fs/bcachefs/btree_types.h

   1 /* SPDX-License-Identifier: GPL-2.0 */
   2 #ifndef _BCACHEFS_BTREE_TYPES_H
   3 #define _BCACHEFS_BTREE_TYPES_H
   4
   5 #include <linux/list.h>
   6 #include <linux/rhashtable.h>
   7
   8 #include "bbpos_types.h"
   9 #include "btree_key_cache_types.h"
  10 #include "buckets_types.h"
  11 #include "darray.h"
  12 #include "errcode.h"
  13 #include "journal_types.h"
  14 #include "replicas_types.h"
  15 #include "six.h"
  16
  17 struct open_bucket;
  18 struct btree_update;
  19 struct btree_trans;
  20
  21 #define MAX_BSETS               3U
  22
  23 struct btree_nr_keys {
  24
  25         /*
  26          * Amount of live metadata (i.e. size of node after a compaction) in
  27          * units of u64s
  28          */
  29         u16                     live_u64s;
  30         u16                     bset_u64s[MAX_BSETS];
  31
  32         /* live keys only: */
  33         u16                     packed_keys;
  34         u16                     unpacked_keys;
  35 };
  36
  37 struct bset_tree {
  38         /*
  39          * We construct a binary tree in an array as if the array
  40          * started at 1, so that things line up on the same cachelines
  41          * better: see comments in bset.c at cacheline_to_bkey() for
  42          * details
  43          */
  44
  45         /* size of the binary tree and prev array */
  46         u16                     size;
  47
  48         /* function of size - precalculated for to_inorder() */
  49         u16                     extra;
  50
  51         u16                     data_offset;
  52         u16                     aux_data_offset;
  53         u16                     end_offset;
  54 };
  55
  56 struct btree_write {
  57         struct journal_entry_pin        journal;
  58 };
  59
  60 struct btree_alloc {
  61         struct open_buckets     ob;
  62         __BKEY_PADDED(k, BKEY_BTREE_PTR_VAL_U64s_MAX);
  63 };
  64
  65 struct btree_bkey_cached_common {
  66         struct six_lock         lock;
  67         u8                      level;
  68         u8                      btree_id;
  69         bool                    cached;
  70 };
  71
  72 struct btree {
  73         struct btree_bkey_cached_common c;
  74
  75         struct rhash_head       hash;
  76         u64                     hash_val;
  77
  78         unsigned long           flags;
  79         u16                     written;
  80         u8                      nsets;
  81         u8                      nr_key_bits;
  82         u16                     version_ondisk;
  83
  84         struct bkey_format      format;
  85
  86         struct btree_node       *data;
  87         void                    *aux_data;
  88
  89         /*
  90          * Sets of sorted keys - the real btree node - plus a binary search tree
  91          *
  92          * set[0] is special; set[0]->tree, set[0]->prev and set[0]->data point
  93          * to the memory we have allocated for this btree node. Additionally,
  94          * set[0]->data points to the entire btree node as it exists on disk.
