count(c, bucket_alloc, ca,
 588                                 bch2_watermarks[watermark],
 589                                 ob->bucket,
 590                                 usage->d[BCH_DATA_free].buckets,
 591                                 avail,
 592                                 bch2_copygc_wait_amount(c),
 593                                 c->copygc_wait - atomic64_read(&c->io_clock[WRITE].now),
 594                                 &s,
 595                                 cl == NULL,
 596                                 "");
 597         else if (!bch2_err_matches(PTR_ERR(ob), BCH_ERR_transaction_restart))
 598                 trace_and_count(c, bucket_alloc_fail, ca,
 599                                 bch2_watermarks[watermark],
 600                                 0,
 601                                 usage->d[BCH_DATA_free].buckets,
 602                                 avail,
 603                                 bch2_copygc_wait_amount(c),
 604                                 c->copygc_wait - atomic64_read(&c->io_clock[WRITE].now),
 605                                 &s,
 606                                 cl == NULL,
 607                                 bch2_err_str(PTR_ERR(ob)));
 608
 609         return ob;
 610 }
 611
 612 struct open_bucket *bch2_bucket_alloc(struct bch_fs *c, struct bch_dev *ca,
 613                                       enum bch_watermark watermark,
 614                                       struct closure *cl)
 615 {
 616         struct bch_dev_usage usage;
 617         struct open_bucket *ob;
 618
 619         bch2_trans_do(c, NULL, NULL, 0,
 620                       PTR_ERR_OR_ZERO(ob = bch2_bucket_alloc_trans(trans, ca, watermark,
 621                                                         cl, &usage)));
 622         return ob;
 623 }
 624
 625 static int __dev_stripe_cmp(struct dev_stripe_state *stripe,
 626                             unsigned l, unsigned r)
 627 {
 628         return ((stripe->next_alloc[l] > stripe->next_alloc[r]) -
 629                 (stripe->next_alloc[l] < stripe->next_alloc[r]));
 630 }
 631
 632 #define dev_stripe_cmp(l, r) __dev_stripe_cmp(stripe, l, r)
 633
 634 struct dev_alloc_list bch2_dev_alloc_list(struct bch_fs *c,
 635                                           struct dev_stripe_state *stripe,
 636                                           struct bch_devs_mask *devs)
 637 {
 638         struct dev_alloc_list ret = { .nr = 0 };
 639         unsigned i;
 640
 641         for_each_set_bit(i, devs->d, BCH_SB_MEMBERS_MAX)
 642                 ret.devs[ret.nr++] = i;
 643
 644         bubble_sort(ret.devs, ret.nr, dev_stripe_cmp);
 645         return ret;
 646 }
 647
 648 static inline void bch2_dev_stripe_increment_inlined(struct bch_dev *ca,
 649                                struct dev_stripe_state *stripe,
 650                                struct bch_dev_usage *usage)
 651 {
 652         u64 *v = stripe->next_alloc + ca->dev_idx;
 653         u64 free_space = dev_buckets_available(ca, BCH_WATERMARK_normal);
 654         u64 free_space_inv = free_space
 655                 ? div64_u64(1ULL << 48, free_space)
 656                 : 1ULL << 48;
 657         u64 scale = *v / 4;
 658
 659         if (*v + free_space_inv >= *v)
 660                 *v += free_space_inv;
 661         else
 662                 *v = U64_MAX;
 663
 664         for (v = stripe->next_alloc;
 665              v < stripe->next_alloc + ARRAY_SIZE(stripe->next_alloc); v++)
 666                 *v = *v < scale ? 0 : *v - scale;
 667 }
 668
 669 void bch2_dev_stripe_increment(struct bch_dev *ca,
 670                                struct dev_stripe_state *stripe)
 671 {
 672         struct bch_dev_usage usage;
 673
 674         bch2_dev_usage_read_fast(ca, &usage);
 675         bch2_dev_stripe_increment_inlined(ca, stripe, &usage);
 676 }
 677
 678 static int add_new_bucket(struct bch_fs *c,
 679                            struct open_buckets *ptrs,
 680                            struct bch_devs_mask *devs_may_alloc,
 681                            unsigned nr_replicas,
 682                            unsigned *nr_effective,
 683                            bool *have_cache,
 684                            unsigned flags,
 685                            struct open_bucket *ob)
 686 {
 687         unsigned durability =
 688                 bch_dev_bkey_exists(c, ob->dev)->mi.durability;
 689
 690         BUG_ON(*nr_effective >= nr_replicas);
 691
 692         __clear_bit(ob->dev, devs_may_alloc->d);
 693         *nr_effective   += durability;
 694         *have_cache     |= !durability;
 695
 696         ob_push(c, ptrs, ob);
 697
 698         if (*nr_effective >= nr_replicas)
 699                 return 1;
 700         if (ob->ec)
 701                 return 1;
 702         return 0;
 703 }
 704
 705 int bch2_bucket_alloc_set_trans(struct btree_trans *trans,
 706                       struct open_buckets *ptrs,
 707                       struct dev_stripe_state *stripe,
 708                       struct bch_devs_mask *devs_may_alloc,
 709                       unsigned nr_replicas,
 710                       unsigned *nr_effective,
