1 // SPDX-License-Identifier: GPL-2.0 2 3 #include "bcachefs.h" 4 #include "backpointers.h" 5 #include "bkey_methods.h" 6 #include "btree_types.h" 7 #include "alloc_background.h" 8 #include "dirent.h" 9 #include "ec.h" 10 #include "error.h" 11 #include "extents.h" 12 #include "inode.h" 13 #include "lru.h" 14 #include "quota.h" 15 #include "reflink.h" 16 #include "subvolume.h" 17 #include "xattr.h" 18 19 const char * const bch2_bkey_types[] = { 20 #define x(name, nr) #name, 21 BCH_BKEY_TYPES() 22 #undef x 23 NULL 24 }; 25 26 static int deleted_key_invalid(const struct bch_fs *c, struct bkey_s_c k, 27 unsigned flags, struct printbuf *err) 28 { 29 return 0; 30 } 31 32 #define bch2_bkey_ops_deleted ((struct bkey_ops) { \ 33 .key_invalid = deleted_key_invalid, \ 34 }) 35 36 #define bch2_bkey_ops_whiteout ((struct bkey_ops) { \ 37 .key_invalid = deleted_key_invalid, \ 38 }) 39 40 static int empty_val_key_invalid(const struct bch_fs *c, struct bkey_s_c k, 41 unsigned flags, struct printbuf *err) 42 { 43 if (bkey_val_bytes(k.k)) { 44 prt_printf(err, "incorrect value size (%zu != 0)", 45 bkey_val_bytes(k.k)); 46 return -BCH_ERR_invalid_bkey; 47 } 48 49 return 0; 50 } 51 52 #define bch2_bkey_ops_error ((struct bkey_ops) { \ 53 .key_invalid = empty_val_key_invalid, \ 54 }) 55 56 static int key_type_cookie_invalid(const struct bch_fs *c, struct bkey_s_c k, 57 unsigned flags, struct printbuf *err) 58 { 59 if (bkey_val_bytes(k.k) != sizeof(struct bch_cookie)) { 60 prt_printf(err, "incorrect value size (%zu != %zu)", 61 bkey_val_bytes(k.k), sizeof(struct bch_cookie)); 62 return -BCH_ERR_invalid_bkey; 63 } 64 65 return 0; 66 } 67 68 #define bch2_bkey_ops_cookie ((struct bkey_ops) { \ 69 .key_invalid = key_type_cookie_invalid, \ 70 }) 71 72 #define bch2_bkey_ops_hash_whiteout ((struct bkey_ops) {\ 73 .key_invalid = empty_val_key_invalid, \ 74 }) 75 76 static int key_type_inline_data_invalid(const struct bch_fs *c, struct bkey_s_c k, 77 unsigned flags, struct printbuf *err) 78 { 79 return 0; 80 } 81 82 static void key_type_inline_data_to_text(struct printbuf *out, struct bch_fs *c, 83 struct bkey_s_c k) 84 { 85 struct bkey_s_c_inline_data d = bkey_s_c_to_inline_data(k); 86 unsigned datalen = bkey_inline_data_bytes(k.k); 87 88 prt_printf(out, "datalen %u: %*phN", 89 datalen, min(datalen, 32U), d.v->data); 90 } 91 92 #define bch2_bkey_ops_inline_data ((struct bkey_ops) { \ 93 .key_invalid = key_type_inline_data_invalid, \ 94 .val_to_text = key_type_inline_data_to_text, \ 95 }) 96 97 static int key_type_set_invalid(const struct bch_fs *c, struct bkey_s_c k, 98 unsigned flags, struct printbuf *err) 99 { 100 if (bkey_val_bytes(k.k)) { 101 prt_printf(err, "incorrect value size (%zu != %zu)", 102 bkey_val_bytes(k.k), sizeof(struct bch_cookie)); 103 return -BCH_ERR_invalid_bkey; 