aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorChris Mason <chris.mason@oracle.com>2012-01-16 15:27:58 -0500
committerChris Mason <chris.mason@oracle.com>2012-01-16 15:27:58 -0500
commitc126dea771be1b3c370c0ffc4a09e6a82d492a49 (patch)
tree99fc723ba2e89d767e260244cf8d19467bc68c8b /fs
parent9785dbdf265ddc47d5c88267d89a97648c0dc14b (diff)
parent21adbd5cbb5344a3fca6bb7ddb2ab6cb03c44546 (diff)
Merge branch 'integrity-check-patch-v2' of git://btrfs.giantdisaster.de/git/btrfs into integration
Conflicts: fs/btrfs/ctree.h fs/btrfs/super.c Signed-off-by: Chris Mason <chris.mason@oracle.com>
Diffstat (limited to 'fs')
-rw-r--r--fs/btrfs/Kconfig19
-rw-r--r--fs/btrfs/Makefile1
-rw-r--r--fs/btrfs/check-integrity.c3068
-rw-r--r--fs/btrfs/check-integrity.h36
-rw-r--r--fs/btrfs/ctree.h8
-rw-r--r--fs/btrfs/disk-io.c26
-rw-r--r--fs/btrfs/extent_io.c5
-rw-r--r--fs/btrfs/scrub.c5
-rw-r--r--fs/btrfs/super.c36
-rw-r--r--fs/btrfs/volumes.c7
10 files changed, 3201 insertions, 10 deletions
diff --git a/fs/btrfs/Kconfig b/fs/btrfs/Kconfig
index ecb9fd3be143..d33f01c08b60 100644
--- a/fs/btrfs/Kconfig
+++ b/fs/btrfs/Kconfig
@@ -31,3 +31,22 @@ config BTRFS_FS_POSIX_ACL
31 Linux website <http://acl.bestbits.at/>. 31 Linux website <http://acl.bestbits.at/>.
32 32
33 If you don't know what Access Control Lists are, say N 33 If you don't know what Access Control Lists are, say N
34
35config BTRFS_FS_CHECK_INTEGRITY
36 bool "Btrfs with integrity check tool compiled in (DANGEROUS)"
37 depends on BTRFS_FS
38 help
39 Adds code that examines all block write requests (including
40 writes of the super block). The goal is to verify that the
41 state of the filesystem on disk is always consistent, i.e.,
42 after a power-loss or kernel panic event the filesystem is
43 in a consistent state.
44
45 If the integrity check tool is included and activated in
46 the mount options, plenty of kernel memory is used, and
47 plenty of additional CPU cycles are spent. Enabling this
48 functionality is not intended for normal use.
49
50 In most cases, unless you are a btrfs developer who needs
51 to verify the integrity of (super)-block write requests
52 during the run of a regression test, say N
diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 70798407b9a2..0c4fa2befae7 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -11,3 +11,4 @@ btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
11 reada.o backref.o ulist.o 11 reada.o backref.o ulist.o
12 12
13btrfs-$(CONFIG_BTRFS_FS_POSIX_ACL) += acl.o 13btrfs-$(CONFIG_BTRFS_FS_POSIX_ACL) += acl.o
14btrfs-$(CONFIG_BTRFS_FS_CHECK_INTEGRITY) += check-integrity.o
diff --git a/fs/btrfs/check-integrity.c b/fs/btrfs/check-integrity.c
new file mode 100644
index 000000000000..ad0b3ba735b7
--- /dev/null
+++ b/fs/btrfs/check-integrity.c
@@ -0,0 +1,3068 @@
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 a FLUSH request to the device where
41 * these blocks are located was received and completed.
42 * 2b. All referenced blocks need to have a generation
43 * number which is equal to the parent's number.
44 *
45 * One issue that was found using this module was that the log
46 * tree on disk became temporarily corrupted because disk blocks
47 * that had been in use for the log tree had been freed and
48 * reused too early, while being referenced by the written super
49 * block.
50 *
51 * The search term in the kernel log that can be used to filter
52 * on the existence of detected integrity issues is
53 * "btrfs: attempt".
54 *
55 * The integrity check is enabled via mount options. These
56 * mount options are only supported if the integrity check
57 * tool is compiled by defining BTRFS_FS_CHECK_INTEGRITY.
58 *
59 * Example #1, apply integrity checks to all metadata:
60 * mount /dev/sdb1 /mnt -o check_int
61 *
62 * Example #2, apply integrity checks to all metadata and
63 * to data extents:
64 * mount /dev/sdb1 /mnt -o check_int_data
65 *
66 * Example #3, apply integrity checks to all metadata and dump
67 * the tree that the super block references to kernel messages
68 * each time after a super block was written:
69 * mount /dev/sdb1 /mnt -o check_int,check_int_print_mask=263
70 *
71 * If the integrity check tool is included and activated in
72 * the mount options, plenty of kernel memory is used, and
73 * plenty of additional CPU cycles are spent. Enabling this
74 * functionality is not intended for normal use. In most
75 * cases, unless you are a btrfs developer who needs to verify
76 * the integrity of (super)-block write requests, do not
77 * enable the config option BTRFS_FS_CHECK_INTEGRITY to
78 * include and compile the integrity check tool.
79 */
80
81#include <linux/sched.h>
82#include <linux/slab.h>
83#include <linux/buffer_head.h>
84#include <linux/mutex.h>
85#include <linux/crc32c.h>
86#include <linux/genhd.h>
87#include <linux/blkdev.h>
88#include "ctree.h"
89#include "disk-io.h"
90#include "transaction.h"
91#include "extent_io.h"
92#include "disk-io.h"
93#include "volumes.h"
94#include "print-tree.h"
95#include "locking.h"
96#include "check-integrity.h"
97
98#define BTRFSIC_BLOCK_HASHTABLE_SIZE 0x10000
99#define BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE 0x10000
100#define BTRFSIC_DEV2STATE_HASHTABLE_SIZE 0x100
101#define BTRFSIC_BLOCK_MAGIC_NUMBER 0x14491051
102#define BTRFSIC_BLOCK_LINK_MAGIC_NUMBER 0x11070807
103#define BTRFSIC_DEV2STATE_MAGIC_NUMBER 0x20111530
104#define BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER 20111300
105#define BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL (200 - 6) /* in characters,
106 * excluding " [...]" */
107#define BTRFSIC_BLOCK_SIZE PAGE_SIZE
108
109#define BTRFSIC_GENERATION_UNKNOWN ((u64)-1)
110
111/*
112 * The definition of the bitmask fields for the print_mask.
113 * They are specified with the mount option check_integrity_print_mask.
114 */
115#define BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE 0x00000001
116#define BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION 0x00000002
117#define BTRFSIC_PRINT_MASK_TREE_AFTER_SB_WRITE 0x00000004
118#define BTRFSIC_PRINT_MASK_TREE_BEFORE_SB_WRITE 0x00000008
119#define BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH 0x00000010
120#define BTRFSIC_PRINT_MASK_END_IO_BIO_BH 0x00000020
121#define BTRFSIC_PRINT_MASK_VERBOSE 0x00000040
122#define BTRFSIC_PRINT_MASK_VERY_VERBOSE 0x00000080
123#define BTRFSIC_PRINT_MASK_INITIAL_TREE 0x00000100
124#define BTRFSIC_PRINT_MASK_INITIAL_ALL_TREES 0x00000200
125#define BTRFSIC_PRINT_MASK_INITIAL_DATABASE 0x00000400
126#define BTRFSIC_PRINT_MASK_NUM_COPIES 0x00000800
127#define BTRFSIC_PRINT_MASK_TREE_WITH_ALL_MIRRORS 0x00001000
128
129struct btrfsic_dev_state;
130struct btrfsic_state;
131
132struct btrfsic_block {
133 u32 magic_num; /* only used for debug purposes */
134 unsigned int is_metadata:1; /* if it is meta-data, not data-data */
135 unsigned int is_superblock:1; /* if it is one of the superblocks */
136 unsigned int is_iodone:1; /* if is done by lower subsystem */
137 unsigned int iodone_w_error:1; /* error was indicated to endio */
138 unsigned int never_written:1; /* block was added because it was
139 * referenced, not because it was
140 * written */
141 unsigned int mirror_num:2; /* large enough to hold
142 * BTRFS_SUPER_MIRROR_MAX */
143 struct btrfsic_dev_state *dev_state;
144 u64 dev_bytenr; /* key, physical byte num on disk */
145 u64 logical_bytenr; /* logical byte num on disk */
146 u64 generation;
147 struct btrfs_disk_key disk_key; /* extra info to print in case of
148 * issues, will not always be correct */
149 struct list_head collision_resolving_node; /* list node */
150 struct list_head all_blocks_node; /* list node */
151
152 /* the following two lists contain block_link items */
153 struct list_head ref_to_list; /* list */
154 struct list_head ref_from_list; /* list */
155 struct btrfsic_block *next_in_same_bio;
156 void *orig_bio_bh_private;
157 union {
158 bio_end_io_t *bio;
159 bh_end_io_t *bh;
160 } orig_bio_bh_end_io;
161 int submit_bio_bh_rw;
162 u64 flush_gen; /* only valid if !never_written */
163};
164
165/*
166 * Elements of this type are allocated dynamically and required because
167 * each block object can refer to and can be ref from multiple blocks.
168 * The key to lookup them in the hashtable is the dev_bytenr of
169 * the block ref to plus the one from the block refered from.
170 * The fact that they are searchable via a hashtable and that a
171 * ref_cnt is maintained is not required for the btrfs integrity
172 * check algorithm itself, it is only used to make the output more
173 * beautiful in case that an error is detected (an error is defined
174 * as a write operation to a block while that block is still referenced).
175 */
176struct btrfsic_block_link {
177 u32 magic_num; /* only used for debug purposes */
178 u32 ref_cnt;
179 struct list_head node_ref_to; /* list node */
180 struct list_head node_ref_from; /* list node */
181 struct list_head collision_resolving_node; /* list node */
182 struct btrfsic_block *block_ref_to;
183 struct btrfsic_block *block_ref_from;
184 u64 parent_generation;
185};
186
187struct btrfsic_dev_state {
188 u32 magic_num; /* only used for debug purposes */
189 struct block_device *bdev;
190 struct btrfsic_state *state;
191 struct list_head collision_resolving_node; /* list node */
192 struct btrfsic_block dummy_block_for_bio_bh_flush;
193 u64 last_flush_gen;
194 char name[BDEVNAME_SIZE];
195};
196
197struct btrfsic_block_hashtable {
198 struct list_head table[BTRFSIC_BLOCK_HASHTABLE_SIZE];
199};
200
201struct btrfsic_block_link_hashtable {
202 struct list_head table[BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE];
203};
204
205struct btrfsic_dev_state_hashtable {
206 struct list_head table[BTRFSIC_DEV2STATE_HASHTABLE_SIZE];
207};
208
209struct btrfsic_block_data_ctx {
210 u64 start; /* virtual bytenr */
211 u64 dev_bytenr; /* physical bytenr on device */
212 u32 len;
213 struct btrfsic_dev_state *dev;
214 char *data;
215 struct buffer_head *bh; /* do not use if set to NULL */
216};
217
218/* This structure is used to implement recursion without occupying
219 * any stack space, refer to btrfsic_process_metablock() */
220struct btrfsic_stack_frame {
221 u32 magic;
222 u32 nr;
223 int error;
224 int i;
225 int limit_nesting;
226 int num_copies;
227 int mirror_num;
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;
234};
235
236/* Some state per mounted filesystem */
237struct btrfsic_state {
238 u32 print_mask;
239 int include_extent_data;
240 int csum_size;
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;
247};
248
249static void btrfsic_block_init(struct btrfsic_block *b);
250static struct btrfsic_block *btrfsic_block_alloc(void);
251static void btrfsic_block_free(struct btrfsic_block *b);
252static void btrfsic_block_link_init(struct btrfsic_block_link *n);
253static struct btrfsic_block_link *btrfsic_block_link_alloc(void);
254static void btrfsic_block_link_free(struct btrfsic_block_link *n);
255static void btrfsic_dev_state_init(struct btrfsic_dev_state *ds);
256static struct btrfsic_dev_state *btrfsic_dev_state_alloc(void);
257static void btrfsic_dev_state_free(struct btrfsic_dev_state *ds);
258static void btrfsic_block_hashtable_init(struct btrfsic_block_hashtable *h);
259static void btrfsic_block_hashtable_add(struct btrfsic_block *b,
260 struct btrfsic_block_hashtable *h);
261static void btrfsic_block_hashtable_remove(struct btrfsic_block *b);
262static struct btrfsic_block *btrfsic_block_hashtable_lookup(
263 struct block_device *bdev,
264 u64 dev_bytenr,
265 struct btrfsic_block_hashtable *h);
266static void btrfsic_block_link_hashtable_init(
267 struct btrfsic_block_link_hashtable *h);
268static void btrfsic_block_link_hashtable_add(
269 struct btrfsic_block_link *l,
270 struct btrfsic_block_link_hashtable *h);
271static void btrfsic_block_link_hashtable_remove(struct btrfsic_block_link *l);
272static struct btrfsic_block_link *btrfsic_block_link_hashtable_lookup(
273 struct block_device *bdev_ref_to,
274 u64 dev_bytenr_ref_to,
275 struct block_device *bdev_ref_from,
276 u64 dev_bytenr_ref_from,
277 struct btrfsic_block_link_hashtable *h);
278static void btrfsic_dev_state_hashtable_init(
279 struct btrfsic_dev_state_hashtable *h);
280static void btrfsic_dev_state_hashtable_add(
281 struct btrfsic_dev_state *ds,
282 struct btrfsic_dev_state_hashtable *h);
283static void btrfsic_dev_state_hashtable_remove(struct btrfsic_dev_state *ds);
284static struct btrfsic_dev_state *btrfsic_dev_state_hashtable_lookup(
285 struct block_device *bdev,
286 struct btrfsic_dev_state_hashtable *h);
287static struct btrfsic_stack_frame *btrfsic_stack_frame_alloc(void);
288static void btrfsic_stack_frame_free(struct btrfsic_stack_frame *sf);
289static int btrfsic_process_superblock(struct btrfsic_state *state,
290 struct btrfs_fs_devices *fs_devices);
291static int btrfsic_process_metablock(struct btrfsic_state *state,
292 struct btrfsic_block *block,
293 struct btrfsic_block_data_ctx *block_ctx,
294 struct btrfs_header *hdr,
295 int limit_nesting, int force_iodone_flag);
296static int btrfsic_create_link_to_next_block(
297 struct btrfsic_state *state,
298 struct btrfsic_block *block,
299 struct btrfsic_block_data_ctx
300 *block_ctx, u64 next_bytenr,
301 int limit_nesting,
302 struct btrfsic_block_data_ctx *next_block_ctx,
303 struct btrfsic_block **next_blockp,
304 int force_iodone_flag,
305 int *num_copiesp, int *mirror_nump,
306 struct btrfs_disk_key *disk_key,
307 u64 parent_generation);
308static int btrfsic_handle_extent_data(struct btrfsic_state *state,
309 struct btrfsic_block *block,
310 struct btrfsic_block_data_ctx *block_ctx,
311 u32 item_offset, int force_iodone_flag);
312static int btrfsic_map_block(struct btrfsic_state *state, u64 bytenr, u32 len,
313 struct btrfsic_block_data_ctx *block_ctx_out,
314 int mirror_num);
315static int btrfsic_map_superblock(struct btrfsic_state *state, u64 bytenr,
316 u32 len, struct block_device *bdev,
317 struct btrfsic_block_data_ctx *block_ctx_out);
318static void btrfsic_release_block_ctx(struct btrfsic_block_data_ctx *block_ctx);
319static int btrfsic_read_block(struct btrfsic_state *state,
320 struct btrfsic_block_data_ctx *block_ctx);
321static void btrfsic_dump_database(struct btrfsic_state *state);
322static int btrfsic_test_for_metadata(struct btrfsic_state *state,
323 const u8 *data, unsigned int size);
324static void btrfsic_process_written_block(struct btrfsic_dev_state *dev_state,
325 u64 dev_bytenr, u8 *mapped_data,
326 unsigned int len, struct bio *bio,
327 int *bio_is_patched,
328 struct buffer_head *bh,
329 int submit_bio_bh_rw);