 711                       bool *have_cache,
 712                       unsigned flags,
 713                       enum bch_data_type data_type,
 714                       enum bch_watermark watermark,
 715                       struct closure *cl)
 716 {
 717         struct bch_fs *c = trans->c;
 718         struct dev_alloc_list devs_sorted =
 719                 bch2_dev_alloc_list(c, stripe, devs_may_alloc);
 720         unsigned dev;
 721         struct bch_dev *ca;
 722         int ret = -BCH_ERR_insufficient_devices;
 723         unsigned i;
 724
 725         BUG_ON(*nr_effective >= nr_replicas);
 726
 727         for (i = 0; i < devs_sorted.nr; i++) {
 728                 struct bch_dev_usage usage;
 729                 struct open_bucket *ob;
 730
 731                 dev = devs_sorted.devs[i];
 732
 733                 rcu_read_lock();
 734                 ca = rcu_dereference(c->devs[dev]);
 735                 if (ca)
 736                         percpu_ref_get(&ca->ref);
 737                 rcu_read_unlock();
 738
 739                 if (!ca)
 740                         continue;
 741
 742                 if (!ca->mi.durability && *have_cache) {
 743                         percpu_ref_put(&ca->ref);
 744                         continue;
 745                 }
 746
 747                 ob = bch2_bucket_alloc_trans(trans, ca, watermark, cl, &usage);
 748                 if (!IS_ERR(ob))
 749                         bch2_dev_stripe_increment_inlined(ca, stripe, &usage);
 750                 percpu_ref_put(&ca->ref);
 751
 752                 if (IS_ERR(ob)) {
 753                         ret = PTR_ERR(ob);
 754                         if (bch2_err_matches(ret, BCH_ERR_transaction_restart) || cl)
 755                                 break;
 756                         continue;
 757                 }
 758
 759                 ob->data_type = data_type;
 760
 761                 if (add_new_bucket(c, ptrs, devs_may_alloc,
 762                                    nr_replicas, nr_effective,
 763                                    have_cache, flags, ob)) {
 764                         ret = 0;
 765                         break;
 766                 }
 767         }
 768
 769         return ret;
 770 }
 771
 772 /* Allocate from stripes: */
 773
 774 /*
 775  * if we can't allocate a new stripe because there are already too many
 776  * partially filled stripes, force allocating from an existing stripe even when
 777  * it's to a device we don't want:
 778  */
 779
 780 static int bucket_alloc_from_stripe(struct btree_trans *trans,
 781                          struct open_buckets *ptrs,
 782                          struct write_point *wp,
 783                          struct bch_devs_mask *devs_may_alloc,
 784                          u16 target,
 785                          unsigned nr_replicas,
 786                          unsigned *nr_effective,
 787                          bool *have_cache,
 788                          enum bch_watermark watermark,
 789                          unsigned flags,
 790                          struct closure *cl)
 791 {
 792         struct bch_fs *c = trans->c;
 793         struct dev_alloc_list devs_sorted;
 794         struct ec_stripe_head *h;
 795         struct open_bucket *ob;
 796         unsigned i, ec_idx;
 797         int ret = 0;
 798
 799         if (nr_replicas < 2)
 800                 return 0;
 801
 802         if (ec_open_bucket(c, ptrs))
 803                 return 0;
 804
 805         h = bch2_ec_stripe_head_get(trans, target, 0, nr_replicas - 1, watermark, cl);
 806         if (IS_ERR(h))
 807                 return PTR_ERR(h);
 808         if (!h)
 809                 return 0;
 810
 811         devs_sorted = bch2_dev_alloc_list(c, &wp->stripe, devs_may_alloc);
 812
 813         for (i = 0; i < devs_sorted.nr; i++)
 814                 for (ec_idx = 0; ec_idx < h->s->nr_data; ec_idx++) {
 815                         if (!h->s->blocks[ec_idx])
 816                                 continue;
 817
 818                         ob = c->open_buckets + h->s->blocks[ec_idx];
 819                         if (ob->dev == devs_sorted.devs[i] &&
 820                             !test_and_set_bit(ec_idx, h->s->blocks_allocated))
 821                                 goto got_bucket;
 822                 }
 823         goto out_put_head;
 824 got_bucket:
 825         ob->ec_idx      = ec_idx;
 826         ob->ec          = h->s;
 827         ec_stripe_new_get(h->s, STRIPE_REF_io);
 828
 829         ret = add_new_bucket(c, ptrs, devs_may_alloc,
 830                              nr_replicas, nr_effective,
 831                              have_cache, flags, ob);
 832 out_put_head:
 833         bch2_ec_stripe_head_put(c, h);
 834         return ret;
 835 }
 836
 837 /* Sector allocator */
 838
 839 static bool want_bucket(struct bch_fs *c,
 840                         struct write_point *wp,
 841                         struct bch_devs_mask *devs_may_alloc,
 842                         bool *have_cache, bool ec,
 843                         struct open_bucket *ob)
 844 {
 845         struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
 846