104 } 105 106 return 0; 107 } 108 109 static bool key_type_set_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r) 110 { 111 bch2_key_resize(l.k, l.k->size + r.k->size); 112 return true; 113 } 114 115 #define bch2_bkey_ops_set ((struct bkey_ops) { \ 116 .key_invalid = key_type_set_invalid, \ 117 .key_merge = key_type_set_merge, \ 118 }) 119 120 const struct bkey_ops bch2_bkey_ops[] = { 121 #define x(name, nr) [KEY_TYPE_##name] = bch2_bkey_ops_##name, 122 BCH_BKEY_TYPES() 123 #undef x 124 }; 125 126 int bch2_bkey_val_invalid(struct bch_fs *c, struct bkey_s_c k, 127 unsigned flags, struct printbuf *err) 128 { 129 if (k.k->type >= KEY_TYPE_MAX) { 130 prt_printf(err, "invalid type (%u >= %u)", k.k->type, KEY_TYPE_MAX); 131 return -BCH_ERR_invalid_bkey; 132 } 133 134 return bch2_bkey_ops[k.k->type].key_invalid(c, k, flags, err); 135 } 136 137 static unsigned bch2_key_types_allowed[] = { 138 [BKEY_TYPE_extents] = 139 (1U << KEY_TYPE_deleted)| 140 (1U << KEY_TYPE_whiteout)| 141 (1U << KEY_TYPE_error)| 142 (1U << KEY_TYPE_cookie)| 143 (1U << KEY_TYPE_extent)| 144 (1U << KEY_TYPE_reservation)| 145 (1U << KEY_TYPE_reflink_p)| 146 (1U << KEY_TYPE_inline_data), 147 [BKEY_TYPE_inodes] = 148 (1U << KEY_TYPE_deleted)| 149 (1U << KEY_TYPE_whiteout)| 150 (1U << KEY_TYPE_inode)| 151 (1U << KEY_TYPE_inode_v2)| 152 (1U << KEY_TYPE_inode_v3)| 153 (1U << KEY_TYPE_inode_generation), 154 [BKEY_TYPE_dirents] = 155 (1U << KEY_TYPE_deleted)| 156 (1U << KEY_TYPE_whiteout)| 157 (1U << KEY_TYPE_hash_whiteout)| 158 (1U << KEY_TYPE_dirent), 159 [BKEY_TYPE_xattrs] = 160 (1U << KEY_TYPE_deleted)| 161 (1U << KEY_TYPE_whiteout)| 162 (1U << KEY_TYPE_cookie)| 163 (1U << KEY_TYPE_hash_whiteout)| 164 (1U << KEY_TYPE_xattr), 165 [BKEY_TYPE_alloc] = 166 (1U << KEY_TYPE_deleted)| 167 (1U << KEY_TYPE_alloc)| 168 (1U << KEY_TYPE_alloc_v2)| 169 (1U << KEY_TYPE_alloc_v3)| 170 (1U << KEY_TYPE_alloc_v4), 171 [BKEY_TYPE_quotas] = 172 (1U << KEY_TYPE_deleted)| 173 (1U << KEY_TYPE_quota), 174 [BKEY_TYPE_stripes] = 175 (1U << KEY_TYPE_deleted)| 176 (1U << KEY_TYPE_stripe), 177 [BKEY_TYPE_reflink] = 178 (1U << KEY_TYPE_deleted)| 179 (1U << KEY_TYPE_reflink_v)| 180 (1U << KEY_TYPE_indirect_inline_data), 181 [BKEY_TYPE_subvolumes] = 182 (1U << KEY_TYPE_deleted)| 183 (1U << KEY_TYPE_subvolume), 184 [BKEY_TYPE_snapshots] = 185 (1U << KEY_TYPE_deleted)| 186 (1U << KEY_TYPE_snapshot), 187 [BKEY_TYPE_lru] = 188 (1U << KEY_TYPE_deleted)| 189 (1U << KEY_TYPE_set), 190 [BKEY_TYPE_freespace] = 191 (1U << KEY_TYPE_deleted)| 192 (1U << KEY_TYPE_set), 193 [BKEY_TYPE_need_discard] = 194 (1U << KEY_TYPE_deleted)| 195 (1U << KEY_TYPE_set), 196 [BKEY_TYPE_backpointers] = 197 (1U << KEY_TYPE_deleted)| 198 (1U << KEY_TYPE_backpointer), 199 [BKEY_TYPE_bucket_gens] = 200 (1U << KEY_TYPE_deleted)| 201 (1U << KEY_TYPE_bucket_gens), 202 [BKEY_TYPE_btree] = 203 (1U << KEY_TYPE_deleted)| 204 (1U << KEY_TYPE_btree_ptr)| 205 (1U << KEY_TYPE_btree_ptr_v2), 206 }; 207 208 int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k, 209 enum btree_node_type type, 210 unsigned flags, struct printbuf *err) 211 { 212 if (k.k->u64s < BKEY_U64s) { 213 prt_printf(err, "u64s too small (%u < %zu)", k.k->u64s, BKEY_U64s); 214 return -BCH_ERR_invalid_bkey; 215 } 216 217 if (!(bch2_key_types_allowed[type] & (1U << k.k->type))) { 218 prt_printf(err, "invalid key type for btree %s (%s)", 219 bch2_btree_ids[type], bch2_bkey_types[k.k->type]); 220 return -BCH_ERR_invalid_bkey; 221 } 222 223 if (btree_node_type_is_extents(type) && !bkey_whiteout(k.k)) { 224 if (k.k->size == 0) { 225 prt_printf(err, "size == 0"); 226 return -BCH_ERR_invalid_bkey; 227 } 228 229 if (k.k->size > k.k->p.offset) { 230 prt_printf(err, "size greater than offset (%u > %llu)", 231 k.k->size, k.k->p.offset); 232 return -BCH_ERR_invalid_bkey; 233 } 234 } else { 235 if (k.k->size) { 236 prt_printf(err, "size != 0"); 237 return -BCH_ERR_invalid_bkey; 238 } 239 } 240 241 if (type != BKEY_TYPE_btree && 242 !btree_type_has_snapshots(type) && 243 k.k->p.snapshot) { 244 prt_printf(err, "nonzero snapshot"); 245 return -BCH_ERR_invalid_bkey; 246 } 247 248 if (type != BKEY_TYPE_btree && 249 btree_type_has_snapshots(type) && 250 !k.k->p.snapshot) { 251 prt_printf(err, "snapshot == 0"); 252 return -BCH_ERR_invalid_bkey; 253 } 254 255 if (type != BKEY_TYPE_btree && 256 bkey_eq(k.k->p, POS_MAX)) { 257 prt_printf(err, "key at POS_MAX"); 258 return -BCH_ERR_invalid_bkey; 259 } 260 261 return 0; 262 } 263 264 int bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k, 265 enum btree_node_type type, 266 unsigned flags, struct printbuf *err) 267 { 268 return __bch2_bkey_invalid(c, k, type, flags, err) ?: 269 bch2_bkey_val_invalid(c, k, flags, err); 270 } 271 272 int bch2_bkey_in_btree_node(struct btree *b, struct bkey_s_c k, 273 struct printbuf *err) 274 { 275 if (bpos_lt(k.k->p, b->data->min_key)) { 276 prt_printf(err, "key before start of btree node"); 277 return -BCH_ERR_invalid_bkey; 278 } 279 280 if (bpos_gt(k.k->p, b->data->max_key)) { 281 prt_printf(err, "key past end of btree node"); 282 return -BCH_ERR_invalid_bkey; 283 } 284 285 return 0; 286 } 287 288 void bch2_bpos_to_text(struct printbuf *out, struct bpos pos) 289 { 290 if (bpos_eq(pos, POS_MIN)) 291 prt_printf(out, "POS_MIN"); 292 else if (bpos_eq(pos, POS_MAX)) 293 prt_printf(out, "POS_MAX"); 294 else if (bpos_eq(pos, SPOS_MAX)) 295 prt_printf(out, "SPOS_MAX"); 296 else { 297 if (pos.inode == U64_MAX) 298 prt_printf(out, "U64_MAX"); 299 else 300 prt_printf(out, "%llu", pos.inode); 301 prt_printf(out, ":"); 302 if (pos.offset == U64_MAX) 303 prt_printf(out, "U64_MAX"); 