1 // SPDX-License-Identifier: GPL-2.0 2 3 #include "bcachefs.h" 4 #include "backpointers.h" 5 #include "bkey_methods.h" 6 #include "btree_cache.h" 7 #include "btree_types.h" 8 #include "alloc_background.h" 9 #include "dirent.h" 10 #include "disk_accounting.h" 11 #include "ec.h" 12 #include "error.h" 13 #include "extents.h" 14 #include "inode.h" 15 #include "io_misc.h" 16 #include "lru.h" 17 #include "quota.h" 18 #include "reflink.h" 19 #include "snapshot.h" 20 #include "subvolume.h" 21 #include "xattr.h" 22 23 const char * const bch2_bkey_types[] = { 24 #define x(name, nr) #name, 25 BCH_BKEY_TYPES() 26 #undef x 27 NULL 28 }; 29 30 static int deleted_key_validate(struct bch_fs *c, struct bkey_s_c k, 31 struct bkey_validate_context from) 32 { 33 return 0; 34 } 35 36 #define bch2_bkey_ops_deleted ((struct bkey_ops) { \ 37 .key_validate = deleted_key_validate, \ 38 }) 39 40 #define bch2_bkey_ops_whiteout ((struct bkey_ops) { \ 41 .key_validate = deleted_key_validate, \ 42 }) 43 44 static int empty_val_key_validate(struct bch_fs *c, struct bkey_s_c k, 45 struct bkey_validate_context from) 46 { 47 int ret = 0; 48 49 bkey_fsck_err_on(bkey_val_bytes(k.k), 50 c, bkey_val_size_nonzero, 51 "incorrect value size (%zu != 0)", 52 bkey_val_bytes(k.k)); 53 fsck_err: 54 return ret; 55 } 56 57 #define bch2_bkey_ops_error ((struct bkey_ops) { \ 58 .key_validate = empty_val_key_validate, \ 59 }) 60 61 static int key_type_cookie_validate(struct bch_fs *c, struct bkey_s_c k, 62 struct bkey_validate_context from) 63 { 64 return 0; 65 } 66 67 static void key_type_cookie_to_text(struct printbuf *out, struct bch_fs *c, 68 struct bkey_s_c k) 69 { 70 struct bkey_s_c_cookie ck = bkey_s_c_to_cookie(k); 71 72 prt_printf(out, "%llu", le64_to_cpu(ck.v->cookie)); 73 } 74 75 #define bch2_bkey_ops_cookie ((struct bkey_ops) { \ 76 .key_validate = key_type_cookie_validate, \ 77 .val_to_text = key_type_cookie_to_text, \ 78 .min_val_size = 8, \ 79 }) 80 81 #define bch2_bkey_ops_hash_whiteout ((struct bkey_ops) {\ 82 .key_validate = empty_val_key_validate, \ 83 }) 84 85 static int key_type_inline_data_validate(struct bch_fs *c, struct bkey_s_c k, 86 struct bkey_validate_context from) 87 { 88 return 0; 89 } 90 91 static void key_type_inline_data_to_text(struct printbuf *out, struct bch_fs *c, 92 struct bkey_s_c k) 93 { 94 struct bkey_s_c_inline_data d = bkey_s_c_to_inline_data(k); 95 unsigned datalen = bkey_inline_data_bytes(k.k); 96 97 prt_printf(out, "datalen %u: %*phN", 98 datalen, min(datalen, 32U), d.v->data); 99 } 100 101 #define bch2_bkey_ops_inline_data ((struct bkey_ops) { \ 102 .key_validate = key_type_inline_data_validate, \ 103 .val_to_text = key_type_inline_data_to_text, \ 104 }) 105 106 static bool key_type_set_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r) 107 { 108 bch2_key_resize(l.k, l.k->size + r.k->size); 109 return true; 110 } 111 112 #define bch2_bkey_ops_set ((struct bkey_ops) { \ 113 .key_validate = empty_val_key_validate, \ 114 .key_merge = key_type_set_merge, \ 115 }) 116 117 const struct bkey_ops bch2_bkey_ops[] = { 118 #define x(name, nr) [KEY_TYPE_##name] = bch2_bkey_ops_##name, 119 BCH_BKEY_TYPES() 120 #undef x 121 }; 122 123 const struct bkey_ops bch2_bkey_null_ops = { 124 }; 125 126 int bch2_bkey_val_validate(struct bch_fs *c, struct bkey_s_c k, 127 struct bkey_validate_context from) 128 { 129 if (test_bit(BCH_FS_no_invalid_checks, &c->flags)) 130 return 0; 131 132 const struct bkey_ops *ops = bch2_bkey_type_ops(k.k->type); 133 int ret = 0; 134 135 bkey_fsck_err_on(bkey_val_bytes(k.k) < ops->min_val_size, 136 c, bkey_val_size_too_small, 137 "bad val size (%zu < %u)", 138 bkey_val_bytes(k.k), ops->min_val_size); 139 140 if (!ops->key_validate) 141 return 0; 142 143 ret = ops->key_validate(c, k, from); 144 fsck_err: 145 return ret; 146 } 147 148 static u64 bch2_key_types_allowed[] = { 149 [BKEY_TYPE_btree] = 150 BIT_ULL(KEY_TYPE_deleted)| 151 BIT_ULL(KEY_TYPE_btree_ptr)| 152 BIT_ULL(KEY_TYPE_btree_ptr_v2), 153 #define x(name, nr, flags, keys) [BKEY_TYPE_##name] = BIT_ULL(KEY_TYPE_deleted)|keys, 154 BCH_BTREE_IDS() 155 #undef x 156 }; 157 158 const char *bch2_btree_node_type_str(enum btree_node_type type) 159 { 160 return type == BKEY_TYPE_btree ? "internal btree node" : bch2_btree_id_str(type - 1); 161 } 162 163 int __bch2_bkey_validate(struct bch_fs *c, struct bkey_s_c k, 164 struct bkey_validate_context from) 165 { 166 enum btree_node_type type = __btree_node_type(from.level, from.btree); 167 168 if (test_bit(BCH_FS_no_invalid_checks, &c->flags)) 169 return 0; 170 171 int ret = 0; 172 173 bkey_fsck_err_on(k.k->u64s < BKEY_U64s, 174 c, bkey_u64s_too_small, 175 "u64s too small (%u < %zu)", k.k->u64s, BKEY_U64s); 176 177 if (type >= BKEY_TYPE_NR) 178 return 0; 179 180 bkey_fsck_err_on(k.k->type < KEY_TYPE_MAX && 181 (type == BKEY_TYPE_btree || (from.flags & BCH_VALIDATE_commit)) && 182 !(bch2_key_types_allowed[type] & BIT_ULL(k.k->type)), 183 c, bkey_invalid_type_for_btree, 184 "invalid key type for btree %s (%s)", 185 bch2_btree_node_type_str(type), 186 k.k->type < KEY_TYPE_MAX 187 ? bch2_bkey_types[k.k->type] 188 : "(unknown)"); 189 190 if (btree_node_type_is_extents(type) && !bkey_whiteout(k.k)) { 191 bkey_fsck_err_on(k.k->size == 0, 192 c, bkey_extent_size_zero, 193 "size == 0"); 194 195 bkey_fsck_err_on(k.k->size > k.k->p.offset, 196 c, bkey_extent_size_greater_than_offset, 197 "size greater than offset (%u > %llu)", 198 k.k->size, k.k->p.offset); 199 } else { 200 bkey_fsck_err_on(k.k->size, 201 c, bkey_size_nonzero, 202 "size != 0"); 203 } 204 205 if (type != BKEY_TYPE_btree) { 206 enum btree_id btree = type - 1; 207 208 if (btree_type_has_snapshots(btree)) { 209 bkey_fsck_err_on(!k.k->p.snapshot, 210 c, bkey_snapshot_zero, 211 "snapshot == 0"); 212 } else if (!btree_type_has_snapshot_field(btree)) { 213 bkey_fsck_err_on(k.k->p.snapshot, 214 c, bkey_snapshot_nonzero, 215 "nonzero snapshot"); 216 } else { 217 /* 218 * btree uses snapshot field but it's not required to be 219 * nonzero 220 */ 221 } 222 223 bkey_fsck_err_on(bkey_eq(k.k->p, POS_MAX), 224 c, bkey_at_pos_max, 225 "key at POS_MAX"); 226 } 227 fsck_err: 228 return ret; 229 } 230 231 int bch2_bkey_validate(struct bch_fs *c, struct bkey_s_c k, 232 struct bkey_validate_context from) 233 { 234 return __bch2_bkey_validate(c, k, from) ?: 235 bch2_bkey_val_validate(c, k, from); 236 } 237 238 int bch2_bkey_in_btree_node(struct bch_fs *c, struct btree *b, 239 struct bkey_s_c k, 240 struct