1 /* 2 * Copyright (C) STRATO AG 2011. All rights reserved. 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of the GNU General Public 6 * License v2 as published by the Free Software Foundation. 7 * 8 * This program is distributed in the hope that it will be useful, 9 * but WITHOUT ANY WARRANTY; without even the implied warranty of 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 11 * General Public License for more details. 12 * 13 * You should have received a copy of the GNU General Public 14 * License along with this program; if not, write to the 15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 16 * Boston, MA 021110-1307, USA. 17 */ 18 19 /* 20 * This module can be used to catch cases when the btrfs kernel 21 * code executes write requests to the disk that bring the file 22 * system in an inconsistent state. In such a state, a power-loss 23 * or kernel panic event would cause that the data on disk is 24 * lost or at least damaged. 25 * 26 * Code is added that examines all block write requests during 27 * runtime (including writes of the super block). Three rules 28 * are verified and an error is printed on violation of the 29 * rules: 30 * 1. It is not allowed to write a disk block which is 31 * currently referenced by the super block (either directly 32 * or indirectly). 33 * 2. When a super block is written, it is verified that all 34 * referenced (directly or indirectly) blocks fulfill the 35 * following requirements: 36 * 2a. All referenced blocks have either been present when 37 * the file system was mounted, (i.e., they have been 38 * referenced by the super block) or they have been 39 * written since then and the write completion callback 40 * was called and no write error was indicated and a 41 * FLUSH request to the device where these blocks are 42 * located was received and completed. 43 * 2b. All referenced blocks need to have a generation 44 * number which is equal to the parent's number. 45 * 46 * One issue that was found using this module was that the log 47 * tree on disk became temporarily corrupted because disk blocks 48 * that had been in use for the log tree had been freed and 49 * reused too early, while being referenced by the written super 50 * block. 51 * 52 * The search term in the kernel log that can be used to filter 53 * on the existence of detected integrity issues is 54 * "btrfs: attempt". 55 * 56 * The integrity check is enabled via mount options. These 57 * mount options are only supported if the integrity check 58 * tool is compiled by defining BTRFS_FS_CHECK_INTEGRITY. 59 * 60 * Example #1, apply integrity checks to all metadata: 61 * mount /dev/sdb1 /mnt -o check_int 62 * 63 * Example #2, apply integrity checks to all metadata and 64 * to data extents: 65 * mount /dev/sdb1 /mnt -o check_int_data 66 * 67 * Example #3, apply integrity checks to all metadata and dump 68 * the tree that the super block references to kernel messages 69 * each time after a super block was written: 70 * mount /dev/sdb1 /mnt -o check_int,check_int_print_mask=263 71 * 72 * If the integrity check tool is included and activated in 73 * the mount options, plenty of kernel memory is used, and 74 * plenty of additional CPU cycles are spent. Enabling this 75 * functionality is not intended for normal use. In most 76 * cases, unless you are a btrfs developer who needs to verify 77 * the integrity of (super)-block write requests, do not 78 * enable the config option BTRFS_FS_CHECK_INTEGRITY to 79 * include and compile the integrity check tool. 80 * 81 * Expect millions of lines of information in the kernel log with an 82 * enabled check_int_print_mask. Therefore set LOG_BUF_SHIFT in the 83 * kernel config to at least 26 (which is 64MB). Usually the value is 84 * limited to 21 (which is 2MB) in init/Kconfig. The file needs to be 85 * changed like this before LOG_BUF_SHIFT can be set to a high value: 86 * config LOG_BUF_SHIFT 87 * int "Kernel log buffer size (16 => 64KB, 17 => 128KB)" 88 * range 12 30 89 */ 90 91 #include <linux/sched.h> 92 #include <linux/slab.h> 93 #include <linux/buffer_head.h> 94 #include <linux/mutex.h> 95 #include <linux/genhd.h> 96 #include <linux/blkdev.h> 97 #include <linux/vmalloc.h> 98 #include <linux/string.h> 99 #include "ctree.h" 100 #include "disk-io.h" 101 #include "hash.h" 102 #include "transaction.h" 103 #include "extent_io.h" 104 #include "volumes.h" 105 #include "print-tree.h" 106 #include "locking.h" 107 #include "check-integrity.h" 108 #include "rcu-string.h" 109 #include "compression.h" 110 111 #define BTRFSIC_BLOCK_HASHTABLE_SIZE 0x10000 112 #define BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE 0x10000 113 #define BTRFSIC_DEV2STATE_HASHTABLE_SIZE 0x100 114 #define BTRFSIC_BLOCK_MAGIC_NUMBER 0x14491051 115 #define BTRFSIC_BLOCK_LINK_MAGIC_NUMBER 0x11070807 116 #define BTRFSIC_DEV2STATE_MAGIC_NUMBER 0x20111530 117 #define BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER 20111300 118 #define BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL (200 - 6) /* in characters, 119 * excluding " [...]" */ 120 #define BTRFSIC_GENERATION_UNKNOWN ((u64)-1) 121 122 /* 123 * The definition of the bitmask fields for the print_mask. 124 * They are specified with the mount option check_integrity_print_mask. 125 */ 126 #define BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE 0x00000001 127 #define BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION 0x00000002 128 #define BTRFSIC_PRINT_MASK_TREE_AFTER_SB_WRITE 0x00000004 129 #define BTRFSIC_PRINT_MASK_TREE_BEFORE_SB_WRITE 0x00000008 130 #define BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH 0x00000010 131 #define BTRFSIC_PRINT_MASK_END_IO_BIO_BH 0x00000020 132 #define BTRFSIC_PRINT_MASK_VERBOSE 0x00000040 133 #define BTRFSIC_PRINT_MASK_VERY_VERBOSE 0x00000080 134 #define BTRFSIC_PRINT_MASK_INITIAL_TREE 0x00000100 135 #define BTRFSIC_PRINT_MASK_INITIAL_ALL_TREES 0x00000200 136 #define BTRFSIC_PRINT_MASK_INITIAL_DATABASE 0x00000400 137 #define BTRFSIC_PRINT_MASK_NUM_COPIES 0x00000800 138 #define BTRFSIC_PRINT_MASK_TREE_WITH_ALL_MIRRORS 0x00001000 139 #define BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH_VERBOSE 0x00002000 140 141 struct btrfsic_dev_state; 142 struct btrfsic_state; 143 144 struct btrfsic_block { 145 u32 magic_num; /* only used for debug purposes */ 146 unsigned int is_metadata:1; /* if it is meta-data, not data-data */ 147 unsigned int is_superblock:1; /* if it is one of the superblocks */ 148 unsigned int is_iodone:1; /* if is done by lower subsystem */ 149 unsigned int iodone_w_error:1; /* error was indicated to endio */ 150 unsigned int never_written:1; /* block was added because it was 151 * referenced, not because it was 152 * written */ 153 unsigned int mirror_num; /* large enough to hold 154 * BTRFS_SUPER_MIRROR_MAX */ 155 struct btrfsic_dev_state *dev_state; 156 u64 dev_bytenr; /* key, physical byte num on disk */ 157 u64 logical_bytenr; /* logical byte num on disk */ 158 u64 generation; 159 struct btrfs_disk_key disk_key; /* extra info to print in case of 160 * issues, will not always be correct */ 161 struct list_head collision_resolving_node; /* list node */ 162 struct list_head all_blocks_node; /* list node */ 163 164 /* the following two lists contain block_link items */ 165 struct list_head ref_to_list; /* list */ 166 struct list_head ref_from_list; /* list */ 167 struct btrfsic_block *next_in_same_bio; 168 void *orig_bio_bh_private; 169 union { 170 bio_end_io_t *bio; 171 bh_end_io_t *bh; 172 } orig_bio_bh_end_io; 173 int submit_bio_bh_rw; 174 u64 flush_gen; /* only valid if !never_written */ 175 }; 176 177 /* 178 * Elements of this type are allocated dynamically and required because 179 * each block object can refer to and can be ref from multiple blocks. 180 * The key to lookup them in the hashtable is the dev_bytenr of 181 * the block ref to plus the one from the block referred from. 182 * The fact that they are searchable via a hashtable and that a 183 * ref_cnt is maintained is not required for the btrfs integrity 184 * check algorithm itself, it is only used to make the output more 185 * beautiful in case that an error is detected (an error is defined 186 * as a write operation to a block while that block is still referenced). 187 */ 188 struct btrfsic_block_link { 189 u32 magic_num; /* only used for debug purposes */ 190 u32 ref_cnt; 191 struct list_head node_ref_to; /* list node */ 192 struct list_head node_ref_from; /* list node */ 193 struct list_head collision_resolving_node; /* list node */ 194 struct btrfsic_block *block_ref_to; 195 struct btrfsic_block *block_ref_from; 196 u64 parent_generation; 197 }; 198 199 struct btrfsic_dev_state { 200 u32 magic_num; /* only used for debug purposes */ 201 struct block_device *bdev; 202 struct btrfsic_state *state; 203 struct list_head collision_resolving_node; /* list node */ 204 struct btrfsic_block dummy_block_for_bio_bh_flush; 205 u64 last_flush_gen; 206 char name[BDEVNAME_SIZE]; 207 }; 208 209 struct btrfsic_block_hashtable { 210 struct list_head table[BTRFSIC_BLOCK_HASHTABLE_SIZE]; 211 }; 212 213 struct btrfsic_block_link_hashtable { 214 struct list_head table[BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE]; 215 }; 216 217 struct btrfsic_dev_state_hashtable { 218 struct list_head table[BTRFSIC_DEV2STATE_HASHTABLE_SIZE]; 219 }; 220 221 struct btrfsic_block_data_ctx { 222 u64 start; /* virtual bytenr */ 223 u64 dev_bytenr; /* physical bytenr on device */ 224 u32 len; 225 struct btrfsic_dev_state *dev; 226 char **datav; 227 struct page **pagev; 228 void *mem_to_free; 229 }; 230 231 /* This structure is used to implement recursion without occupying 232 * any stack space, refer to btrfsic_process_metablock() */ 233 struct btrfsic_stack_frame { 234 u32 magic; 235 u32 nr; 236 int error; 237 int i; 238 int limit_nesting; 239 int num_copies; 240 int mirror_num; 241 struct btrfsic_block *block; 242 struct btrfsic_block_data_ctx *block_ctx; 243 struct btrfsic_block *next_block; 244 struct btrfsic_block_data_ctx next_block_ctx; 245 struct btrfs_header *hdr; 246 struct btrfsic_stack_frame *prev; 247 }; 248 249 /* Some state per mounted filesystem */ 250 struct btrfsic_state { 251 u32 print_mask; 252 int include_extent_data; 253 int csum_size; 254 struct list_head all_blocks_list; 255 struct btrfsic_block_hashtable block_hashtable; 256 struct btrfsic_block_link_hashtable block_link_hashtable; 257 struct btrfs_root *root; 258 u64 max_superblock_generation; 259 struct btrfsic_block *latest_superblock; 260 u32 metablock_size; 261 u32 datablock_size; 262 }; 263 264 static void btrfsic_block_init(struct btrfsic_block *b); 265 static struct btrfsic_block *btrfsic_block_alloc(void); 266 static void btrfsic_block_free(struct btrfsic_block *b); 267 static void btrfsic_block_link_init(struct btrfsic_block_link *n); 268 static struct btrfsic_block_link *btrfsic_block_link_alloc(void); 269 static void btrfsic_block_link_free(struct btrfsic_block_link *n); 270 static void btrfsic_dev_state_init(struct btrfsic_dev_state *ds); 271 static struct btrfsic_dev_state *btrfsic_dev_state_alloc(void); 272 static void btrfsic_dev_state_free(struct btrfsic_dev_state *ds); 273 static void btrfsic_block_hashtable_init(struct btrfsic_block_hashtable *h); 274 static void btrfsic_block_hashtable_add(struct btrfsic_block *b, 275 struct btrfsic_block_hashtable *h); 276 static void btrfsic_block_hashtable_remove(struct btrfsic_block *b); 277 static struct btrfsic_block *btrfsic_block_hashtable_lookup( 278 struct block_device *bdev, 279 u64 dev_bytenr, 280 struct btrfsic_block_hashtable *h); 281 static void btrfsic_block_link_hashtable_init( 282 struct btrfsic_block_link_hashtable *h); 283 static void btrfsic_block_link_hashtable_add( 284 struct btrfsic_block_link *l, 285 struct btrfsic_block_link_hashtable *h); 286 static void btrfsic_block_link_hashtable_remove(struct btrfsic_block_link *l); 287 static struct btrfsic_block_link *btrfsic_block_link_hashtable_lookup( 288 struct block_device *bdev_ref_to, 289 u64 dev_bytenr_ref_to, 290 struct block_device *bdev_ref_from, 291 u64 dev_bytenr_ref_from, 292 struct btrfsic_block_link_hashtable *h); 293 static void btrfsic_dev_state_hashtable_init( 294 struct btrfsic_dev_state_hashtable *h); 295 static void btrfsic_dev_state_hashtable_add( 296 struct btrfsic_dev_state *ds, 297 struct btrfsic_dev_state_hashtable *h); 298 static void btrfsic_dev_state_hashtable_remove(struct btrfsic_dev_state *ds); 299 static struct btrfsic_dev_state *btrfsic_dev_state_hashtable_lookup( 300 struct block_device *bdev, 301 struct btrfsic_dev_state_hashtable *h); 302 static struct btrfsic_stack_frame *btrfsic_stack_frame_alloc(void); 303 static void btrfsic_stack_frame_free(struct btrfsic_stack_frame *sf); 304 static int btrfsic_process_superblock(struct btrfsic_state *state, 305 struct btrfs_fs_devices *fs_devices); 306 static int btrfsic_process_metablock(struct btrfsic_state *state, 307 struct btrfsic_block *block, 308 struct btrfsic_block_data_ctx *block_ctx, 309 int limit_nesting, int force_iodone_flag); 310 static void btrfsic_read_from_block_data( 311 struct btrfsic_block_data_ctx *block_ctx, 312 void *dst, u32 offset, size_t len); 313 static int btrfsic_create_link_to_next_block( 314 struct btrfsic_state *state, 315 struct btrfsic_block *block, 316 struct btrfsic_block_data_ctx 317 *block_ctx, u64 next_bytenr, 318 int limit_nesting, 319 struct btrfsic_block_data_ctx *next_block_ctx, 320 struct btrfsic_block **next_blockp, 321 int force_iodone_flag, 322 int *num_copiesp, int *mirror_nump, 323 struct btrfs_disk_key *disk_key, 324 u64 parent_generation); 325 static int btrfsic_handle_extent_data(struct btrfsic_state *state, 326 struct btrfsic_block *block, 327 struct btrfsic_block_data_ctx *block_ctx, 328 u32 item_offset, int force_iodone_flag); 329 static int btrfsic_map_block(struct btrfsic_state *state, u64 bytenr, u32 len, 330 struct btrfsic_block_data_ctx *block_ctx_out, 331 int mirror_num); 332 static void btrfsic_release_block_ctx(struct btrfsic_block_data_ctx *block_ctx); 333 static int btrfsic_read_block(struct btrfsic_state *state, 334 struct btrfsic_block_data_ctx *block_ctx); 335 static void btrfsic_dump_database(struct btrfsic_state *state); 336 static int btrfsic_test_for_metadata(struct btrfsic_state *state, 337 char **datav, unsigned int num_pages); 338 static void btrfsic_process_written_block(struct btrfsic_dev_state *dev_state, 339 u64 dev_bytenr, char **mapped_datav, 340 unsigned int num_pages, 341 struct bio *bio, int *bio_is_patched, 342 struct buffer_head *bh, 343 int submit_bio_bh_rw); 344 static int btrfsic_process_written_superblock( 345 struct btrfsic_state *state, 346 struct btrfsic_block *const block, 347 struct btrfs_super_block *const super_hdr); 348 static void btrfsic_bio_end_io(struct bio *bp); 349 static void btrfsic_bh_end_io(struct buffer_head *bh, int uptodate); 350 static int btrfsic_is_block_ref_by_superblock(const struct btrfsic_state *state, 351 const struct btrfsic_block *block, 352 int recursion_level); 353 static int btrfsic_check_all_ref_blocks(struct btrfsic_state *state, 354 struct btrfsic_block *const block, 355 int recursion_level); 356 static void btrfsic_print_add_link(const struct btrfsic_state *state, 357 const struct btrfsic_block_link *l); 358 static void btrfsic_print_rem_link(const struct btrfsic_state *state, 359 const struct btrfsic_block_link *l); 360 static char btrfsic_get_block_type(const struct btrfsic_state *state, 361 const struct btrfsic_block *block); 362 static void btrfsic_dump_tree(const struct btrfsic_state *state); 363 static void btrfsic_dump_tree_sub(const struct btrfsic_state *state, 364 const struct btrfsic_block *block, 365 int indent_level); 366 static struct btrfsic_block_link *btrfsic_block_link_lookup_or_add( 367 struct btrfsic_state *state, 368 struct btrfsic_block_data_ctx *next_block_ctx, 369 struct btrfsic_block *next_block, 370 struct btrfsic_block *from_block, 371 u64 parent_generation); 372 static struct btrfsic_block *btrfsic_block_lookup_or_add( 373 struct btrfsic_state *state, 374 struct btrfsic_block_data_ctx *block_ctx, 375 const char *additional_string, 376 int is_metadata, 377 int is_iodone, 378 int never_written, 379 int mirror_num, 380 int *was_created); 381 static int btrfsic_process_superblock_dev_mirror( 382 struct btrfsic_state *state, 383 struct btrfsic_dev_state *dev_state, 384 struct btrfs_device *device, 385 int superblock_mirror_num, 386 struct btrfsic_dev_state **selected_dev_state, 387 struct btrfs_super_block *selected_super); 388 static struct btrfsic_dev_state *btrfsic_dev_state_lookup( 389 struct block_device *bdev); 390 static void btrfsic_cmp_log_and_dev_bytenr(struct btrfsic_state *state, 391 u64 bytenr, 392 struct btrfsic_dev_state *dev_state, 393 u64 dev_bytenr); 394 395 static struct mutex btrfsic_mutex; 396 static int btrfsic_is_initialized; 397 static struct btrfsic_dev_state_hashtable btrfsic_dev_state_hashtable; 398 399 400 static void btrfsic_block_init(struct btrfsic_block *b) 401 { 402 b->magic_num = BTRFSIC_BLOCK_MAGIC_NUMBER; 403 b->dev_state = NULL; 404 b->dev_bytenr = 0; 405 b->logical_bytenr = 0; 406 b->generation = BTRFSIC_GENERATION_UNKNOWN; 407 b->disk_key.objectid = 0; 408 b->disk_key.type = 0; 409 b->disk_key.offset = 0; 410 b->is_metadata = 0; 411 b->is_superblock = 0; 412 b->is_iodone = 0; 413 b->iodone_w_error = 0; 414 b->never_written = 0; 415 b->mirror_num = 0; 416 b->next_in_same_bio = NULL; 417 b->orig_bio_bh_private = NULL; 418 b->orig_bio_bh_end_io.bio = NULL; 419 INIT_LIST_HEAD(&b->collision_resolving_node); 420 INIT_LIST_HEAD(&b->all_blocks_node); 421 INIT_LIST_HEAD(&b->ref_to_list); 422 INIT_LIST_HEAD(&b->ref_from_list); 423 b->submit_bio_bh_rw = 0; 424 b->flush_gen = 0; 425 } 426 427 static struct btrfsic_block *btrfsic_block_alloc(void) 428 { 429 struct btrfsic_block *b; 430 431 b = kzalloc(sizeof(*b), GFP_NOFS); 432 if (NULL != b) 433 btrfsic_block_init(b); 434 435 return b; 436 } 437 438 static void btrfsic_block_free(struct btrfsic_block *b) 439 { 440 BUG_ON(!(NULL == b || BTRFSIC_BLOCK_MAGIC_NUMBER == b->magic_num)); 441 kfree(b); 442 } 443 444 static void btrfsic_block_link_init(struct btrfsic_block_link *l) 445 { 446 l->magic_num = BTRFSIC_BLOCK_LINK_MAGIC_NUMBER; 447 l->ref_cnt = 1; 448 INIT_LIST_HEAD(&l->node_ref_to); 449 INIT_LIST_HEAD(&l->node_ref_from); 450 INIT_LIST_HEAD(&l->collision_resolving_node); 451 l->block_ref_to = NULL; 452 l->block_ref_from = NULL; 453 } 454 455 static struct btrfsic_block_link *btrfsic_block_link_alloc(void) 456 { 457 struct btrfsic_block_link *l; 458 459 l = kzalloc(sizeof(*l), GFP_NOFS); 460 if (NULL != l) 461 btrfsic_block_link_init(l); 462 463 return l; 464 } 465 466 static void btrfsic_block_link_free(struct btrfsic_block_link *l) 467 { 468 BUG_ON(!(NULL == l || BTRFSIC_BLOCK_LINK_MAGIC_NUMBER == l->magic_num)); 469 kfree(l); 470 } 471 472 static void btrfsic_dev_state_init(struct btrfsic_dev_state *ds) 473 { 474 ds->magic_num = BTRFSIC_DEV2STATE_MAGIC_NUMBER; 475 ds->bdev = NULL; 476 ds->state = NULL; 477 ds->name[0] = '\0'; 478 INIT_LIST_HEAD(&ds->collision_resolving_node); 479 ds->last_flush_gen = 0; 480 btrfsic_block_init(&ds->dummy_block_for_bio_bh_flush); 481 ds->dummy_block_for_bio_bh_flush.is_iodone = 1; 482 ds->dummy_block_for_bio_bh_flush.dev_state = ds; 483 } 484 485 static struct btrfsic_dev_state *btrfsic_dev_state_alloc(void) 486 { 487 struct btrfsic_dev_state *ds; 488 489 ds = kzalloc(sizeof(*ds), GFP_NOFS); 490 if (NULL != ds) 491 btrfsic_dev_state_init(ds); 492 493 return ds; 494 } 495 496 static void btrfsic_dev_state_free(struct btrfsic_dev_state *ds) 497 { 498 BUG_ON(!(NULL == ds || 499 BTRFSIC_DEV2STATE_MAGIC_NUMBER == ds->magic_num)); 500 kfree(ds); 501 } 502 503 static void btrfsic_block_hashtable_init(struct btrfsic_block_hashtable *h) 504 { 505 int i; 506 507 for (i = 0; i < BTRFSIC_BLOCK_HASHTABLE_SIZE; i++) 508 INIT_LIST_HEAD(h->table + i); 509 } 510 511 static void btrfsic_block_hashtable_add(struct btrfsic_block *b, 512 struct btrfsic_block_hashtable *h) 513 { 514 const unsigned int hashval = 515 (((unsigned int)(b->dev_bytenr >> 16)) ^ 516 ((unsigned int)((uintptr_t)b->dev_state->bdev))) & 517 (BTRFSIC_BLOCK_HASHTABLE_SIZE - 1); 518 519 list_add(&b->collision_resolving_node, h->table + hashval); 520 } 521 522 static void btrfsic_block_hashtable_remove(struct btrfsic_block *b) 523 { 524 list_del(&b->collision_resolving_node); 525 } 526 527 static struct btrfsic_block *btrfsic_block_hashtable_lookup( 528 struct block_device *bdev, 529 u64 dev_bytenr, 530 struct btrfsic_block_hashtable *h) 531 { 532 const unsigned int hashval = 533 (((unsigned int)(dev_bytenr >> 16)) ^ 534 ((unsigned int)((uintptr_t)bdev))) & 535 (BTRFSIC_BLOCK_HASHTABLE_SIZE - 1); 536 struct btrfsic_block *b; 537 538 list_for_each_entry(b, h->table + hashval, collision_resolving_node) { 539 if (b->dev_state->bdev == bdev && b->dev_bytenr == dev_bytenr) 540 return b; 541 } 542 543 return NULL; 544 } 545 546 static void btrfsic_block_link_hashtable_init( 547 struct btrfsic_block_link_hashtable *h) 548 { 549 int i; 550 551 for (i = 0; i < BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE; i++) 552 INIT_LIST_HEAD(h->table + i); 553 } 554 555 static void btrfsic_block_link_hashtable_add( 556 struct btrfsic_block_link *l, 557 struct btrfsic_block_link_hashtable *h) 558 { 559 const unsigned int hashval = 560 (((unsigned int)(l->block_ref_to->dev_bytenr >> 16)) ^ 561 ((unsigned int)(l->block_ref_from->dev_bytenr >> 16)) ^ 562 ((unsigned int)((uintptr_t)l->block_ref_to->dev_state->bdev)) ^ 563 ((unsigned int)((uintptr_t)l->block_ref_from->dev_state->bdev))) 564 & (BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE - 1); 565 566 BUG_ON(NULL == l->block_ref_to); 567 BUG_ON(NULL == l->block_ref_from); 568 list_add(&l->collision_resolving_node, h->table + hashval); 569 } 570 571 static void btrfsic_block_link_hashtable_remove(struct btrfsic_block_link *l) 572 { 573 list_del(&l->collision_resolving_node); 574 } 575 576 static struct btrfsic_block_link *btrfsic_block_link_hashtable_lookup( 577 struct block_device *bdev_ref_to, 578 u64 dev_bytenr_ref_to, 579 struct block_device *bdev_ref_from, 580 u64 dev_bytenr_ref_from, 581 struct btrfsic_block_link_hashtable *h) 582 { 583 const unsigned int hashval = 584 (((unsigned int)(dev_bytenr_ref_to >> 16)) ^ 585 ((unsigned int)(dev_bytenr_ref_from >> 16)) ^ 586 ((unsigned int)((uintptr_t)bdev_ref_to)) ^ 587 ((unsigned int)((uintptr_t)bdev_ref_from))) & 588 (BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE - 1); 589 struct btrfsic_block_link *l; 590 591 list_for_each_entry(l, h->table + hashval, collision_resolving_node) { 592 BUG_ON(NULL == l->block_ref_to); 593 BUG_ON(NULL == l->block_ref_from); 594 if (l->block_ref_to->dev_state->bdev == bdev_ref_to && 595 l->block_ref_to->dev_bytenr == dev_bytenr_ref_to && 596 l->block_ref_from->dev_state->bdev == bdev_ref_from && 597 l->block_ref_from->dev_bytenr == dev_bytenr_ref_from) 598 return l; 599 } 600 601 return NULL; 602 } 603 604 static void btrfsic_dev_state_hashtable_init( 605 struct btrfsic_dev_state_hashtable *h) 606 { 607 int i; 608 609 for (i = 0; i < BTRFSIC_DEV2STATE_HASHTABLE_SIZE; i++) 610 INIT_LIST_HEAD(h->table + i); 611 } 612 613 static void btrfsic_dev_state_hashtable_add( 614 struct btrfsic_dev_state *ds, 615 struct btrfsic_dev_state_hashtable *h) 616 { 617 const unsigned int hashval = 618 (((unsigned int)((uintptr_t)ds->bdev)) & 619 (BTRFSIC_DEV2STATE_HASHTABLE_SIZE - 1)); 620 621 list_add(&ds->collision_resolving_node, h->table + hashval); 622 } 623 624 static void btrfsic_dev_state_hashtable_remove(struct btrfsic_dev_state *ds) 625 { 626 list_del(&ds->collision_resolving_node); 627 } 628 629 static struct btrfsic_dev_state *btrfsic_dev_state_hashtable_lookup( 630 struct block_device *bdev, 631 struct btrfsic_dev_state_hashtable *h) 632 { 633 const unsigned int hashval = 634 (((unsigned int)((uintptr_t)bdev)) & 635 (BTRFSIC_DEV2STATE_HASHTABLE_SIZE - 1)); 636 struct btrfsic_dev_state *ds; 637 638 list_for_each_entry(ds, h->table + hashval, collision_resolving_node) { 639 if (ds->bdev == bdev) 640 return ds; 641 } 642 643 return NULL; 644 } 645 646 static int btrfsic_process_superblock(struct btrfsic_state *state, 647 struct btrfs_fs_devices *fs_devices) 648 { 649 int ret = 0; 650 struct btrfs_super_block *selected_super; 651 struct list_head *dev_head = &fs_devices->devices; 652 struct btrfs_device *device; 653 struct btrfsic_dev_state *selected_dev_state = NULL; 654 int pass; 655 656 BUG_ON(NULL == state); 657 selected_super = kzalloc(sizeof(*selected_super), GFP_NOFS); 658 if (NULL == selected_super) { 659 pr_info("btrfsic: error, kmalloc failed!\n"); 660 return -ENOMEM; 661 } 662 663 list_for_each_entry(device, dev_head, dev_list) { 664 int i; 665 struct btrfsic_dev_state *dev_state; 666 667 if (!device->bdev || !device->name) 668 continue; 669 670 dev_state = btrfsic_dev_state_lookup(device->bdev); 671 BUG_ON(NULL == dev_state); 672 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) { 673 ret = btrfsic_process_superblock_dev_mirror( 674 state, dev_state, device, i, 675 &selected_dev_state, selected_super); 676 if (0 != ret && 0 == i) { 677 kfree(selected_super); 678 return ret; 679 } 680 } 681 } 682 683 if (NULL == state->latest_superblock) { 684 pr_info("btrfsic: no superblock found!