| 1 | |
| 2 | |
| 3 | * Copyright (C) 2011 STRATO. All rights reserved.
|
| 4 |
|
| 5 |
|
| 6 | #include <linux/mm.h>
|
| 7 | #include <linux/rbtree.h>
|
| 8 | #include <trace/events/btrfs.h>
|
| 9 | #include "ctree.h"
|
| 10 | #include "disk-io.h"
|
| 11 | #include "backref.h"
|
| 12 | #include "ulist.h"
|
| 13 | #include "transaction.h"
|
| 14 | #include "delayed-ref.h"
|
| 15 | #include "locking.h"
|
| 16 | #include "misc.h"
|
| 17 | #include "tree-mod-log.h"
|
| 18 | #include "fs.h"
|
| 19 | #include "accessors.h"
|
| 20 | #include "extent-tree.h"
|
| 21 | #include "relocation.h"
|
| 22 | #include "tree-checker.h"
|
| 23 |
|
| 24 |
|
| 25 | #define BACKREF_FOUND_SHARED 6
|
| 26 | #define BACKREF_FOUND_NOT_SHARED 7
|
| 27 |
|
| 28 | struct extent_inode_elem {
|
| 29 | u64 inum;
|
| 30 | u64 offset;
|
| 31 | u64 num_bytes;
|
| 32 | struct extent_inode_elem *next;
|
| 33 | };
|
| 34 |
|
| 35 | static int check_extent_in_eb(struct btrfs_backref_walk_ctx *ctx,
|
| 36 | const struct btrfs_key *key,
|
| 37 | const struct extent_buffer *eb,
|
| 38 | const struct btrfs_file_extent_item *fi,
|
| 39 | struct extent_inode_elem **eie)
|
| 40 | {
|
| 41 | const u64 data_len = btrfs_file_extent_num_bytes(eb, fi);
|
| 42 | u64 offset = key->offset;
|
| 43 | struct extent_inode_elem *e;
|
| 44 | const u64 *root_ids;
|
| 45 | int root_count;
|
| 46 | bool cached;
|
| 47 |
|
| 48 | if (!ctx->ignore_extent_item_pos &&
|
| 49 | !btrfs_file_extent_compression(eb, fi) &&
|
| 50 | !btrfs_file_extent_encryption(eb, fi) &&
|
| 51 | !btrfs_file_extent_other_encoding(eb, fi)) {
|
| 52 | u64 data_offset;
|
| 53 |
|
| 54 | data_offset = btrfs_file_extent_offset(eb, fi);
|
| 55 |
|
| 56 | if (ctx->extent_item_pos < data_offset ||
|
| 57 | ctx->extent_item_pos >= data_offset + data_len)
|
| 58 | return 1;
|
| 59 | offset += ctx->extent_item_pos - data_offset;
|
| 60 | }
|
| 61 |
|
| 62 | if (!ctx->indirect_ref_iterator || !ctx->cache_lookup)
|
| 63 | goto add_inode_elem;
|
| 64 |
|
| 65 | cached = ctx->cache_lookup(eb->start, ctx->user_ctx, &root_ids,
|
| 66 | &root_count);
|
| 67 | if (!cached)
|
| 68 | goto add_inode_elem;
|
| 69 |
|
| 70 | for (int i = 0; i < root_count; i++) {
|
| 71 | int ret;
|
| 72 |
|
| 73 | ret = ctx->indirect_ref_iterator(key->objectid, offset,
|
| 74 | data_len, root_ids[i],
|
| 75 | ctx->user_ctx);
|
| 76 | if (ret)
|
| 77 | return ret;
|
| 78 | }
|
| 79 |
|
| 80 | add_inode_elem:
|
| 81 | e = kmalloc_obj(*e, GFP_NOFS);
|
| 82 | if (!e)
|
| 83 | return -ENOMEM;
|
| 84 |
|
| 85 | e->next = *eie;
|
| 86 | e->inum = key->objectid;
|
| 87 | e->offset = offset;
|
| 88 | e->num_bytes = data_len;
|
| 89 | *eie = e;
|
| 90 |
|
| 91 | return 0;
|
| 92 | }
|
| 93 |
|
| 94 | static void free_inode_elem_list(struct extent_inode_elem *eie)
|
| 95 | {
|
| 96 | struct extent_inode_elem *eie_next;
|
| 97 |
|
| 98 | for (; eie; eie = eie_next) {
|
| 99 | eie_next = eie->next;
|
| 100 | kfree(eie);
|
| 101 | }
|
| 102 | }
|
| 103 |
|
| 104 | static int find_extent_in_eb(struct btrfs_backref_walk_ctx *ctx,
|
| 105 | const struct extent_buffer *eb,
|
| 106 | struct extent_inode_elem **eie)
|
| 107 | {
|
| 108 | u64 disk_byte;
|
| 109 | struct btrfs_key key;
|
| 110 | struct btrfs_file_extent_item *fi;
|
| 111 | int slot;
|
| 112 | int nritems;
|
| 113 | int extent_type;
|
| 114 | int ret;
|
| 115 |
|
| 116 | |
| 117 | * from the shared data ref, we only have the leaf but we need
|
| 118 | * the key. thus, we must look into all items and see that we
|
| 119 | * find one (some) with a reference to our extent item.
|
| 120 |
|
| 121 | nritems = btrfs_header_nritems(eb);
|
| 122 | for (slot = 0; slot < nritems; ++slot) {
|
| 123 | btrfs_item_key_to_cpu(eb, &key, slot);
|
| 124 | if (key.type != BTRFS_EXTENT_DATA_KEY)
|
| 125 | continue;
|
| 126 | fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
|
| 127 | extent_type = btrfs_file_extent_type(eb, fi);
|
| 128 | if (extent_type == BTRFS_FILE_EXTENT_INLINE)
|
| 129 | continue;
|
| 130 |
|
| 131 | disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
|
| 132 | if (disk_byte != ctx->bytenr)
|
| 133 | continue;
|
| 134 |
|
| 135 | ret = check_extent_in_eb(ctx, &key, eb, fi, eie);
|
| 136 | if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP || ret < 0)
|
| 137 | return ret;
|
| 138 | }
|
| 139 |
|
| 140 | return 0;
|
| 141 | }
|
| 142 |
|
| 143 | struct preftree {
|
| 144 | struct rb_root_cached root;
|
| 145 | unsigned int count;
|
| 146 | };
|
| 147 |
|
| 148 | #define PREFTREE_INIT { .root = RB_ROOT_CACHED, .count = 0 }
|
| 149 |
|
| 150 | struct preftrees {
|
| 151 | struct preftree direct;
|
| 152 | struct preftree indirect;
|
| 153 | struct preftree indirect_missing_keys;
|
| 154 | };
|
| 155 |
|
| 156 | |
| 157 | * Checks for a shared extent during backref search.
|
| 158 | *
|
| 159 | * The share_count tracks prelim_refs (direct and indirect) having a
|
| 160 | * ref->count >0:
|
| 161 | * - incremented when a ref->count transitions to >0
|
| 162 | * - decremented when a ref->count transitions to <1
|
| 163 |
|
| 164 | struct share_check {
|
| 165 | struct btrfs_backref_share_check_ctx *ctx;
|
| 166 | struct btrfs_root *root;
|
| 167 | u64 inum;
|
| 168 | u64 data_bytenr;
|
| 169 | u64 data_extent_gen;
|
| 170 | |
| 171 | * Counts number of inodes that refer to an extent (different inodes in
|
| 172 | * the same root or different roots) that we could find. The sharedness
|
| 173 | * check typically stops once this counter gets greater than 1, so it
|
| 174 | * may not reflect the total number of inodes.
|
| 175 |
|
| 176 | int share_count;
|
| 177 | |
| 178 | * The number of times we found our inode refers to the data extent we
|
| 179 | * are determining the sharedness. In other words, how many file extent
|
| 180 | * items we could find for our inode that point to our target data
|
| 181 | * extent. The value we get here after finishing the extent sharedness
|
| 182 | * check may be smaller than reality, but if it ends up being greater
|
| 183 | * than 1, then we know for sure the inode has multiple file extent
|
| 184 | * items that point to our inode, and we can safely assume it's useful
|
| 185 | * to cache the sharedness check result.
|
| 186 |
|
| 187 | int self_ref_count;
|
| 188 | bool have_delayed_delete_refs;
|
| 189 | };
|
| 190 |
|
| 191 | static inline int extent_is_shared(struct share_check *sc)
|
| 192 | {
|
| 193 | return (sc && sc->share_count > 1) ? BACKREF_FOUND_SHARED : 0;
|
| 194 | }
|
| 195 |
|
| 196 | static struct kmem_cache *btrfs_prelim_ref_cache;
|
| 197 |
|
| 198 | int __init btrfs_prelim_ref_init(void)
|
| 199 | {
|
| 200 | btrfs_prelim_ref_cache = kmem_cache_create("btrfs_prelim_ref",
|
| 201 | sizeof(struct prelim_ref), 0, 0, NULL);
|
| 202 | if (!btrfs_prelim_ref_cache)
|
| 203 | return -ENOMEM;
|
| 204 | return 0;
|
| 205 | }
|
| 206 |
|
| 207 | void __cold btrfs_prelim_ref_exit(void)
|
| 208 | {
|
| 209 | kmem_cache_destroy(btrfs_prelim_ref_cache);
|
| 210 | }
|
| 211 |
|
| 212 | static void free_pref(struct prelim_ref *ref)
|
| 213 | {
|
| 214 | kmem_cache_free(btrfs_prelim_ref_cache, ref);
|
| 215 | }
|
| 216 |
|
| 217 | |
| 218 | * Return 0 when both refs are for the same block (and can be merged).
|
| 219 | * A -1 return indicates ref1 is a 'lower' block than ref2, while 1
|
| 220 | * indicates a 'higher' block.
|
| 221 |
|
| 222 | static int prelim_ref_compare(const struct prelim_ref *ref1,
|
| 223 | const struct prelim_ref *ref2)
|
| 224 | {
|
| 225 | if (ref1->level < ref2->level)
|
| 226 | return -1;
|
| 227 | if (ref1->level > ref2->level)
|
| 228 | return 1;
|
| 229 | if (ref1->root_id < ref2->root_id)
|
| 230 | return -1;
|
| 231 | if (ref1->root_id > ref2->root_id)
|
| 232 | return 1;
|
| 233 | if (ref1->key_for_search.type < ref2->key_for_search.type)
|
| 234 | return -1;
|
| 235 | if (ref1->key_for_search.type > ref2->key_for_search.type)
|
| 236 | return 1;
|
| 237 | if (ref1->key_for_search.objectid < ref2->key_for_search.objectid)
|
| 238 | return -1;
|
| 239 | if (ref1->key_for_search.objectid > ref2->key_for_search.objectid)
|
| 240 | return 1;
|
| 241 | if (ref1->key_for_search.offset < ref2->key_for_search.offset)
|
| 242 | return -1;
|
| 243 | if (ref1->key_for_search.offset > ref2->key_for_search.offset)
|
| 244 | return 1;
|
| 245 | if (ref1->parent < ref2->parent)
|
| 246 | return -1;
|
| 247 | if (ref1->parent > ref2->parent)
|
| 248 | return 1;
|
| 249 |
|
| 250 | return 0;
|
| 251 | }
|
| 252 |
|
| 253 | static int prelim_ref_rb_add_cmp(const struct rb_node *new,
|
| 254 | const struct rb_node *exist)
|
| 255 | {
|
| 256 | const struct prelim_ref *ref_new =
|
| 257 | rb_entry(new, struct prelim_ref, rbnode);
|
| 258 | const struct prelim_ref *ref_exist =
|
| 259 | rb_entry(exist, struct prelim_ref, rbnode);
|
| 260 |
|
| 261 | |
| 262 | * prelim_ref_compare() expects the first parameter as the existing one,
|
| 263 | * different from the rb_find_add_cached() order.
|
| 264 |
|
| 265 | return prelim_ref_compare(ref_exist, ref_new);
|
| 266 | }
|
| 267 |
|
| 268 | static void update_share_count(struct share_check *sc, int oldcount,
|
| 269 | int newcount, const struct prelim_ref *newref)
|
| 270 | {
|
| 271 | if ((!sc) || (oldcount == 0 && newcount < 1))
|
| 272 | return;
|
| 273 |
|
| 274 | if (oldcount > 0 && newcount < 1)
|
| 275 | sc->share_count--;
|
| 276 | else if (oldcount < 1 && newcount > 0)
|
| 277 | sc->share_count++;
|
| 278 |
|
| 279 | if (newref->root_id == btrfs_root_id(sc->root) &&
|
| 280 | newref->wanted_disk_byte == sc->data_bytenr &&
|
| 281 | newref->key_for_search.objectid == sc->inum)
|
| 282 | sc->self_ref_count += newref->count;
|
| 283 | }
|
| 284 |
|
| 285 | |
| 286 | * Add @newref to the @root rbtree, merging identical refs.
|
| 287 | *
|
| 288 | * Callers should assume that newref has been freed after calling.
|
| 289 |
|
| 290 | static void prelim_ref_insert(const struct btrfs_fs_info *fs_info,
|
| 291 | struct preftree *preftree,
|
| 292 | struct prelim_ref *newref,
|
| 293 | struct share_check *sc)
|
| 294 | {
|
| 295 | struct rb_root_cached *root;
|
| 296 | struct rb_node *exist;
|
| 297 |
|
| 298 | root = &preftree->root;
|
| 299 | exist = rb_find_add_cached(&newref->rbnode, root, prelim_ref_rb_add_cmp);
|
| 300 | if (exist) {
|
| 301 | struct prelim_ref *ref = rb_entry(exist, struct prelim_ref, rbnode);
|
| 302 |
|
| 303 | struct extent_inode_elem *eie = ref->inode_list;
|
| 304 |
|
| 305 | while (eie && eie->next)
|
| 306 | eie = eie->next;
|
| 307 |
|
| 308 | if (!eie)
|
| 309 | ref->inode_list = newref->inode_list;
|
| 310 | else
|
| 311 | eie->next = newref->inode_list;
|
| 312 | trace_btrfs_prelim_ref_merge(fs_info, ref, newref,
|
| 313 | preftree->count);
|
| 314 | |
| 315 | * A delayed ref can have newref->count < 0.
|
| 316 | * The ref->count is updated to follow any
|
| 317 | * BTRFS_[ADD|DROP]_DELAYED_REF actions.
|
| 318 |
|
| 319 | update_share_count(sc, ref->count,
|
| 320 | ref->count + newref->count, newref);
|
| 321 | ref->count += newref->count;
|
| 322 | free_pref(newref);
|
| 323 | return;
|
| 324 | }
|
| 325 |
|
| 326 | update_share_count(sc, 0, newref->count, newref);
|
| 327 | preftree->count++;
|
| 328 | trace_btrfs_prelim_ref_insert(fs_info, newref, NULL, preftree->count);
|
| 329 | }
|
| 330 |
|
| 331 | |
| 332 | * Release the entire tree. We don't care about internal consistency so
|
| 333 | * just free everything and then reset the tree root.
|
| 334 |
|
| 335 | static void prelim_release(struct preftree *preftree)
|
| 336 | {
|
| 337 | struct prelim_ref *ref, *next_ref;
|
| 338 |
|
| 339 | rbtree_postorder_for_each_entry_safe(ref, next_ref,
|
| 340 | &preftree->root.rb_root, rbnode) {
|
| 341 | free_inode_elem_list(ref->inode_list);
|
| 342 | free_pref(ref);
|
| 343 | }
|
| 344 |
|
| 345 | preftree->root = RB_ROOT_CACHED;
|
| 346 | preftree->count = 0;
|
| 347 | }
|
| 348 |
|
| 349 | |
| 350 | * the rules for all callers of this function are:
|
| 351 | * - obtaining the parent is the goal
|
| 352 | * - if you add a key, you must know that it is a correct key
|
| 353 | * - if you cannot add the parent or a correct key, then we will look into the
|
| 354 | * block later to set a correct key
|
| 355 | *
|
| 356 | * delayed refs
|
| 357 | * ============
|
| 358 | * backref type | shared | indirect | shared | indirect
|
| 359 | * information | tree | tree | data | data
|
| 360 | * --------------------+--------+----------+--------+----------
|
| 361 | * parent logical | y | - | - | -
|
| 362 | * key to resolve | - | y | y | y
|
| 363 | * tree block logical | - | - | - | -
|
| 364 | * root for resolving | y | y | y | y
|
| 365 | *
|
| 366 | * - column 1: we've the parent -> done
|
| 367 | * - column 2, 3, 4: we use the key to find the parent
|
| 368 | *
|
| 369 | * on disk refs (inline or keyed)
|
| 370 | * ==============================
|
| 371 | * backref type | shared | indirect | shared | indirect
|
| 372 | * information | tree | tree | data | data
|
| 373 | * --------------------+--------+----------+--------+----------
|
| 374 | * parent logical | y | - | y | -
|
| 375 | * key to resolve | - | - | - | y
|
| 376 | * tree block logical | y | y | y | y
|
| 377 | * root for resolving | - | y | y | y
|
| 378 | *
|
| 379 | * - column 1, 3: we've the parent -> done
|
| 380 | * - column 2: we take the first key from the block to find the parent
|
| 381 | * (see add_missing_keys)
|
| 382 | * - column 4: we use the key to find the parent
|
| 383 | *
|
| 384 | * additional information that's available but not required to find the parent
|
| 385 | * block might help in merging entries to gain some speed.
|
| 386 |
|
| 387 | static int add_prelim_ref(const struct btrfs_fs_info *fs_info,
|
| 388 | struct preftree *preftree, u64 root_id,
|
| 389 | const struct btrfs_key *key, int level, u64 parent,
|
| 390 | u64 wanted_disk_byte, int count,
|
| 391 | struct share_check *sc, gfp_t gfp_mask)
|
| 392 | {
|
| 393 | struct prelim_ref *ref;
|
| 394 |
|
| 395 | if (root_id == BTRFS_DATA_RELOC_TREE_OBJECTID)
|
| 396 | return 0;
|
| 397 |
|
| 398 | ref = kmem_cache_alloc(btrfs_prelim_ref_cache, gfp_mask);
|
| 399 | if (!ref)
|
| 400 | return -ENOMEM;
|
| 401 |
|
| 402 | ref->root_id = root_id;
|
| 403 | if (key)
|
| 404 | ref->key_for_search = *key;
|
| 405 | else
|
| 406 | memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
|
| 407 |
|
| 408 | ref->inode_list = NULL;
|
| 409 | ref->level = level;
|
| 410 | ref->count = count;
|
| 411 | ref->parent = parent;
|
| 412 | ref->wanted_disk_byte = wanted_disk_byte;
|
| 413 | prelim_ref_insert(fs_info, preftree, ref, sc);
|
| 414 | return extent_is_shared(sc);
|
| 415 | }
|
| 416 |
|
| 417 |
|
| 418 | static int add_direct_ref(const struct btrfs_fs_info *fs_info,
|
| 419 | struct preftrees *preftrees, int level, u64 parent,
|
| 420 | u64 wanted_disk_byte, int count,
|
| 421 | struct share_check *sc, gfp_t gfp_mask)
|
| 422 | {
|
| 423 | return add_prelim_ref(fs_info, &preftrees->direct, 0, NULL, level,
|
| 424 | parent, wanted_disk_byte, count, sc, gfp_mask);
|
| 425 | }
|
| 426 |
|
| 427 |
|
| 428 | static int add_indirect_ref(const struct btrfs_fs_info *fs_info,
|
| 429 | struct preftrees *preftrees, u64 root_id,
|
| 430 | const struct btrfs_key *key, int level,
|
| 431 | u64 wanted_disk_byte, int count,
|
| 432 | struct share_check *sc, gfp_t gfp_mask)
|
| 433 | {
|
| 434 | struct preftree *tree = &preftrees->indirect;
|
| 435 |
|
| 436 | if (!key)
|
| 437 | tree = &preftrees->indirect_missing_keys;
|
| 438 | return add_prelim_ref(fs_info, tree, root_id, key, level, 0,
|
| 439 | wanted_disk_byte, count, sc, gfp_mask);
|
| 440 | }
|
| 441 |
|
| 442 | static int is_shared_data_backref(struct preftrees *preftrees, u64 bytenr)
|
| 443 | {
|
| 444 | struct rb_node **p = &preftrees->direct.root.rb_root.rb_node;
|
| 445 | struct rb_node *parent = NULL;
|
| 446 | struct prelim_ref *ref = NULL;
|
| 447 | struct prelim_ref target = {};
|
| 448 | int result;
|
| 449 |
|
| 450 | target.parent = bytenr;
|
| 451 |
|
| 452 | while (*p) {
|
| 453 | parent = *p;
|
| 454 | ref = rb_entry(parent, struct prelim_ref, rbnode);
|
| 455 | result = prelim_ref_compare(ref, &target);
|
| 456 |
|
| 457 | if (result < 0)
|
| 458 | p = &(*p)->rb_left;
|
| 459 | else if (result > 0)
|
| 460 | p = &(*p)->rb_right;
|
| 461 | else
|
| 462 | return 1;
|
| 463 | }
|
| 464 | return 0;
|
| 465 | }
|
| 466 |
|
| 467 | static int add_all_parents(struct btrfs_backref_walk_ctx *ctx,
|
| 468 | struct btrfs_root *root, struct btrfs_path *path,
|
| 469 | struct ulist *parents,
|
| 470 | struct preftrees *preftrees, struct prelim_ref *ref,
|
| 471 | int level)
|
| 472 | {
|
| 473 | int ret = 0;
|
| 474 | int slot;
|
| 475 | struct extent_buffer *eb;
|
| 476 | struct btrfs_key key;
|
| 477 | struct btrfs_key *key_for_search = &ref->key_for_search;
|
| 478 | struct btrfs_file_extent_item *fi;
|
| 479 | struct extent_inode_elem *eie = NULL, *old = NULL;
|
| 480 | u64 disk_byte;
|
| 481 | u64 wanted_disk_byte = ref->wanted_disk_byte;
|
| 482 | u64 count = 0;
|
| 483 | u64 data_offset;
|
| 484 | u8 type;
|
| 485 |
|
| 486 | if (level != 0) {
|
| 487 | eb = path->nodes[level];
|
| 488 | ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
|
| 489 | if (ret < 0)
|
| 490 | return ret;
|
| 491 | return 0;
|
| 492 | }
|
| 493 |
|
| 494 | |
| 495 | * 1. We normally enter this function with the path already pointing to
|
| 496 | * the first item to check. But sometimes, we may enter it with
|
| 497 | * slot == nritems.