  95          */
  96         struct bset_tree        set[MAX_BSETS];
  97
  98         struct btree_nr_keys    nr;
  99         u16                     sib_u64s[2];
 100         u16                     whiteout_u64s;
 101         u8                      byte_order;
 102         u8                      unpack_fn_len;
 103
 104         struct btree_write      writes[2];
 105
 106         /* Key/pointer for this btree node */
 107         __BKEY_PADDED(key, BKEY_BTREE_PTR_VAL_U64s_MAX);
 108
 109         /*
 110          * XXX: add a delete sequence number, so when bch2_btree_node_relock()
 111          * fails because the lock sequence number has changed - i.e. the
 112          * contents were modified - we can still relock the node if it's still
 113          * the one we want, without redoing the traversal
 114          */
 115
 116         /*
 117          * For asynchronous splits/interior node updates:
 118          * When we do a split, we allocate new child nodes and update the parent
 119          * node to point to them: we update the parent in memory immediately,
 120          * but then we must wait until the children have been written out before
 121          * the update to the parent can be written - this is a list of the
 122          * btree_updates that are blocking this node from being
 123          * written:
 124          */
 125         struct list_head        write_blocked;
 126
 127         /*
 128          * Also for asynchronous splits/interior node updates:
 129          * If a btree node isn't reachable yet, we don't want to kick off
 130          * another write - because that write also won't yet be reachable and
 131          * marking it as completed before it's reachable would be incorrect:
 132          */
 133         unsigned long           will_make_reachable;
 134
 135         struct open_buckets     ob;
 136
 137         /* lru list */
 138         struct list_head        list;
 139 };
 140
 141 #define BCH_BTREE_CACHE_NOT_FREED_REASONS()     \
 142         x(lock_intent)                          \
 143         x(lock_write)                           \
 144         x(dirty)                                \
 145         x(read_in_flight)                       \
 146         x(write_in_flight)                      \
 147         x(noevict)                              \
 148         x(write_blocked)                        \
 149         x(will_make_reachable)                  \
 150         x(access_bit)
 151
 152 enum bch_btree_cache_not_freed_reasons {
 153 #define x(n) BCH_BTREE_CACHE_NOT_FREED_##n,
 154         BCH_BTREE_CACHE_NOT_FREED_REASONS()
 155 #undef x
 156         BCH_BTREE_CACHE_NOT_FREED_REASONS_NR,
 157 };
 158
 159 struct btree_cache_list {
 160         unsigned                idx;
 161         struct shrinker         *shrink;
 162         struct list_head        list;
 163         size_t                  nr;
 164 };
 165
 166 struct btree_cache {
 167         struct rhashtable       table;
 168         bool                    table_init_done;
 169         /*
 170          * We never free a struct btree, except on shutdown - we just put it on
 171          * the btree_cache_freed list and reuse it later. This simplifies the
 172          * code, and it doesn't cost us much memory as the memory usage is
 173          * dominated by buffers that hold the actual btree node data and those
 174          * can be freed - and the number of struct btrees allocated is
 175          * effectively bounded.
 176          *
 177          * btree_cache_freeable effectively is a small cache - we use it because
 178          * high order page allocations can be rather expensive, and it's quite
 179          * common to delete and allocate btree nodes in quick succession. It
 180          * should never grow past ~2-3 nodes in practice.
 181          */
 182         struct mutex            lock;
 183         struct list_head        freeable;
 184         struct list_head        freed_pcpu;
 185         struct list_head        freed_nonpcpu;
 186         struct btree_cache_list live[2];
 187
 188         size_t                  nr_freeable;
 189         size_t                  nr_reserve;
 190         size_t                  nr_by_btree[BTREE_ID_NR];
 191         atomic_long_t           nr_dirty;
 192
 193         /* shrinker stats */
 194         size_t                  nr_freed;
 195         u64                     not_freed[BCH_BTREE_CACHE_NOT_FREED_REASONS_NR];
 196
 197         /*
 198          * If we need to allocate memory for a new btree node and that
 199          * allocation fails, we can cannibalize another node in the btree cache
 200          * to satisfy the allocation - lock to guarantee only one thread does
 201          * this at a time:
