2 * Copyright (C) STRATO AG 2011. All rights reserved.
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.
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.
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.
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.
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
30 * 1. It is not allowed to write a disk block which is
31 * currently referenced by the super block (either directly
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.
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
52 * The search term in the kernel log that can be used to filter
53 * on the existence of detected integrity issues is
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.
60 * Example #1, apply integrity checks to all metadata:
61 * mount /dev/sdb1 /mnt -o check_int
63 * Example #2, apply integrity checks to all metadata and
65 * mount /dev/sdb1 /mnt -o check_int_data
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
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.
82 #include <linux/sched.h>
83 #include <linux/slab.h>
84 #include <linux/buffer_head.h>
85 #include <linux/mutex.h>
86 #include <linux/crc32c.h>
87 #include <linux/genhd.h>
88 #include <linux/blkdev.h>
91 #include "transaction.h"
92 #include "extent_io.h"
94 #include "print-tree.h"
96 #include "check-integrity.h"
97 #include "rcu-string.h"
99 #define BTRFSIC_BLOCK_HASHTABLE_SIZE 0x10000
100 #define BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE 0x10000
101 #define BTRFSIC_DEV2STATE_HASHTABLE_SIZE 0x100
102 #define BTRFSIC_BLOCK_MAGIC_NUMBER 0x14491051
103 #define BTRFSIC_BLOCK_LINK_MAGIC_NUMBER 0x11070807
104 #define BTRFSIC_DEV2STATE_MAGIC_NUMBER 0x20111530
105 #define BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER 20111300
106 #define BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL (200 - 6) /* in characters,
107 * excluding " [...]" */
108 #define BTRFSIC_GENERATION_UNKNOWN ((u64)-1)
111 * The definition of the bitmask fields for the print_mask.
112 * They are specified with the mount option check_integrity_print_mask.
114 #define BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE 0x00000001
115 #define BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION 0x00000002
116 #define BTRFSIC_PRINT_MASK_TREE_AFTER_SB_WRITE 0x00000004
117 #define BTRFSIC_PRINT_MASK_TREE_BEFORE_SB_WRITE 0x00000008
118 #define BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH 0x00000010
119 #define BTRFSIC_PRINT_MASK_END_IO_BIO_BH 0x00000020
120 #define BTRFSIC_PRINT_MASK_VERBOSE 0x00000040
121 #define BTRFSIC_PRINT_MASK_VERY_VERBOSE 0x00000080
122 #define BTRFSIC_PRINT_MASK_INITIAL_TREE 0x00000100
123 #define BTRFSIC_PRINT_MASK_INITIAL_ALL_TREES 0x00000200
124 #define BTRFSIC_PRINT_MASK_INITIAL_DATABASE 0x00000400
125 #define BTRFSIC_PRINT_MASK_NUM_COPIES 0x00000800
126 #define BTRFSIC_PRINT_MASK_TREE_WITH_ALL_MIRRORS 0x00001000
128 struct btrfsic_dev_state;
129 struct btrfsic_state;
131 struct btrfsic_block {
132 u32 magic_num; /* only used for debug purposes */
133 unsigned int is_metadata:1; /* if it is meta-data, not data-data */
134 unsigned int is_superblock:1; /* if it is one of the superblocks */
135 unsigned int is_iodone:1; /* if is done by lower subsystem */
136 unsigned int iodone_w_error:1; /* error was indicated to endio */
137 unsigned int never_written:1; /* block was added because it was
138 * referenced, not because it was
140 unsigned int mirror_num; /* large enough to hold
141 * BTRFS_SUPER_MIRROR_MAX */
142 struct btrfsic_dev_state *dev_state;
143 u64 dev_bytenr; /* key, physical byte num on disk */
144 u64 logical_bytenr; /* logical byte num on disk */
146 struct btrfs_disk_key disk_key; /* extra info to print in case of
147 * issues, will not always be correct */
148 struct list_head collision_resolving_node; /* list node */
149 struct list_head all_blocks_node; /* list node */
151 /* the following two lists contain block_link items */
152 struct list_head ref_to_list; /* list */
153 struct list_head ref_from_list; /* list */
154 struct btrfsic_block *next_in_same_bio;
155 void *orig_bio_bh_private;
159 } orig_bio_bh_end_io;
160 int submit_bio_bh_rw;
161 u64 flush_gen; /* only valid if !never_written */
165 * Elements of this type are allocated dynamically and required because
166 * each block object can refer to and can be ref from multiple blocks.
167 * The key to lookup them in the hashtable is the dev_bytenr of
168 * the block ref to plus the one from the block refered from.
169 * The fact that they are searchable via a hashtable and that a
170 * ref_cnt is maintained is not required for the btrfs integrity
171 * check algorithm itself, it is only used to make the output more
172 * beautiful in case that an error is detected (an error is defined
173 * as a write operation to a block while that block is still referenced).
175 struct btrfsic_block_link {
176 u32 magic_num; /* only used for debug purposes */
178 struct list_head node_ref_to; /* list node */
179 struct list_head node_ref_from; /* list node */
180 struct list_head collision_resolving_node; /* list node */
181 struct btrfsic_block *block_ref_to;
182 struct btrfsic_block *block_ref_from;
183 u64 parent_generation;
186 struct btrfsic_dev_state {
187 u32 magic_num; /* only used for debug purposes */
188 struct block_device *bdev;
189 struct btrfsic_state *state;
190 struct list_head collision_resolving_node; /* list node */
191 struct btrfsic_block dummy_block_for_bio_bh_flush;
193 char name[BDEVNAME_SIZE];
196 struct btrfsic_block_hashtable {
197 struct list_head table[BTRFSIC_BLOCK_HASHTABLE_SIZE];
200 struct btrfsic_block_link_hashtable {
201 struct list_head table[BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE];
204 struct btrfsic_dev_state_hashtable {
205 struct list_head table[BTRFSIC_DEV2STATE_HASHTABLE_SIZE];
208 struct btrfsic_block_data_ctx {
209 u64 start; /* virtual bytenr */
210 u64 dev_bytenr; /* physical bytenr on device */
212 struct btrfsic_dev_state *dev;
218 /* This structure is used to implement recursion without occupying
219 * any stack space, refer to btrfsic_process_metablock() */
220 struct btrfsic_stack_frame {
228 struct btrfsic_block *block;
229 struct btrfsic_block_data_ctx *block_ctx;
230 struct btrfsic_block *next_block;
231 struct btrfsic_block_data_ctx next_block_ctx;
232 struct btrfs_header *hdr;
233 struct btrfsic_stack_frame *prev;
236 /* Some state per mounted filesystem */
237 struct btrfsic_state {
239 int include_extent_data;
241 struct list_head all_blocks_list;
242 struct btrfsic_block_hashtable block_hashtable;
243 struct btrfsic_block_link_hashtable block_link_hashtable;
244 struct btrfs_root *root;
245 u64 max_superblock_generation;
246 struct btrfsic_block *latest_superblock;
251 static void btrfsic_block_init(struct btrfsic_block *b);
252 static struct btrfsic_block *btrfsic_block_alloc(void);
253 static void btrfsic_block_free(struct btrfsic_block *b);
254 static void btrfsic_block_link_init(struct btrfsic_block_link *n);
255 static struct btrfsic_block_link *btrfsic_block_link_alloc(void);
256 static void btrfsic_block_link_free(struct btrfsic_block_link *n);
257 static void btrfsic_dev_state_init(struct btrfsic_dev_state *ds);
258 static struct btrfsic_dev_state *btrfsic_dev_state_alloc(void);
259 static void btrfsic_dev_state_free(struct btrfsic_dev_state *ds);
260 static void btrfsic_block_hashtable_init(struct btrfsic_block_hashtable *h);
261 static void btrfsic_block_hashtable_add(struct btrfsic_block *b,
262 struct btrfsic_block_hashtable *h);
263 static void btrfsic_block_hashtable_remove(struct btrfsic_block *b);
264 static struct btrfsic_block *btrfsic_block_hashtable_lookup(
265 struct block_device *bdev,
267 struct btrfsic_block_hashtable *h);
268 static void btrfsic_block_link_hashtable_init(
269 struct btrfsic_block_link_hashtable *h);
270 static void btrfsic_block_link_hashtable_add(
271 struct btrfsic_block_link *l,
272 struct btrfsic_block_link_hashtable *h);
273 static void btrfsic_block_link_hashtable_remove(struct btrfsic_block_link *l);
274 static struct btrfsic_block_link *btrfsic_block_link_hashtable_lookup(
275 struct block_device *bdev_ref_to,
276 u64 dev_bytenr_ref_to,
277 struct block_device *bdev_ref_from,
278 u64 dev_bytenr_ref_from,
279 struct btrfsic_block_link_hashtable *h);
280 static void btrfsic_dev_state_hashtable_init(
281 struct btrfsic_dev_state_hashtable *h);
282 static void btrfsic_dev_state_hashtable_add(
283 struct btrfsic_dev_state *ds,
284 struct btrfsic_dev_state_hashtable *h);
285 static void btrfsic_dev_state_hashtable_remove(struct btrfsic_dev_state *ds);
286 static struct btrfsic_dev_state *btrfsic_dev_state_hashtable_lookup(
287 struct block_device *bdev,
288 struct btrfsic_dev_state_hashtable *h);
289 static struct btrfsic_stack_frame *btrfsic_stack_frame_alloc(void);
290 static void btrfsic_stack_frame_free(struct btrfsic_stack_frame *sf);
291 static int btrfsic_process_superblock(struct btrfsic_state *state,
292 struct btrfs_fs_devices *fs_devices);
293 static int btrfsic_process_metablock(struct btrfsic_state *state,
294 struct btrfsic_block *block,
295 struct btrfsic_block_data_ctx *block_ctx,
296 int limit_nesting, int force_iodone_flag);
297 static void btrfsic_read_from_block_data(
298 struct btrfsic_block_data_ctx *block_ctx,
299 void *dst, u32 offset, size_t len);
300 static int btrfsic_create_link_to_next_block(
301 struct btrfsic_state *state,
302 struct btrfsic_block *block,
303 struct btrfsic_block_data_ctx
304 *block_ctx, u64 next_bytenr,
306 struct btrfsic_block_data_ctx *next_block_ctx,
307 struct btrfsic_block **next_blockp,
308 int force_iodone_flag,
309 int *num_copiesp, int *mirror_nump,
310 struct btrfs_disk_key *disk_key,
311 u64 parent_generation);
312 static int btrfsic_handle_extent_data(struct btrfsic_state *state,
313 struct btrfsic_block *block,
314 struct btrfsic_block_data_ctx *block_ctx,
315 u32 item_offset, int force_iodone_flag);
316 static int btrfsic_map_block(struct btrfsic_state *state, u64 bytenr, u32 len,
317 struct btrfsic_block_data_ctx *block_ctx_out,
319 static int btrfsic_map_superblock(struct btrfsic_state *state, u64 bytenr,
320 u32 len, struct block_device *bdev,
321 struct btrfsic_block_data_ctx *block_ctx_out);
322 static void btrfsic_release_block_ctx(struct btrfsic_block_data_ctx *block_ctx);
323 static int btrfsic_read_block(struct btrfsic_state *state,
324 struct btrfsic_block_data_ctx *block_ctx);
325 static void btrfsic_dump_database(struct btrfsic_state *state);
326 static void btrfsic_complete_bio_end_io(struct bio *bio, int err);
327 static int btrfsic_test_for_metadata(struct btrfsic_state *state,
328 char **datav, unsigned int num_pages);
329 static void btrfsic_process_written_block(struct btrfsic_dev_state *dev_state,
330 u64 dev_bytenr, char **mapped_datav,
331 unsigned int num_pages,
332 struct bio *bio, int *bio_is_patched,
333 struct buffer_head *bh,