\n"); 685 kfree(selected_super); 686 return -1; 687 } 688 689 state->csum_size = btrfs_super_csum_size(selected_super); 690 691 for (pass = 0; pass < 3; pass++) { 692 int num_copies; 693 int mirror_num; 694 u64 next_bytenr; 695 696 switch (pass) { 697 case 0: 698 next_bytenr = btrfs_super_root(selected_super); 699 if (state->print_mask & 700 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION) 701 pr_info("root@%llu\n", next_bytenr); 702 break; 703 case 1: 704 next_bytenr = btrfs_super_chunk_root(selected_super); 705 if (state->print_mask & 706 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION) 707 pr_info("chunk@%llu\n", next_bytenr); 708 break; 709 case 2: 710 next_bytenr = btrfs_super_log_root(selected_super); 711 if (0 == next_bytenr) 712 continue; 713 if (state->print_mask & 714 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION) 715 pr_info("log@%llu\n", next_bytenr); 716 break; 717 } 718 719 num_copies = 720 btrfs_num_copies(state->root->fs_info, 721 next_bytenr, state->metablock_size); 722 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES) 723 pr_info("num_copies(log_bytenr=%llu) = %d\n", 724 next_bytenr, num_copies); 725 726 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) { 727 struct btrfsic_block *next_block; 728 struct btrfsic_block_data_ctx tmp_next_block_ctx; 729 struct btrfsic_block_link *l; 730 731 ret = btrfsic_map_block(state, next_bytenr, 732 state->metablock_size, 733 &tmp_next_block_ctx, 734 mirror_num); 735 if (ret) { 736 pr_info("btrfsic: btrfsic_map_block(root @%llu, mirror %d) failed!\n", 737 next_bytenr, mirror_num); 738 kfree(selected_super); 739 return -1; 740 } 741 742 next_block = btrfsic_block_hashtable_lookup( 743 tmp_next_block_ctx.dev->bdev, 744 tmp_next_block_ctx.dev_bytenr, 745 &state->block_hashtable); 746 BUG_ON(NULL == next_block); 747 748 l = btrfsic_block_link_hashtable_lookup( 749 tmp_next_block_ctx.dev->bdev, 750 tmp_next_block_ctx.dev_bytenr, 751 state->latest_superblock->dev_state-> 752 bdev, 753 state->latest_superblock->dev_bytenr, 754 &state->block_link_hashtable); 755 BUG_ON(NULL == l); 756 757 ret = btrfsic_read_block(state, &tmp_next_block_ctx); 758 if (ret < (int)PAGE_SIZE) { 759 pr_info("btrfsic: read @logical %llu failed!\n", 760 tmp_next_block_ctx.start); 761 btrfsic_release_block_ctx(&tmp_next_block_ctx); 762 kfree(selected_super); 763 return -1; 764 } 765 766 ret = btrfsic_process_metablock(state, 767 next_block, 768 &tmp_next_block_ctx, 769 BTRFS_MAX_LEVEL + 3, 1); 770 btrfsic_release_block_ctx(&tmp_next_block_ctx); 771 } 772 } 773 774 kfree(selected_super); 775 return ret; 776 } 777 778 static int btrfsic_process_superblock_dev_mirror( 779 struct btrfsic_state *state, 780 struct btrfsic_dev_state *dev_state, 781 struct btrfs_device *device, 782 int superblock_mirror_num, 783 struct btrfsic_dev_state **selected_dev_state, 784 struct btrfs_super_block *selected_super) 785 { 786 struct btrfs_super_block *super_tmp; 787 u64 dev_bytenr; 788 struct buffer_head *bh; 789 struct btrfsic_block *superblock_tmp; 790 int pass; 791 struct block_device *const superblock_bdev = device->bdev; 792 793 /* super block bytenr is always the unmapped device bytenr */ 794 dev_bytenr = btrfs_sb_offset(superblock_mirror_num); 795 if (dev_bytenr + BTRFS_SUPER_INFO_SIZE > device->commit_total_bytes) 796 return -1; 797 bh = __bread(superblock_bdev, dev_bytenr / 4096, 798 BTRFS_SUPER_INFO_SIZE); 799 if (NULL == bh) 800 return -1; 801 super_tmp = (struct btrfs_super_block *) 802 (bh->b_data + (dev_bytenr & 4095)); 803 804 if (btrfs_super_bytenr(super_tmp) != dev_bytenr || 805 btrfs_super_magic(super_tmp) != BTRFS_MAGIC || 806 memcmp(device->uuid, super_tmp->dev_item.uuid, BTRFS_UUID_SIZE) || 807 btrfs_super_nodesize(super_tmp) != state->metablock_size || 808 btrfs_super_sectorsize(super_tmp) != state->datablock_size) { 809 brelse(bh); 810 return 0; 811 } 812 813 superblock_tmp = 814 btrfsic_block_hashtable_lookup(superblock_bdev, 815 dev_bytenr, 816 &state->block_hashtable); 817 if (NULL == superblock_tmp) { 818 superblock_tmp = btrfsic_block_alloc(); 819 if (NULL == superblock_tmp) { 820 pr_info("btrfsic: error, kmalloc failed!\n"); 821 brelse(bh); 822 return -1; 823 } 824 /* for superblock, only the dev_bytenr makes sense */ 825 superblock_tmp->dev_bytenr = dev_bytenr; 826 superblock_tmp->dev_state = dev_state; 827 superblock_tmp->logical_bytenr = dev_bytenr; 828 superblock_tmp->generation = btrfs_super_generation(super_tmp); 829 superblock_tmp->is_metadata = 1; 830 superblock_tmp->is_superblock = 1; 831 superblock_tmp->is_iodone = 1; 832 superblock_tmp->never_written = 0; 833 superblock_tmp->mirror_num = 1 + superblock_mirror_num; 834 if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE) 835 btrfs_info_in_rcu(device->dev_root->fs_info, 836 "new initial S-block (bdev %p, %s) @%llu (%s/%llu/%d)", 837 superblock_bdev, 838 rcu_str_deref(device->name), dev_bytenr, 839 dev_state->name, dev_bytenr, 840 superblock_mirror_num); 841 list_add(&superblock_tmp->all_blocks_node, 842 &state->all_blocks_list); 843 btrfsic_block_hashtable_add(superblock_tmp, 844 &state->block_hashtable); 845 } 846 847 /* select the one with the highest generation field */ 848 if (btrfs_super_generation(super_tmp) > 849 state->max_superblock_generation || 850 0 == state->max_superblock_generation) { 851 memcpy(selected_super, super_tmp, sizeof(*selected_super)); 852 *selected_dev_state = dev_state; 853 state->max_superblock_generation = 854 btrfs_super_generation(super_tmp); 855 state->latest_superblock = superblock_tmp; 856 } 857 858 for (pass = 0; pass < 3; pass++) { 859 u64 next_bytenr; 860 int num_copies; 861 int mirror_num; 862 const char *additional_string = NULL; 863 struct btrfs_disk_key tmp_disk_key; 864 865 tmp_disk_key.type = BTRFS_ROOT_ITEM_KEY; 866 tmp_disk_key.offset = 0; 867 switch (pass) { 868 case 0: 869 btrfs_set_disk_key_objectid(&tmp_disk_key, 870 BTRFS_ROOT_TREE_OBJECTID); 871 additional_string = "initial root "; 872 next_bytenr = btrfs_super_root(super_tmp); 873 break; 874 case 1: 875 btrfs_set_disk_key_objectid(&tmp_disk_key, 876 BTRFS_CHUNK_TREE_OBJECTID); 877 additional_string = "initial chunk "; 878 next_bytenr = btrfs_super_chunk_root(super_tmp); 879 break; 880 case 2: 881 btrfs_set_disk_key_objectid(&tmp_disk_key, 882 BTRFS_TREE_LOG_OBJECTID); 883 additional_string = "initial log "; 884 next_bytenr = btrfs_super_log_root(super_tmp); 885 if (0 == next_bytenr) 886 continue; 887 break; 888 } 889 890 num_copies = 891 btrfs_num_copies(state->root->fs_info, 892 next_bytenr, state->metablock_size); 893 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES) 894 pr_info("num_copies(log_bytenr=%llu) = %d\n", 895 next_bytenr, num_copies); 896 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) { 897 struct btrfsic_block *next_block; 898 struct btrfsic_block_data_ctx tmp_next_block_ctx; 899 struct btrfsic_block_link *l; 900 901 if (btrfsic_map_block(state, next_bytenr, 902 state->metablock_size, 903 &tmp_next_block_ctx, 904 mirror_num)) { 905 pr_info("btrfsic: btrfsic_map_block(bytenr @%llu, mirror %d) failed!\n", 906 next_bytenr, mirror_num); 907 brelse(bh); 908 return -1; 909 } 910 911 next_block = btrfsic_block_lookup_or_add( 912 state, &tmp_next_block_ctx, 913 additional_string, 1, 1, 0, 914 mirror_num, NULL); 915 if (NULL == next_block) { 916 btrfsic_release_block_ctx(&tmp_next_block_ctx); 917 brelse(bh); 918 return -1; 919 } 920 921 next_block->disk_key = tmp_disk_key; 922 next_block->generation = BTRFSIC_GENERATION_UNKNOWN; 923 l = btrfsic_block_link_lookup_or_add( 924 state, &tmp_next_block_ctx, 925 next_block, superblock_tmp, 926 BTRFSIC_GENERATION_UNKNOWN); 927 btrfsic_release_block_ctx(&tmp_next_block_ctx); 928 if (NULL == l) { 929 brelse(bh); 930 return -1; 931 } 932 } 933 } 934 if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_ALL_TREES) 935 btrfsic_dump_tree_sub(state, superblock_tmp, 0); 936 937 brelse(bh); 938 return 0; 939 } 940 941 static struct btrfsic_stack_frame *btrfsic_stack_frame_alloc(void) 942 { 943 struct btrfsic_stack_frame *sf; 944 945 sf = kzalloc(sizeof(*sf), GFP_NOFS); 946 if (NULL == sf) 947 pr_info("btrfsic: alloc memory failed!\n"); 948 else 949 sf->magic = BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER; 950 return sf; 951 } 952 953 static void btrfsic_stack_frame_free(struct btrfsic_stack_frame *sf) 954 { 955 BUG_ON(!(NULL == sf || 956 BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER == sf->magic)); 957 kfree(sf); 958 } 959 960 static int btrfsic_process_metablock( 961 struct btrfsic_state *state, 962 struct btrfsic_block *const first_block, 963 struct btrfsic_block_data_ctx *const first_block_ctx, 964 int first_limit_nesting, int force_iodone_flag) 965 { 966 struct btrfsic_stack_frame initial_stack_frame = { 0 }; 967 struct btrfsic_stack_frame *sf; 968 struct btrfsic_stack_frame *next_stack; 969 struct btrfs_header *const first_hdr = 970 (struct btrfs_header *)first_block_ctx->datav[0]; 971 972 BUG_ON(!first_hdr); 973 sf = &initial_stack_frame; 974 sf->error = 0; 975 sf->i = -1; 976 sf->limit_nesting = first_limit_nesting; 977 sf->block = first_block; 978 sf->block_ctx = first_block_ctx; 979 sf->next_block = NULL; 980 sf->hdr = first_hdr; 981 sf->prev = NULL; 982 983 continue_with_new_stack_frame: 984 sf->block->generation = le64_to_cpu(sf->hdr->generation); 985 if (0 == sf->hdr->level) { 986 struct btrfs_leaf *const leafhdr = 987 (struct btrfs_leaf *)sf->hdr; 988 989 if (-1 == sf->i) { 990 sf->nr = btrfs_stack_header_nritems(&leafhdr->header); 991 992 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 993 pr_info("leaf %llu items %d generation %llu owner %llu\n", 994 sf->block_ctx->start, sf->nr, 995 btrfs_stack_header_generation( 996 &leafhdr->header), 997 btrfs_stack_header_owner( 998 &leafhdr->header)); 999 } 1000 1001 continue_with_current_leaf_stack_frame: 1002 if (0 == sf->num_copies || sf->mirror_num > sf->num_copies) { 1003 sf->i++; 1004 sf->num_copies = 0; 1005 } 1006 1007 if (sf->i < sf->nr) { 1008 struct btrfs_item disk_item; 1009 u32 disk_item_offset = 1010 (uintptr_t)(leafhdr->items + sf->i) - 1011 (uintptr_t)leafhdr; 1012 struct btrfs_disk_key *disk_key; 1013 u8 type; 1014 u32 item_offset; 1015 u32 item_size; 1016 1017 if (disk_item_offset + sizeof(struct btrfs_item) > 1018 sf->block_ctx->len) { 1019 leaf_item_out_of_bounce_error: 1020 pr_info("btrfsic: leaf item out of bounce at logical %llu, dev %s\n", 1021 sf->block_ctx->start, 1022 sf->block_ctx->dev->name); 1023 goto one_stack_frame_backwards; 1024 } 1025 btrfsic_read_from_block_data(sf->block_ctx, 1026 &disk_item, 1027 disk_item_offset, 1028 sizeof(struct btrfs_item)); 1029 item_offset = btrfs_stack_item_offset(&disk_item); 1030 item_size = btrfs_stack_item_size(&disk_item); 1031 disk_key = &disk_item.key; 1032 type = btrfs_disk_key_type(disk_key); 1033 1034 if (BTRFS_ROOT_ITEM_KEY == type) { 1035 struct btrfs_root_item root_item; 1036 u32 root_item_offset; 1037 u64 next_bytenr; 1038 1039 root_item_offset = item_offset + 1040 offsetof(struct btrfs_leaf, items); 1041 if (root_item_offset + item_size > 1042 sf->block_ctx->len) 1043 goto leaf_item_out_of_bounce_error; 1044 btrfsic_read_from_block_data( 1045 sf->block_ctx, &root_item, 1046 root_item_offset, 1047 item_size); 1048 next_bytenr = btrfs_root_bytenr(&root_item); 1049 1050 sf->error = 1051 btrfsic_create_link_to_next_block( 1052 state, 1053 sf->block, 1054 sf->block_ctx, 1055 next_bytenr, 1056 sf->limit_nesting, 1057 &sf->next_block_ctx, 1058 &sf->next_block, 1059 force_iodone_flag, 1060 &sf->num_copies, 1061 &sf->mirror_num, 1062 disk_key, 1063 btrfs_root_generation( 1064 &root_item)); 1065 if (sf->error) 1066 goto one_stack_frame_backwards; 1067 1068 if (NULL != sf->next_block) { 1069 struct btrfs_header *const next_hdr = 1070 (struct btrfs_header *) 1071 sf->next_block_ctx.datav[0]; 1072 1073 next_stack = 1074 btrfsic_stack_frame_alloc(); 1075 if (NULL == next_stack) { 1076 sf->error = -1; 1077 btrfsic_release_block_ctx( 1078 &sf-> 1079 next_block_ctx); 1080 goto one_stack_frame_backwards; 1081 } 1082 1083 next_stack->i = -1; 1084 next_stack->block = sf->next_block; 1085 next_stack->block_ctx = 1086 &sf->next_block_ctx; 1087 next_stack->next_block = NULL; 1088 next_stack->hdr = next_hdr; 1089 next_stack->limit_nesting = 1090 sf->limit_nesting - 1; 1091 next_stack->prev = sf; 1092 sf = next_stack; 1093 goto continue_with_new_stack_frame; 1094 } 1095 } else if (BTRFS_EXTENT_DATA_KEY == type && 1096 state->include_extent_data) { 1097 sf->error = btrfsic_handle_extent_data( 1098 state, 1099 sf->block, 1100 sf->block_ctx, 1101 item_offset, 1102 force_iodone_flag); 1103 if (sf->error) 1104 goto one_stack_frame_backwards; 1105 } 1106 1107 goto continue_with_current_leaf_stack_frame; 1108 } 1109 } else { 1110 struct btrfs_node *const nodehdr = (struct btrfs_node *)sf->hdr; 1111 1112 if (-1 == sf->i) { 1113 sf->nr = btrfs_stack_header_nritems(&nodehdr->header); 1114 1115 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 1116 pr_info("node %llu level %d items %d generation %llu owner %llu\n", 1117 sf->block_ctx->start, 1118 nodehdr->header.level, sf->nr, 1119 btrfs_stack_header_generation( 1120 &nodehdr->header), 1121 btrfs_stack_header_owner( 1122 &nodehdr->header)); 1123 } 1124 1125 continue_with_current_node_stack_frame: 1126 if (0 == sf->num_copies || sf->mirror_num > sf->num_copies) { 1127 sf->i++; 1128 sf->num_copies = 0; 1129 } 1130 1131 if (sf->i < sf->nr) { 1132 struct btrfs_key_ptr key_ptr; 1133 u32 key_ptr_offset; 1134 u64 next_bytenr; 1135 1136 key_ptr_offset = (uintptr_t)(nodehdr->ptrs + sf->i) - 1137 (uintptr_t)nodehdr; 1138 if (key_ptr_offset + sizeof(struct btrfs_key_ptr) > 1139 sf->block_ctx->len) { 1140 pr_info("btrfsic: node item out of bounce at logical %llu, dev %s\n", 1141 sf->block_ctx->start, 1142 sf->block_ctx->dev->name); 1143 goto one_stack_frame_backwards; 1144 } 1145 btrfsic_read_from_block_data( 1146 sf->block_ctx, &key_ptr, key_ptr_offset, 1147 sizeof(struct btrfs_key_ptr)); 1148 next_bytenr = btrfs_stack_key_blockptr(&key_ptr); 1149 1150 sf->error = btrfsic_create_link_to_next_block( 1151 state, 1152 sf->block, 1153 sf->block_ctx, 1154 next_bytenr, 1155 sf->limit_nesting, 1156 &sf->next_block_ctx, 1157 &sf->next_block, 1158 force_iodone_flag, 1159 &sf->num_copies, 1160 &sf->mirror_num, 1161 &key_ptr.key, 1162 btrfs_stack_key_generation(&key_ptr)); 1163 if (sf->error) 1164 goto one_stack_frame_backwards; 1165 1166 if (NULL != sf->next_block) { 1167 struct btrfs_header *const next_hdr = 1168 (struct btrfs_header *) 1169 sf->next_block_ctx.datav[0]; 1170 1171 next_stack = btrfsic_stack_frame_alloc(); 1172 if (NULL == next_stack) { 1173 sf->error = -1; 1174 goto one_stack_frame_backwards; 1175 } 1176 1177 next_stack->i = -1; 1178 next_stack->block = sf->next_block; 1179 next_stack->block_ctx = &sf->next_block_ctx; 1180 next_stack->next_block = NULL; 1181 next_stack->hdr = next_hdr; 1182 next_stack->limit_nesting = 1183 sf->limit_nesting - 1; 1184 next_stack->prev = sf; 1185 sf = next_stack; 1186 goto continue_with_new_stack_frame; 1187 } 1188 1189 goto continue_with_current_node_stack_frame; 1190 } 1191 } 1192 1193 one_stack_frame_backwards: 1194 if (NULL != sf->prev) { 1195 struct btrfsic_stack_frame *const prev = sf->prev; 1196 1197 /* the one for the initial block is freed in the caller */ 1198 btrfsic_release_block_ctx(sf->block_ctx); 1199 1200 if (sf->error) { 1201 prev->error = sf->error; 1202 btrfsic_stack_frame_free(sf); 1203 sf = prev; 1204 goto one_stack_frame_backwards; 1205 } 1206 1207 btrfsic_stack_frame_free(sf); 1208 sf = prev; 1209 goto continue_with_new_stack_frame; 1210 } else { 1211 BUG_ON(&initial_stack_frame != sf); 1212 } 1213 1214 return sf->error; 1215 } 1216 1217 static void btrfsic_read_from_block_data( 1218 struct btrfsic_block_data_ctx *block_ctx, 1219 void *dstv, u32 offset, size_t len) 1220 { 1221 size_t cur; 1222 size_t offset_in_page; 1223 char *kaddr; 1224 char *dst = (char *)dstv; 1225 size_t start_offset = block_ctx->start & ((u64)PAGE_SIZE - 1); 1226 unsigned long i = (start_offset + offset) >> PAGE_SHIFT; 1227 1228 WARN_ON(offset + len > block_ctx->len); 1229 offset_in_page = (start_offset + offset) & (PAGE_SIZE - 1); 1230 1231 while (len > 0) { 1232 cur = min(len, ((size_t)PAGE_SIZE - offset_in_page)); 1233 BUG_ON(i >= DIV_ROUND_UP(block_ctx->len, PAGE_SIZE)); 1234 kaddr = block_ctx->datav[i]; 1235 memcpy(dst, kaddr + offset_in_page, cur); 1236 1237 dst += cur; 1238 len -= cur; 1239 offset_in_page = 0; 1240 i++; 1241 } 1242 } 1243 1244 static int btrfsic_create_link_to_next_block( 1245 struct btrfsic_state *state, 1246 struct btrfsic_block *block, 1247 struct btrfsic_block_data_ctx *block_ctx, 1248 u64 next_bytenr, 1249 int limit_nesting, 1250 struct btrfsic_block_data_ctx *next_block_ctx, 1251 struct btrfsic_block **next_blockp, 1252 int force_iodone_flag, 1253 int *num_copiesp, int *mirror_nump, 1254 struct btrfs_disk_key *disk_key, 1255 u64 parent_generation) 1256 { 1257 struct btrfsic_block *next_block = NULL; 1258 int ret; 1259 struct btrfsic_block_link *l; 1260 int did_alloc_block_link; 1261 int block_was_created; 1262 1263 *next_blockp = NULL; 1264 if (0 == *num_copiesp) { 1265 *num_copiesp = 1266 btrfs_num_copies(state->root->fs_info, 1267 next_bytenr, state->metablock_size); 1268 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES) 1269 pr_info("num_copies(log_bytenr=%llu) = %d\n", 1270 next_bytenr, *num_copiesp); 1271 *mirror_nump = 1; 1272 } 1273 1274 if (*mirror_nump > *num_copiesp) 1275 return 0; 1276 1277 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 1278 pr_info("btrfsic_create_link_to_next_block(mirror_num=%d)\n", 1279 *mirror_nump); 1280 ret = btrfsic_map_block(state, next_bytenr, 1281 state->metablock_size, 1282 next_block_ctx, *mirror_nump); 1283 if (ret) { 1284 pr_info("btrfsic: btrfsic_map_block(@%llu, mirror=%d) failed!\n", 1285 next_bytenr, *mirror_nump); 1286 btrfsic_release_block_ctx(next_block_ctx); 1287 *next_blockp = NULL; 1288 return -1; 1289 } 1290 1291 next_block = btrfsic_block_lookup_or_add(state, 1292 next_block_ctx, "referenced ", 1293 1, force_iodone_flag, 1294 !force_iodone_flag, 1295 *mirror_nump, 1296 &block_was_created); 1297 if (NULL == next_block) { 1298 btrfsic_release_block_ctx(next_block_ctx); 1299 *next_blockp = NULL; 1300 return -1; 1301 } 1302 if (block_was_created) { 1303 l = NULL; 1304 next_block->generation = BTRFSIC_GENERATION_UNKNOWN; 1305 } else { 1306 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) { 1307 if (next_block->logical_bytenr != next_bytenr && 1308 !(!next_block->is_metadata && 1309 0 == next_block->logical_bytenr)) 1310 pr_info("Referenced block @%llu (%s/%llu/%d) found in hash table, %c, bytenr mismatch (!= stored %llu).\n", 1311 next_bytenr, next_block_ctx->dev->name, 1312 next_block_ctx->dev_bytenr, *mirror_nump, 1313 btrfsic_get_block_type(state, 1314 next_block), 1315 next_block->logical_bytenr); 1316 else 1317 pr_info("Referenced block @%llu (%s/%llu/%d) found in hash table, %c.\n", 1318 next_bytenr, next_block_ctx->dev->name, 1319 next_block_ctx->dev_bytenr, *mirror_nump, 1320 btrfsic_get_block_type(state, 1321 next_block)); 1322 } 1323 next_block->logical_bytenr = next_bytenr; 1324 1325 next_block->mirror_num = *mirror_nump; 1326 l = btrfsic_block_link_hashtable_lookup( 1327 next_block_ctx->dev->bdev, 1328 next_block_ctx->dev_bytenr, 1329 block_ctx->dev->bdev, 1330 block_ctx->dev_bytenr, 1331 &state->block_link_hashtable); 1332 } 1333 1334 next_block->disk_key = *disk_key; 1335 if (NULL == l) { 1336 l = btrfsic_block_link_alloc(); 1337 if (NULL == l) { 1338 pr_info("btrfsic: error, kmalloc failed!\n"); 1339 btrfsic_release_block_ctx(next_block_ctx); 1340 *next_blockp = NULL; 1341 return -1; 1342 } 1343 1344 did_alloc_block_link = 1; 1345 l->block_ref_to = next_block; 1346 l->block_ref_from = block; 1347 l->ref_cnt = 1; 1348 l->parent_generation = parent_generation; 1349 1350 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 1351 btrfsic_print_add_link(state, l); 1352 1353 list_add(&l->node_ref_to, &block->ref_to_list); 1354 list_add(&l->node_ref_from, &next_block->ref_from_list); 1355 1356 btrfsic_block_link_hashtable_add(l, 1357 &state->block_link_hashtable); 1358 } else { 1359 did_alloc_block_link = 0; 1360 if (0 == limit_nesting) { 1361 l->ref_cnt++; 1362 l->parent_generation = parent_generation; 1363 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 1364 btrfsic_print_add_link(state, l); 1365 } 1366 } 1367 1368 if (limit_nesting > 0 && did_alloc_block_link) { 1369 ret = btrfsic_read_block(state, next_block_ctx); 1370 if (ret < (int)next_block_ctx->len) { 1371 pr_info("btrfsic: read block @logical %llu failed!\n", 1372 next_bytenr); 1373 btrfsic_release_block_ctx(next_block_ctx); 1374 *next_blockp = NULL; 1375 return -1; 1376 } 1377 1378 *next_blockp = next_block; 1379 } else { 1380 *next_blockp = NULL; 1381 } 1382 (*mirror_nump)++; 1383 1384 return 0; 1385 } 1386 1387 static int btrfsic_handle_extent_data( 1388 struct btrfsic_state *state, 1389 struct btrfsic_block *block, 1390 struct btrfsic_block_data_ctx *block_ctx, 1391 u32 item_offset, int force_iodone_flag) 1392 { 1393 int ret; 1394 struct btrfs_file_extent_item file_extent_item; 1395 u64 file_extent_item_offset; 1396 u64 next_bytenr; 1397 u64 num_bytes; 1398 u64 generation; 1399 struct btrfsic_block_link *l; 1400 1401 file_extent_item_offset = offsetof(struct btrfs_leaf, items) + 1402 item_offset; 1403 if (file_extent_item_offset + 1404 offsetof(struct btrfs_file_extent_item, disk_num_bytes) > 1405 block_ctx->len) { 1406 pr_info("btrfsic: file item out of bounce at logical %llu, dev %s\n", 1407 block_ctx->start, block_ctx->dev->name); 1408 return -1; 1409 } 1410 1411 btrfsic_read_from_block_data(block_ctx, &file_extent_item, 1412 file_extent_item_offset, 1413 offsetof(struct btrfs_file_extent_item, disk_num_bytes)); 1414 if (BTRFS_FILE_EXTENT_REG != file_extent_item.type || 1415 btrfs_stack_file_extent_disk_bytenr(&file_extent_item) == 0) { 1416 if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE) 1417 pr_info("extent_data: type %u, disk_bytenr = %llu\n", 1418 file_extent_item.type, 1419 btrfs_stack_file_extent_disk_bytenr( 1420 &file_extent_item)); 1421 return 0; 1422 } 1423 1424 if (file_extent_item_offset + sizeof(struct btrfs_file_extent_item) > 1425 block_ctx->len) { 1426 pr_info("btrfsic: file item out of bounce at logical %llu, dev %s\n", 1427 block_ctx->start, block_ctx->dev->name); 1428 return -1; 1429 } 1430 btrfsic_read_from_block_data(block_ctx, &file_extent_item, 1431 file_extent_item_offset, 1432 sizeof(struct btrfs_file_extent_item)); 1433 next_bytenr = btrfs_stack_file_extent_disk_bytenr(&file_extent_item); 1434 if (btrfs_stack_file_extent_compression(&file_extent_item) == 1435 BTRFS_COMPRESS_NONE) { 1436 next_bytenr += btrfs_stack_file_extent_offset(&file_extent_item); 1437 num_bytes = btrfs_stack_file_extent_num_bytes(&file_extent_item); 1438 } else { 1439 num_bytes = btrfs_stack_file_extent_disk_num_bytes(&file_extent_item); 1440 } 1441 generation = btrfs_stack_file_extent_generation(&file_extent_item); 1442 1443 if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE) 1444 pr_info("extent_data: type %u, disk_bytenr = %llu, offset = %llu, num_bytes = %llu\n", 1445 file_extent_item.type, 1446 btrfs_stack_file_extent_disk_bytenr(&file_extent_item), 1447 btrfs_stack_file_extent_offset(&file_extent_item), 1448 num_bytes); 1449 while (num_bytes > 0) { 1450 u32 chunk_len; 1451 int num_copies; 1452 int mirror_num; 1453 1454 if (num_bytes > state->datablock_size) 1455 chunk_len = state->datablock_size; 1456 else 1457 chunk_len = num_bytes; 1458 1459 num_copies = 1460 btrfs_num_copies(state->root->fs_info, 1461 next_bytenr, state->datablock_size); 1462 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES) 1463 pr_info("num_copies(log_bytenr=%llu) = %d\n", 1464 next_bytenr, num_copies); 1465 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) { 1466 struct btrfsic_block_data_ctx next_block_ctx; 1467 struct btrfsic_block *next_block; 1468 int block_was_created; 1469 1470 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 1471 pr_info("btrfsic_handle_extent_data(mirror_num=%d)\n", 1472 mirror_num); 1473 if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE) 1474 pr_info("\tdisk_bytenr = %llu, num_bytes %u\n", 1475 next_bytenr, chunk_len); 1476 ret = btrfsic_map_block(state, next_bytenr, 1477 chunk_len, &next_block_ctx, 1478 mirror_num); 1479 if (ret) { 1480 pr_info("btrfsic: btrfsic_map_block(@%llu, mirror=%d) failed!\n", 1481 next_bytenr, mirror_num); 1482 return -1; 1483 } 1484 1485 next_block = btrfsic_block_lookup_or_add( 1486 state, 1487 &next_block_ctx, 1488 "referenced ", 1489 0, 1490 force_iodone_flag, 1491 !force_iodone_flag, 1492 mirror_num, 1493 &block_was_created); 1494 if (NULL == next_block) { 1495 pr_info("btrfsic: error, kmalloc failed!\n"); 1496 btrfsic_release_block_ctx(&next_block_ctx); 1497 return -1; 1498 } 1499 if (!block_was_created) { 1500 if ((state->print_mask & 1501 BTRFSIC_PRINT_MASK_VERBOSE) && 1502 next_block->logical_bytenr != next_bytenr && 1503 !(!next_block->is_metadata && 1504 0 == next_block->logical_bytenr)) { 1505 pr_info("Referenced block @%llu (%s/%llu/%d) found in hash table, D, bytenr mismatch (!= stored %llu).