 202          */
 203         struct task_struct      *alloc_lock;
 204         struct closure_waitlist alloc_wait;
 205
 206         struct bbpos            pinned_nodes_start;
 207         struct bbpos            pinned_nodes_end;
 208         /* btree id mask: 0 for leaves, 1 for interior */
 209         u64                     pinned_nodes_mask[2];
 210 };
 211
 212 struct btree_node_iter {
 213         struct btree_node_iter_set {
 214                 u16     k, end;
 215         } data[MAX_BSETS];
 216 };
 217
 218 #define BTREE_ITER_FLAGS()                      \
 219         x(slots)                                \
 220         x(intent)                               \
 221         x(prefetch)                             \
 222         x(is_extents)                           \
 223         x(not_extents)                          \
 224         x(cached)                               \
 225         x(with_key_cache)                       \
 226         x(with_updates)                         \
 227         x(with_journal)                         \
 228         x(snapshot_field)                       \
 229         x(all_snapshots)                        \
 230         x(filter_snapshots)                     \
 231         x(nopreserve)                           \
 232         x(cached_nofill)                        \
 233         x(key_cache_fill)                       \
 234
 235 #define STR_HASH_FLAGS()                        \
 236         x(must_create)                          \
 237         x(must_replace)
 238
 239 #define BTREE_UPDATE_FLAGS()                    \
 240         x(internal_snapshot_node)               \
 241         x(nojournal)                            \
 242         x(key_cache_reclaim)
 243
 244
 245 /*
 246  * BTREE_TRIGGER_norun - don't run triggers at all
 247  *
 248  * BTREE_TRIGGER_transactional - we're running transactional triggers as part of
 249  * a transaction commit: triggers may generate new updates
 250  *
 251  * BTREE_TRIGGER_atomic - we're running atomic triggers during a transaction
 252  * commit: we have our journal reservation, we're holding btree node write
 253  * locks, and we know the transaction is going to commit (returning an error
 254  * here is a fatal error, causing us to go emergency read-only)
 255  *
 256  * BTREE_TRIGGER_gc - we're in gc/fsck: running triggers to recalculate e.g. disk usage
 257  *
 258  * BTREE_TRIGGER_insert - @new is entering the btree
 259  * BTREE_TRIGGER_overwrite - @old is leaving the btree
 260  *
 261  * BTREE_TRIGGER_bucket_invalidate - signal from bucket invalidate path to alloc
 262  * trigger
 263  */
 264 #define BTREE_TRIGGER_FLAGS()                   \
 265         x(norun)                                \
 266         x(transactional)                        \
 267         x(atomic)                               \
 268         x(check_repair)                         \
 269         x(gc)                                   \
 270         x(insert)                               \
 271         x(overwrite)                            \
 272         x(is_root)                              \
 273         x(bucket_invalidate)
 274
 275 enum {
 276 #define x(n) BTREE_ITER_FLAG_BIT_##n,
 277         BTREE_ITER_FLAGS()
 278         STR_HASH_FLAGS()
 279         BTREE_UPDATE_FLAGS()
 280         BTREE_TRIGGER_FLAGS()
 281 #undef x
 282 };
 283
 284 /* iter flags must fit in a u16: */
 285 //BUILD_BUG_ON(BTREE_ITER_FLAG_BIT_key_cache_fill > 15);
 286
 287 enum btree_iter_update_trigger_flags {
 288 #define x(n) BTREE_ITER_##n     = 1U << BTREE_ITER_FLAG_BIT_##n,
 289         BTREE_ITER_FLAGS()
 290 #undef x
 291 #define x(n) STR_HASH_##n       = 1U << BTREE_ITER_FLAG_BIT_##n,
 292         STR_HASH_FLAGS()
 293 #undef x
 294 #define x(n) BTREE_UPDATE_##n   = 1U << BTREE_ITER_FLAG_BIT_##n,
 295         BTREE_UPDATE_FLAGS()
 296 #undef x
 297 #define x(n) BTREE_TRIGGER_##n  = 1U << BTREE_ITER_FLAG_BIT_##n,
 298         BTREE_TRIGGER_FLAGS()
 299 #undef x
 300 };
 301
 302 enum btree_path_uptodate {
 303         BTREE_ITER_UPTODATE             = 0,
 304         BTREE_ITER_NEED_RELOCK          = 1,
 305         BTREE_ITER_NEED_TRAVERSE        = 2,
 306 };
 307
 308 #if defined(CONFIG_BCACHEFS_LOCK_TIME_STATS) || defined(CONFIG_BCACHEFS_DEBUG)
 309 #define TRACK_PATH_ALLOCATED
 310 #endif
 311
 312 typedef u16 btree_path_idx_t;
 313
 314 struct btree_path {
 315         btree_path_idx_t        sorted_idx;
 316         u8                      ref;
 317         u8                      intent_ref;
 318
 319         /* btree_iter_copy starts here: */
 320         struct bpos             pos;
 321
 322         enum btree_id           btree_id:5;
 323         bool                    cached:1;
