Line data Source code
1 : /*
2 : * Copyright (C) 2007 Oracle. 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 : #include <linux/fs.h>
20 : #include <linux/blkdev.h>
21 : #include <linux/scatterlist.h>
22 : #include <linux/swap.h>
23 : #include <linux/radix-tree.h>
24 : #include <linux/writeback.h>
25 : #include <linux/buffer_head.h>
26 : #include <linux/workqueue.h>
27 : #include <linux/kthread.h>
28 : #include <linux/freezer.h>
29 : #include <linux/slab.h>
30 : #include <linux/migrate.h>
31 : #include <linux/ratelimit.h>
32 : #include <linux/uuid.h>
33 : #include <linux/semaphore.h>
34 : #include <asm/unaligned.h>
35 : #include "ctree.h"
36 : #include "disk-io.h"
37 : #include "hash.h"
38 : #include "transaction.h"
39 : #include "btrfs_inode.h"
40 : #include "volumes.h"
41 : #include "print-tree.h"
42 : #include "locking.h"
43 : #include "tree-log.h"
44 : #include "free-space-cache.h"
45 : #include "inode-map.h"
46 : #include "check-integrity.h"
47 : #include "rcu-string.h"
48 : #include "dev-replace.h"
49 : #include "raid56.h"
50 : #include "sysfs.h"
51 : #include "qgroup.h"
52 :
53 : #ifdef CONFIG_X86
54 : #include <asm/cpufeature.h>
55 : #endif
56 :
57 : static struct extent_io_ops btree_extent_io_ops;
58 : static void end_workqueue_fn(struct btrfs_work *work);
59 : static void free_fs_root(struct btrfs_root *root);
60 : static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
61 : int read_only);
62 : static void btrfs_destroy_ordered_extents(struct btrfs_root *root);
63 : static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
64 : struct btrfs_root *root);
65 : static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root);
66 : static int btrfs_destroy_marked_extents(struct btrfs_root *root,
67 : struct extent_io_tree *dirty_pages,
68 : int mark);
69 : static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
70 : struct extent_io_tree *pinned_extents);
71 : static int btrfs_cleanup_transaction(struct btrfs_root *root);
72 : static void btrfs_error_commit_super(struct btrfs_root *root);
73 :
74 : /*
75 : * end_io_wq structs are used to do processing in task context when an IO is
76 : * complete. This is used during reads to verify checksums, and it is used
77 : * by writes to insert metadata for new file extents after IO is complete.
78 : */
79 : struct end_io_wq {
80 : struct bio *bio;
81 : bio_end_io_t *end_io;
82 : void *private;
83 : struct btrfs_fs_info *info;
84 : int error;
85 : int metadata;
86 : struct list_head list;
87 : struct btrfs_work work;
88 : };
89 :
90 : /*
91 : * async submit bios are used to offload expensive checksumming
92 : * onto the worker threads. They checksum file and metadata bios
93 : * just before they are sent down the IO stack.
94 : */
95 : struct async_submit_bio {
96 : struct inode *inode;
97 : struct bio *bio;
98 : struct list_head list;
99 : extent_submit_bio_hook_t *submit_bio_start;
100 : extent_submit_bio_hook_t *submit_bio_done;
101 : int rw;
102 : int mirror_num;
103 : unsigned long bio_flags;
104 : /*
105 : * bio_offset is optional, can be used if the pages in the bio
106 : * can't tell us where in the file the bio should go
107 : */
108 : u64 bio_offset;
109 : struct btrfs_work work;
110 : int error;
111 : };
112 :
113 : /*
114 : * Lockdep class keys for extent_buffer->lock's in this root. For a given
115 : * eb, the lockdep key is determined by the btrfs_root it belongs to and
116 : * the level the eb occupies in the tree.
117 : *
118 : * Different roots are used for different purposes and may nest inside each
119 : * other and they require separate keysets. As lockdep keys should be
120 : * static, assign keysets according to the purpose of the root as indicated
121 : * by btrfs_root->objectid. This ensures that all special purpose roots
122 : * have separate keysets.
123 : *
124 : * Lock-nesting across peer nodes is always done with the immediate parent
125 : * node locked thus preventing deadlock. As lockdep doesn't know this, use
126 : * subclass to avoid triggering lockdep warning in such cases.
127 : *
128 : * The key is set by the readpage_end_io_hook after the buffer has passed
129 : * csum validation but before the pages are unlocked. It is also set by
130 : * btrfs_init_new_buffer on freshly allocated blocks.
131 : *
132 : * We also add a check to make sure the highest level of the tree is the
133 : * same as our lockdep setup here. If BTRFS_MAX_LEVEL changes, this code
134 : * needs update as well.