334 int submit_bio_bh_rw);
335 static int btrfsic_process_written_superblock(
336 struct btrfsic_state *state,
337 struct btrfsic_block *const block,
338 struct btrfs_super_block *const super_hdr);
339 static void btrfsic_bio_end_io(struct bio *bp, int bio_error_status);
340 static void btrfsic_bh_end_io(struct buffer_head *bh, int uptodate);
341 static int btrfsic_is_block_ref_by_superblock(const struct btrfsic_state *state,
342 const struct btrfsic_block *block,
343 int recursion_level);
344 static int btrfsic_check_all_ref_blocks(struct btrfsic_state *state,
345 struct btrfsic_block *const block,
346 int recursion_level);
347 static void btrfsic_print_add_link(const struct btrfsic_state *state,
348 const struct btrfsic_block_link *l);
349 static void btrfsic_print_rem_link(const struct btrfsic_state *state,
350 const struct btrfsic_block_link *l);
351 static char btrfsic_get_block_type(const struct btrfsic_state *state,
352 const struct btrfsic_block *block);
353 static void btrfsic_dump_tree(const struct btrfsic_state *state);
354 static void btrfsic_dump_tree_sub(const struct btrfsic_state *state,
355 const struct btrfsic_block *block,
357 static struct btrfsic_block_link *btrfsic_block_link_lookup_or_add(
358 struct btrfsic_state *state,
359 struct btrfsic_block_data_ctx *next_block_ctx,
360 struct btrfsic_block *next_block,
361 struct btrfsic_block *from_block,
362 u64 parent_generation);
363 static struct btrfsic_block *btrfsic_block_lookup_or_add(
364 struct btrfsic_state *state,
365 struct btrfsic_block_data_ctx *block_ctx,
366 const char *additional_string,
372 static int btrfsic_process_superblock_dev_mirror(
373 struct btrfsic_state *state,
374 struct btrfsic_dev_state *dev_state,
375 struct btrfs_device *device,
376 int superblock_mirror_num,
377 struct btrfsic_dev_state **selected_dev_state,
378 struct btrfs_super_block *selected_super);
379 static struct btrfsic_dev_state *btrfsic_dev_state_lookup(
380 struct block_device *bdev);
381 static void btrfsic_cmp_log_and_dev_bytenr(struct btrfsic_state *state,
383 struct btrfsic_dev_state *dev_state,
386 static struct mutex btrfsic_mutex;
387 static int btrfsic_is_initialized;
388 static struct btrfsic_dev_state_hashtable btrfsic_dev_state_hashtable;
391 static void btrfsic_block_init(struct btrfsic_block *b)
393 b->magic_num = BTRFSIC_BLOCK_MAGIC_NUMBER;
396 b->logical_bytenr = 0;
397 b->generation = BTRFSIC_GENERATION_UNKNOWN;
398 b->disk_key.objectid = 0;
399 b->disk_key.type = 0;
400 b->disk_key.offset = 0;
402 b->is_superblock = 0;
404 b->iodone_w_error = 0;
405 b->never_written = 0;
407 b->next_in_same_bio = NULL;
408 b->orig_bio_bh_private = NULL;
409 b->orig_bio_bh_end_io.bio = NULL;
410 INIT_LIST_HEAD(&b->collision_resolving_node);
411 INIT_LIST_HEAD(&b->all_blocks_node);
412 INIT_LIST_HEAD(&b->ref_to_list);
413 INIT_LIST_HEAD(&b->ref_from_list);
414 b->submit_bio_bh_rw = 0;
418 static struct btrfsic_block *btrfsic_block_alloc(void)
420 struct btrfsic_block *b;
422 b = kzalloc(sizeof(*b), GFP_NOFS);
424 btrfsic_block_init(b);
429 static void btrfsic_block_free(struct btrfsic_block *b)
431 BUG_ON(!(NULL == b || BTRFSIC_BLOCK_MAGIC_NUMBER == b->magic_num));
435 static void btrfsic_block_link_init(struct btrfsic_block_link *l)
437 l->magic_num = BTRFSIC_BLOCK_LINK_MAGIC_NUMBER;
439 INIT_LIST_HEAD(&l->node_ref_to);
440 INIT_LIST_HEAD(&l->node_ref_from);
441 INIT_LIST_HEAD(&l->collision_resolving_node);
442 l->block_ref_to = NULL;
443 l->block_ref_from = NULL;
446 static struct btrfsic_block_link *btrfsic_block_link_alloc(void)
448 struct btrfsic_block_link *l;
450 l = kzalloc(sizeof(*l), GFP_NOFS);
452 btrfsic_block_link_init(l);
457 static void btrfsic_block_link_free(struct btrfsic_block_link *l)
459 BUG_ON(!(NULL == l || BTRFSIC_BLOCK_LINK_MAGIC_NUMBER == l->magic_num));
463 static void btrfsic_dev_state_init(struct btrfsic_dev_state *ds)
465 ds->magic_num = BTRFSIC_DEV2STATE_MAGIC_NUMBER;
469 INIT_LIST_HEAD(&ds->collision_resolving_node);
470 ds->last_flush_gen = 0;
471 btrfsic_block_init(&ds->dummy_block_for_bio_bh_flush);
472 ds->dummy_block_for_bio_bh_flush.is_iodone = 1;
473 ds->dummy_block_for_bio_bh_flush.dev_state = ds;
476 static struct btrfsic_dev_state *btrfsic_dev_state_alloc(void)
478 struct btrfsic_dev_state *ds;
480 ds = kzalloc(sizeof(*ds), GFP_NOFS);
482 btrfsic_dev_state_init(ds);
487 static void btrfsic_dev_state_free(struct btrfsic_dev_state *ds)
489 BUG_ON(!(NULL == ds ||
490 BTRFSIC_DEV2STATE_MAGIC_NUMBER == ds->magic_num));
494 static void btrfsic_block_hashtable_init(struct btrfsic_block_hashtable *h)
498 for (i = 0; i < BTRFSIC_BLOCK_HASHTABLE_SIZE; i++)
499 INIT_LIST_HEAD(h->table + i);
502 static void btrfsic_block_hashtable_add(struct btrfsic_block *b,
503 struct btrfsic_block_hashtable *h)
505 const unsigned int hashval =
506 (((unsigned int)(b->dev_bytenr >> 16)) ^
507 ((unsigned int)((uintptr_t)b->dev_state->bdev))) &
508 (BTRFSIC_BLOCK_HASHTABLE_SIZE - 1);
510 list_add(&b->collision_resolving_node, h->table + hashval);
513 static void btrfsic_block_hashtable_remove(struct btrfsic_block *b)
515 list_del(&b->collision_resolving_node);
518 static struct btrfsic_block *btrfsic_block_hashtable_lookup(
519 struct block_device *bdev,
521 struct btrfsic_block_hashtable *h)
523 const unsigned int hashval =
524 (((unsigned int)(dev_bytenr >> 16)) ^
525 ((unsigned int)((uintptr_t)bdev))) &
526 (BTRFSIC_BLOCK_HASHTABLE_SIZE - 1);
527 struct list_head *elem;
529 list_for_each(elem, h->table + hashval) {
530 struct btrfsic_block *const b =
531 list_entry(elem, struct btrfsic_block,
532 collision_resolving_node);
534 if (b->dev_state->bdev == bdev && b->dev_bytenr == dev_bytenr)
541 static void btrfsic_block_link_hashtable_init(
542 struct btrfsic_block_link_hashtable *h)
546 for (i = 0; i < BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE; i++)
547 INIT_LIST_HEAD(h->table + i);
550 static void btrfsic_block_link_hashtable_add(
551 struct btrfsic_block_link *l,
552 struct btrfsic_block_link_hashtable *h)
554 const unsigned int hashval =
555 (((unsigned int)(l->block_ref_to->dev_bytenr >> 16)) ^
556 ((unsigned int)(l->block_ref_from->dev_bytenr >> 16)) ^
557 ((unsigned int)((uintptr_t)l->block_ref_to->dev_state->bdev)) ^
558 ((unsigned int)((uintptr_t)l->block_ref_from->dev_state->bdev)))
559 & (BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE - 1);
561 BUG_ON(NULL == l->block_ref_to);
562 BUG_ON(NULL == l->block_ref_from);
563 list_add(&l->collision_resolving_node, h->table + hashval);
566 static void btrfsic_block_link_hashtable_remove(struct btrfsic_block_link *l)
568 list_del(&l->collision_resolving_node);
571 static struct btrfsic_block_link *btrfsic_block_link_hashtable_lookup(
572 struct block_device *bdev_ref_to,
573 u64 dev_bytenr_ref_to,
574 struct block_device *bdev_ref_from,
575 u64 dev_bytenr_ref_from,
576 struct btrfsic_block_link_hashtable *h)
578 const unsigned int hashval =
579 (((unsigned int)(dev_bytenr_ref_to >> 16)) ^
580 ((unsigned int)(dev_bytenr_ref_from >> 16)) ^
581 ((unsigned int)((uintptr_t)bdev_ref_to)) ^
582 ((unsigned int)((uintptr_t)bdev_ref_from))) &
583 (BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE - 1);
584 struct list_head *elem;
586 list_for_each(elem, h->table + hashval) {
587 struct btrfsic_block_link *const l =
588 list_entry(elem, struct btrfsic_block_link,
589 collision_resolving_node);
591 BUG_ON(NULL == l->block_ref_to);
592 BUG_ON(NULL == l->block_ref_from);
593 if (l->block_ref_to->dev_state->bdev == bdev_ref_to &&
594 l->block_ref_to->dev_bytenr == dev_bytenr_ref_to &&
595 l->block_ref_from->dev_state->bdev == bdev_ref_from &&
596 l->block_ref_from->dev_bytenr == dev_bytenr_ref_from)
603 static void btrfsic_dev_state_hashtable_init(
604 struct btrfsic_dev_state_hashtable *h)
608 for (i = 0; i < BTRFSIC_DEV2STATE_HASHTABLE_SIZE; i++)
609 INIT_LIST_HEAD(h->table + i);
612 static void btrfsic_dev_state_hashtable_add(
613 struct btrfsic_dev_state *ds,
614 struct btrfsic_dev_state_hashtable *h)
616 const unsigned int hashval =
617 (((unsigned int)((uintptr_t)ds->bdev)) &
618 (BTRFSIC_DEV2STATE_HASHTABLE_SIZE - 1));
620 list_add(&ds->collision_resolving_node, h->table + hashval);
623 static void btrfsic_dev_state_hashtable_remove(struct btrfsic_dev_state *ds)
625 list_del(&ds->collision_resolving_node);
628 static struct btrfsic_dev_state *btrfsic_dev_state_hashtable_lookup(
629 struct block_device *bdev,
630 struct btrfsic_dev_state_hashtable *h)
632 const unsigned int hashval =
633 (((unsigned int)((uintptr_t)bdev)) &
634 (BTRFSIC_DEV2STATE_HASHTABLE_SIZE - 1));
635 struct list_head *elem;
637 list_for_each(elem, h->table + hashval) {
638 struct btrfsic_dev_state *const ds =
639 list_entry(elem, struct btrfsic_dev_state,
640 collision_resolving_node);
642 if (ds->bdev == bdev)
649 static int btrfsic_process_superblock(struct btrfsic_state *state,
650 struct btrfs_fs_devices *fs_devices)
653 struct btrfs_super_block *selected_super;
654 struct list_head *dev_head = &fs_devices->devices;
655 struct btrfs_device *device;
656 struct btrfsic_dev_state *selected_dev_state = NULL;
659 BUG_ON(NULL == state);
660 selected_super = kzalloc(sizeof(*selected_super), GFP_NOFS);
661 if (NULL == selected_super) {
662 printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
666 list_for_each_entry(device, dev_head, dev_list) {
668 struct btrfsic_dev_state *dev_state;
670 if (!device->bdev || !device->name)
673 dev_state = btrfsic_dev_state_lookup(device->bdev);
674 BUG_ON(NULL == dev_state);
675 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
676 ret = btrfsic_process_superblock_dev_mirror(
677 state, dev_state, device, i,
678 &selected_dev_state, selected_super);
679 if (0 != ret && 0 == i) {
680 kfree(selected_super);
686 if (NULL == state->latest_superblock) {
687 printk(KERN_INFO "btrfsic: no superblock found!\n");
688 kfree(selected_super);
692 state->csum_size = btrfs_super_csum_size(selected_super);
694 for (pass = 0; pass < 3; pass++) {
701 next_bytenr = btrfs_super_root(selected_super);
702 if (state->print_mask &
703 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
704 printk(KERN_INFO "root@%llu\n", next_bytenr);
707 next_bytenr = btrfs_super_chunk_root(selected_super);
708 if (state->print_mask &
709 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
710 printk(KERN_INFO "chunk@%llu\n", next_bytenr);
713 next_bytenr = btrfs_super_log_root(selected_super);
714 if (0 == next_bytenr)
716 if (state->print_mask &
717 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
718 printk(KERN_INFO "log@%llu\n", next_bytenr);
723 btrfs_num_copies(state->root->fs_info,
724 next_bytenr, state->metablock_size);
725 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
726 printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
727 next_bytenr, num_copies);
729 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
730 struct btrfsic_block *next_block;
731 struct btrfsic_block_data_ctx tmp_next_block_ctx;