bkey_validate_context from) 241 { 242 int ret = 0; 243 244 bkey_fsck_err_on(bpos_lt(k.k->p, b->data->min_key), 245 c, bkey_before_start_of_btree_node, 246 "key before start of btree node"); 247 248 bkey_fsck_err_on(bpos_gt(k.k->p, b->data->max_key), 249 c, bkey_after_end_of_btree_node, 250 "key past end of btree node"); 251 fsck_err: 252 return ret; 253 } 254 255 void bch2_bpos_to_text(struct printbuf *out, struct bpos pos) 256 { 257 if (bpos_eq(pos, POS_MIN)) 258 prt_printf(out, "POS_MIN"); 259 else if (bpos_eq(pos, POS_MAX)) 260 prt_printf(out, "POS_MAX"); 261 else if (bpos_eq(pos, SPOS_MAX)) 262 prt_printf(out, "SPOS_MAX"); 263 else { 264 if (pos.inode == U64_MAX) 265 prt_printf(out, "U64_MAX"); 266 else 267 prt_printf(out, "%llu", pos.inode); 268 prt_printf(out, ":"); 269 if (pos.offset == U64_MAX) 270 prt_printf(out, "U64_MAX"); 271 else 272 prt_printf(out, "%llu", pos.offset); 273 prt_printf(out, ":"); 274 if (pos.snapshot == U32_MAX) 275 prt_printf(out, "U32_MAX"); 276 else 277 prt_printf(out, "%u", pos.snapshot); 278 } 279 } 280 281 void bch2_bkey_to_text(struct printbuf *out, const struct bkey *k) 282 { 283 if (k) { 284 prt_printf(out, "u64s %u type ", k->u64s); 285 286 if (k->type < KEY_TYPE_MAX) 287 prt_printf(out, "%s ", bch2_bkey_types[k->type]); 288 else 289 prt_printf(out, "%u ", k->type); 290 291 bch2_bpos_to_text(out, k->p); 292 293 prt_printf(out, " len %u ver %llu", k->size, k->bversion.lo); 294 } else { 295 prt_printf(out, "(null)"); 296 } 297 } 298 299 void bch2_val_to_text(struct printbuf *out, struct bch_fs *c, 300 struct bkey_s_c k) 301 { 302 const struct bkey_ops *ops = bch2_bkey_type_ops(k.k->type); 303 304 if (likely(ops->val_to_text)) 305 ops->val_to_text(out, c, k); 306 } 307 308 void bch2_bkey_val_to_text(struct printbuf *out, struct bch_fs *c, 309 struct bkey_s_c k) 310 { 311 bch2_bkey_to_text(out, k.k); 312 313 if (bkey_val_bytes(k.k)) { 314 prt_printf(out, ": "); 315 bch2_val_to_text(out, c, k); 316 } 317 } 318 319 void bch2_bkey_swab_val(struct bkey_s k) 320 { 321 const struct bkey_ops *ops = bch2_bkey_type_ops(k.k->type); 322 323 if (ops->swab) 324 ops->swab(k); 325 } 326 327 bool bch2_bkey_normalize(struct bch_fs *c, struct bkey_s k) 328 { 329 const struct bkey_ops *ops = bch2_bkey_type_ops(k.k->type); 330 331 return ops->key_normalize 332 ? ops->key_normalize(c, k) 333 : false; 334 } 335 336 bool bch2_bkey_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r) 337 { 338 const struct bkey_ops *ops = bch2_bkey_type_ops(l.k->type); 339 340 return ops->key_merge && 341 bch2_bkey_maybe_mergable(l.k, r.k) && 342 (u64) l.k->size + r.k->size <= KEY_SIZE_MAX && 343 !bch2_key_merging_disabled && 344 ops->key_merge(c, l, r); 345 } 346 347 static const struct old_bkey_type { 348 u8 btree_node_type; 349 u8 old; 350 u8 new; 351 } bkey_renumber_table[] = { 352 {BKEY_TYPE_btree, 128, KEY_TYPE_btree_ptr }, 353 {BKEY_TYPE_extents, 128, KEY_TYPE_extent }, 354 {BKEY_TYPE_extents, 129, KEY_TYPE_extent }, 355 {BKEY_TYPE_extents, 130, KEY_TYPE_reservation }, 356 {BKEY_TYPE_inodes, 128, KEY_TYPE_inode }, 357 {BKEY_TYPE_inodes, 130, KEY_TYPE_inode_generation }, 358 {BKEY_TYPE_dirents, 128, KEY_TYPE_dirent }, 359 {BKEY_TYPE_dirents, 129, KEY_TYPE_hash_whiteout }, 360 {BKEY_TYPE_xattrs, 128, KEY_TYPE_xattr }, 361 {BKEY_TYPE_xattrs, 129, KEY_TYPE_hash_whiteout }, 362 {BKEY_TYPE_alloc, 128, KEY_TYPE_alloc }, 363 {BKEY_TYPE_quotas, 128, KEY_TYPE_quota }, 364 }; 365 366 void bch2_bkey_renumber(enum btree_node_type btree_node_type, 367 struct bkey_packed *k, 368 int write) 369 { 370 const struct old_bkey_type *i; 371 372 for (i = bkey_renumber_table; 373 i < bkey_renumber_table + ARRAY_SIZE(bkey_renumber_table); 374 i++) 375 if (btree_node_type == i->btree_node_type && 376 k->type == (write ? i->new : i->old)) { 377 k->type = write ? i->old : i->new; 378 break; 379 } 380 } 381 382 void __bch2_bkey_compat(unsigned level, enum btree_id btree_id, 383 unsigned version, unsigned big_endian, 384 int write, 385 struct bkey_format *f, 386 struct bkey_packed *k) 387 { 388 const struct bkey_ops *ops; 389 struct bkey uk; 390 unsigned nr_compat = 5; 391 int i; 392 393 /* 394 * Do these operations in reverse order in the write path: 395 */ 396 397 for (i = 0; i < nr_compat; i++) 398 switch (!write ? i : nr_compat - 1 - i) { 399 case 0: 400 if (big_endian != CPU_BIG_ENDIAN) { 401 bch2_bkey_swab_key(f, k); 402 } else if (IS_ENABLED(CONFIG_BCACHEFS_DEBUG)) { 403 bch2_bkey_swab_key(f, k); 404 bch2_bkey_swab_key(f, k); 405 } 406 break; 407 case 1: 408 if (version < bcachefs_metadata_version_bkey_renumber) 409 bch2_bkey_renumber(__btree_node_type(level, btree_id), k, write); 410 break; 411 case 2: 412 if (version < bcachefs_metadata_version_inode_btree_change && 413 btree_id == BTREE_ID_inodes) { 414 if (!bkey_packed(k)) { 415 struct bkey_i *u = packed_to_bkey(k); 416 417 swap(u->k.p.inode, u->k.p.offset); 418 } else if (f->bits_per_field[BKEY_FIELD_INODE] && 419 f->bits_per_field[BKEY_FIELD_OFFSET]) { 420 struct bkey_format tmp = *f, *in = f, *out = &tmp; 421 422 swap(tmp.bits_per_field[BKEY_FIELD_INODE], 423 tmp.bits_per_field[BKEY_FIELD_OFFSET]); 424 swap(tmp.field_offset[BKEY_FIELD_INODE], 425 tmp.field_offset[BKEY_FIELD_OFFSET]); 426 427 if (!write) 428 swap(in, out); 429 430 uk = __bch2_bkey_unpack_key(in, k); 431 swap(uk.p.inode, uk.p.offset); 432 BUG_ON(!bch2_bkey_pack_key(k, &uk, out)); 433 } 434 } 435 break; 436 case 3: 437 if (version < bcachefs_metadata_version_snapshot && 438 (level || btree_type_has_snapshots(btree_id))) { 439 struct bkey_i *u = packed_to_bkey(k); 440 441 if (u) { 442 u->k.p.snapshot = write 443 ? 0 : U32_MAX; 444 } else { 445 u64 min_packed = le64_to_cpu(f->field_offset[BKEY_FIELD_SNAPSHOT]); 446 u64 max_packed = min_packed + 447 ~(~0ULL << f->bits_per_field[BKEY_FIELD_SNAPSHOT]); 448 449 uk = __bch2_bkey_unpack_key(f, k); 450 uk.p.snapshot = write 451 ? min_packed : min_t(u64, U32_MAX, max_packed); 452 453 BUG_ON(!bch2_bkey_pack_key(k, &uk, f)); 454 } 455 } 456 457 break; 458 case 4: { 459 struct bkey_s u; 460 461 if (!bkey_packed(k)) { 462 u = bkey_i_to_s(packed_to_bkey(k)); 463 } else { 464 uk = __bch2_bkey_unpack_key(f, k); 465 u.k = &uk; 466 u.v = bkeyp_val(f, k); 467 } 468 469 if (big_endian != CPU_BIG_ENDIAN) 470 bch2_bkey_swab_val(u); 471 472 ops = bch2_bkey_type_ops(k->type); 473 474 if (ops->compat) 475 ops->compat(btree_id, version, big_endian, write, u); 476 break; 477 } 478 default: 479 BUG(); 480 } 481 } 482