304 else 305 prt_printf(out, "%llu", pos.offset); 306 prt_printf(out, ":"); 307 if (pos.snapshot == U32_MAX) 308 prt_printf(out, "U32_MAX"); 309 else 310 prt_printf(out, "%u", pos.snapshot); 311 } 312 } 313 314 void bch2_bkey_to_text(struct printbuf *out, const struct bkey *k) 315 { 316 if (k) { 317 prt_printf(out, "u64s %u type ", k->u64s); 318 319 if (k->type < KEY_TYPE_MAX) 320 prt_printf(out, "%s ", bch2_bkey_types[k->type]); 321 else 322 prt_printf(out, "%u ", k->type); 323 324 bch2_bpos_to_text(out, k->p); 325 326 prt_printf(out, " len %u ver %llu", k->size, k->version.lo); 327 } else { 328 prt_printf(out, "(null)"); 329 } 330 } 331 332 void bch2_val_to_text(struct printbuf *out, struct bch_fs *c, 333 struct bkey_s_c k) 334 { 335 if (k.k->type < KEY_TYPE_MAX) { 336 const struct bkey_ops *ops = &bch2_bkey_ops[k.k->type]; 337 338 if (likely(ops->val_to_text)) 339 ops->val_to_text(out, c, k); 340 } else { 341 prt_printf(out, "(invalid type %u)", k.k->type); 342 } 343 } 344 345 void bch2_bkey_val_to_text(struct printbuf *out, struct bch_fs *c, 346 struct bkey_s_c k) 347 { 348 bch2_bkey_to_text(out, k.k); 349 350 if (bkey_val_bytes(k.k)) { 351 prt_printf(out, ": "); 352 bch2_val_to_text(out, c, k); 353 } 354 } 355 356 void bch2_bkey_swab_val(struct bkey_s k) 357 { 358 const struct bkey_ops *ops = &bch2_bkey_ops[k.k->type]; 359 360 if (ops->swab) 361 ops->swab(k); 362 } 363 364 bool bch2_bkey_normalize(struct bch_fs *c, struct bkey_s k) 365 { 366 const struct bkey_ops *ops = &bch2_bkey_ops[k.k->type]; 367 368 return ops->key_normalize 369 ? ops->key_normalize(c, k) 370 : false; 371 } 372 373 bool bch2_bkey_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r) 374 { 375 const struct bkey_ops *ops = &bch2_bkey_ops[l.k->type]; 376 377 return bch2_bkey_maybe_mergable(l.k, r.k) && 378 (u64) l.k->size + r.k->size <= KEY_SIZE_MAX && 379 bch2_bkey_ops[l.k->type].key_merge && 380 !bch2_key_merging_disabled && 381 ops->key_merge(c, l, r); 382 } 383 384 static const struct old_bkey_type { 385 u8 btree_node_type; 386 u8 old; 387 u8 new; 388 } bkey_renumber_table[] = { 389 {BKEY_TYPE_btree, 128, KEY_TYPE_btree_ptr }, 390 {BKEY_TYPE_extents, 128, KEY_TYPE_extent }, 391 {BKEY_TYPE_extents, 129, KEY_TYPE_extent }, 392 {BKEY_TYPE_extents, 130, KEY_TYPE_reservation }, 393 {BKEY_TYPE_inodes, 128, KEY_TYPE_inode }, 394 {BKEY_TYPE_inodes, 130, KEY_TYPE_inode_generation }, 395 {BKEY_TYPE_dirents, 128, KEY_TYPE_dirent }, 396 {BKEY_TYPE_dirents, 129, KEY_TYPE_hash_whiteout }, 397 {BKEY_TYPE_xattrs, 128, KEY_TYPE_xattr }, 398 {BKEY_TYPE_xattrs, 129, KEY_TYPE_hash_whiteout }, 399 {BKEY_TYPE_alloc, 128, KEY_TYPE_alloc }, 400 {BKEY_TYPE_quotas, 128, KEY_TYPE_quota }, 401 }; 402 403 void bch2_bkey_renumber(enum btree_node_type btree_node_type, 404 struct