732 struct btrfsic_block_link *l;
734 ret = btrfsic_map_block(state, next_bytenr,
735 state->metablock_size,
739 printk(KERN_INFO "btrfsic:"
740 " btrfsic_map_block(root @%llu,"
741 " mirror %d) failed!\n",
742 next_bytenr, mirror_num);
743 kfree(selected_super);
747 next_block = btrfsic_block_hashtable_lookup(
748 tmp_next_block_ctx.dev->bdev,
749 tmp_next_block_ctx.dev_bytenr,
750 &state->block_hashtable);
751 BUG_ON(NULL == next_block);
753 l = btrfsic_block_link_hashtable_lookup(
754 tmp_next_block_ctx.dev->bdev,
755 tmp_next_block_ctx.dev_bytenr,
756 state->latest_superblock->dev_state->
758 state->latest_superblock->dev_bytenr,
759 &state->block_link_hashtable);
762 ret = btrfsic_read_block(state, &tmp_next_block_ctx);
763 if (ret < (int)PAGE_CACHE_SIZE) {
765 "btrfsic: read @logical %llu failed!\n",
766 tmp_next_block_ctx.start);
767 btrfsic_release_block_ctx(&tmp_next_block_ctx);
768 kfree(selected_super);
772 ret = btrfsic_process_metablock(state,
775 BTRFS_MAX_LEVEL + 3, 1);
776 btrfsic_release_block_ctx(&tmp_next_block_ctx);
780 kfree(selected_super);
784 static int btrfsic_process_superblock_dev_mirror(
785 struct btrfsic_state *state,
786 struct btrfsic_dev_state *dev_state,
787 struct btrfs_device *device,
788 int superblock_mirror_num,
789 struct btrfsic_dev_state **selected_dev_state,
790 struct btrfs_super_block *selected_super)
792 struct btrfs_super_block *super_tmp;
794 struct buffer_head *bh;
795 struct btrfsic_block *superblock_tmp;
797 struct block_device *const superblock_bdev = device->bdev;
799 /* super block bytenr is always the unmapped device bytenr */
800 dev_bytenr = btrfs_sb_offset(superblock_mirror_num);
801 if (dev_bytenr + BTRFS_SUPER_INFO_SIZE > device->total_bytes)
803 bh = __bread(superblock_bdev, dev_bytenr / 4096,
804 BTRFS_SUPER_INFO_SIZE);
807 super_tmp = (struct btrfs_super_block *)
808 (bh->b_data + (dev_bytenr & 4095));
810 if (btrfs_super_bytenr(super_tmp) != dev_bytenr ||
811 btrfs_super_magic(super_tmp) != BTRFS_MAGIC ||
812 memcmp(device->uuid, super_tmp->dev_item.uuid, BTRFS_UUID_SIZE) ||
813 btrfs_super_nodesize(super_tmp) != state->metablock_size ||
814 btrfs_super_leafsize(super_tmp) != state->metablock_size ||
815 btrfs_super_sectorsize(super_tmp) != state->datablock_size) {
821 btrfsic_block_hashtable_lookup(superblock_bdev,
823 &state->block_hashtable);
824 if (NULL == superblock_tmp) {
825 superblock_tmp = btrfsic_block_alloc();
826 if (NULL == superblock_tmp) {
827 printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
831 /* for superblock, only the dev_bytenr makes sense */
832 superblock_tmp->dev_bytenr = dev_bytenr;
833 superblock_tmp->dev_state = dev_state;
834 superblock_tmp->logical_bytenr = dev_bytenr;
835 superblock_tmp->generation = btrfs_super_generation(super_tmp);
836 superblock_tmp->is_metadata = 1;
837 superblock_tmp->is_superblock = 1;
838 superblock_tmp->is_iodone = 1;
839 superblock_tmp->never_written = 0;
840 superblock_tmp->mirror_num = 1 + superblock_mirror_num;
841 if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
842 printk_in_rcu(KERN_INFO "New initial S-block (bdev %p, %s)"
843 " @%llu (%s/%llu/%d)\n",
845 rcu_str_deref(device->name), dev_bytenr,
846 dev_state->name, dev_bytenr,
847 superblock_mirror_num);
848 list_add(&superblock_tmp->all_blocks_node,
849 &state->all_blocks_list);
850 btrfsic_block_hashtable_add(superblock_tmp,
851 &state->block_hashtable);
854 /* select the one with the highest generation field */
855 if (btrfs_super_generation(super_tmp) >
856 state->max_superblock_generation ||
857 0 == state->max_superblock_generation) {
858 memcpy(selected_super, super_tmp, sizeof(*selected_super));
859 *selected_dev_state = dev_state;
860 state->max_superblock_generation =
861 btrfs_super_generation(super_tmp);
862 state->latest_superblock = superblock_tmp;
865 for (pass = 0; pass < 3; pass++) {
869 const char *additional_string = NULL;
870 struct btrfs_disk_key tmp_disk_key;
872 tmp_disk_key.type = BTRFS_ROOT_ITEM_KEY;
873 tmp_disk_key.offset = 0;
876 btrfs_set_disk_key_objectid(&tmp_disk_key,
877 BTRFS_ROOT_TREE_OBJECTID);
878 additional_string = "initial root ";
879 next_bytenr = btrfs_super_root(super_tmp);
882 btrfs_set_disk_key_objectid(&tmp_disk_key,
883 BTRFS_CHUNK_TREE_OBJECTID);
884 additional_string = "initial chunk ";
885 next_bytenr = btrfs_super_chunk_root(super_tmp);
888 btrfs_set_disk_key_objectid(&tmp_disk_key,
889 BTRFS_TREE_LOG_OBJECTID);
890 additional_string = "initial log ";
891 next_bytenr = btrfs_super_log_root(super_tmp);
892 if (0 == next_bytenr)
898 btrfs_num_copies(state->root->fs_info,
899 next_bytenr, state->metablock_size);
900 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
901 printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
902 next_bytenr, num_copies);
903 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
904 struct btrfsic_block *next_block;
905 struct btrfsic_block_data_ctx tmp_next_block_ctx;
906 struct btrfsic_block_link *l;
908 if (btrfsic_map_block(state, next_bytenr,
909 state->metablock_size,
912 printk(KERN_INFO "btrfsic: btrfsic_map_block("
913 "bytenr @%llu, mirror %d) failed!\n",
914 next_bytenr, mirror_num);
919 next_block = btrfsic_block_lookup_or_add(
920 state, &tmp_next_block_ctx,
921 additional_string, 1, 1, 0,
923 if (NULL == next_block) {
924 btrfsic_release_block_ctx(&tmp_next_block_ctx);
929 next_block->disk_key = tmp_disk_key;
930 next_block->generation = BTRFSIC_GENERATION_UNKNOWN;
931 l = btrfsic_block_link_lookup_or_add(
932 state, &tmp_next_block_ctx,
933 next_block, superblock_tmp,
934 BTRFSIC_GENERATION_UNKNOWN);
935 btrfsic_release_block_ctx(&tmp_next_block_ctx);
942 if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_ALL_TREES)
943 btrfsic_dump_tree_sub(state, superblock_tmp, 0);
949 static struct btrfsic_stack_frame *btrfsic_stack_frame_alloc(void)
951 struct btrfsic_stack_frame *sf;
953 sf = kzalloc(sizeof(*sf), GFP_NOFS);
955 printk(KERN_INFO "btrfsic: alloc memory failed!\n");
957 sf->magic = BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER;
961 static void btrfsic_stack_frame_free(struct btrfsic_stack_frame *sf)
963 BUG_ON(!(NULL == sf ||
964 BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER == sf->magic));
968 static int btrfsic_process_metablock(
969 struct btrfsic_state *state,
970 struct btrfsic_block *const first_block,
971 struct btrfsic_block_data_ctx *const first_block_ctx,
972 int first_limit_nesting, int force_iodone_flag)
974 struct btrfsic_stack_frame initial_stack_frame = { 0 };
975 struct btrfsic_stack_frame *sf;
976 struct btrfsic_stack_frame *next_stack;
977 struct btrfs_header *const first_hdr =
978 (struct btrfs_header *)first_block_ctx->datav[0];
981 sf = &initial_stack_frame;
984 sf->limit_nesting = first_limit_nesting;
985 sf->block = first_block;
986 sf->block_ctx = first_block_ctx;
987 sf->next_block = NULL;
991 continue_with_new_stack_frame:
992 sf->block->generation = le64_to_cpu(sf->hdr->generation);
993 if (0 == sf->hdr->level) {
994 struct btrfs_leaf *const leafhdr =
995 (struct btrfs_leaf *)sf->hdr;
998 sf->nr = btrfs_stack_header_nritems(&leafhdr->header);
1000 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1002 "leaf %llu items %d generation %llu"
1004 sf->block_ctx->start, sf->nr,
1005 btrfs_stack_header_generation(
1007 btrfs_stack_header_owner(
1011 continue_with_current_leaf_stack_frame:
1012 if (0 == sf->num_copies || sf->mirror_num > sf->num_copies) {
1017 if (sf->i < sf->nr) {
1018 struct btrfs_item disk_item;
1019 u32 disk_item_offset =
1020 (uintptr_t)(leafhdr->items + sf->i) -
1022 struct btrfs_disk_key *disk_key;
1027 if (disk_item_offset + sizeof(struct btrfs_item) >
1028 sf->block_ctx->len) {
1029 leaf_item_out_of_bounce_error:
1031 "btrfsic: leaf item out of bounce at logical %llu, dev %s\n",
1032 sf->block_ctx->start,
1033 sf->block_ctx->dev->name);
1034 goto one_stack_frame_backwards;
1036 btrfsic_read_from_block_data(sf->block_ctx,
1039 sizeof(struct btrfs_item));
1040 item_offset = btrfs_stack_item_offset(&disk_item);
1041 item_size = btrfs_stack_item_size(&disk_item);
1042 disk_key = &disk_item.key;
1043 type = btrfs_disk_key_type(disk_key);
1045 if (BTRFS_ROOT_ITEM_KEY == type) {
1046 struct btrfs_root_item root_item;
1047 u32 root_item_offset;
1050 root_item_offset = item_offset +
1051 offsetof(struct btrfs_leaf, items);
1052 if (root_item_offset + item_size >
1054 goto leaf_item_out_of_bounce_error;
1055 btrfsic_read_from_block_data(
1056 sf->block_ctx, &root_item,
1059 next_bytenr = btrfs_root_bytenr(&root_item);
1062 btrfsic_create_link_to_next_block(
1068 &sf->next_block_ctx,
1074 btrfs_root_generation(
1077 goto one_stack_frame_backwards;
1079 if (NULL != sf->next_block) {
1080 struct btrfs_header *const next_hdr =
1081 (struct btrfs_header *)
1082 sf->next_block_ctx.datav[0];
1085 btrfsic_stack_frame_alloc();
1086 if (NULL == next_stack) {
1087 btrfsic_release_block_ctx(
1090 goto one_stack_frame_backwards;
1094 next_stack->block = sf->next_block;
1095 next_stack->block_ctx =
1096 &sf->next_block_ctx;
1097 next_stack->next_block = NULL;
1098 next_stack->hdr = next_hdr;
1099 next_stack->limit_nesting =
1100 sf->limit_nesting - 1;
1101 next_stack->prev = sf;
1103 goto continue_with_new_stack_frame;
1105 } else if (BTRFS_EXTENT_DATA_KEY == type &&
1106 state->include_extent_data) {
1107 sf->error = btrfsic_handle_extent_data(
1114 goto one_stack_frame_backwards;
1117 goto continue_with_current_leaf_stack_frame;
1120 struct btrfs_node *const nodehdr = (struct btrfs_node *)sf->hdr;
1123 sf->nr = btrfs_stack_header_nritems(&nodehdr->header);
1125 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1126 printk(KERN_INFO "node %llu level %d items %d"
1127 " generation %llu owner %llu\n",
1128 sf->block_ctx->start,
1129 nodehdr->header.level, sf->nr,
1130 btrfs_stack_header_generation(
1132 btrfs_stack_header_owner(
1136 continue_with_current_node_stack_frame:
1137 if (0 == sf->num_copies || sf->mirror_num > sf->num_copies) {
1142 if (sf->i < sf->nr) {
1143 struct btrfs_key_ptr key_ptr;
1147 key_ptr_offset = (uintptr_t)(nodehdr->ptrs + sf->i) -
1149 if (key_ptr_offset + sizeof(struct btrfs_key_ptr) >
1150 sf->block_ctx->len) {
1152 "btrfsic: node item out of bounce at logical %llu, dev %s\n",
1153 sf->block_ctx->start,
1154 sf->block_ctx->dev->name);
1155 goto one_stack_frame_backwards;
1157 btrfsic_read_from_block_data(
1158 sf->block_ctx, &key_ptr, key_ptr_offset,
1159 sizeof(struct btrfs_key_ptr));
1160 next_bytenr = btrfs_stack_key_blockptr(&key_ptr);
1162 sf->error = btrfsic_create_link_to_next_block(
1168 &sf->next_block_ctx,
1174 btrfs_stack_key_generation(&key_ptr));
1176 goto one_stack_frame_backwards;
1178 if (NULL != sf->next_block) {
1179 struct btrfs_header *const next_hdr =
1180 (struct btrfs_header *)
1181 sf->next_block_ctx.datav[0];
1183 next_stack = btrfsic_stack_frame_alloc();
1184 if (NULL == next_stack)
1185 goto one_stack_frame_backwards;
1188 next_stack->block = sf->next_block;