 324         bool                    preserve:1;
 325         enum btree_path_uptodate uptodate:2;
 326         /*
 327          * When true, failing to relock this path will cause the transaction to
 328          * restart:
 329          */
 330         bool                    should_be_locked:1;
 331         unsigned                level:3,
 332                                 locks_want:3;
 333         u8                      nodes_locked;
 334
 335         struct btree_path_level {
 336                 struct btree    *b;
 337                 struct btree_node_iter iter;
 338                 u32             lock_seq;
 339 #ifdef CONFIG_BCACHEFS_LOCK_TIME_STATS
 340                 u64             lock_taken_time;
 341 #endif
 342         }                       l[BTREE_MAX_DEPTH];
 343 #ifdef TRACK_PATH_ALLOCATED
 344         unsigned long           ip_allocated;
 345 #endif
 346 };
 347
 348 static inline struct btree_path_level *path_l(struct btree_path *path)
 349 {
 350         return path->l + path->level;
 351 }
 352
 353 static inline unsigned long btree_path_ip_allocated(struct btree_path *path)
 354 {
 355 #ifdef TRACK_PATH_ALLOCATED
 356         return path->ip_allocated;
 357 #else
 358         return _THIS_IP_;
 359 #endif
 360 }
 361
 362 /*
 363  * @pos                 - iterator's current position
 364  * @level               - current btree depth
 365  * @locks_want          - btree level below which we start taking intent locks
 366  * @nodes_locked        - bitmask indicating which nodes in @nodes are locked
 367  * @nodes_intent_locked - bitmask indicating which locks are intent locks
 368  */
 369 struct btree_iter {
 370         struct btree_trans      *trans;
 371         btree_path_idx_t        path;
 372         btree_path_idx_t        update_path;
 373         btree_path_idx_t        key_cache_path;
 374
 375         enum btree_id           btree_id:8;
 376         u8                      min_depth;
 377
 378         /* btree_iter_copy starts here: */
 379         u16                     flags;
 380
 381         /* When we're filtering by snapshot, the snapshot ID we're looking for: */
 382         unsigned                snapshot;
 383
 384         struct bpos             pos;
 385         /*
 386          * Current unpacked key - so that bch2_btree_iter_next()/
 387          * bch2_btree_iter_next_slot() can correctly advance pos.
 388          */
 389         struct bkey             k;
 390
 391         /* BTREE_ITER_with_journal: */
 392         size_t                  journal_idx;
 393 #ifdef TRACK_PATH_ALLOCATED
 394         unsigned long           ip_allocated;
 395 #endif
 396 };
 397
 398 #define BKEY_CACHED_ACCESSED            0
 399 #define BKEY_CACHED_DIRTY               1
 400
 401 struct bkey_cached {
 402         struct btree_bkey_cached_common c;
 403
 404         unsigned long           flags;
 405         u16                     u64s;
 406         struct bkey_cached_key  key;
 407
 408         struct rhash_head       hash;
 409
 410         struct journal_entry_pin journal;
 411         u64                     seq;
 412
 413         struct bkey_i           *k;
 414         struct rcu_head         rcu;
 415 };
 416
 417 static inline struct bpos btree_node_pos(struct btree_bkey_cached_common *b)
 418 {
 419         return !b->cached
 420                 ? container_of(b, struct btree, c)->key.k.p
 421                 : container_of(b, struct bkey_cached, c)->key.pos;
 422 }
 423
 424 struct btree_insert_entry {
 425         unsigned                flags;
 426         u8                      bkey_type;
 427         enum btree_id           btree_id:8;
 428         u8                      level:4;
 429         bool                    cached:1;
 430         bool                    insert_trigger_run:1;
 431         bool                    overwrite_trigger_run:1;
 432         bool                    key_cache_already_flushed:1;
 433         /*
 434          * @old_k may be a key from the journal; @old_btree_u64s always refers
 435          * to the size of the key being overwritten in the btree:
 436          */
 437         u8                      old_btree_u64s;
 438         btree_path_idx_t        path;
 439         struct bkey_i           *k;
 440         /* key being overwritten: */
 441         struct bkey             old_k;
 442         const struct bch_val    *old_v;
 443         unsigned long           ip_allocated;
 444 };
 445
 446 /* Number of btree paths we preallocate, usually enough */
 447 #define BTREE_ITER_INITIAL              64
 448 /*
 449  * Lmiit for btree_trans_too_many_iters(); this is enough that almost all code
 450  * paths should run inside this limit, and if they don't it usually indicates a
 451  * bug (leaking/duplicated btree paths).