330static int btrfsic_process_written_superblock(
331 struct btrfsic_state *state,
332 struct btrfsic_block *const block,
333 struct btrfs_super_block *const super_hdr);
334static void btrfsic_bio_end_io(struct bio *bp, int bio_error_status);
335static void btrfsic_bh_end_io(struct buffer_head *bh, int uptodate);
336static int btrfsic_is_block_ref_by_superblock(const struct btrfsic_state *state,
337 const struct btrfsic_block *block,
338 int recursion_level);
339static int btrfsic_check_all_ref_blocks(struct btrfsic_state *state,
340 struct btrfsic_block *const block,
341 int recursion_level);
342static void btrfsic_print_add_link(const struct btrfsic_state *state,
343 const struct btrfsic_block_link *l);
344static void btrfsic_print_rem_link(const struct btrfsic_state *state,
345 const struct btrfsic_block_link *l);
346static char btrfsic_get_block_type(const struct btrfsic_state *state,
347 const struct btrfsic_block *block);
348static void btrfsic_dump_tree(const struct btrfsic_state *state);
349static void btrfsic_dump_tree_sub(const struct btrfsic_state *state,
350 const struct btrfsic_block *block,
351 int indent_level);
352static struct btrfsic_block_link *btrfsic_block_link_lookup_or_add(
353 struct btrfsic_state *state,
354 struct btrfsic_block_data_ctx *next_block_ctx,
355 struct btrfsic_block *next_block,
356 struct btrfsic_block *from_block,
357 u64 parent_generation);
358static struct btrfsic_block *btrfsic_block_lookup_or_add(
359 struct btrfsic_state *state,
360 struct btrfsic_block_data_ctx *block_ctx,
361 const char *additional_string,
362 int is_metadata,
363 int is_iodone,
364 int never_written,
365 int mirror_num,
366 int *was_created);
367static int btrfsic_process_superblock_dev_mirror(
368 struct btrfsic_state *state,
369 struct btrfsic_dev_state *dev_state,
370 struct btrfs_device *device,
371 int superblock_mirror_num,
372 struct btrfsic_dev_state **selected_dev_state,
373 struct btrfs_super_block *selected_super);
374static struct btrfsic_dev_state *btrfsic_dev_state_lookup(
375 struct block_device *bdev);
376static void btrfsic_cmp_log_and_dev_bytenr(struct btrfsic_state *state,
377 u64 bytenr,
378 struct btrfsic_dev_state *dev_state,
379 u64 dev_bytenr, char *data);
380
381static struct mutex btrfsic_mutex;
382static int btrfsic_is_initialized;
383static struct btrfsic_dev_state_hashtable btrfsic_dev_state_hashtable;
384
385
386static void btrfsic_block_init(struct btrfsic_block *b)
387{
388 b->magic_num = BTRFSIC_BLOCK_MAGIC_NUMBER;
389 b->dev_state = NULL;
390 b->dev_bytenr = 0;
391 b->logical_bytenr = 0;
392 b->generation = BTRFSIC_GENERATION_UNKNOWN;
393 b->disk_key.objectid = 0;
394 b->disk_key.type = 0;
395 b->disk_key.offset = 0;
396 b->is_metadata = 0;
397 b->is_superblock = 0;
398 b->is_iodone = 0;
399 b->iodone_w_error = 0;
400 b->never_written = 0;
401 b->mirror_num = 0;
402 b->next_in_same_bio = NULL;
403 b->orig_bio_bh_private = NULL;
404 b->orig_bio_bh_end_io.bio = NULL;
405 INIT_LIST_HEAD(&b->collision_resolving_node);
406 INIT_LIST_HEAD(&b->all_blocks_node);
407 INIT_LIST_HEAD(&b->ref_to_list);
408 INIT_LIST_HEAD(&b->ref_from_list);
409 b->submit_bio_bh_rw = 0;
410 b->flush_gen = 0;
411}
412
413static struct btrfsic_block *btrfsic_block_alloc(void)
414{
415 struct btrfsic_block *b;
416
417 b = kzalloc(sizeof(*b), GFP_NOFS);
418 if (NULL != b)
419 btrfsic_block_init(b);
420
421 return b;
422}
423
424static void btrfsic_block_free(struct btrfsic_block *b)
425{
426 BUG_ON(!(NULL == b || BTRFSIC_BLOCK_MAGIC_NUMBER == b->magic_num));
427 kfree(b);
428}
429
430static void btrfsic_block_link_init(struct btrfsic_block_link *l)
431{
432 l->magic_num = BTRFSIC_BLOCK_LINK_MAGIC_NUMBER;
433 l->ref_cnt = 1;
434 INIT_LIST_HEAD(&l->node_ref_to);
435 INIT_LIST_HEAD(&l->node_ref_from);
436 INIT_LIST_HEAD(&l->collision_resolving_node);
437 l->block_ref_to = NULL;
438 l->block_ref_from = NULL;
439}
440
441static struct btrfsic_block_link *btrfsic_block_link_alloc(void)
442{
443 struct btrfsic_block_link *l;
444
445 l = kzalloc(sizeof(*l), GFP_NOFS);
446 if (NULL != l)
447 btrfsic_block_link_init(l);
448
449 return l;
450}
451
452static void btrfsic_block_link_free(struct btrfsic_block_link *l)
453{
454 BUG_ON(!(NULL == l || BTRFSIC_BLOCK_LINK_MAGIC_NUMBER == l->magic_num));
455 kfree(l);
456}
457
458static void btrfsic_dev_state_init(struct btrfsic_dev_state *ds)
459{
460 ds->magic_num = BTRFSIC_DEV2STATE_MAGIC_NUMBER;
461 ds->bdev = NULL;
462 ds->state = NULL;
463 ds->name[0] = '\0';
464 INIT_LIST_HEAD(&ds->collision_resolving_node);
465 ds->last_flush_gen = 0;
466 btrfsic_block_init(&ds->dummy_block_for_bio_bh_flush);
467 ds->dummy_block_for_bio_bh_flush.is_iodone = 1;
468 ds->dummy_block_for_bio_bh_flush.dev_state = ds;
469}
470
471static struct btrfsic_dev_state *btrfsic_dev_state_alloc(void)
472{
473 struct btrfsic_dev_state *ds;
474
475 ds = kzalloc(sizeof(*ds), GFP_NOFS);
476 if (NULL != ds)
477 btrfsic_dev_state_init(ds);
478
479 return ds;
480}
481
482static void btrfsic_dev_state_free(struct btrfsic_dev_state *ds)
483{
484 BUG_ON(!(NULL == ds ||
485 BTRFSIC_DEV2STATE_MAGIC_NUMBER == ds->magic_num));
486 kfree(ds);
487}
488
489static void btrfsic_block_hashtable_init(struct btrfsic_block_hashtable *h)
490{
491 int i;
492
493 for (i = 0; i < BTRFSIC_BLOCK_HASHTABLE_SIZE; i++)
494 INIT_LIST_HEAD(h->table + i);
495}
496
497static void btrfsic_block_hashtable_add(struct btrfsic_block *b,
498 struct btrfsic_block_hashtable *h)
499{
500 const unsigned int hashval =
501 (((unsigned int)(b->dev_bytenr >> 16)) ^
502 ((unsigned int)((uintptr_t)b->dev_state->bdev))) &
503 (BTRFSIC_BLOCK_HASHTABLE_SIZE - 1);
504
505 list_add(&b->collision_resolving_node, h->table + hashval);
506}
507
508static void btrfsic_block_hashtable_remove(struct btrfsic_block *b)
509{
510 list_del(&b->collision_resolving_node);
511}
512
513static struct btrfsic_block *btrfsic_block_hashtable_lookup(
514 struct block_device *bdev,
515 u64 dev_bytenr,
516 struct btrfsic_block_hashtable *h)
517{
518 const unsigned int hashval =
519 (((unsigned int)(dev_bytenr >> 16)) ^
520 ((unsigned int)((uintptr_t)bdev))) &
521 (BTRFSIC_BLOCK_HASHTABLE_SIZE - 1);
522 struct list_head *elem;
523
524 list_for_each(elem, h->table + hashval) {
525 struct btrfsic_block *const b =
526 list_entry(elem, struct btrfsic_block,
527 collision_resolving_node);
528
529 if (b->dev_state->bdev == bdev && b->dev_bytenr == dev_bytenr)
530 return b;
531 }
532
533 return NULL;
534}
535
536static void btrfsic_block_link_hashtable_init(
537 struct btrfsic_block_link_hashtable *h)
538{
539 int i;
540
541 for (i = 0; i < BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE; i++)
542 INIT_LIST_HEAD(h->table + i);
543}
544
545static void btrfsic_block_link_hashtable_add(
546 struct btrfsic_block_link *l,
547 struct btrfsic_block_link_hashtable *h)
548{
549 const unsigned int hashval =
550 (((unsigned int)(l->block_ref_to->dev_bytenr >> 16)) ^
551 ((unsigned int)(l->block_ref_from->dev_bytenr >> 16)) ^
552 ((unsigned int)((uintptr_t)l->block_ref_to->dev_state->bdev)) ^
553 ((unsigned int)((uintptr_t)l->block_ref_from->dev_state->bdev)))
554 & (BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE - 1);
555
556 BUG_ON(NULL == l->block_ref_to);
557 BUG_ON(NULL == l->block_ref_from);
558 list_add(&l->collision_resolving_node, h->table + hashval);
559}
560
561static void btrfsic_block_link_hashtable_remove(struct btrfsic_block_link *l)
562{
563 list_del(&l->collision_resolving_node);
564}
565
566static struct btrfsic_block_link *btrfsic_block_link_hashtable_lookup(
567 struct block_device *bdev_ref_to,
568 u64 dev_bytenr_ref_to,
569 struct block_device *bdev_ref_from,
570 u64 dev_bytenr_ref_from,
571 struct btrfsic_block_link_hashtable *h)
572{
573 const unsigned int hashval =
574 (((unsigned int)(dev_bytenr_ref_to >> 16)) ^
575 ((unsigned int)(dev_bytenr_ref_from >> 16)) ^
576 ((unsigned int)((uintptr_t)bdev_ref_to)) ^
577 ((unsigned int)((uintptr_t)bdev_ref_from))) &
578 (BTRFSIC_BLOCK_LINK_HASHTABLE_SIZE - 1);
579 struct list_head *elem;
580
581 list_for_each(elem, h->table + hashval) {
582 struct btrfsic_block_link *const l =
583 list_entry(elem, struct btrfsic_block_link,
584 collision_resolving_node);
585
586 BUG_ON(NULL == l->block_ref_to);
587 BUG_ON(NULL == l->block_ref_from);
588 if (l->block_ref_to->dev_state->bdev == bdev_ref_to &&
589 l->block_ref_to->dev_bytenr == dev_bytenr_ref_to &&
590 l->block_ref_from->dev_state->bdev == bdev_ref_from &&
591 l->block_ref_from->dev_bytenr == dev_bytenr_ref_from)
592 return l;
593 }
594
595 return NULL;
596}
597
598static void btrfsic_dev_state_hashtable_init(
599 struct btrfsic_dev_state_hashtable *h)
600{
601 int i;
602
603 for (i = 0; i < BTRFSIC_DEV2STATE_HASHTABLE_SIZE; i++)
604 INIT_LIST_HEAD(h->table + i);
605}
606
607static void btrfsic_dev_state_hashtable_add(
608 struct btrfsic_dev_state *ds,
609 struct btrfsic_dev_state_hashtable *h)
610{
611 const unsigned int hashval =
612 (((unsigned int)((uintptr_t)ds->bdev)) &
613 (BTRFSIC_DEV2STATE_HASHTABLE_SIZE - 1));
614
615 list_add(&ds->collision_resolving_node, h->table + hashval);
616}
617
618static void btrfsic_dev_state_hashtable_remove(struct btrfsic_dev_state *ds)
619{
620 list_del(&ds->collision_resolving_node);
621}
622
623static struct btrfsic_dev_state *btrfsic_dev_state_hashtable_lookup(
624 struct block_device *bdev,
625 struct btrfsic_dev_state_hashtable *h)
626{
627 const unsigned int hashval =
628 (((unsigned int)((uintptr_t)bdev)) &
629 (BTRFSIC_DEV2STATE_HASHTABLE_SIZE - 1));
630 struct list_head *elem;
631
632 list_for_each(elem, h->table + hashval) {
633 struct btrfsic_dev_state *const ds =
634 list_entry(elem, struct btrfsic_dev_state,
635 collision_resolving_node);
636
637 if (ds->bdev == bdev)
638 return ds;
639 }
640
641 return NULL;
642}
643
644static int btrfsic_process_superblock(struct btrfsic_state *state,
645 struct btrfs_fs_devices *fs_devices)
646{
647 int ret;
648 struct btrfs_super_block *selected_super;
649 struct list_head *dev_head = &fs_devices->devices;
650 struct btrfs_device *device;
651 struct btrfsic_dev_state *selected_dev_state = NULL;
652 int pass;
653
654 BUG_ON(NULL == state);
655 selected_super = kmalloc(sizeof(*selected_super), GFP_NOFS);
656 if (NULL == selected_super) {
657 printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
658 return -1;
659 }
660
661 list_for_each_entry(device, dev_head, dev_list) {
662 int i;
663 struct btrfsic_dev_state *dev_state;
664
665 if (!device->bdev || !device->name)
666 continue;
667
668 dev_state = btrfsic_dev_state_lookup(device->bdev);
669 BUG_ON(NULL == dev_state);
670 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
671 ret = btrfsic_process_superblock_dev_mirror(
672 state, dev_state, device, i,
673 &selected_dev_state, selected_super);
674 if (0 != ret && 0 == i) {
675 kfree(selected_super);
676 return ret;
677 }
678 }
679 }
680
681 if (NULL == state->latest_superblock) {
682 printk(KERN_INFO "btrfsic: no superblock found!\n");
683 kfree(selected_super);
684 return -1;
685 }
686
687 state->csum_size = btrfs_super_csum_size(selected_super);
688
689 for (pass = 0; pass < 3; pass++) {
690 int num_copies;
691 int mirror_num;
692 u64 next_bytenr;
693
694 switch (pass) {
695 case 0:
696 next_bytenr = btrfs_super_root(selected_super);
697 if (state->print_mask &
698 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
699 printk(KERN_INFO "root@%llu\n",
700 (unsigned long long)next_bytenr);
701 break;
702 case 1:
703 next_bytenr = btrfs_super_chunk_root(selected_super);
704 if (state->print_mask &
705 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
706 printk(KERN_INFO "chunk@%llu\n",
707 (unsigned long long)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 printk(KERN_INFO "log@%llu\n",
716 (unsigned long long)next_bytenr);
717 break;
718 }
719
720 num_copies =
721 btrfs_num_copies(&state->root->fs_info->mapping_tree,
722 next_bytenr, PAGE_SIZE);
723 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
724 printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
725 (unsigned long long)next_bytenr, num_copies);
726
727 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
728 struct btrfsic_block *next_block;
729 struct btrfsic_block_data_ctx tmp_next_block_ctx;
730 struct btrfsic_block_link *l;
731 struct btrfs_header *hdr;
732
733 ret = btrfsic_map_block(state, next_bytenr, PAGE_SIZE,
734 &tmp_next_block_ctx,
735 mirror_num);
736 if (ret) {
737 printk(KERN_INFO "btrfsic:"
738 " btrfsic_map_block(root @%llu,"
739 " mirror %d) failed!\n",
740 (unsigned long long)next_bytenr,
741 mirror_num);
742 kfree(selected_super);
743 return -1;
744 }
745
746 next_block = btrfsic_block_hashtable_lookup(
747 tmp_next_block_ctx.dev->bdev,
748 tmp_next_block_ctx.dev_bytenr,
749 &state->block_hashtable);
750 BUG_ON(NULL == next_block);
751
752 l = btrfsic_block_link_hashtable_lookup(
753 tmp_next_block_ctx.dev->bdev,
754 tmp_next_block_ctx.dev_bytenr,
755 state->latest_superblock->dev_state->
756 bdev,
757 state->latest_superblock->dev_bytenr,
758 &state->block_link_hashtable);
759 BUG_ON(NULL == l);
760
761 ret = btrfsic_read_block(state, &tmp_next_block_ctx);
762 if (ret < (int)BTRFSIC_BLOCK_SIZE) {
763 printk(KERN_INFO
764 "btrfsic: read @logical %llu failed!\n",
765 (unsigned long long)
766 tmp_next_block_ctx.start);
767 btrfsic_release_block_ctx(&tmp_next_block_ctx);
768 kfree(selected_super);
769 return -1;
770 }
771
772 hdr = (struct btrfs_header *)tmp_next_block_ctx.data;
773 ret = btrfsic_process_metablock(state,
774 next_block,
775 &tmp_next_block_ctx,
776 hdr,
777 BTRFS_MAX_LEVEL + 3, 1);
778 btrfsic_release_block_ctx(&tmp_next_block_ctx);
779 }
780 }
781
782 kfree(selected_super);
783 return ret;
784}
785
786static int btrfsic_process_superblock_dev_mirror(
787 struct btrfsic_state *state,
788 struct btrfsic_dev_state *dev_state,
789 struct btrfs_device *device,
790 int superblock_mirror_num,
791 struct btrfsic_dev_state **selected_dev_state,
792 struct btrfs_super_block *selected_super)
793{
794 struct btrfs_super_block *super_tmp;
795 u64 dev_bytenr;
796 struct buffer_head *bh;
797 struct btrfsic_block *superblock_tmp;
798 int pass;
799 struct block_device *const superblock_bdev = device->bdev;
800
801 /* super block bytenr is always the unmapped device bytenr */
802 dev_bytenr = btrfs_sb_offset(superblock_mirror_num);
803 bh = __bread(superblock_bdev, dev_bytenr / 4096, 4096);
804 if (NULL == bh)
805 return -1;
806 super_tmp = (struct btrfs_super_block *)
807 (bh->b_data + (dev_bytenr & 4095));
808
809 if (btrfs_super_bytenr(super_tmp) != dev_bytenr ||
810 strncmp((char *)(&(super_tmp->magic)), BTRFS_MAGIC,