1189 next_stack->block_ctx = &sf->next_block_ctx;
1190 next_stack->next_block = NULL;
1191 next_stack->hdr = next_hdr;
1192 next_stack->limit_nesting =
1193 sf->limit_nesting - 1;
1194 next_stack->prev = sf;
1196 goto continue_with_new_stack_frame;
1199 goto continue_with_current_node_stack_frame;
1203 one_stack_frame_backwards:
1204 if (NULL != sf->prev) {
1205 struct btrfsic_stack_frame *const prev = sf->prev;
1207 /* the one for the initial block is freed in the caller */
1208 btrfsic_release_block_ctx(sf->block_ctx);
1211 prev->error = sf->error;
1212 btrfsic_stack_frame_free(sf);
1214 goto one_stack_frame_backwards;
1217 btrfsic_stack_frame_free(sf);
1219 goto continue_with_new_stack_frame;
1221 BUG_ON(&initial_stack_frame != sf);
1227 static void btrfsic_read_from_block_data(
1228 struct btrfsic_block_data_ctx *block_ctx,
1229 void *dstv, u32 offset, size_t len)
1232 size_t offset_in_page;
1234 char *dst = (char *)dstv;
1235 size_t start_offset = block_ctx->start & ((u64)PAGE_CACHE_SIZE - 1);
1236 unsigned long i = (start_offset + offset) >> PAGE_CACHE_SHIFT;
1238 WARN_ON(offset + len > block_ctx->len);
1239 offset_in_page = (start_offset + offset) & (PAGE_CACHE_SIZE - 1);
1242 cur = min(len, ((size_t)PAGE_CACHE_SIZE - offset_in_page));
1243 BUG_ON(i >= (block_ctx->len + PAGE_CACHE_SIZE - 1) >>
1245 kaddr = block_ctx->datav[i];
1246 memcpy(dst, kaddr + offset_in_page, cur);
1255 static int btrfsic_create_link_to_next_block(
1256 struct btrfsic_state *state,
1257 struct btrfsic_block *block,
1258 struct btrfsic_block_data_ctx *block_ctx,
1261 struct btrfsic_block_data_ctx *next_block_ctx,
1262 struct btrfsic_block **next_blockp,
1263 int force_iodone_flag,
1264 int *num_copiesp, int *mirror_nump,
1265 struct btrfs_disk_key *disk_key,
1266 u64 parent_generation)
1268 struct btrfsic_block *next_block = NULL;
1270 struct btrfsic_block_link *l;
1271 int did_alloc_block_link;
1272 int block_was_created;
1274 *next_blockp = NULL;
1275 if (0 == *num_copiesp) {
1277 btrfs_num_copies(state->root->fs_info,
1278 next_bytenr, state->metablock_size);
1279 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
1280 printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
1281 next_bytenr, *num_copiesp);
1285 if (*mirror_nump > *num_copiesp)
1288 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1290 "btrfsic_create_link_to_next_block(mirror_num=%d)\n",
1292 ret = btrfsic_map_block(state, next_bytenr,
1293 state->metablock_size,
1294 next_block_ctx, *mirror_nump);
1297 "btrfsic: btrfsic_map_block(@%llu, mirror=%d) failed!\n",
1298 next_bytenr, *mirror_nump);
1299 btrfsic_release_block_ctx(next_block_ctx);
1300 *next_blockp = NULL;
1304 next_block = btrfsic_block_lookup_or_add(state,
1305 next_block_ctx, "referenced ",
1306 1, force_iodone_flag,
1309 &block_was_created);
1310 if (NULL == next_block) {
1311 btrfsic_release_block_ctx(next_block_ctx);
1312 *next_blockp = NULL;
1315 if (block_was_created) {
1317 next_block->generation = BTRFSIC_GENERATION_UNKNOWN;
1319 if (next_block->logical_bytenr != next_bytenr &&
1320 !(!next_block->is_metadata &&
1321 0 == next_block->logical_bytenr)) {
1323 "Referenced block @%llu (%s/%llu/%d)"
1324 " found in hash table, %c,"
1325 " bytenr mismatch (!= stored %llu).\n",
1326 next_bytenr, next_block_ctx->dev->name,
1327 next_block_ctx->dev_bytenr, *mirror_nump,
1328 btrfsic_get_block_type(state, next_block),
1329 next_block->logical_bytenr);
1330 } else if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1332 "Referenced block @%llu (%s/%llu/%d)"
1333 " found in hash table, %c.\n",
1334 next_bytenr, next_block_ctx->dev->name,
1335 next_block_ctx->dev_bytenr, *mirror_nump,
1336 btrfsic_get_block_type(state, next_block));
1337 next_block->logical_bytenr = next_bytenr;
1339 next_block->mirror_num = *mirror_nump;
1340 l = btrfsic_block_link_hashtable_lookup(
1341 next_block_ctx->dev->bdev,
1342 next_block_ctx->dev_bytenr,
1343 block_ctx->dev->bdev,
1344 block_ctx->dev_bytenr,
1345 &state->block_link_hashtable);
1348 next_block->disk_key = *disk_key;
1350 l = btrfsic_block_link_alloc();
1352 printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
1353 btrfsic_release_block_ctx(next_block_ctx);
1354 *next_blockp = NULL;
1358 did_alloc_block_link = 1;
1359 l->block_ref_to = next_block;
1360 l->block_ref_from = block;
1362 l->parent_generation = parent_generation;
1364 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1365 btrfsic_print_add_link(state, l);
1367 list_add(&l->node_ref_to, &block->ref_to_list);
1368 list_add(&l->node_ref_from, &next_block->ref_from_list);
1370 btrfsic_block_link_hashtable_add(l,
1371 &state->block_link_hashtable);
1373 did_alloc_block_link = 0;
1374 if (0 == limit_nesting) {
1376 l->parent_generation = parent_generation;
1377 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1378 btrfsic_print_add_link(state, l);
1382 if (limit_nesting > 0 && did_alloc_block_link) {
1383 ret = btrfsic_read_block(state, next_block_ctx);
1384 if (ret < (int)next_block_ctx->len) {
1386 "btrfsic: read block @logical %llu failed!\n",
1388 btrfsic_release_block_ctx(next_block_ctx);
1389 *next_blockp = NULL;
1393 *next_blockp = next_block;
1395 *next_blockp = NULL;
1402 static int btrfsic_handle_extent_data(
1403 struct btrfsic_state *state,
1404 struct btrfsic_block *block,
1405 struct btrfsic_block_data_ctx *block_ctx,
1406 u32 item_offset, int force_iodone_flag)
1409 struct btrfs_file_extent_item file_extent_item;
1410 u64 file_extent_item_offset;
1414 struct btrfsic_block_link *l;
1416 file_extent_item_offset = offsetof(struct btrfs_leaf, items) +
1418 if (file_extent_item_offset +
1419 offsetof(struct btrfs_file_extent_item, disk_num_bytes) >
1422 "btrfsic: file item out of bounce at logical %llu, dev %s\n",
1423 block_ctx->start, block_ctx->dev->name);
1427 btrfsic_read_from_block_data(block_ctx, &file_extent_item,
1428 file_extent_item_offset,
1429 offsetof(struct btrfs_file_extent_item, disk_num_bytes));
1430 if (BTRFS_FILE_EXTENT_REG != file_extent_item.type ||
1431 btrfs_stack_file_extent_disk_bytenr(&file_extent_item) == 0) {
1432 if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE)
1433 printk(KERN_INFO "extent_data: type %u, disk_bytenr = %llu\n",
1434 file_extent_item.type,
1435 btrfs_stack_file_extent_disk_bytenr(
1436 &file_extent_item));
1440 if (file_extent_item_offset + sizeof(struct btrfs_file_extent_item) >
1443 "btrfsic: file item out of bounce at logical %llu, dev %s\n",
1444 block_ctx->start, block_ctx->dev->name);
1447 btrfsic_read_from_block_data(block_ctx, &file_extent_item,
1448 file_extent_item_offset,
1449 sizeof(struct btrfs_file_extent_item));
1450 next_bytenr = btrfs_stack_file_extent_disk_bytenr(&file_extent_item) +
1451 btrfs_stack_file_extent_offset(&file_extent_item);
1452 generation = btrfs_stack_file_extent_generation(&file_extent_item);
1453 num_bytes = btrfs_stack_file_extent_num_bytes(&file_extent_item);
1454 generation = btrfs_stack_file_extent_generation(&file_extent_item);
1456 if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE)
1457 printk(KERN_INFO "extent_data: type %u, disk_bytenr = %llu,"
1458 " offset = %llu, num_bytes = %llu\n",
1459 file_extent_item.type,
1460 btrfs_stack_file_extent_disk_bytenr(&file_extent_item),
1461 btrfs_stack_file_extent_offset(&file_extent_item),
1463 while (num_bytes > 0) {
1468 if (num_bytes > state->datablock_size)
1469 chunk_len = state->datablock_size;
1471 chunk_len = num_bytes;
1474 btrfs_num_copies(state->root->fs_info,
1475 next_bytenr, state->datablock_size);
1476 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
1477 printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
1478 next_bytenr, num_copies);
1479 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
1480 struct btrfsic_block_data_ctx next_block_ctx;
1481 struct btrfsic_block *next_block;
1482 int block_was_created;
1484 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1485 printk(KERN_INFO "btrfsic_handle_extent_data("
1486 "mirror_num=%d)\n", mirror_num);
1487 if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE)
1489 "\tdisk_bytenr = %llu, num_bytes %u\n",
1490 next_bytenr, chunk_len);
1491 ret = btrfsic_map_block(state, next_bytenr,
1492 chunk_len, &next_block_ctx,
1496 "btrfsic: btrfsic_map_block(@%llu,"
1497 " mirror=%d) failed!\n",
1498 next_bytenr, mirror_num);
1502 next_block = btrfsic_block_lookup_or_add(
1510 &block_was_created);
1511 if (NULL == next_block) {
1513 "btrfsic: error, kmalloc failed!\n");
1514 btrfsic_release_block_ctx(&next_block_ctx);
1517 if (!block_was_created) {
1518 if (next_block->logical_bytenr != next_bytenr &&
1519 !(!next_block->is_metadata &&
1520 0 == next_block->logical_bytenr)) {
1523 " @%llu (%s/%llu/%d)"
1524 " found in hash table, D,"
1526 " (!= stored %llu).\n",
1528 next_block_ctx.dev->name,
1529 next_block_ctx.dev_bytenr,
1531 next_block->logical_bytenr);
1533 next_block->logical_bytenr = next_bytenr;
1534 next_block->mirror_num = mirror_num;
1537 l = btrfsic_block_link_lookup_or_add(state,
1541 btrfsic_release_block_ctx(&next_block_ctx);
1546 next_bytenr += chunk_len;
1547 num_bytes -= chunk_len;
1553 static int btrfsic_map_block(struct btrfsic_state *state, u64 bytenr, u32 len,
1554 struct btrfsic_block_data_ctx *block_ctx_out,
1559 struct btrfs_bio *multi = NULL;
1560 struct btrfs_device *device;
1563 ret = btrfs_map_block(state->root->fs_info, READ,
1564 bytenr, &length, &multi, mirror_num);
1567 block_ctx_out->start = 0;
1568 block_ctx_out->dev_bytenr = 0;
1569 block_ctx_out->len = 0;
1570 block_ctx_out->dev = NULL;
1571 block_ctx_out->datav = NULL;
1572 block_ctx_out->pagev = NULL;
1573 block_ctx_out->mem_to_free = NULL;
1578 device = multi->stripes[0].dev;
1579 block_ctx_out->dev = btrfsic_dev_state_lookup(device->bdev);
1580 block_ctx_out->dev_bytenr = multi->stripes[0].physical;
1581 block_ctx_out->start = bytenr;
1582 block_ctx_out->len = len;
1583 block_ctx_out->datav = NULL;
1584 block_ctx_out->pagev = NULL;
1585 block_ctx_out->mem_to_free = NULL;
1588 if (NULL == block_ctx_out->dev) {
1590 printk(KERN_INFO "btrfsic: error, cannot lookup dev (#1)!\n");
1596 static int btrfsic_map_superblock(struct btrfsic_state *state, u64 bytenr,
1597 u32 len, struct block_device *bdev,
1598 struct btrfsic_block_data_ctx *block_ctx_out)
1600 block_ctx_out->dev = btrfsic_dev_state_lookup(bdev);
1601 block_ctx_out->dev_bytenr = bytenr;
1602 block_ctx_out->start = bytenr;
1603 block_ctx_out->len = len;
1604 block_ctx_out->datav = NULL;
1605 block_ctx_out->pagev = NULL;
1606 block_ctx_out->mem_to_free = NULL;
1607 if (NULL != block_ctx_out->dev) {
1610 printk(KERN_INFO "btrfsic: error, cannot lookup dev (#2)!\n");
1615 static void btrfsic_release_block_ctx(struct btrfsic_block_data_ctx *block_ctx)
1617 if (block_ctx->mem_to_free) {
1618 unsigned int num_pages;
1620 BUG_ON(!block_ctx->datav);
1621 BUG_ON(!block_ctx->pagev);
1622 num_pages = (block_ctx->len + (u64)PAGE_CACHE_SIZE - 1) >>
1624 while (num_pages > 0) {