 452  *
 453  * exception: some fsck paths
 454  *
 455  * bugs with excessive path usage seem to have possibly been eliminated now, so
 456  * we might consider eliminating this (and btree_trans_too_many_iter()) at some
 457  * point.
 458  */
 459 #define BTREE_ITER_NORMAL_LIMIT         256
 460 /* never exceed limit */
 461 #define BTREE_ITER_MAX                  (1U << 10)
 462
 463 struct btree_trans_commit_hook;
 464 typedef int (btree_trans_commit_hook_fn)(struct btree_trans *, struct btree_trans_commit_hook *);
 465
 466 struct btree_trans_commit_hook {
 467         btree_trans_commit_hook_fn      *fn;
 468         struct btree_trans_commit_hook  *next;
 469 };
 470
 471 #define BTREE_TRANS_MEM_MAX     (1U << 16)
 472
 473 #define BTREE_TRANS_MAX_LOCK_HOLD_TIME_NS       10000
 474
 475 struct btree_trans_paths {
 476         unsigned long           nr_paths;
 477         struct btree_path       paths[];
 478 };
 479
 480 struct btree_trans {
 481         struct bch_fs           *c;
 482
 483         unsigned long           *paths_allocated;
 484         struct btree_path       *paths;
 485         btree_path_idx_t        *sorted;
 486         struct btree_insert_entry *updates;
 487
 488         void                    *mem;
 489         unsigned                mem_top;
 490         unsigned                mem_bytes;
 491
 492         btree_path_idx_t        nr_sorted;
 493         btree_path_idx_t        nr_paths;
 494         btree_path_idx_t        nr_paths_max;
 495         btree_path_idx_t        nr_updates;
 496         u8                      fn_idx;
 497         u8                      lock_must_abort;
 498         bool                    lock_may_not_fail:1;
 499         bool                    srcu_held:1;
 500         bool                    locked:1;
 501         bool                    pf_memalloc_nofs:1;
 502         bool                    write_locked:1;
 503         bool                    used_mempool:1;
 504         bool                    in_traverse_all:1;
 505         bool                    paths_sorted:1;
 506         bool                    memory_allocation_failure:1;
 507         bool                    journal_transaction_names:1;
 508         bool                    journal_replay_not_finished:1;
 509         bool                    notrace_relock_fail:1;
 510         enum bch_errcode        restarted:16;
 511         u32                     restart_count;
 512
 513         u64                     last_begin_time;
 514         unsigned long           last_begin_ip;
 515         unsigned long           last_restarted_ip;
 516         unsigned long           last_unlock_ip;
 517         unsigned long           srcu_lock_time;
 518
 519         const char              *fn;
 520         struct btree_bkey_cached_common *locking;
 521         struct six_lock_waiter  locking_wait;
 522         int                     srcu_idx;
 523
 524         /* update path: */
 525         u16                     journal_entries_u64s;
 526         u16                     journal_entries_size;
 527         struct jset_entry       *journal_entries;
 528
 529         struct btree_trans_commit_hook *hooks;
 530         struct journal_entry_pin *journal_pin;
 531
 532         struct journal_res      journal_res;
 533         u64                     *journal_seq;
 534         struct disk_reservation *disk_res;
 535
 536         struct bch_fs_usage_base fs_usage_delta;
 537
 538         unsigned                journal_u64s;
 539         unsigned                extra_disk_res; /* XXX kill */
 540
 541 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 542         struct lockdep_map      dep_map;
 543 #endif
 544         /* Entries before this are zeroed out on every bch2_trans_get() call */
 545
 546         struct list_head        list;
 547         struct closure          ref;
 548
 549         unsigned long           _paths_allocated[BITS_TO_LONGS(BTREE_ITER_INITIAL)];
 550         struct btree_trans_paths trans_paths;