\n", 1506 next_bytenr, 1507 next_block_ctx.dev->name, 1508 next_block_ctx.dev_bytenr, 1509 mirror_num, 1510 next_block->logical_bytenr); 1511 } 1512 next_block->logical_bytenr = next_bytenr; 1513 next_block->mirror_num = mirror_num; 1514 } 1515 1516 l = btrfsic_block_link_lookup_or_add(state, 1517 &next_block_ctx, 1518 next_block, block, 1519 generation); 1520 btrfsic_release_block_ctx(&next_block_ctx); 1521 if (NULL == l) 1522 return -1; 1523 } 1524 1525 next_bytenr += chunk_len; 1526 num_bytes -= chunk_len; 1527 } 1528 1529 return 0; 1530 } 1531 1532 static int btrfsic_map_block(struct btrfsic_state *state, u64 bytenr, u32 len, 1533 struct btrfsic_block_data_ctx *block_ctx_out, 1534 int mirror_num) 1535 { 1536 int ret; 1537 u64 length; 1538 struct btrfs_bio *multi = NULL; 1539 struct btrfs_device *device; 1540 1541 length = len; 1542 ret = btrfs_map_block(state->root->fs_info, READ, 1543 bytenr, &length, &multi, mirror_num); 1544 1545 if (ret) { 1546 block_ctx_out->start = 0; 1547 block_ctx_out->dev_bytenr = 0; 1548 block_ctx_out->len = 0; 1549 block_ctx_out->dev = NULL; 1550 block_ctx_out->datav = NULL; 1551 block_ctx_out->pagev = NULL; 1552 block_ctx_out->mem_to_free = NULL; 1553 1554 return ret; 1555 } 1556 1557 device = multi->stripes[0].dev; 1558 block_ctx_out->dev = btrfsic_dev_state_lookup(device->bdev); 1559 block_ctx_out->dev_bytenr = multi->stripes[0].physical; 1560 block_ctx_out->start = bytenr; 1561 block_ctx_out->len = len; 1562 block_ctx_out->datav = NULL; 1563 block_ctx_out->pagev = NULL; 1564 block_ctx_out->mem_to_free = NULL; 1565 1566 kfree(multi); 1567 if (NULL == block_ctx_out->dev) { 1568 ret = -ENXIO; 1569 pr_info("btrfsic: error, cannot lookup dev (#1)!\n"); 1570 } 1571 1572 return ret; 1573 } 1574 1575 static void btrfsic_release_block_ctx(struct btrfsic_block_data_ctx *block_ctx) 1576 { 1577 if (block_ctx->mem_to_free) { 1578 unsigned int num_pages; 1579 1580 BUG_ON(!block_ctx->datav); 1581 BUG_ON(!block_ctx->pagev); 1582 num_pages = (block_ctx->len + (u64)PAGE_SIZE - 1) >> 1583 PAGE_SHIFT; 1584 while (num_pages > 0) { 1585 num_pages--; 1586 if (block_ctx->datav[num_pages]) { 1587 kunmap(block_ctx->pagev[num_pages]); 1588 block_ctx->datav[num_pages] = NULL; 1589 } 1590 if (block_ctx->pagev[num_pages]) { 1591 __free_page(block_ctx->pagev[num_pages]); 1592 block_ctx->pagev[num_pages] = NULL; 1593 } 1594 } 1595 1596 kfree(block_ctx->mem_to_free); 1597 block_ctx->mem_to_free = NULL; 1598 block_ctx->pagev = NULL; 1599 block_ctx->datav = NULL; 1600 } 1601 } 1602 1603 static int btrfsic_read_block(struct btrfsic_state *state, 1604 struct btrfsic_block_data_ctx *block_ctx) 1605 { 1606 unsigned int num_pages; 1607 unsigned int i; 1608 u64 dev_bytenr; 1609 int ret; 1610 1611 BUG_ON(block_ctx->datav); 1612 BUG_ON(block_ctx->pagev); 1613 BUG_ON(block_ctx->mem_to_free); 1614 if (block_ctx->dev_bytenr & ((u64)PAGE_SIZE - 1)) { 1615 pr_info("btrfsic: read_block() with unaligned bytenr %llu\n", 1616 block_ctx->dev_bytenr); 1617 return -1; 1618 } 1619 1620 num_pages = (block_ctx->len + (u64)PAGE_SIZE - 1) >> 1621 PAGE_SHIFT; 1622 block_ctx->mem_to_free = kzalloc((sizeof(*block_ctx->datav) + 1623 sizeof(*block_ctx->pagev)) * 1624 num_pages, GFP_NOFS); 1625 if (!block_ctx->mem_to_free) 1626 return -ENOMEM; 1627 block_ctx->datav = block_ctx->mem_to_free; 1628 block_ctx->pagev = (struct page **)(block_ctx->datav + num_pages); 1629 for (i = 0; i < num_pages; i++) { 1630 block_ctx->pagev[i] = alloc_page(GFP_NOFS); 1631 if (!block_ctx->pagev[i]) 1632 return -1; 1633 } 1634 1635 dev_bytenr = block_ctx->dev_bytenr; 1636 for (i = 0; i < num_pages;) { 1637 struct bio *bio; 1638 unsigned int j; 1639 1640 bio = btrfs_io_bio_alloc(GFP_NOFS, num_pages - i); 1641 if (!bio) { 1642 pr_info("btrfsic: bio_alloc() for %u pages failed!\n", 1643 num_pages - i); 1644 return -1; 1645 } 1646 bio->bi_bdev = block_ctx->dev->bdev; 1647 bio->bi_iter.bi_sector = dev_bytenr >> 9; 1648 bio_set_op_attrs(bio, REQ_OP_READ, 0); 1649 1650 for (j = i; j < num_pages; j++) { 1651 ret = bio_add_page(bio, block_ctx->pagev[j], 1652 PAGE_SIZE, 0); 1653 if (PAGE_SIZE != ret) 1654 break; 1655 } 1656 if (j == i) { 1657 pr_info("btrfsic: error, failed to add a single page!\n"); 1658 return -1; 1659 } 1660 if (submit_bio_wait(bio)) { 1661 pr_info("btrfsic: read error at logical %llu dev %s!\n", 1662 block_ctx->start, block_ctx->dev->name); 1663 bio_put(bio); 1664 return -1; 1665 } 1666 bio_put(bio); 1667 dev_bytenr += (j - i) * PAGE_SIZE; 1668 i = j; 1669 } 1670 for (i = 0; i < num_pages; i++) { 1671 block_ctx->datav[i] = kmap(block_ctx->pagev[i]); 1672 if (!block_ctx->datav[i]) { 1673 pr_info("btrfsic: kmap() failed (dev %s)!\n", 1674 block_ctx->dev->name); 1675 return -1; 1676 } 1677 } 1678 1679 return block_ctx->len; 1680 } 1681 1682 static void btrfsic_dump_database(struct btrfsic_state *state) 1683 { 1684 const struct btrfsic_block *b_all; 1685 1686 BUG_ON(NULL == state); 1687 1688 pr_info("all_blocks_list:\n"); 1689 list_for_each_entry(b_all, &state->all_blocks_list, all_blocks_node) { 1690 const struct btrfsic_block_link *l; 1691 1692 pr_info("%c-block @%llu (%s/%llu/%d)\n", 1693 btrfsic_get_block_type(state, b_all), 1694 b_all->logical_bytenr, b_all->dev_state->name, 1695 b_all->dev_bytenr, b_all->mirror_num); 1696 1697 list_for_each_entry(l, &b_all->ref_to_list, node_ref_to) { 1698 pr_info(" %c @%llu (%s/%llu/%d) refers %u* to %c @%llu (%s/%llu/%d)\n", 1699 btrfsic_get_block_type(state, b_all), 1700 b_all->logical_bytenr, b_all->dev_state->name, 1701 b_all->dev_bytenr, b_all->mirror_num, 1702 l->ref_cnt, 1703 btrfsic_get_block_type(state, l->block_ref_to), 1704 l->block_ref_to->logical_bytenr, 1705 l->block_ref_to->dev_state->name, 1706 l->block_ref_to->dev_bytenr, 1707 l->block_ref_to->mirror_num); 1708 } 1709 1710 list_for_each_entry(l, &b_all->ref_from_list, node_ref_from) { 1711 pr_info(" %c @%llu (%s/%llu/%d) is ref %u* from %c @%llu (%s/%llu/%d)\n", 1712 btrfsic_get_block_type(state, b_all), 1713 b_all->logical_bytenr, b_all->dev_state->name, 1714 b_all->dev_bytenr, b_all->mirror_num, 1715 l->ref_cnt, 1716 btrfsic_get_block_type(state, l->block_ref_from), 1717 l->block_ref_from->logical_bytenr, 1718 l->block_ref_from->dev_state->name, 1719 l->block_ref_from->dev_bytenr, 1720 l->block_ref_from->mirror_num); 1721 } 1722 1723 pr_info("\n"); 1724 } 1725 } 1726 1727 /* 1728 * Test whether the disk block contains a tree block (leaf or node) 1729 * (note that this test fails for the super block) 1730 */ 1731 static int btrfsic_test_for_metadata(struct btrfsic_state *state, 1732 char **datav, unsigned int num_pages) 1733 { 1734 struct btrfs_header *h; 1735 u8 csum[BTRFS_CSUM_SIZE]; 1736 u32 crc = ~(u32)0; 1737 unsigned int i; 1738 1739 if (num_pages * PAGE_SIZE < state->metablock_size) 1740 return 1; /* not metadata */ 1741 num_pages = state->metablock_size >> PAGE_SHIFT; 1742 h = (struct btrfs_header *)datav[0]; 1743 1744 if (memcmp(h->fsid, state->root->fs_info->fsid, BTRFS_UUID_SIZE)) 1745 return 1; 1746 1747 for (i = 0; i < num_pages; i++) { 1748 u8 *data = i ? datav[i] : (datav[i] + BTRFS_CSUM_SIZE); 1749 size_t sublen = i ? PAGE_SIZE : 1750 (PAGE_SIZE - BTRFS_CSUM_SIZE); 1751 1752 crc = btrfs_crc32c(crc, data, sublen); 1753 } 1754 btrfs_csum_final(crc, csum); 1755 if (memcmp(csum, h->csum, state->csum_size)) 1756 return 1; 1757 1758 return 0; /* is metadata */ 1759 } 1760 1761 static void btrfsic_process_written_block(struct btrfsic_dev_state *dev_state, 1762 u64 dev_bytenr, char **mapped_datav, 1763 unsigned int num_pages, 1764 struct bio *bio, int *bio_is_patched, 1765 struct buffer_head *bh, 1766 int submit_bio_bh_rw) 1767 { 1768 int is_metadata; 1769 struct btrfsic_block *block; 1770 struct btrfsic_block_data_ctx block_ctx; 1771 int ret; 1772 struct btrfsic_state *state = dev_state->state; 1773 struct block_device *bdev = dev_state->bdev; 1774 unsigned int processed_len; 1775 1776 if (NULL != bio_is_patched) 1777 *bio_is_patched = 0; 1778 1779 again: 1780 if (num_pages == 0) 1781 return; 1782 1783 processed_len = 0; 1784 is_metadata = (0 == btrfsic_test_for_metadata(state, mapped_datav, 1785 num_pages)); 1786 1787 block = btrfsic_block_hashtable_lookup(bdev, dev_bytenr, 1788 &state->block_hashtable); 1789 if (NULL != block) { 1790 u64 bytenr = 0; 1791 struct btrfsic_block_link *l, *tmp; 1792 1793 if (block->is_superblock) { 1794 bytenr = btrfs_super_bytenr((struct btrfs_super_block *) 1795 mapped_datav[0]); 1796 if (num_pages * PAGE_SIZE < 1797 BTRFS_SUPER_INFO_SIZE) { 1798 pr_info("btrfsic: cannot work with too short bios!\n"); 1799 return; 1800 } 1801 is_metadata = 1; 1802 BUG_ON(BTRFS_SUPER_INFO_SIZE & (PAGE_SIZE - 1)); 1803 processed_len = BTRFS_SUPER_INFO_SIZE; 1804 if (state->print_mask & 1805 BTRFSIC_PRINT_MASK_TREE_BEFORE_SB_WRITE) { 1806 pr_info("[before new superblock is written]:\n"); 1807 btrfsic_dump_tree_sub(state, block, 0); 1808 } 1809 } 1810 if (is_metadata) { 1811 if (!block->is_superblock) { 1812 if (num_pages * PAGE_SIZE < 1813 state->metablock_size) { 1814 pr_info("btrfsic: cannot work with too short bios!\n"); 1815 return; 1816 } 1817 processed_len = state->metablock_size; 1818 bytenr = btrfs_stack_header_bytenr( 1819 (struct btrfs_header *) 1820 mapped_datav[0]); 1821 btrfsic_cmp_log_and_dev_bytenr(state, bytenr, 1822 dev_state, 1823 dev_bytenr); 1824 } 1825 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) { 1826 if (block->logical_bytenr != bytenr && 1827 !(!block->is_metadata && 1828 block->logical_bytenr == 0)) 1829 pr_info("Written block @%llu (%s/%llu/%d) found in hash table, %c, bytenr mismatch (!= stored %llu).\n", 1830 bytenr, dev_state->name, 1831 dev_bytenr, 1832 block->mirror_num, 1833 btrfsic_get_block_type(state, 1834 block), 1835 block->logical_bytenr); 1836 else 1837 pr_info("Written block @%llu (%s/%llu/%d) found in hash table, %c.\n", 1838 bytenr, dev_state->name, 1839 dev_bytenr, block->mirror_num, 1840 btrfsic_get_block_type(state, 1841 block)); 1842 } 1843 block->logical_bytenr = bytenr; 1844 } else { 1845 if (num_pages * PAGE_SIZE < 1846 state->datablock_size) { 1847 pr_info("btrfsic: cannot work with too short bios!\n"); 1848 return; 1849 } 1850 processed_len = state->datablock_size; 1851 bytenr = block->logical_bytenr; 1852 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 1853 pr_info("Written block @%llu (%s/%llu/%d) found in hash table, %c.\n", 1854 bytenr, dev_state->name, dev_bytenr, 1855 block->mirror_num, 1856 btrfsic_get_block_type(state, block)); 1857 } 1858 1859 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 1860 pr_info("ref_to_list: %cE, ref_from_list: %cE\n", 1861 list_empty(&block->ref_to_list) ? ' ' : '!', 1862 list_empty(&block->ref_from_list) ? ' ' : '!'); 1863 if (btrfsic_is_block_ref_by_superblock(state, block, 0)) { 1864 pr_info("btrfs: attempt to overwrite %c-block @%llu (%s/%llu/%d), old(gen=%llu, objectid=%llu, type=%d, offset=%llu), new(gen=%llu), which is referenced by most recent superblock (superblockgen=%llu)!\n", 1865 btrfsic_get_block_type(state, block), bytenr, 1866 dev_state->name, dev_bytenr, block->mirror_num, 1867 block->generation, 1868 btrfs_disk_key_objectid(&block->disk_key), 1869 block->disk_key.type, 1870 btrfs_disk_key_offset(&block->disk_key), 1871 btrfs_stack_header_generation( 1872 (struct btrfs_header *) mapped_datav[0]), 1873 state->max_superblock_generation); 1874 btrfsic_dump_tree(state); 1875 } 1876 1877 if (!block->is_iodone && !block->never_written) { 1878 pr_info("btrfs: attempt to overwrite %c-block @%llu (%s/%llu/%d), oldgen=%llu, newgen=%llu, which is not yet iodone!\n", 1879 btrfsic_get_block_type(state, block), bytenr, 1880 dev_state->name, dev_bytenr, block->mirror_num, 1881 block->generation, 1882 btrfs_stack_header_generation( 1883 (struct btrfs_header *) 1884 mapped_datav[0])); 1885 /* it would not be safe to go on */ 1886 btrfsic_dump_tree(state); 1887 goto continue_loop; 1888 } 1889 1890 /* 1891 * Clear all references of this block. Do not free 1892 * the block itself even if is not referenced anymore 1893 * because it still carries valuable information 1894 * like whether it was ever written and IO completed. 1895 */ 1896 list_for_each_entry_safe(l, tmp, &block->ref_to_list, 1897 node_ref_to) { 1898 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 1899 btrfsic_print_rem_link(state, l); 1900 l->ref_cnt--; 1901 if (0 == l->ref_cnt) { 1902 list_del(&l->node_ref_to); 1903 list_del(&l->node_ref_from); 1904 btrfsic_block_link_hashtable_remove(l); 1905 btrfsic_block_link_free(l); 1906 } 1907 } 1908 1909 block_ctx.dev = dev_state; 1910 block_ctx.dev_bytenr = dev_bytenr; 1911 block_ctx.start = bytenr; 1912 block_ctx.len = processed_len; 1913 block_ctx.pagev = NULL; 1914 block_ctx.mem_to_free = NULL; 1915 block_ctx.datav = mapped_datav; 1916 1917 if (is_metadata || state->include_extent_data) { 1918 block->never_written = 0; 1919 block->iodone_w_error = 0; 1920 if (NULL != bio) { 1921 block->is_iodone = 0; 1922 BUG_ON(NULL == bio_is_patched); 1923 if (!