fs/bcachefs/btree_update_interior.h

   1 /* SPDX-License-Identifier: GPL-2.0 */
   2 #ifndef _BCACHEFS_BTREE_UPDATE_INTERIOR_H
   3 #define _BCACHEFS_BTREE_UPDATE_INTERIOR_H
   4
   5 #include "btree_cache.h"
   6 #include "btree_locking.h"
   7 #include "btree_update.h"
   8
   9 #define BTREE_UPDATE_NODES_MAX          ((BTREE_MAX_DEPTH - 2) * 2 + GC_MERGE_NODES)
  10
  11 #define BTREE_UPDATE_JOURNAL_RES        (BTREE_UPDATE_NODES_MAX * (BKEY_BTREE_PTR_U64s_MAX + 1))
  12
  13 int bch2_btree_node_check_topology(struct btree_trans *, struct btree *);
  14
  15 #define BTREE_UPDATE_MODES()    \
  16         x(none)                 \
  17         x(node)                 \
  18         x(root)                 \
  19         x(update)
  20
  21 enum btree_update_mode {
  22 #define x(n)    BTREE_UPDATE_##n,
  23         BTREE_UPDATE_MODES()
  24 #undef x
  25 };
  26
  27 /*
  28  * Tracks an in progress split/rewrite of a btree node and the update to the
  29  * parent node:
  30  *
  31  * When we split/rewrite a node, we do all the updates in memory without
  32  * waiting for any writes to complete - we allocate the new node(s) and update
  33  * the parent node, possibly recursively up to the root.
  34  *
  35  * The end result is that we have one or more new nodes being written -
  36  * possibly several, if there were multiple splits - and then a write (updating
  37  * an interior node) which will make all these new nodes visible.
  38  *
  39  * Additionally, as we split/rewrite nodes we free the old nodes - but the old
  40  * nodes can't be freed (their space on disk can't be reclaimed) until the
  41  * update to the interior node that makes the new node visible completes -
  42  * until then, the old nodes are still reachable on disk.
  43  *
  44  */
  45 struct btree_update {
  46         struct closure                  cl;
  47         struct bch_fs                   *c;
  48         u64                             start_time;
  49         unsigned long                   ip_started;
  50
  51         struct list_head                list;
  52         struct list_head                unwritten_list;
  53
  54         enum btree_update_mode          mode;
  55         enum bch_watermark              watermark;
  56         unsigned                        nodes_written:1;
  57         unsigned                        took_gc_lock:1;
  58
  59         enum btree_id                   btree_id;
  60         unsigned                        update_level_start;
  61         unsigned                        update_level_end;
  62
  63         struct disk_reservation         disk_res;
  64
  65         /*
  66          * BTREE_UPDATE_node:
  67          * The update that made the new nodes visible was a regular update to an
  68          * existing interior node - @b. We can't write out the update to @b
  69          * until the new nodes we created are finished writing, so we block @b
  70          * from writing by putting this btree_interior update on the
  71          * @b->write_blocked list with @write_blocked_list:
  72          */
  73         struct btree                    *b;
  74         struct list_head                write_blocked_list;
  75
  76         /*
  77          * We may be freeing nodes that were dirty, and thus had journal entries
  78          * pinned: we need to transfer the oldest of those pins to the
  79          * btree_update operation, and release it when the new node(s)
  80          * are all persistent and reachable:
  81          */
  82         struct journal_entry_pin        journal;
  83
  84         /* Preallocated nodes we reserve when we start the update: */
  85         struct prealloc_nodes {
  86                 struct btree            *b[BTREE_UPDATE_NODES_MAX];
  87                 unsigned                nr;
  88         }                               prealloc_nodes[2];
  89
  90         /* Nodes being freed: */
  91         struct keylist                  old_keys;
  92         u64                             _old_keys[BTREE_UPDATE_NODES_MAX *
  93                                                   BKEY_BTREE_PTR_U64s_MAX];
  94
  95         /* Nodes being added: */
  96         struct keylist                  new_keys;
  97         u64                             _new_keys[BTREE_UPDATE_NODES_MAX *
  98                                                   BKEY_BTREE_PTR_U64s_MAX];
  99
 100         /* New nodes, that will be made reachable by this update: */
 101         struct btree                    *new_nodes[BTREE_UPDATE_NODES_MAX];
 102         unsigned                        nr_new_nodes;
 103
 104         struct btree                    *old_nodes[BTREE_UPDATE_NODES_MAX];