 847         if (!test_bit(ob->dev, devs_may_alloc->d))
 848                 return false;
 849
 850         if (ob->data_type != wp->data_type)
 851                 return false;
 852
 853         if (!ca->mi.durability &&
 854             (wp->data_type == BCH_DATA_btree || ec || *have_cache))
 855                 return false;
 856
 857         if (ec != (ob->ec != NULL))
 858                 return false;
 859
 860         return true;
 861 }
 862
 863 static int bucket_alloc_set_writepoint(struct bch_fs *c,
 864                                        struct open_buckets *ptrs,
 865                                        struct write_point *wp,
 866                                        struct bch_devs_mask *devs_may_alloc,
 867                                        unsigned nr_replicas,
 868                                        unsigned *nr_effective,
 869                                        bool *have_cache,
 870                                        bool ec, unsigned flags)
 871 {
 872         struct open_buckets ptrs_skip = { .nr = 0 };
 873         struct open_bucket *ob;
 874         unsigned i;
 875         int ret = 0;
 876
 877         open_bucket_for_each(c, &wp->ptrs, ob, i) {
 878                 if (!ret && want_bucket(c, wp, devs_may_alloc,
 879                                         have_cache, ec, ob))
 880                         ret = add_new_bucket(c, ptrs, devs_may_alloc,
 881                                        nr_replicas, nr_effective,
 882                                        have_cache, flags, ob);
 883                 else
 884                         ob_push(c, &ptrs_skip, ob);
 885         }
 886         wp->ptrs = ptrs_skip;
 887
 888         return ret;
 889 }
 890
 891 static int bucket_alloc_set_partial(struct bch_fs *c,
 892                                     struct open_buckets *ptrs,
 893                                     struct write_point *wp,
 894                                     struct bch_devs_mask *devs_may_alloc,
 895                                     unsigned nr_replicas,
 896                                     unsigned *nr_effective,
 897                                     bool *have_cache, bool ec,
 898                                     enum bch_watermark watermark,
 899                                     unsigned flags)
 900 {
 901         int i, ret = 0;
 902
 903         if (!c->open_buckets_partial_nr)
 904                 return 0;
 905
 906         spin_lock(&c->freelist_lock);
 907
 908         if (!c->open_buckets_partial_nr)
 909                 goto unlock;
 910
 911         for (i = c->open_buckets_partial_nr - 1; i >= 0; --i) {
 912                 struct open_bucket *ob = c->open_buckets + c->open_buckets_partial[i];
 913
 914                 if (want_bucket(c, wp, devs_may_alloc, have_cache, ec, ob)) {
 915                         struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
 916                         struct bch_dev_usage usage;
 917                         u64 avail;
 918
 919                         bch2_dev_usage_read_fast(ca, &usage);
 920                         avail = dev_buckets_free(ca, usage, watermark);
 921                         if (!avail)
 922                                 continue;
 923
 924                         array_remove_item(c->open_buckets_partial,
 925                                           c->open_buckets_partial_nr,
 926                                           i);
 927                         ob->on_partial_list = false;
 928
 929                         ret = add_new_bucket(c, ptrs, devs_may_alloc,
 930                                              nr_replicas, nr_effective,
 931                                              have_cache, flags, ob);
 932                         if (ret)
 933                                 break;
 934                 }
 935         }
 936 unlock:
 937         spin_unlock(&c->freelist_lock);
 938         return ret;
 939 }
 940
 941 static int __open_bucket_add_buckets(struct btree_trans *trans,
 942                         struct open_buckets *ptrs,
 943                         struct write_point *wp,
 944                         struct bch_devs_list *devs_have,
 945                         u16 target,
 946                         bool erasure_code,
 947                         unsigned nr_replicas,
 948                         unsigned *nr_effective,
 949                         bool *have_cache,
 950                         enum bch_watermark watermark,
 951                         unsigned flags,
 952                         struct closure *_cl)
 953 {
 954         struct bch_fs *c = trans->c;
 955         struct bch_devs_mask devs;
 956         struct open_bucket *ob;
 957         struct closure *cl = NULL;
 958         unsigned i;
 959         int ret;
 960
 961         devs = target_rw_devs(c, wp->data_type, target);
 962
 963         /* Don't allocate from devices we already have pointers to: */
 964         for (i = 0; i < devs_have->nr; i++)
 965                 __clear_bit(devs_have->devs[i], devs.d);
 966
 967         open_bucket_for_each(c, ptrs, ob, i)
 968                 __clear_bit(ob->dev, devs.d);
 969