1626 if (block_ctx->datav[num_pages]) {
1627 kunmap(block_ctx->pagev[num_pages]);
1628 block_ctx->datav[num_pages] = NULL;
1630 if (block_ctx->pagev[num_pages]) {
1631 __free_page(block_ctx->pagev[num_pages]);
1632 block_ctx->pagev[num_pages] = NULL;
1636 kfree(block_ctx->mem_to_free);
1637 block_ctx->mem_to_free = NULL;
1638 block_ctx->pagev = NULL;
1639 block_ctx->datav = NULL;
1643 static int btrfsic_read_block(struct btrfsic_state *state,
1644 struct btrfsic_block_data_ctx *block_ctx)
1646 unsigned int num_pages;
1651 BUG_ON(block_ctx->datav);
1652 BUG_ON(block_ctx->pagev);
1653 BUG_ON(block_ctx->mem_to_free);
1654 if (block_ctx->dev_bytenr & ((u64)PAGE_CACHE_SIZE - 1)) {
1656 "btrfsic: read_block() with unaligned bytenr %llu\n",
1657 block_ctx->dev_bytenr);
1661 num_pages = (block_ctx->len + (u64)PAGE_CACHE_SIZE - 1) >>
1663 block_ctx->mem_to_free = kzalloc((sizeof(*block_ctx->datav) +
1664 sizeof(*block_ctx->pagev)) *
1665 num_pages, GFP_NOFS);
1666 if (!block_ctx->mem_to_free)
1668 block_ctx->datav = block_ctx->mem_to_free;
1669 block_ctx->pagev = (struct page **)(block_ctx->datav + num_pages);
1670 for (i = 0; i < num_pages; i++) {
1671 block_ctx->pagev[i] = alloc_page(GFP_NOFS);
1672 if (!block_ctx->pagev[i])
1676 dev_bytenr = block_ctx->dev_bytenr;
1677 for (i = 0; i < num_pages;) {
1680 DECLARE_COMPLETION_ONSTACK(complete);
1682 bio = btrfs_io_bio_alloc(GFP_NOFS, num_pages - i);
1685 "btrfsic: bio_alloc() for %u pages failed!\n",
1689 bio->bi_bdev = block_ctx->dev->bdev;
1690 bio->bi_sector = dev_bytenr >> 9;
1691 bio->bi_end_io = btrfsic_complete_bio_end_io;
1692 bio->bi_private = &complete;
1694 for (j = i; j < num_pages; j++) {
1695 ret = bio_add_page(bio, block_ctx->pagev[j],
1696 PAGE_CACHE_SIZE, 0);
1697 if (PAGE_CACHE_SIZE != ret)
1702 "btrfsic: error, failed to add a single page!\n");
1705 submit_bio(READ, bio);
1707 /* this will also unplug the queue */
1708 wait_for_completion(&complete);
1710 if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) {
1712 "btrfsic: read error at logical %llu dev %s!\n",
1713 block_ctx->start, block_ctx->dev->name);
1718 dev_bytenr += (j - i) * PAGE_CACHE_SIZE;
1721 for (i = 0; i < num_pages; i++) {
1722 block_ctx->datav[i] = kmap(block_ctx->pagev[i]);
1723 if (!block_ctx->datav[i]) {
1724 printk(KERN_INFO "btrfsic: kmap() failed (dev %s)!\n",
1725 block_ctx->dev->name);
1730 return block_ctx->len;
1733 static void btrfsic_complete_bio_end_io(struct bio *bio, int err)
1735 complete((struct completion *)bio->bi_private);
1738 static void btrfsic_dump_database(struct btrfsic_state *state)
1740 struct list_head *elem_all;
1742 BUG_ON(NULL == state);
1744 printk(KERN_INFO "all_blocks_list:\n");
1745 list_for_each(elem_all, &state->all_blocks_list) {
1746 const struct btrfsic_block *const b_all =
1747 list_entry(elem_all, struct btrfsic_block,
1749 struct list_head *elem_ref_to;
1750 struct list_head *elem_ref_from;
1752 printk(KERN_INFO "%c-block @%llu (%s/%llu/%d)\n",
1753 btrfsic_get_block_type(state, b_all),
1754 b_all->logical_bytenr, b_all->dev_state->name,
1755 b_all->dev_bytenr, b_all->mirror_num);
1757 list_for_each(elem_ref_to, &b_all->ref_to_list) {
1758 const struct btrfsic_block_link *const l =
1759 list_entry(elem_ref_to,
1760 struct btrfsic_block_link,
1763 printk(KERN_INFO " %c @%llu (%s/%llu/%d)"
1765 " %c @%llu (%s/%llu/%d)\n",
1766 btrfsic_get_block_type(state, b_all),
1767 b_all->logical_bytenr, b_all->dev_state->name,
1768 b_all->dev_bytenr, b_all->mirror_num,
1770 btrfsic_get_block_type(state, l->block_ref_to),
1771 l->block_ref_to->logical_bytenr,
1772 l->block_ref_to->dev_state->name,
1773 l->block_ref_to->dev_bytenr,
1774 l->block_ref_to->mirror_num);
1777 list_for_each(elem_ref_from, &b_all->ref_from_list) {
1778 const struct btrfsic_block_link *const l =
1779 list_entry(elem_ref_from,
1780 struct btrfsic_block_link,
1783 printk(KERN_INFO " %c @%llu (%s/%llu/%d)"
1785 " %c @%llu (%s/%llu/%d)\n",
1786 btrfsic_get_block_type(state, b_all),
1787 b_all->logical_bytenr, b_all->dev_state->name,
1788 b_all->dev_bytenr, b_all->mirror_num,
1790 btrfsic_get_block_type(state, l->block_ref_from),
1791 l->block_ref_from->logical_bytenr,
1792 l->block_ref_from->dev_state->name,
1793 l->block_ref_from->dev_bytenr,
1794 l->block_ref_from->mirror_num);
1797 printk(KERN_INFO "\n");
1802 * Test whether the disk block contains a tree block (leaf or node)
1803 * (note that this test fails for the super block)
1805 static int btrfsic_test_for_metadata(struct btrfsic_state *state,
1806 char **datav, unsigned int num_pages)
1808 struct btrfs_header *h;
1809 u8 csum[BTRFS_CSUM_SIZE];
1813 if (num_pages * PAGE_CACHE_SIZE < state->metablock_size)
1814 return 1; /* not metadata */
1815 num_pages = state->metablock_size >> PAGE_CACHE_SHIFT;
1816 h = (struct btrfs_header *)datav[0];
1818 if (memcmp(h->fsid, state->root->fs_info->fsid, BTRFS_UUID_SIZE))
1821 for (i = 0; i < num_pages; i++) {
1822 u8 *data = i ? datav[i] : (datav[i] + BTRFS_CSUM_SIZE);
1823 size_t sublen = i ? PAGE_CACHE_SIZE :
1824 (PAGE_CACHE_SIZE - BTRFS_CSUM_SIZE);
1826 crc = crc32c(crc, data, sublen);
1828 btrfs_csum_final(crc, csum);
1829 if (memcmp(csum, h->csum, state->csum_size))
1832 return 0; /* is metadata */
1835 static void btrfsic_process_written_block(struct btrfsic_dev_state *dev_state,
1836 u64 dev_bytenr, char **mapped_datav,
1837 unsigned int num_pages,
1838 struct bio *bio, int *bio_is_patched,
1839 struct buffer_head *bh,
1840 int submit_bio_bh_rw)
1843 struct btrfsic_block *block;
1844 struct btrfsic_block_data_ctx block_ctx;
1846 struct btrfsic_state *state = dev_state->state;
1847 struct block_device *bdev = dev_state->bdev;
1848 unsigned int processed_len;
1850 if (NULL != bio_is_patched)
1851 *bio_is_patched = 0;
1858 is_metadata = (0 == btrfsic_test_for_metadata(state, mapped_datav,
1861 block = btrfsic_block_hashtable_lookup(bdev, dev_bytenr,
1862 &state->block_hashtable);
1863 if (NULL != block) {
1865 struct list_head *elem_ref_to;
1866 struct list_head *tmp_ref_to;
1868 if (block->is_superblock) {
1869 bytenr = btrfs_super_bytenr((struct btrfs_super_block *)
1871 if (num_pages * PAGE_CACHE_SIZE <
1872 BTRFS_SUPER_INFO_SIZE) {
1874 "btrfsic: cannot work with too short bios!\n");
1878 BUG_ON(BTRFS_SUPER_INFO_SIZE & (PAGE_CACHE_SIZE - 1));
1879 processed_len = BTRFS_SUPER_INFO_SIZE;
1880 if (state->print_mask &
1881 BTRFSIC_PRINT_MASK_TREE_BEFORE_SB_WRITE) {
1883 "[before new superblock is written]:\n");
1884 btrfsic_dump_tree_sub(state, block, 0);
1888 if (!block->is_superblock) {
1889 if (num_pages * PAGE_CACHE_SIZE <
1890 state->metablock_size) {
1892 "btrfsic: cannot work with too short bios!\n");
1895 processed_len = state->metablock_size;
1896 bytenr = btrfs_stack_header_bytenr(
1897 (struct btrfs_header *)
1899 btrfsic_cmp_log_and_dev_bytenr(state, bytenr,
1903 if (block->logical_bytenr != bytenr &&
1904 !(!block->is_metadata &&
1905 block->logical_bytenr == 0))
1907 "Written block @%llu (%s/%llu/%d)"
1908 " found in hash table, %c,"
1910 " (!= stored %llu).\n",
1911 bytenr, dev_state->name, dev_bytenr,
1913 btrfsic_get_block_type(state, block),
1914 block->logical_bytenr);
1915 else if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1917 "Written block @%llu (%s/%llu/%d)"
1918 " found in hash table, %c.\n",
1919 bytenr, dev_state->name, dev_bytenr,
1921 btrfsic_get_block_type(state, block));
1922 block->logical_bytenr = bytenr;
1924 if (num_pages * PAGE_CACHE_SIZE <
1925 state->datablock_size) {
1927 "btrfsic: cannot work with too short bios!\n");
1930 processed_len = state->datablock_size;
1931 bytenr = block->logical_bytenr;
1932 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1934 "Written block @%llu (%s/%llu/%d)"
1935 " found in hash table, %c.\n",
1936 bytenr, dev_state->name, dev_bytenr,
1938 btrfsic_get_block_type(state, block));
1941 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1943 "ref_to_list: %cE, ref_from_list: %cE\n",
1944 list_empty(&block->ref_to_list) ? ' ' : '!',
1945 list_empty(&block->ref_from_list) ? ' ' : '!');
1946 if (btrfsic_is_block_ref_by_superblock(state, block, 0)) {
1947 printk(KERN_INFO "btrfs: attempt to overwrite %c-block"
1948 " @%llu (%s/%llu/%d), old(gen=%llu,"
1949 " objectid=%llu, type=%d, offset=%llu),"
1951 " which is referenced by most recent superblock"
1952 " (superblockgen=%llu)!\n",
1953 btrfsic_get_block_type(state, block), bytenr,
1954 dev_state->name, dev_bytenr, block->mirror_num,
1956 btrfs_disk_key_objectid(&block->disk_key),
1957 block->disk_key.type,
1958 btrfs_disk_key_offset(&block->disk_key),
1959 btrfs_stack_header_generation(
1960 (struct btrfs_header *) mapped_datav[0]),
1961 state->max_superblock_generation);
1962 btrfsic_dump_tree(state);
1965 if (!block->is_iodone && !block->never_written) {
1966 printk(KERN_INFO "btrfs: attempt to overwrite %c-block"
1967 " @%llu (%s/%llu/%d), oldgen=%llu, newgen=%llu,"
1968 " which is not yet iodone!\n",
1969 btrfsic_get_block_type(state, block), bytenr,
1970 dev_state->name, dev_bytenr, block->mirror_num,
1972 btrfs_stack_header_generation(
1973 (struct btrfs_header *)
1975 /* it would not be safe to go on */
1976 btrfsic_dump_tree(state);
1981 * Clear all references of this block. Do not free
1982 * the block itself even if is not referenced anymore
1983 * because it still carries valueable information
1984 * like whether it was ever written and IO completed.
1986 list_for_each_safe(elem_ref_to, tmp_ref_to,
1987 &block->ref_to_list) {
1988 struct btrfsic_block_link *const l =
1989 list_entry(elem_ref_to,
1990 struct btrfsic_block_link,
1993 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1994 btrfsic_print_rem_link(state, l);
1996 if (0 == l->ref_cnt) {
1997 list_del(&l->node_ref_to);
1998 list_del(&l->node_ref_from);
1999 btrfsic_block_link_hashtable_remove(l);
2000 btrfsic_block_link_free(l);
2004 if (block->is_superblock)
2005 ret = btrfsic_map_superblock(state, bytenr,
2009 ret = btrfsic_map_block(state, bytenr, processed_len,
2013 "btrfsic: btrfsic_map_block(root @%llu)"
2014 " failed!\n", bytenr);
2017 block_ctx.datav = mapped_datav;
2018 /* the following is required in case of writes to mirrors,
2019 * use the same that was used for the lookup */
2020 block_ctx.dev = dev_state;
2021 block_ctx.dev_bytenr = dev_bytenr;
2023 if (is_metadata || state->include_extent_data) {
2024 block->never_written = 0;
2025 block->iodone_w_error = 0;
2027 block->is_iodone = 0;
2028 BUG_ON(NULL == bio_is_patched);
2029 if (!*bio_is_patched) {
2030 block->orig_bio_bh_private =
2032 block->orig_bio_bh_end_io.bio =
2034 block->next_in_same_bio = NULL;
2035 bio->bi_private = block;
2036 bio->bi_end_io = btrfsic_bio_end_io;
2037 *bio_is_patched = 1;
2039 struct btrfsic_block *chained_block =
2040 (struct btrfsic_block *)
2043 BUG_ON(NULL == chained_block);
2044 block->orig_bio_bh_private =
2045 chained_block->orig_bio_bh_private;
2046 block->orig_bio_bh_end_io.bio =
2047 chained_block->orig_bio_bh_end_io.