 105         __le64                          old_nodes_seq[BTREE_UPDATE_NODES_MAX];
 106         unsigned                        nr_old_nodes;
 107
 108         open_bucket_idx_t               open_buckets[BTREE_UPDATE_NODES_MAX *
 109                                                      BCH_REPLICAS_MAX];
 110         open_bucket_idx_t               nr_open_buckets;
 111
 112         unsigned                        journal_u64s;
 113         u64                             journal_entries[BTREE_UPDATE_JOURNAL_RES];
 114
 115         /* Only here to reduce stack usage on recursive splits: */
 116         struct keylist                  parent_keys;
 117         /*
 118          * Enough room for btree_split's keys without realloc - btree node
 119          * pointers never have crc/compression info, so we only need to acount
 120          * for the pointers for three keys
 121          */
 122         u64                             inline_keys[BKEY_BTREE_PTR_U64s_MAX * 3];
 123 };
 124
 125 struct btree *__bch2_btree_node_alloc_replacement(struct btree_update *,
 126                                                   struct btree_trans *,
 127                                                   struct btree *,
 128                                                   struct bkey_format);
 129
 130 int bch2_btree_split_leaf(struct btree_trans *, btree_path_idx_t, unsigned);
 131
 132 int bch2_btree_increase_depth(struct btree_trans *, btree_path_idx_t, unsigned);
 133
 134 int __bch2_foreground_maybe_merge(struct btree_trans *, btree_path_idx_t,
 135                                   unsigned, unsigned, enum btree_node_sibling);
 136
 137 static inline int bch2_foreground_maybe_merge_sibling(struct btree_trans *trans,
 138                                         btree_path_idx_t path_idx,
 139                                         unsigned level, unsigned flags,
 140                                         enum btree_node_sibling sib)
 141 {
 142         struct btree_path *path = trans->paths + path_idx;
 143         struct btree *b;
 144
 145         EBUG_ON(!btree_node_locked(path, level));
 146
 147         b = path->l[level].b;
 148         if (b->sib_u64s[sib] > trans->c->btree_foreground_merge_threshold)
 149                 return 0;
 150
 151         return __bch2_foreground_maybe_merge(trans, path_idx, level, flags, sib);
 152 }
 153
 154 static inline int bch2_foreground_maybe_merge(struct btree_trans *trans,
 155                                               btree_path_idx_t path,
 156                                               unsigned level,
 157                                               unsigned flags)
 158 {
 159         return  bch2_foreground_maybe_merge_sibling(trans, path, level, flags,
 160                                                     btree_prev_sib) ?:
 161                 bch2_foreground_maybe_merge_sibling(trans, path, level, flags,
 162                                                     btree_next_sib);
 163 }
 164
 165 int bch2_btree_node_rewrite(struct btree_trans *, struct btree_iter *,
 166                             struct btree *, unsigned);
 167 void bch2_btree_node_rewrite_async(struct bch_fs *, struct btree *);
 168 int bch2_btree_node_update_key(struct btree_trans *, struct btree_iter *,
 169                                struct btree *, struct bkey_i *,
 170                                unsigned, bool);
 171 int bch2_btree_node_update_key_get_iter(struct btree_trans *, struct btree *,
 172                                         struct bkey_i *, unsigned, bool);
 173
 174 void bch2_btree_set_root_for_read(struct bch_fs *, struct btree *);
 175 void bch2_btree_root_alloc_fake(struct bch_fs *, enum btree_id, unsigned);
 176
 177 static inline unsigned btree_update_reserve_required(struct bch_fs *c,
 178                                                      struct btree *b)
 179 {
 180         unsigned depth = btree_node_root(c, b)->c.level + 1;
 181
 182         /*
 183          * Number of nodes we might have to allocate in a worst case btree
 184          * split operation - we split all the way up to the root, then allocate
 185          * a new root, unless we're already at max depth:
 186          */
 187         if (depth < BTREE_MAX_DEPTH)
 188                 return (depth - b->c.level) * 2 + 1;
 189         else
 190                 return (depth - b->c.level) * 2 - 1;
 191 }
 192
 193 static inline void btree_node_reset_sib_u64s(struct btree *b)
 194 {
 195         b->sib_u64s[0] = b->nr.live_u64s;
 196         b->sib_u64s[1] = b->nr.live_u64s;
 197 }
 198
 199 static inline void *btree_data_end(struct btree *b)
 200 {