 970         if (erasure_code && ec_open_bucket(c, ptrs))
 971                 return 0;
 972
 973         ret = bucket_alloc_set_writepoint(c, ptrs, wp, &devs,
 974                                  nr_replicas, nr_effective,
 975                                  have_cache, erasure_code, flags);
 976         if (ret)
 977                 return ret;
 978
 979         ret = bucket_alloc_set_partial(c, ptrs, wp, &devs,
 980                                  nr_replicas, nr_effective,
 981                                  have_cache, erasure_code, watermark, flags);
 982         if (ret)
 983                 return ret;
 984
 985         if (erasure_code) {
 986                 ret = bucket_alloc_from_stripe(trans, ptrs, wp, &devs,
 987                                          target,
 988                                          nr_replicas, nr_effective,
 989                                          have_cache,
 990                                          watermark, flags, _cl);
 991         } else {
 992 retry_blocking:
 993                 /*
 994                  * Try nonblocking first, so that if one device is full we'll try from
 995                  * other devices:
 996                  */
 997                 ret = bch2_bucket_alloc_set_trans(trans, ptrs, &wp->stripe, &devs,
 998                                         nr_replicas, nr_effective, have_cache,
 999                                         flags, wp->data_type, watermark, cl);
1000                 if (ret &&
1001                     !bch2_err_matches(ret, BCH_ERR_transaction_restart) &&
1002                     !bch2_err_matches(ret, BCH_ERR_insufficient_devices) &&
1003                     !cl && _cl) {
1004                         cl = _cl;
1005                         goto retry_blocking;
1006                 }
1007         }
1008
1009         return ret;
1010 }
1011
1012 static int open_bucket_add_buckets(struct btree_trans *trans,
1013                         struct open_buckets *ptrs,
1014                         struct write_point *wp,
1015                         struct bch_devs_list *devs_have,
1016                         u16 target,
1017                         unsigned erasure_code,
1018                         unsigned nr_replicas,
1019                         unsigned *nr_effective,
1020                         bool *have_cache,
1021                         enum bch_watermark watermark,
1022                         unsigned flags,
1023                         struct closure *cl)
1024 {
1025         int ret;
1026
1027         if (erasure_code) {
1028                 ret = __open_bucket_add_buckets(trans, ptrs, wp,
1029                                 devs_have, target, erasure_code,
1030                                 nr_replicas, nr_effective, have_cache,
1031                                 watermark, flags, cl);
1032                 if (bch2_err_matches(ret, BCH_ERR_transaction_restart) ||
1033                     bch2_err_matches(ret, BCH_ERR_operation_blocked) ||
1034                     bch2_err_matches(ret, BCH_ERR_freelist_empty) ||
1035                     bch2_err_matches(ret, BCH_ERR_open_buckets_empty))
1036                         return ret;
1037                 if (*nr_effective >= nr_replicas)
1038                         return 0;
1039         }
1040
1041         ret = __open_bucket_add_buckets(trans, ptrs, wp,
1042                         devs_have, target, false,
1043                         nr_replicas, nr_effective, have_cache,
1044                         watermark, flags, cl);
1045         return ret < 0 ? ret : 0;
1046 }
1047
1048 /**
1049  * should_drop_bucket - check if this is open_bucket should go away
1050  * @ob:         open_bucket to predicate on
1051  * @c:          filesystem handle
1052  * @ca:         if set, we're killing buckets for a particular device
1053  * @ec:         if true, we're shutting down erasure coding and killing all ec
1054  *              open_buckets
1055  *              otherwise, return true
1056  * Returns: true if we should kill this open_bucket
1057  *
1058  * We're killing open_buckets because we're shutting down a device, erasure
1059  * coding, or the entire filesystem - check if this open_bucket matches:
1060  */
1061 static bool should_drop_bucket(struct open_bucket *ob, struct bch_fs *c,
1062                                struct bch_dev *ca, bool ec)
1063 {
1064         if (ec) {
1065                 return ob->ec != NULL;
1066         } else if (ca) {
1067                 bool drop = ob->dev == ca->dev_idx;
1068                 struct open_bucket *ob2;
1069                 unsigned i;
1070
1071                 if (!drop && ob->ec) {
1072                         unsigned nr_blocks;
1073
1074                         mutex_lock(&ob->ec->lock);
1075                         nr_blocks = bkey_i_to_stripe(&ob->ec->new_stripe.key)->v.nr_blocks;
1076
1077                         for (i = 0; i < nr_blocks; i++) {
1078                                 if (!ob->ec->blocks[i])
1079                                         continue;
1080
1081                                 ob2 = c->open_buckets + ob->ec->blocks[i];
1082                                 drop |= ob2->dev == ca->dev_idx;