|
| 498 | * 2. We are searching for normal backref but bytenr of this leaf
|
| 499 | * matches shared data backref
|
| 500 | * 3. The leaf owner is not equal to the root we are searching
|
| 501 | *
|
| 502 | * For these cases, go to the next leaf before we continue.
|
| 503 |
|
| 504 | eb = path->nodes[0];
|
| 505 | if (path->slots[0] >= btrfs_header_nritems(eb) ||
|
| 506 | is_shared_data_backref(preftrees, eb->start) ||
|
| 507 | ref->root_id != btrfs_header_owner(eb)) {
|
| 508 | if (ctx->time_seq == BTRFS_SEQ_LAST)
|
| 509 | ret = btrfs_next_leaf(root, path);
|
| 510 | else
|
| 511 | ret = btrfs_next_old_leaf(root, path, ctx->time_seq);
|
| 512 | }
|
| 513 |
|
| 514 | while (!ret && count < ref->count) {
|
| 515 | eb = path->nodes[0];
|
| 516 | slot = path->slots[0];
|
| 517 |
|
| 518 | btrfs_item_key_to_cpu(eb, &key, slot);
|
| 519 |
|
| 520 | if (key.objectid != key_for_search->objectid ||
|
| 521 | key.type != BTRFS_EXTENT_DATA_KEY)
|
| 522 | break;
|
| 523 |
|
| 524 | |
| 525 | * We are searching for normal backref but bytenr of this leaf
|
| 526 | * matches shared data backref, OR
|
| 527 | * the leaf owner is not equal to the root we are searching for
|
| 528 |
|
| 529 | if (slot == 0 &&
|
| 530 | (is_shared_data_backref(preftrees, eb->start) ||
|
| 531 | ref->root_id != btrfs_header_owner(eb))) {
|
| 532 | if (ctx->time_seq == BTRFS_SEQ_LAST)
|
| 533 | ret = btrfs_next_leaf(root, path);
|
| 534 | else
|
| 535 | ret = btrfs_next_old_leaf(root, path, ctx->time_seq);
|
| 536 | continue;
|
| 537 | }
|
| 538 | fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
|
| 539 | type = btrfs_file_extent_type(eb, fi);
|
| 540 | if (type == BTRFS_FILE_EXTENT_INLINE)
|
| 541 | goto next;
|
| 542 | disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
|
| 543 | data_offset = btrfs_file_extent_offset(eb, fi);
|
| 544 |
|
| 545 | if (disk_byte == wanted_disk_byte) {
|
| 546 | eie = NULL;
|
| 547 | old = NULL;
|
| 548 | if (ref->key_for_search.offset == key.offset - data_offset)
|
| 549 | count++;
|
| 550 | else
|
| 551 | goto next;
|
| 552 | if (!ctx->skip_inode_ref_list) {
|
| 553 | ret = check_extent_in_eb(ctx, &key, eb, fi, &eie);
|
| 554 | if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP ||
|
| 555 | ret < 0)
|
| 556 | break;
|
| 557 | }
|
| 558 | if (ret > 0)
|
| 559 | goto next;
|
| 560 | ret = ulist_add_merge_ptr(parents, eb->start,
|
| 561 | eie, (void **)&old, GFP_NOFS);
|
| 562 | if (ret < 0)
|
| 563 | break;
|
| 564 | if (!ret && !ctx->skip_inode_ref_list) {
|
| 565 | while (old->next)
|
| 566 | old = old->next;
|
| 567 | old->next = eie;
|
| 568 | }
|
| 569 | eie = NULL;
|
| 570 | }
|
| 571 | next:
|
| 572 | if (ctx->time_seq == BTRFS_SEQ_LAST)
|
| 573 | ret = btrfs_next_item(root, path);
|
| 574 | else
|
| 575 | ret = btrfs_next_old_item(root, path, ctx->time_seq);
|
| 576 | }
|
| 577 |
|
| 578 | if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP || ret < 0)
|
| 579 | free_inode_elem_list(eie);
|
| 580 | else if (ret > 0)
|
| 581 | ret = 0;
|
| 582 |
|
| 583 | return ret;
|
| 584 | }
|
| 585 |
|
| 586 | |
| 587 | * resolve an indirect backref in the form (root_id, key, level)
|
| 588 | * to a logical address
|
| 589 |
|
| 590 | static int resolve_indirect_ref(struct btrfs_backref_walk_ctx *ctx,
|
| 591 | struct btrfs_path *path,
|
| 592 | struct preftrees *preftrees,
|
| 593 | struct prelim_ref *ref, struct ulist *parents)
|
| 594 | {
|
| 595 | struct btrfs_root *root;
|
| 596 | struct extent_buffer *eb;
|
| 597 | int ret = 0;
|
| 598 | int root_level;
|
| 599 | int level = ref->level;
|
| 600 | struct btrfs_key search_key = ref->key_for_search;
|
| 601 |
|
| 602 | |
| 603 | * If we're search_commit_root we could possibly be holding locks on
|
| 604 | * other tree nodes. This happens when qgroups does backref walks when
|
| 605 | * adding new delayed refs. To deal with this we need to look in cache
|
| 606 | * for the root, and if we don't find it then we need to search the
|
| 607 | * tree_root's commit root, thus the btrfs_get_fs_root_commit_root usage
|
| 608 | * here.
|
| 609 |
|
| 610 | if (path->search_commit_root)
|
| 611 | root = btrfs_get_fs_root_commit_root(ctx->fs_info, path, ref->root_id);
|
| 612 | else
|
| 613 | root = btrfs_get_fs_root(ctx->fs_info, ref->root_id, false);
|
| 614 | if (IS_ERR(root)) {
|
| 615 | ret = PTR_ERR(root);
|
| 616 | goto out_free;
|
| 617 | }
|
| 618 |
|
| 619 | if (!path->search_commit_root &&
|
| 620 | test_bit(BTRFS_ROOT_DELETING, &root->state)) {
|
| 621 | ret = -ENOENT;
|
| 622 | goto out;
|
| 623 | }
|
| 624 |
|
| 625 | if (btrfs_is_testing(ctx->fs_info)) {
|
| 626 | ret = -ENOENT;
|
| 627 | goto out;
|
| 628 | }
|
| 629 |
|
| 630 | if (path->search_commit_root)
|
| 631 | root_level = btrfs_header_level(root->commit_root);
|
| 632 | else if (ctx->time_seq == BTRFS_SEQ_LAST)
|
| 633 | root_level = btrfs_header_level(root->node);
|
| 634 | else
|
| 635 | root_level = btrfs_old_root_level(root, ctx->time_seq);
|
| 636 |
|
| 637 | if (root_level + 1 == level)
|
| 638 | goto out;
|
| 639 |
|
| 640 | |
| 641 | * We can often find data backrefs with an offset that is too large
|
| 642 | * (>= LLONG_MAX, maximum allowed file offset) due to underflows when
|
| 643 | * subtracting a file's offset with the data offset of its
|
| 644 | * corresponding extent data item. This can happen for example in the
|
| 645 | * clone ioctl.
|
| 646 | *
|
| 647 | * So if we detect such case we set the search key's offset to zero to
|
| 648 | * make sure we will find the matching file extent item at
|
| 649 | * add_all_parents(), otherwise we will miss it because the offset
|
| 650 | * taken form the backref is much larger then the offset of the file
|
| 651 | * extent item. This can make us scan a very large number of file
|
| 652 | * extent items, but at least it will not make us miss any.
|
| 653 | *
|
| 654 | * This is an ugly workaround for a behaviour that should have never
|
| 655 | * existed, but it does and a fix for the clone ioctl would touch a lot
|
| 656 | * of places, cause backwards incompatibility and would not fix the
|
| 657 | * problem for extents cloned with older kernels.
|
| 658 |
|
| 659 | if (search_key.type == BTRFS_EXTENT_DATA_KEY &&
|
| 660 | search_key.offset >= LLONG_MAX)
|
| 661 | search_key.offset = 0;
|
| 662 | path->lowest_level = level;
|
| 663 | if (ctx->time_seq == BTRFS_SEQ_LAST)
|
| 664 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0);
|
| 665 | else
|
| 666 | ret = btrfs_search_old_slot(root, &search_key, path, ctx->time_seq);
|
| 667 |
|
| 668 | btrfs_debug(ctx->fs_info,
|
| 669 | "search slot in root %llu (level %d, ref count %d) returned %d for key " BTRFS_KEY_FMT,
|
| 670 | ref->root_id, level, ref->count, ret,
|
| 671 | BTRFS_KEY_FMT_VALUE(&ref->key_for_search));
|
| 672 | if (ret < 0)
|
| 673 | goto out;
|
| 674 |
|
| 675 | eb = path->nodes[level];
|
| 676 | while (!eb) {
|
| 677 | if (WARN_ON(!level)) {
|
| 678 | ret = 1;
|
| 679 | goto out;
|
| 680 | }
|
| 681 | level--;
|
| 682 | eb = path->nodes[level];
|
| 683 | }
|
| 684 |
|
| 685 | ret = add_all_parents(ctx, root, path, parents, preftrees, ref, level);
|
| 686 | out:
|
| 687 | btrfs_put_root(root);
|
| 688 | out_free:
|
| 689 | path->lowest_level = 0;
|
| 690 | btrfs_release_path(path);
|
| 691 | return ret;
|
| 692 | }
|
| 693 |
|
| 694 | static struct extent_inode_elem *
|
| 695 | unode_aux_to_inode_list(struct ulist_node *node)
|
| 696 | {
|
| 697 | if (!node)
|
| 698 | return NULL;
|
| 699 | return (struct extent_inode_elem *)(uintptr_t)node->aux;
|
| 700 | }
|
| 701 |
|
| 702 | static void free_leaf_list(struct ulist *ulist)
|
| 703 | {
|
| 704 | struct ulist_node *node;
|
| 705 | struct ulist_iterator uiter;
|
| 706 |
|
| 707 | ULIST_ITER_INIT(&uiter);
|
| 708 | while ((node = ulist_next(ulist, &uiter)))
|
| 709 | free_inode_elem_list(unode_aux_to_inode_list(node));
|
| 710 |
|
| 711 | ulist_free(ulist);
|
| 712 | }
|
| 713 |
|
| 714 | |
| 715 | * We maintain three separate rbtrees: one for direct refs, one for
|
| 716 | * indirect refs which have a key, and one for indirect refs which do not
|
| 717 | * have a key. Each tree does merge on insertion.
|
| 718 | *
|
| 719 | * Once all of the references are located, we iterate over the tree of
|
| 720 | * indirect refs with missing keys. An appropriate key is located and
|
| 721 | * the ref is moved onto the tree for indirect refs. After all missing
|
| 722 | * keys are thus located, we iterate over the indirect ref tree, resolve
|
| 723 | * each reference, and then insert the resolved reference onto the
|
| 724 | * direct tree (merging there too).
|
| 725 | *
|
| 726 | * New backrefs (i.e., for parent nodes) are added to the appropriate
|
| 727 | * rbtree as they are encountered. The new backrefs are subsequently
|
| 728 | * resolved as above.
|
| 729 |
|
| 730 | static int resolve_indirect_refs(struct btrfs_backref_walk_ctx *ctx,
|
| 731 | struct btrfs_path *path,
|
| 732 | struct preftrees *preftrees,
|
| 733 | struct share_check *sc)
|
| 734 | {
|
| 735 | int ret = 0;
|
| 736 | struct ulist *parents;
|
| 737 | struct ulist_node *node;
|
| 738 | struct ulist_iterator uiter;
|
| 739 | struct rb_node *rnode;
|
| 740 |
|
| 741 | parents = ulist_alloc(GFP_NOFS);
|
| 742 | if (!parents)
|
| 743 | return -ENOMEM;
|
| 744 |
|
| 745 | |
| 746 | * We could trade memory usage for performance here by iterating
|
| 747 | * the tree, allocating new refs for each insertion, and then
|
| 748 | * freeing the entire indirect tree when we're done. In some test
|
| 749 | * cases, the tree can grow quite large (~200k objects).
|
| 750 |
|
| 751 | while ((rnode = rb_first_cached(&preftrees->indirect.root))) {
|
| 752 | struct prelim_ref *ref;
|
| 753 | int ret2;
|
| 754 |
|
| 755 | ref = rb_entry(rnode, struct prelim_ref, rbnode);
|
| 756 | if (WARN(ref->parent,
|
| 757 | "BUG: direct ref found in indirect tree")) {
|
| 758 | ret = -EINVAL;
|
| 759 | goto out;
|
| 760 | }
|
| 761 |
|
| 762 | rb_erase_cached(&ref->rbnode, &preftrees->indirect.root);
|
| 763 | preftrees->indirect.count--;
|
| 764 |
|
| 765 | if (ref->count == 0) {
|
| 766 | free_pref(ref);
|
| 767 | continue;
|
| 768 | }
|
| 769 |
|
| 770 | if (sc && ref->root_id != btrfs_root_id(sc->root)) {
|
| 771 | free_pref(ref);
|
| 772 | ret = BACKREF_FOUND_SHARED;
|
| 773 | goto out;
|
| 774 | }
|
| 775 | ret2 = resolve_indirect_ref(ctx, path, preftrees, ref, parents);
|
| 776 | |
| 777 | * we can only tolerate ENOENT,otherwise,we should catch error
|
| 778 | * and return directly.
|
| 779 |
|
| 780 | if (ret2 == -ENOENT) {
|
| 781 | prelim_ref_insert(ctx->fs_info, &preftrees->direct, ref,
|
| 782 | NULL);
|
| 783 | continue;
|
| 784 | } else if (ret2) {
|
| 785 | free_pref(ref);
|
| 786 | ret = ret2;
|
| 787 | goto out;
|
| 788 | }
|
| 789 |
|
| 790 |
|
| 791 | ULIST_ITER_INIT(&uiter);
|
| 792 | node = ulist_next(parents, &uiter);
|
| 793 | ref->parent = node ? node->val : 0;
|
| 794 | ref->inode_list = unode_aux_to_inode_list(node);
|
| 795 |
|
| 796 |
|
| 797 | while ((node = ulist_next(parents, &uiter))) {
|
| 798 | struct prelim_ref *new_ref;
|
| 799 |
|
| 800 | new_ref = kmem_cache_alloc(btrfs_prelim_ref_cache,
|
| 801 | GFP_NOFS);
|
| 802 | if (!new_ref) {
|
| 803 | free_pref(ref);
|
| 804 | ret = -ENOMEM;
|
| 805 | goto out;
|
| 806 | }
|
| 807 | memcpy(new_ref, ref, sizeof(*ref));
|
| 808 | new_ref->parent = node->val;
|
| 809 | new_ref->inode_list = unode_aux_to_inode_list(node);
|
| 810 | prelim_ref_insert(ctx->fs_info, &preftrees->direct,
|
| 811 | new_ref, NULL);
|
| 812 | }
|
| 813 |
|
| 814 | |
| 815 | * Now it's a direct ref, put it in the direct tree. We must
|
| 816 | * do this last because the ref could be merged/freed here.
|
| 817 |
|
| 818 | prelim_ref_insert(ctx->fs_info, &preftrees->direct, ref, NULL);
|
| 819 |
|
| 820 | ulist_reinit(parents);
|
| 821 | cond_resched();
|
| 822 | }
|
| 823 | out:
|
| 824 | |
| 825 | * We may have inode lists attached to refs in the parents ulist, so we
|
| 826 | * must free them before freeing the ulist and its refs.
|
| 827 |
|
| 828 | free_leaf_list(parents);
|
| 829 | return ret;
|
| 830 | }
|
| 831 |
|
| 832 | |
| 833 | * read tree blocks and add keys where required.
|
| 834 |
|
| 835 | static int add_missing_keys(struct btrfs_fs_info *fs_info,
|
| 836 | struct preftrees *preftrees, bool lock)
|
| 837 | {
|
| 838 | struct prelim_ref *ref;
|
| 839 | struct extent_buffer *eb;
|
| 840 | struct preftree *tree = &preftrees->indirect_missing_keys;
|
| 841 | struct rb_node *node;
|
| 842 |
|
| 843 | while ((node = rb_first_cached(&tree->root))) {
|
| 844 | struct btrfs_tree_parent_check check = { 0 };
|
| 845 |
|
| 846 | ref = rb_entry(node, struct prelim_ref, rbnode);
|
| 847 | rb_erase_cached(node, &tree->root);
|
| 848 |
|
| 849 | BUG_ON(ref->parent);
|
| 850 | BUG_ON(ref->key_for_search.type);
|
| 851 | BUG_ON(!ref->wanted_disk_byte);
|
| 852 |
|
| 853 | check.level = ref->level - 1;
|
| 854 | check.owner_root = ref->root_id;
|
| 855 |
|
| 856 | eb = read_tree_block(fs_info, ref->wanted_disk_byte, &check);
|
| 857 | if (IS_ERR(eb)) {
|
| 858 | free_pref(ref);
|
| 859 | return PTR_ERR(eb);
|
| 860 | }
|
| 861 |
|
| 862 | if (lock)
|
| 863 | btrfs_tree_read_lock(eb);
|
| 864 | if (btrfs_header_level(eb) == 0)
|
| 865 | btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
|
| 866 | else
|
| 867 | btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
|
| 868 | if (lock)
|
| 869 | btrfs_tree_read_unlock(eb);
|
| 870 | free_extent_buffer(eb);
|
| 871 | prelim_ref_insert(fs_info, &preftrees->indirect, ref, NULL);
|
| 872 | cond_resched();
|
| 873 | }
|
| 874 | return 0;
|
| 875 | }
|
| 876 |
|
| 877 | |
| 878 | * add all currently queued delayed refs from this head whose seq nr is
|
| 879 | * smaller or equal that seq to the list
|
| 880 |
|
| 881 | static int add_delayed_refs(const struct btrfs_fs_info *fs_info,
|
| 882 | struct btrfs_delayed_ref_head *head, u64 seq,
|
| 883 | struct preftrees *preftrees, struct share_check *sc)
|
| 884 | {
|
| 885 | struct btrfs_delayed_ref_node *node;
|
| 886 | struct btrfs_key key;
|
| 887 | struct rb_node *n;
|
| 888 | int count;
|
| 889 | int ret = 0;
|
| 890 |
|
| 891 | spin_lock(&head->lock);
|
| 892 | for (n = rb_first_cached(&head->ref_tree); n; n = rb_next(n)) {
|
| 893 | node = rb_entry(n, struct btrfs_delayed_ref_node,
|
| 894 | ref_node);
|
| 895 | if (node->seq > seq)
|
| 896 | continue;
|
| 897 |
|
| 898 | switch (node->action) {
|
| 899 | case BTRFS_ADD_DELAYED_EXTENT:
|
| 900 | case BTRFS_UPDATE_DELAYED_HEAD:
|
| 901 | WARN_ON(1);
|
| 902 | continue;
|
| 903 | case BTRFS_ADD_DELAYED_REF:
|
| 904 | count = node->ref_mod;
|
| 905 | break;
|
| 906 | case BTRFS_DROP_DELAYED_REF:
|
| 907 | count = node->ref_mod * -1;
|
| 908 | break;
|
| 909 | default:
|
| 910 | BUG();
|
| 911 | }
|
| 912 | switch (node->type) {
|
| 913 | case BTRFS_TREE_BLOCK_REF_KEY: {
|
| 914 |
|
| 915 | struct btrfs_key *key_ptr = NULL;
|
| 916 |
|
| 917 | int level = btrfs_delayed_ref_owner(node);
|
| 918 |
|
| 919 | if (head->extent_op && head->extent_op->update_key) {
|
| 920 | btrfs_disk_key_to_cpu(&key, &head->extent_op->key);
|
| 921 | key_ptr = &key;
|
| 922 | }
|
| 923 |
|
| 924 | ret = add_indirect_ref(fs_info, preftrees, node->ref_root,
|
| 925 | key_ptr, level + 1, node->bytenr,
|
| 926 | count, sc, GFP_ATOMIC);
|
| 927 | break;
|
| 928 | }
|
| 929 | case BTRFS_SHARED_BLOCK_REF_KEY: {
|
| 930 | |
| 931 | * SHARED DIRECT METADATA backref
|
| 932 | *
|
| 933 | * The owner of a tree block ref is the level.
|
| 934 |
|
| 935 | int level = btrfs_delayed_ref_owner(node);
|
| 936 |
|
| 937 | ret = add_direct_ref(fs_info, preftrees, level + 1,
|
| 938 | node->parent, node->bytenr, count,
|
| 939 | sc, GFP_ATOMIC);
|
| 940 | break;
|
| 941 | }
|
| 942 | case BTRFS_EXTENT_DATA_REF_KEY: {
|
| 943 |
|
| 944 | key.objectid = btrfs_delayed_ref_owner(node);
|
| 945 | key.type = BTRFS_EXTENT_DATA_KEY;
|
| 946 | key.offset = btrfs_delayed_ref_offset(node);
|
| 947 |
|
| 948 | |
| 949 | * If we have a share check context and a reference for
|
| 950 | * another inode, we can't exit immediately. This is
|
| 951 | * because even if this is a BTRFS_ADD_DELAYED_REF
|
| 952 | * reference we may find next a BTRFS_DROP_DELAYED_REF
|
| 953 | * which cancels out this ADD reference.
|
| 954 | *
|
| 955 | * If this is a DROP reference and there was no previous
|
| 956 | * ADD reference, then we need to signal that when we
|
| 957 | * process references from the extent tree (through
|
| 958 | * add_inline_refs() and add_keyed_refs()), we should
|
| 959 | * not exit early if we find a reference for another
|
| 960 | * inode, because one of the delayed DROP references
|
| 961 | * may cancel that reference in the extent tree.