bkey_packed *k, 405 int write) 406 { 407 const struct old_bkey_type *i; 408 409 for (i = bkey_renumber_table; 410 i < bkey_renumber_table + ARRAY_SIZE(bkey_renumber_table); 411 i++) 412 if (btree_node_type == i->btree_node_type && 413 k->type == (write ? i->new : i->old)) { 414 k->type = write ? i->old : i->new; 415 break; 416 } 417 } 418 419 void __bch2_bkey_compat(unsigned level, enum btree_id btree_id, 420 unsigned version, unsigned big_endian, 421 int write, 422 struct bkey_format *f, 423 struct bkey_packed *k) 424 { 425 const struct bkey_ops *ops; 426 struct bkey uk; 427 struct bkey_s u; 428 unsigned nr_compat = 5; 429 int i; 430 431 /* 432 * Do these operations in reverse order in the write path: 433 */ 434 435 for (i = 0; i < nr_compat; i++) 436 switch (!write ? i : nr_compat - 1 - i) { 437 case 0: 438 if (big_endian != CPU_BIG_ENDIAN) 439 bch2_bkey_swab_key(f, k); 440 break; 441 case 1: 442 if (version < bcachefs_metadata_version_bkey_renumber) 443 bch2_bkey_renumber(__btree_node_type(level, btree_id), k, write); 444 break; 445 case 2: 446 if (version < bcachefs_metadata_version_inode_btree_change && 447 btree_id == BTREE_ID_inodes) { 448 if (!bkey_packed(k)) { 449 struct bkey_i *u = packed_to_bkey(k); 450 451 swap(u->k.p.inode, u->k.p.offset); 452 } else if (f->bits_per_field[BKEY_FIELD_INODE] && 453 f->bits_per_field[BKEY_FIELD_OFFSET]) { 454 struct bkey_format tmp = *f, *in = f, *out = &tmp; 455 456 swap(tmp.bits_per_field[BKEY_FIELD_INODE], 457 tmp.bits_per_field[BKEY_FIELD_OFFSET]); 458 swap(tmp.field_offset[BKEY_FIELD_INODE], 459 tmp.field_offset[BKEY_FIELD_OFFSET]); 460 461 if (!write) 462 swap(in, out); 463 464 uk = __bch2_bkey_unpack_key(in, k); 465 swap(uk.p.inode, uk.p.offset); 466 BUG_ON(!bch2_bkey_pack_key(k, &uk, out)); 467 } 468 } 469 break; 470 case 3: 471 if (version < bcachefs_metadata_version_snapshot && 472 (level || btree_type_has_snapshots(btree_id))) { 473 struct bkey_i *u = packed_to_bkey(k); 474 475 if (u) { 476 u->k.p.snapshot = write 477 ? 0 : U32_MAX; 478 } else { 479 u64 min_packed = f->field_offset[BKEY_FIELD_SNAPSHOT]; 480 u64 max_packed = min_packed + 481 ~(~0ULL << f->bits_per_field[BKEY_FIELD_SNAPSHOT]); 482 483 uk = __bch2_bkey_unpack_key(f, k); 484 uk.p.snapshot = write 485 ? min_packed : min_t(u64, U32_MAX, max_packed); 486 487 BUG_ON(!bch2_bkey_pack_key(k, &uk, f)); 488 } 489 } 490 491 break; 492 case 4: 493 if (!bkey_packed(k)) { 494 u = bkey_i_to_s(packed_to_bkey(k)); 495 } else { 496 uk = __bch2_bkey_unpack_key(f, k); 497 u.k = &uk; 498 u.v = bkeyp_val(f, k); 499 } 500 501 if (big_endian != CPU_BIG_ENDIAN) 502 bch2_bkey_swab_val(u); 503 504 ops = &bch2_bkey_ops[k->type]; 505 506 if (ops->compat) 507 ops->compat(btree_id, version, big_endian, write, u); 508 break; 509 default: 510 BUG(); 511 } 512 } 513