1083                         }
1084                         mutex_unlock(&ob->ec->lock);
1085                 }
1086
1087                 return drop;
1088         } else {
1089                 return true;
1090         }
1091 }
1092
1093 static void bch2_writepoint_stop(struct bch_fs *c, struct bch_dev *ca,
1094                                  bool ec, struct write_point *wp)
1095 {
1096         struct open_buckets ptrs = { .nr = 0 };
1097         struct open_bucket *ob;
1098         unsigned i;
1099
1100         mutex_lock(&wp->lock);
1101         open_bucket_for_each(c, &wp->ptrs, ob, i)
1102                 if (should_drop_bucket(ob, c, ca, ec))
1103                         bch2_open_bucket_put(c, ob);
1104                 else
1105                         ob_push(c, &ptrs, ob);
1106         wp->ptrs = ptrs;
1107         mutex_unlock(&wp->lock);
1108 }
1109
1110 void bch2_open_buckets_stop(struct bch_fs *c, struct bch_dev *ca,
1111                             bool ec)
1112 {
1113         unsigned i;
1114
1115         /* Next, close write points that point to this device... */
1116         for (i = 0; i < ARRAY_SIZE(c->write_points); i++)
1117                 bch2_writepoint_stop(c, ca, ec, &c->write_points[i]);
1118
1119         bch2_writepoint_stop(c, ca, ec, &c->copygc_write_point);
1120         bch2_writepoint_stop(c, ca, ec, &c->rebalance_write_point);
1121         bch2_writepoint_stop(c, ca, ec, &c->btree_write_point);
1122
1123         mutex_lock(&c->btree_reserve_cache_lock);
1124         while (c->btree_reserve_cache_nr) {
1125                 struct btree_alloc *a =
1126                         &c->btree_reserve_cache[--c->btree_reserve_cache_nr];
1127
1128                 bch2_open_buckets_put(c, &a->ob);
1129         }
1130         mutex_unlock(&c->btree_reserve_cache_lock);
1131
1132         spin_lock(&c->freelist_lock);
1133         i = 0;
1134         while (i < c->open_buckets_partial_nr) {
1135                 struct open_bucket *ob =
1136                         c->open_buckets + c->open_buckets_partial[i];
1137
1138                 if (should_drop_bucket(ob, c, ca, ec)) {
1139                         --c->open_buckets_partial_nr;
1140                         swap(c->open_buckets_partial[i],
1141                              c->open_buckets_partial[c->open_buckets_partial_nr]);
1142                         ob->on_partial_list = false;
1143                         spin_unlock(&c->freelist_lock);
1144                         bch2_open_bucket_put(c, ob);
1145                         spin_lock(&c->freelist_lock);
1146                 } else {
1147                         i++;
1148                 }
1149         }
1150         spin_unlock(&c->freelist_lock);
1151
1152         bch2_ec_stop_dev(c, ca);
1153 }
1154
1155 static inline struct hlist_head *writepoint_hash(struct bch_fs *c,
1156                                                  unsigned long write_point)
1157 {
1158         unsigned hash =
1159                 hash_long(write_point, ilog2(ARRAY_SIZE(c->write_points_hash)));
1160
1161         return &c->write_points_hash[hash];
1162 }
1163
1164 static struct write_point *__writepoint_find(struct hlist_head *head,
1165                                              unsigned long write_point)
1166 {
1167         struct write_point *wp;
1168
1169         rcu_read_lock();
1170         hlist_for_each_entry_rcu(wp, head, node)
1171                 if (wp->write_point == write_point)
1172                         goto out;
1173         wp = NULL;
1174 out:
1175         rcu_read_unlock();
1176         return wp;
1177 }
1178
1179 static inline bool too_many_writepoints(struct bch_fs *c, unsigned factor)
1180 {
1181         u64 stranded    = c->write_points_nr * c->bucket_size_max;
1182         u64 free        = bch2_fs_usage_read_short(c).free;
1183
1184         return stranded * factor > free;
1185 }
1186
1187 static bool try_increase_writepoints(struct bch_fs *c)
1188 {
1189         struct write_point *wp;
1190
1191         if (c->write_points_nr == ARRAY_SIZE(c->write_points) ||
1192             too_many_writepoints(c, 32))
1193                 return false;
1194
1195         wp = c->write_points + c->write_points_nr++;
1196         hlist_add_head_rcu(&wp->node, writepoint_hash(c, wp->write_point));
1197         return true;
1198 }
1199
1200 static bool try_decrease_writepoints(struct btree_trans *trans, unsigned old_nr)
1201 {
1202         struct bch_fs *c = trans->c;
1203         struct write_point *wp;
1204         struct open_bucket *ob;
1205         unsigned i;
1206
1207         mutex_lock(&c->write_points_hash_lock);
1208         if (c->write_points_nr < old_nr) {
1209                 mutex_unlock(&c->write_points_hash_lock);
1210                 return true;
1211         }
1212
1213         if (c->write_points_nr == 1 ||
1214             !too_many_writepoints(c, 8)) {
1215                 mutex_unlock(&c->write_points_hash_lock);
1216                 return false;
1217         }
1218
1219         wp = c->write_points + --c->write_points_nr;
1220
1221         hlist_del_rcu(&wp->node);
1222         mutex_unlock(&c->write_points_hash_lock);
1223