|
| 962 |
|
| 963 | if (sc && count < 0)
|
| 964 | sc->have_delayed_delete_refs = true;
|
| 965 |
|
| 966 | ret = add_indirect_ref(fs_info, preftrees, node->ref_root,
|
| 967 | &key, 0, node->bytenr, count, sc,
|
| 968 | GFP_ATOMIC);
|
| 969 | break;
|
| 970 | }
|
| 971 | case BTRFS_SHARED_DATA_REF_KEY: {
|
| 972 |
|
| 973 | ret = add_direct_ref(fs_info, preftrees, 0, node->parent,
|
| 974 | node->bytenr, count, sc,
|
| 975 | GFP_ATOMIC);
|
| 976 | break;
|
| 977 | }
|
| 978 | default:
|
| 979 | WARN_ON(1);
|
| 980 | }
|
| 981 | |
| 982 | * We must ignore BACKREF_FOUND_SHARED until all delayed
|
| 983 | * refs have been checked.
|
| 984 |
|
| 985 | if (ret && (ret != BACKREF_FOUND_SHARED))
|
| 986 | break;
|
| 987 | }
|
| 988 | if (!ret)
|
| 989 | ret = extent_is_shared(sc);
|
| 990 |
|
| 991 | spin_unlock(&head->lock);
|
| 992 | return ret;
|
| 993 | }
|
| 994 |
|
| 995 | |
| 996 | * add all inline backrefs for bytenr to the list
|
| 997 | *
|
| 998 | * Returns 0 on success, <0 on error, or BACKREF_FOUND_SHARED.
|
| 999 |
|
| 1000 | static int add_inline_refs(struct btrfs_backref_walk_ctx *ctx,
|
| 1001 | struct btrfs_path *path,
|
| 1002 | int *info_level, struct preftrees *preftrees,
|
| 1003 | struct share_check *sc)
|
| 1004 | {
|
| 1005 | int ret = 0;
|
| 1006 | int slot;
|
| 1007 | struct extent_buffer *leaf;
|
| 1008 | struct btrfs_key key;
|
| 1009 | struct btrfs_key found_key;
|
| 1010 | unsigned long ptr;
|
| 1011 | unsigned long end;
|
| 1012 | struct btrfs_extent_item *ei;
|
| 1013 | u64 flags;
|
| 1014 | u64 item_size;
|
| 1015 |
|
| 1016 | |
| 1017 | * enumerate all inline refs
|
| 1018 |
|
| 1019 | leaf = path->nodes[0];
|
| 1020 | slot = path->slots[0];
|
| 1021 |
|
| 1022 | item_size = btrfs_item_size(leaf, slot);
|
| 1023 | ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
|
| 1024 |
|
| 1025 | if (ctx->check_extent_item) {
|
| 1026 | ret = ctx->check_extent_item(ctx->bytenr, ei, leaf, ctx->user_ctx);
|
| 1027 | if (ret)
|
| 1028 | return ret;
|
| 1029 | }
|
| 1030 |
|
| 1031 | flags = btrfs_extent_flags(leaf, ei);
|
| 1032 | btrfs_item_key_to_cpu(leaf, &found_key, slot);
|
| 1033 |
|
| 1034 | ptr = (unsigned long)(ei + 1);
|
| 1035 | end = (unsigned long)ei + item_size;
|
| 1036 |
|
| 1037 | if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
|
| 1038 | flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
|
| 1039 | struct btrfs_tree_block_info *info;
|
| 1040 |
|
| 1041 | info = (struct btrfs_tree_block_info *)ptr;
|
| 1042 | *info_level = btrfs_tree_block_level(leaf, info);
|
| 1043 | ptr += sizeof(struct btrfs_tree_block_info);
|
| 1044 | BUG_ON(ptr > end);
|
| 1045 | } else if (found_key.type == BTRFS_METADATA_ITEM_KEY) {
|
| 1046 | *info_level = found_key.offset;
|
| 1047 | } else {
|
| 1048 | BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
|
| 1049 | }
|
| 1050 |
|
| 1051 | while (ptr < end) {
|
| 1052 | struct btrfs_extent_inline_ref *iref;
|
| 1053 | u64 offset;
|
| 1054 | int type;
|
| 1055 |
|
| 1056 | iref = (struct btrfs_extent_inline_ref *)ptr;
|
| 1057 | type = btrfs_get_extent_inline_ref_type(leaf, iref,
|
| 1058 | BTRFS_REF_TYPE_ANY);
|
| 1059 | if (unlikely(type == BTRFS_REF_TYPE_INVALID))
|
| 1060 | return -EUCLEAN;
|
| 1061 |
|
| 1062 | offset = btrfs_extent_inline_ref_offset(leaf, iref);
|
| 1063 |
|
| 1064 | switch (type) {
|
| 1065 | case BTRFS_SHARED_BLOCK_REF_KEY:
|
| 1066 | ret = add_direct_ref(ctx->fs_info, preftrees,
|
| 1067 | *info_level + 1, offset,
|
| 1068 | ctx->bytenr, 1, NULL, GFP_NOFS);
|
| 1069 | break;
|
| 1070 | case BTRFS_SHARED_DATA_REF_KEY: {
|
| 1071 | struct btrfs_shared_data_ref *sdref;
|
| 1072 | int count;
|
| 1073 |
|
| 1074 | sdref = (struct btrfs_shared_data_ref *)(iref + 1);
|
| 1075 | count = btrfs_shared_data_ref_count(leaf, sdref);
|
| 1076 |
|
| 1077 | ret = add_direct_ref(ctx->fs_info, preftrees, 0, offset,
|
| 1078 | ctx->bytenr, count, sc, GFP_NOFS);
|
| 1079 | break;
|
| 1080 | }
|
| 1081 | case BTRFS_TREE_BLOCK_REF_KEY:
|
| 1082 | ret = add_indirect_ref(ctx->fs_info, preftrees, offset,
|
| 1083 | NULL, *info_level + 1,
|
| 1084 | ctx->bytenr, 1, NULL, GFP_NOFS);
|
| 1085 | break;
|
| 1086 | case BTRFS_EXTENT_DATA_REF_KEY: {
|
| 1087 | struct btrfs_extent_data_ref *dref;
|
| 1088 | int count;
|
| 1089 | u64 root;
|
| 1090 |
|
| 1091 | dref = (struct btrfs_extent_data_ref *)(&iref->offset);
|
| 1092 | count = btrfs_extent_data_ref_count(leaf, dref);
|
| 1093 | key.objectid = btrfs_extent_data_ref_objectid(leaf,
|
| 1094 | dref);
|
| 1095 | key.type = BTRFS_EXTENT_DATA_KEY;
|
| 1096 | key.offset = btrfs_extent_data_ref_offset(leaf, dref);
|
| 1097 |
|
| 1098 | if (sc && key.objectid != sc->inum &&
|
| 1099 | !sc->have_delayed_delete_refs) {
|
| 1100 | ret = BACKREF_FOUND_SHARED;
|
| 1101 | break;
|
| 1102 | }
|
| 1103 |
|
| 1104 | root = btrfs_extent_data_ref_root(leaf, dref);
|
| 1105 |
|
| 1106 | if (!ctx->skip_data_ref ||
|
| 1107 | !ctx->skip_data_ref(root, key.objectid, key.offset,
|
| 1108 | ctx->user_ctx))
|
| 1109 | ret = add_indirect_ref(ctx->fs_info, preftrees,
|
| 1110 | root, &key, 0, ctx->bytenr,
|
| 1111 | count, sc, GFP_NOFS);
|
| 1112 | break;
|
| 1113 | }
|
| 1114 | case BTRFS_EXTENT_OWNER_REF_KEY:
|
| 1115 | ASSERT(btrfs_fs_incompat(ctx->fs_info, SIMPLE_QUOTA));
|
| 1116 | break;
|
| 1117 | default:
|
| 1118 | WARN_ON(1);
|
| 1119 | }
|
| 1120 | if (ret)
|
| 1121 | return ret;
|
| 1122 | ptr += btrfs_extent_inline_ref_size(type);
|
| 1123 | }
|
| 1124 |
|
| 1125 | return 0;
|
| 1126 | }
|
| 1127 |
|
| 1128 | |
| 1129 | * add all non-inline backrefs for bytenr to the list
|
| 1130 | *
|
| 1131 | * Returns 0 on success, <0 on error, or BACKREF_FOUND_SHARED.
|
| 1132 |
|
| 1133 | static int add_keyed_refs(struct btrfs_backref_walk_ctx *ctx,
|
| 1134 | struct btrfs_root *extent_root,
|
| 1135 | struct btrfs_path *path,
|
| 1136 | int info_level, struct preftrees *preftrees,
|
| 1137 | struct share_check *sc)
|
| 1138 | {
|
| 1139 | struct btrfs_fs_info *fs_info = extent_root->fs_info;
|
| 1140 | int ret;
|
| 1141 | int slot;
|
| 1142 | struct extent_buffer *leaf;
|
| 1143 | struct btrfs_key key;
|
| 1144 |
|
| 1145 | while (1) {
|
| 1146 | ret = btrfs_next_item(extent_root, path);
|
| 1147 | if (ret < 0)
|
| 1148 | break;
|
| 1149 | if (ret) {
|
| 1150 | ret = 0;
|
| 1151 | break;
|
| 1152 | }
|
| 1153 |
|
| 1154 | slot = path->slots[0];
|
| 1155 | leaf = path->nodes[0];
|
| 1156 | btrfs_item_key_to_cpu(leaf, &key, slot);
|
| 1157 |
|
| 1158 | if (key.objectid != ctx->bytenr)
|
| 1159 | break;
|
| 1160 | if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
|
| 1161 | continue;
|
| 1162 | if (key.type > BTRFS_SHARED_DATA_REF_KEY)
|
| 1163 | break;
|
| 1164 |
|
| 1165 | switch (key.type) {
|
| 1166 | case BTRFS_SHARED_BLOCK_REF_KEY:
|
| 1167 |
|
| 1168 | ret = add_direct_ref(fs_info, preftrees,
|
| 1169 | info_level + 1, key.offset,
|
| 1170 | ctx->bytenr, 1, NULL, GFP_NOFS);
|
| 1171 | break;
|
| 1172 | case BTRFS_SHARED_DATA_REF_KEY: {
|
| 1173 |
|
| 1174 | struct btrfs_shared_data_ref *sdref;
|
| 1175 | int count;
|
| 1176 |
|
| 1177 | sdref = btrfs_item_ptr(leaf, slot,
|
| 1178 | struct btrfs_shared_data_ref);
|
| 1179 | count = btrfs_shared_data_ref_count(leaf, sdref);
|
| 1180 | ret = add_direct_ref(fs_info, preftrees, 0,
|
| 1181 | key.offset, ctx->bytenr, count,
|
| 1182 | sc, GFP_NOFS);
|
| 1183 | break;
|
| 1184 | }
|
| 1185 | case BTRFS_TREE_BLOCK_REF_KEY:
|
| 1186 |
|
| 1187 | ret = add_indirect_ref(fs_info, preftrees, key.offset,
|
| 1188 | NULL, info_level + 1, ctx->bytenr,
|
| 1189 | 1, NULL, GFP_NOFS);
|
| 1190 | break;
|
| 1191 | case BTRFS_EXTENT_DATA_REF_KEY: {
|
| 1192 |
|
| 1193 | struct btrfs_extent_data_ref *dref;
|
| 1194 | int count;
|
| 1195 | u64 root;
|
| 1196 |
|
| 1197 | dref = btrfs_item_ptr(leaf, slot,
|
| 1198 | struct btrfs_extent_data_ref);
|
| 1199 | count = btrfs_extent_data_ref_count(leaf, dref);
|
| 1200 | key.objectid = btrfs_extent_data_ref_objectid(leaf,
|
| 1201 | dref);
|
| 1202 | key.type = BTRFS_EXTENT_DATA_KEY;
|
| 1203 | key.offset = btrfs_extent_data_ref_offset(leaf, dref);
|
| 1204 |
|
| 1205 | if (sc && key.objectid != sc->inum &&
|
| 1206 | !sc->have_delayed_delete_refs) {
|
| 1207 | ret = BACKREF_FOUND_SHARED;
|
| 1208 | break;
|
| 1209 | }
|
| 1210 |
|
| 1211 | root = btrfs_extent_data_ref_root(leaf, dref);
|
| 1212 |
|
| 1213 | if (!ctx->skip_data_ref ||
|
| 1214 | !ctx->skip_data_ref(root, key.objectid, key.offset,
|
| 1215 | ctx->user_ctx))
|
| 1216 | ret = add_indirect_ref(fs_info, preftrees, root,
|
| 1217 | &key, 0, ctx->bytenr,
|
| 1218 | count, sc, GFP_NOFS);
|
| 1219 | break;
|
| 1220 | }
|
| 1221 | default:
|
| 1222 | WARN_ON(1);
|
| 1223 | }
|
| 1224 | if (ret)
|
| 1225 | return ret;
|
| 1226 |
|
| 1227 | }
|
| 1228 |
|
| 1229 | return ret;
|
| 1230 | }
|
| 1231 |
|
| 1232 | |
| 1233 | * The caller has joined a transaction or is holding a read lock on the
|
| 1234 | * fs_info->commit_root_sem semaphore, so no need to worry about the root's last
|
| 1235 | * snapshot field changing while updating or checking the cache.
|
| 1236 |
|
| 1237 | static bool lookup_backref_shared_cache(struct btrfs_backref_share_check_ctx *ctx,
|
| 1238 | struct btrfs_root *root,
|
| 1239 | u64 bytenr, int level, bool *is_shared)
|
| 1240 | {
|
| 1241 | const struct btrfs_fs_info *fs_info = root->fs_info;
|
| 1242 | struct btrfs_backref_shared_cache_entry *entry;
|
| 1243 |
|
| 1244 | if (!current->journal_info)
|
| 1245 | lockdep_assert_held(&fs_info->commit_root_sem);
|
| 1246 |
|
| 1247 | if (!ctx->use_path_cache)
|
| 1248 | return false;
|
| 1249 |
|
| 1250 | if (WARN_ON_ONCE(level >= BTRFS_MAX_LEVEL))
|
| 1251 | return false;
|
| 1252 |
|
| 1253 | |
| 1254 | * Level -1 is used for the data extent, which is not reliable to cache
|
| 1255 | * because its reference count can increase or decrease without us
|
| 1256 | * realizing. We cache results only for extent buffers that lead from
|
| 1257 | * the root node down to the leaf with the file extent item.
|
| 1258 |
|
| 1259 | ASSERT(level >= 0, "level=%d", level);
|
| 1260 |
|
| 1261 | entry = &ctx->path_cache_entries[level];
|
| 1262 |
|
| 1263 |
|
| 1264 | if (entry->bytenr != bytenr)
|
| 1265 | return false;
|
| 1266 |
|
| 1267 | |
| 1268 | * We cached a false result, but the last snapshot generation of the
|
| 1269 | * root changed, so we now have a snapshot. Don't trust the result.
|
| 1270 |
|
| 1271 | if (!entry->is_shared &&
|
| 1272 | entry->gen != btrfs_root_last_snapshot(&root->root_item))
|
| 1273 | return false;
|
| 1274 |
|
| 1275 | |
| 1276 | * If we cached a true result and the last generation used for dropping
|
| 1277 | * a root changed, we can not trust the result, because the dropped root
|
| 1278 | * could be a snapshot sharing this extent buffer.
|
| 1279 |
|
| 1280 | if (entry->is_shared &&
|
| 1281 | entry->gen != btrfs_get_last_root_drop_gen(fs_info))
|
| 1282 | return false;
|
| 1283 |
|
| 1284 | *is_shared = entry->is_shared;
|
| 1285 | |
| 1286 | * If the node at this level is shared, than all nodes below are also
|
| 1287 | * shared. Currently some of the nodes below may be marked as not shared
|
| 1288 | * because we have just switched from one leaf to another, and switched
|
| 1289 | * also other nodes above the leaf and below the current level, so mark
|
| 1290 | * them as shared.
|
| 1291 |
|
| 1292 | if (*is_shared) {
|
| 1293 | for (int i = 0; i < level; i++) {
|
| 1294 | ctx->path_cache_entries[i].is_shared = true;
|
| 1295 | ctx->path_cache_entries[i].gen = entry->gen;
|
| 1296 | }
|
| 1297 | }
|
| 1298 |
|
| 1299 | return true;
|
| 1300 | }
|
| 1301 |
|
| 1302 | |
| 1303 | * The caller has joined a transaction or is holding a read lock on the
|
| 1304 | * fs_info->commit_root_sem semaphore, so no need to worry about the root's last
|
| 1305 | * snapshot field changing while updating or checking the cache.
|
| 1306 |
|
| 1307 | static void store_backref_shared_cache(struct btrfs_backref_share_check_ctx *ctx,
|
| 1308 | struct btrfs_root *root,
|
| 1309 | u64 bytenr, int level, bool is_shared)
|
| 1310 | {
|
| 1311 | const struct btrfs_fs_info *fs_info = root->fs_info;
|
| 1312 | struct btrfs_backref_shared_cache_entry *entry;
|
| 1313 | u64 gen;
|
| 1314 |
|
| 1315 | if (!current->journal_info)
|
| 1316 | lockdep_assert_held(&fs_info->commit_root_sem);
|
| 1317 |
|
| 1318 | if (!ctx->use_path_cache)
|
| 1319 | return;
|
| 1320 |
|
| 1321 | if (WARN_ON_ONCE(level >= BTRFS_MAX_LEVEL))
|
| 1322 | return;
|
| 1323 |
|
| 1324 | |
| 1325 | * Level -1 is used for the data extent, which is not reliable to cache
|
| 1326 | * because its reference count can increase or decrease without us
|
| 1327 | * realizing. We cache results only for extent buffers that lead from
|
| 1328 | * the root node down to the leaf with the file extent item.
|
| 1329 |
|
| 1330 | ASSERT(level >= 0, "level=%d", level);
|
| 1331 |
|
| 1332 | if (is_shared)
|
| 1333 | gen = btrfs_get_last_root_drop_gen(fs_info);
|
| 1334 | else
|
| 1335 | gen = btrfs_root_last_snapshot(&root->root_item);
|
| 1336 |
|
| 1337 | entry = &ctx->path_cache_entries[level];
|
| 1338 | entry->bytenr = bytenr;
|
| 1339 | entry->is_shared = is_shared;
|
| 1340 | entry->gen = gen;
|
| 1341 |
|
| 1342 | |
| 1343 | * If we found an extent buffer is shared, set the cache result for all
|
| 1344 | * extent buffers below it to true. As nodes in the path are COWed,
|
| 1345 | * their sharedness is moved to their children, and if a leaf is COWed,
|
| 1346 | * then the sharedness of a data extent becomes direct, the refcount of
|
| 1347 | * data extent is increased in the extent item at the extent tree.
|
| 1348 |
|
| 1349 | if (is_shared) {
|
| 1350 | for (int i = 0; i < level; i++) {
|
| 1351 | entry = &ctx->path_cache_entries[i];
|
| 1352 | entry->is_shared = is_shared;
|
| 1353 | entry->gen = gen;
|
| 1354 | }
|
| 1355 | }
|
| 1356 | }
|
| 1357 |
|
| 1358 | |
| 1359 | * this adds all existing backrefs (inline backrefs, backrefs and delayed
|
| 1360 | * refs) for the given bytenr to the refs list, merges duplicates and resolves
|
| 1361 | * indirect refs to their parent bytenr.
|
| 1362 | * When roots are found, they're added to the roots list
|
| 1363 | *
|
| 1364 | * @ctx: Backref walking context object, must be not NULL.
|
| 1365 | * @sc: If !NULL, then immediately return BACKREF_FOUND_SHARED when a
|
| 1366 | * shared extent is detected.
|
| 1367 | *
|
| 1368 | * Otherwise this returns 0 for success and <0 for an error.
|
| 1369 | *
|
| 1370 | * FIXME some caching might speed things up
|
| 1371 |
|
| 1372 | static int find_parent_nodes(struct btrfs_backref_walk_ctx *ctx,
|
| 1373 | struct share_check *sc)
|
| 1374 | {
|
| 1375 | struct btrfs_root *root = btrfs_extent_root(ctx->fs_info, ctx->bytenr);
|
| 1376 | struct btrfs_key key;
|
| 1377 | struct btrfs_path *path;
|
| 1378 | struct btrfs_delayed_ref_root *delayed_refs = NULL;
|
| 1379 | struct btrfs_delayed_ref_head *head;
|
| 1380 | int info_level = 0;
|
| 1381 | int ret;
|
| 1382 | struct prelim_ref *ref;
|
| 1383 | struct rb_node *node;
|
| 1384 | struct extent_inode_elem *eie = NULL;
|
| 1385 | struct preftrees preftrees = {
|
| 1386 | .direct = PREFTREE_INIT,
|
| 1387 | .indirect = PREFTREE_INIT,
|
| 1388 | .indirect_missing_keys = PREFTREE_INIT
|
| 1389 | };
|
| 1390 |
|
| 1391 | if (unlikely(!root)) {
|
| 1392 | btrfs_err(ctx->fs_info,
|
| 1393 | "missing extent root for extent at bytenr %llu",
|
| 1394 | ctx->bytenr);
|
| 1395 | return -EUCLEAN;
|
| 1396 | }
|
| 1397 |
|
| 1398 |
|
| 1399 | if (sc)
|
| 1400 | ASSERT(ctx->roots == NULL);
|
| 1401 |
|
| 1402 | key.objectid = ctx->bytenr;
|
| 1403 | if (btrfs_fs_incompat(ctx->fs_info, SKINNY_METADATA))
|
| 1404 | key.type = BTRFS_METADATA_ITEM_KEY;
|
| 1405 | else
|
| 1406 | key.type = BTRFS_EXTENT_ITEM_KEY;
|
| 1407 | key.offset = (u64)-1;
|
| 1408 |
|
| 1409 | path = btrfs_alloc_path();
|
| 1410 | if (!path)
|
| 1411 | return -ENOMEM;
|
| 1412 | if (!ctx->trans) {
|
| 1413 | path->search_commit_root = true;
|
| 1414 | path->skip_locking = true;
|
| 1415 | }
|
| 1416 |
|
| 1417 | if (ctx->time_seq == BTRFS_SEQ_LAST)
|
| 1418 | path->skip_locking = true;
|
| 1419 |
|
| 1420 | again:
|
| 1421 | head = NULL;
|
| 1422 |
|
| 1423 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
|
| 1424 | if (ret < 0)
|
| 1425 | goto out;
|
| 1426 | if (unlikely(ret == 0)) {
|
| 1427 | |
| 1428 | * Key with offset -1 found, there would have to exist an extent
|
| 1429 | * item with such offset, but this is out of the valid range.