 551         struct btree_path       _paths[BTREE_ITER_INITIAL];
 552         btree_path_idx_t        _sorted[BTREE_ITER_INITIAL + 4];
 553         struct btree_insert_entry _updates[BTREE_ITER_INITIAL];
 554 };
 555
 556 static inline struct btree_path *btree_iter_path(struct btree_trans *trans, struct btree_iter *iter)
 557 {
 558         return trans->paths + iter->path;
 559 }
 560
 561 static inline struct btree_path *btree_iter_key_cache_path(struct btree_trans *trans, struct btree_iter *iter)
 562 {
 563         return iter->key_cache_path
 564                 ? trans->paths + iter->key_cache_path
 565                 : NULL;
 566 }
 567
 568 #define BCH_BTREE_WRITE_TYPES()                                         \
 569         x(initial,              0)                                      \
 570         x(init_next_bset,       1)                                      \
 571         x(cache_reclaim,        2)                                      \
 572         x(journal_reclaim,      3)                                      \
 573         x(interior,             4)
 574
 575 enum btree_write_type {
 576 #define x(t, n) BTREE_WRITE_##t,
 577         BCH_BTREE_WRITE_TYPES()
 578 #undef x
 579         BTREE_WRITE_TYPE_NR,
 580 };
 581
 582 #define BTREE_WRITE_TYPE_MASK   (roundup_pow_of_two(BTREE_WRITE_TYPE_NR) - 1)
 583 #define BTREE_WRITE_TYPE_BITS   ilog2(roundup_pow_of_two(BTREE_WRITE_TYPE_NR))
 584
 585 #define BTREE_FLAGS()                                                   \
 586         x(read_in_flight)                                               \
 587         x(read_error)                                                   \
 588         x(dirty)                                                        \
 589         x(need_write)                                                   \
 590         x(write_blocked)                                                \
 591         x(will_make_reachable)                                          \
 592         x(noevict)                                                      \
 593         x(write_idx)                                                    \
 594         x(accessed)                                                     \
 595         x(write_in_flight)                                              \
 596         x(write_in_flight_inner)                                        \
 597         x(just_written)                                                 \
 598         x(dying)                                                        \
 599         x(fake)                                                         \
 600         x(need_rewrite)                                                 \
 601         x(never_write)                                                  \
 602         x(pinned)
 603
 604 enum btree_flags {
 605         /* First bits for btree node write type */
 606         BTREE_NODE_FLAGS_START = BTREE_WRITE_TYPE_BITS - 1,
 607 #define x(flag) BTREE_NODE_##flag,
 608         BTREE_FLAGS()
 609 #undef x
 610 };
 611
 612 #define x(flag)                                                         \
 613 static inline bool btree_node_ ## flag(struct btree *b)                 \
 614 {       return test_bit(BTREE_NODE_ ## flag, &b->flags); }              \
 615                                                                         \
 616 static inline void set_btree_node_ ## flag(struct btree *b)             \
 617 {       set_bit(BTREE_NODE_ ## flag, &b->flags); }                      \
 618                                                                         \
 619 static inline void clear_btree_node_ ## flag(struct btree *b)           \
 620 {       clear_bit(BTREE_NODE_ ## flag, &b->flags); }
 621
 622 BTREE_FLAGS()
 623 #undef x
 624
 625 static inline struct btree_write *btree_current_write(struct btree *b)
 626 {
 627         return b->writes + btree_node_write_idx(b);
 628 }
 629
 630 static inline struct btree_write *btree_prev_write(struct btree *b)