*bio_is_patched) { 1924 block->orig_bio_bh_private = 1925 bio->bi_private; 1926 block->orig_bio_bh_end_io.bio = 1927 bio->bi_end_io; 1928 block->next_in_same_bio = NULL; 1929 bio->bi_private = block; 1930 bio->bi_end_io = btrfsic_bio_end_io; 1931 *bio_is_patched = 1; 1932 } else { 1933 struct btrfsic_block *chained_block = 1934 (struct btrfsic_block *) 1935 bio->bi_private; 1936 1937 BUG_ON(NULL == chained_block); 1938 block->orig_bio_bh_private = 1939 chained_block->orig_bio_bh_private; 1940 block->orig_bio_bh_end_io.bio = 1941 chained_block->orig_bio_bh_end_io. 1942 bio; 1943 block->next_in_same_bio = chained_block; 1944 bio->bi_private = block; 1945 } 1946 } else if (NULL != bh) { 1947 block->is_iodone = 0; 1948 block->orig_bio_bh_private = bh->b_private; 1949 block->orig_bio_bh_end_io.bh = bh->b_end_io; 1950 block->next_in_same_bio = NULL; 1951 bh->b_private = block; 1952 bh->b_end_io = btrfsic_bh_end_io; 1953 } else { 1954 block->is_iodone = 1; 1955 block->orig_bio_bh_private = NULL; 1956 block->orig_bio_bh_end_io.bio = NULL; 1957 block->next_in_same_bio = NULL; 1958 } 1959 } 1960 1961 block->flush_gen = dev_state->last_flush_gen + 1; 1962 block->submit_bio_bh_rw = submit_bio_bh_rw; 1963 if (is_metadata) { 1964 block->logical_bytenr = bytenr; 1965 block->is_metadata = 1; 1966 if (block->is_superblock) { 1967 BUG_ON(PAGE_SIZE != 1968 BTRFS_SUPER_INFO_SIZE); 1969 ret = btrfsic_process_written_superblock( 1970 state, 1971 block, 1972 (struct btrfs_super_block *) 1973 mapped_datav[0]); 1974 if (state->print_mask & 1975 BTRFSIC_PRINT_MASK_TREE_AFTER_SB_WRITE) { 1976 pr_info("[after new superblock is written]:\n"); 1977 btrfsic_dump_tree_sub(state, block, 0); 1978 } 1979 } else { 1980 block->mirror_num = 0; /* unknown */ 1981 ret = btrfsic_process_metablock( 1982 state, 1983 block, 1984 &block_ctx, 1985 0, 0); 1986 } 1987 if (ret) 1988 pr_info("btrfsic: btrfsic_process_metablock(root @%llu) failed!\n", 1989 dev_bytenr); 1990 } else { 1991 block->is_metadata = 0; 1992 block->mirror_num = 0; /* unknown */ 1993 block->generation = BTRFSIC_GENERATION_UNKNOWN; 1994 if (!state->include_extent_data 1995 && list_empty(&block->ref_from_list)) { 1996 /* 1997 * disk block is overwritten with extent 1998 * data (not meta data) and we are configured 1999 * to not include extent data: take the 2000 * chance and free the block's memory 2001 */ 2002 btrfsic_block_hashtable_remove(block); 2003 list_del(&block->all_blocks_node); 2004 btrfsic_block_free(block); 2005 } 2006 } 2007 btrfsic_release_block_ctx(&block_ctx); 2008 } else { 2009 /* block has not been found in hash table */ 2010 u64 bytenr; 2011 2012 if (!is_metadata) { 2013 processed_len = state->datablock_size; 2014 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2015 pr_info("Written block (%s/%llu/?) !found in hash table, D.\n", 2016 dev_state->name, dev_bytenr); 2017 if (!state->include_extent_data) { 2018 /* ignore that written D block */ 2019 goto continue_loop; 2020 } 2021 2022 /* this is getting ugly for the 2023 * include_extent_data case... */ 2024 bytenr = 0; /* unknown */ 2025 } else { 2026 processed_len = state->metablock_size; 2027 bytenr = btrfs_stack_header_bytenr( 2028 (struct btrfs_header *) 2029 mapped_datav[0]); 2030 btrfsic_cmp_log_and_dev_bytenr(state, bytenr, dev_state, 2031 dev_bytenr); 2032 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2033 pr_info("Written block @%llu (%s/%llu/?) !found in hash table, M.\n", 2034 bytenr, dev_state->name, dev_bytenr); 2035 } 2036 2037 block_ctx.dev = dev_state; 2038 block_ctx.dev_bytenr = dev_bytenr; 2039 block_ctx.start = bytenr; 2040 block_ctx.len = processed_len; 2041 block_ctx.pagev = NULL; 2042 block_ctx.mem_to_free = NULL; 2043 block_ctx.datav = mapped_datav; 2044 2045 block = btrfsic_block_alloc(); 2046 if (NULL == block) { 2047 pr_info("btrfsic: error, kmalloc failed!\n"); 2048 btrfsic_release_block_ctx(&block_ctx); 2049 goto continue_loop; 2050 } 2051 block->dev_state = dev_state; 2052 block->dev_bytenr = dev_bytenr; 2053 block->logical_bytenr = bytenr; 2054 block->is_metadata = is_metadata; 2055 block->never_written = 0; 2056 block->iodone_w_error = 0; 2057 block->mirror_num = 0; /* unknown */ 2058 block->flush_gen = dev_state->last_flush_gen + 1; 2059 block->submit_bio_bh_rw = submit_bio_bh_rw; 2060 if (NULL != bio) { 2061 block->is_iodone = 0; 2062 BUG_ON(NULL == bio_is_patched); 2063 if (!*bio_is_patched) { 2064 block->orig_bio_bh_private = bio->bi_private; 2065 block->orig_bio_bh_end_io.bio = bio->bi_end_io; 2066 block->next_in_same_bio = NULL; 2067 bio->bi_private = block; 2068 bio->bi_end_io = btrfsic_bio_end_io; 2069 *bio_is_patched = 1; 2070 } else { 2071 struct btrfsic_block *chained_block = 2072 (struct btrfsic_block *) 2073 bio->bi_private; 2074 2075 BUG_ON(NULL == chained_block); 2076 block->orig_bio_bh_private = 2077 chained_block->orig_bio_bh_private; 2078 block->orig_bio_bh_end_io.bio = 2079 chained_block->orig_bio_bh_end_io.bio; 2080 block->next_in_same_bio = chained_block; 2081 bio->bi_private = block; 2082 } 2083 } else if (NULL != bh) { 2084 block->is_iodone = 0; 2085 block->orig_bio_bh_private = bh->b_private; 2086 block->orig_bio_bh_end_io.bh = bh->b_end_io; 2087 block->next_in_same_bio = NULL; 2088 bh->b_private = block; 2089 bh->b_end_io = btrfsic_bh_end_io; 2090 } else { 2091 block->is_iodone = 1; 2092 block->orig_bio_bh_private = NULL; 2093 block->orig_bio_bh_end_io.bio = NULL; 2094 block->next_in_same_bio = NULL; 2095 } 2096 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2097 pr_info("New written %c-block @%llu (%s/%llu/%d)\n", 2098 is_metadata ? 'M' : 'D', 2099 block->logical_bytenr, block->dev_state->name, 2100 block->dev_bytenr, block->mirror_num); 2101 list_add(&block->all_blocks_node, &state->all_blocks_list); 2102 btrfsic_block_hashtable_add(block, &state->block_hashtable); 2103 2104 if (is_metadata) { 2105 ret = btrfsic_process_metablock(state, block, 2106 &block_ctx, 0, 0); 2107 if (ret) 2108 pr_info("btrfsic: process_metablock(root @%llu) failed!\n", 2109 dev_bytenr); 2110 } 2111 btrfsic_release_block_ctx(&block_ctx); 2112 } 2113 2114 continue_loop: 2115 BUG_ON(!processed_len); 2116 dev_bytenr += processed_len; 2117 mapped_datav += processed_len >> PAGE_SHIFT; 2118 num_pages -= processed_len >> PAGE_SHIFT; 2119 goto again; 2120 } 2121 2122 static void btrfsic_bio_end_io(struct bio *bp) 2123 { 2124 struct btrfsic_block *block = (struct btrfsic_block *)bp->bi_private; 2125 int iodone_w_error; 2126 2127 /* mutex is not held! This is not save if IO is not yet completed 2128 * on umount */ 2129 iodone_w_error = 0; 2130 if (bp->bi_error) 2131 iodone_w_error = 1; 2132 2133 BUG_ON(NULL == block); 2134 bp->bi_private = block->orig_bio_bh_private; 2135 bp->bi_end_io = block->orig_bio_bh_end_io.bio; 2136 2137 do { 2138 struct btrfsic_block *next_block; 2139 struct btrfsic_dev_state *const dev_state = block->dev_state; 2140 2141 if ((dev_state->state->print_mask & 2142 BTRFSIC_PRINT_MASK_END_IO_BIO_BH)) 2143 pr_info("bio_end_io(err=%d) for %c @%llu (%s/%llu/%d)\n", 2144 bp->bi_error, 2145 btrfsic_get_block_type(dev_state->state, block), 2146 block->logical_bytenr, dev_state->name, 2147 block->dev_bytenr, block->mirror_num); 2148 next_block = block->next_in_same_bio; 2149 block->iodone_w_error = iodone_w_error; 2150 if (block->submit_bio_bh_rw & REQ_PREFLUSH) { 2151 dev_state->last_flush_gen++; 2152 if ((dev_state->state->print_mask & 2153 BTRFSIC_PRINT_MASK_END_IO_BIO_BH)) 2154 pr_info("bio_end_io() new %s flush_gen=%llu\n", 2155 dev_state->name, 2156 dev_state->last_flush_gen); 2157 } 2158 if (block->submit_bio_bh_rw & REQ_FUA) 2159 block->flush_gen = 0; /* FUA completed means block is 2160 * on disk */ 2161 block->is_iodone = 1; /* for FLUSH, this releases the block */ 2162 block = next_block; 2163 } while (NULL != block); 2164 2165 bp->bi_end_io(bp); 2166 } 2167 2168 static void btrfsic_bh_end_io(struct buffer_head *bh, int uptodate) 2169 { 2170 struct btrfsic_block *block = (struct btrfsic_block *)bh->b_private; 2171 int iodone_w_error = !uptodate; 2172 struct btrfsic_dev_state *dev_state; 2173 2174 BUG_ON(NULL == block); 2175 dev_state = block->dev_state; 2176 if ((dev_state->state->print_mask & BTRFSIC_PRINT_MASK_END_IO_BIO_BH)) 2177 pr_info("bh_end_io(error=%d) for %c @%llu (%s/%llu/%d)\n", 2178 iodone_w_error, 2179 btrfsic_get_block_type(dev_state->state, block), 2180 block->logical_bytenr, block->dev_state->name, 2181 block->dev_bytenr, block->mirror_num); 2182 2183 block->iodone_w_error = iodone_w_error; 2184 if (block->submit_bio_bh_rw & REQ_PREFLUSH) { 2185 dev_state->last_flush_gen++; 2186 if ((dev_state->state->print_mask & 2187 BTRFSIC_PRINT_MASK_END_IO_BIO_BH)) 2188 pr_info("bh_end_io() new %s flush_gen=%llu\n", 2189 dev_state->name, dev_state->last_flush_gen); 2190 } 2191 if (block->submit_bio_bh_rw & REQ_FUA) 2192 block->flush_gen = 0; /* FUA completed means block is on disk */ 2193 2194 bh->b_private = block->orig_bio_bh_private; 2195 bh->b_end_io = block->orig_bio_bh_end_io.bh; 2196 block->is_iodone = 1; /* for FLUSH, this releases the block */ 2197 bh->b_end_io(bh, uptodate); 2198 } 2199 2200 static int btrfsic_process_written_superblock( 2201 struct btrfsic_state *state, 2202 struct btrfsic_block *const superblock, 2203 struct btrfs_super_block *const super_hdr) 2204 { 2205 int pass; 2206 2207 superblock->generation = btrfs_super_generation(super_hdr); 2208 if (!(superblock->generation > state->max_superblock_generation || 2209 0 == state->max_superblock_generation)) { 2210 if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE) 2211 pr_info("btrfsic: superblock @%llu (%s/%llu/%d) with old gen %llu <= %llu\n", 2212 superblock->logical_bytenr, 2213 superblock->dev_state->name, 2214 superblock->dev_bytenr, superblock->mirror_num, 2215 btrfs_super_generation(super_hdr), 2216 state->max_superblock_generation); 2217 } else { 2218 if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE) 2219 pr_info("btrfsic: got new superblock @%llu (%s/%llu/%d) with new gen %llu > %llu\n", 2220 superblock->logical_bytenr, 2221 superblock->dev_state->name, 2222 superblock->dev_bytenr, superblock->mirror_num, 2223 btrfs_super_generation(super_hdr), 2224 state->max_superblock_generation); 2225 2226 state->max_superblock_generation = 2227 btrfs_super_generation(super_hdr); 2228 state->latest_superblock = superblock; 2229 } 2230 2231 for (pass = 0; pass < 3; pass++) { 2232 int ret; 2233 u64 next_bytenr; 2234 struct btrfsic_block *next_block; 2235 struct btrfsic_block_data_ctx tmp_next_block_ctx; 2236 struct btrfsic_block_link *l; 2237 int num_copies; 2238 int mirror_num; 2239 const char *additional_string = NULL; 2240 struct btrfs_disk_key tmp_disk_key = {0}; 2241 2242 btrfs_set_disk_key_objectid(&tmp_disk_key, 2243 BTRFS_ROOT_ITEM_KEY); 2244 btrfs_set_disk_key_objectid(&tmp_disk_key, 0); 2245 2246 switch (pass) { 2247 case 0: 2248 btrfs_set_disk_key_objectid(&tmp_disk_key, 2249 BTRFS_ROOT_TREE_OBJECTID); 2250 additional_string = "root "; 2251 next_bytenr = btrfs_super_root(super_hdr); 2252 if (state->print_mask & 2253 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION) 2254 pr_info("root@%llu\n", next_bytenr); 2255 break; 2256 case 1: 2257 btrfs_set_disk_key_objectid(&tmp_disk_key, 2258 BTRFS_CHUNK_TREE_OBJECTID); 2259 additional_string = "chunk "; 2260 next_bytenr = btrfs_super_chunk_root(super_hdr); 2261 if (state->print_mask & 2262 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION) 2263 pr_info("chunk@%llu\n", next_bytenr); 2264 break; 2265 case 2: 2266 btrfs_set_disk_key_objectid(&tmp_disk_key, 2267 BTRFS_TREE_LOG_OBJECTID); 2268 additional_string = "log "; 2269 next_bytenr = btrfs_super_log_root(super_hdr); 2270 if (0 == next_bytenr) 2271 continue; 2272 if (state->print_mask & 2273 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION) 2274 pr_info("log@%llu\n", next_bytenr); 2275 break; 2276 } 2277 2278 num_copies = 2279 btrfs_num_copies(state->root->fs_info, 2280 next_bytenr, BTRFS_SUPER_INFO_SIZE); 2281 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES) 2282 pr_info("num_copies(log_bytenr=%llu) = %d\n", 2283 next_bytenr, num_copies); 2284 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) { 2285 int was_created; 2286 2287 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2288 pr_info("btrfsic_process_written_superblock(mirror_num=%d)\n", mirror_num); 2289 ret = btrfsic_map_block(state, next_bytenr, 2290 BTRFS_SUPER_INFO_SIZE, 2291 &tmp_next_block_ctx, 2292 mirror_num); 2293 if (ret) { 2294 pr_info("btrfsic: btrfsic_map_block(@%llu, mirror=%d) failed!\n", 2295 next_bytenr, mirror_num); 2296 return -1; 2297 } 2298 2299 next_block = btrfsic_block_lookup_or_add( 2300 state, 2301 &tmp_next_block_ctx, 2302 additional_string, 2303 1, 0, 1, 2304 mirror_num, 2305 &was_created); 2306 if (NULL == next_block) { 2307 pr_info("btrfsic: error, kmalloc failed!