2049 block->next_in_same_bio = chained_block;
2050 bio->bi_private = block;
2052 } else if (NULL != bh) {
2053 block->is_iodone = 0;
2054 block->orig_bio_bh_private = bh->b_private;
2055 block->orig_bio_bh_end_io.bh = bh->b_end_io;
2056 block->next_in_same_bio = NULL;
2057 bh->b_private = block;
2058 bh->b_end_io = btrfsic_bh_end_io;
2060 block->is_iodone = 1;
2061 block->orig_bio_bh_private = NULL;
2062 block->orig_bio_bh_end_io.bio = NULL;
2063 block->next_in_same_bio = NULL;
2067 block->flush_gen = dev_state->last_flush_gen + 1;
2068 block->submit_bio_bh_rw = submit_bio_bh_rw;
2070 block->logical_bytenr = bytenr;
2071 block->is_metadata = 1;
2072 if (block->is_superblock) {
2073 BUG_ON(PAGE_CACHE_SIZE !=
2074 BTRFS_SUPER_INFO_SIZE);
2075 ret = btrfsic_process_written_superblock(
2078 (struct btrfs_super_block *)
2080 if (state->print_mask &
2081 BTRFSIC_PRINT_MASK_TREE_AFTER_SB_WRITE) {
2083 "[after new superblock is written]:\n");
2084 btrfsic_dump_tree_sub(state, block, 0);
2087 block->mirror_num = 0; /* unknown */
2088 ret = btrfsic_process_metablock(
2096 "btrfsic: btrfsic_process_metablock"
2097 "(root @%llu) failed!\n",
2100 block->is_metadata = 0;
2101 block->mirror_num = 0; /* unknown */
2102 block->generation = BTRFSIC_GENERATION_UNKNOWN;
2103 if (!state->include_extent_data
2104 && list_empty(&block->ref_from_list)) {
2106 * disk block is overwritten with extent
2107 * data (not meta data) and we are configured
2108 * to not include extent data: take the
2109 * chance and free the block's memory
2111 btrfsic_block_hashtable_remove(block);
2112 list_del(&block->all_blocks_node);
2113 btrfsic_block_free(block);
2116 btrfsic_release_block_ctx(&block_ctx);
2118 /* block has not been found in hash table */
2122 processed_len = state->datablock_size;
2123 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2124 printk(KERN_INFO "Written block (%s/%llu/?)"
2125 " !found in hash table, D.\n",
2126 dev_state->name, dev_bytenr);
2127 if (!state->include_extent_data) {
2128 /* ignore that written D block */
2132 /* this is getting ugly for the
2133 * include_extent_data case... */
2134 bytenr = 0; /* unknown */
2135 block_ctx.start = bytenr;
2136 block_ctx.len = processed_len;
2137 block_ctx.mem_to_free = NULL;
2138 block_ctx.pagev = NULL;
2140 processed_len = state->metablock_size;
2141 bytenr = btrfs_stack_header_bytenr(
2142 (struct btrfs_header *)
2144 btrfsic_cmp_log_and_dev_bytenr(state, bytenr, dev_state,
2146 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2148 "Written block @%llu (%s/%llu/?)"
2149 " !found in hash table, M.\n",
2150 bytenr, dev_state->name, dev_bytenr);
2152 ret = btrfsic_map_block(state, bytenr, processed_len,
2156 "btrfsic: btrfsic_map_block(root @%llu)"
2162 block_ctx.datav = mapped_datav;
2163 /* the following is required in case of writes to mirrors,
2164 * use the same that was used for the lookup */
2165 block_ctx.dev = dev_state;
2166 block_ctx.dev_bytenr = dev_bytenr;
2168 block = btrfsic_block_alloc();
2169 if (NULL == block) {
2170 printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
2171 btrfsic_release_block_ctx(&block_ctx);
2174 block->dev_state = dev_state;
2175 block->dev_bytenr = dev_bytenr;
2176 block->logical_bytenr = bytenr;
2177 block->is_metadata = is_metadata;
2178 block->never_written = 0;
2179 block->iodone_w_error = 0;
2180 block->mirror_num = 0; /* unknown */
2181 block->flush_gen = dev_state->last_flush_gen + 1;
2182 block->submit_bio_bh_rw = submit_bio_bh_rw;
2184 block->is_iodone = 0;
2185 BUG_ON(NULL == bio_is_patched);
2186 if (!*bio_is_patched) {
2187 block->orig_bio_bh_private = bio->bi_private;
2188 block->orig_bio_bh_end_io.bio = bio->bi_end_io;
2189 block->next_in_same_bio = NULL;
2190 bio->bi_private = block;
2191 bio->bi_end_io = btrfsic_bio_end_io;
2192 *bio_is_patched = 1;
2194 struct btrfsic_block *chained_block =
2195 (struct btrfsic_block *)
2198 BUG_ON(NULL == chained_block);
2199 block->orig_bio_bh_private =
2200 chained_block->orig_bio_bh_private;
2201 block->orig_bio_bh_end_io.bio =
2202 chained_block->orig_bio_bh_end_io.bio;
2203 block->next_in_same_bio = chained_block;
2204 bio->bi_private = block;
2206 } else if (NULL != bh) {
2207 block->is_iodone = 0;
2208 block->orig_bio_bh_private = bh->b_private;
2209 block->orig_bio_bh_end_io.bh = bh->b_end_io;
2210 block->next_in_same_bio = NULL;
2211 bh->b_private = block;
2212 bh->b_end_io = btrfsic_bh_end_io;
2214 block->is_iodone = 1;
2215 block->orig_bio_bh_private = NULL;
2216 block->orig_bio_bh_end_io.bio = NULL;
2217 block->next_in_same_bio = NULL;
2219 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2221 "New written %c-block @%llu (%s/%llu/%d)\n",
2222 is_metadata ? 'M' : 'D',
2223 block->logical_bytenr, block->dev_state->name,
2224 block->dev_bytenr, block->mirror_num);
2225 list_add(&block->all_blocks_node, &state->all_blocks_list);
2226 btrfsic_block_hashtable_add(block, &state->block_hashtable);
2229 ret = btrfsic_process_metablock(state, block,
2233 "btrfsic: process_metablock(root @%llu)"
2237 btrfsic_release_block_ctx(&block_ctx);
2241 BUG_ON(!processed_len);
2242 dev_bytenr += processed_len;
2243 mapped_datav += processed_len >> PAGE_CACHE_SHIFT;
2244 num_pages -= processed_len >> PAGE_CACHE_SHIFT;
2248 static void btrfsic_bio_end_io(struct bio *bp, int bio_error_status)
2250 struct btrfsic_block *block = (struct btrfsic_block *)bp->bi_private;
2253 /* mutex is not held! This is not save if IO is not yet completed
2256 if (bio_error_status)
2259 BUG_ON(NULL == block);
2260 bp->bi_private = block->orig_bio_bh_private;
2261 bp->bi_end_io = block->orig_bio_bh_end_io.bio;
2264 struct btrfsic_block *next_block;
2265 struct btrfsic_dev_state *const dev_state = block->dev_state;
2267 if ((dev_state->state->print_mask &
2268 BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
2270 "bio_end_io(err=%d) for %c @%llu (%s/%llu/%d)\n",
2272 btrfsic_get_block_type(dev_state->state, block),
2273 block->logical_bytenr, dev_state->name,
2274 block->dev_bytenr, block->mirror_num);
2275 next_block = block->next_in_same_bio;
2276 block->iodone_w_error = iodone_w_error;
2277 if (block->submit_bio_bh_rw & REQ_FLUSH) {
2278 dev_state->last_flush_gen++;
2279 if ((dev_state->state->print_mask &
2280 BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
2282 "bio_end_io() new %s flush_gen=%llu\n",
2284 dev_state->last_flush_gen);
2286 if (block->submit_bio_bh_rw & REQ_FUA)
2287 block->flush_gen = 0; /* FUA completed means block is
2289 block->is_iodone = 1; /* for FLUSH, this releases the block */
2291 } while (NULL != block);
2293 bp->bi_end_io(bp, bio_error_status);
2296 static void btrfsic_bh_end_io(struct buffer_head *bh, int uptodate)
2298 struct btrfsic_block *block = (struct btrfsic_block *)bh->b_private;
2299 int iodone_w_error = !uptodate;
2300 struct btrfsic_dev_state *dev_state;
2302 BUG_ON(NULL == block);
2303 dev_state = block->dev_state;
2304 if ((dev_state->state->print_mask & BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
2306 "bh_end_io(error=%d) for %c @%llu (%s/%llu/%d)\n",
2308 btrfsic_get_block_type(dev_state->state, block),
2309 block->logical_bytenr, block->dev_state->name,
2310 block->dev_bytenr, block->mirror_num);
2312 block->iodone_w_error = iodone_w_error;
2313 if (block->submit_bio_bh_rw & REQ_FLUSH) {
2314 dev_state->last_flush_gen++;
2315 if ((dev_state->state->print_mask &
2316 BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
2318 "bh_end_io() new %s flush_gen=%llu\n",
2319 dev_state->name, dev_state->last_flush_gen);
2321 if (block->submit_bio_bh_rw & REQ_FUA)
2322 block->flush_gen = 0; /* FUA completed means block is on disk */
2324 bh->b_private = block->orig_bio_bh_private;
2325 bh->b_end_io = block->orig_bio_bh_end_io.bh;
2326 block->is_iodone = 1; /* for FLUSH, this releases the block */
2327 bh->b_end_io(bh, uptodate);
2330 static int btrfsic_process_written_superblock(
2331 struct btrfsic_state *state,
2332 struct btrfsic_block *const superblock,
2333 struct btrfs_super_block *const super_hdr)
2337 superblock->generation = btrfs_super_generation(super_hdr);
2338 if (!(superblock->generation > state->max_superblock_generation ||
2339 0 == state->max_superblock_generation)) {
2340 if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
2342 "btrfsic: superblock @%llu (%s/%llu/%d)"
2343 " with old gen %llu <= %llu\n",
2344 superblock->logical_bytenr,
2345 superblock->dev_state->name,
2346 superblock->dev_bytenr, superblock->mirror_num,
2347 btrfs_super_generation(super_hdr),
2348 state->max_superblock_generation);
2350 if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
2352 "btrfsic: got new superblock @%llu (%s/%llu/%d)"
2353 " with new gen %llu > %llu\n",
2354 superblock->logical_bytenr,
2355 superblock->dev_state->name,
2356 superblock->dev_bytenr, superblock->mirror_num,
2357 btrfs_super_generation(super_hdr),
2358 state->max_superblock_generation);
2360 state->max_superblock_generation =
2361 btrfs_super_generation(super_hdr);
2362 state->latest_superblock = superblock;
2365 for (pass = 0; pass < 3; pass++) {
2368 struct btrfsic_block *next_block;
2369 struct btrfsic_block_data_ctx tmp_next_block_ctx;
2370 struct btrfsic_block_link *l;
2373 const char *additional_string = NULL;
2374 struct btrfs_disk_key tmp_disk_key = {0};
2376 btrfs_set_disk_key_objectid(&tmp_disk_key,
2377 BTRFS_ROOT_ITEM_KEY);
2378 btrfs_set_disk_key_objectid(&tmp_disk_key, 0);
2382 btrfs_set_disk_key_objectid(&tmp_disk_key,
2383 BTRFS_ROOT_TREE_OBJECTID);
2384 additional_string = "root ";
2385 next_bytenr = btrfs_super_root(super_hdr);
2386 if (state->print_mask &
2387 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
2388 printk(KERN_INFO "root@%llu\n", next_bytenr);
2391 btrfs_set_disk_key_objectid(&tmp_disk_key,
2392 BTRFS_CHUNK_TREE_OBJECTID);
2393 additional_string = "chunk ";
2394 next_bytenr = btrfs_super_chunk_root(super_hdr);
2395 if (state->print_mask &
2396 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
2397 printk(KERN_INFO "chunk@%llu\n", next_bytenr);
2400 btrfs_set_disk_key_objectid(&tmp_disk_key,
2401 BTRFS_TREE_LOG_OBJECTID);
2402 additional_string = "log ";
2403 next_bytenr = btrfs_super_log_root(super_hdr);
2404 if (0 == next_bytenr)
2406 if (state->print_mask &
2407 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
2408 printk(KERN_INFO "log@%llu\n", next_bytenr);
2413 btrfs_num_copies(state->root->fs_info,
2414 next_bytenr, BTRFS_SUPER_INFO_SIZE);
2415 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
2416 printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
2417 next_bytenr, num_copies);
2418 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
2421 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2423 "btrfsic_process_written_superblock("
2424 "mirror_num=%d)\n", mirror_num);
2425 ret = btrfsic_map_block(state, next_bytenr,
2426 BTRFS_SUPER_INFO_SIZE,
2427 &tmp_next_block_ctx,
2431 "btrfsic: btrfsic_map_block(@%llu,"
2432 " mirror=%d) failed!\n",
2433 next_bytenr, mirror_num);
2437 next_block = btrfsic_block_lookup_or_add(
2439 &tmp_next_block_ctx,
2444 if (NULL == next_block) {
2446 "btrfsic: error, kmalloc failed!\n");
2447 btrfsic_release_block_ctx(&tmp_next_block_ctx);
2451 next_block->disk_key = tmp_disk_key;
2453 next_block->generation =
2454 BTRFSIC_GENERATION_UNKNOWN;
2455 l = btrfsic_block_link_lookup_or_add(
2457 &tmp_next_block_ctx,
2460 BTRFSIC_GENERATION_UNKNOWN);
2461 btrfsic_release_block_ctx(&tmp_next_block_ctx);
2467 if (WARN_ON(-1 == btrfsic_check_all_ref_blocks(state, superblock, 0)))
2468 btrfsic_dump_tree(state);
2473 static int btrfsic_check_all_ref_blocks(struct btrfsic_state *state,
2474 struct btrfsic_block *const block,
2475 int recursion_level)
2477 struct list_head *elem_ref_to;
2480 if (recursion_level >= 3 + BTRFS_MAX_LEVEL) {
2482 * Note that this situation can happen and does not
2483 * indicate an error in regular cases. It happens
2484 * when disk blocks are freed and later reused.