|
| 1430 |
|
| 1431 | ret = -EUCLEAN;
|
| 1432 | goto out;
|
| 1433 | }
|
| 1434 |
|
| 1435 | if (ctx->trans && likely(!btrfs_is_dummy_transaction(ctx->trans)) &&
|
| 1436 | ctx->time_seq != BTRFS_SEQ_LAST) {
|
| 1437 | |
| 1438 | * We have a specific time_seq we care about and trans which
|
| 1439 | * means we have the path lock, we need to grab the ref head and
|
| 1440 | * lock it so we have a consistent view of the refs at the given
|
| 1441 | * time.
|
| 1442 |
|
| 1443 | delayed_refs = &ctx->trans->transaction->delayed_refs;
|
| 1444 | spin_lock(&delayed_refs->lock);
|
| 1445 | head = btrfs_find_delayed_ref_head(ctx->fs_info, delayed_refs,
|
| 1446 | ctx->bytenr);
|
| 1447 | if (head) {
|
| 1448 | if (!mutex_trylock(&head->mutex)) {
|
| 1449 | refcount_inc(&head->refs);
|
| 1450 | spin_unlock(&delayed_refs->lock);
|
| 1451 |
|
| 1452 | btrfs_release_path(path);
|
| 1453 |
|
| 1454 | |
| 1455 | * Mutex was contended, block until it's
|
| 1456 | * released and try again
|
| 1457 |
|
| 1458 | mutex_lock(&head->mutex);
|
| 1459 | mutex_unlock(&head->mutex);
|
| 1460 | btrfs_put_delayed_ref_head(head);
|
| 1461 | goto again;
|
| 1462 | }
|
| 1463 | spin_unlock(&delayed_refs->lock);
|
| 1464 | ret = add_delayed_refs(ctx->fs_info, head, ctx->time_seq,
|
| 1465 | &preftrees, sc);
|
| 1466 | mutex_unlock(&head->mutex);
|
| 1467 | if (ret)
|
| 1468 | goto out;
|
| 1469 | } else {
|
| 1470 | spin_unlock(&delayed_refs->lock);
|
| 1471 | }
|
| 1472 | }
|
| 1473 |
|
| 1474 | if (path->slots[0]) {
|
| 1475 | struct extent_buffer *leaf;
|
| 1476 | int slot;
|
| 1477 |
|
| 1478 | path->slots[0]--;
|
| 1479 | leaf = path->nodes[0];
|
| 1480 | slot = path->slots[0];
|
| 1481 | btrfs_item_key_to_cpu(leaf, &key, slot);
|
| 1482 | if (key.objectid == ctx->bytenr &&
|
| 1483 | (key.type == BTRFS_EXTENT_ITEM_KEY ||
|
| 1484 | key.type == BTRFS_METADATA_ITEM_KEY)) {
|
| 1485 | ret = add_inline_refs(ctx, path, &info_level,
|
| 1486 | &preftrees, sc);
|
| 1487 | if (ret)
|
| 1488 | goto out;
|
| 1489 | ret = add_keyed_refs(ctx, root, path, info_level,
|
| 1490 | &preftrees, sc);
|
| 1491 | if (ret)
|
| 1492 | goto out;
|
| 1493 | }
|
| 1494 | }
|
| 1495 |
|
| 1496 | |
| 1497 | * If we have a share context and we reached here, it means the extent
|
| 1498 | * is not directly shared (no multiple reference items for it),
|
| 1499 | * otherwise we would have exited earlier with a return value of
|
| 1500 | * BACKREF_FOUND_SHARED after processing delayed references or while
|
| 1501 | * processing inline or keyed references from the extent tree.
|
| 1502 | * The extent may however be indirectly shared through shared subtrees
|
| 1503 | * as a result from creating snapshots, so we determine below what is
|
| 1504 | * its parent node, in case we are dealing with a metadata extent, or
|
| 1505 | * what's the leaf (or leaves), from a fs tree, that has a file extent
|
| 1506 | * item pointing to it in case we are dealing with a data extent.
|
| 1507 |
|
| 1508 | ASSERT(extent_is_shared(sc) == 0);
|
| 1509 |
|
| 1510 | |
| 1511 | * If we are here for a data extent and we have a share_check structure
|
| 1512 | * it means the data extent is not directly shared (does not have
|
| 1513 | * multiple reference items), so we have to check if a path in the fs
|
| 1514 | * tree (going from the root node down to the leaf that has the file
|
| 1515 | * extent item pointing to the data extent) is shared, that is, if any
|
| 1516 | * of the extent buffers in the path is referenced by other trees.
|
| 1517 |
|
| 1518 | if (sc && ctx->bytenr == sc->data_bytenr) {
|
| 1519 | |
| 1520 | * If our data extent is from a generation more recent than the
|
| 1521 | * last generation used to snapshot the root, then we know that
|
| 1522 | * it can not be shared through subtrees, so we can skip
|
| 1523 | * resolving indirect references, there's no point in
|
| 1524 | * determining the extent buffers for the path from the fs tree
|
| 1525 | * root node down to the leaf that has the file extent item that
|
| 1526 | * points to the data extent.
|
| 1527 |
|
| 1528 | if (sc->data_extent_gen >
|
| 1529 | btrfs_root_last_snapshot(&sc->root->root_item)) {
|
| 1530 | ret = BACKREF_FOUND_NOT_SHARED;
|
| 1531 | goto out;
|
| 1532 | }
|
| 1533 |
|
| 1534 | |
| 1535 | * If we are only determining if a data extent is shared or not
|
| 1536 | * and the corresponding file extent item is located in the same
|
| 1537 | * leaf as the previous file extent item, we can skip resolving
|
| 1538 | * indirect references for a data extent, since the fs tree path
|
| 1539 | * is the same (same leaf, so same path). We skip as long as the
|
| 1540 | * cached result for the leaf is valid and only if there's only
|
| 1541 | * one file extent item pointing to the data extent, because in
|
| 1542 | * the case of multiple file extent items, they may be located
|
| 1543 | * in different leaves and therefore we have multiple paths.
|
| 1544 |
|
| 1545 | if (sc->ctx->curr_leaf_bytenr == sc->ctx->prev_leaf_bytenr &&
|
| 1546 | sc->self_ref_count == 1) {
|
| 1547 | bool cached;
|
| 1548 | bool is_shared;
|
| 1549 |
|
| 1550 | cached = lookup_backref_shared_cache(sc->ctx, sc->root,
|
| 1551 | sc->ctx->curr_leaf_bytenr,
|
| 1552 | 0, &is_shared);
|
| 1553 | if (cached) {
|
| 1554 | if (is_shared)
|
| 1555 | ret = BACKREF_FOUND_SHARED;
|
| 1556 | else
|
| 1557 | ret = BACKREF_FOUND_NOT_SHARED;
|
| 1558 | goto out;
|
| 1559 | }
|
| 1560 | }
|
| 1561 | }
|
| 1562 |
|
| 1563 | btrfs_release_path(path);
|
| 1564 |
|
| 1565 | ret = add_missing_keys(ctx->fs_info, &preftrees, !path->skip_locking);
|
| 1566 | if (ret)
|
| 1567 | goto out;
|
| 1568 |
|
| 1569 | WARN_ON(!RB_EMPTY_ROOT(&preftrees.indirect_missing_keys.root.rb_root));
|
| 1570 |
|
| 1571 | ret = resolve_indirect_refs(ctx, path, &preftrees, sc);
|
| 1572 | if (ret)
|
| 1573 | goto out;
|
| 1574 |
|
| 1575 | WARN_ON(!RB_EMPTY_ROOT(&preftrees.indirect.root.rb_root));
|
| 1576 |
|
| 1577 | |
| 1578 | * This walks the tree of merged and resolved refs. Tree blocks are
|
| 1579 | * read in as needed. Unique entries are added to the ulist, and
|
| 1580 | * the list of found roots is updated.
|
| 1581 | *
|
| 1582 | * We release the entire tree in one go before returning.
|
| 1583 |
|
| 1584 | node = rb_first_cached(&preftrees.direct.root);
|
| 1585 | while (node) {
|
| 1586 | ref = rb_entry(node, struct prelim_ref, rbnode);
|
| 1587 | node = rb_next(&ref->rbnode);
|
| 1588 | |
| 1589 | * ref->count < 0 can happen here if there are delayed
|
| 1590 | * refs with a node->action of BTRFS_DROP_DELAYED_REF.
|
| 1591 | * prelim_ref_insert() relies on this when merging
|
| 1592 | * identical refs to keep the overall count correct.
|
| 1593 | * prelim_ref_insert() will merge only those refs
|
| 1594 | * which compare identically. Any refs having
|
| 1595 | * e.g. different offsets would not be merged,
|
| 1596 | * and would retain their original ref->count < 0.
|
| 1597 |
|
| 1598 | if (ctx->roots && ref->count && ref->root_id && ref->parent == 0) {
|
| 1599 |
|
| 1600 | ret = ulist_add(ctx->roots, ref->root_id, 0, GFP_NOFS);
|
| 1601 | if (ret < 0)
|
| 1602 | goto out;
|
| 1603 | }
|
| 1604 | if (ref->count && ref->parent) {
|
| 1605 | if (!ctx->skip_inode_ref_list && !ref->inode_list &&
|
| 1606 | ref->level == 0) {
|
| 1607 | struct btrfs_tree_parent_check check = { 0 };
|
| 1608 | struct extent_buffer *eb;
|
| 1609 |
|
| 1610 | check.level = ref->level;
|
| 1611 |
|
| 1612 | eb = read_tree_block(ctx->fs_info, ref->parent,
|
| 1613 | &check);
|
| 1614 | if (IS_ERR(eb)) {
|
| 1615 | ret = PTR_ERR(eb);
|
| 1616 | goto out;
|
| 1617 | }
|
| 1618 |
|
| 1619 | if (!path->skip_locking)
|
| 1620 | btrfs_tree_read_lock(eb);
|
| 1621 | ret = find_extent_in_eb(ctx, eb, &eie);
|
| 1622 | if (!path->skip_locking)
|
| 1623 | btrfs_tree_read_unlock(eb);
|
| 1624 | free_extent_buffer(eb);
|
| 1625 | if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP ||
|
| 1626 | ret < 0)
|
| 1627 | goto out;
|
| 1628 | ref->inode_list = eie;
|
| 1629 | |
| 1630 | * We transferred the list ownership to the ref,
|
| 1631 | * so set to NULL to avoid a double free in case
|
| 1632 | * an error happens after this.
|
| 1633 |
|
| 1634 | eie = NULL;
|
| 1635 | }
|
| 1636 | ret = ulist_add_merge_ptr(ctx->refs, ref->parent,
|
| 1637 | ref->inode_list,
|
| 1638 | (void **)&eie, GFP_NOFS);
|
| 1639 | if (ret < 0)
|
| 1640 | goto out;
|
| 1641 | if (!ret && !ctx->skip_inode_ref_list) {
|
| 1642 | |
| 1643 | * We've recorded that parent, so we must extend
|
| 1644 | * its inode list here.
|
| 1645 | *
|
| 1646 | * However if there was corruption we may not
|
| 1647 | * have found an eie, return an error in this
|
| 1648 | * case.
|
| 1649 |
|
| 1650 | ASSERT(eie);
|
| 1651 | if (unlikely(!eie)) {
|
| 1652 | ret = -EUCLEAN;
|
| 1653 | goto out;
|
| 1654 | }
|
| 1655 | while (eie->next)
|
| 1656 | eie = eie->next;
|
| 1657 | eie->next = ref->inode_list;
|
| 1658 | }
|
| 1659 | eie = NULL;
|
| 1660 | |
| 1661 | * We have transferred the inode list ownership from
|
| 1662 | * this ref to the ref we added to the 'refs' ulist.
|
| 1663 | * So set this ref's inode list to NULL to avoid
|
| 1664 | * use-after-free when our caller uses it or double
|
| 1665 | * frees in case an error happens before we return.
|
| 1666 |
|
| 1667 | ref->inode_list = NULL;
|
| 1668 | }
|
| 1669 | cond_resched();
|
| 1670 | }
|
| 1671 |
|
| 1672 | out:
|
| 1673 | btrfs_free_path(path);
|
| 1674 |
|
| 1675 | prelim_release(&preftrees.direct);
|
| 1676 | prelim_release(&preftrees.indirect);
|
| 1677 | prelim_release(&preftrees.indirect_missing_keys);
|
| 1678 |
|
| 1679 | if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP || ret < 0)
|
| 1680 | free_inode_elem_list(eie);
|
| 1681 | return ret;
|
| 1682 | }
|
| 1683 |
|
| 1684 | |
| 1685 | * Finds all leaves with a reference to the specified combination of
|
| 1686 | * @ctx->bytenr and @ctx->extent_item_pos. The bytenr of the found leaves are
|
| 1687 | * added to the ulist at @ctx->refs, and that ulist is allocated by this
|
| 1688 | * function. The caller should free the ulist with free_leaf_list() if
|
| 1689 | * @ctx->ignore_extent_item_pos is false, otherwise a simple ulist_free() is
|
| 1690 | * enough.
|
| 1691 | *
|
| 1692 | * Returns 0 on success and < 0 on error. On error @ctx->refs is not allocated.
|
| 1693 |
|
| 1694 | int btrfs_find_all_leafs(struct btrfs_backref_walk_ctx *ctx)
|
| 1695 | {
|
| 1696 | int ret;
|
| 1697 |
|
| 1698 | ASSERT(ctx->refs == NULL);
|
| 1699 |
|
| 1700 | ctx->refs = ulist_alloc(GFP_NOFS);
|
| 1701 | if (!ctx->refs)
|
| 1702 | return -ENOMEM;
|
| 1703 |
|
| 1704 | ret = find_parent_nodes(ctx, NULL);
|
| 1705 | if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP ||
|
| 1706 | (ret < 0 && ret != -ENOENT)) {
|
| 1707 | free_leaf_list(ctx->refs);
|
| 1708 | ctx->refs = NULL;
|
| 1709 | return ret;
|
| 1710 | }
|
| 1711 |
|
| 1712 | return 0;
|
| 1713 | }
|
| 1714 |
|
| 1715 | |
| 1716 | * Walk all backrefs for a given extent to find all roots that reference this
|
| 1717 | * extent. Walking a backref means finding all extents that reference this
|
| 1718 | * extent and in turn walk the backrefs of those, too. Naturally this is a
|
| 1719 | * recursive process, but here it is implemented in an iterative fashion: We
|
| 1720 | * find all referencing extents for the extent in question and put them on a
|
| 1721 | * list. In turn, we find all referencing extents for those, further appending
|
| 1722 | * to the list. The way we iterate the list allows adding more elements after
|
| 1723 | * the current while iterating. The process stops when we reach the end of the
|
| 1724 | * list.
|
| 1725 | *
|
| 1726 | * Found roots are added to @ctx->roots, which is allocated by this function if
|
| 1727 | * it points to NULL, in which case the caller is responsible for freeing it
|
| 1728 | * after it's not needed anymore.
|
| 1729 | * This function requires @ctx->refs to be NULL, as it uses it for allocating a
|
| 1730 | * ulist to do temporary work, and frees it before returning.
|
| 1731 | *
|
| 1732 | * Returns 0 on success, < 0 on error.
|
| 1733 |
|
| 1734 | static int btrfs_find_all_roots_safe(struct btrfs_backref_walk_ctx *ctx)
|
| 1735 | {
|
| 1736 | const u64 orig_bytenr = ctx->bytenr;
|
| 1737 | const bool orig_skip_inode_ref_list = ctx->skip_inode_ref_list;
|
| 1738 | bool roots_ulist_allocated = false;
|
| 1739 | struct ulist_iterator uiter;
|
| 1740 | int ret = 0;
|
| 1741 |
|
| 1742 | ASSERT(ctx->refs == NULL);
|
| 1743 |
|
| 1744 | ctx->refs = ulist_alloc(GFP_NOFS);
|
| 1745 | if (!ctx->refs)
|
| 1746 | return -ENOMEM;
|
| 1747 |
|
| 1748 | if (!ctx->roots) {
|
| 1749 | ctx->roots = ulist_alloc(GFP_NOFS);
|
| 1750 | if (!ctx->roots) {
|
| 1751 | ulist_free(ctx->refs);
|
| 1752 | ctx->refs = NULL;
|
| 1753 | return -ENOMEM;
|
| 1754 | }
|
| 1755 | roots_ulist_allocated = true;
|
| 1756 | }
|
| 1757 |
|
| 1758 | ctx->skip_inode_ref_list = true;
|
| 1759 |
|
| 1760 | ULIST_ITER_INIT(&uiter);
|
| 1761 | while (1) {
|
| 1762 | struct ulist_node *node;
|
| 1763 |
|
| 1764 | ret = find_parent_nodes(ctx, NULL);
|
| 1765 | if (ret < 0 && ret != -ENOENT) {
|
| 1766 | if (roots_ulist_allocated) {
|
| 1767 | ulist_free(ctx->roots);
|
| 1768 | ctx->roots = NULL;
|
| 1769 | }
|
| 1770 | break;
|
| 1771 | }
|
| 1772 | ret = 0;
|
| 1773 | node = ulist_next(ctx->refs, &uiter);
|
| 1774 | if (!node)
|
| 1775 | break;
|
| 1776 | ctx->bytenr = node->val;
|
| 1777 | cond_resched();
|
| 1778 | }
|
| 1779 |
|
| 1780 | ulist_free(ctx->refs);
|
| 1781 | ctx->refs = NULL;
|
| 1782 | ctx->bytenr = orig_bytenr;
|
| 1783 | ctx->skip_inode_ref_list = orig_skip_inode_ref_list;
|
| 1784 |
|
| 1785 | return ret;
|
| 1786 | }
|
| 1787 |
|
| 1788 | int btrfs_find_all_roots(struct btrfs_backref_walk_ctx *ctx,
|
| 1789 | bool skip_commit_root_sem)
|
| 1790 | {
|
| 1791 | int ret;
|
| 1792 |
|
| 1793 | if (!ctx->trans && !skip_commit_root_sem)
|
| 1794 | down_read(&ctx->fs_info->commit_root_sem);
|
| 1795 | ret = btrfs_find_all_roots_safe(ctx);
|
| 1796 | if (!ctx->trans && !skip_commit_root_sem)
|
| 1797 | up_read(&ctx->fs_info->commit_root_sem);
|
| 1798 | return ret;
|
| 1799 | }
|
| 1800 |
|
| 1801 | struct btrfs_backref_share_check_ctx *btrfs_alloc_backref_share_check_ctx(void)
|
| 1802 | {
|
| 1803 | struct btrfs_backref_share_check_ctx *ctx;
|
| 1804 |
|
| 1805 | ctx = kzalloc_obj(*ctx);
|
| 1806 | if (!ctx)
|
| 1807 | return NULL;
|
| 1808 |
|
| 1809 | ulist_init(&ctx->refs);
|
| 1810 |
|
| 1811 | return ctx;
|
| 1812 | }
|
| 1813 |
|
| 1814 | void btrfs_free_backref_share_ctx(struct btrfs_backref_share_check_ctx *ctx)
|
| 1815 | {
|
| 1816 | if (!ctx)
|
| 1817 | return;
|
| 1818 |
|
| 1819 | ulist_release(&ctx->refs);
|
| 1820 | kfree(ctx);
|
| 1821 | }
|
| 1822 |
|
| 1823 | |
| 1824 | * Check if a data extent is shared or not.
|
| 1825 | *
|
| 1826 | * @inode: The inode whose extent we are checking.
|
| 1827 | * @bytenr: Logical bytenr of the extent we are checking.
|
| 1828 | * @extent_gen: Generation of the extent (file extent item) or 0 if it is
|
| 1829 | * not known.
|
| 1830 | * @ctx: A backref sharedness check context.
|
| 1831 | *
|
| 1832 | * btrfs_is_data_extent_shared uses the backref walking code but will short
|
| 1833 | * circuit as soon as it finds a root or inode that doesn't match the
|
| 1834 | * one passed in. This provides a significant performance benefit for
|
| 1835 | * callers (such as fiemap) which want to know whether the extent is
|
| 1836 | * shared but do not need a ref count.
|
| 1837 | *
|
| 1838 | * This attempts to attach to the running transaction in order to account for
|
| 1839 | * delayed refs, but continues on even when no running transaction exists.
|
| 1840 | *
|
| 1841 | * Return: 0 if extent is not shared, 1 if it is shared, < 0 on error.