\n"); 2308 btrfsic_release_block_ctx(&tmp_next_block_ctx); 2309 return -1; 2310 } 2311 2312 next_block->disk_key = tmp_disk_key; 2313 if (was_created) 2314 next_block->generation = 2315 BTRFSIC_GENERATION_UNKNOWN; 2316 l = btrfsic_block_link_lookup_or_add( 2317 state, 2318 &tmp_next_block_ctx, 2319 next_block, 2320 superblock, 2321 BTRFSIC_GENERATION_UNKNOWN); 2322 btrfsic_release_block_ctx(&tmp_next_block_ctx); 2323 if (NULL == l) 2324 return -1; 2325 } 2326 } 2327 2328 if (WARN_ON(-1 == btrfsic_check_all_ref_blocks(state, superblock, 0))) 2329 btrfsic_dump_tree(state); 2330 2331 return 0; 2332 } 2333 2334 static int btrfsic_check_all_ref_blocks(struct btrfsic_state *state, 2335 struct btrfsic_block *const block, 2336 int recursion_level) 2337 { 2338 const struct btrfsic_block_link *l; 2339 int ret = 0; 2340 2341 if (recursion_level >= 3 + BTRFS_MAX_LEVEL) { 2342 /* 2343 * Note that this situation can happen and does not 2344 * indicate an error in regular cases. It happens 2345 * when disk blocks are freed and later reused. 2346 * The check-integrity module is not aware of any 2347 * block free operations, it just recognizes block 2348 * write operations. Therefore it keeps the linkage 2349 * information for a block until a block is 2350 * rewritten. This can temporarily cause incorrect 2351 * and even circular linkage informations. This 2352 * causes no harm unless such blocks are referenced 2353 * by the most recent super block. 2354 */ 2355 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2356 pr_info("btrfsic: abort cyclic linkage (case 1).\n"); 2357 2358 return ret; 2359 } 2360 2361 /* 2362 * This algorithm is recursive because the amount of used stack 2363 * space is very small and the max recursion depth is limited. 2364 */ 2365 list_for_each_entry(l, &block->ref_to_list, node_ref_to) { 2366 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2367 pr_info("rl=%d, %c @%llu (%s/%llu/%d) %u* refers to %c @%llu (%s/%llu/%d)\n", 2368 recursion_level, 2369 btrfsic_get_block_type(state, block), 2370 block->logical_bytenr, block->dev_state->name, 2371 block->dev_bytenr, block->mirror_num, 2372 l->ref_cnt, 2373 btrfsic_get_block_type(state, l->block_ref_to), 2374 l->block_ref_to->logical_bytenr, 2375 l->block_ref_to->dev_state->name, 2376 l->block_ref_to->dev_bytenr, 2377 l->block_ref_to->mirror_num); 2378 if (l->block_ref_to->never_written) { 2379 pr_info("btrfs: attempt to write superblock which references block %c @%llu (%s/%llu/%d) which is never written!\n", 2380 btrfsic_get_block_type(state, l->block_ref_to), 2381 l->block_ref_to->logical_bytenr, 2382 l->block_ref_to->dev_state->name, 2383 l->block_ref_to->dev_bytenr, 2384 l->block_ref_to->mirror_num); 2385 ret = -1; 2386 } else if (!l->block_ref_to->is_iodone) { 2387 pr_info("btrfs: attempt to write superblock which references block %c @%llu (%s/%llu/%d) which is not yet iodone!\n", 2388 btrfsic_get_block_type(state, l->block_ref_to), 2389 l->block_ref_to->logical_bytenr, 2390 l->block_ref_to->dev_state->name, 2391 l->block_ref_to->dev_bytenr, 2392 l->block_ref_to->mirror_num); 2393 ret = -1; 2394 } else if (l->block_ref_to->iodone_w_error) { 2395 pr_info("btrfs: attempt to write superblock which references block %c @%llu (%s/%llu/%d) which has write error!\n", 2396 btrfsic_get_block_type(state, l->block_ref_to), 2397 l->block_ref_to->logical_bytenr, 2398 l->block_ref_to->dev_state->name, 2399 l->block_ref_to->dev_bytenr, 2400 l->block_ref_to->mirror_num); 2401 ret = -1; 2402 } else if (l->parent_generation != 2403 l->block_ref_to->generation && 2404 BTRFSIC_GENERATION_UNKNOWN != 2405 l->parent_generation && 2406 BTRFSIC_GENERATION_UNKNOWN != 2407 l->block_ref_to->generation) { 2408 pr_info("btrfs: attempt to write superblock which references block %c @%llu (%s/%llu/%d) with generation %llu != parent generation %llu!\n", 2409 btrfsic_get_block_type(state, l->block_ref_to), 2410 l->block_ref_to->logical_bytenr, 2411 l->block_ref_to->dev_state->name, 2412 l->block_ref_to->dev_bytenr, 2413 l->block_ref_to->mirror_num, 2414 l->block_ref_to->generation, 2415 l->parent_generation); 2416 ret = -1; 2417 } else if (l->block_ref_to->flush_gen > 2418 l->block_ref_to->dev_state->last_flush_gen) { 2419 pr_info("btrfs: attempt to write superblock which references block %c @%llu (%s/%llu/%d) which is not flushed out of disk's write cache (block flush_gen=%llu, dev->flush_gen=%llu)!\n", 2420 btrfsic_get_block_type(state, l->block_ref_to), 2421 l->block_ref_to->logical_bytenr, 2422 l->block_ref_to->dev_state->name, 2423 l->block_ref_to->dev_bytenr, 2424 l->block_ref_to->mirror_num, block->flush_gen, 2425 l->block_ref_to->dev_state->last_flush_gen); 2426 ret = -1; 2427 } else if (-1 == btrfsic_check_all_ref_blocks(state, 2428 l->block_ref_to, 2429 recursion_level + 2430 1)) { 2431 ret = -1; 2432 } 2433 } 2434 2435 return ret; 2436 } 2437 2438 static int btrfsic_is_block_ref_by_superblock( 2439 const struct btrfsic_state *state, 2440 const struct btrfsic_block *block, 2441 int recursion_level) 2442 { 2443 const struct btrfsic_block_link *l; 2444 2445 if (recursion_level >= 3 + BTRFS_MAX_LEVEL) { 2446 /* refer to comment at "abort cyclic linkage (case 1)" */ 2447 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2448 pr_info("btrfsic: abort cyclic linkage (case 2).\n"); 2449 2450 return 0; 2451 } 2452 2453 /* 2454 * This algorithm is recursive because the amount of used stack space 2455 * is very small and the max recursion depth is limited. 2456 */ 2457 list_for_each_entry(l, &block->ref_from_list, node_ref_from) { 2458 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2459 pr_info("rl=%d, %c @%llu (%s/%llu/%d) is ref %u* from %c @%llu (%s/%llu/%d)\n", 2460 recursion_level, 2461 btrfsic_get_block_type(state, block), 2462 block->logical_bytenr, block->dev_state->name, 2463 block->dev_bytenr, block->mirror_num, 2464 l->ref_cnt, 2465 btrfsic_get_block_type(state, l->block_ref_from), 2466 l->block_ref_from->logical_bytenr, 2467 l->block_ref_from->dev_state->name, 2468 l->block_ref_from->dev_bytenr, 2469 l->block_ref_from->mirror_num); 2470 if (l->block_ref_from->is_superblock && 2471 state->latest_superblock->dev_bytenr == 2472 l->block_ref_from->dev_bytenr && 2473 state->latest_superblock->dev_state->bdev == 2474 l->block_ref_from->dev_state->bdev) 2475 return 1; 2476 else if (btrfsic_is_block_ref_by_superblock(state, 2477 l->block_ref_from, 2478 recursion_level + 2479 1)) 2480 return 1; 2481 } 2482 2483 return 0; 2484 } 2485 2486 static void btrfsic_print_add_link(const struct btrfsic_state *state, 2487 const struct btrfsic_block_link *l) 2488 { 2489 pr_info("Add %u* link from %c @%llu (%s/%llu/%d) to %c @%llu (%s/%llu/%d).\n", 2490 l->ref_cnt, 2491 btrfsic_get_block_type(state, l->block_ref_from), 2492 l->block_ref_from->logical_bytenr, 2493 l->block_ref_from->dev_state->name, 2494 l->block_ref_from->dev_bytenr, l->block_ref_from->mirror_num, 2495 btrfsic_get_block_type(state, l->block_ref_to), 2496 l->block_ref_to->logical_bytenr, 2497 l->block_ref_to->dev_state->name, l->block_ref_to->dev_bytenr, 2498 l->block_ref_to->mirror_num); 2499 } 2500 2501 static void btrfsic_print_rem_link(const struct btrfsic_state *state, 2502 const struct btrfsic_block_link *l) 2503 { 2504 pr_info("Rem %u* link from %c @%llu (%s/%llu/%d) to %c @%llu (%s/%llu/%d).\n", 2505 l->ref_cnt, 2506 btrfsic_get_block_type(state, l->block_ref_from), 2507 l->block_ref_from->logical_bytenr, 2508 l->block_ref_from->dev_state->name, 2509 l->block_ref_from->dev_bytenr, l->block_ref_from->mirror_num, 2510 btrfsic_get_block_type(state, l->block_ref_to), 2511 l->block_ref_to->logical_bytenr, 2512 l->block_ref_to->dev_state->name, l->block_ref_to->dev_bytenr, 2513 l->block_ref_to->mirror_num); 2514 } 2515 2516 static char btrfsic_get_block_type(const struct btrfsic_state *state, 2517 const struct btrfsic_block *block) 2518 { 2519 if (block->is_superblock && 2520 state->latest_superblock->dev_bytenr == block->dev_bytenr && 2521 state->latest_superblock->dev_state->bdev == block->dev_state->bdev) 2522 return 'S'; 2523 else if (block->is_superblock) 2524 return 's'; 2525 else if (block->is_metadata) 2526 return 'M'; 2527 else 2528 return 'D'; 2529 } 2530 2531 static void btrfsic_dump_tree(const struct btrfsic_state *state) 2532 { 2533 btrfsic_dump_tree_sub(state, state->latest_superblock, 0); 2534 } 2535 2536 static void btrfsic_dump_tree_sub(const struct btrfsic_state *state, 2537 const struct btrfsic_block *block, 2538 int indent_level) 2539 { 2540 const struct btrfsic_block_link *l; 2541 int indent_add; 2542 static char buf[80]; 2543 int cursor_position; 2544 2545 /* 2546 * Should better fill an on-stack buffer with a complete line and 2547 * dump it at once when it is time to print a newline character. 2548 */ 2549 2550 /* 2551 * This algorithm is recursive because the amount of used stack space 2552 * is very small and the max recursion depth is limited. 2553 */ 2554 indent_add = sprintf(buf, "%c-%llu(%s/%llu/%u)", 2555 btrfsic_get_block_type(state, block), 2556 block->logical_bytenr, block->dev_state->name, 2557 block->dev_bytenr, block->mirror_num); 2558 if (indent_level + indent_add > BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL) { 2559 printk("[...]\n"); 2560 return; 2561 } 2562 printk(buf); 2563 indent_level += indent_add; 2564 if (list_empty(&block->ref_to_list)) { 2565 printk("\n"); 2566 return; 2567 } 2568 if (block->mirror_num > 1 && 2569 !(state->print_mask & BTRFSIC_PRINT_MASK_TREE_WITH_ALL_MIRRORS)) { 2570 printk(" [...]\n"); 2571 return; 2572 } 2573 2574 cursor_position = indent_level; 2575 list_for_each_entry(l, &block->ref_to_list, node_ref_to) { 2576 while (cursor_position < indent_level) { 2577 printk(" "); 2578 cursor_position++; 2579 } 2580 if (l->ref_cnt > 1) 2581 indent_add = sprintf(buf, " %d*--> ", l->ref_cnt); 2582 else 2583 indent_add = sprintf(buf, " --> "); 2584 if (indent_level + indent_add > 2585 BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL) { 2586 printk("[...]\n"); 2587 cursor_position = 0; 2588 continue; 2589 } 2590 2591 printk(buf); 2592 2593 btrfsic_dump_tree_sub(state, l->block_ref_to, 2594 indent_level + indent_add); 2595 cursor_position = 0; 2596 } 2597 } 2598 2599 static struct btrfsic_block_link *btrfsic_block_link_lookup_or_add( 2600 struct btrfsic_state *state, 2601 struct btrfsic_block_data_ctx *next_block_ctx, 2602 struct btrfsic_block *next_block, 2603 struct btrfsic_block *from_block, 2604 u64 parent_generation) 2605 { 2606 struct btrfsic_block_link *l; 2607 2608 l = btrfsic_block_link_hashtable_lookup(next_block_ctx->dev->bdev, 2609 next_block_ctx->dev_bytenr, 2610 from_block->dev_state->bdev, 2611 from_block->dev_bytenr, 2612 &state->block_link_hashtable); 2613 if (NULL == l) { 2614 l = btrfsic_block_link_alloc(); 2615 if (NULL == l) { 2616 pr_info("btrfsic: error, kmalloc failed!\n"); 2617 return NULL; 2618 } 2619 2620 l->block_ref_to = next_block; 2621 l->block_ref_from = from_block; 2622 l->ref_cnt = 1; 2623 l->parent_generation = parent_generation; 2624 2625 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2626 btrfsic_print_add_link(state, l); 2627 2628 list_add(&l->node_ref_to, &from_block->ref_to_list); 2629 list_add(&l->node_ref_from, &next_block->ref_from_list); 2630 2631 btrfsic_block_link_hashtable_add(l, 2632 &state->block_link_hashtable); 2633 } else { 2634 l->ref_cnt++; 2635 l->parent_generation = parent_generation; 2636 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2637 btrfsic_print_add_link(state, l); 2638 } 2639 2640 return l; 2641 } 2642 2643 static struct btrfsic_block *btrfsic_block_lookup_or_add( 2644 struct btrfsic_state *state, 2645 struct btrfsic_block_data_ctx *block_ctx, 2646 const char *additional_string, 2647 int is_metadata, 2648 int is_iodone, 2649 int never_written, 2650 int mirror_num, 2651 int *was_created) 2652 { 2653 struct btrfsic_block *block; 2654 2655 block = btrfsic_block_hashtable_lookup(block_ctx->dev->bdev, 2656 block_ctx->dev_bytenr, 2657 &state->block_hashtable); 2658 if (NULL == block) { 2659 struct btrfsic_dev_state *dev_state; 2660 2661 block = btrfsic_block_alloc(); 2662 if (NULL == block) { 2663 pr_info("btrfsic: error, kmalloc failed!\n"); 2664 return NULL; 2665 } 2666 dev_state = btrfsic_dev_state_lookup(block_ctx->dev->bdev); 2667 if (NULL == dev_state) { 2668 pr_info("btrfsic: error, lookup dev_state failed!