811 sizeof(super_tmp->magic)) ||
812 memcmp(device->uuid, super_tmp->dev_item.uuid, BTRFS_UUID_SIZE)) {
813 brelse(bh);
814 return 0;
815 }
816
817 superblock_tmp =
818 btrfsic_block_hashtable_lookup(superblock_bdev,
819 dev_bytenr,
820 &state->block_hashtable);
821 if (NULL == superblock_tmp) {
822 superblock_tmp = btrfsic_block_alloc();
823 if (NULL == superblock_tmp) {
824 printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
825 brelse(bh);
826 return -1;
827 }
828 /* for superblock, only the dev_bytenr makes sense */
829 superblock_tmp->dev_bytenr = dev_bytenr;
830 superblock_tmp->dev_state = dev_state;
831 superblock_tmp->logical_bytenr = dev_bytenr;
832 superblock_tmp->generation = btrfs_super_generation(super_tmp);
833 superblock_tmp->is_metadata = 1;
834 superblock_tmp->is_superblock = 1;
835 superblock_tmp->is_iodone = 1;
836 superblock_tmp->never_written = 0;
837 superblock_tmp->mirror_num = 1 + superblock_mirror_num;
838 if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
839 printk(KERN_INFO "New initial S-block (bdev %p, %s)"
840 " @%llu (%s/%llu/%d)\n",
841 superblock_bdev, device->name,
842 (unsigned long long)dev_bytenr,
843 dev_state->name,
844 (unsigned long long)dev_bytenr,
845 superblock_mirror_num);
846 list_add(&superblock_tmp->all_blocks_node,
847 &state->all_blocks_list);
848 btrfsic_block_hashtable_add(superblock_tmp,
849 &state->block_hashtable);
850 }
851
852 /* select the one with the highest generation field */
853 if (btrfs_super_generation(super_tmp) >
854 state->max_superblock_generation ||
855 0 == state->max_superblock_generation) {
856 memcpy(selected_super, super_tmp, sizeof(*selected_super));
857 *selected_dev_state = dev_state;
858 state->max_superblock_generation =
859 btrfs_super_generation(super_tmp);
860 state->latest_superblock = superblock_tmp;
861 }
862
863 for (pass = 0; pass < 3; pass++) {
864 u64 next_bytenr;
865 int num_copies;
866 int mirror_num;
867 const char *additional_string = NULL;
868 struct btrfs_disk_key tmp_disk_key;
869
870 tmp_disk_key.type = BTRFS_ROOT_ITEM_KEY;
871 tmp_disk_key.offset = 0;
872 switch (pass) {
873 case 0:
874 tmp_disk_key.objectid =
875 cpu_to_le64(BTRFS_ROOT_TREE_OBJECTID);
876 additional_string = "initial root ";
877 next_bytenr = btrfs_super_root(super_tmp);
878 break;
879 case 1:
880 tmp_disk_key.objectid =
881 cpu_to_le64(BTRFS_CHUNK_TREE_OBJECTID);
882 additional_string = "initial chunk ";
883 next_bytenr = btrfs_super_chunk_root(super_tmp);
884 break;
885 case 2:
886 tmp_disk_key.objectid =
887 cpu_to_le64(BTRFS_TREE_LOG_OBJECTID);
888 additional_string = "initial log ";
889 next_bytenr = btrfs_super_log_root(super_tmp);
890 if (0 == next_bytenr)
891 continue;
892 break;
893 }
894
895 num_copies =
896 btrfs_num_copies(&state->root->fs_info->mapping_tree,
897 next_bytenr, PAGE_SIZE);
898 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
899 printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
900 (unsigned long long)next_bytenr, num_copies);
901 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
902 struct btrfsic_block *next_block;
903 struct btrfsic_block_data_ctx tmp_next_block_ctx;
904 struct btrfsic_block_link *l;
905
906 if (btrfsic_map_block(state, next_bytenr, PAGE_SIZE,
907 &tmp_next_block_ctx,
908 mirror_num)) {
909 printk(KERN_INFO "btrfsic: btrfsic_map_block("
910 "bytenr @%llu, mirror %d) failed!\n",
911 (unsigned long long)next_bytenr,
912 mirror_num);
913 brelse(bh);
914 return -1;
915 }
916
917 next_block = btrfsic_block_lookup_or_add(
918 state, &tmp_next_block_ctx,
919 additional_string, 1, 1, 0,
920 mirror_num, NULL);
921 if (NULL == next_block) {
922 btrfsic_release_block_ctx(&tmp_next_block_ctx);
923 brelse(bh);
924 return -1;
925 }
926
927 next_block->disk_key = tmp_disk_key;
928 next_block->generation = BTRFSIC_GENERATION_UNKNOWN;
929 l = btrfsic_block_link_lookup_or_add(
930 state, &tmp_next_block_ctx,
931 next_block, superblock_tmp,
932 BTRFSIC_GENERATION_UNKNOWN);
933 btrfsic_release_block_ctx(&tmp_next_block_ctx);
934 if (NULL == l) {
935 brelse(bh);
936 return -1;
937 }
938 }
939 }
940 if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_ALL_TREES)
941 btrfsic_dump_tree_sub(state, superblock_tmp, 0);
942
943 brelse(bh);
944 return 0;
945}
946
947static struct btrfsic_stack_frame *btrfsic_stack_frame_alloc(void)
948{
949 struct btrfsic_stack_frame *sf;
950
951 sf = kzalloc(sizeof(*sf), GFP_NOFS);
952 if (NULL == sf)
953 printk(KERN_INFO "btrfsic: alloc memory failed!\n");
954 else
955 sf->magic = BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER;
956 return sf;
957}
958
959static void btrfsic_stack_frame_free(struct btrfsic_stack_frame *sf)
960{
961 BUG_ON(!(NULL == sf ||
962 BTRFSIC_BLOCK_STACK_FRAME_MAGIC_NUMBER == sf->magic));
963 kfree(sf);
964}
965
966static int btrfsic_process_metablock(
967 struct btrfsic_state *state,
968 struct btrfsic_block *const first_block,
969 struct btrfsic_block_data_ctx *const first_block_ctx,
970 struct btrfs_header *const first_hdr,
971 int first_limit_nesting, int force_iodone_flag)
972{
973 struct btrfsic_stack_frame initial_stack_frame = { 0 };
974 struct btrfsic_stack_frame *sf;
975 struct btrfsic_stack_frame *next_stack;
976
977 sf = &initial_stack_frame;
978 sf->error = 0;
979 sf->i = -1;
980 sf->limit_nesting = first_limit_nesting;
981 sf->block = first_block;
982 sf->block_ctx = first_block_ctx;
983 sf->next_block = NULL;
984 sf->hdr = first_hdr;
985 sf->prev = NULL;
986
987continue_with_new_stack_frame:
988 sf->block->generation = le64_to_cpu(sf->hdr->generation);
989 if (0 == sf->hdr->level) {
990 struct btrfs_leaf *const leafhdr =
991 (struct btrfs_leaf *)sf->hdr;
992
993 if (-1 == sf->i) {
994 sf->nr = le32_to_cpu(leafhdr->header.nritems);
995
996 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
997 printk(KERN_INFO
998 "leaf %llu items %d generation %llu"
999 " owner %llu\n",
1000 (unsigned long long)
1001 sf->block_ctx->start,
1002 sf->nr,
1003 (unsigned long long)
1004 le64_to_cpu(leafhdr->header.generation),
1005 (unsigned long long)
1006 le64_to_cpu(leafhdr->header.owner));
1007 }
1008
1009continue_with_current_leaf_stack_frame:
1010 if (0 == sf->num_copies || sf->mirror_num > sf->num_copies) {
1011 sf->i++;
1012 sf->num_copies = 0;
1013 }
1014
1015 if (sf->i < sf->nr) {
1016 struct btrfs_item *disk_item = leafhdr->items + sf->i;
1017 struct btrfs_disk_key *disk_key = &disk_item->key;
1018 u8 type;
1019 const u32 item_offset = le32_to_cpu(disk_item->offset);
1020
1021 type = disk_key->type;
1022
1023 if (BTRFS_ROOT_ITEM_KEY == type) {
1024 const struct btrfs_root_item *const root_item =
1025 (struct btrfs_root_item *)
1026 (sf->block_ctx->data +
1027 offsetof(struct btrfs_leaf, items) +
1028 item_offset);
1029 const u64 next_bytenr =
1030 le64_to_cpu(root_item->bytenr);
1031
1032 sf->error =
1033 btrfsic_create_link_to_next_block(
1034 state,
1035 sf->block,
1036 sf->block_ctx,
1037 next_bytenr,
1038 sf->limit_nesting,
1039 &sf->next_block_ctx,
1040 &sf->next_block,
1041 force_iodone_flag,
1042 &sf->num_copies,
1043 &sf->mirror_num,
1044 disk_key,
1045 le64_to_cpu(root_item->
1046 generation));
1047 if (sf->error)
1048 goto one_stack_frame_backwards;
1049
1050 if (NULL != sf->next_block) {
1051 struct btrfs_header *const next_hdr =
1052 (struct btrfs_header *)
1053 sf->next_block_ctx.data;
1054
1055 next_stack =
1056 btrfsic_stack_frame_alloc();
1057 if (NULL == next_stack) {
1058 btrfsic_release_block_ctx(
1059 &sf->
1060 next_block_ctx);
1061 goto one_stack_frame_backwards;
1062 }
1063
1064 next_stack->i = -1;
1065 next_stack->block = sf->next_block;
1066 next_stack->block_ctx =
1067 &sf->next_block_ctx;
1068 next_stack->next_block = NULL;
1069 next_stack->hdr = next_hdr;
1070 next_stack->limit_nesting =
1071 sf->limit_nesting - 1;
1072 next_stack->prev = sf;
1073 sf = next_stack;
1074 goto continue_with_new_stack_frame;
1075 }
1076 } else if (BTRFS_EXTENT_DATA_KEY == type &&
1077 state->include_extent_data) {
1078 sf->error = btrfsic_handle_extent_data(
1079 state,
1080 sf->block,
1081 sf->block_ctx,
1082 item_offset,
1083 force_iodone_flag);
1084 if (sf->error)
1085 goto one_stack_frame_backwards;
1086 }
1087
1088 goto continue_with_current_leaf_stack_frame;
1089 }
1090 } else {
1091 struct btrfs_node *const nodehdr = (struct btrfs_node *)sf->hdr;
1092
1093 if (-1 == sf->i) {
1094 sf->nr = le32_to_cpu(nodehdr->header.nritems);
1095
1096 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1097 printk(KERN_INFO "node %llu level %d items %d"
1098 " generation %llu owner %llu\n",
1099 (unsigned long long)
1100 sf->block_ctx->start,
1101 nodehdr->header.level, sf->nr,
1102 (unsigned long long)
1103 le64_to_cpu(nodehdr->header.generation),
1104 (unsigned long long)
1105 le64_to_cpu(nodehdr->header.owner));
1106 }
1107
1108continue_with_current_node_stack_frame:
1109 if (0 == sf->num_copies || sf->mirror_num > sf->num_copies) {
1110 sf->i++;
1111 sf->num_copies = 0;
1112 }
1113
1114 if (sf->i < sf->nr) {
1115 struct btrfs_key_ptr *disk_key_ptr =
1116 nodehdr->ptrs + sf->i;
1117 const u64 next_bytenr =
1118 le64_to_cpu(disk_key_ptr->blockptr);
1119
1120 sf->error = btrfsic_create_link_to_next_block(
1121 state,
1122 sf->block,
1123 sf->block_ctx,
1124 next_bytenr,
1125 sf->limit_nesting,
1126 &sf->next_block_ctx,
1127 &sf->next_block,
1128 force_iodone_flag,
1129 &sf->num_copies,
1130 &sf->mirror_num,
1131 &disk_key_ptr->key,
1132 le64_to_cpu(disk_key_ptr->generation));
1133 if (sf->error)
1134 goto one_stack_frame_backwards;
1135
1136 if (NULL != sf->next_block) {
1137 struct btrfs_header *const next_hdr =
1138 (struct btrfs_header *)
1139 sf->next_block_ctx.data;
1140
1141 next_stack = btrfsic_stack_frame_alloc();
1142 if (NULL == next_stack)
1143 goto one_stack_frame_backwards;
1144
1145 next_stack->i = -1;
1146 next_stack->block = sf->next_block;
1147 next_stack->block_ctx = &sf->next_block_ctx;
1148 next_stack->next_block = NULL;
1149 next_stack->hdr = next_hdr;
1150 next_stack->limit_nesting =
1151 sf->limit_nesting - 1;
1152 next_stack->prev = sf;
1153 sf = next_stack;
1154 goto continue_with_new_stack_frame;
1155 }
1156
1157 goto continue_with_current_node_stack_frame;
1158 }
1159 }
1160
1161one_stack_frame_backwards:
1162 if (NULL != sf->prev) {
1163 struct btrfsic_stack_frame *const prev = sf->prev;
1164
1165 /* the one for the initial block is freed in the caller */
1166 btrfsic_release_block_ctx(sf->block_ctx);
1167
1168 if (sf->error) {
1169 prev->error = sf->error;
1170 btrfsic_stack_frame_free(sf);
1171 sf = prev;
1172 goto one_stack_frame_backwards;
1173 }
1174
1175 btrfsic_stack_frame_free(sf);
1176 sf = prev;
1177 goto continue_with_new_stack_frame;
1178 } else {
1179 BUG_ON(&initial_stack_frame != sf);
1180 }
1181
1182 return sf->error;
1183}
1184
1185static int btrfsic_create_link_to_next_block(
1186 struct btrfsic_state *state,
1187 struct btrfsic_block *block,
1188 struct btrfsic_block_data_ctx *block_ctx,
1189 u64 next_bytenr,
1190 int limit_nesting,
1191 struct btrfsic_block_data_ctx *next_block_ctx,
1192 struct btrfsic_block **next_blockp,
1193 int force_iodone_flag,
1194 int *num_copiesp, int *mirror_nump,
1195 struct btrfs_disk_key *disk_key,
1196 u64 parent_generation)
1197{
1198 struct btrfsic_block *next_block = NULL;
1199 int ret;
1200 struct btrfsic_block_link *l;
1201 int did_alloc_block_link;
1202 int block_was_created;
1203
1204 *next_blockp = NULL;
1205 if (0 == *num_copiesp) {
1206 *num_copiesp =
1207 btrfs_num_copies(&state->root->fs_info->mapping_tree,
1208 next_bytenr, PAGE_SIZE);
1209 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
1210 printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
1211 (unsigned long long)next_bytenr, *num_copiesp);
1212 *mirror_nump = 1;
1213 }
1214
1215 if (*mirror_nump > *num_copiesp)
1216 return 0;
1217
1218 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1219 printk(KERN_INFO
1220 "btrfsic_create_link_to_next_block(mirror_num=%d)\n",
1221 *mirror_nump);
1222 ret = btrfsic_map_block(state, next_bytenr,
1223 BTRFSIC_BLOCK_SIZE,
1224 next_block_ctx, *mirror_nump);
1225 if (ret) {
1226 printk(KERN_INFO
1227 "btrfsic: btrfsic_map_block(@%llu, mirror=%d) failed!\n",
1228 (unsigned long long)next_bytenr, *mirror_nump);
1229 btrfsic_release_block_ctx(next_block_ctx);
1230 *next_blockp = NULL;
1231 return -1;
1232 }
1233
1234 next_block = btrfsic_block_lookup_or_add(state,
1235 next_block_ctx, "referenced ",
1236 1, force_iodone_flag,
1237 !force_iodone_flag,
1238 *mirror_nump,
1239 &block_was_created);
1240 if (NULL == next_block) {
1241 btrfsic_release_block_ctx(next_block_ctx);
1242 *next_blockp = NULL;
1243 return -1;
1244 }
1245 if (block_was_created) {
1246 l = NULL;
1247 next_block->generation = BTRFSIC_GENERATION_UNKNOWN;
1248 } else {
1249 if (next_block->logical_bytenr != next_bytenr &&
1250 !(!next_block->is_metadata &&
1251 0 == next_block->logical_bytenr)) {
1252 printk(KERN_INFO
1253 "Referenced block @%llu (%s/%llu/%d)"
1254 " found in hash table, %c,"
1255 " bytenr mismatch (!= stored %llu).\n",
1256 (unsigned long long)next_bytenr,
1257 next_block_ctx->dev->name,
1258 (unsigned long long)next_block_ctx->dev_bytenr,
1259 *mirror_nump,
1260 btrfsic_get_block_type(state, next_block),
1261 (unsigned long long)next_block->logical_bytenr);
1262 } else if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1263 printk(KERN_INFO
1264 "Referenced block @%llu (%s/%llu/%d)"
1265 " found in hash table, %c.\n",
1266 (unsigned long long)next_bytenr,
1267 next_block_ctx->dev->name,
1268 (unsigned long long)next_block_ctx->dev_bytenr,
1269 *mirror_nump,
1270 btrfsic_get_block_type(state, next_block));
1271 next_block->logical_bytenr = next_bytenr;
1272
1273 next_block->mirror_num = *mirror_nump;
1274 l = btrfsic_block_link_hashtable_lookup(
1275 next_block_ctx->dev->bdev,
1276 next_block_ctx->dev_bytenr,
1277 block_ctx->dev->bdev,
1278 block_ctx->dev_bytenr,
1279 &state->block_link_hashtable);
1280 }
1281
1282 next_block->disk_key = *disk_key;
1283 if (NULL == l) {
1284 l = btrfsic_block_link_alloc();
1285 if (NULL == l) {
1286 printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
1287 btrfsic_release_block_ctx(next_block_ctx);
1288 *next_blockp = NULL;
1289 return -1;
1290 }
1291
1292 did_alloc_block_link = 1;
1293 l->block_ref_to = next_block;
1294 l->block_ref_from = block;
1295 l->ref_cnt = 1;
1296 l->parent_generation = parent_generation;
1297
1298 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1299 btrfsic_print_add_link(state, l);
1300
1301 list_add(&l->node_ref_to, &block->ref_to_list);
1302 list_add(&l->node_ref_from, &next_block->ref_from_list);
1303
1304 btrfsic_block_link_hashtable_add(l,