|
| 1842 |
|
| 1843 | int btrfs_is_data_extent_shared(struct btrfs_inode *inode, u64 bytenr,
|
| 1844 | u64 extent_gen,
|
| 1845 | struct btrfs_backref_share_check_ctx *ctx)
|
| 1846 | {
|
| 1847 | struct btrfs_backref_walk_ctx walk_ctx = { 0 };
|
| 1848 | struct btrfs_root *root = inode->root;
|
| 1849 | struct btrfs_fs_info *fs_info = root->fs_info;
|
| 1850 | struct btrfs_trans_handle *trans;
|
| 1851 | struct ulist_iterator uiter;
|
| 1852 | struct ulist_node *node;
|
| 1853 | struct btrfs_seq_list elem = BTRFS_SEQ_LIST_INIT(elem);
|
| 1854 | int ret = 0;
|
| 1855 | struct share_check shared = {
|
| 1856 | .ctx = ctx,
|
| 1857 | .root = root,
|
| 1858 | .inum = btrfs_ino(inode),
|
| 1859 | .data_bytenr = bytenr,
|
| 1860 | .data_extent_gen = extent_gen,
|
| 1861 | .share_count = 0,
|
| 1862 | .self_ref_count = 0,
|
| 1863 | .have_delayed_delete_refs = false,
|
| 1864 | };
|
| 1865 | int level;
|
| 1866 | bool leaf_cached;
|
| 1867 | bool leaf_is_shared;
|
| 1868 |
|
| 1869 | for (int i = 0; i < BTRFS_BACKREF_CTX_PREV_EXTENTS_SIZE; i++) {
|
| 1870 | if (ctx->prev_extents_cache[i].bytenr == bytenr)
|
| 1871 | return ctx->prev_extents_cache[i].is_shared;
|
| 1872 | }
|
| 1873 |
|
| 1874 | ulist_init(&ctx->refs);
|
| 1875 |
|
| 1876 | trans = btrfs_join_transaction_nostart(root);
|
| 1877 | if (IS_ERR(trans)) {
|
| 1878 | if (PTR_ERR(trans) != -ENOENT && PTR_ERR(trans) != -EROFS) {
|
| 1879 | ret = PTR_ERR(trans);
|
| 1880 | goto out;
|
| 1881 | }
|
| 1882 | trans = NULL;
|
| 1883 | down_read(&fs_info->commit_root_sem);
|
| 1884 | } else {
|
| 1885 | btrfs_get_tree_mod_seq(fs_info, &elem);
|
| 1886 | walk_ctx.time_seq = elem.seq;
|
| 1887 | }
|
| 1888 |
|
| 1889 | ctx->use_path_cache = true;
|
| 1890 |
|
| 1891 | |
| 1892 | * We may have previously determined that the current leaf is shared.
|
| 1893 | * If it is, then we have a data extent that is shared due to a shared
|
| 1894 | * subtree (caused by snapshotting) and we don't need to check for data
|
| 1895 | * backrefs. If the leaf is not shared, then we must do backref walking
|
| 1896 | * to determine if the data extent is shared through reflinks.
|
| 1897 |
|
| 1898 | leaf_cached = lookup_backref_shared_cache(ctx, root,
|
| 1899 | ctx->curr_leaf_bytenr, 0,
|
| 1900 | &leaf_is_shared);
|
| 1901 | if (leaf_cached && leaf_is_shared) {
|
| 1902 | ret = 1;
|
| 1903 | goto out_trans;
|
| 1904 | }
|
| 1905 |
|
| 1906 | walk_ctx.skip_inode_ref_list = true;
|
| 1907 | walk_ctx.trans = trans;
|
| 1908 | walk_ctx.fs_info = fs_info;
|
| 1909 | walk_ctx.refs = &ctx->refs;
|
| 1910 |
|
| 1911 |
|
| 1912 | level = -1;
|
| 1913 | ULIST_ITER_INIT(&uiter);
|
| 1914 | while (1) {
|
| 1915 | const unsigned long prev_ref_count = ctx->refs.nnodes;
|
| 1916 |
|
| 1917 | walk_ctx.bytenr = bytenr;
|
| 1918 | ret = find_parent_nodes(&walk_ctx, &shared);
|
| 1919 | if (ret == BACKREF_FOUND_SHARED ||
|
| 1920 | ret == BACKREF_FOUND_NOT_SHARED) {
|
| 1921 |
|
| 1922 | ret = (ret == BACKREF_FOUND_SHARED) ? 1 : 0;
|
| 1923 | if (level >= 0)
|
| 1924 | store_backref_shared_cache(ctx, root, bytenr,
|
| 1925 | level, ret == 1);
|
| 1926 | break;
|
| 1927 | }
|
| 1928 | if (ret < 0 && ret != -ENOENT)
|
| 1929 | break;
|
| 1930 | ret = 0;
|
| 1931 |
|
| 1932 | |
| 1933 | * More than one extent buffer (bytenr) may have been added to
|
| 1934 | * the ctx->refs ulist, in which case we have to check multiple
|
| 1935 | * tree paths in case the first one is not shared, so we can not
|
| 1936 | * use the path cache which is made for a single path. Multiple
|
| 1937 | * extent buffers at the current level happen when:
|
| 1938 | *
|
| 1939 | * 1) level -1, the data extent: If our data extent was not
|
| 1940 | * directly shared (without multiple reference items), then
|
| 1941 | * it might have a single reference item with a count > 1 for
|
| 1942 | * the same offset, which means there are 2 (or more) file
|
| 1943 | * extent items that point to the data extent - this happens
|
| 1944 | * when a file extent item needs to be split and then one
|
| 1945 | * item gets moved to another leaf due to a b+tree leaf split
|
| 1946 | * when inserting some item. In this case the file extent
|
| 1947 | * items may be located in different leaves and therefore
|
| 1948 | * some of the leaves may be referenced through shared
|
| 1949 | * subtrees while others are not. Since our extent buffer
|
| 1950 | * cache only works for a single path (by far the most common
|
| 1951 | * case and simpler to deal with), we can not use it if we
|
| 1952 | * have multiple leaves (which implies multiple paths).
|
| 1953 | *
|
| 1954 | * 2) level >= 0, a tree node/leaf: We can have a mix of direct
|
| 1955 | * and indirect references on a b+tree node/leaf, so we have
|
| 1956 | * to check multiple paths, and the extent buffer (the
|
| 1957 | * current bytenr) may be shared or not. One example is
|
| 1958 | * during relocation as we may get a shared tree block ref
|
| 1959 | * (direct ref) and a non-shared tree block ref (indirect
|
| 1960 | * ref) for the same node/leaf.
|
| 1961 |
|
| 1962 | if ((ctx->refs.nnodes - prev_ref_count) > 1)
|
| 1963 | ctx->use_path_cache = false;
|
| 1964 |
|
| 1965 | if (level >= 0)
|
| 1966 | store_backref_shared_cache(ctx, root, bytenr,
|
| 1967 | level, false);
|
| 1968 | node = ulist_next(&ctx->refs, &uiter);
|
| 1969 | if (!node)
|
| 1970 | break;
|
| 1971 | bytenr = node->val;
|
| 1972 | if (ctx->use_path_cache) {
|
| 1973 | bool is_shared;
|
| 1974 | bool cached;
|
| 1975 |
|
| 1976 | level++;
|
| 1977 | cached = lookup_backref_shared_cache(ctx, root, bytenr,
|
| 1978 | level, &is_shared);
|
| 1979 | if (cached) {
|
| 1980 | ret = (is_shared ? 1 : 0);
|
| 1981 | break;
|
| 1982 | }
|
| 1983 | }
|
| 1984 | shared.share_count = 0;
|
| 1985 | shared.have_delayed_delete_refs = false;
|
| 1986 | cond_resched();
|
| 1987 | }
|
| 1988 |
|
| 1989 | |
| 1990 | * If the path cache is disabled, then it means at some tree level we
|
| 1991 | * got multiple parents due to a mix of direct and indirect backrefs or
|
| 1992 | * multiple leaves with file extent items pointing to the same data
|
| 1993 | * extent. We have to invalidate the cache and cache only the sharedness
|
| 1994 | * result for the levels where we got only one node/reference.
|
| 1995 |
|
| 1996 | if (!ctx->use_path_cache) {
|
| 1997 | int i = 0;
|
| 1998 |
|
| 1999 | level--;
|
| 2000 | if (ret >= 0 && level >= 0) {
|
| 2001 | bytenr = ctx->path_cache_entries[level].bytenr;
|
| 2002 | ctx->use_path_cache = true;
|
| 2003 | store_backref_shared_cache(ctx, root, bytenr, level, ret);
|
| 2004 | i = level + 1;
|
| 2005 | }
|
| 2006 |
|
| 2007 | for ( ; i < BTRFS_MAX_LEVEL; i++)
|
| 2008 | ctx->path_cache_entries[i].bytenr = 0;
|
| 2009 | }
|
| 2010 |
|
| 2011 | |
| 2012 | * Cache the sharedness result for the data extent if we know our inode
|
| 2013 | * has more than 1 file extent item that refers to the data extent.
|
| 2014 |
|
| 2015 | if (ret >= 0 && shared.self_ref_count > 1) {
|
| 2016 | int slot = ctx->prev_extents_cache_slot;
|
| 2017 |
|
| 2018 | ctx->prev_extents_cache[slot].bytenr = shared.data_bytenr;
|
| 2019 | ctx->prev_extents_cache[slot].is_shared = (ret == 1);
|
| 2020 |
|
| 2021 | slot = (slot + 1) % BTRFS_BACKREF_CTX_PREV_EXTENTS_SIZE;
|
| 2022 | ctx->prev_extents_cache_slot = slot;
|
| 2023 | }
|
| 2024 |
|
| 2025 | out_trans:
|
| 2026 | if (trans) {
|
| 2027 | btrfs_put_tree_mod_seq(fs_info, &elem);
|
| 2028 | btrfs_end_transaction(trans);
|
| 2029 | } else {
|
| 2030 | up_read(&fs_info->commit_root_sem);
|
| 2031 | }
|
| 2032 | out:
|
| 2033 | ulist_release(&ctx->refs);
|
| 2034 | ctx->prev_leaf_bytenr = ctx->curr_leaf_bytenr;
|
| 2035 |
|
| 2036 | return ret;
|
| 2037 | }
|
| 2038 |
|
| 2039 | int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
|
| 2040 | u64 start_off, struct btrfs_path *path,
|
| 2041 | struct btrfs_inode_extref **ret_extref,
|
| 2042 | u64 *found_off)
|
| 2043 | {
|
| 2044 | int ret, slot;
|
| 2045 | struct btrfs_key key;
|
| 2046 | struct btrfs_key found_key;
|
| 2047 | struct btrfs_inode_extref *extref;
|
| 2048 | const struct extent_buffer *leaf;
|
| 2049 | unsigned long ptr;
|
| 2050 |
|
| 2051 | key.objectid = inode_objectid;
|
| 2052 | key.type = BTRFS_INODE_EXTREF_KEY;
|
| 2053 | key.offset = start_off;
|
| 2054 |
|
| 2055 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
|
| 2056 | if (ret < 0)
|
| 2057 | return ret;
|
| 2058 |
|
| 2059 | while (1) {
|
| 2060 | leaf = path->nodes[0];
|
| 2061 | slot = path->slots[0];
|
| 2062 | if (slot >= btrfs_header_nritems(leaf)) {
|
| 2063 | |
| 2064 | * If the item at offset is not found,
|
| 2065 | * btrfs_search_slot will point us to the slot
|
| 2066 | * where it should be inserted. In our case
|
| 2067 | * that will be the slot directly before the
|
| 2068 | * next INODE_REF_KEY_V2 item. In the case
|
| 2069 | * that we're pointing to the last slot in a
|
| 2070 | * leaf, we must move one leaf over.
|
| 2071 |
|
| 2072 | ret = btrfs_next_leaf(root, path);
|
| 2073 | if (ret) {
|
| 2074 | if (ret >= 1)
|
| 2075 | ret = -ENOENT;
|
| 2076 | break;
|
| 2077 | }
|
| 2078 | continue;
|
| 2079 | }
|
| 2080 |
|
| 2081 | btrfs_item_key_to_cpu(leaf, &found_key, slot);
|
| 2082 |
|
| 2083 | |
| 2084 | * Check that we're still looking at an extended ref key for
|
| 2085 | * this particular objectid. If we have different
|
| 2086 | * objectid or type then there are no more to be found
|
| 2087 | * in the tree and we can exit.
|
| 2088 |
|
| 2089 | ret = -ENOENT;
|
| 2090 | if (found_key.objectid != inode_objectid)
|
| 2091 | break;
|
| 2092 | if (found_key.type != BTRFS_INODE_EXTREF_KEY)
|
| 2093 | break;
|
| 2094 |
|
| 2095 | ret = 0;
|
| 2096 | ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
|
| 2097 | extref = (struct btrfs_inode_extref *)ptr;
|
| 2098 | *ret_extref = extref;
|
| 2099 | if (found_off)
|
| 2100 | *found_off = found_key.offset;
|
| 2101 | break;
|
| 2102 | }
|
| 2103 |
|
| 2104 | return ret;
|
| 2105 | }
|
| 2106 |
|
| 2107 | |
| 2108 | * this iterates to turn a name (from iref/extref) into a full filesystem path.
|
| 2109 | * Elements of the path are separated by '/' and the path is guaranteed to be
|
| 2110 | * 0-terminated. the path is only given within the current file system.
|
| 2111 | * Therefore, it never starts with a '/'. the caller is responsible to provide
|
| 2112 | * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
|
| 2113 | * the start point of the resulting string is returned. this pointer is within
|
| 2114 | * dest, normally.
|
| 2115 | * in case the path buffer would overflow, the pointer is decremented further
|
| 2116 | * as if output was written to the buffer, though no more output is actually
|
| 2117 | * generated. that way, the caller can determine how much space would be
|
| 2118 | * required for the path to fit into the buffer. in that case, the returned
|
| 2119 | * value will be smaller than dest. callers must check this!
|
| 2120 |
|
| 2121 | char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
|
| 2122 | u32 name_len, unsigned long name_off,
|
| 2123 | struct extent_buffer *eb_in, u64 parent,
|
| 2124 | char *dest, u32 size)
|
| 2125 | {
|
| 2126 | int slot;
|
| 2127 | u64 next_inum;
|
| 2128 | int ret;
|
| 2129 | s64 bytes_left = ((s64)size) - 1;
|
| 2130 | struct extent_buffer *eb = eb_in;
|
| 2131 | struct btrfs_key found_key;
|
| 2132 | struct btrfs_inode_ref *iref;
|
| 2133 |
|
| 2134 | if (bytes_left >= 0)
|
| 2135 | dest[bytes_left] = '\0';
|
| 2136 |
|
| 2137 | while (1) {
|
| 2138 | bytes_left -= name_len;
|
| 2139 | if (bytes_left >= 0)
|
| 2140 | read_extent_buffer(eb, dest + bytes_left,
|
| 2141 | name_off, name_len);
|
| 2142 | if (eb != eb_in) {
|
| 2143 | if (!path->skip_locking)
|
| 2144 | btrfs_tree_read_unlock(eb);
|
| 2145 | free_extent_buffer(eb);
|
| 2146 | }
|
| 2147 | ret = btrfs_find_item(fs_root, path, parent, 0,
|
| 2148 | BTRFS_INODE_REF_KEY, &found_key);
|
| 2149 | if (ret > 0)
|
| 2150 | ret = -ENOENT;
|
| 2151 | if (ret)
|
| 2152 | break;
|
| 2153 |
|
| 2154 | next_inum = found_key.offset;
|
| 2155 |
|
| 2156 |
|
| 2157 | if (parent == next_inum)
|
| 2158 | break;
|
| 2159 |
|
| 2160 | slot = path->slots[0];
|
| 2161 | eb = path->nodes[0];
|
| 2162 |
|
| 2163 | if (eb != eb_in) {
|
| 2164 | path->nodes[0] = NULL;
|
| 2165 | path->locks[0] = 0;
|
| 2166 | }
|
| 2167 | btrfs_release_path(path);
|
| 2168 | iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
|
| 2169 |
|
| 2170 | name_len = btrfs_inode_ref_name_len(eb, iref);
|
| 2171 | name_off = (unsigned long)(iref + 1);
|
| 2172 |
|
| 2173 | parent = next_inum;
|
| 2174 | --bytes_left;
|
| 2175 | if (bytes_left >= 0)
|
| 2176 | dest[bytes_left] = '/';
|
| 2177 | }
|
| 2178 |
|
| 2179 | btrfs_release_path(path);
|
| 2180 |
|
| 2181 | if (ret)
|
| 2182 | return ERR_PTR(ret);
|
| 2183 |
|
| 2184 | return dest + bytes_left;
|
| 2185 | }
|
| 2186 |
|
| 2187 | |
| 2188 | * this makes the path point to (logical EXTENT_ITEM *)
|
| 2189 | * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
|
| 2190 | * tree blocks and <0 on error.
|
| 2191 |
|
| 2192 | int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
|
| 2193 | struct btrfs_path *path, struct btrfs_key *found_key,
|
| 2194 | u64 *flags_ret)
|
| 2195 | {
|
| 2196 | struct btrfs_root *extent_root = btrfs_extent_root(fs_info, logical);
|
| 2197 | int ret;
|
| 2198 | u64 flags;
|
| 2199 | u64 size = 0;
|
| 2200 | const struct extent_buffer *eb;
|
| 2201 | struct btrfs_extent_item *ei;
|
| 2202 | struct btrfs_key key;
|
| 2203 |
|
| 2204 | if (unlikely(!extent_root)) {
|
| 2205 | btrfs_err(fs_info,
|
| 2206 | "missing extent root for extent at bytenr %llu",
|
| 2207 | logical);
|
| 2208 | return -EUCLEAN;
|
| 2209 | }
|
| 2210 |
|
| 2211 | key.objectid = logical;
|
| 2212 | if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
|
| 2213 | key.type = BTRFS_METADATA_ITEM_KEY;
|
| 2214 | else
|
| 2215 | key.type = BTRFS_EXTENT_ITEM_KEY;
|
| 2216 | key.offset = (u64)-1;
|
| 2217 |
|
| 2218 | ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
|
| 2219 | if (ret < 0)
|
| 2220 | return ret;
|
| 2221 | if (unlikely(ret == 0)) {
|
| 2222 | |
| 2223 | * Key with offset -1 found, there would have to exist an extent
|
| 2224 | * item with such offset, but this is out of the valid range.
|
| 2225 |
|
| 2226 | return -EUCLEAN;
|
| 2227 | }
|
| 2228 |
|
| 2229 | ret = btrfs_previous_extent_item(extent_root, path, 0);
|
| 2230 | if (ret) {
|
| 2231 | if (ret > 0)
|
| 2232 | ret = -ENOENT;
|
| 2233 | return ret;
|
| 2234 | }
|
| 2235 | btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
|
| 2236 | if (found_key->type == BTRFS_METADATA_ITEM_KEY)
|
| 2237 | size = fs_info->nodesize;
|
| 2238 | else if (found_key->type == BTRFS_EXTENT_ITEM_KEY)
|
| 2239 | size = found_key->offset;
|
| 2240 |
|
| 2241 | if (found_key->objectid > logical ||
|
| 2242 | found_key->objectid + size <= logical) {
|
| 2243 | btrfs_debug(fs_info,
|
| 2244 | "logical %llu is not within any extent", logical);
|
| 2245 | return -ENOENT;
|
| 2246 | }
|
| 2247 |
|
| 2248 | eb = path->nodes[0];
|
| 2249 |
|
| 2250 | ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
|
| 2251 | flags = btrfs_extent_flags(eb, ei);
|
| 2252 |
|
| 2253 | btrfs_debug(fs_info,
|
| 2254 | "logical %llu is at position %llu within the extent (%llu EXTENT_ITEM %llu) flags %#llx size %u",
|
| 2255 | logical, logical - found_key->objectid, found_key->objectid,
|
| 2256 | found_key->offset, flags, btrfs_item_size(eb, path->slots[0]));
|
| 2257 |
|
| 2258 | WARN_ON(!flags_ret);
|
| 2259 | if (flags_ret) {
|
| 2260 | if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
|
| 2261 | *flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK;
|
| 2262 | else if (flags & BTRFS_EXTENT_FLAG_DATA)
|
| 2263 | *flags_ret = BTRFS_EXTENT_FLAG_DATA;
|
| 2264 | else
|
| 2265 | BUG();
|
| 2266 | return 0;
|
| 2267 | }
|
| 2268 |
|
| 2269 | return -EIO;
|
| 2270 | }
|
| 2271 |
|
| 2272 | |
| 2273 | * helper function to iterate extent inline refs. ptr must point to a 0 value
|
| 2274 | * for the first call and may be modified. it is used to track state.
|
| 2275 | * if more refs exist, 0 is returned and the next call to
|
| 2276 | * get_extent_inline_ref must pass the modified ptr parameter to get the
|
| 2277 | * next ref. after the last ref was processed, 1 is returned.
|
| 2278 | * returns <0 on error
|
| 2279 |
|
| 2280 | static int get_extent_inline_ref(unsigned long *ptr,
|
| 2281 | const struct extent_buffer *eb,
|
| 2282 | const struct btrfs_key *key,
|
| 2283 | const struct btrfs_extent_item *ei,
|
| 2284 | u32 item_size,
|
| 2285 | struct btrfs_extent_inline_ref **out_eiref,
|
| 2286 | int *out_type)
|
| 2287 | {
|
| 2288 | unsigned long end;
|
| 2289 | u64 flags;
|
| 2290 | struct btrfs_tree_block_info *info;
|
| 2291 |
|
| 2292 | if (!*ptr) {
|
| 2293 |
|
| 2294 | flags = btrfs_extent_flags(eb, ei);
|
| 2295 | if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
|
| 2296 | if (key->type == BTRFS_METADATA_ITEM_KEY) {
|
| 2297 |
|
| 2298 | *out_eiref =
|
| 2299 | (struct btrfs_extent_inline_ref *)(ei + 1);
|
| 2300 | } else {
|
| 2301 | WARN_ON(key->type != BTRFS_EXTENT_ITEM_KEY);
|
| 2302 | info = (struct btrfs_tree_block_info *)(ei + 1);
|
| 2303 | *out_eiref =
|
| 2304 | (struct btrfs_extent_inline_ref *)(info + 1);
|
| 2305 | }
|
| 2306 | } else {
|
| 2307 | *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
|
| 2308 | }
|
| 2309 | *ptr = (unsigned long)*out_eiref;
|
| 2310 | if ((unsigned long)(*ptr) >= (unsigned long)ei + item_size)
|
| 2311 | return -ENOENT;
|
| 2312 | }
|
| 2313 |
|
| 2314 | end = (unsigned long)ei + item_size;
|
| 2315 | *out_eiref = (struct btrfs_extent_inline_ref *)(*ptr);
|
| 2316 | *out_type = btrfs_get_extent_inline_ref_type(eb, *out_eiref,
|
| 2317 | BTRFS_REF_TYPE_ANY);
|
| 2318 | if (unlikely(*out_type == BTRFS_REF_TYPE_INVALID))
|
| 2319 | return -EUCLEAN;
|
| 2320 |
|
| 2321 | *ptr += btrfs_extent_inline_ref_size(*out_type);
|
| 2322 | WARN_ON(*ptr > end);
|
| 2323 | if (*ptr == end)
|
| 2324 | return 1;
|
| 2325 |
|
| 2326 | return 0;
|
| 2327 | }
|
| 2328 |
|
| 2329 | |
| 2330 | * reads the tree block backref for an extent. tree level and root are returned
|
| 2331 | * through out_level and out_root. ptr must point to a 0 value for the first
|
| 2332 | * call and may be modified (see get_extent_inline_ref comment).