\n"); 2669 btrfsic_block_free(block); 2670 return NULL; 2671 } 2672 block->dev_state = dev_state; 2673 block->dev_bytenr = block_ctx->dev_bytenr; 2674 block->logical_bytenr = block_ctx->start; 2675 block->is_metadata = is_metadata; 2676 block->is_iodone = is_iodone; 2677 block->never_written = never_written; 2678 block->mirror_num = mirror_num; 2679 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 2680 pr_info("New %s%c-block @%llu (%s/%llu/%d)\n", 2681 additional_string, 2682 btrfsic_get_block_type(state, block), 2683 block->logical_bytenr, dev_state->name, 2684 block->dev_bytenr, mirror_num); 2685 list_add(&block->all_blocks_node, &state->all_blocks_list); 2686 btrfsic_block_hashtable_add(block, &state->block_hashtable); 2687 if (NULL != was_created) 2688 *was_created = 1; 2689 } else { 2690 if (NULL != was_created) 2691 *was_created = 0; 2692 } 2693 2694 return block; 2695 } 2696 2697 static void btrfsic_cmp_log_and_dev_bytenr(struct btrfsic_state *state, 2698 u64 bytenr, 2699 struct btrfsic_dev_state *dev_state, 2700 u64 dev_bytenr) 2701 { 2702 int num_copies; 2703 int mirror_num; 2704 int ret; 2705 struct btrfsic_block_data_ctx block_ctx; 2706 int match = 0; 2707 2708 num_copies = btrfs_num_copies(state->root->fs_info, 2709 bytenr, state->metablock_size); 2710 2711 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) { 2712 ret = btrfsic_map_block(state, bytenr, state->metablock_size, 2713 &block_ctx, mirror_num); 2714 if (ret) { 2715 pr_info("btrfsic: btrfsic_map_block(logical @%llu, mirror %d) failed!\n", 2716 bytenr, mirror_num); 2717 continue; 2718 } 2719 2720 if (dev_state->bdev == block_ctx.dev->bdev && 2721 dev_bytenr == block_ctx.dev_bytenr) { 2722 match++; 2723 btrfsic_release_block_ctx(&block_ctx); 2724 break; 2725 } 2726 btrfsic_release_block_ctx(&block_ctx); 2727 } 2728 2729 if (WARN_ON(!match)) { 2730 pr_info("btrfs: attempt to write M-block which contains logical bytenr that doesn't map to dev+physical bytenr of submit_bio, buffer->log_bytenr=%llu, submit_bio(bdev=%s, phys_bytenr=%llu)!\n", 2731 bytenr, dev_state->name, dev_bytenr); 2732 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) { 2733 ret = btrfsic_map_block(state, bytenr, 2734 state->metablock_size, 2735 &block_ctx, mirror_num); 2736 if (ret) 2737 continue; 2738 2739 pr_info("Read logical bytenr @%llu maps to (%s/%llu/%d)\n", 2740 bytenr, block_ctx.dev->name, 2741 block_ctx.dev_bytenr, mirror_num); 2742 } 2743 } 2744 } 2745 2746 static struct btrfsic_dev_state *btrfsic_dev_state_lookup( 2747 struct block_device *bdev) 2748 { 2749 return btrfsic_dev_state_hashtable_lookup(bdev, 2750 &btrfsic_dev_state_hashtable); 2751 } 2752 2753 int btrfsic_submit_bh(int op, int op_flags, struct buffer_head *bh) 2754 { 2755 struct btrfsic_dev_state *dev_state; 2756 2757 if (!btrfsic_is_initialized) 2758 return submit_bh(op, op_flags, bh); 2759 2760 mutex_lock(&btrfsic_mutex); 2761 /* since btrfsic_submit_bh() might also be called before 2762 * btrfsic_mount(), this might return NULL */ 2763 dev_state = btrfsic_dev_state_lookup(bh->b_bdev); 2764 2765 /* Only called to write the superblock (incl. FLUSH/FUA) */ 2766 if (NULL != dev_state && 2767 (op == REQ_OP_WRITE) && bh->b_size > 0) { 2768 u64 dev_bytenr; 2769 2770 dev_bytenr = 4096 * bh->b_blocknr; 2771 if (dev_state->state->print_mask & 2772 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH) 2773 pr_info("submit_bh(op=0x%x,0x%x, blocknr=%llu (bytenr %llu), size=%zu, data=%p, bdev=%p)\n", 2774 op, op_flags, (unsigned long long)bh->b_blocknr, 2775 dev_bytenr, bh->b_size, bh->b_data, bh->b_bdev); 2776 btrfsic_process_written_block(dev_state, dev_bytenr, 2777 &bh->b_data, 1, NULL, 2778 NULL, bh, op_flags); 2779 } else if (NULL != dev_state && (op_flags & REQ_PREFLUSH)) { 2780 if (dev_state->state->print_mask & 2781 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH) 2782 pr_info("submit_bh(op=0x%x,0x%x FLUSH, bdev=%p)\n", 2783 op, op_flags, bh->b_bdev); 2784 if (!dev_state->dummy_block_for_bio_bh_flush.is_iodone) { 2785 if ((dev_state->state->print_mask & 2786 (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH | 2787 BTRFSIC_PRINT_MASK_VERBOSE))) 2788 pr_info("btrfsic_submit_bh(%s) with FLUSH but dummy block already in use (ignored)!\n", 2789 dev_state->name); 2790 } else { 2791 struct btrfsic_block *const block = 2792 &dev_state->dummy_block_for_bio_bh_flush; 2793 2794 block->is_iodone = 0; 2795 block->never_written = 0; 2796 block->iodone_w_error = 0; 2797 block->flush_gen = dev_state->last_flush_gen + 1; 2798 block->submit_bio_bh_rw = op_flags; 2799 block->orig_bio_bh_private = bh->b_private; 2800 block->orig_bio_bh_end_io.bh = bh->b_end_io; 2801 block->next_in_same_bio = NULL; 2802 bh->b_private = block; 2803 bh->b_end_io = btrfsic_bh_end_io; 2804 } 2805 } 2806 mutex_unlock(&btrfsic_mutex); 2807 return submit_bh(op, op_flags, bh); 2808 } 2809 2810 static void __btrfsic_submit_bio(struct bio *bio) 2811 { 2812 struct btrfsic_dev_state *dev_state; 2813 2814 if (!btrfsic_is_initialized) 2815 return; 2816 2817 mutex_lock(&btrfsic_mutex); 2818 /* since btrfsic_submit_bio() is also called before 2819 * btrfsic_mount(), this might return NULL */ 2820 dev_state = btrfsic_dev_state_lookup(bio->bi_bdev); 2821 if (NULL != dev_state && 2822 (bio_op(bio) == REQ_OP_WRITE) && NULL != bio->bi_io_vec) { 2823 unsigned int i; 2824 u64 dev_bytenr; 2825 u64 cur_bytenr; 2826 int bio_is_patched; 2827 char **mapped_datav; 2828 2829 dev_bytenr = 512 * bio->bi_iter.bi_sector; 2830 bio_is_patched = 0; 2831 if (dev_state->state->print_mask & 2832 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH) 2833 pr_info("submit_bio(rw=%d,0x%x, bi_vcnt=%u, bi_sector=%llu (bytenr %llu), bi_bdev=%p)\n", 2834 bio_op(bio), bio->bi_opf, bio->bi_vcnt, 2835 (unsigned long long)bio->bi_iter.bi_sector, 2836 dev_bytenr, bio->bi_bdev); 2837 2838 mapped_datav = kmalloc_array(bio->bi_vcnt, 2839 sizeof(*mapped_datav), GFP_NOFS); 2840 if (!mapped_datav) 2841 goto leave; 2842 cur_bytenr = dev_bytenr; 2843 for (i = 0; i < bio->bi_vcnt; i++) { 2844 BUG_ON(bio->bi_io_vec[i].bv_len != PAGE_SIZE); 2845 mapped_datav[i] = kmap(bio->bi_io_vec[i].bv_page); 2846 if (!mapped_datav[i]) { 2847 while (i > 0) { 2848 i--; 2849 kunmap(bio->bi_io_vec[i].bv_page); 2850 } 2851 kfree(mapped_datav); 2852 goto leave; 2853 } 2854 if (dev_state->state->print_mask & 2855 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH_VERBOSE) 2856 pr_info("#%u: bytenr=%llu, len=%u, offset=%u\n", 2857 i, cur_bytenr, bio->bi_io_vec[i].bv_len, 2858 bio->bi_io_vec[i].bv_offset); 2859 cur_bytenr += bio->bi_io_vec[i].bv_len; 2860 } 2861 btrfsic_process_written_block(dev_state, dev_bytenr, 2862 mapped_datav, bio->bi_vcnt, 2863 bio, &bio_is_patched, 2864 NULL, bio->bi_opf); 2865 while (i > 0) { 2866 i--; 2867 kunmap(bio->bi_io_vec[i].bv_page); 2868 } 2869 kfree(mapped_datav); 2870 } else if (NULL != dev_state && (bio->bi_opf & REQ_PREFLUSH)) { 2871 if (dev_state->state->print_mask & 2872 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH) 2873 pr_info("submit_bio(rw=%d,0x%x FLUSH, bdev=%p)\n", 2874 bio_op(bio), bio->bi_opf, bio->bi_bdev); 2875 if (!dev_state->dummy_block_for_bio_bh_flush.is_iodone) { 2876 if ((dev_state->state->print_mask & 2877 (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH | 2878 BTRFSIC_PRINT_MASK_VERBOSE))) 2879 pr_info("btrfsic_submit_bio(%s) with FLUSH but dummy block already in use (ignored)!\n", 2880 dev_state->name); 2881 } else { 2882 struct btrfsic_block *const block = 2883 &dev_state->dummy_block_for_bio_bh_flush; 2884 2885 block->is_iodone = 0; 2886 block->never_written = 0; 2887 block->iodone_w_error = 0; 2888 block->flush_gen = dev_state->last_flush_gen + 1; 2889 block->submit_bio_bh_rw = bio->bi_opf; 2890 block->orig_bio_bh_private = bio->bi_private; 2891 block->orig_bio_bh_end_io.bio = bio->bi_end_io; 2892 block->next_in_same_bio = NULL; 2893 bio->bi_private = block; 2894 bio->bi_end_io = btrfsic_bio_end_io; 2895 } 2896 } 2897 leave: 2898 mutex_unlock(&btrfsic_mutex); 2899 } 2900 2901 void btrfsic_submit_bio(struct bio *bio) 2902 { 2903 __btrfsic_submit_bio(bio); 2904 submit_bio(bio); 2905 } 2906 2907 int btrfsic_submit_bio_wait(struct bio *bio) 2908 { 2909 __btrfsic_submit_bio(bio); 2910 return submit_bio_wait(bio); 2911 } 2912 2913 int btrfsic_mount(struct btrfs_root *root, 2914 struct btrfs_fs_devices *fs_devices, 2915 int including_extent_data, u32 print_mask) 2916 { 2917 int ret; 2918 struct btrfsic_state *state; 2919 struct list_head *dev_head = &fs_devices->devices; 2920 struct btrfs_device *device; 2921 2922 if (root->nodesize & ((u64)PAGE_SIZE - 1)) { 2923 pr_info("btrfsic: cannot handle nodesize %d not being a multiple of PAGE_SIZE %ld!\n", 2924 root->nodesize, PAGE_SIZE); 2925 return -1; 2926 } 2927 if (root->sectorsize & ((u64)PAGE_SIZE - 1)) { 2928 pr_info("btrfsic: cannot handle sectorsize %d not being a multiple of PAGE_SIZE %ld!\n", 2929 root->sectorsize, PAGE_SIZE); 2930 return -1; 2931 } 2932 state = kzalloc(sizeof(*state), GFP_KERNEL | __GFP_NOWARN | __GFP_REPEAT); 2933 if (!state) { 2934 state = vzalloc(sizeof(*state)); 2935 if (!state) { 2936 pr_info("btrfs check-integrity: vzalloc() failed!\n"); 2937 return -1; 2938 } 2939 } 2940 2941 if (!btrfsic_is_initialized) { 2942 mutex_init(&btrfsic_mutex); 2943 btrfsic_dev_state_hashtable_init(&btrfsic_dev_state_hashtable); 2944 btrfsic_is_initialized = 1; 2945 } 2946 mutex_lock(&btrfsic_mutex); 2947 state->root = root; 2948 state->print_mask = print_mask; 2949 state->include_extent_data = including_extent_data; 2950 state->csum_size = 0; 2951 state->metablock_size = root->nodesize; 2952 state->datablock_size = root->sectorsize; 2953 INIT_LIST_HEAD(&state->all_blocks_list); 2954 btrfsic_block_hashtable_init(&state->block_hashtable); 2955 btrfsic_block_link_hashtable_init(&state->block_link_hashtable); 2956 state->max_superblock_generation = 0; 2957 state->latest_superblock = NULL; 2958 2959 list_for_each_entry(device, dev_head, dev_list) { 2960 struct btrfsic_dev_state *ds; 2961 const char *p; 2962 2963 if (!device->bdev || !device->name) 2964 continue; 2965 2966 ds = btrfsic_dev_state_alloc(); 2967 if (NULL == ds) { 2968 pr_info("btrfs check-integrity: kmalloc() failed!\n"); 2969 mutex_unlock(&btrfsic_mutex); 2970 return -1; 2971 } 2972 ds->bdev = device->bdev; 2973 ds->state = state; 2974 bdevname(ds->bdev, ds->name); 2975 ds->name[BDEVNAME_SIZE - 1] = '\0'; 2976 p = kbasename(ds->name); 2977 strlcpy(ds->name, p, sizeof(ds->name)); 2978 btrfsic_dev_state_hashtable_add(ds, 2979 &btrfsic_dev_state_hashtable); 2980 } 2981 2982 ret = btrfsic_process_superblock(state, fs_devices); 2983 if (0 != ret) { 2984 mutex_unlock(&btrfsic_mutex); 2985 btrfsic_unmount(root, fs_devices); 2986 return ret; 2987 } 2988 2989 if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_DATABASE) 2990 btrfsic_dump_database(state); 2991 if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_TREE) 2992 btrfsic_dump_tree(state); 2993 2994 mutex_unlock(&btrfsic_mutex); 2995 return 0; 2996 } 2997 2998 void btrfsic_unmount(struct btrfs_root *root, 2999 struct btrfs_fs_devices *fs_devices) 3000 { 3001 struct btrfsic_block *b_all, *tmp_all; 3002 struct btrfsic_state *state; 3003 struct list_head *dev_head = &fs_devices->devices; 3004 struct btrfs_device *device; 3005 3006 if (!btrfsic_is_initialized) 3007 return; 3008 3009 mutex_lock(&btrfsic_mutex); 3010 3011 state = NULL; 3012 list_for_each_entry(device, dev_head, dev_list) { 3013 struct btrfsic_dev_state *ds; 3014 3015 if (!device->bdev || !device->name) 3016 continue; 3017 3018 ds = btrfsic_dev_state_hashtable_lookup( 3019 device->bdev, 3020 &btrfsic_dev_state_hashtable); 3021 if (NULL != ds) { 3022 state = ds->state; 3023 btrfsic_dev_state_hashtable_remove(ds); 3024 btrfsic_dev_state_free(ds); 3025 } 3026 } 3027 3028 if (NULL == state) { 3029 pr_info("btrfsic: error, cannot find state information on umount!\n"); 3030 mutex_unlock(&btrfsic_mutex); 3031 return; 3032 } 3033 3034 /* 3035 * Don't care about keeping the lists' state up to date, 3036 * just free all memory that was allocated dynamically. 3037 * Free the blocks and the block_links. 3038 */ 3039 list_for_each_entry_safe(b_all, tmp_all, &state->all_blocks_list, 3040 all_blocks_node) { 3041 struct btrfsic_block_link *l, *tmp; 3042 3043 list_for_each_entry_safe(l, tmp, &b_all->ref_to_list, 3044 node_ref_to) { 3045 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE) 3046 btrfsic_print_rem_link(state, l); 3047 3048 l->ref_cnt--; 3049 if (0 == l->ref_cnt) 3050 btrfsic_block_link_free(l); 3051 } 3052 3053 if (b_all->is_iodone || b_all->never_written) 3054 btrfsic_block_free(b_all); 3055 else 3056 pr_info("btrfs: attempt to free %c-block @%llu (%s/%llu/%d) on umount which is not yet iodone!\n", 3057 btrfsic_get_block_type(state, b_all), 3058 b_all->logical_bytenr, b_all->dev_state->name, 3059 b_all->dev_bytenr, b_all->mirror_num); 3060 } 3061 3062 mutex_unlock(&btrfsic_mutex); 3063 3064 kvfree(state); 3065 } 3066