1305 &state->block_link_hashtable);
1306 } else {
1307 did_alloc_block_link = 0;
1308 if (0 == limit_nesting) {
1309 l->ref_cnt++;
1310 l->parent_generation = parent_generation;
1311 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1312 btrfsic_print_add_link(state, l);
1313 }
1314 }
1315
1316 if (limit_nesting > 0 && did_alloc_block_link) {
1317 ret = btrfsic_read_block(state, next_block_ctx);
1318 if (ret < (int)BTRFSIC_BLOCK_SIZE) {
1319 printk(KERN_INFO
1320 "btrfsic: read block @logical %llu failed!\n",
1321 (unsigned long long)next_bytenr);
1322 btrfsic_release_block_ctx(next_block_ctx);
1323 *next_blockp = NULL;
1324 return -1;
1325 }
1326
1327 *next_blockp = next_block;
1328 } else {
1329 *next_blockp = NULL;
1330 }
1331 (*mirror_nump)++;
1332
1333 return 0;
1334}
1335
1336static int btrfsic_handle_extent_data(
1337 struct btrfsic_state *state,
1338 struct btrfsic_block *block,
1339 struct btrfsic_block_data_ctx *block_ctx,
1340 u32 item_offset, int force_iodone_flag)
1341{
1342 int ret;
1343 struct btrfs_file_extent_item *file_extent_item =
1344 (struct btrfs_file_extent_item *)(block_ctx->data +
1345 offsetof(struct btrfs_leaf,
1346 items) + item_offset);
1347 u64 next_bytenr =
1348 le64_to_cpu(file_extent_item->disk_bytenr) +
1349 le64_to_cpu(file_extent_item->offset);
1350 u64 num_bytes = le64_to_cpu(file_extent_item->num_bytes);
1351 u64 generation = le64_to_cpu(file_extent_item->generation);
1352 struct btrfsic_block_link *l;
1353
1354 if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE)
1355 printk(KERN_INFO "extent_data: type %u, disk_bytenr = %llu,"
1356 " offset = %llu, num_bytes = %llu\n",
1357 file_extent_item->type,
1358 (unsigned long long)
1359 le64_to_cpu(file_extent_item->disk_bytenr),
1360 (unsigned long long)
1361 le64_to_cpu(file_extent_item->offset),
1362 (unsigned long long)
1363 le64_to_cpu(file_extent_item->num_bytes));
1364 if (BTRFS_FILE_EXTENT_REG != file_extent_item->type ||
1365 ((u64)0) == le64_to_cpu(file_extent_item->disk_bytenr))
1366 return 0;
1367 while (num_bytes > 0) {
1368 u32 chunk_len;
1369 int num_copies;
1370 int mirror_num;
1371
1372 if (num_bytes > BTRFSIC_BLOCK_SIZE)
1373 chunk_len = BTRFSIC_BLOCK_SIZE;
1374 else
1375 chunk_len = num_bytes;
1376
1377 num_copies =
1378 btrfs_num_copies(&state->root->fs_info->mapping_tree,
1379 next_bytenr, PAGE_SIZE);
1380 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
1381 printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
1382 (unsigned long long)next_bytenr, num_copies);
1383 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
1384 struct btrfsic_block_data_ctx next_block_ctx;
1385 struct btrfsic_block *next_block;
1386 int block_was_created;
1387
1388 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1389 printk(KERN_INFO "btrfsic_handle_extent_data("
1390 "mirror_num=%d)\n", mirror_num);
1391 if (state->print_mask & BTRFSIC_PRINT_MASK_VERY_VERBOSE)
1392 printk(KERN_INFO
1393 "\tdisk_bytenr = %llu, num_bytes %u\n",
1394 (unsigned long long)next_bytenr,
1395 chunk_len);
1396 ret = btrfsic_map_block(state, next_bytenr,
1397 chunk_len, &next_block_ctx,
1398 mirror_num);
1399 if (ret) {
1400 printk(KERN_INFO
1401 "btrfsic: btrfsic_map_block(@%llu,"
1402 " mirror=%d) failed!\n",
1403 (unsigned long long)next_bytenr,
1404 mirror_num);
1405 return -1;
1406 }
1407
1408 next_block = btrfsic_block_lookup_or_add(
1409 state,
1410 &next_block_ctx,
1411 "referenced ",
1412 0,
1413 force_iodone_flag,
1414 !force_iodone_flag,
1415 mirror_num,
1416 &block_was_created);
1417 if (NULL == next_block) {
1418 printk(KERN_INFO
1419 "btrfsic: error, kmalloc failed!\n");
1420 btrfsic_release_block_ctx(&next_block_ctx);
1421 return -1;
1422 }
1423 if (!block_was_created) {
1424 if (next_block->logical_bytenr != next_bytenr &&
1425 !(!next_block->is_metadata &&
1426 0 == next_block->logical_bytenr)) {
1427 printk(KERN_INFO
1428 "Referenced block"
1429 " @%llu (%s/%llu/%d)"
1430 " found in hash table, D,"
1431 " bytenr mismatch"
1432 " (!= stored %llu).\n",
1433 (unsigned long long)next_bytenr,
1434 next_block_ctx.dev->name,
1435 (unsigned long long)
1436 next_block_ctx.dev_bytenr,
1437 mirror_num,
1438 (unsigned long long)
1439 next_block->logical_bytenr);
1440 }
1441 next_block->logical_bytenr = next_bytenr;
1442 next_block->mirror_num = mirror_num;
1443 }
1444
1445 l = btrfsic_block_link_lookup_or_add(state,
1446 &next_block_ctx,
1447 next_block, block,
1448 generation);
1449 btrfsic_release_block_ctx(&next_block_ctx);
1450 if (NULL == l)
1451 return -1;
1452 }
1453
1454 next_bytenr += chunk_len;
1455 num_bytes -= chunk_len;
1456 }
1457
1458 return 0;
1459}
1460
1461static int btrfsic_map_block(struct btrfsic_state *state, u64 bytenr, u32 len,
1462 struct btrfsic_block_data_ctx *block_ctx_out,
1463 int mirror_num)
1464{
1465 int ret;
1466 u64 length;
1467 struct btrfs_bio *multi = NULL;
1468 struct btrfs_device *device;
1469
1470 length = len;
1471 ret = btrfs_map_block(&state->root->fs_info->mapping_tree, READ,
1472 bytenr, &length, &multi, mirror_num);
1473
1474 device = multi->stripes[0].dev;
1475 block_ctx_out->dev = btrfsic_dev_state_lookup(device->bdev);
1476 block_ctx_out->dev_bytenr = multi->stripes[0].physical;
1477 block_ctx_out->start = bytenr;
1478 block_ctx_out->len = len;
1479 block_ctx_out->data = NULL;
1480 block_ctx_out->bh = NULL;
1481
1482 if (0 == ret)
1483 kfree(multi);
1484 if (NULL == block_ctx_out->dev) {
1485 ret = -ENXIO;
1486 printk(KERN_INFO "btrfsic: error, cannot lookup dev (#1)!\n");
1487 }
1488
1489 return ret;
1490}
1491
1492static int btrfsic_map_superblock(struct btrfsic_state *state, u64 bytenr,
1493 u32 len, struct block_device *bdev,
1494 struct btrfsic_block_data_ctx *block_ctx_out)
1495{
1496 block_ctx_out->dev = btrfsic_dev_state_lookup(bdev);
1497 block_ctx_out->dev_bytenr = bytenr;
1498 block_ctx_out->start = bytenr;
1499 block_ctx_out->len = len;
1500 block_ctx_out->data = NULL;
1501 block_ctx_out->bh = NULL;
1502 if (NULL != block_ctx_out->dev) {
1503 return 0;
1504 } else {
1505 printk(KERN_INFO "btrfsic: error, cannot lookup dev (#2)!\n");
1506 return -ENXIO;
1507 }
1508}
1509
1510static void btrfsic_release_block_ctx(struct btrfsic_block_data_ctx *block_ctx)
1511{
1512 if (NULL != block_ctx->bh) {
1513 brelse(block_ctx->bh);
1514 block_ctx->bh = NULL;
1515 }
1516}
1517
1518static int btrfsic_read_block(struct btrfsic_state *state,
1519 struct btrfsic_block_data_ctx *block_ctx)
1520{
1521 block_ctx->bh = NULL;
1522 if (block_ctx->dev_bytenr & 4095) {
1523 printk(KERN_INFO
1524 "btrfsic: read_block() with unaligned bytenr %llu\n",
1525 (unsigned long long)block_ctx->dev_bytenr);
1526 return -1;
1527 }
1528 if (block_ctx->len > 4096) {
1529 printk(KERN_INFO
1530 "btrfsic: read_block() with too huge size %d\n",
1531 block_ctx->len);
1532 return -1;
1533 }
1534
1535 block_ctx->bh = __bread(block_ctx->dev->bdev,
1536 block_ctx->dev_bytenr >> 12, 4096);
1537 if (NULL == block_ctx->bh)
1538 return -1;
1539 block_ctx->data = block_ctx->bh->b_data;
1540
1541 return block_ctx->len;
1542}
1543
1544static void btrfsic_dump_database(struct btrfsic_state *state)
1545{
1546 struct list_head *elem_all;
1547
1548 BUG_ON(NULL == state);
1549
1550 printk(KERN_INFO "all_blocks_list:\n");
1551 list_for_each(elem_all, &state->all_blocks_list) {
1552 const struct btrfsic_block *const b_all =
1553 list_entry(elem_all, struct btrfsic_block,
1554 all_blocks_node);
1555 struct list_head *elem_ref_to;
1556 struct list_head *elem_ref_from;
1557
1558 printk(KERN_INFO "%c-block @%llu (%s/%llu/%d)\n",
1559 btrfsic_get_block_type(state, b_all),
1560 (unsigned long long)b_all->logical_bytenr,
1561 b_all->dev_state->name,
1562 (unsigned long long)b_all->dev_bytenr,
1563 b_all->mirror_num);
1564
1565 list_for_each(elem_ref_to, &b_all->ref_to_list) {
1566 const struct btrfsic_block_link *const l =
1567 list_entry(elem_ref_to,
1568 struct btrfsic_block_link,
1569 node_ref_to);
1570
1571 printk(KERN_INFO " %c @%llu (%s/%llu/%d)"
1572 " refers %u* to"
1573 " %c @%llu (%s/%llu/%d)\n",
1574 btrfsic_get_block_type(state, b_all),
1575 (unsigned long long)b_all->logical_bytenr,
1576 b_all->dev_state->name,
1577 (unsigned long long)b_all->dev_bytenr,
1578 b_all->mirror_num,
1579 l->ref_cnt,
1580 btrfsic_get_block_type(state, l->block_ref_to),
1581 (unsigned long long)
1582 l->block_ref_to->logical_bytenr,
1583 l->block_ref_to->dev_state->name,
1584 (unsigned long long)l->block_ref_to->dev_bytenr,
1585 l->block_ref_to->mirror_num);
1586 }
1587
1588 list_for_each(elem_ref_from, &b_all->ref_from_list) {
1589 const struct btrfsic_block_link *const l =
1590 list_entry(elem_ref_from,
1591 struct btrfsic_block_link,
1592 node_ref_from);
1593
1594 printk(KERN_INFO " %c @%llu (%s/%llu/%d)"
1595 " is ref %u* from"
1596 " %c @%llu (%s/%llu/%d)\n",
1597 btrfsic_get_block_type(state, b_all),
1598 (unsigned long long)b_all->logical_bytenr,
1599 b_all->dev_state->name,
1600 (unsigned long long)b_all->dev_bytenr,
1601 b_all->mirror_num,
1602 l->ref_cnt,
1603 btrfsic_get_block_type(state, l->block_ref_from),
1604 (unsigned long long)
1605 l->block_ref_from->logical_bytenr,
1606 l->block_ref_from->dev_state->name,
1607 (unsigned long long)
1608 l->block_ref_from->dev_bytenr,
1609 l->block_ref_from->mirror_num);
1610 }
1611
1612 printk(KERN_INFO "\n");
1613 }
1614}
1615
1616/*
1617 * Test whether the disk block contains a tree block (leaf or node)
1618 * (note that this test fails for the super block)
1619 */
1620static int btrfsic_test_for_metadata(struct btrfsic_state *state,
1621 const u8 *data, unsigned int size)
1622{
1623 struct btrfs_header *h;
1624 u8 csum[BTRFS_CSUM_SIZE];
1625 u32 crc = ~(u32)0;
1626 int fail = 0;
1627 int crc_fail = 0;
1628
1629 h = (struct btrfs_header *)data;
1630
1631 if (memcmp(h->fsid, state->root->fs_info->fsid, BTRFS_UUID_SIZE))
1632 fail++;
1633
1634 crc = crc32c(crc, data + BTRFS_CSUM_SIZE, PAGE_SIZE - BTRFS_CSUM_SIZE);
1635 btrfs_csum_final(crc, csum);
1636 if (memcmp(csum, h->csum, state->csum_size))
1637 crc_fail++;
1638
1639 return fail || crc_fail;
1640}
1641
1642static void btrfsic_process_written_block(struct btrfsic_dev_state *dev_state,
1643 u64 dev_bytenr,
1644 u8 *mapped_data, unsigned int len,
1645 struct bio *bio,
1646 int *bio_is_patched,
1647 struct buffer_head *bh,
1648 int submit_bio_bh_rw)
1649{
1650 int is_metadata;
1651 struct btrfsic_block *block;
1652 struct btrfsic_block_data_ctx block_ctx;
1653 int ret;
1654 struct btrfsic_state *state = dev_state->state;
1655 struct block_device *bdev = dev_state->bdev;
1656
1657 WARN_ON(len > PAGE_SIZE);
1658 is_metadata = (0 == btrfsic_test_for_metadata(state, mapped_data, len));
1659 if (NULL != bio_is_patched)
1660 *bio_is_patched = 0;
1661
1662 block = btrfsic_block_hashtable_lookup(bdev, dev_bytenr,
1663 &state->block_hashtable);
1664 if (NULL != block) {
1665 u64 bytenr;
1666 struct list_head *elem_ref_to;
1667 struct list_head *tmp_ref_to;
1668
1669 if (block->is_superblock) {
1670 bytenr = le64_to_cpu(((struct btrfs_super_block *)
1671 mapped_data)->bytenr);
1672 is_metadata = 1;
1673 if (state->print_mask &
1674 BTRFSIC_PRINT_MASK_TREE_BEFORE_SB_WRITE) {
1675 printk(KERN_INFO
1676 "[before new superblock is written]:\n");
1677 btrfsic_dump_tree_sub(state, block, 0);
1678 }
1679 }
1680 if (is_metadata) {
1681 if (!block->is_superblock) {
1682 bytenr = le64_to_cpu(((struct btrfs_header *)
1683 mapped_data)->bytenr);
1684 btrfsic_cmp_log_and_dev_bytenr(state, bytenr,
1685 dev_state,
1686 dev_bytenr,
1687 mapped_data);
1688 }
1689 if (block->logical_bytenr != bytenr) {
1690 printk(KERN_INFO
1691 "Written block @%llu (%s/%llu/%d)"
1692 " found in hash table, %c,"
1693 " bytenr mismatch"
1694 " (!= stored %llu).\n",
1695 (unsigned long long)bytenr,
1696 dev_state->name,
1697 (unsigned long long)dev_bytenr,
1698 block->mirror_num,
1699 btrfsic_get_block_type(state, block),
1700 (unsigned long long)
1701 block->logical_bytenr);
1702 block->logical_bytenr = bytenr;
1703 } else if (state->print_mask &
1704 BTRFSIC_PRINT_MASK_VERBOSE)
1705 printk(KERN_INFO
1706 "Written block @%llu (%s/%llu/%d)"
1707 " found in hash table, %c.\n",
1708 (unsigned long long)bytenr,
1709 dev_state->name,
1710 (unsigned long long)dev_bytenr,
1711 block->mirror_num,
1712 btrfsic_get_block_type(state, block));
1713 } else {
1714 bytenr = block->logical_bytenr;
1715 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1716 printk(KERN_INFO
1717 "Written block @%llu (%s/%llu/%d)"
1718 " found in hash table, %c.\n",
1719 (unsigned long long)bytenr,
1720 dev_state->name,
1721 (unsigned long long)dev_bytenr,
1722 block->mirror_num,
1723 btrfsic_get_block_type(state, block));
1724 }
1725
1726 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1727 printk(KERN_INFO
1728 "ref_to_list: %cE, ref_from_list: %cE\n",
1729 list_empty(&block->ref_to_list) ? ' ' : '!',
1730 list_empty(&block->ref_from_list) ? ' ' : '!');
1731 if (btrfsic_is_block_ref_by_superblock(state, block, 0)) {
1732 printk(KERN_INFO "btrfs: attempt to overwrite %c-block"
1733 " @%llu (%s/%llu/%d), old(gen=%llu,"
1734 " objectid=%llu, type=%d, offset=%llu),"
1735 " new(gen=%llu),"
1736 " which is referenced by most recent superblock"
1737 " (superblockgen=%llu)!\n",
1738 btrfsic_get_block_type(state, block),
1739 (unsigned long long)bytenr,
1740 dev_state->name,
1741 (unsigned long long)dev_bytenr,
1742 block->mirror_num,
1743 (unsigned long long)block->generation,
1744 (unsigned long long)
1745 le64_to_cpu(block->disk_key.objectid),
1746 block->disk_key.type,
1747 (unsigned long long)
1748 le64_to_cpu(block->disk_key.offset),
1749 (unsigned long long)
1750 le64_to_cpu(((struct btrfs_header *)
1751 mapped_data)->generation),
1752 (unsigned long long)
1753 state->max_superblock_generation);
1754 btrfsic_dump_tree(state);
1755 }
1756
1757 if (!block->is_iodone && !block->never_written) {
1758 printk(KERN_INFO "btrfs: attempt to overwrite %c-block"
1759 " @%llu (%s/%llu/%d), oldgen=%llu, newgen=%llu,"
1760 " which is not yet iodone!\n",
1761 btrfsic_get_block_type(state, block),
1762 (unsigned long long)bytenr,
1763 dev_state->name,
1764 (unsigned long long)dev_bytenr,
1765 block->mirror_num,
1766 (unsigned long long)block->generation,
1767 (unsigned long long)
1768 le64_to_cpu(((struct btrfs_header *)
1769 mapped_data)->generation));
1770 /* it would not be safe to go on */
1771 btrfsic_dump_tree(state);
1772 return;
1773 }
1774
1775 /*
1776 * Clear all references of this block. Do not free
1777 * the block itself even if is not referenced anymore
1778 * because it still carries valueable information
1779 * like whether it was ever written and IO completed.