|
| 2333 | * returns 0 if data was provided, 1 if there was no more data to provide or
|
| 2334 | * <0 on error.
|
| 2335 |
|
| 2336 | int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
|
| 2337 | struct btrfs_key *key, struct btrfs_extent_item *ei,
|
| 2338 | u32 item_size, u64 *out_root, u8 *out_level)
|
| 2339 | {
|
| 2340 | int ret;
|
| 2341 | int type;
|
| 2342 | struct btrfs_extent_inline_ref *eiref;
|
| 2343 |
|
| 2344 | if (*ptr == (unsigned long)-1)
|
| 2345 | return 1;
|
| 2346 |
|
| 2347 | while (1) {
|
| 2348 | ret = get_extent_inline_ref(ptr, eb, key, ei, item_size,
|
| 2349 | &eiref, &type);
|
| 2350 | if (ret < 0)
|
| 2351 | return ret;
|
| 2352 |
|
| 2353 | if (type == BTRFS_TREE_BLOCK_REF_KEY ||
|
| 2354 | type == BTRFS_SHARED_BLOCK_REF_KEY)
|
| 2355 | break;
|
| 2356 |
|
| 2357 | if (ret == 1)
|
| 2358 | return 1;
|
| 2359 | }
|
| 2360 |
|
| 2361 |
|
| 2362 | *out_root = btrfs_extent_inline_ref_offset(eb, eiref);
|
| 2363 |
|
| 2364 | if (key->type == BTRFS_EXTENT_ITEM_KEY) {
|
| 2365 | struct btrfs_tree_block_info *info;
|
| 2366 |
|
| 2367 | info = (struct btrfs_tree_block_info *)(ei + 1);
|
| 2368 | *out_level = btrfs_tree_block_level(eb, info);
|
| 2369 | } else {
|
| 2370 | ASSERT(key->type == BTRFS_METADATA_ITEM_KEY, "key->type=%hhu", key->type);
|
| 2371 | *out_level = (u8)key->offset;
|
| 2372 | }
|
| 2373 |
|
| 2374 | if (ret == 1)
|
| 2375 | *ptr = (unsigned long)-1;
|
| 2376 |
|
| 2377 | return 0;
|
| 2378 | }
|
| 2379 |
|
| 2380 | static int iterate_leaf_refs(struct btrfs_fs_info *fs_info,
|
| 2381 | struct extent_inode_elem *inode_list,
|
| 2382 | u64 root, u64 extent_item_objectid,
|
| 2383 | iterate_extent_inodes_t *iterate, void *ctx)
|
| 2384 | {
|
| 2385 | struct extent_inode_elem *eie;
|
| 2386 | int ret = 0;
|
| 2387 |
|
| 2388 | for (eie = inode_list; eie; eie = eie->next) {
|
| 2389 | btrfs_debug(fs_info,
|
| 2390 | "ref for %llu resolved, key (%llu EXTEND_DATA %llu), root %llu",
|
| 2391 | extent_item_objectid, eie->inum,
|
| 2392 | eie->offset, root);
|
| 2393 | ret = iterate(eie->inum, eie->offset, eie->num_bytes, root, ctx);
|
| 2394 | if (ret) {
|
| 2395 | btrfs_debug(fs_info,
|
| 2396 | "stopping iteration for %llu due to ret=%d",
|
| 2397 | extent_item_objectid, ret);
|
| 2398 | break;
|
| 2399 | }
|
| 2400 | }
|
| 2401 |
|
| 2402 | return ret;
|
| 2403 | }
|
| 2404 |
|
| 2405 | |
| 2406 | * calls iterate() for every inode that references the extent identified by
|
| 2407 | * the given parameters.
|
| 2408 | * when the iterator function returns a non-zero value, iteration stops.
|
| 2409 |
|
| 2410 | int iterate_extent_inodes(struct btrfs_backref_walk_ctx *ctx,
|
| 2411 | bool search_commit_root,
|
| 2412 | iterate_extent_inodes_t *iterate, void *user_ctx)
|
| 2413 | {
|
| 2414 | int ret;
|
| 2415 | struct ulist *refs;
|
| 2416 | struct ulist_node *ref_node;
|
| 2417 | struct btrfs_seq_list seq_elem = BTRFS_SEQ_LIST_INIT(seq_elem);
|
| 2418 | struct ulist_iterator ref_uiter;
|
| 2419 |
|
| 2420 | btrfs_debug(ctx->fs_info, "resolving all inodes for extent %llu",
|
| 2421 | ctx->bytenr);
|
| 2422 |
|
| 2423 | ASSERT(ctx->trans == NULL);
|
| 2424 | ASSERT(ctx->roots == NULL);
|
| 2425 |
|
| 2426 | if (!search_commit_root) {
|
| 2427 | struct btrfs_trans_handle *trans;
|
| 2428 |
|
| 2429 | trans = btrfs_attach_transaction(ctx->fs_info->tree_root);
|
| 2430 | if (IS_ERR(trans)) {
|
| 2431 | if (PTR_ERR(trans) != -ENOENT &&
|
| 2432 | PTR_ERR(trans) != -EROFS)
|
| 2433 | return PTR_ERR(trans);
|
| 2434 | trans = NULL;
|
| 2435 | }
|
| 2436 | ctx->trans = trans;
|
| 2437 | }
|
| 2438 |
|
| 2439 | if (ctx->trans) {
|
| 2440 | btrfs_get_tree_mod_seq(ctx->fs_info, &seq_elem);
|
| 2441 | ctx->time_seq = seq_elem.seq;
|
| 2442 | } else {
|
| 2443 | down_read(&ctx->fs_info->commit_root_sem);
|
| 2444 | }
|
| 2445 |
|
| 2446 | ret = btrfs_find_all_leafs(ctx);
|
| 2447 | if (ret)
|
| 2448 | goto out;
|
| 2449 | refs = ctx->refs;
|
| 2450 | ctx->refs = NULL;
|
| 2451 |
|
| 2452 | ULIST_ITER_INIT(&ref_uiter);
|
| 2453 | while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
|
| 2454 | const u64 leaf_bytenr = ref_node->val;
|
| 2455 | struct ulist_node *root_node;
|
| 2456 | struct ulist_iterator root_uiter;
|
| 2457 | struct extent_inode_elem *inode_list;
|
| 2458 |
|
| 2459 | inode_list = (struct extent_inode_elem *)(uintptr_t)ref_node->aux;
|
| 2460 |
|
| 2461 | if (ctx->cache_lookup) {
|
| 2462 | const u64 *root_ids;
|
| 2463 | int root_count;
|
| 2464 | bool cached;
|
| 2465 |
|
| 2466 | cached = ctx->cache_lookup(leaf_bytenr, ctx->user_ctx,
|
| 2467 | &root_ids, &root_count);
|
| 2468 | if (cached) {
|
| 2469 | for (int i = 0; i < root_count; i++) {
|
| 2470 | ret = iterate_leaf_refs(ctx->fs_info,
|
| 2471 | inode_list,
|
| 2472 | root_ids[i],
|
| 2473 | leaf_bytenr,
|
| 2474 | iterate,
|
| 2475 | user_ctx);
|
| 2476 | if (ret)
|
| 2477 | break;
|
| 2478 | }
|
| 2479 | continue;
|
| 2480 | }
|
| 2481 | }
|
| 2482 |
|
| 2483 | if (!ctx->roots) {
|
| 2484 | ctx->roots = ulist_alloc(GFP_NOFS);
|
| 2485 | if (!ctx->roots) {
|
| 2486 | ret = -ENOMEM;
|
| 2487 | break;
|
| 2488 | }
|
| 2489 | }
|
| 2490 |
|
| 2491 | ctx->bytenr = leaf_bytenr;
|
| 2492 | ret = btrfs_find_all_roots_safe(ctx);
|
| 2493 | if (ret)
|
| 2494 | break;
|
| 2495 |
|
| 2496 | if (ctx->cache_store)
|
| 2497 | ctx->cache_store(leaf_bytenr, ctx->roots, ctx->user_ctx);
|
| 2498 |
|
| 2499 | ULIST_ITER_INIT(&root_uiter);
|
| 2500 | while (!ret && (root_node = ulist_next(ctx->roots, &root_uiter))) {
|
| 2501 | btrfs_debug(ctx->fs_info,
|
| 2502 | "root %llu references leaf %llu, data list %#llx",
|
| 2503 | root_node->val, ref_node->val,
|
| 2504 | ref_node->aux);
|
| 2505 | ret = iterate_leaf_refs(ctx->fs_info, inode_list,
|
| 2506 | root_node->val, ctx->bytenr,
|
| 2507 | iterate, user_ctx);
|
| 2508 | }
|
| 2509 | ulist_reinit(ctx->roots);
|
| 2510 | }
|
| 2511 |
|
| 2512 | free_leaf_list(refs);
|
| 2513 | out:
|
| 2514 | if (ctx->trans) {
|
| 2515 | btrfs_put_tree_mod_seq(ctx->fs_info, &seq_elem);
|
| 2516 | btrfs_end_transaction(ctx->trans);
|
| 2517 | ctx->trans = NULL;
|
| 2518 | } else {
|
| 2519 | up_read(&ctx->fs_info->commit_root_sem);
|
| 2520 | }
|
| 2521 |
|
| 2522 | ulist_free(ctx->roots);
|
| 2523 | ctx->roots = NULL;
|
| 2524 |
|
| 2525 | if (ret == BTRFS_ITERATE_EXTENT_INODES_STOP)
|
| 2526 | ret = 0;
|
| 2527 |
|
| 2528 | return ret;
|
| 2529 | }
|
| 2530 |
|
| 2531 | static int build_ino_list(u64 inum, u64 offset, u64 num_bytes, u64 root, void *ctx)
|
| 2532 | {
|
| 2533 | struct btrfs_data_container *inodes = ctx;
|
| 2534 | const size_t c = 3 * sizeof(u64);
|
| 2535 |
|
| 2536 | if (inodes->bytes_left >= c) {
|
| 2537 | inodes->bytes_left -= c;
|
| 2538 | inodes->val[inodes->elem_cnt] = inum;
|
| 2539 | inodes->val[inodes->elem_cnt + 1] = offset;
|
| 2540 | inodes->val[inodes->elem_cnt + 2] = root;
|
| 2541 | inodes->elem_cnt += 3;
|
| 2542 | } else {
|
| 2543 | inodes->bytes_missing += c - inodes->bytes_left;
|
| 2544 | inodes->bytes_left = 0;
|
| 2545 | inodes->elem_missed += 3;
|
| 2546 | }
|
| 2547 |
|
| 2548 | return 0;
|
| 2549 | }
|
| 2550 |
|
| 2551 | int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
|
| 2552 | void *ctx, bool ignore_offset)
|
| 2553 | {
|
| 2554 | struct btrfs_backref_walk_ctx walk_ctx = { 0 };
|
| 2555 | int ret;
|
| 2556 | u64 flags = 0;
|
| 2557 | struct btrfs_key found_key;
|
| 2558 | struct btrfs_path *path;
|
| 2559 |
|
| 2560 | path = btrfs_alloc_path();
|
| 2561 | if (!path)
|
| 2562 | return -ENOMEM;
|
| 2563 |
|
| 2564 | ret = extent_from_logical(fs_info, logical, path, &found_key, &flags);
|
| 2565 | btrfs_free_path(path);
|
| 2566 | if (ret < 0)
|
| 2567 | return ret;
|
| 2568 | if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
|
| 2569 | return -EINVAL;
|
| 2570 |
|
| 2571 | walk_ctx.bytenr = found_key.objectid;
|
| 2572 | if (ignore_offset)
|
| 2573 | walk_ctx.ignore_extent_item_pos = true;
|
| 2574 | else
|
| 2575 | walk_ctx.extent_item_pos = logical - found_key.objectid;
|
| 2576 | walk_ctx.fs_info = fs_info;
|
| 2577 |
|
| 2578 | return iterate_extent_inodes(&walk_ctx, false, build_ino_list, ctx);
|
| 2579 | }
|
| 2580 |
|
| 2581 | static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
|
| 2582 | struct extent_buffer *eb, struct inode_fs_paths *ipath);
|
| 2583 |
|
| 2584 | static int iterate_inode_refs(u64 inum, struct inode_fs_paths *ipath)
|
| 2585 | {
|
| 2586 | int ret = 0;
|
| 2587 | int slot;
|
| 2588 | u32 cur;
|
| 2589 | u32 len;
|
| 2590 | u32 name_len;
|
| 2591 | u64 parent = 0;
|
| 2592 | int found = 0;
|
| 2593 | struct btrfs_root *fs_root = ipath->fs_root;
|
| 2594 | struct btrfs_path *path = ipath->btrfs_path;
|
| 2595 | struct extent_buffer *eb;
|
| 2596 | struct btrfs_inode_ref *iref;
|
| 2597 | struct btrfs_key found_key;
|
| 2598 |
|
| 2599 | while (!ret) {
|
| 2600 | ret = btrfs_find_item(fs_root, path, inum,
|
| 2601 | parent ? parent + 1 : 0, BTRFS_INODE_REF_KEY,
|
| 2602 | &found_key);
|
| 2603 |
|
| 2604 | if (ret < 0)
|
| 2605 | break;
|
| 2606 | if (ret) {
|
| 2607 | ret = found ? 0 : -ENOENT;
|
| 2608 | break;
|
| 2609 | }
|
| 2610 | ++found;
|
| 2611 |
|
| 2612 | parent = found_key.offset;
|
| 2613 | slot = path->slots[0];
|
| 2614 | eb = btrfs_clone_extent_buffer(path->nodes[0]);
|
| 2615 | if (!eb) {
|
| 2616 | ret = -ENOMEM;
|
| 2617 | break;
|
| 2618 | }
|
| 2619 | btrfs_release_path(path);
|
| 2620 |
|
| 2621 | iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
|
| 2622 |
|
| 2623 | for (cur = 0; cur < btrfs_item_size(eb, slot); cur += len) {
|
| 2624 | name_len = btrfs_inode_ref_name_len(eb, iref);
|
| 2625 |
|
| 2626 | btrfs_debug(fs_root->fs_info,
|
| 2627 | "following ref at offset %u for inode %llu in tree %llu",
|
| 2628 | cur, found_key.objectid,
|
| 2629 | btrfs_root_id(fs_root));
|
| 2630 | ret = inode_to_path(parent, name_len,
|
| 2631 | (unsigned long)(iref + 1), eb, ipath);
|
| 2632 | if (ret)
|
| 2633 | break;
|
| 2634 | len = sizeof(*iref) + name_len;
|
| 2635 | iref = (struct btrfs_inode_ref *)((char *)iref + len);
|
| 2636 | }
|
| 2637 | free_extent_buffer(eb);
|
| 2638 | }
|
| 2639 |
|
| 2640 | btrfs_release_path(path);
|
| 2641 |
|
| 2642 | return ret;
|
| 2643 | }
|
| 2644 |
|
| 2645 | static int iterate_inode_extrefs(u64 inum, struct inode_fs_paths *ipath)
|
| 2646 | {
|
| 2647 | int ret;
|
| 2648 | int slot;
|
| 2649 | u64 offset = 0;
|
| 2650 | u64 parent;
|
| 2651 | int found = 0;
|
| 2652 | struct btrfs_root *fs_root = ipath->fs_root;
|
| 2653 | struct btrfs_path *path = ipath->btrfs_path;
|
| 2654 | struct extent_buffer *eb;
|
| 2655 | struct btrfs_inode_extref *extref;
|
| 2656 | u32 item_size;
|
| 2657 | u32 cur_offset;
|
| 2658 | unsigned long ptr;
|
| 2659 |
|
| 2660 | while (1) {
|
| 2661 | ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref,
|
| 2662 | &offset);
|
| 2663 | if (ret < 0)
|
| 2664 | break;
|
| 2665 | if (ret) {
|
| 2666 | ret = found ? 0 : -ENOENT;
|
| 2667 | break;
|
| 2668 | }
|
| 2669 | ++found;
|
| 2670 |
|
| 2671 | slot = path->slots[0];
|
| 2672 | eb = btrfs_clone_extent_buffer(path->nodes[0]);
|
| 2673 | if (!eb) {
|
| 2674 | ret = -ENOMEM;
|
| 2675 | break;
|
| 2676 | }
|
| 2677 | btrfs_release_path(path);
|
| 2678 |
|
| 2679 | item_size = btrfs_item_size(eb, slot);
|
| 2680 | ptr = btrfs_item_ptr_offset(eb, slot);
|
| 2681 | cur_offset = 0;
|
| 2682 |
|
| 2683 | while (cur_offset < item_size) {
|
| 2684 | u32 name_len;
|
| 2685 |
|
| 2686 | extref = (struct btrfs_inode_extref *)(ptr + cur_offset);
|
| 2687 | parent = btrfs_inode_extref_parent(eb, extref);
|
| 2688 | name_len = btrfs_inode_extref_name_len(eb, extref);
|
| 2689 | ret = inode_to_path(parent, name_len,
|
| 2690 | (unsigned long)&extref->name, eb, ipath);
|
| 2691 | if (ret)
|
| 2692 | break;
|
| 2693 |
|
| 2694 | cur_offset += btrfs_inode_extref_name_len(eb, extref);
|
| 2695 | cur_offset += sizeof(*extref);
|
| 2696 | }
|
| 2697 | free_extent_buffer(eb);
|
| 2698 |
|
| 2699 | offset++;
|
| 2700 | }
|
| 2701 |
|
| 2702 | btrfs_release_path(path);
|
| 2703 |
|
| 2704 | return ret;
|
| 2705 | }
|
| 2706 |
|
| 2707 | |
| 2708 | * returns 0 if the path could be dumped (probably truncated)
|
| 2709 | * returns <0 in case of an error
|
| 2710 |
|
| 2711 | static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
|
| 2712 | struct extent_buffer *eb, struct inode_fs_paths *ipath)
|
| 2713 | {
|
| 2714 | char *fspath;
|
| 2715 | char *fspath_min;
|
| 2716 | int i = ipath->fspath->elem_cnt;
|
| 2717 | const int s_ptr = sizeof(char *);
|
| 2718 | u32 bytes_left;
|
| 2719 |
|
| 2720 | bytes_left = ipath->fspath->bytes_left > s_ptr ?
|
| 2721 | ipath->fspath->bytes_left - s_ptr : 0;
|
| 2722 |
|
| 2723 | fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
|
| 2724 | fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len,
|
| 2725 | name_off, eb, inum, fspath_min, bytes_left);
|
| 2726 | if (IS_ERR(fspath))
|
| 2727 | return PTR_ERR(fspath);
|
| 2728 |
|
| 2729 | if (fspath > fspath_min) {
|
| 2730 | ipath->fspath->val[i] = (u64)(unsigned long)fspath;
|
| 2731 | ++ipath->fspath->elem_cnt;
|
| 2732 | ipath->fspath->bytes_left = fspath - fspath_min;
|
| 2733 | } else {
|
| 2734 | ++ipath->fspath->elem_missed;
|
| 2735 | ipath->fspath->bytes_missing += fspath_min - fspath;
|
| 2736 | ipath->fspath->bytes_left = 0;
|
| 2737 | }
|
| 2738 |
|
| 2739 | return 0;
|
| 2740 | }
|
| 2741 |
|
| 2742 | |
| 2743 | * this dumps all file system paths to the inode into the ipath struct, provided
|
| 2744 | * is has been created large enough. each path is zero-terminated and accessed
|
| 2745 | * from ipath->fspath->val[i].
|
| 2746 | * when it returns, there are ipath->fspath->elem_cnt number of paths available
|
| 2747 | * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
|
| 2748 | * number of missed paths is recorded in ipath->fspath->elem_missed, otherwise,
|
| 2749 | * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
|
| 2750 | * have been needed to return all paths.
|
| 2751 |
|
| 2752 | int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
|
| 2753 | {
|
| 2754 | int ret;
|
| 2755 | int found_refs = 0;
|
| 2756 |
|
| 2757 | ret = iterate_inode_refs(inum, ipath);
|
| 2758 | if (!ret)
|
| 2759 | ++found_refs;
|
| 2760 | else if (ret != -ENOENT)
|
| 2761 | return ret;
|
| 2762 |
|
| 2763 | ret = iterate_inode_extrefs(inum, ipath);
|
| 2764 | if (ret == -ENOENT && found_refs)
|
| 2765 | return 0;
|
| 2766 |
|
| 2767 | return ret;
|
| 2768 | }
|
| 2769 |
|
| 2770 | struct btrfs_data_container *init_data_container(u32 total_bytes)
|
| 2771 | {
|
| 2772 | struct btrfs_data_container *data;
|
| 2773 | size_t alloc_bytes;
|
| 2774 |
|
| 2775 | alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
|
| 2776 | data = kvzalloc(alloc_bytes, GFP_KERNEL);
|
| 2777 | if (!data)
|
| 2778 | return ERR_PTR(-ENOMEM);
|
| 2779 |
|
| 2780 | if (total_bytes >= sizeof(*data))
|
| 2781 | data->bytes_left = total_bytes - sizeof(*data);
|
| 2782 | else
|
| 2783 | data->bytes_missing = sizeof(*data) - total_bytes;
|
| 2784 |
|
| 2785 | return data;
|
| 2786 | }
|
| 2787 |
|
| 2788 | |
| 2789 | * allocates space to return multiple file system paths for an inode.
|
| 2790 | * total_bytes to allocate are passed, note that space usable for actual path
|
| 2791 | * information will be total_bytes - sizeof(struct inode_fs_paths).
|
| 2792 | * the returned pointer must be freed with __free_inode_fs_paths() in the end.