 631 {
 632         return b->writes + (btree_node_write_idx(b) ^ 1);
 633 }
 634
 635 static inline struct bset_tree *bset_tree_last(struct btree *b)
 636 {
 637         EBUG_ON(!b->nsets);
 638         return b->set + b->nsets - 1;
 639 }
 640
 641 static inline void *
 642 __btree_node_offset_to_ptr(const struct btree *b, u16 offset)
 643 {
 644         return (void *) ((u64 *) b->data + 1 + offset);
 645 }
 646
 647 static inline u16
 648 __btree_node_ptr_to_offset(const struct btree *b, const void *p)
 649 {
 650         u16 ret = (u64 *) p - 1 - (u64 *) b->data;
 651
 652         EBUG_ON(__btree_node_offset_to_ptr(b, ret) != p);
 653         return ret;
 654 }
 655
 656 static inline struct bset *bset(const struct btree *b,
 657                                 const struct bset_tree *t)
 658 {
 659         return __btree_node_offset_to_ptr(b, t->data_offset);
 660 }
 661
 662 static inline void set_btree_bset_end(struct btree *b, struct bset_tree *t)
 663 {
 664         t->end_offset =
 665                 __btree_node_ptr_to_offset(b, vstruct_last(bset(b, t)));
 666 }
 667
 668 static inline void set_btree_bset(struct btree *b, struct bset_tree *t,
 669                                   const struct bset *i)
 670 {
 671         t->data_offset = __btree_node_ptr_to_offset(b, i);
 672         set_btree_bset_end(b, t);
 673 }
 674
 675 static inline struct bset *btree_bset_first(struct btree *b)
 676 {
 677         return bset(b, b->set);
 678 }
 679
 680 static inline struct bset *btree_bset_last(struct btree *b)
 681 {
 682         return bset(b, bset_tree_last(b));
 683 }
 684
 685 static inline u16
 686 __btree_node_key_to_offset(const struct btree *b, const struct bkey_packed *k)
 687 {
 688         return __btree_node_ptr_to_offset(b, k);
 689 }
 690
 691 static inline struct bkey_packed *
 692 __btree_node_offset_to_key(const struct btree *b, u16 k)
 693 {
 694         return __btree_node_offset_to_ptr(b, k);
 695 }
 696
 697 static inline unsigned btree_bkey_first_offset(const struct bset_tree *t)
 698 {
 699         return t->data_offset + offsetof(struct bset, _data) / sizeof(u64);
 700 }
 701
 702 #define btree_bkey_first(_b, _t)                                        \
 703 ({                                                                      \
 704         EBUG_ON(bset(_b, _t)->start !=                                  \
 705                 __btree_node_offset_to_key(_b, btree_bkey_first_offset(_t)));\
 706                                                                         \
 707         bset(_b, _t)->start;                                            \
 708 })
 709
 710 #define btree_bkey_last(_b, _t)                                         \
 711 ({                                                                      \
 712         EBUG_ON(__btree_node_offset_to_key(_b, (_t)->end_offset) !=     \
 713                 vstruct_last(bset(_b, _t)));                            \
 714                                                                         \
 715         __btree_node_offset_to_key(_b, (_t)->end_offset);               \
 716 })
 717
 718 static inline unsigned bset_u64s(struct bset_tree *t)
 719 {
 720         return t->end_offset - t->data_offset -
 721                 sizeof(struct bset) / sizeof(u64);
 722 }
 723
 724 static inline unsigned bset_dead_u64s(struct btree *b, struct bset_tree *t)
 725 {
 726         return bset_u64s(t) - b->nr.bset_u64s[t - b->set];
 727 }
 728
 729 static inline unsigned bset_byte_offset(struct btree *b, void *i)
 730 {
 731         return i - (void *) b->data;
 732 }
 733
 734 enum btree_node_type {
 735         BKEY_TYPE_btree,
 736 #define x(kwd, val, ...) BKEY_TYPE_##kwd = val + 1,
 737         BCH_BTREE_IDS()
 738 #undef x
 739         BKEY_TYPE_NR
 740 };
 741
 742 /* Type of a key in btree @id at level @level: */
 743 static inline enum btree_node_type __btree_node_type(unsigned level, enum btree_id id)
 744 {