1780 */
1781 list_for_each_safe(elem_ref_to, tmp_ref_to,
1782 &block->ref_to_list) {
1783 struct btrfsic_block_link *const l =
1784 list_entry(elem_ref_to,
1785 struct btrfsic_block_link,
1786 node_ref_to);
1787
1788 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1789 btrfsic_print_rem_link(state, l);
1790 l->ref_cnt--;
1791 if (0 == l->ref_cnt) {
1792 list_del(&l->node_ref_to);
1793 list_del(&l->node_ref_from);
1794 btrfsic_block_link_hashtable_remove(l);
1795 btrfsic_block_link_free(l);
1796 }
1797 }
1798
1799 if (block->is_superblock)
1800 ret = btrfsic_map_superblock(state, bytenr, len,
1801 bdev, &block_ctx);
1802 else
1803 ret = btrfsic_map_block(state, bytenr, len,
1804 &block_ctx, 0);
1805 if (ret) {
1806 printk(KERN_INFO
1807 "btrfsic: btrfsic_map_block(root @%llu)"
1808 " failed!\n", (unsigned long long)bytenr);
1809 return;
1810 }
1811 block_ctx.data = mapped_data;
1812 /* the following is required in case of writes to mirrors,
1813 * use the same that was used for the lookup */
1814 block_ctx.dev = dev_state;
1815 block_ctx.dev_bytenr = dev_bytenr;
1816
1817 if (is_metadata || state->include_extent_data) {
1818 block->never_written = 0;
1819 block->iodone_w_error = 0;
1820 if (NULL != bio) {
1821 block->is_iodone = 0;
1822 BUG_ON(NULL == bio_is_patched);
1823 if (!*bio_is_patched) {
1824 block->orig_bio_bh_private =
1825 bio->bi_private;
1826 block->orig_bio_bh_end_io.bio =
1827 bio->bi_end_io;
1828 block->next_in_same_bio = NULL;
1829 bio->bi_private = block;
1830 bio->bi_end_io = btrfsic_bio_end_io;
1831 *bio_is_patched = 1;
1832 } else {
1833 struct btrfsic_block *chained_block =
1834 (struct btrfsic_block *)
1835 bio->bi_private;
1836
1837 BUG_ON(NULL == chained_block);
1838 block->orig_bio_bh_private =
1839 chained_block->orig_bio_bh_private;
1840 block->orig_bio_bh_end_io.bio =
1841 chained_block->orig_bio_bh_end_io.
1842 bio;
1843 block->next_in_same_bio = chained_block;
1844 bio->bi_private = block;
1845 }
1846 } else if (NULL != bh) {
1847 block->is_iodone = 0;
1848 block->orig_bio_bh_private = bh->b_private;
1849 block->orig_bio_bh_end_io.bh = bh->b_end_io;
1850 block->next_in_same_bio = NULL;
1851 bh->b_private = block;
1852 bh->b_end_io = btrfsic_bh_end_io;
1853 } else {
1854 block->is_iodone = 1;
1855 block->orig_bio_bh_private = NULL;
1856 block->orig_bio_bh_end_io.bio = NULL;
1857 block->next_in_same_bio = NULL;
1858 }
1859 }
1860
1861 block->flush_gen = dev_state->last_flush_gen + 1;
1862 block->submit_bio_bh_rw = submit_bio_bh_rw;
1863 if (is_metadata) {
1864 block->logical_bytenr = bytenr;
1865 block->is_metadata = 1;
1866 if (block->is_superblock) {
1867 ret = btrfsic_process_written_superblock(
1868 state,
1869 block,
1870 (struct btrfs_super_block *)
1871 mapped_data);
1872 if (state->print_mask &
1873 BTRFSIC_PRINT_MASK_TREE_AFTER_SB_WRITE) {
1874 printk(KERN_INFO
1875 "[after new superblock is written]:\n");
1876 btrfsic_dump_tree_sub(state, block, 0);
1877 }
1878 } else {
1879 block->mirror_num = 0; /* unknown */
1880 ret = btrfsic_process_metablock(
1881 state,
1882 block,
1883 &block_ctx,
1884 (struct btrfs_header *)
1885 block_ctx.data,
1886 0, 0);
1887 }
1888 if (ret)
1889 printk(KERN_INFO
1890 "btrfsic: btrfsic_process_metablock"
1891 "(root @%llu) failed!\n",
1892 (unsigned long long)dev_bytenr);
1893 } else {
1894 block->is_metadata = 0;
1895 block->mirror_num = 0; /* unknown */
1896 block->generation = BTRFSIC_GENERATION_UNKNOWN;
1897 if (!state->include_extent_data
1898 && list_empty(&block->ref_from_list)) {
1899 /*
1900 * disk block is overwritten with extent
1901 * data (not meta data) and we are configured
1902 * to not include extent data: take the
1903 * chance and free the block's memory
1904 */
1905 btrfsic_block_hashtable_remove(block);
1906 list_del(&block->all_blocks_node);
1907 btrfsic_block_free(block);
1908 }
1909 }
1910 btrfsic_release_block_ctx(&block_ctx);
1911 } else {
1912 /* block has not been found in hash table */
1913 u64 bytenr;
1914
1915 if (!is_metadata) {
1916 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1917 printk(KERN_INFO "Written block (%s/%llu/?)"
1918 " !found in hash table, D.\n",
1919 dev_state->name,
1920 (unsigned long long)dev_bytenr);
1921 if (!state->include_extent_data)
1922 return; /* ignore that written D block */
1923
1924 /* this is getting ugly for the
1925 * include_extent_data case... */
1926 bytenr = 0; /* unknown */
1927 block_ctx.start = bytenr;
1928 block_ctx.len = len;
1929 block_ctx.bh = NULL;
1930 } else {
1931 bytenr = le64_to_cpu(((struct btrfs_header *)
1932 mapped_data)->bytenr);
1933 btrfsic_cmp_log_and_dev_bytenr(state, bytenr, dev_state,
1934 dev_bytenr,
1935 mapped_data);
1936 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
1937 printk(KERN_INFO
1938 "Written block @%llu (%s/%llu/?)"
1939 " !found in hash table, M.\n",
1940 (unsigned long long)bytenr,
1941 dev_state->name,
1942 (unsigned long long)dev_bytenr);
1943
1944 ret = btrfsic_map_block(state, bytenr, len, &block_ctx,
1945 0);
1946 if (ret) {
1947 printk(KERN_INFO
1948 "btrfsic: btrfsic_map_block(root @%llu)"
1949 " failed!\n",
1950 (unsigned long long)dev_bytenr);
1951 return;
1952 }
1953 }
1954 block_ctx.data = mapped_data;
1955 /* the following is required in case of writes to mirrors,
1956 * use the same that was used for the lookup */
1957 block_ctx.dev = dev_state;
1958 block_ctx.dev_bytenr = dev_bytenr;
1959
1960 block = btrfsic_block_alloc();
1961 if (NULL == block) {
1962 printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
1963 btrfsic_release_block_ctx(&block_ctx);
1964 return;
1965 }
1966 block->dev_state = dev_state;
1967 block->dev_bytenr = dev_bytenr;
1968 block->logical_bytenr = bytenr;
1969 block->is_metadata = is_metadata;
1970 block->never_written = 0;
1971 block->iodone_w_error = 0;
1972 block->mirror_num = 0; /* unknown */
1973 block->flush_gen = dev_state->last_flush_gen + 1;
1974 block->submit_bio_bh_rw = submit_bio_bh_rw;
1975 if (NULL != bio) {
1976 block->is_iodone = 0;
1977 BUG_ON(NULL == bio_is_patched);
1978 if (!*bio_is_patched) {
1979 block->orig_bio_bh_private = bio->bi_private;
1980 block->orig_bio_bh_end_io.bio = bio->bi_end_io;
1981 block->next_in_same_bio = NULL;
1982 bio->bi_private = block;
1983 bio->bi_end_io = btrfsic_bio_end_io;
1984 *bio_is_patched = 1;
1985 } else {
1986 struct btrfsic_block *chained_block =
1987 (struct btrfsic_block *)
1988 bio->bi_private;
1989
1990 BUG_ON(NULL == chained_block);
1991 block->orig_bio_bh_private =
1992 chained_block->orig_bio_bh_private;
1993 block->orig_bio_bh_end_io.bio =
1994 chained_block->orig_bio_bh_end_io.bio;
1995 block->next_in_same_bio = chained_block;
1996 bio->bi_private = block;
1997 }
1998 } else if (NULL != bh) {
1999 block->is_iodone = 0;
2000 block->orig_bio_bh_private = bh->b_private;
2001 block->orig_bio_bh_end_io.bh = bh->b_end_io;
2002 block->next_in_same_bio = NULL;
2003 bh->b_private = block;
2004 bh->b_end_io = btrfsic_bh_end_io;
2005 } else {
2006 block->is_iodone = 1;
2007 block->orig_bio_bh_private = NULL;
2008 block->orig_bio_bh_end_io.bio = NULL;
2009 block->next_in_same_bio = NULL;
2010 }
2011 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2012 printk(KERN_INFO
2013 "New written %c-block @%llu (%s/%llu/%d)\n",
2014 is_metadata ? 'M' : 'D',
2015 (unsigned long long)block->logical_bytenr,
2016 block->dev_state->name,
2017 (unsigned long long)block->dev_bytenr,
2018 block->mirror_num);
2019 list_add(&block->all_blocks_node, &state->all_blocks_list);
2020 btrfsic_block_hashtable_add(block, &state->block_hashtable);
2021
2022 if (is_metadata) {
2023 ret = btrfsic_process_metablock(state, block,
2024 &block_ctx,
2025 (struct btrfs_header *)
2026 block_ctx.data, 0, 0);
2027 if (ret)
2028 printk(KERN_INFO
2029 "btrfsic: process_metablock(root @%llu)"
2030 " failed!\n",
2031 (unsigned long long)dev_bytenr);
2032 }
2033 btrfsic_release_block_ctx(&block_ctx);
2034 }
2035}
2036
2037static void btrfsic_bio_end_io(struct bio *bp, int bio_error_status)
2038{
2039 struct btrfsic_block *block = (struct btrfsic_block *)bp->bi_private;
2040 int iodone_w_error;
2041
2042 /* mutex is not held! This is not save if IO is not yet completed
2043 * on umount */
2044 iodone_w_error = 0;
2045 if (bio_error_status)
2046 iodone_w_error = 1;
2047
2048 BUG_ON(NULL == block);
2049 bp->bi_private = block->orig_bio_bh_private;
2050 bp->bi_end_io = block->orig_bio_bh_end_io.bio;
2051
2052 do {
2053 struct btrfsic_block *next_block;
2054 struct btrfsic_dev_state *const dev_state = block->dev_state;
2055
2056 if ((dev_state->state->print_mask &
2057 BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
2058 printk(KERN_INFO
2059 "bio_end_io(err=%d) for %c @%llu (%s/%llu/%d)\n",
2060 bio_error_status,
2061 btrfsic_get_block_type(dev_state->state, block),
2062 (unsigned long long)block->logical_bytenr,
2063 dev_state->name,
2064 (unsigned long long)block->dev_bytenr,
2065 block->mirror_num);
2066 next_block = block->next_in_same_bio;
2067 block->iodone_w_error = iodone_w_error;
2068 if (block->submit_bio_bh_rw & REQ_FLUSH) {
2069 dev_state->last_flush_gen++;
2070 if ((dev_state->state->print_mask &
2071 BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
2072 printk(KERN_INFO
2073 "bio_end_io() new %s flush_gen=%llu\n",
2074 dev_state->name,
2075 (unsigned long long)
2076 dev_state->last_flush_gen);
2077 }
2078 if (block->submit_bio_bh_rw & REQ_FUA)
2079 block->flush_gen = 0; /* FUA completed means block is
2080 * on disk */
2081 block->is_iodone = 1; /* for FLUSH, this releases the block */
2082 block = next_block;
2083 } while (NULL != block);
2084
2085 bp->bi_end_io(bp, bio_error_status);
2086}
2087
2088static void btrfsic_bh_end_io(struct buffer_head *bh, int uptodate)
2089{
2090 struct btrfsic_block *block = (struct btrfsic_block *)bh->b_private;
2091 int iodone_w_error = !uptodate;
2092 struct btrfsic_dev_state *dev_state;
2093
2094 BUG_ON(NULL == block);
2095 dev_state = block->dev_state;
2096 if ((dev_state->state->print_mask & BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
2097 printk(KERN_INFO
2098 "bh_end_io(error=%d) for %c @%llu (%s/%llu/%d)\n",
2099 iodone_w_error,
2100 btrfsic_get_block_type(dev_state->state, block),
2101 (unsigned long long)block->logical_bytenr,
2102 block->dev_state->name,
2103 (unsigned long long)block->dev_bytenr,
2104 block->mirror_num);
2105
2106 block->iodone_w_error = iodone_w_error;
2107 if (block->submit_bio_bh_rw & REQ_FLUSH) {
2108 dev_state->last_flush_gen++;
2109 if ((dev_state->state->print_mask &
2110 BTRFSIC_PRINT_MASK_END_IO_BIO_BH))
2111 printk(KERN_INFO
2112 "bh_end_io() new %s flush_gen=%llu\n",
2113 dev_state->name,
2114 (unsigned long long)dev_state->last_flush_gen);
2115 }
2116 if (block->submit_bio_bh_rw & REQ_FUA)
2117 block->flush_gen = 0; /* FUA completed means block is on disk */
2118
2119 bh->b_private = block->orig_bio_bh_private;
2120 bh->b_end_io = block->orig_bio_bh_end_io.bh;
2121 block->is_iodone = 1; /* for FLUSH, this releases the block */
2122 bh->b_end_io(bh, uptodate);
2123}
2124
2125static int btrfsic_process_written_superblock(
2126 struct btrfsic_state *state,
2127 struct btrfsic_block *const superblock,
2128 struct btrfs_super_block *const super_hdr)
2129{
2130 int pass;
2131
2132 superblock->generation = btrfs_super_generation(super_hdr);
2133 if (!(superblock->generation > state->max_superblock_generation ||
2134 0 == state->max_superblock_generation)) {
2135 if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
2136 printk(KERN_INFO
2137 "btrfsic: superblock @%llu (%s/%llu/%d)"
2138 " with old gen %llu <= %llu\n",
2139 (unsigned long long)superblock->logical_bytenr,
2140 superblock->dev_state->name,
2141 (unsigned long long)superblock->dev_bytenr,
2142 superblock->mirror_num,
2143 (unsigned long long)
2144 btrfs_super_generation(super_hdr),
2145 (unsigned long long)
2146 state->max_superblock_generation);
2147 } else {
2148 if (state->print_mask & BTRFSIC_PRINT_MASK_SUPERBLOCK_WRITE)
2149 printk(KERN_INFO
2150 "btrfsic: got new superblock @%llu (%s/%llu/%d)"
2151 " with new gen %llu > %llu\n",
2152 (unsigned long long)superblock->logical_bytenr,
2153 superblock->dev_state->name,
2154 (unsigned long long)superblock->dev_bytenr,
2155 superblock->mirror_num,
2156 (unsigned long long)
2157 btrfs_super_generation(super_hdr),
2158 (unsigned long long)
2159 state->max_superblock_generation);
2160