2485 * The check-integrity module is not aware of any
2486 * block free operations, it just recognizes block
2487 * write operations. Therefore it keeps the linkage
2488 * information for a block until a block is
2489 * rewritten. This can temporarily cause incorrect
2490 * and even circular linkage informations. This
2491 * causes no harm unless such blocks are referenced
2492 * by the most recent super block.
2494 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2496 "btrfsic: abort cyclic linkage (case 1).\n");
2502 * This algorithm is recursive because the amount of used stack
2503 * space is very small and the max recursion depth is limited.
2505 list_for_each(elem_ref_to, &block->ref_to_list) {
2506 const struct btrfsic_block_link *const l =
2507 list_entry(elem_ref_to, struct btrfsic_block_link,
2510 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2512 "rl=%d, %c @%llu (%s/%llu/%d)"
2513 " %u* refers to %c @%llu (%s/%llu/%d)\n",
2515 btrfsic_get_block_type(state, block),
2516 block->logical_bytenr, block->dev_state->name,
2517 block->dev_bytenr, block->mirror_num,
2519 btrfsic_get_block_type(state, l->block_ref_to),
2520 l->block_ref_to->logical_bytenr,
2521 l->block_ref_to->dev_state->name,
2522 l->block_ref_to->dev_bytenr,
2523 l->block_ref_to->mirror_num);
2524 if (l->block_ref_to->never_written) {
2525 printk(KERN_INFO "btrfs: attempt to write superblock"
2526 " which references block %c @%llu (%s/%llu/%d)"
2527 " which is never written!\n",
2528 btrfsic_get_block_type(state, l->block_ref_to),
2529 l->block_ref_to->logical_bytenr,
2530 l->block_ref_to->dev_state->name,
2531 l->block_ref_to->dev_bytenr,
2532 l->block_ref_to->mirror_num);
2534 } else if (!l->block_ref_to->is_iodone) {
2535 printk(KERN_INFO "btrfs: attempt to write superblock"
2536 " which references block %c @%llu (%s/%llu/%d)"
2537 " which is not yet iodone!\n",
2538 btrfsic_get_block_type(state, l->block_ref_to),
2539 l->block_ref_to->logical_bytenr,
2540 l->block_ref_to->dev_state->name,
2541 l->block_ref_to->dev_bytenr,
2542 l->block_ref_to->mirror_num);
2544 } else if (l->block_ref_to->iodone_w_error) {
2545 printk(KERN_INFO "btrfs: attempt to write superblock"
2546 " which references block %c @%llu (%s/%llu/%d)"
2547 " which has write error!\n",
2548 btrfsic_get_block_type(state, l->block_ref_to),
2549 l->block_ref_to->logical_bytenr,
2550 l->block_ref_to->dev_state->name,
2551 l->block_ref_to->dev_bytenr,
2552 l->block_ref_to->mirror_num);
2554 } else if (l->parent_generation !=
2555 l->block_ref_to->generation &&
2556 BTRFSIC_GENERATION_UNKNOWN !=
2557 l->parent_generation &&
2558 BTRFSIC_GENERATION_UNKNOWN !=
2559 l->block_ref_to->generation) {
2560 printk(KERN_INFO "btrfs: attempt to write superblock"
2561 " which references block %c @%llu (%s/%llu/%d)"
2562 " with generation %llu !="
2563 " parent generation %llu!\n",
2564 btrfsic_get_block_type(state, l->block_ref_to),
2565 l->block_ref_to->logical_bytenr,
2566 l->block_ref_to->dev_state->name,
2567 l->block_ref_to->dev_bytenr,
2568 l->block_ref_to->mirror_num,
2569 l->block_ref_to->generation,
2570 l->parent_generation);
2572 } else if (l->block_ref_to->flush_gen >
2573 l->block_ref_to->dev_state->last_flush_gen) {
2574 printk(KERN_INFO "btrfs: attempt to write superblock"
2575 " which references block %c @%llu (%s/%llu/%d)"
2576 " which is not flushed out of disk's write cache"
2577 " (block flush_gen=%llu,"
2578 " dev->flush_gen=%llu)!\n",
2579 btrfsic_get_block_type(state, l->block_ref_to),
2580 l->block_ref_to->logical_bytenr,
2581 l->block_ref_to->dev_state->name,
2582 l->block_ref_to->dev_bytenr,
2583 l->block_ref_to->mirror_num, block->flush_gen,
2584 l->block_ref_to->dev_state->last_flush_gen);
2586 } else if (-1 == btrfsic_check_all_ref_blocks(state,
2597 static int btrfsic_is_block_ref_by_superblock(
2598 const struct btrfsic_state *state,
2599 const struct btrfsic_block *block,
2600 int recursion_level)
2602 struct list_head *elem_ref_from;
2604 if (recursion_level >= 3 + BTRFS_MAX_LEVEL) {
2605 /* refer to comment at "abort cyclic linkage (case 1)" */
2606 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2608 "btrfsic: abort cyclic linkage (case 2).\n");
2614 * This algorithm is recursive because the amount of used stack space
2615 * is very small and the max recursion depth is limited.
2617 list_for_each(elem_ref_from, &block->ref_from_list) {
2618 const struct btrfsic_block_link *const l =
2619 list_entry(elem_ref_from, struct btrfsic_block_link,
2622 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2624 "rl=%d, %c @%llu (%s/%llu/%d)"
2625 " is ref %u* from %c @%llu (%s/%llu/%d)\n",
2627 btrfsic_get_block_type(state, block),
2628 block->logical_bytenr, block->dev_state->name,
2629 block->dev_bytenr, block->mirror_num,
2631 btrfsic_get_block_type(state, l->block_ref_from),
2632 l->block_ref_from->logical_bytenr,
2633 l->block_ref_from->dev_state->name,
2634 l->block_ref_from->dev_bytenr,
2635 l->block_ref_from->mirror_num);
2636 if (l->block_ref_from->is_superblock &&
2637 state->latest_superblock->dev_bytenr ==
2638 l->block_ref_from->dev_bytenr &&
2639 state->latest_superblock->dev_state->bdev ==
2640 l->block_ref_from->dev_state->bdev)
2642 else if (btrfsic_is_block_ref_by_superblock(state,
2652 static void btrfsic_print_add_link(const struct btrfsic_state *state,
2653 const struct btrfsic_block_link *l)
2656 "Add %u* link from %c @%llu (%s/%llu/%d)"
2657 " to %c @%llu (%s/%llu/%d).\n",
2659 btrfsic_get_block_type(state, l->block_ref_from),
2660 l->block_ref_from->logical_bytenr,
2661 l->block_ref_from->dev_state->name,
2662 l->block_ref_from->dev_bytenr, l->block_ref_from->mirror_num,
2663 btrfsic_get_block_type(state, l->block_ref_to),
2664 l->block_ref_to->logical_bytenr,
2665 l->block_ref_to->dev_state->name, l->block_ref_to->dev_bytenr,
2666 l->block_ref_to->mirror_num);
2669 static void btrfsic_print_rem_link(const struct btrfsic_state *state,
2670 const struct btrfsic_block_link *l)
2673 "Rem %u* link from %c @%llu (%s/%llu/%d)"
2674 " to %c @%llu (%s/%llu/%d).\n",
2676 btrfsic_get_block_type(state, l->block_ref_from),
2677 l->block_ref_from->logical_bytenr,
2678 l->block_ref_from->dev_state->name,
2679 l->block_ref_from->dev_bytenr, l->block_ref_from->mirror_num,
2680 btrfsic_get_block_type(state, l->block_ref_to),
2681 l->block_ref_to->logical_bytenr,
2682 l->block_ref_to->dev_state->name, l->block_ref_to->dev_bytenr,
2683 l->block_ref_to->mirror_num);
2686 static char btrfsic_get_block_type(const struct btrfsic_state *state,
2687 const struct btrfsic_block *block)
2689 if (block->is_superblock &&
2690 state->latest_superblock->dev_bytenr == block->dev_bytenr &&
2691 state->latest_superblock->dev_state->bdev == block->dev_state->bdev)
2693 else if (block->is_superblock)
2695 else if (block->is_metadata)
2701 static void btrfsic_dump_tree(const struct btrfsic_state *state)
2703 btrfsic_dump_tree_sub(state, state->latest_superblock, 0);
2706 static void btrfsic_dump_tree_sub(const struct btrfsic_state *state,
2707 const struct btrfsic_block *block,
2710 struct list_head *elem_ref_to;
2712 static char buf[80];
2713 int cursor_position;
2716 * Should better fill an on-stack buffer with a complete line and
2717 * dump it at once when it is time to print a newline character.
2721 * This algorithm is recursive because the amount of used stack space
2722 * is very small and the max recursion depth is limited.
2724 indent_add = sprintf(buf, "%c-%llu(%s/%llu/%d)",
2725 btrfsic_get_block_type(state, block),
2726 block->logical_bytenr, block->dev_state->name,
2727 block->dev_bytenr, block->mirror_num);
2728 if (indent_level + indent_add > BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL) {
2733 indent_level += indent_add;
2734 if (list_empty(&block->ref_to_list)) {
2738 if (block->mirror_num > 1 &&
2739 !(state->print_mask & BTRFSIC_PRINT_MASK_TREE_WITH_ALL_MIRRORS)) {
2744 cursor_position = indent_level;
2745 list_for_each(elem_ref_to, &block->ref_to_list) {
2746 const struct btrfsic_block_link *const l =
2747 list_entry(elem_ref_to, struct btrfsic_block_link,
2750 while (cursor_position < indent_level) {
2755 indent_add = sprintf(buf, " %d*--> ", l->ref_cnt);
2757 indent_add = sprintf(buf, " --> ");
2758 if (indent_level + indent_add >
2759 BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL) {
2761 cursor_position = 0;
2767 btrfsic_dump_tree_sub(state, l->block_ref_to,
2768 indent_level + indent_add);
2769 cursor_position = 0;
2773 static struct btrfsic_block_link *btrfsic_block_link_lookup_or_add(
2774 struct btrfsic_state *state,
2775 struct btrfsic_block_data_ctx *next_block_ctx,
2776 struct btrfsic_block *next_block,
2777 struct btrfsic_block *from_block,
2778 u64 parent_generation)
2780 struct btrfsic_block_link *l;
2782 l = btrfsic_block_link_hashtable_lookup(next_block_ctx->dev->bdev,
2783 next_block_ctx->dev_bytenr,
2784 from_block->dev_state->bdev,
2785 from_block->dev_bytenr,
2786 &state->block_link_hashtable);
2788 l = btrfsic_block_link_alloc();
2791 "btrfsic: error, kmalloc" " failed!\n");
2795 l->block_ref_to = next_block;
2796 l->block_ref_from = from_block;
2798 l->parent_generation = parent_generation;
2800 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2801 btrfsic_print_add_link(state, l);
2803 list_add(&l->node_ref_to, &from_block->ref_to_list);
2804 list_add(&l->node_ref_from, &next_block->ref_from_list);
2806 btrfsic_block_link_hashtable_add(l,
2807 &state->block_link_hashtable);
2810 l->parent_generation = parent_generation;
2811 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2812 btrfsic_print_add_link(state, l);
2818 static struct btrfsic_block *btrfsic_block_lookup_or_add(
2819 struct btrfsic_state *state,
2820 struct btrfsic_block_data_ctx *block_ctx,
2821 const char *additional_string,
2828 struct btrfsic_block *block;
2830 block = btrfsic_block_hashtable_lookup(block_ctx->dev->bdev,
2831 block_ctx->dev_bytenr,
2832 &state->block_hashtable);
2833 if (NULL == block) {
2834 struct btrfsic_dev_state *dev_state;
2836 block = btrfsic_block_alloc();
2837 if (NULL == block) {
2838 printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
2841 dev_state = btrfsic_dev_state_lookup(block_ctx->dev->bdev);
2842 if (NULL == dev_state) {
2844 "btrfsic: error, lookup dev_state failed!\n");
2845 btrfsic_block_free(block);
2848 block->dev_state = dev_state;
2849 block->dev_bytenr = block_ctx->dev_bytenr;
2850 block->logical_bytenr = block_ctx->start;
2851 block->is_metadata = is_metadata;
2852 block->is_iodone = is_iodone;
2853 block->never_written = never_written;
2854 block->mirror_num = mirror_num;
2855 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2857 "New %s%c-block @%llu (%s/%llu/%d)\n",
2859 btrfsic_get_block_type(state, block),
2860 block->logical_bytenr, dev_state->name,
2861 block->dev_bytenr, mirror_num);