 745         return level ? BKEY_TYPE_btree : (unsigned) id + 1;
 746 }
 747
 748 /* Type of keys @b contains: */
 749 static inline enum btree_node_type btree_node_type(struct btree *b)
 750 {
 751         return __btree_node_type(b->c.level, b->c.btree_id);
 752 }
 753
 754 const char *bch2_btree_node_type_str(enum btree_node_type);
 755
 756 #define BTREE_NODE_TYPE_HAS_TRANS_TRIGGERS              \
 757         (BIT_ULL(BKEY_TYPE_extents)|                    \
 758          BIT_ULL(BKEY_TYPE_alloc)|                      \
 759          BIT_ULL(BKEY_TYPE_inodes)|                     \
 760          BIT_ULL(BKEY_TYPE_stripes)|                    \
 761          BIT_ULL(BKEY_TYPE_reflink)|                    \
 762          BIT_ULL(BKEY_TYPE_subvolumes)|                 \
 763          BIT_ULL(BKEY_TYPE_btree))
 764
 765 #define BTREE_NODE_TYPE_HAS_ATOMIC_TRIGGERS             \
 766         (BIT_ULL(BKEY_TYPE_alloc)|                      \
 767          BIT_ULL(BKEY_TYPE_inodes)|                     \
 768          BIT_ULL(BKEY_TYPE_stripes)|                    \
 769          BIT_ULL(BKEY_TYPE_snapshots))
 770
 771 #define BTREE_NODE_TYPE_HAS_TRIGGERS                    \
 772         (BTREE_NODE_TYPE_HAS_TRANS_TRIGGERS|            \
 773          BTREE_NODE_TYPE_HAS_ATOMIC_TRIGGERS)
 774
 775 static inline bool btree_node_type_has_trans_triggers(enum btree_node_type type)
 776 {
 777         return BIT_ULL(type) & BTREE_NODE_TYPE_HAS_TRANS_TRIGGERS;
 778 }
 779
 780 static inline bool btree_node_type_has_atomic_triggers(enum btree_node_type type)
 781 {
 782         return BIT_ULL(type) & BTREE_NODE_TYPE_HAS_ATOMIC_TRIGGERS;
 783 }
 784
 785 static inline bool btree_node_type_has_triggers(enum btree_node_type type)
 786 {
 787         return BIT_ULL(type) & BTREE_NODE_TYPE_HAS_TRIGGERS;
 788 }
 789
 790 static inline bool btree_node_type_is_extents(enum btree_node_type type)
 791 {
 792         const u64 mask = 0
 793 #define x(name, nr, flags, ...) |((!!((flags) & BTREE_ID_EXTENTS)) << (nr + 1))
 794         BCH_BTREE_IDS()
 795 #undef x
 796         ;
 797
 798         return BIT_ULL(type) & mask;
 799 }
 800
 801 static inline bool btree_id_is_extents(enum btree_id btree)
 802 {
 803         return btree_node_type_is_extents(__btree_node_type(0, btree));
 804 }
 805
 806 static inline bool btree_type_has_snapshots(enum btree_id id)
 807 {
 808         const u64 mask = 0
 809 #define x(name, nr, flags, ...) |((!!((flags) & BTREE_ID_SNAPSHOTS)) << nr)
 810         BCH_BTREE_IDS()
 811 #undef x
 812         ;
 813
 814         return BIT_ULL(id) & mask;
 815 }
 816
 817 static inline bool btree_type_has_snapshot_field(enum btree_id id)
 818 {
 819         const u64 mask = 0
 820 #define x(name, nr, flags, ...) |((!!((flags) & (BTREE_ID_SNAPSHOT_FIELD|BTREE_ID_SNAPSHOTS))) << nr)
 821         BCH_BTREE_IDS()
 822 #undef x
 823         ;
 824
 825         return BIT_ULL(id) & mask;
 826 }
 827
 828 static inline bool btree_type_has_ptrs(enum btree_id id)
 829 {
 830         const u64 mask = 0
 831 #define x(name, nr, flags, ...) |((!!((flags) & BTREE_ID_DATA)) << nr)
 832         BCH_BTREE_IDS()
 833 #undef x
 834         ;
 835
 836         return BIT_ULL(id) & mask;
 837 }
 838
 839 struct btree_root {
 840         struct btree            *b;
 841
 842         /* On disk root - see async splits: */
 843         __BKEY_PADDED(key, BKEY_BTREE_PTR_VAL_U64s_MAX);
 844         u8                      level;
 845         u8                      alive;
 846         s16                     error;
 847 };
 848
 849 enum btree_gc_coalesce_fail_reason {
 850         BTREE_GC_COALESCE_FAIL_RESERVE_GET,
 851         BTREE_GC_COALESCE_FAIL_KEYLIST_REALLOC,
 852         BTREE_GC_COALESCE_FAIL_FORMAT_FITS,
 853 };
 854
 855 enum btree_node_sibling {
 856         btree_prev_sib,
 857         btree_next_sib,
 858 };
 859
 860 struct get_locks_fail {
 861         unsigned        l;
 862         struct btree    *b;
 863 };
 864
 865 #endif /* _BCACHEFS_BTREE_TYPES_H */