 201         return (void *) b->data + btree_buf_bytes(b);
 202 }
 203
 204 static inline struct bkey_packed *unwritten_whiteouts_start(struct btree *b)
 205 {
 206         return (void *) ((u64 *) btree_data_end(b) - b->whiteout_u64s);
 207 }
 208
 209 static inline struct bkey_packed *unwritten_whiteouts_end(struct btree *b)
 210 {
 211         return btree_data_end(b);
 212 }
 213
 214 static inline void *write_block(struct btree *b)
 215 {
 216         return (void *) b->data + (b->written << 9);
 217 }
 218
 219 static inline bool __btree_addr_written(struct btree *b, void *p)
 220 {
 221         return p < write_block(b);
 222 }
 223
 224 static inline bool bset_written(struct btree *b, struct bset *i)
 225 {
 226         return __btree_addr_written(b, i);
 227 }
 228
 229 static inline bool bkey_written(struct btree *b, struct bkey_packed *k)
 230 {
 231         return __btree_addr_written(b, k);
 232 }
 233
 234 static inline ssize_t __bch2_btree_u64s_remaining(struct btree *b, void *end)
 235 {
 236         ssize_t used = bset_byte_offset(b, end) / sizeof(u64) +
 237                 b->whiteout_u64s;
 238         ssize_t total = btree_buf_bytes(b) >> 3;
 239
 240         /* Always leave one extra u64 for bch2_varint_decode: */
 241         used++;
 242
 243         return total - used;
 244 }
 245
 246 static inline size_t bch2_btree_keys_u64s_remaining(struct btree *b)
 247 {
 248         ssize_t remaining = __bch2_btree_u64s_remaining(b,
 249                                 btree_bkey_last(b, bset_tree_last(b)));
 250
 251         BUG_ON(remaining < 0);
 252
 253         if (bset_written(b, btree_bset_last(b)))
 254                 return 0;
 255
 256         return remaining;
 257 }
 258
 259 #define BTREE_WRITE_SET_U64s_BITS       9
 260
 261 static inline unsigned btree_write_set_buffer(struct btree *b)
 262 {
 263         /*
 264          * Could buffer up larger amounts of keys for btrees with larger keys,
 265          * pending benchmarking:
 266          */
 267         return 8 << BTREE_WRITE_SET_U64s_BITS;
 268 }
 269
 270 static inline struct btree_node_entry *want_new_bset(struct bch_fs *c, struct btree *b)
 271 {
 272         struct bset_tree *t = bset_tree_last(b);
 273         struct btree_node_entry *bne = max(write_block(b),
 274                         (void *) btree_bkey_last(b, bset_tree_last(b)));
 275         ssize_t remaining_space =
 276                 __bch2_btree_u64s_remaining(b, bne->keys.start);
 277
 278         if (unlikely(bset_written(b, bset(b, t)))) {
 279                 if (remaining_space > (ssize_t) (block_bytes(c) >> 3))
 280                         return bne;
 281         } else {
 282                 if (unlikely(bset_u64s(t) * sizeof(u64) > btree_write_set_buffer(b)) &&
 283                     remaining_space > (ssize_t) (btree_write_set_buffer(b) >> 3))
 284                         return bne;
 285         }
 286
 287         return NULL;
 288 }
 289
 290 static inline void push_whiteout(struct btree *b, struct bpos pos)
 291 {
 292         struct bkey_packed k;
 293
 294         BUG_ON(bch2_btree_keys_u64s_remaining(b) < BKEY_U64s);
 295         EBUG_ON(btree_node_just_written(b));
 296
 297         if (!bkey_pack_pos(&k, pos, b)) {
 298                 struct bkey *u = (void *) &k;
 299
 300                 bkey_init(u);
 301                 u->p = pos;
 302         }
 303
 304         k.needs_whiteout = true;
 305
 306         b->whiteout_u64s += k.u64s;
 307         bkey_p_copy(unwritten_whiteouts_start(b), &k);
 308 }
 309
 310 /*
 311  * write lock must be held on @b (else the dirty bset that we were going to
 312  * insert into could be written out from under us)
 313  */
 314 static inline bool bch2_btree_node_insert_fits(struct btree *b, unsigned u64s)
 315 {
 316         if (unlikely(btree_node_need_rewrite(b)))
 317                 return false;
 318
 319         return u64s <= bch2_btree_keys_u64s_remaining(b);
 320 }
 321
 322 void bch2_btree_updates_to_text(struct printbuf *, struct bch_fs *);
 323
 324 bool bch2_btree_interior_updates_flush(struct bch_fs *);
 325
 326 void bch2_journal_entry_to_btree_root(struct bch_fs *, struct jset_entry *);
 327 struct jset_entry *bch2_btree_roots_to_journal_entries(struct bch_fs *,
 328                                         struct jset_entry *, unsigned long);
 329
 330 void bch2_do_pending_node_rewrites(struct bch_fs *);
 331 void bch2_free_pending_node_rewrites(struct bch_fs *);
 332
 333 void bch2_fs_btree_interior_update_exit(struct bch_fs *);
 334 void bch2_fs_btree_interior_update_init_early(struct bch_fs *);
 335 int bch2_fs_btree_interior_update_init(struct bch_fs *);
 336
 337 #endif /* _BCACHEFS_BTREE_UPDATE_INTERIOR_H */