2161 state->max_superblock_generation =
2162 btrfs_super_generation(super_hdr);
2163 state->latest_superblock = superblock;
2164 }
2165
2166 for (pass = 0; pass < 3; pass++) {
2167 int ret;
2168 u64 next_bytenr;
2169 struct btrfsic_block *next_block;
2170 struct btrfsic_block_data_ctx tmp_next_block_ctx;
2171 struct btrfsic_block_link *l;
2172 int num_copies;
2173 int mirror_num;
2174 const char *additional_string = NULL;
2175 struct btrfs_disk_key tmp_disk_key;
2176
2177 tmp_disk_key.type = BTRFS_ROOT_ITEM_KEY;
2178 tmp_disk_key.offset = 0;
2179
2180 switch (pass) {
2181 case 0:
2182 tmp_disk_key.objectid =
2183 cpu_to_le64(BTRFS_ROOT_TREE_OBJECTID);
2184 additional_string = "root ";
2185 next_bytenr = btrfs_super_root(super_hdr);
2186 if (state->print_mask &
2187 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
2188 printk(KERN_INFO "root@%llu\n",
2189 (unsigned long long)next_bytenr);
2190 break;
2191 case 1:
2192 tmp_disk_key.objectid =
2193 cpu_to_le64(BTRFS_CHUNK_TREE_OBJECTID);
2194 additional_string = "chunk ";
2195 next_bytenr = btrfs_super_chunk_root(super_hdr);
2196 if (state->print_mask &
2197 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
2198 printk(KERN_INFO "chunk@%llu\n",
2199 (unsigned long long)next_bytenr);
2200 break;
2201 case 2:
2202 tmp_disk_key.objectid =
2203 cpu_to_le64(BTRFS_TREE_LOG_OBJECTID);
2204 additional_string = "log ";
2205 next_bytenr = btrfs_super_log_root(super_hdr);
2206 if (0 == next_bytenr)
2207 continue;
2208 if (state->print_mask &
2209 BTRFSIC_PRINT_MASK_ROOT_CHUNK_LOG_TREE_LOCATION)
2210 printk(KERN_INFO "log@%llu\n",
2211 (unsigned long long)next_bytenr);
2212 break;
2213 }
2214
2215 num_copies =
2216 btrfs_num_copies(&state->root->fs_info->mapping_tree,
2217 next_bytenr, PAGE_SIZE);
2218 if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES)
2219 printk(KERN_INFO "num_copies(log_bytenr=%llu) = %d\n",
2220 (unsigned long long)next_bytenr, num_copies);
2221 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
2222 int was_created;
2223
2224 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2225 printk(KERN_INFO
2226 "btrfsic_process_written_superblock("
2227 "mirror_num=%d)\n", mirror_num);
2228 ret = btrfsic_map_block(state, next_bytenr, PAGE_SIZE,
2229 &tmp_next_block_ctx,
2230 mirror_num);
2231 if (ret) {
2232 printk(KERN_INFO
2233 "btrfsic: btrfsic_map_block(@%llu,"
2234 " mirror=%d) failed!\n",
2235 (unsigned long long)next_bytenr,
2236 mirror_num);
2237 return -1;
2238 }
2239
2240 next_block = btrfsic_block_lookup_or_add(
2241 state,
2242 &tmp_next_block_ctx,
2243 additional_string,
2244 1, 0, 1,
2245 mirror_num,
2246 &was_created);
2247 if (NULL == next_block) {
2248 printk(KERN_INFO
2249 "btrfsic: error, kmalloc failed!\n");
2250 btrfsic_release_block_ctx(&tmp_next_block_ctx);
2251 return -1;
2252 }
2253
2254 next_block->disk_key = tmp_disk_key;
2255 if (was_created)
2256 next_block->generation =
2257 BTRFSIC_GENERATION_UNKNOWN;
2258 l = btrfsic_block_link_lookup_or_add(
2259 state,
2260 &tmp_next_block_ctx,
2261 next_block,
2262 superblock,
2263 BTRFSIC_GENERATION_UNKNOWN);
2264 btrfsic_release_block_ctx(&tmp_next_block_ctx);
2265 if (NULL == l)
2266 return -1;
2267 }
2268 }
2269
2270 if (-1 == btrfsic_check_all_ref_blocks(state, superblock, 0)) {
2271 WARN_ON(1);
2272 btrfsic_dump_tree(state);
2273 }
2274
2275 return 0;
2276}
2277
2278static int btrfsic_check_all_ref_blocks(struct btrfsic_state *state,
2279 struct btrfsic_block *const block,
2280 int recursion_level)
2281{
2282 struct list_head *elem_ref_to;
2283 int ret = 0;
2284
2285 if (recursion_level >= 3 + BTRFS_MAX_LEVEL) {
2286 /*
2287 * Note that this situation can happen and does not
2288 * indicate an error in regular cases. It happens
2289 * when disk blocks are freed and later reused.
2290 * The check-integrity module is not aware of any
2291 * block free operations, it just recognizes block
2292 * write operations. Therefore it keeps the linkage
2293 * information for a block until a block is
2294 * rewritten. This can temporarily cause incorrect
2295 * and even circular linkage informations. This
2296 * causes no harm unless such blocks are referenced
2297 * by the most recent super block.
2298 */
2299 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2300 printk(KERN_INFO
2301 "btrfsic: abort cyclic linkage (case 1).\n");
2302
2303 return ret;
2304 }
2305
2306 /*
2307 * This algorithm is recursive because the amount of used stack
2308 * space is very small and the max recursion depth is limited.
2309 */
2310 list_for_each(elem_ref_to, &block->ref_to_list) {
2311 const struct btrfsic_block_link *const l =
2312 list_entry(elem_ref_to, struct btrfsic_block_link,
2313 node_ref_to);
2314
2315 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2316 printk(KERN_INFO
2317 "rl=%d, %c @%llu (%s/%llu/%d)"
2318 " %u* refers to %c @%llu (%s/%llu/%d)\n",
2319 recursion_level,
2320 btrfsic_get_block_type(state, block),
2321 (unsigned long long)block->logical_bytenr,
2322 block->dev_state->name,
2323 (unsigned long long)block->dev_bytenr,
2324 block->mirror_num,
2325 l->ref_cnt,
2326 btrfsic_get_block_type(state, l->block_ref_to),
2327 (unsigned long long)
2328 l->block_ref_to->logical_bytenr,
2329 l->block_ref_to->dev_state->name,
2330 (unsigned long long)l->block_ref_to->dev_bytenr,
2331 l->block_ref_to->mirror_num);
2332 if (l->block_ref_to->never_written) {
2333 printk(KERN_INFO "btrfs: attempt to write superblock"
2334 " which references block %c @%llu (%s/%llu/%d)"
2335 " which is never written!\n",
2336 btrfsic_get_block_type(state, l->block_ref_to),
2337 (unsigned long long)
2338 l->block_ref_to->logical_bytenr,
2339 l->block_ref_to->dev_state->name,
2340 (unsigned long long)l->block_ref_to->dev_bytenr,
2341 l->block_ref_to->mirror_num);
2342 ret = -1;
2343 } else if (!l->block_ref_to->is_iodone) {
2344 printk(KERN_INFO "btrfs: attempt to write superblock"
2345 " which references block %c @%llu (%s/%llu/%d)"
2346 " which is not yet iodone!\n",
2347 btrfsic_get_block_type(state, l->block_ref_to),
2348 (unsigned long long)
2349 l->block_ref_to->logical_bytenr,
2350 l->block_ref_to->dev_state->name,
2351 (unsigned long long)l->block_ref_to->dev_bytenr,
2352 l->block_ref_to->mirror_num);
2353 ret = -1;
2354 } else if (l->parent_generation !=
2355 l->block_ref_to->generation &&
2356 BTRFSIC_GENERATION_UNKNOWN !=
2357 l->parent_generation &&
2358 BTRFSIC_GENERATION_UNKNOWN !=
2359 l->block_ref_to->generation) {
2360 printk(KERN_INFO "btrfs: attempt to write superblock"
2361 " which references block %c @%llu (%s/%llu/%d)"
2362 " with generation %llu !="
2363 " parent generation %llu!\n",
2364 btrfsic_get_block_type(state, l->block_ref_to),
2365 (unsigned long long)
2366 l->block_ref_to->logical_bytenr,
2367 l->block_ref_to->dev_state->name,
2368 (unsigned long long)l->block_ref_to->dev_bytenr,
2369 l->block_ref_to->mirror_num,
2370 (unsigned long long)l->block_ref_to->generation,
2371 (unsigned long long)l->parent_generation);
2372 ret = -1;
2373 } else if (l->block_ref_to->flush_gen >
2374 l->block_ref_to->dev_state->last_flush_gen) {
2375 printk(KERN_INFO "btrfs: attempt to write superblock"
2376 " which references block %c @%llu (%s/%llu/%d)"
2377 " which is not flushed out of disk's write cache"
2378 " (block flush_gen=%llu,"
2379 " dev->flush_gen=%llu)!\n",
2380 btrfsic_get_block_type(state, l->block_ref_to),
2381 (unsigned long long)
2382 l->block_ref_to->logical_bytenr,
2383 l->block_ref_to->dev_state->name,
2384 (unsigned long long)l->block_ref_to->dev_bytenr,
2385 l->block_ref_to->mirror_num,
2386 (unsigned long long)block->flush_gen,
2387 (unsigned long long)
2388 l->block_ref_to->dev_state->last_flush_gen);
2389 ret = -1;
2390 } else if (-1 == btrfsic_check_all_ref_blocks(state,
2391 l->block_ref_to,
2392 recursion_level +
2393 1)) {
2394 ret = -1;
2395 }
2396 }
2397
2398 return ret;
2399}
2400
2401static int btrfsic_is_block_ref_by_superblock(
2402 const struct btrfsic_state *state,
2403 const struct btrfsic_block *block,
2404 int recursion_level)
2405{
2406 struct list_head *elem_ref_from;
2407
2408 if (recursion_level >= 3 + BTRFS_MAX_LEVEL) {
2409 /* refer to comment at "abort cyclic linkage (case 1)" */
2410 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2411 printk(KERN_INFO
2412 "btrfsic: abort cyclic linkage (case 2).\n");
2413
2414 return 0;
2415 }
2416
2417 /*
2418 * This algorithm is recursive because the amount of used stack space
2419 * is very small and the max recursion depth is limited.
2420 */
2421 list_for_each(elem_ref_from, &block->ref_from_list) {
2422 const struct btrfsic_block_link *const l =
2423 list_entry(elem_ref_from, struct btrfsic_block_link,
2424 node_ref_from);
2425
2426 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2427 printk(KERN_INFO
2428 "rl=%d, %c @%llu (%s/%llu/%d)"
2429 " is ref %u* from %c @%llu (%s/%llu/%d)\n",
2430 recursion_level,
2431 btrfsic_get_block_type(state, block),
2432 (unsigned long long)block->logical_bytenr,
2433 block->dev_state->name,
2434 (unsigned long long)block->dev_bytenr,
2435 block->mirror_num,
2436 l->ref_cnt,
2437 btrfsic_get_block_type(state, l->block_ref_from),
2438 (unsigned long long)
2439 l->block_ref_from->logical_bytenr,
2440 l->block_ref_from->dev_state->name,
2441 (unsigned long long)
2442 l->block_ref_from->dev_bytenr,
2443 l->block_ref_from->mirror_num);
2444 if (l->block_ref_from->is_superblock &&
2445 state->latest_superblock->dev_bytenr ==
2446 l->block_ref_from->dev_bytenr &&
2447 state->latest_superblock->dev_state->bdev ==
2448 l->block_ref_from->dev_state->bdev)
2449 return 1;
2450 else if (btrfsic_is_block_ref_by_superblock(state,
2451 l->block_ref_from,
2452 recursion_level +
2453 1))
2454 return 1;
2455 }
2456
2457 return 0;
2458}
2459
2460static void btrfsic_print_add_link(const struct btrfsic_state *state,
2461 const struct btrfsic_block_link *l)
2462{
2463 printk(KERN_INFO
2464 "Add %u* link from %c @%llu (%s/%llu/%d)"
2465 " to %c @%llu (%s/%llu/%d).\n",
2466 l->ref_cnt,
2467 btrfsic_get_block_type(state, l->block_ref_from),
2468 (unsigned long long)l->block_ref_from->logical_bytenr,
2469 l->block_ref_from->dev_state->name,
2470 (unsigned long long)l->block_ref_from->dev_bytenr,
2471 l->block_ref_from->mirror_num,
2472 btrfsic_get_block_type(state, l->block_ref_to),
2473 (unsigned long long)l->block_ref_to->logical_bytenr,
2474 l->block_ref_to->dev_state->name,
2475 (unsigned long long)l->block_ref_to->dev_bytenr,
2476 l->block_ref_to->mirror_num);
2477}
2478
2479static void btrfsic_print_rem_link(const struct btrfsic_state *state,
2480 const struct btrfsic_block_link *l)
2481{
2482 printk(KERN_INFO
2483 "Rem %u* link from %c @%llu (%s/%llu/%d)"
2484 " to %c @%llu (%s/%llu/%d).\n",
2485 l->ref_cnt,
2486 btrfsic_get_block_type(state, l->block_ref_from),
2487 (unsigned long long)l->block_ref_from->logical_bytenr,
2488 l->block_ref_from->dev_state->name,
2489 (unsigned long long)l->block_ref_from->dev_bytenr,
2490 l->block_ref_from->mirror_num,
2491 btrfsic_get_block_type(state, l->block_ref_to),
2492 (unsigned long long)l->block_ref_to->logical_bytenr,
2493 l->block_ref_to->dev_state->name,
2494 (unsigned long long)l->block_ref_to->dev_bytenr,
2495 l->block_ref_to->mirror_num);
2496}
2497
2498static char btrfsic_get_block_type(const struct btrfsic_state *state,
2499 const struct btrfsic_block *block)
2500{
2501 if (block->is_superblock &&
2502 state->latest_superblock->dev_bytenr == block->dev_bytenr &&
2503 state->latest_superblock->dev_state->bdev == block->dev_state->bdev)
2504 return 'S';
2505 else if (block->is_superblock)
2506 return 's';
2507 else if (block->is_metadata)
2508 return 'M';
2509 else
2510 return 'D';
2511}
2512
2513static void btrfsic_dump_tree(const struct btrfsic_state *state)
2514{
2515 btrfsic_dump_tree_sub(state, state->latest_superblock, 0);
2516}
2517
2518static void btrfsic_dump_tree_sub(const struct btrfsic_state *state,
2519 const struct btrfsic_block *block,
2520 int indent_level)
2521{
2522 struct list_head *elem_ref_to;
2523 int indent_add;
2524 static char buf[80];
2525 int cursor_position;
2526
2527 /*
2528 * Should better fill an on-stack buffer with a complete line and
2529 * dump it at once when it is time to print a newline character.
2530 */
2531
2532 /*
2533 * This algorithm is recursive because the amount of used stack space
2534 * is very small and the max recursion depth is limited.