2862 list_add(&block->all_blocks_node, &state->all_blocks_list);
2863 btrfsic_block_hashtable_add(block, &state->block_hashtable);
2864 if (NULL != was_created)
2867 if (NULL != was_created)
2874 static void btrfsic_cmp_log_and_dev_bytenr(struct btrfsic_state *state,
2876 struct btrfsic_dev_state *dev_state,
2882 struct btrfsic_block_data_ctx block_ctx;
2885 num_copies = btrfs_num_copies(state->root->fs_info,
2886 bytenr, state->metablock_size);
2888 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
2889 ret = btrfsic_map_block(state, bytenr, state->metablock_size,
2890 &block_ctx, mirror_num);
2892 printk(KERN_INFO "btrfsic:"
2893 " btrfsic_map_block(logical @%llu,"
2894 " mirror %d) failed!\n",
2895 bytenr, mirror_num);
2899 if (dev_state->bdev == block_ctx.dev->bdev &&
2900 dev_bytenr == block_ctx.dev_bytenr) {
2902 btrfsic_release_block_ctx(&block_ctx);
2905 btrfsic_release_block_ctx(&block_ctx);
2908 if (WARN_ON(!match)) {
2909 printk(KERN_INFO "btrfs: attempt to write M-block which contains logical bytenr that doesn't map to dev+physical bytenr of submit_bio,"
2910 " buffer->log_bytenr=%llu, submit_bio(bdev=%s,"
2911 " phys_bytenr=%llu)!\n",
2912 bytenr, dev_state->name, dev_bytenr);
2913 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
2914 ret = btrfsic_map_block(state, bytenr,
2915 state->metablock_size,
2916 &block_ctx, mirror_num);
2920 printk(KERN_INFO "Read logical bytenr @%llu maps to"
2922 bytenr, block_ctx.dev->name,
2923 block_ctx.dev_bytenr, mirror_num);
2928 static struct btrfsic_dev_state *btrfsic_dev_state_lookup(
2929 struct block_device *bdev)
2931 struct btrfsic_dev_state *ds;
2933 ds = btrfsic_dev_state_hashtable_lookup(bdev,
2934 &btrfsic_dev_state_hashtable);
2938 int btrfsic_submit_bh(int rw, struct buffer_head *bh)
2940 struct btrfsic_dev_state *dev_state;
2942 if (!btrfsic_is_initialized)
2943 return submit_bh(rw, bh);
2945 mutex_lock(&btrfsic_mutex);
2946 /* since btrfsic_submit_bh() might also be called before
2947 * btrfsic_mount(), this might return NULL */
2948 dev_state = btrfsic_dev_state_lookup(bh->b_bdev);
2950 /* Only called to write the superblock (incl. FLUSH/FUA) */
2951 if (NULL != dev_state &&
2952 (rw & WRITE) && bh->b_size > 0) {
2955 dev_bytenr = 4096 * bh->b_blocknr;
2956 if (dev_state->state->print_mask &
2957 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
2959 "submit_bh(rw=0x%x, blocknr=%llu (bytenr %llu),"
2960 " size=%zu, data=%p, bdev=%p)\n",
2961 rw, (unsigned long long)bh->b_blocknr,
2962 dev_bytenr, bh->b_size, bh->b_data, bh->b_bdev);
2963 btrfsic_process_written_block(dev_state, dev_bytenr,
2964 &bh->b_data, 1, NULL,
2966 } else if (NULL != dev_state && (rw & REQ_FLUSH)) {
2967 if (dev_state->state->print_mask &
2968 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
2970 "submit_bh(rw=0x%x FLUSH, bdev=%p)\n",
2972 if (!dev_state->dummy_block_for_bio_bh_flush.is_iodone) {
2973 if ((dev_state->state->print_mask &
2974 (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
2975 BTRFSIC_PRINT_MASK_VERBOSE)))
2977 "btrfsic_submit_bh(%s) with FLUSH"
2978 " but dummy block already in use"
2982 struct btrfsic_block *const block =
2983 &dev_state->dummy_block_for_bio_bh_flush;
2985 block->is_iodone = 0;
2986 block->never_written = 0;
2987 block->iodone_w_error = 0;
2988 block->flush_gen = dev_state->last_flush_gen + 1;
2989 block->submit_bio_bh_rw = rw;
2990 block->orig_bio_bh_private = bh->b_private;
2991 block->orig_bio_bh_end_io.bh = bh->b_end_io;
2992 block->next_in_same_bio = NULL;
2993 bh->b_private = block;
2994 bh->b_end_io = btrfsic_bh_end_io;
2997 mutex_unlock(&btrfsic_mutex);
2998 return submit_bh(rw, bh);
3001 void btrfsic_submit_bio(int rw, struct bio *bio)
3003 struct btrfsic_dev_state *dev_state;
3005 if (!btrfsic_is_initialized) {
3006 submit_bio(rw, bio);
3010 mutex_lock(&btrfsic_mutex);
3011 /* since btrfsic_submit_bio() is also called before
3012 * btrfsic_mount(), this might return NULL */
3013 dev_state = btrfsic_dev_state_lookup(bio->bi_bdev);
3014 if (NULL != dev_state &&
3015 (rw & WRITE) && NULL != bio->bi_io_vec) {
3019 char **mapped_datav;
3021 dev_bytenr = 512 * bio->bi_sector;
3023 if (dev_state->state->print_mask &
3024 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
3026 "submit_bio(rw=0x%x, bi_vcnt=%u,"
3027 " bi_sector=%llu (bytenr %llu), bi_bdev=%p)\n",
3029 (unsigned long long)bio->bi_sector, dev_bytenr,
3032 mapped_datav = kmalloc(sizeof(*mapped_datav) * bio->bi_vcnt,
3036 for (i = 0; i < bio->bi_vcnt; i++) {
3037 BUG_ON(bio->bi_io_vec[i].bv_len != PAGE_CACHE_SIZE);
3038 mapped_datav[i] = kmap(bio->bi_io_vec[i].bv_page);
3039 if (!mapped_datav[i]) {
3042 kunmap(bio->bi_io_vec[i].bv_page);
3044 kfree(mapped_datav);
3047 if ((BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
3048 BTRFSIC_PRINT_MASK_VERBOSE) ==
3049 (dev_state->state->print_mask &
3050 (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
3051 BTRFSIC_PRINT_MASK_VERBOSE)))
3053 "#%u: page=%p, len=%u, offset=%u\n",
3054 i, bio->bi_io_vec[i].bv_page,
3055 bio->bi_io_vec[i].bv_len,
3056 bio->bi_io_vec[i].bv_offset);
3058 btrfsic_process_written_block(dev_state, dev_bytenr,
3059 mapped_datav, bio->bi_vcnt,
3060 bio, &bio_is_patched,
3064 kunmap(bio->bi_io_vec[i].bv_page);
3066 kfree(mapped_datav);
3067 } else if (NULL != dev_state && (rw & REQ_FLUSH)) {
3068 if (dev_state->state->print_mask &
3069 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
3071 "submit_bio(rw=0x%x FLUSH, bdev=%p)\n",
3073 if (!dev_state->dummy_block_for_bio_bh_flush.is_iodone) {
3074 if ((dev_state->state->print_mask &
3075 (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
3076 BTRFSIC_PRINT_MASK_VERBOSE)))
3078 "btrfsic_submit_bio(%s) with FLUSH"
3079 " but dummy block already in use"
3083 struct btrfsic_block *const block =
3084 &dev_state->dummy_block_for_bio_bh_flush;
3086 block->is_iodone = 0;
3087 block->never_written = 0;
3088 block->iodone_w_error = 0;
3089 block->flush_gen = dev_state->last_flush_gen + 1;
3090 block->submit_bio_bh_rw = rw;
3091 block->orig_bio_bh_private = bio->bi_private;
3092 block->orig_bio_bh_end_io.bio = bio->bi_end_io;
3093 block->next_in_same_bio = NULL;
3094 bio->bi_private = block;
3095 bio->bi_end_io = btrfsic_bio_end_io;
3099 mutex_unlock(&btrfsic_mutex);
3101 submit_bio(rw, bio);
3104 int btrfsic_mount(struct btrfs_root *root,
3105 struct btrfs_fs_devices *fs_devices,
3106 int including_extent_data, u32 print_mask)
3109 struct btrfsic_state *state;
3110 struct list_head *dev_head = &fs_devices->devices;
3111 struct btrfs_device *device;
3113 if (root->nodesize != root->leafsize) {
3115 "btrfsic: cannot handle nodesize %d != leafsize %d!\n",
3116 root->nodesize, root->leafsize);
3119 if (root->nodesize & ((u64)PAGE_CACHE_SIZE - 1)) {
3121 "btrfsic: cannot handle nodesize %d not being a multiple of PAGE_CACHE_SIZE %ld!\n",
3122 root->nodesize, PAGE_CACHE_SIZE);
3125 if (root->leafsize & ((u64)PAGE_CACHE_SIZE - 1)) {
3127 "btrfsic: cannot handle leafsize %d not being a multiple of PAGE_CACHE_SIZE %ld!\n",
3128 root->leafsize, PAGE_CACHE_SIZE);
3131 if (root->sectorsize & ((u64)PAGE_CACHE_SIZE - 1)) {
3133 "btrfsic: cannot handle sectorsize %d not being a multiple of PAGE_CACHE_SIZE %ld!\n",
3134 root->sectorsize, PAGE_CACHE_SIZE);
3137 state = kzalloc(sizeof(*state), GFP_NOFS);
3138 if (NULL == state) {
3139 printk(KERN_INFO "btrfs check-integrity: kmalloc() failed!\n");
3143 if (!btrfsic_is_initialized) {
3144 mutex_init(&btrfsic_mutex);
3145 btrfsic_dev_state_hashtable_init(&btrfsic_dev_state_hashtable);
3146 btrfsic_is_initialized = 1;
3148 mutex_lock(&btrfsic_mutex);
3150 state->print_mask = print_mask;
3151 state->include_extent_data = including_extent_data;
3152 state->csum_size = 0;
3153 state->metablock_size = root->nodesize;
3154 state->datablock_size = root->sectorsize;
3155 INIT_LIST_HEAD(&state->all_blocks_list);
3156 btrfsic_block_hashtable_init(&state->block_hashtable);
3157 btrfsic_block_link_hashtable_init(&state->block_link_hashtable);
3158 state->max_superblock_generation = 0;
3159 state->latest_superblock = NULL;
3161 list_for_each_entry(device, dev_head, dev_list) {
3162 struct btrfsic_dev_state *ds;
3165 if (!device->bdev || !device->name)
3168 ds = btrfsic_dev_state_alloc();
3171 "btrfs check-integrity: kmalloc() failed!\n");
3172 mutex_unlock(&btrfsic_mutex);
3175 ds->bdev = device->bdev;
3177 bdevname(ds->bdev, ds->name);
3178 ds->name[BDEVNAME_SIZE - 1] = '\0';
3179 for (p = ds->name; *p != '\0'; p++);
3180 while (p > ds->name && *p != '/')
3184 strlcpy(ds->name, p, sizeof(ds->name));
3185 btrfsic_dev_state_hashtable_add(ds,
3186 &btrfsic_dev_state_hashtable);
3189 ret = btrfsic_process_superblock(state, fs_devices);
3191 mutex_unlock(&btrfsic_mutex);
3192 btrfsic_unmount(root, fs_devices);
3196 if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_DATABASE)
3197 btrfsic_dump_database(state);
3198 if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_TREE)
3199 btrfsic_dump_tree(state);
3201 mutex_unlock(&btrfsic_mutex);
3205 void btrfsic_unmount(struct btrfs_root *root,
3206 struct btrfs_fs_devices *fs_devices)
3208 struct list_head *elem_all;
3209 struct list_head *tmp_all;
3210 struct btrfsic_state *state;
3211 struct list_head *dev_head = &fs_devices->devices;
3212 struct btrfs_device *device;
3214 if (!btrfsic_is_initialized)
3217 mutex_lock(&btrfsic_mutex);
3220 list_for_each_entry(device, dev_head, dev_list) {
3221 struct btrfsic_dev_state *ds;
3223 if (!device->bdev || !device->name)
3226 ds = btrfsic_dev_state_hashtable_lookup(
3228 &btrfsic_dev_state_hashtable);
3231 btrfsic_dev_state_hashtable_remove(ds);
3232 btrfsic_dev_state_free(ds);
3236 if (NULL == state) {
3238 "btrfsic: error, cannot find state information"
3240 mutex_unlock(&btrfsic_mutex);
3245 * Don't care about keeping the lists' state up to date,
3246 * just free all memory that was allocated dynamically.
3247 * Free the blocks and the block_links.
3249 list_for_each_safe(elem_all, tmp_all, &state->all_blocks_list) {
3250 struct btrfsic_block *const b_all =
3251 list_entry(elem_all, struct btrfsic_block,
3253 struct list_head *elem_ref_to;
3254 struct list_head *tmp_ref_to;
3256 list_for_each_safe(elem_ref_to, tmp_ref_to,
3257 &b_all->ref_to_list) {
3258 struct btrfsic_block_link *const l =
3259 list_entry(elem_ref_to,
3260 struct btrfsic_block_link,
3263 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
3264 btrfsic_print_rem_link(state, l);
3267 if (0 == l->ref_cnt)
3268 btrfsic_block_link_free(l);
3271 if (b_all->is_iodone || b_all->never_written)
3272 btrfsic_block_free(b_all);
3274 printk(KERN_INFO "btrfs: attempt to free %c-block"
3275 " @%llu (%s/%llu/%d) on umount which is"
3276 " not yet iodone!\n",
3277 btrfsic_get_block_type(state, b_all),
3278 b_all->logical_bytenr, b_all->dev_state->name,
3279 b_all->dev_bytenr, b_all->mirror_num);
3282 mutex_unlock(&btrfsic_mutex);