|
| 2793 |
|
| 2794 | struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
|
| 2795 | struct btrfs_path *path)
|
| 2796 | {
|
| 2797 | struct inode_fs_paths *ifp;
|
| 2798 | struct btrfs_data_container *fspath;
|
| 2799 |
|
| 2800 | fspath = init_data_container(total_bytes);
|
| 2801 | if (IS_ERR(fspath))
|
| 2802 | return ERR_CAST(fspath);
|
| 2803 |
|
| 2804 | ifp = kmalloc_obj(*ifp);
|
| 2805 | if (!ifp) {
|
| 2806 | kvfree(fspath);
|
| 2807 | return ERR_PTR(-ENOMEM);
|
| 2808 | }
|
| 2809 |
|
| 2810 | ifp->btrfs_path = path;
|
| 2811 | ifp->fspath = fspath;
|
| 2812 | ifp->fs_root = fs_root;
|
| 2813 |
|
| 2814 | return ifp;
|
| 2815 | }
|
| 2816 |
|
| 2817 | int btrfs_backref_iter_init(struct btrfs_backref_iter *iter)
|
| 2818 | {
|
| 2819 | iter->path = btrfs_alloc_path();
|
| 2820 | if (!iter->path)
|
| 2821 | return -ENOMEM;
|
| 2822 |
|
| 2823 |
|
| 2824 | iter->path->search_commit_root = true;
|
| 2825 | iter->path->skip_locking = true;
|
| 2826 |
|
| 2827 | return 0;
|
| 2828 | }
|
| 2829 |
|
| 2830 | static void btrfs_backref_iter_release(struct btrfs_backref_iter *iter)
|
| 2831 | {
|
| 2832 | iter->bytenr = 0;
|
| 2833 | iter->item_ptr = 0;
|
| 2834 | iter->cur_ptr = 0;
|
| 2835 | iter->end_ptr = 0;
|
| 2836 | btrfs_release_path(iter->path);
|
| 2837 | memset(&iter->cur_key, 0, sizeof(iter->cur_key));
|
| 2838 | }
|
| 2839 |
|
| 2840 | int btrfs_backref_iter_start(struct btrfs_fs_info *fs_info, struct btrfs_backref_iter *iter, u64 bytenr)
|
| 2841 | {
|
| 2842 | struct btrfs_root *extent_root = btrfs_extent_root(fs_info, bytenr);
|
| 2843 | struct btrfs_path *path = iter->path;
|
| 2844 | struct btrfs_extent_item *ei;
|
| 2845 | struct btrfs_key key;
|
| 2846 | int ret;
|
| 2847 |
|
| 2848 | if (unlikely(!extent_root)) {
|
| 2849 | btrfs_err(fs_info,
|
| 2850 | "missing extent root for extent at bytenr %llu",
|
| 2851 | bytenr);
|
| 2852 | return -EUCLEAN;
|
| 2853 | }
|
| 2854 |
|
| 2855 | key.objectid = bytenr;
|
| 2856 | key.type = BTRFS_METADATA_ITEM_KEY;
|
| 2857 | key.offset = (u64)-1;
|
| 2858 | iter->bytenr = bytenr;
|
| 2859 |
|
| 2860 | ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
|
| 2861 | if (ret < 0)
|
| 2862 | return ret;
|
| 2863 | if (unlikely(ret == 0)) {
|
| 2864 | |
| 2865 | * Key with offset -1 found, there would have to exist an extent
|
| 2866 | * item with such offset, but this is out of the valid range.
|
| 2867 |
|
| 2868 | ret = -EUCLEAN;
|
| 2869 | goto release;
|
| 2870 | }
|
| 2871 | if (unlikely(path->slots[0] == 0)) {
|
| 2872 | DEBUG_WARN();
|
| 2873 | ret = -EUCLEAN;
|
| 2874 | goto release;
|
| 2875 | }
|
| 2876 | path->slots[0]--;
|
| 2877 |
|
| 2878 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
|
| 2879 | if ((key.type != BTRFS_EXTENT_ITEM_KEY &&
|
| 2880 | key.type != BTRFS_METADATA_ITEM_KEY) || key.objectid != bytenr) {
|
| 2881 | ret = -ENOENT;
|
| 2882 | goto release;
|
| 2883 | }
|
| 2884 | memcpy(&iter->cur_key, &key, sizeof(key));
|
| 2885 | iter->item_ptr = (u32)btrfs_item_ptr_offset(path->nodes[0],
|
| 2886 | path->slots[0]);
|
| 2887 | iter->end_ptr = (u32)(iter->item_ptr +
|
| 2888 | btrfs_item_size(path->nodes[0], path->slots[0]));
|
| 2889 | ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
|
| 2890 | struct btrfs_extent_item);
|
| 2891 |
|
| 2892 | |
| 2893 | * Only support iteration on tree backref yet.
|
| 2894 | *
|
| 2895 | * This is an extra precaution for non skinny-metadata, where
|
| 2896 | * EXTENT_ITEM is also used for tree blocks, that we can only use
|
| 2897 | * extent flags to determine if it's a tree block.
|
| 2898 |
|
| 2899 | if (btrfs_extent_flags(path->nodes[0], ei) & BTRFS_EXTENT_FLAG_DATA) {
|
| 2900 | ret = -ENOTSUPP;
|
| 2901 | goto release;
|
| 2902 | }
|
| 2903 | iter->cur_ptr = (u32)(iter->item_ptr + sizeof(*ei));
|
| 2904 |
|
| 2905 |
|
| 2906 | if (iter->cur_ptr >= iter->end_ptr) {
|
| 2907 | ret = btrfs_next_item(extent_root, path);
|
| 2908 |
|
| 2909 |
|
| 2910 | if (ret > 0) {
|
| 2911 | ret = -ENOENT;
|
| 2912 | goto release;
|
| 2913 | }
|
| 2914 | if (ret < 0)
|
| 2915 | goto release;
|
| 2916 |
|
| 2917 | btrfs_item_key_to_cpu(path->nodes[0], &iter->cur_key,
|
| 2918 | path->slots[0]);
|
| 2919 | if (iter->cur_key.objectid != bytenr ||
|
| 2920 | (iter->cur_key.type != BTRFS_SHARED_BLOCK_REF_KEY &&
|
| 2921 | iter->cur_key.type != BTRFS_TREE_BLOCK_REF_KEY)) {
|
| 2922 | ret = -ENOENT;
|
| 2923 | goto release;
|
| 2924 | }
|
| 2925 | iter->cur_ptr = (u32)btrfs_item_ptr_offset(path->nodes[0],
|
| 2926 | path->slots[0]);
|
| 2927 | iter->item_ptr = iter->cur_ptr;
|
| 2928 | iter->end_ptr = (u32)(iter->item_ptr + btrfs_item_size(
|
| 2929 | path->nodes[0], path->slots[0]));
|
| 2930 | }
|
| 2931 |
|
| 2932 | return 0;
|
| 2933 | release:
|
| 2934 | btrfs_backref_iter_release(iter);
|
| 2935 | return ret;
|
| 2936 | }
|
| 2937 |
|
| 2938 | static bool btrfs_backref_iter_is_inline_ref(struct btrfs_backref_iter *iter)
|
| 2939 | {
|
| 2940 | if (iter->cur_key.type == BTRFS_EXTENT_ITEM_KEY ||
|
| 2941 | iter->cur_key.type == BTRFS_METADATA_ITEM_KEY)
|
| 2942 | return true;
|
| 2943 | return false;
|
| 2944 | }
|
| 2945 |
|
| 2946 | |
| 2947 | * Go to the next backref item of current bytenr, can be either inlined or
|
| 2948 | * keyed.
|
| 2949 | *
|
| 2950 | * Caller needs to check whether it's inline ref or not by iter->cur_key.
|
| 2951 | *
|
| 2952 | * Return 0 if we get next backref without problem.
|
| 2953 | * Return >0 if there is no extra backref for this bytenr.
|
| 2954 | * Return <0 if there is something wrong happened.
|
| 2955 |
|
| 2956 | int btrfs_backref_iter_next(struct btrfs_fs_info *fs_info, struct btrfs_backref_iter *iter)
|
| 2957 | {
|
| 2958 | struct extent_buffer *eb = iter->path->nodes[0];
|
| 2959 | struct btrfs_root *extent_root;
|
| 2960 | struct btrfs_path *path = iter->path;
|
| 2961 | struct btrfs_extent_inline_ref *iref;
|
| 2962 | int ret;
|
| 2963 | u32 size;
|
| 2964 |
|
| 2965 | if (btrfs_backref_iter_is_inline_ref(iter)) {
|
| 2966 |
|
| 2967 | ASSERT(iter->cur_ptr < iter->end_ptr,
|
| 2968 | "iter->cur_ptr=%u iter->end_ptr=%u",
|
| 2969 | iter->cur_ptr, iter->end_ptr);
|
| 2970 |
|
| 2971 | if (btrfs_backref_has_tree_block_info(iter)) {
|
| 2972 |
|
| 2973 | size = sizeof(struct btrfs_tree_block_info);
|
| 2974 | } else {
|
| 2975 |
|
| 2976 | int type;
|
| 2977 |
|
| 2978 | iref = (struct btrfs_extent_inline_ref *)
|
| 2979 | ((unsigned long)iter->cur_ptr);
|
| 2980 | type = btrfs_extent_inline_ref_type(eb, iref);
|
| 2981 |
|
| 2982 | size = btrfs_extent_inline_ref_size(type);
|
| 2983 | }
|
| 2984 | iter->cur_ptr += size;
|
| 2985 | if (iter->cur_ptr < iter->end_ptr)
|
| 2986 | return 0;
|
| 2987 |
|
| 2988 |
|
| 2989 | }
|
| 2990 |
|
| 2991 |
|
| 2992 | extent_root = btrfs_extent_root(fs_info, iter->bytenr);
|
| 2993 | if (unlikely(!extent_root)) {
|
| 2994 | btrfs_err(fs_info, "missing extent root for extent at bytenr %llu",
|
| 2995 | iter->bytenr);
|
| 2996 | return -EUCLEAN;
|
| 2997 | }
|
| 2998 |
|
| 2999 | ret = btrfs_next_item(extent_root, iter->path);
|
| 3000 | if (ret)
|
| 3001 | return ret;
|
| 3002 |
|
| 3003 | btrfs_item_key_to_cpu(path->nodes[0], &iter->cur_key, path->slots[0]);
|
| 3004 | if (iter->cur_key.objectid != iter->bytenr ||
|
| 3005 | (iter->cur_key.type != BTRFS_TREE_BLOCK_REF_KEY &&
|
| 3006 | iter->cur_key.type != BTRFS_SHARED_BLOCK_REF_KEY))
|
| 3007 | return 1;
|
| 3008 | iter->item_ptr = (u32)btrfs_item_ptr_offset(path->nodes[0],
|
| 3009 | path->slots[0]);
|
| 3010 | iter->cur_ptr = iter->item_ptr;
|
| 3011 | iter->end_ptr = iter->item_ptr + (u32)btrfs_item_size(path->nodes[0],
|
| 3012 | path->slots[0]);
|
| 3013 | return 0;
|
| 3014 | }
|
| 3015 |
|
| 3016 | void btrfs_backref_init_cache(struct btrfs_fs_info *fs_info,
|
| 3017 | struct btrfs_backref_cache *cache, bool is_reloc)
|
| 3018 | {
|
| 3019 | int i;
|
| 3020 |
|
| 3021 | cache->rb_root = RB_ROOT;
|
| 3022 | for (i = 0; i < BTRFS_MAX_LEVEL; i++)
|
| 3023 | INIT_LIST_HEAD(&cache->pending[i]);
|
| 3024 | INIT_LIST_HEAD(&cache->pending_edge);
|
| 3025 | INIT_LIST_HEAD(&cache->useless_node);
|
| 3026 | cache->fs_info = fs_info;
|
| 3027 | cache->is_reloc = is_reloc;
|
| 3028 | }
|
| 3029 |
|
| 3030 | struct btrfs_backref_node *btrfs_backref_alloc_node(
|
| 3031 | struct btrfs_backref_cache *cache, u64 bytenr, int level)
|
| 3032 | {
|
| 3033 | struct btrfs_backref_node *node;
|
| 3034 |
|
| 3035 | ASSERT(level >= 0 && level < BTRFS_MAX_LEVEL, "level=%d", level);
|
| 3036 | node = kzalloc_obj(*node, GFP_NOFS);
|
| 3037 | if (!node)
|
| 3038 | return node;
|
| 3039 |
|
| 3040 | INIT_LIST_HEAD(&node->list);
|
| 3041 | INIT_LIST_HEAD(&node->upper);
|
| 3042 | INIT_LIST_HEAD(&node->lower);
|
| 3043 | RB_CLEAR_NODE(&node->rb_node);
|
| 3044 | cache->nr_nodes++;
|
| 3045 | node->level = level;
|
| 3046 | node->bytenr = bytenr;
|
| 3047 |
|
| 3048 | return node;
|
| 3049 | }
|
| 3050 |
|
| 3051 | void btrfs_backref_free_node(struct btrfs_backref_cache *cache,
|
| 3052 | struct btrfs_backref_node *node)
|
| 3053 | {
|
| 3054 | if (node) {
|
| 3055 | ASSERT(list_empty(&node->list));
|
| 3056 | ASSERT(list_empty(&node->lower));
|
| 3057 | ASSERT(node->eb == NULL, "node->eb->start=%llu level=%d owner=%llu",
|
| 3058 | node->eb ? node->eb->start : 0,
|
| 3059 | node->eb ? btrfs_header_level(node->eb) : 0,
|
| 3060 | node->eb ? btrfs_header_owner(node->eb) : 0);
|
| 3061 | cache->nr_nodes--;
|
| 3062 | btrfs_put_root(node->root);
|
| 3063 | kfree(node);
|
| 3064 | }
|
| 3065 | }
|
| 3066 |
|
| 3067 | struct btrfs_backref_edge *btrfs_backref_alloc_edge(
|
| 3068 | struct btrfs_backref_cache *cache)
|
| 3069 | {
|
| 3070 | struct btrfs_backref_edge *edge;
|
| 3071 |
|
| 3072 | edge = kzalloc_obj(*edge, GFP_NOFS);
|
| 3073 | if (edge)
|
| 3074 | cache->nr_edges++;
|
| 3075 | return edge;
|
| 3076 | }
|
| 3077 |
|
| 3078 | void btrfs_backref_free_edge(struct btrfs_backref_cache *cache,
|
| 3079 | struct btrfs_backref_edge *edge)
|
| 3080 | {
|
| 3081 | if (edge) {
|
| 3082 | cache->nr_edges--;
|
| 3083 | kfree(edge);
|
| 3084 | }
|
| 3085 | }
|
| 3086 |
|
| 3087 | void btrfs_backref_unlock_node_buffer(struct btrfs_backref_node *node)
|
| 3088 | {
|
| 3089 | if (node->locked) {
|
| 3090 | btrfs_tree_unlock(node->eb);
|
| 3091 | node->locked = 0;
|
| 3092 | }
|
| 3093 | }
|
| 3094 |
|
| 3095 | void btrfs_backref_drop_node_buffer(struct btrfs_backref_node *node)
|
| 3096 | {
|
| 3097 | if (node->eb) {
|
| 3098 | btrfs_backref_unlock_node_buffer(node);
|
| 3099 | free_extent_buffer(node->eb);
|
| 3100 | node->eb = NULL;
|
| 3101 | }
|
| 3102 | }
|
| 3103 |
|
| 3104 | |
| 3105 | * Drop the backref node from cache without cleaning up its children
|
| 3106 | * edges.
|
| 3107 | *
|
| 3108 | * This can only be called on node without parent edges.
|
| 3109 | * The children edges are still kept as is.
|
| 3110 |
|
| 3111 | void btrfs_backref_drop_node(struct btrfs_backref_cache *tree,
|
| 3112 | struct btrfs_backref_node *node)
|
| 3113 | {
|
| 3114 | ASSERT(list_empty(&node->upper));
|
| 3115 |
|
| 3116 | btrfs_backref_drop_node_buffer(node);
|
| 3117 | list_del_init(&node->list);
|
| 3118 | list_del_init(&node->lower);
|
| 3119 | if (!RB_EMPTY_NODE(&node->rb_node))
|
| 3120 | rb_erase(&node->rb_node, &tree->rb_root);
|
| 3121 | btrfs_backref_free_node(tree, node);
|
| 3122 | }
|
| 3123 |
|
| 3124 | |
| 3125 | * Drop the backref node from cache, also cleaning up all its
|
| 3126 | * upper edges and any uncached nodes in the path.
|
| 3127 | *
|
| 3128 | * This cleanup happens bottom up, thus the node should either
|
| 3129 | * be the lowest node in the cache or a detached node.
|
| 3130 |
|
| 3131 | void btrfs_backref_cleanup_node(struct btrfs_backref_cache *cache,
|
| 3132 | struct btrfs_backref_node *node)
|
| 3133 | {
|
| 3134 | struct btrfs_backref_edge *edge;
|
| 3135 |
|
| 3136 | if (!node)
|
| 3137 | return;
|
| 3138 |
|
| 3139 | while (!list_empty(&node->upper)) {
|
| 3140 | edge = list_first_entry(&node->upper, struct btrfs_backref_edge,
|
| 3141 | list[LOWER]);
|
| 3142 | list_del(&edge->list[LOWER]);
|
| 3143 | list_del(&edge->list[UPPER]);
|
| 3144 | btrfs_backref_free_edge(cache, edge);
|
| 3145 | }
|
| 3146 |
|
| 3147 | btrfs_backref_drop_node(cache, node);
|
| 3148 | }
|
| 3149 |
|
| 3150 | |
| 3151 | * Release all nodes/edges from current cache
|
| 3152 |
|
| 3153 | void btrfs_backref_release_cache(struct btrfs_backref_cache *cache)
|
| 3154 | {
|
| 3155 | struct btrfs_backref_node *node;
|
| 3156 |
|
| 3157 | while ((node = rb_entry_safe(rb_first(&cache->rb_root),
|
| 3158 | struct btrfs_backref_node, rb_node)))
|
| 3159 | btrfs_backref_cleanup_node(cache, node);
|
| 3160 |
|
| 3161 | ASSERT(list_empty(&cache->pending_edge));
|
| 3162 | ASSERT(list_empty(&cache->useless_node));
|
| 3163 | ASSERT(!cache->nr_nodes, "cache->nr_nodes=%d", cache->nr_nodes);
|
| 3164 | ASSERT(!cache->nr_edges, "cache->nr_edges=%d", cache->nr_edges);
|
| 3165 | }
|
| 3166 |
|
| 3167 | static void btrfs_backref_link_edge(struct btrfs_backref_edge *edge,
|
| 3168 | struct btrfs_backref_node *lower,
|
| 3169 | struct btrfs_backref_node *upper)
|
| 3170 | {
|
| 3171 | ASSERT(upper != NULL);
|
| 3172 | ASSERT(lower != NULL);
|
| 3173 | ASSERT(upper->level == lower->level + 1, "upper->level=%d lower->level=%d",
|
| 3174 | upper->level, lower->level);
|
| 3175 | edge->node[LOWER] = lower;
|
| 3176 | edge->node[UPPER] = upper;
|
| 3177 | list_add_tail(&edge->list[LOWER], &lower->upper);
|
| 3178 | }
|
| 3179 | |
| 3180 | * Handle direct tree backref
|
| 3181 | *
|
| 3182 | * Direct tree backref means, the backref item shows its parent bytenr
|
| 3183 | * directly. This is for SHARED_BLOCK_REF backref (keyed or inlined).
|
| 3184 | *
|
| 3185 | * @ref_key: The converted backref key.
|
| 3186 | * For keyed backref, it's the item key.
|
| 3187 | * For inlined backref, objectid is the bytenr,
|
| 3188 | * type is btrfs_inline_ref_type, offset is
|
| 3189 | * btrfs_inline_ref_offset.
|
| 3190 |
|
| 3191 | static int handle_direct_tree_backref(struct btrfs_backref_cache *cache,
|
| 3192 | struct btrfs_key *ref_key,
|
| 3193 | struct btrfs_backref_node *cur)
|
| 3194 | {
|
| 3195 | struct btrfs_backref_edge *edge;
|
| 3196 | struct btrfs_backref_node *upper;
|
| 3197 | struct rb_node *rb_node;
|
| 3198 |
|
| 3199 | ASSERT(ref_key->type == BTRFS_SHARED_BLOCK_REF_KEY, "ref_key->type=%hhu", ref_key->type);
|
| 3200 |
|
| 3201 |
|
| 3202 | if (ref_key->objectid == ref_key->offset) {
|
| 3203 | struct btrfs_root *root;
|
| 3204 |
|
| 3205 | cur->is_reloc_root = 1;
|
| 3206 |
|
| 3207 | if (cache->is_reloc) {
|
| 3208 | root = find_reloc_root(cache->fs_info, cur->bytenr);
|
| 3209 | if (!root)
|
| 3210 | return -ENOENT;
|
| 3211 | cur->root = root;
|
| 3212 | } else {
|
| 3213 | |
| 3214 | * For generic purpose backref cache, reloc root node
|
| 3215 | * is useless.
|
| 3216 |
|
| 3217 | list_add(&cur->list, &cache->useless_node);
|
| 3218 | }
|
| 3219 | return 0;
|
| 3220 | }
|
| 3221 |
|
| 3222 | edge = btrfs_backref_alloc_edge(cache);
|
| 3223 | if (!edge)
|
| 3224 | return -ENOMEM;
|
| 3225 |
|
| 3226 | rb_node = rb_simple_search(&cache->rb_root, ref_key->offset);
|
| 3227 | if (!rb_node) {
|
| 3228 |
|
| 3229 | upper = btrfs_backref_alloc_node(cache, ref_key->offset,
|
| 3230 | cur->level + 1);
|
| 3231 | if (!upper) {
|
| 3232 | btrfs_backref_free_edge(cache, edge);
|
| 3233 | return -ENOMEM;
|
| 3234 | }
|
| 3235 |
|
| 3236 | |
| 3237 | * Backrefs for the upper level block isn't cached, add the
|
| 3238 | * block to pending list
|
| 3239 |
|
| 3240 | list_add_tail(&edge->list[UPPER], &cache->pending_edge);
|
| 3241 | } else {
|
| 3242 |
|
| 3243 | upper = rb_entry(rb_node, struct btrfs_backref_node, rb_node);
|
| 3244 | ASSERT(upper->checked);
|
| 3245 | INIT_LIST_HEAD(&edge->list[UPPER]);
|
| 3246 | }
|
| 3247 | btrfs_backref_link_edge(edge, cur, upper);
|
| 3248 | return 0;
|
| 3249 | }
|
| 3250 |
|
| 3251 | |
| 3252 | * Handle indirect tree backref
|
| 3253 | *
|
| 3254 | * Indirect tree backref means, we only know which tree the node belongs to.