1224         bch2_trans_mutex_lock_norelock(trans, &wp->lock);
1225         open_bucket_for_each(c, &wp->ptrs, ob, i)
1226                 open_bucket_free_unused(c, ob);
1227         wp->ptrs.nr = 0;
1228         mutex_unlock(&wp->lock);
1229         return true;
1230 }
1231
1232 static struct write_point *writepoint_find(struct btree_trans *trans,
1233                                            unsigned long write_point)
1234 {
1235         struct bch_fs *c = trans->c;
1236         struct write_point *wp, *oldest;
1237         struct hlist_head *head;
1238
1239         if (!(write_point & 1UL)) {
1240                 wp = (struct write_point *) write_point;
1241                 bch2_trans_mutex_lock_norelock(trans, &wp->lock);
1242                 return wp;
1243         }
1244
1245         head = writepoint_hash(c, write_point);
1246 restart_find:
1247         wp = __writepoint_find(head, write_point);
1248         if (wp) {
1249 lock_wp:
1250                 bch2_trans_mutex_lock_norelock(trans, &wp->lock);
1251                 if (wp->write_point == write_point)
1252                         goto out;
1253                 mutex_unlock(&wp->lock);
1254                 goto restart_find;
1255         }
1256 restart_find_oldest:
1257         oldest = NULL;
1258         for (wp = c->write_points;
1259              wp < c->write_points + c->write_points_nr; wp++)
1260                 if (!oldest || time_before64(wp->last_used, oldest->last_used))
1261                         oldest = wp;
1262
1263         bch2_trans_mutex_lock_norelock(trans, &oldest->lock);
1264         bch2_trans_mutex_lock_norelock(trans, &c->write_points_hash_lock);
1265         if (oldest >= c->write_points + c->write_points_nr ||
1266             try_increase_writepoints(c)) {
1267                 mutex_unlock(&c->write_points_hash_lock);
1268                 mutex_unlock(&oldest->lock);
1269                 goto restart_find_oldest;
1270         }
1271
1272         wp = __writepoint_find(head, write_point);
1273         if (wp && wp != oldest) {
1274                 mutex_unlock(&c->write_points_hash_lock);
1275                 mutex_unlock(&oldest->lock);
1276                 goto lock_wp;
1277         }
1278
1279         wp = oldest;
1280         hlist_del_rcu(&wp->node);
1281         wp->write_point = write_point;
1282         hlist_add_head_rcu(&wp->node, head);
1283         mutex_unlock(&c->write_points_hash_lock);
1284 out:
1285         wp->last_used = local_clock();
1286         return wp;
1287 }
1288
1289 static noinline void
1290 deallocate_extra_replicas(struct bch_fs *c,
1291                           struct open_buckets *ptrs,
1292                           struct open_buckets *ptrs_no_use,
1293                           unsigned extra_replicas)
1294 {
1295         struct open_buckets ptrs2 = { 0 };
1296         struct open_bucket *ob;
1297         unsigned i;
1298
1299         open_bucket_for_each(c, ptrs, ob, i) {
1300                 unsigned d = bch_dev_bkey_exists(c, ob->dev)->mi.durability;
1301
1302                 if (d && d <= extra_replicas) {
1303                         extra_replicas -= d;
1304                         ob_push(c, ptrs_no_use, ob);
1305                 } else {
1306                         ob_push(c, &ptrs2, ob);
1307                 }
1308         }
1309
1310         *ptrs = ptrs2;
1311 }
1312
1313 /*
1314  * Get us an open_bucket we can allocate from, return with it locked:
1315  */
1316 int bch2_alloc_sectors_start_trans(struct btree_trans *trans,
1317                              unsigned target,
1318                              unsigned erasure_code,
1319                              struct write_point_specifier write_point,
1320                              struct bch_devs_list *devs_have,
1321                              unsigned nr_replicas,
1322                              unsigned nr_replicas_required,
1323                              enum bch_watermark watermark,
1324                              unsigned flags,
1325                              struct closure *cl,
1326                              struct write_point **wp_ret)
1327 {
1328         struct bch_fs *c = trans->c;
1329         struct write_point *wp;
1330         struct open_bucket *ob;
1331         struct open_buckets ptrs;
1332         unsigned nr_effective, write_points_nr;
1333         bool have_cache;
1334         int ret;
1335         int i;
1336
1337         if (!IS_ENABLED(CONFIG_BCACHEFS_ERASURE_CODING))
1338                 erasure_code = false;
1339
1340         BUG_ON(flags & BCH_WRITE_ONLY_SPECIFIED_DEVS);
1341
1342         BUG_ON(!nr_replicas || !nr_replicas_required);
1343 retry:
1344         ptrs.nr         = 0;
1345         nr_effective    = 0;
1346         write_points_nr = c->write_points_nr;
1347         have_cache      = false;
1348
1349         *wp_ret = wp = writepoint_find(trans, write_point.v);
1350
1351         /* metadata may not allocate on cache devices: */
1352         if (wp->data_type != BCH_DATA_user)
1353                 have_cache = true;
1354
1355         if (target && !(flags & BCH_WRITE_ONLY_SPECIFIED_DEVS)) {
1356                 ret = open_bucket_add_buckets(trans, &ptrs, wp, devs_have,