2535 */
2536 indent_add = sprintf(buf, "%c-%llu(%s/%llu/%d)",
2537 btrfsic_get_block_type(state, block),
2538 (unsigned long long)block->logical_bytenr,
2539 block->dev_state->name,
2540 (unsigned long long)block->dev_bytenr,
2541 block->mirror_num);
2542 if (indent_level + indent_add > BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL) {
2543 printk("[...]\n");
2544 return;
2545 }
2546 printk(buf);
2547 indent_level += indent_add;
2548 if (list_empty(&block->ref_to_list)) {
2549 printk("\n");
2550 return;
2551 }
2552 if (block->mirror_num > 1 &&
2553 !(state->print_mask & BTRFSIC_PRINT_MASK_TREE_WITH_ALL_MIRRORS)) {
2554 printk(" [...]\n");
2555 return;
2556 }
2557
2558 cursor_position = indent_level;
2559 list_for_each(elem_ref_to, &block->ref_to_list) {
2560 const struct btrfsic_block_link *const l =
2561 list_entry(elem_ref_to, struct btrfsic_block_link,
2562 node_ref_to);
2563
2564 while (cursor_position < indent_level) {
2565 printk(" ");
2566 cursor_position++;
2567 }
2568 if (l->ref_cnt > 1)
2569 indent_add = sprintf(buf, " %d*--> ", l->ref_cnt);
2570 else
2571 indent_add = sprintf(buf, " --> ");
2572 if (indent_level + indent_add >
2573 BTRFSIC_TREE_DUMP_MAX_INDENT_LEVEL) {
2574 printk("[...]\n");
2575 cursor_position = 0;
2576 continue;
2577 }
2578
2579 printk(buf);
2580
2581 btrfsic_dump_tree_sub(state, l->block_ref_to,
2582 indent_level + indent_add);
2583 cursor_position = 0;
2584 }
2585}
2586
2587static struct btrfsic_block_link *btrfsic_block_link_lookup_or_add(
2588 struct btrfsic_state *state,
2589 struct btrfsic_block_data_ctx *next_block_ctx,
2590 struct btrfsic_block *next_block,
2591 struct btrfsic_block *from_block,
2592 u64 parent_generation)
2593{
2594 struct btrfsic_block_link *l;
2595
2596 l = btrfsic_block_link_hashtable_lookup(next_block_ctx->dev->bdev,
2597 next_block_ctx->dev_bytenr,
2598 from_block->dev_state->bdev,
2599 from_block->dev_bytenr,
2600 &state->block_link_hashtable);
2601 if (NULL == l) {
2602 l = btrfsic_block_link_alloc();
2603 if (NULL == l) {
2604 printk(KERN_INFO
2605 "btrfsic: error, kmalloc" " failed!\n");
2606 return NULL;
2607 }
2608
2609 l->block_ref_to = next_block;
2610 l->block_ref_from = from_block;
2611 l->ref_cnt = 1;
2612 l->parent_generation = parent_generation;
2613
2614 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2615 btrfsic_print_add_link(state, l);
2616
2617 list_add(&l->node_ref_to, &from_block->ref_to_list);
2618 list_add(&l->node_ref_from, &next_block->ref_from_list);
2619
2620 btrfsic_block_link_hashtable_add(l,
2621 &state->block_link_hashtable);
2622 } else {
2623 l->ref_cnt++;
2624 l->parent_generation = parent_generation;
2625 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2626 btrfsic_print_add_link(state, l);
2627 }
2628
2629 return l;
2630}
2631
2632static struct btrfsic_block *btrfsic_block_lookup_or_add(
2633 struct btrfsic_state *state,
2634 struct btrfsic_block_data_ctx *block_ctx,
2635 const char *additional_string,
2636 int is_metadata,
2637 int is_iodone,
2638 int never_written,
2639 int mirror_num,
2640 int *was_created)
2641{
2642 struct btrfsic_block *block;
2643
2644 block = btrfsic_block_hashtable_lookup(block_ctx->dev->bdev,
2645 block_ctx->dev_bytenr,
2646 &state->block_hashtable);
2647 if (NULL == block) {
2648 struct btrfsic_dev_state *dev_state;
2649
2650 block = btrfsic_block_alloc();
2651 if (NULL == block) {
2652 printk(KERN_INFO "btrfsic: error, kmalloc failed!\n");
2653 return NULL;
2654 }
2655 dev_state = btrfsic_dev_state_lookup(block_ctx->dev->bdev);
2656 if (NULL == dev_state) {
2657 printk(KERN_INFO
2658 "btrfsic: error, lookup dev_state failed!\n");
2659 btrfsic_block_free(block);
2660 return NULL;
2661 }
2662 block->dev_state = dev_state;
2663 block->dev_bytenr = block_ctx->dev_bytenr;
2664 block->logical_bytenr = block_ctx->start;
2665 block->is_metadata = is_metadata;
2666 block->is_iodone = is_iodone;
2667 block->never_written = never_written;
2668 block->mirror_num = mirror_num;
2669 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
2670 printk(KERN_INFO
2671 "New %s%c-block @%llu (%s/%llu/%d)\n",
2672 additional_string,
2673 btrfsic_get_block_type(state, block),
2674 (unsigned long long)block->logical_bytenr,
2675 dev_state->name,
2676 (unsigned long long)block->dev_bytenr,
2677 mirror_num);
2678 list_add(&block->all_blocks_node, &state->all_blocks_list);
2679 btrfsic_block_hashtable_add(block, &state->block_hashtable);
2680 if (NULL != was_created)
2681 *was_created = 1;
2682 } else {
2683 if (NULL != was_created)
2684 *was_created = 0;
2685 }
2686
2687 return block;
2688}
2689
2690static void btrfsic_cmp_log_and_dev_bytenr(struct btrfsic_state *state,
2691 u64 bytenr,
2692 struct btrfsic_dev_state *dev_state,
2693 u64 dev_bytenr, char *data)
2694{
2695 int num_copies;
2696 int mirror_num;
2697 int ret;
2698 struct btrfsic_block_data_ctx block_ctx;
2699 int match = 0;
2700
2701 num_copies = btrfs_num_copies(&state->root->fs_info->mapping_tree,
2702 bytenr, PAGE_SIZE);
2703
2704 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
2705 ret = btrfsic_map_block(state, bytenr, PAGE_SIZE,
2706 &block_ctx, mirror_num);
2707 if (ret) {
2708 printk(KERN_INFO "btrfsic:"
2709 " btrfsic_map_block(logical @%llu,"
2710 " mirror %d) failed!\n",
2711 (unsigned long long)bytenr, mirror_num);
2712 continue;
2713 }
2714
2715 if (dev_state->bdev == block_ctx.dev->bdev &&
2716 dev_bytenr == block_ctx.dev_bytenr) {
2717 match++;
2718 btrfsic_release_block_ctx(&block_ctx);
2719 break;
2720 }
2721 btrfsic_release_block_ctx(&block_ctx);
2722 }
2723
2724 if (!match) {
2725 printk(KERN_INFO "btrfs: attempt to write M-block which contains logical bytenr that doesn't map to dev+physical bytenr of submit_bio,"
2726 " buffer->log_bytenr=%llu, submit_bio(bdev=%s,"
2727 " phys_bytenr=%llu)!\n",
2728 (unsigned long long)bytenr, dev_state->name,
2729 (unsigned long long)dev_bytenr);
2730 for (mirror_num = 1; mirror_num <= num_copies; mirror_num++) {
2731 ret = btrfsic_map_block(state, bytenr, PAGE_SIZE,
2732 &block_ctx, mirror_num);
2733 if (ret)
2734 continue;
2735
2736 printk(KERN_INFO "Read logical bytenr @%llu maps to"
2737 " (%s/%llu/%d)\n",
2738 (unsigned long long)bytenr,
2739 block_ctx.dev->name,
2740 (unsigned long long)block_ctx.dev_bytenr,
2741 mirror_num);
2742 }
2743 WARN_ON(1);
2744 }
2745}
2746
2747static struct btrfsic_dev_state *btrfsic_dev_state_lookup(
2748 struct block_device *bdev)
2749{
2750 struct btrfsic_dev_state *ds;
2751
2752 ds = btrfsic_dev_state_hashtable_lookup(bdev,
2753 &btrfsic_dev_state_hashtable);
2754 return ds;
2755}
2756
2757int btrfsic_submit_bh(int rw, struct buffer_head *bh)
2758{
2759 struct btrfsic_dev_state *dev_state;
2760
2761 if (!btrfsic_is_initialized)
2762 return submit_bh(rw, bh);
2763
2764 mutex_lock(&btrfsic_mutex);
2765 /* since btrfsic_submit_bh() might also be called before
2766 * btrfsic_mount(), this might return NULL */
2767 dev_state = btrfsic_dev_state_lookup(bh->b_bdev);
2768
2769 /* Only called to write the superblock (incl. FLUSH/FUA) */
2770 if (NULL != dev_state &&
2771 (rw & WRITE) && bh->b_size > 0) {
2772 u64 dev_bytenr;
2773
2774 dev_bytenr = 4096 * bh->b_blocknr;
2775 if (dev_state->state->print_mask &
2776 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
2777 printk(KERN_INFO
2778 "submit_bh(rw=0x%x, blocknr=%lu (bytenr %llu),"
2779 " size=%lu, data=%p, bdev=%p)\n",
2780 rw, bh->b_blocknr,
2781 (unsigned long long)dev_bytenr, bh->b_size,
2782 bh->b_data, bh->b_bdev);
2783 btrfsic_process_written_block(dev_state, dev_bytenr,
2784 bh->b_data, bh->b_size, NULL,
2785 NULL, bh, rw);
2786 } else if (NULL != dev_state && (rw & REQ_FLUSH)) {
2787 if (dev_state->state->print_mask &
2788 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
2789 printk(KERN_INFO
2790 "submit_bh(rw=0x%x) FLUSH, bdev=%p)\n",
2791 rw, bh->b_bdev);
2792 if (!dev_state->dummy_block_for_bio_bh_flush.is_iodone) {
2793 if ((dev_state->state->print_mask &
2794 (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
2795 BTRFSIC_PRINT_MASK_VERBOSE)))
2796 printk(KERN_INFO
2797 "btrfsic_submit_bh(%s) with FLUSH"
2798 " but dummy block already in use"
2799 " (ignored)!\n",
2800 dev_state->name);
2801 } else {
2802 struct btrfsic_block *const block =
2803 &dev_state->dummy_block_for_bio_bh_flush;
2804
2805 block->is_iodone = 0;
2806 block->never_written = 0;
2807 block->iodone_w_error = 0;
2808 block->flush_gen = dev_state->last_flush_gen + 1;
2809 block->submit_bio_bh_rw = rw;
2810 block->orig_bio_bh_private = bh->b_private;
2811 block->orig_bio_bh_end_io.bh = bh->b_end_io;
2812 block->next_in_same_bio = NULL;
2813 bh->b_private = block;
2814 bh->b_end_io = btrfsic_bh_end_io;
2815 }
2816 }
2817 mutex_unlock(&btrfsic_mutex);
2818 return submit_bh(rw, bh);
2819}
2820
2821void btrfsic_submit_bio(int rw, struct bio *bio)
2822{
2823 struct btrfsic_dev_state *dev_state;
2824
2825 if (!btrfsic_is_initialized) {
2826 submit_bio(rw, bio);
2827 return;
2828 }
2829
2830 mutex_lock(&btrfsic_mutex);
2831 /* since btrfsic_submit_bio() is also called before
2832 * btrfsic_mount(), this might return NULL */
2833 dev_state = btrfsic_dev_state_lookup(bio->bi_bdev);
2834 if (NULL != dev_state &&
2835 (rw & WRITE) && NULL != bio->bi_io_vec) {
2836 unsigned int i;
2837 u64 dev_bytenr;
2838 int bio_is_patched;
2839
2840 dev_bytenr = 512 * bio->bi_sector;
2841 bio_is_patched = 0;
2842 if (dev_state->state->print_mask &
2843 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
2844 printk(KERN_INFO
2845 "submit_bio(rw=0x%x, bi_vcnt=%u,"
2846 " bi_sector=%lu (bytenr %llu), bi_bdev=%p)\n",
2847 rw, bio->bi_vcnt, bio->bi_sector,
2848 (unsigned long long)dev_bytenr,
2849 bio->bi_bdev);
2850
2851 for (i = 0; i < bio->bi_vcnt; i++) {
2852 u8 *mapped_data;
2853
2854 mapped_data = kmap(bio->bi_io_vec[i].bv_page);
2855 if ((BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
2856 BTRFSIC_PRINT_MASK_VERBOSE) ==
2857 (dev_state->state->print_mask &
2858 (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
2859 BTRFSIC_PRINT_MASK_VERBOSE)))
2860 printk(KERN_INFO
2861 "#%u: page=%p, mapped=%p, len=%u,"
2862 " offset=%u\n",
2863 i, bio->bi_io_vec[i].bv_page,
2864 mapped_data,
2865 bio->bi_io_vec[i].bv_len,
2866 bio->bi_io_vec[i].bv_offset);
2867 btrfsic_process_written_block(dev_state, dev_bytenr,
2868 mapped_data,
2869 bio->bi_io_vec[i].bv_len,
2870 bio, &bio_is_patched,
2871 NULL, rw);
2872 kunmap(bio->bi_io_vec[i].bv_page);
2873 dev_bytenr += bio->bi_io_vec[i].bv_len;
2874 }
2875 } else if (NULL != dev_state && (rw & REQ_FLUSH)) {
2876 if (dev_state->state->print_mask &
2877 BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH)
2878 printk(KERN_INFO
2879 "submit_bio(rw=0x%x) FLUSH, bdev=%p)\n",
2880 rw, bio->bi_bdev);
2881 if (!dev_state->dummy_block_for_bio_bh_flush.is_iodone) {
2882 if ((dev_state->state->print_mask &
2883 (BTRFSIC_PRINT_MASK_SUBMIT_BIO_BH |
2884 BTRFSIC_PRINT_MASK_VERBOSE)))
2885 printk(KERN_INFO
2886 "btrfsic_submit_bio(%s) with FLUSH"
2887 " but dummy block already in use"
2888 " (ignored)!\n",
2889 dev_state->name);
2890 } else {
2891 struct btrfsic_block *const block =
2892 &dev_state->dummy_block_for_bio_bh_flush;
2893
2894 block->is_iodone = 0;
2895 block->never_written = 0;
2896 block->iodone_w_error = 0;
2897 block->flush_gen = dev_state->last_flush_gen + 1;
2898 block->submit_bio_bh_rw = rw;
2899 block->orig_bio_bh_private = bio->bi_private;
2900 block->orig_bio_bh_end_io.bio = bio->bi_end_io;
2901 block->next_in_same_bio = NULL;
2902 bio->bi_private = block;
2903 bio->bi_end_io = btrfsic_bio_end_io;
2904 }
2905 }
2906 mutex_unlock(&btrfsic_mutex);
2907
2908 submit_bio(rw, bio);
2909}
2910
2911int btrfsic_mount(struct btrfs_root *root,
2912 struct btrfs_fs_devices *fs_devices,
2913 int including_extent_data, u32 print_mask)
2914{
2915 int ret;
2916 struct btrfsic_state *state;
2917 struct list_head *dev_head = &fs_devices->devices;
2918 struct btrfs_device *device;
2919
2920 state = kzalloc(sizeof(*state), GFP_NOFS);
2921 if (NULL == state) {
2922 printk(KERN_INFO "btrfs check-integrity: kmalloc() failed!\n");
2923 return -1;
2924 }
2925
2926 if (!btrfsic_is_initialized) {
2927 mutex_init(&btrfsic_mutex);
2928 btrfsic_dev_state_hashtable_init(&btrfsic_dev_state_hashtable);
2929 btrfsic_is_initialized = 1;
2930 }
2931 mutex_lock(&btrfsic_mutex);
2932 state->root = root;
2933 state->print_mask = print_mask;
2934 state->include_extent_data = including_extent_data;
2935 state->csum_size = 0;
2936 INIT_LIST_HEAD(&state->all_blocks_list);
2937 btrfsic_block_hashtable_init(&state->block_hashtable);
2938 btrfsic_block_link_hashtable_init(&state->block_link_hashtable);
2939 state->max_superblock_generation = 0;
2940 state->latest_superblock = NULL;
2941
2942 list_for_each_entry(device, dev_head, dev_list) {
2943 struct btrfsic_dev_state *ds;
2944 char *p;
2945
2946 if (!device->bdev || !device->name)
2947 continue;
2948
2949 ds = btrfsic_dev_state_alloc();
2950 if (NULL == ds) {
2951 printk(KERN_INFO
2952 "btrfs check-integrity: kmalloc() failed!\n");
2953 mutex_unlock(&btrfsic_mutex);
2954 return -1;
2955 }
2956 ds->bdev = device->bdev;
2957 ds->state = state;
2958 bdevname(ds->bdev, ds->name);
2959 ds->name[BDEVNAME_SIZE - 1] = '\0';
2960 for (p = ds->name; *p != '\0'; p++);
2961 while (p > ds->name && *p != '/')
2962 p--;
2963 if (*p == '/')
2964 p++;
2965 strlcpy(ds->name, p, sizeof(ds->name));
2966 btrfsic_dev_state_hashtable_add(ds,
2967 &btrfsic_dev_state_hashtable);
2968 }
2969
2970 ret = btrfsic_process_superblock(state, fs_devices);
2971 if (0 != ret) {
2972 mutex_unlock(&btrfsic_mutex);
2973 btrfsic_unmount(root, fs_devices);
2974 return ret;
2975 }
2976
2977 if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_DATABASE)
2978 btrfsic_dump_database(state);
2979 if (state->print_mask & BTRFSIC_PRINT_MASK_INITIAL_TREE)
2980 btrfsic_dump_tree(state);
2981
2982 mutex_unlock(&btrfsic_mutex);
2983 return 0;
2984}
2985
2986void btrfsic_unmount(struct btrfs_root *root,
2987 struct btrfs_fs_devices *fs_devices)
2988{
2989 struct list_head *elem_all;
2990 struct list_head *tmp_all;
2991 struct btrfsic_state *state;
2992 struct list_head *dev_head = &fs_devices->devices;
2993 struct btrfs_device *device;
2994
2995 if (!btrfsic_is_initialized)
2996 return;
2997
2998 mutex_lock(&btrfsic_mutex);
2999
3000 state = NULL;
3001 list_for_each_entry(device, dev_head, dev_list) {
3002 struct btrfsic_dev_state *ds;
3003
3004 if (!device->bdev || !device->name)
3005 continue;
3006
3007 ds = btrfsic_dev_state_hashtable_lookup(
3008 device->bdev,
3009 &btrfsic_dev_state_hashtable);
3010 if (NULL != ds) {
3011 state = ds->state;
3012 btrfsic_dev_state_hashtable_remove(ds);
3013 btrfsic_dev_state_free(ds);
3014 }
3015 }
3016
3017 if (NULL == state) {
3018 printk(KERN_INFO
3019 "btrfsic: error, cannot find state information"
3020 " on umount!\n");
3021 mutex_unlock(&btrfsic_mutex);
3022 return;
3023 }
3024
3025 /*
3026 * Don't care about keeping the lists' state up to date,
3027 * just free all memory that was allocated dynamically.