|
| 3255 | * We still need to do a tree search to find out the parents. This is for
|
| 3256 | * TREE_BLOCK_REF backref (keyed or inlined).
|
| 3257 | *
|
| 3258 | * @trans: Transaction handle.
|
| 3259 | * @ref_key: The same as @ref_key in handle_direct_tree_backref()
|
| 3260 | * @tree_key: The first key of this tree block.
|
| 3261 | * @path: A clean (released) path, to avoid allocating path every time
|
| 3262 | * the function get called.
|
| 3263 |
|
| 3264 | static int handle_indirect_tree_backref(struct btrfs_trans_handle *trans,
|
| 3265 | struct btrfs_backref_cache *cache,
|
| 3266 | struct btrfs_path *path,
|
| 3267 | struct btrfs_key *ref_key,
|
| 3268 | struct btrfs_key *tree_key,
|
| 3269 | struct btrfs_backref_node *cur)
|
| 3270 | {
|
| 3271 | struct btrfs_fs_info *fs_info = cache->fs_info;
|
| 3272 | struct btrfs_backref_node *upper;
|
| 3273 | struct btrfs_backref_node *lower;
|
| 3274 | struct btrfs_backref_edge *edge;
|
| 3275 | struct extent_buffer *eb;
|
| 3276 | struct btrfs_root *root;
|
| 3277 | struct rb_node *rb_node;
|
| 3278 | int level;
|
| 3279 | bool need_check = true;
|
| 3280 | int ret;
|
| 3281 |
|
| 3282 | root = btrfs_get_fs_root(fs_info, ref_key->offset, false);
|
| 3283 | if (IS_ERR(root))
|
| 3284 | return PTR_ERR(root);
|
| 3285 |
|
| 3286 |
|
| 3287 | if (unlikely(!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))) {
|
| 3288 | btrfs_put_root(root);
|
| 3289 | return -EUCLEAN;
|
| 3290 | }
|
| 3291 |
|
| 3292 | if (btrfs_root_level(&root->root_item) == cur->level) {
|
| 3293 |
|
| 3294 | ASSERT(btrfs_root_bytenr(&root->root_item) == cur->bytenr,
|
| 3295 | "root_bytenr=%llu cur->bytenr=%llu",
|
| 3296 | btrfs_root_bytenr(&root->root_item), cur->bytenr);
|
| 3297 | |
| 3298 | * For reloc backref cache, we may ignore reloc root. But for
|
| 3299 | * general purpose backref cache, we can't rely on
|
| 3300 | * btrfs_should_ignore_reloc_root() as it may conflict with
|
| 3301 | * current running relocation and lead to missing root.
|
| 3302 | *
|
| 3303 | * For general purpose backref cache, reloc root detection is
|
| 3304 | * completely relying on direct backref (key->offset is parent
|
| 3305 | * bytenr), thus only do such check for reloc cache.
|
| 3306 |
|
| 3307 | if (btrfs_should_ignore_reloc_root(root) && cache->is_reloc) {
|
| 3308 | btrfs_put_root(root);
|
| 3309 | list_add(&cur->list, &cache->useless_node);
|
| 3310 | } else {
|
| 3311 | cur->root = root;
|
| 3312 | }
|
| 3313 | return 0;
|
| 3314 | }
|
| 3315 |
|
| 3316 | level = cur->level + 1;
|
| 3317 |
|
| 3318 |
|
| 3319 | path->search_commit_root = true;
|
| 3320 | path->skip_locking = true;
|
| 3321 | path->lowest_level = level;
|
| 3322 | ret = btrfs_search_slot(NULL, root, tree_key, path, 0, 0);
|
| 3323 | path->lowest_level = 0;
|
| 3324 | if (ret < 0) {
|
| 3325 | btrfs_put_root(root);
|
| 3326 | return ret;
|
| 3327 | }
|
| 3328 | if (ret > 0 && path->slots[level] > 0)
|
| 3329 | path->slots[level]--;
|
| 3330 |
|
| 3331 | eb = path->nodes[level];
|
| 3332 | if (btrfs_node_blockptr(eb, path->slots[level]) != cur->bytenr) {
|
| 3333 | btrfs_err(fs_info,
|
| 3334 | "couldn't find block (%llu) (level %d) in tree (%llu) with key " BTRFS_KEY_FMT,
|
| 3335 | cur->bytenr, level - 1, btrfs_root_id(root),
|
| 3336 | BTRFS_KEY_FMT_VALUE(tree_key));
|
| 3337 | btrfs_put_root(root);
|
| 3338 | ret = -ENOENT;
|
| 3339 | goto out;
|
| 3340 | }
|
| 3341 | lower = cur;
|
| 3342 |
|
| 3343 |
|
| 3344 | for (; level < BTRFS_MAX_LEVEL; level++) {
|
| 3345 | if (!path->nodes[level]) {
|
| 3346 | ASSERT(btrfs_root_bytenr(&root->root_item) == lower->bytenr,
|
| 3347 | "root_bytenr=%llu lower->bytenr=%llu",
|
| 3348 | btrfs_root_bytenr(&root->root_item), lower->bytenr);
|
| 3349 |
|
| 3350 | if (btrfs_should_ignore_reloc_root(root) &&
|
| 3351 | cache->is_reloc) {
|
| 3352 | btrfs_put_root(root);
|
| 3353 | list_add(&lower->list, &cache->useless_node);
|
| 3354 | } else {
|
| 3355 | lower->root = root;
|
| 3356 | }
|
| 3357 | break;
|
| 3358 | }
|
| 3359 |
|
| 3360 | edge = btrfs_backref_alloc_edge(cache);
|
| 3361 | if (!edge) {
|
| 3362 | btrfs_put_root(root);
|
| 3363 | ret = -ENOMEM;
|
| 3364 | goto out;
|
| 3365 | }
|
| 3366 |
|
| 3367 | eb = path->nodes[level];
|
| 3368 | rb_node = rb_simple_search(&cache->rb_root, eb->start);
|
| 3369 | if (!rb_node) {
|
| 3370 | upper = btrfs_backref_alloc_node(cache, eb->start,
|
| 3371 | lower->level + 1);
|
| 3372 | if (!upper) {
|
| 3373 | btrfs_put_root(root);
|
| 3374 | btrfs_backref_free_edge(cache, edge);
|
| 3375 | ret = -ENOMEM;
|
| 3376 | goto out;
|
| 3377 | }
|
| 3378 | upper->owner = btrfs_header_owner(eb);
|
| 3379 |
|
| 3380 |
|
| 3381 | if (unlikely(!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))) {
|
| 3382 | btrfs_put_root(root);
|
| 3383 | btrfs_backref_free_edge(cache, edge);
|
| 3384 | btrfs_backref_free_node(cache, upper);
|
| 3385 | ret = -EUCLEAN;
|
| 3386 | goto out;
|
| 3387 | }
|
| 3388 |
|
| 3389 | |
| 3390 | * If we know the block isn't shared we can avoid
|
| 3391 | * checking its backrefs.
|
| 3392 |
|
| 3393 | if (btrfs_block_can_be_shared(trans, root, eb))
|
| 3394 | upper->checked = 0;
|
| 3395 | else
|
| 3396 | upper->checked = 1;
|
| 3397 |
|
| 3398 | |
| 3399 | * Add the block to pending list if we need to check its
|
| 3400 | * backrefs, we only do this once while walking up a
|
| 3401 | * tree as we will catch anything else later on.
|
| 3402 |
|
| 3403 | if (!upper->checked && need_check) {
|
| 3404 | need_check = false;
|
| 3405 | list_add_tail(&edge->list[UPPER],
|
| 3406 | &cache->pending_edge);
|
| 3407 | } else {
|
| 3408 | if (upper->checked)
|
| 3409 | need_check = true;
|
| 3410 | INIT_LIST_HEAD(&edge->list[UPPER]);
|
| 3411 | }
|
| 3412 | } else {
|
| 3413 | upper = rb_entry(rb_node, struct btrfs_backref_node,
|
| 3414 | rb_node);
|
| 3415 | ASSERT(upper->checked);
|
| 3416 | INIT_LIST_HEAD(&edge->list[UPPER]);
|
| 3417 | if (!upper->owner)
|
| 3418 | upper->owner = btrfs_header_owner(eb);
|
| 3419 | }
|
| 3420 | btrfs_backref_link_edge(edge, lower, upper);
|
| 3421 |
|
| 3422 | if (rb_node) {
|
| 3423 | btrfs_put_root(root);
|
| 3424 | break;
|
| 3425 | }
|
| 3426 | lower = upper;
|
| 3427 | upper = NULL;
|
| 3428 | }
|
| 3429 | out:
|
| 3430 | btrfs_release_path(path);
|
| 3431 | return ret;
|
| 3432 | }
|
| 3433 |
|
| 3434 | |
| 3435 | * Add backref node @cur into @cache.
|
| 3436 | *
|
| 3437 | * NOTE: Even if the function returned 0, @cur is not yet cached as its upper
|
| 3438 | * links aren't yet bi-directional. Needs to finish such links.
|
| 3439 | * Use btrfs_backref_finish_upper_links() to finish such linkage.
|
| 3440 | *
|
| 3441 | * @trans: Transaction handle.
|
| 3442 | * @path: Released path for indirect tree backref lookup
|
| 3443 | * @iter: Released backref iter for extent tree search
|
| 3444 | * @node_key: The first key of the tree block
|
| 3445 |
|
| 3446 | int btrfs_backref_add_tree_node(struct btrfs_trans_handle *trans,
|
| 3447 | struct btrfs_backref_cache *cache,
|
| 3448 | struct btrfs_path *path,
|
| 3449 | struct btrfs_backref_iter *iter,
|
| 3450 | struct btrfs_key *node_key,
|
| 3451 | struct btrfs_backref_node *cur)
|
| 3452 | {
|
| 3453 | struct btrfs_backref_edge *edge;
|
| 3454 | struct btrfs_backref_node *exist;
|
| 3455 | int ret;
|
| 3456 |
|
| 3457 | ret = btrfs_backref_iter_start(trans->fs_info, iter, cur->bytenr);
|
| 3458 | if (ret < 0)
|
| 3459 | return ret;
|
| 3460 | |
| 3461 | * We skip the first btrfs_tree_block_info, as we don't use the key
|
| 3462 | * stored in it, but fetch it from the tree block
|
| 3463 |
|
| 3464 | if (btrfs_backref_has_tree_block_info(iter)) {
|
| 3465 | ret = btrfs_backref_iter_next(trans->fs_info, iter);
|
| 3466 | if (ret < 0)
|
| 3467 | goto out;
|
| 3468 |
|
| 3469 | if (unlikely(ret > 0)) {
|
| 3470 | ret = -EUCLEAN;
|
| 3471 | goto out;
|
| 3472 | }
|
| 3473 | }
|
| 3474 | WARN_ON(cur->checked);
|
| 3475 | if (!list_empty(&cur->upper)) {
|
| 3476 | |
| 3477 | * The backref was added previously when processing backref of
|
| 3478 | * type BTRFS_TREE_BLOCK_REF_KEY
|
| 3479 |
|
| 3480 | ASSERT(list_is_singular(&cur->upper));
|
| 3481 | edge = list_first_entry(&cur->upper, struct btrfs_backref_edge,
|
| 3482 | list[LOWER]);
|
| 3483 | ASSERT(list_empty(&edge->list[UPPER]));
|
| 3484 | exist = edge->node[UPPER];
|
| 3485 | |
| 3486 | * Add the upper level block to pending list if we need check
|
| 3487 | * its backrefs
|
| 3488 |
|
| 3489 | if (!exist->checked)
|
| 3490 | list_add_tail(&edge->list[UPPER], &cache->pending_edge);
|
| 3491 | } else {
|
| 3492 | exist = NULL;
|
| 3493 | }
|
| 3494 |
|
| 3495 | for (; ret == 0; ret = btrfs_backref_iter_next(trans->fs_info, iter)) {
|
| 3496 | struct extent_buffer *eb;
|
| 3497 | struct btrfs_key key;
|
| 3498 | int type;
|
| 3499 |
|
| 3500 | cond_resched();
|
| 3501 | eb = iter->path->nodes[0];
|
| 3502 |
|
| 3503 | key.objectid = iter->bytenr;
|
| 3504 | if (btrfs_backref_iter_is_inline_ref(iter)) {
|
| 3505 | struct btrfs_extent_inline_ref *iref;
|
| 3506 |
|
| 3507 |
|
| 3508 | iref = (struct btrfs_extent_inline_ref *)
|
| 3509 | ((unsigned long)iter->cur_ptr);
|
| 3510 | type = btrfs_get_extent_inline_ref_type(eb, iref,
|
| 3511 | BTRFS_REF_TYPE_BLOCK);
|
| 3512 | if (unlikely(type == BTRFS_REF_TYPE_INVALID)) {
|
| 3513 | ret = -EUCLEAN;
|
| 3514 | goto out;
|
| 3515 | }
|
| 3516 | key.type = type;
|
| 3517 | key.offset = btrfs_extent_inline_ref_offset(eb, iref);
|
| 3518 | } else {
|
| 3519 | key.type = iter->cur_key.type;
|
| 3520 | key.offset = iter->cur_key.offset;
|
| 3521 | }
|
| 3522 |
|
| 3523 | |
| 3524 | * Parent node found and matches current inline ref, no need to
|
| 3525 | * rebuild this node for this inline ref
|
| 3526 |
|
| 3527 | if (exist &&
|
| 3528 | ((key.type == BTRFS_TREE_BLOCK_REF_KEY &&
|
| 3529 | exist->owner == key.offset) ||
|
| 3530 | (key.type == BTRFS_SHARED_BLOCK_REF_KEY &&
|
| 3531 | exist->bytenr == key.offset))) {
|
| 3532 | exist = NULL;
|
| 3533 | continue;
|
| 3534 | }
|
| 3535 |
|
| 3536 |
|
| 3537 | if (key.type == BTRFS_SHARED_BLOCK_REF_KEY) {
|
| 3538 | ret = handle_direct_tree_backref(cache, &key, cur);
|
| 3539 | if (ret < 0)
|
| 3540 | goto out;
|
| 3541 | } else if (key.type == BTRFS_TREE_BLOCK_REF_KEY) {
|
| 3542 | |
| 3543 | * key.type == BTRFS_TREE_BLOCK_REF_KEY, inline ref
|
| 3544 | * offset means the root objectid. We need to search
|
| 3545 | * the tree to get its parent bytenr.
|
| 3546 |
|
| 3547 | ret = handle_indirect_tree_backref(trans, cache, path,
|
| 3548 | &key, node_key, cur);
|
| 3549 | if (ret < 0)
|
| 3550 | goto out;
|
| 3551 | }
|
| 3552 | |
| 3553 | * Unrecognized tree backref items (if it can pass tree-checker)
|
| 3554 | * would be ignored.
|
| 3555 |
|
| 3556 | }
|
| 3557 | ret = 0;
|
| 3558 | cur->checked = 1;
|
| 3559 | WARN_ON(exist);
|
| 3560 | out:
|
| 3561 | btrfs_backref_iter_release(iter);
|
| 3562 | return ret;
|
| 3563 | }
|
| 3564 |
|
| 3565 | |
| 3566 | * Finish the upwards linkage created by btrfs_backref_add_tree_node()
|
| 3567 |
|
| 3568 | int btrfs_backref_finish_upper_links(struct btrfs_backref_cache *cache,
|
| 3569 | struct btrfs_backref_node *start)
|
| 3570 | {
|
| 3571 | struct list_head *useless_node = &cache->useless_node;
|
| 3572 | struct btrfs_backref_edge *edge;
|
| 3573 | struct rb_node *rb_node;
|
| 3574 | LIST_HEAD(pending_edge);
|
| 3575 |
|
| 3576 | ASSERT(start->checked);
|
| 3577 |
|
| 3578 | rb_node = rb_simple_insert(&cache->rb_root, &start->simple_node);
|
| 3579 | if (rb_node)
|
| 3580 | btrfs_backref_panic(cache->fs_info, start->bytenr, -EEXIST);
|
| 3581 |
|
| 3582 | |
| 3583 | * Use breadth first search to iterate all related edges.
|
| 3584 | *
|
| 3585 | * The starting points are all the edges of this node
|
| 3586 |
|
| 3587 | list_for_each_entry(edge, &start->upper, list[LOWER])
|
| 3588 | list_add_tail(&edge->list[UPPER], &pending_edge);
|
| 3589 |
|
| 3590 | while (!list_empty(&pending_edge)) {
|
| 3591 | struct btrfs_backref_node *upper;
|
| 3592 | struct btrfs_backref_node *lower;
|
| 3593 |
|
| 3594 | edge = list_first_entry(&pending_edge,
|
| 3595 | struct btrfs_backref_edge, list[UPPER]);
|
| 3596 | list_del_init(&edge->list[UPPER]);
|
| 3597 | upper = edge->node[UPPER];
|
| 3598 | lower = edge->node[LOWER];
|
| 3599 |
|
| 3600 |
|
| 3601 | if (upper->detached) {
|
| 3602 | list_del(&edge->list[LOWER]);
|
| 3603 | btrfs_backref_free_edge(cache, edge);
|
| 3604 |
|
| 3605 |
|
| 3606 | if (list_empty(&lower->upper))
|
| 3607 | list_add(&lower->list, useless_node);
|
| 3608 | continue;
|
| 3609 | }
|
| 3610 |
|
| 3611 | |
| 3612 | * All new nodes added in current build_backref_tree() haven't
|
| 3613 | * been linked to the cache rb tree.
|
| 3614 | * So if we have upper->rb_node populated, this means a cache
|
| 3615 | * hit. We only need to link the edge, as @upper and all its
|
| 3616 | * parents have already been linked.
|
| 3617 |
|
| 3618 | if (!RB_EMPTY_NODE(&upper->rb_node)) {
|
| 3619 | list_add_tail(&edge->list[UPPER], &upper->lower);
|
| 3620 | continue;
|
| 3621 | }
|
| 3622 |
|
| 3623 |
|
| 3624 | if (unlikely(!upper->checked)) {
|
| 3625 | DEBUG_WARN("we should not have any unchecked nodes");
|
| 3626 | return -EUCLEAN;
|
| 3627 | }
|
| 3628 |
|
| 3629 | rb_node = rb_simple_insert(&cache->rb_root, &upper->simple_node);
|
| 3630 | if (unlikely(rb_node))
|
| 3631 | btrfs_backref_panic(cache->fs_info, upper->bytenr, -EEXIST);
|
| 3632 |
|
| 3633 | list_add_tail(&edge->list[UPPER], &upper->lower);
|
| 3634 |
|
| 3635 | |
| 3636 | * Also queue all the parent edges of this uncached node
|
| 3637 | * to finish the upper linkage
|
| 3638 |
|
| 3639 | list_for_each_entry(edge, &upper->upper, list[LOWER])
|
| 3640 | list_add_tail(&edge->list[UPPER], &pending_edge);
|
| 3641 | }
|
| 3642 | return 0;
|
| 3643 | }
|
| 3644 |
|
| 3645 | void btrfs_backref_error_cleanup(struct btrfs_backref_cache *cache,
|
| 3646 | struct btrfs_backref_node *node)
|
| 3647 | {
|
| 3648 | struct btrfs_backref_node *lower;
|
| 3649 | struct btrfs_backref_node *upper;
|
| 3650 | struct btrfs_backref_edge *edge;
|
| 3651 |
|
| 3652 | while (!list_empty(&cache->useless_node)) {
|
| 3653 | lower = list_first_entry(&cache->useless_node,
|
| 3654 | struct btrfs_backref_node, list);
|
| 3655 | list_del_init(&lower->list);
|
| 3656 | }
|
| 3657 | while (!list_empty(&cache->pending_edge)) {
|
| 3658 | edge = list_first_entry(&cache->pending_edge,
|
| 3659 | struct btrfs_backref_edge, list[UPPER]);
|
| 3660 | list_del(&edge->list[UPPER]);
|
| 3661 | list_del(&edge->list[LOWER]);
|
| 3662 | lower = edge->node[LOWER];
|
| 3663 | upper = edge->node[UPPER];
|
| 3664 | btrfs_backref_free_edge(cache, edge);
|
| 3665 |
|
| 3666 | |
| 3667 | * Lower is no longer linked to any upper backref nodes and
|
| 3668 | * isn't in the cache, we can free it ourselves.
|
| 3669 |
|
| 3670 | if (list_empty(&lower->upper) &&
|
| 3671 | RB_EMPTY_NODE(&lower->rb_node))
|
| 3672 | list_add(&lower->list, &cache->useless_node);
|
| 3673 |
|
| 3674 | if (!RB_EMPTY_NODE(&upper->rb_node))
|
| 3675 | continue;
|
| 3676 |
|
| 3677 |
|
| 3678 | list_for_each_entry(edge, &upper->upper, list[LOWER])
|
| 3679 | list_add_tail(&edge->list[UPPER],
|
| 3680 | &cache->pending_edge);
|
| 3681 | if (list_empty(&upper->upper))
|
| 3682 | list_add(&upper->list, &cache->useless_node);
|
| 3683 | }
|
| 3684 |
|
| 3685 | while (!list_empty(&cache->useless_node)) {
|
| 3686 | lower = list_first_entry(&cache->useless_node,
|
| 3687 | struct btrfs_backref_node, list);
|
| 3688 | list_del_init(&lower->list);
|
| 3689 | if (lower == node)
|
| 3690 | node = NULL;
|
| 3691 | btrfs_backref_drop_node(cache, lower);
|
| 3692 | }
|
| 3693 |
|
| 3694 | btrfs_backref_cleanup_node(cache, node);
|
| 3695 | ASSERT(list_empty(&cache->useless_node) &&
|
| 3696 | list_empty(&cache->pending_edge));
|
| 3697 | }
|