1357                                               target, erasure_code,
1358                                               nr_replicas, &nr_effective,
1359                                               &have_cache, watermark,
1360                                               flags, NULL);
1361                 if (!ret ||
1362                     bch2_err_matches(ret, BCH_ERR_transaction_restart))
1363                         goto alloc_done;
1364
1365                 /* Don't retry from all devices if we're out of open buckets: */
1366                 if (bch2_err_matches(ret, BCH_ERR_open_buckets_empty))
1367                         goto allocate_blocking;
1368
1369                 /*
1370                  * Only try to allocate cache (durability = 0 devices) from the
1371                  * specified target:
1372                  */
1373                 have_cache = true;
1374
1375                 ret = open_bucket_add_buckets(trans, &ptrs, wp, devs_have,
1376                                               0, erasure_code,
1377                                               nr_replicas, &nr_effective,
1378                                               &have_cache, watermark,
1379                                               flags, cl);
1380         } else {
1381 allocate_blocking:
1382                 ret = open_bucket_add_buckets(trans, &ptrs, wp, devs_have,
1383                                               target, erasure_code,
1384                                               nr_replicas, &nr_effective,
1385                                               &have_cache, watermark,
1386                                               flags, cl);
1387         }
1388 alloc_done:
1389         BUG_ON(!ret && nr_effective < nr_replicas);
1390
1391         if (erasure_code && !ec_open_bucket(c, &ptrs))
1392                 pr_debug("failed to get ec bucket: ret %u", ret);
1393
1394         if (ret == -BCH_ERR_insufficient_devices &&
1395             nr_effective >= nr_replicas_required)
1396                 ret = 0;
1397
1398         if (ret)
1399                 goto err;
1400
1401         if (nr_effective > nr_replicas)
1402                 deallocate_extra_replicas(c, &ptrs, &wp->ptrs, nr_effective - nr_replicas);
1403
1404         /* Free buckets we didn't use: */
1405         open_bucket_for_each(c, &wp->ptrs, ob, i)
1406                 open_bucket_free_unused(c, ob);
1407
1408         wp->ptrs = ptrs;
1409
1410         wp->sectors_free = UINT_MAX;
1411
1412         open_bucket_for_each(c, &wp->ptrs, ob, i)
1413                 wp->sectors_free = min(wp->sectors_free, ob->sectors_free);
1414
1415         BUG_ON(!wp->sectors_free || wp->sectors_free == UINT_MAX);
1416
1417         return 0;
1418 err:
1419         open_bucket_for_each(c, &wp->ptrs, ob, i)
1420                 if (ptrs.nr < ARRAY_SIZE(ptrs.v))
1421                         ob_push(c, &ptrs, ob);
1422                 else
1423                         open_bucket_free_unused(c, ob);
1424         wp->ptrs = ptrs;
1425
1426         mutex_unlock(&wp->lock);
1427
1428         if (bch2_err_matches(ret, BCH_ERR_freelist_empty) &&
1429             try_decrease_writepoints(trans, write_points_nr))
1430                 goto retry;
1431
1432         if (bch2_err_matches(ret, BCH_ERR_open_buckets_empty) ||
1433             bch2_err_matches(ret, BCH_ERR_freelist_empty))
1434                 return cl
1435                         ? -BCH_ERR_bucket_alloc_blocked
1436                         : -BCH_ERR_ENOSPC_bucket_alloc;
1437
1438         return ret;
1439 }
1440
1441 struct bch_extent_ptr bch2_ob_ptr(struct bch_fs *c, struct open_bucket *ob)
1442 {
1443         struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
1444
1445         return (struct bch_extent_ptr) {
1446                 .type   = 1 << BCH_EXTENT_ENTRY_ptr,
1447                 .gen    = ob->gen,
1448                 .dev    = ob->dev,
1449                 .offset = bucket_to_sector(ca, ob->bucket) +
1450                         ca->mi.bucket_size -
1451                         ob->sectors_free,
1452         };
1453 }
1454
1455 void bch2_alloc_sectors_append_ptrs(struct bch_fs *c, struct write_point *wp,
1456                                     struct bkey_i *k, unsigned sectors,
1457                                     bool cached)
1458 {
1459         bch2_alloc_sectors_append_ptrs_inlined(c, wp, k, sectors, cached);
1460 }
1461
1462 /*
1463  * Append pointers to the space we just allocated to @k, and mark @sectors space
1464  * as allocated out of @ob
1465  */
1466 void bch2_alloc_sectors_done(struct bch_fs *c, struct write_point *wp)
1467 {
1468         bch2_alloc_sectors_done_inlined(c, wp);
1469 }
1470
1471 static inline void writepoint_init(struct write_point *wp,
1472                                    enum bch_data_type type)
1473 {
1474         mutex_init(&wp->lock);
1475         wp->data_type = type;
1476
1477         INIT_WORK(&wp->index_update_work, bch2_write_point_do_index_updates);
1478         INIT_LIST_HEAD(&wp->writes);
1479         spin_lock_init(&wp->writes_lock);
1480 }
1481
1482 void bch2_fs_allocator_foreground_init(struct bch_fs *c)
1483 {