3028 * Free the blocks and the block_links.
3029 */
3030 list_for_each_safe(elem_all, tmp_all, &state->all_blocks_list) {
3031 struct btrfsic_block *const b_all =
3032 list_entry(elem_all, struct btrfsic_block,
3033 all_blocks_node);
3034 struct list_head *elem_ref_to;
3035 struct list_head *tmp_ref_to;
3036
3037 list_for_each_safe(elem_ref_to, tmp_ref_to,
3038 &b_all->ref_to_list) {
3039 struct btrfsic_block_link *const l =
3040 list_entry(elem_ref_to,
3041 struct btrfsic_block_link,
3042 node_ref_to);
3043
3044 if (state->print_mask & BTRFSIC_PRINT_MASK_VERBOSE)
3045 btrfsic_print_rem_link(state, l);
3046
3047 l->ref_cnt--;
3048 if (0 == l->ref_cnt)
3049 btrfsic_block_link_free(l);
3050 }
3051
3052 if (b_all->is_iodone)
3053 btrfsic_block_free(b_all);
3054 else
3055 printk(KERN_INFO "btrfs: attempt to free %c-block"
3056 " @%llu (%s/%llu/%d) on umount which is"
3057 " not yet iodone!\n",
3058 btrfsic_get_block_type(state, b_all),
3059 (unsigned long long)b_all->logical_bytenr,
3060 b_all->dev_state->name,
3061 (unsigned long long)b_all->dev_bytenr,
3062 b_all->mirror_num);
3063 }
3064
3065 mutex_unlock(&btrfsic_mutex);
3066
3067 kfree(state);
3068}
diff --git a/fs/btrfs/check-integrity.h b/fs/btrfs/check-integrity.h
new file mode 100644
index 000000000000..8b59175cc502
--- /dev/null
+++ b/fs/btrfs/check-integrity.h
@@ -0,0 +1,36 @@
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#if !defined(__BTRFS_CHECK_INTEGRITY__)
20#define __BTRFS_CHECK_INTEGRITY__
21
22#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
23int btrfsic_submit_bh(int rw, struct buffer_head *bh);
24void btrfsic_submit_bio(int rw, struct bio *bio);
25#else
26#define btrfsic_submit_bh submit_bh
27#define btrfsic_submit_bio submit_bio
28#endif
29
30int btrfsic_mount(struct btrfs_root *root,
31 struct btrfs_fs_devices *fs_devices,
32 int including_extent_data, u32 print_mask);
33void btrfsic_unmount(struct btrfs_root *root,
34 struct btrfs_fs_devices *fs_devices);
35
36#endif
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index b6d1020c4571..3c2cbf7b6663 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -1041,7 +1041,7 @@ struct btrfs_fs_info {
1041 * is required instead of the faster short fsync log commits 1041 * is required instead of the faster short fsync log commits
1042 */ 1042 */
1043 u64 last_trans_log_full_commit; 1043 u64 last_trans_log_full_commit;
1044 unsigned long mount_opt:20; 1044 unsigned long mount_opt:21;
1045 unsigned long compress_type:4; 1045 unsigned long compress_type:4;
1046 u64 max_inline; 1046 u64 max_inline;
1047 u64 alloc_start; 1047 u64 alloc_start;
@@ -1236,6 +1236,10 @@ struct btrfs_fs_info {
1236 int scrub_workers_refcnt; 1236 int scrub_workers_refcnt;
1237 struct btrfs_workers scrub_workers; 1237 struct btrfs_workers scrub_workers;
1238 1238
1239#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1240 u32 check_integrity_print_mask;
1241#endif
1242
1239 /* filesystem state */ 1243 /* filesystem state */
1240 u64 fs_state; 1244 u64 fs_state;
1241 1245
@@ -1497,6 +1501,8 @@ struct btrfs_ioctl_defrag_range_args {
1497#define BTRFS_MOUNT_INODE_MAP_CACHE (1 << 17) 1501#define BTRFS_MOUNT_INODE_MAP_CACHE (1 << 17)
1498#define BTRFS_MOUNT_RECOVERY (1 << 18) 1502#define BTRFS_MOUNT_RECOVERY (1 << 18)
1499#define BTRFS_MOUNT_SKIP_BALANCE (1 << 19) 1503#define BTRFS_MOUNT_SKIP_BALANCE (1 << 19)
1504#define BTRFS_MOUNT_CHECK_INTEGRITY (1 << 20)
1505#define BTRFS_MOUNT_CHECK_INTEGRITY_INCLUDING_EXTENT_DATA (1 << 21)
1500 1506
1501#define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt) 1507#define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt)
1502#define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) 1508#define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt)
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 9be97716c5e0..da4457f84d78 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -43,6 +43,7 @@
43#include "tree-log.h" 43#include "tree-log.h"
44#include "free-space-cache.h" 44#include "free-space-cache.h"
45#include "inode-map.h" 45#include "inode-map.h"
46#include "check-integrity.h"
46 47
47static struct extent_io_ops btree_extent_io_ops; 48static struct extent_io_ops btree_extent_io_ops;
48static void end_workqueue_fn(struct btrfs_work *work); 49static void end_workqueue_fn(struct btrfs_work *work);
@@ -2002,6 +2003,9 @@ struct btrfs_root *open_ctree(struct super_block *sb,
2002 init_waitqueue_head(&fs_info->scrub_pause_wait); 2003 init_waitqueue_head(&fs_info->scrub_pause_wait);
2003 init_rwsem(&fs_info->scrub_super_lock); 2004 init_rwsem(&fs_info->scrub_super_lock);
2004 fs_info->scrub_workers_refcnt = 0; 2005 fs_info->scrub_workers_refcnt = 0;
2006#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2007 fs_info->check_integrity_print_mask = 0;
2008#endif
2005 2009
2006 spin_lock_init(&fs_info->balance_lock); 2010 spin_lock_init(&fs_info->balance_lock);
2007 mutex_init(&fs_info->balance_mutex); 2011 mutex_init(&fs_info->balance_mutex);
@@ -2360,6 +2364,19 @@ retry_root_backup:
2360 btrfs_set_opt(fs_info->mount_opt, SSD); 2364 btrfs_set_opt(fs_info->mount_opt, SSD);
2361 } 2365 }
2362 2366
2367#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2368 if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2369 ret = btrfsic_mount(tree_root, fs_devices,
2370 btrfs_test_opt(tree_root,
2371 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2372 1 : 0,
2373 fs_info->check_integrity_print_mask);
2374 if (ret)
2375 printk(KERN_WARNING "btrfs: failed to initialize"
2376 " integrity check module %s\n", sb->s_id);
2377 }
2378#endif
2379
2363 /* do not make disk changes in broken FS */ 2380 /* do not make disk changes in broken FS */
2364 if (btrfs_super_log_root(disk_super) != 0 && 2381 if (btrfs_super_log_root(disk_super) != 0 &&
2365 !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) { 2382 !(fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR)) {
@@ -2642,7 +2659,7 @@ static int write_dev_supers(struct btrfs_device *device,
2642 * we fua the first super. The others we allow 2659 * we fua the first super. The others we allow
2643 * to go down lazy. 2660 * to go down lazy.
2644 */ 2661 */
2645 ret = submit_bh(WRITE_FUA, bh); 2662 ret = btrfsic_submit_bh(WRITE_FUA, bh);
2646 if (ret) 2663 if (ret)
2647 errors++; 2664 errors++;
2648 } 2665 }
@@ -2719,7 +2736,7 @@ static int write_dev_flush(struct btrfs_device *device, int wait)
2719 device->flush_bio = bio; 2736 device->flush_bio = bio;
2720 2737
2721 bio_get(bio); 2738 bio_get(bio);
2722 submit_bio(WRITE_FLUSH, bio); 2739 btrfsic_submit_bio(WRITE_FLUSH, bio);
2723 2740
2724 return 0; 2741 return 0;
2725} 2742}
@@ -3068,6 +3085,11 @@ int close_ctree(struct btrfs_root *root)
3068 btrfs_stop_workers(&fs_info->caching_workers); 3085 btrfs_stop_workers(&fs_info->caching_workers);
3069 btrfs_stop_workers(&fs_info->readahead_workers); 3086 btrfs_stop_workers(&fs_info->readahead_workers);
3070 3087
3088#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3089 if (btrfs_test_opt(root, CHECK_INTEGRITY))
3090 btrfsic_unmount(root, fs_info->fs_devices);
3091#endif
3092
3071 btrfs_close_devices(fs_info->fs_devices); 3093 btrfs_close_devices(fs_info->fs_devices);
3072 btrfs_mapping_tree_free(&fs_info->mapping_tree); 3094 btrfs_mapping_tree_free(&fs_info->mapping_tree);
3073 3095
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 3622cc22ff91..9d09a4f81875 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -18,6 +18,7 @@
18#include "ctree.h" 18#include "ctree.h"
19#include "btrfs_inode.h" 19#include "btrfs_inode.h"
20#include "volumes.h" 20#include "volumes.h"
21#include "check-integrity.h"
21 22
22static struct kmem_cache *extent_state_cache; 23static struct kmem_cache *extent_state_cache;
23static struct kmem_cache *extent_buffer_cache; 24static struct kmem_cache *extent_buffer_cache;
@@ -1895,7 +1896,7 @@ int repair_io_failure(struct btrfs_mapping_tree *map_tree, u64 start,
1895 } 1896 }
1896 bio->bi_bdev = dev->bdev; 1897 bio->bi_bdev = dev->bdev;
1897 bio_add_page(bio, page, length, start-page_offset(page)); 1898 bio_add_page(bio, page, length, start-page_offset(page));
1898 submit_bio(WRITE_SYNC, bio); 1899 btrfsic_submit_bio(WRITE_SYNC, bio);
1899 wait_for_completion(&compl); 1900 wait_for_completion(&compl);
1900 1901
1901 if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) { 1902 if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) {
@@ -2393,7 +2394,7 @@ static int submit_one_bio(int rw, struct bio *bio, int mirror_num,
2393 ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio, 2394 ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
2394 mirror_num, bio_flags, start); 2395 mirror_num, bio_flags, start);
2395 else 2396 else
2396 submit_bio(rw, bio); 2397 btrfsic_submit_bio(rw, bio);
2397 2398
2398 if (bio_flagged(bio, BIO_EOPNOTSUPP)) 2399 if (bio_flagged(bio, BIO_EOPNOTSUPP))
2399 ret = -EOPNOTSUPP; 2400 ret = -EOPNOTSUPP;
diff --git a/fs/btrfs/scrub.c b/fs/btrfs/scrub.c
index 6a6a51a809ba..9770cc5bfb76 100644
--- a/fs/btrfs/scrub.c
+++ b/fs/btrfs/scrub.c
@@ -25,6 +25,7 @@
25#include "transaction.h" 25#include "transaction.h"
26#include "backref.h" 26#include "backref.h"
27#include "extent_io.h" 27#include "extent_io.h"
28#include "check-integrity.h"
28 29
29/* 30/*
30 * This is only the first step towards a full-features scrub. It reads all 31 * This is only the first step towards a full-features scrub. It reads all
@@ -733,7 +734,7 @@ static int scrub_fixup_io(int rw, struct block_device *bdev, sector_t sector,
733 bio_add_page(bio, page, PAGE_SIZE, 0); 734 bio_add_page(bio, page, PAGE_SIZE, 0);
734 bio->bi_end_io = scrub_fixup_end_io; 735 bio->bi_end_io = scrub_fixup_end_io;
735 bio->bi_private = &complete; 736 bio->bi_private = &complete;
736 submit_bio(rw, bio); 737 btrfsic_submit_bio(rw, bio);
737 738
738 /* this will also unplug the queue */ 739 /* this will also unplug the queue */
739 wait_for_completion(&complete); 740 wait_for_completion(&complete);
@@ -959,7 +960,7 @@ static int scrub_submit(struct scrub_dev *sdev)
959 sdev->curr = -1; 960 sdev->curr = -1;
960 atomic_inc(&sdev->in_flight); 961 atomic_inc(&sdev->in_flight);
961 962
962 submit_bio(READ, sbio->bio); 963 btrfsic_submit_bio(READ, sbio->bio);
963 964
964 return 0; 965 return 0;
965} 966}
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 5a7227fa9380..61717a4eb14f 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -166,6 +166,8 @@ enum {
166 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed, 166 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
167 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache, 167 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
168 Opt_no_space_cache, Opt_recovery, Opt_skip_balance, 168 Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
169 Opt_check_integrity, Opt_check_integrity_including_extent_data,
170 Opt_check_integrity_print_mask,
169 Opt_err, 171 Opt_err,
170}; 172};
171 173
@@ -202,6 +204,9 @@ static match_table_t tokens = {
202 {Opt_no_space_cache, "nospace_cache"}, 204 {Opt_no_space_cache, "nospace_cache"},
203 {Opt_recovery, "recovery"}, 205 {Opt_recovery, "recovery"},
204 {Opt_skip_balance, "skip_balance"}, 206 {Opt_skip_balance, "skip_balance"},
207 {Opt_check_integrity, "check_int"},
208 {Opt_check_integrity_including_extent_data, "check_int_data"},
209 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
205 {Opt_err, NULL}, 210 {Opt_err, NULL},
206}; 211};
207 212
@@ -403,6 +408,37 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
403 case Opt_skip_balance: 408 case Opt_skip_balance:
404 btrfs_set_opt(info->mount_opt, SKIP_BALANCE); 409 btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
405 break; 410 break;
411#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
412 case Opt_check_integrity_including_extent_data:
413 printk(KERN_INFO "btrfs: enabling check integrity"
414 " including extent data\n");
415 btrfs_set_opt(info->mount_opt,
416 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
417 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
418 break;
419 case Opt_check_integrity:
420 printk(KERN_INFO "btrfs: enabling check integrity\n");
421 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
422 break;
423 case Opt_check_integrity_print_mask:
424 intarg = 0;
425 match_int(&args[0], &intarg);
426 if (intarg) {
427 info->check_integrity_print_mask = intarg;
428 printk(KERN_INFO "btrfs:"
429 " check_integrity_print_mask 0x%x\n",
430 info->check_integrity_print_mask);
431 }
432 break;
433#else
434 case Opt_check_integrity_including_extent_data:
435 case Opt_check_integrity:
436 case Opt_check_integrity_print_mask:
437 printk(KERN_ERR "btrfs: support for check_integrity*"
438 " not compiled in!\n");
439 ret = -EINVAL;
440 goto out;
441#endif
406 case Opt_err: 442 case Opt_err:
407 printk(KERN_INFO "btrfs: unrecognized mount option " 443 printk(KERN_INFO "btrfs: unrecognized mount option "
408 "'%s'\n", p); 444 "'%s'\n", p);
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index e0b7bb92a170..59e878f9fdcc 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -33,6 +33,7 @@
33#include "print-tree.h" 33#include "print-tree.h"
34#include "volumes.h" 34#include "volumes.h"
35#include "async-thread.h" 35#include "async-thread.h"
36#include "check-integrity.h"
36 37
37static int init_first_rw_device(struct btrfs_trans_handle *trans, 38static int init_first_rw_device(struct btrfs_trans_handle *trans,
38 struct btrfs_root *root, 39 struct btrfs_root *root,
@@ -247,7 +248,7 @@ loop_lock:
247 sync_pending = 0; 248 sync_pending = 0;
248 } 249 }
249 250
250 submit_bio(cur->bi_rw, cur); 251 btrfsic_submit_bio(cur->bi_rw, cur);
251 num_run++; 252 num_run++;
252 batch_run++; 253 batch_run++;
253 if (need_resched()) 254 if (need_resched())
@@ -3962,7 +3963,7 @@ static noinline int schedule_bio(struct btrfs_root *root,
3962 /* don't bother with additional async steps for reads, right now */ 3963 /* don't bother with additional async steps for reads, right now */
3963 if (!(rw & REQ_WRITE)) { 3964 if (!(rw & REQ_WRITE)) {
3964 bio_get(bio); 3965 bio_get(bio);
3965 submit_bio(rw, bio); 3966 btrfsic_submit_bio(rw, bio);
3966 bio_put(bio); 3967 bio_put(bio);
3967 return 0; 3968 return 0;
3968 } 3969 }
@@ -4057,7 +4058,7 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
4057 if (async_submit) 4058 if (async_submit)
4058 schedule_bio(root, dev, rw, bio); 4059 schedule_bio(root, dev, rw, bio);
4059 else 4060 else
4060 submit_bio(rw, bio); 4061 btrfsic_submit_bio(rw, bio);
4061 } else { 4062 } else {
4062 bio->bi_bdev = root->fs_info->fs_devices->latest_bdev; 4063 bio->bi_bdev = root->fs_info->fs_devices->latest_bdev;
4063 bio->bi_sector = logical >> 9; 4064 bio->bi_sector = logical >> 9;
generated by cgit v1.2.2 (git 2.25.0) at 2024-12-17 17:24:42 -0500