1484         struct open_bucket *ob;
1485         struct write_point *wp;
1486
1487         mutex_init(&c->write_points_hash_lock);
1488         c->write_points_nr = ARRAY_SIZE(c->write_points);
1489
1490         /* open bucket 0 is a sentinal NULL: */
1491         spin_lock_init(&c->open_buckets[0].lock);
1492
1493         for (ob = c->open_buckets + 1;
1494              ob < c->open_buckets + ARRAY_SIZE(c->open_buckets); ob++) {
1495                 spin_lock_init(&ob->lock);
1496                 c->open_buckets_nr_free++;
1497
1498                 ob->freelist = c->open_buckets_freelist;
1499                 c->open_buckets_freelist = ob - c->open_buckets;
1500         }
1501
1502         writepoint_init(&c->btree_write_point,          BCH_DATA_btree);
1503         writepoint_init(&c->rebalance_write_point,      BCH_DATA_user);
1504         writepoint_init(&c->copygc_write_point,         BCH_DATA_user);
1505
1506         for (wp = c->write_points;
1507              wp < c->write_points + c->write_points_nr; wp++) {
1508                 writepoint_init(wp, BCH_DATA_user);
1509
1510                 wp->last_used   = local_clock();
1511                 wp->write_point = (unsigned long) wp;
1512                 hlist_add_head_rcu(&wp->node,
1513                                    writepoint_hash(c, wp->write_point));
1514         }
1515 }
1516
1517 static void bch2_open_bucket_to_text(struct printbuf *out, struct bch_fs *c, struct open_bucket *ob)
1518 {
1519         struct bch_dev *ca = bch_dev_bkey_exists(c, ob->dev);
1520         unsigned data_type = ob->data_type;
1521         barrier(); /* READ_ONCE() doesn't work on bitfields */
1522
1523         prt_printf(out, "%zu ref %u %s %u:%llu gen %u allocated %u/%u",
1524                    ob - c->open_buckets,
1525                    atomic_read(&ob->pin),
1526                    data_type < BCH_DATA_NR ? bch2_data_types[data_type] : "invalid data type",
1527                    ob->dev, ob->bucket, ob->gen,
1528                    ca->mi.bucket_size - ob->sectors_free, ca->mi.bucket_size);
1529         if (ob->ec)
1530                 prt_printf(out, " ec idx %llu", ob->ec->idx);
1531         if (ob->on_partial_list)
1532                 prt_str(out, " partial");
1533         prt_newline(out);
1534 }
1535
1536 void bch2_open_buckets_to_text(struct printbuf *out, struct bch_fs *c)
1537 {
1538         struct open_bucket *ob;
1539
1540         out->atomic++;
1541
1542         for (ob = c->open_buckets;
1543              ob < c->open_buckets + ARRAY_SIZE(c->open_buckets);
1544              ob++) {
1545                 spin_lock(&ob->lock);
1546                 if (ob->valid && !ob->on_partial_list)
1547                         bch2_open_bucket_to_text(out, c, ob);
1548                 spin_unlock(&ob->lock);
1549         }
1550
1551         --out->atomic;
1552 }
1553
1554 void bch2_open_buckets_partial_to_text(struct printbuf *out, struct bch_fs *c)
1555 {
1556         unsigned i;
1557
1558         out->atomic++;
1559         spin_lock(&c->freelist_lock);
1560
1561         for (i = 0; i < c->open_buckets_partial_nr; i++)
1562                 bch2_open_bucket_to_text(out, c,
1563                                 c->open_buckets + c->open_buckets_partial[i]);
1564
1565         spin_unlock(&c->freelist_lock);
1566         --out->atomic;
1567 }
1568
1569 static const char * const bch2_write_point_states[] = {
1570 #define x(n)    #n,
1571         WRITE_POINT_STATES()
1572 #undef x
1573         NULL
1574 };
1575
1576 static void bch2_write_point_to_text(struct printbuf *out, struct bch_fs *c,
1577                                      struct write_point *wp)
1578 {
1579         struct open_bucket *ob;
1580         unsigned i;
1581
1582         prt_printf(out, "%lu: ", wp->write_point);
1583         prt_human_readable_u64(out, wp->sectors_allocated);
1584
1585         prt_printf(out, " last wrote: ");
1586         bch2_pr_time_units(out, sched_clock() - wp->last_used);
1587
1588         for (i = 0; i < WRITE_POINT_STATE_NR; i++) {
1589                 prt_printf(out, " %s: ", bch2_write_point_states[i]);
1590                 bch2_pr_time_units(out, wp->time[i]);
1591         }
1592
1593         prt_newline(out);
1594
1595         printbuf_indent_add(out, 2);
1596         open_bucket_for_each(c, &wp->ptrs, ob, i)
1597                 bch2_open_bucket_to_text(out, c, ob);
1598         printbuf_indent_sub(out, 2);
1599 }
1600
1601 void bch2_write_points_to_text(struct printbuf *out, struct bch_fs *c)
1602 {
1603         struct write_point *wp;
1604
1605         prt_str(out, "Foreground write points\n");
1606         for (wp = c->write_points;
1607              wp < c->write_points + ARRAY_SIZE(c->write_points);
1608              wp++)
1609                 bch2_write_point_to_text(out, c, wp);
1610
1611         prt_str(out, "Copygc write point\n");
1612         bch2_write_point_to_text(out, c, &c->copygc_write_point);
1613
1614         prt_str(out, "Rebalance write point\n");
1615         bch2_write_point_to_text(out, c, &c->rebalance_write_point);
1616
1617         prt_str(out, "Btree write point\n");
1618         bch2_write_point_to_text(out, c, &c->btree_write_point);
1619 }