135 : */
136 : #ifdef CONFIG_DEBUG_LOCK_ALLOC
137 : # if BTRFS_MAX_LEVEL != 8
138 : # error
139 : # endif
140 :
141 : static struct btrfs_lockdep_keyset {
142 : u64 id; /* root objectid */
143 : const char *name_stem; /* lock name stem */
144 : char names[BTRFS_MAX_LEVEL + 1][20];
145 : struct lock_class_key keys[BTRFS_MAX_LEVEL + 1];
146 : } btrfs_lockdep_keysets[] = {
147 : { .id = BTRFS_ROOT_TREE_OBJECTID, .name_stem = "root" },
148 : { .id = BTRFS_EXTENT_TREE_OBJECTID, .name_stem = "extent" },
149 : { .id = BTRFS_CHUNK_TREE_OBJECTID, .name_stem = "chunk" },
150 : { .id = BTRFS_DEV_TREE_OBJECTID, .name_stem = "dev" },
151 : { .id = BTRFS_FS_TREE_OBJECTID, .name_stem = "fs" },
152 : { .id = BTRFS_CSUM_TREE_OBJECTID, .name_stem = "csum" },
153 : { .id = BTRFS_QUOTA_TREE_OBJECTID, .name_stem = "quota" },
154 : { .id = BTRFS_TREE_LOG_OBJECTID, .name_stem = "log" },
155 : { .id = BTRFS_TREE_RELOC_OBJECTID, .name_stem = "treloc" },
156 : { .id = BTRFS_DATA_RELOC_TREE_OBJECTID, .name_stem = "dreloc" },
157 : { .id = BTRFS_UUID_TREE_OBJECTID, .name_stem = "uuid" },
158 : { .id = 0, .name_stem = "tree" },
159 : };
160 :
161 : void __init btrfs_init_lockdep(void)
162 : {
163 : int i, j;
164 :
165 : /* initialize lockdep class names */
166 : for (i = 0; i < ARRAY_SIZE(btrfs_lockdep_keysets); i++) {
167 : struct btrfs_lockdep_keyset *ks = &btrfs_lockdep_keysets[i];
168 :
169 : for (j = 0; j < ARRAY_SIZE(ks->names); j++)
170 : snprintf(ks->names[j], sizeof(ks->names[j]),
171 : "btrfs-%s-%02d", ks->name_stem, j);
172 : }
173 : }
174 :
175 : void btrfs_set_buffer_lockdep_class(u64 objectid, struct extent_buffer *eb,
176 : int level)
177 : {
178 : struct btrfs_lockdep_keyset *ks;
179 :
180 : BUG_ON(level >= ARRAY_SIZE(ks->keys));
181 :
182 : /* find the matching keyset, id 0 is the default entry */
183 : for (ks = btrfs_lockdep_keysets; ks->id; ks++)
184 : if (ks->id == objectid)
185 : break;
186 :
187 : lockdep_set_class_and_name(&eb->lock,
188 : &ks->keys[level], ks->names[level]);
189 : }
190 :
191 : #endif
192 :
193 : /*
194 : * extents on the btree inode are pretty simple, there's one extent
195 : * that covers the entire device
196 : */
197 8277 : static struct extent_map *btree_get_extent(struct inode *inode,
198 : struct page *page, size_t pg_offset, u64 start, u64 len,
199 : int create)
200 : {
201 8277 : struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
202 : struct extent_map *em;
203 : int ret;
204 :
205 8277 : read_lock(&em_tree->lock);
206 8277 : em = lookup_extent_mapping(em_tree, start, len);
207 8277 : if (em) {
208 8056 : em->bdev =
209 8056 : BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
210 : read_unlock(&em_tree->lock);
211 : goto out;
212 : }
213 : read_unlock(&em_tree->lock);
214 :
215 221 : em = alloc_extent_map();
216 221 : if (!em) {
217 : em = ERR_PTR(-ENOMEM);
218 : goto out;
219 : }
220 221 : em->start = 0;
221 221 : em->len = (u64)-1;
222 221 : em->block_len = (u64)-1;
223 221 : em->block_start = 0;
224 221 : em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
225 :
226 221 : write_lock(&em_tree->lock);
227 221 : ret = add_extent_mapping(em_tree, em, 0);
228 221 : if (ret == -EEXIST) {
229 0 : free_extent_map(em);
230 0 : em = lookup_extent_mapping(em_tree, start, len);
231 0 : if (!em)
232 : em = ERR_PTR(-EIO);
233 221 : } else if (ret) {
234 0 : free_extent_map(em);
235 0 : em = ERR_PTR(ret);
236 : }
237 : write_unlock(&em_tree->lock);
238 :
239 : out:
240 8277 : return em;
241 : }
242 :
243 5553329 : u32 btrfs_csum_data(char *data, u32 seed, size_t len)
244 : {
245 5744709 : return btrfs_crc32c(seed, data, len);
246 : }
247 :
248 5534406 : void btrfs_csum_final(u32 crc, char *result)
249 : {
250 5599672 : put_unaligned_le32(~crc, result);
251 5534406 : }
252 :
253 : /*
254 : * compute the csum for a btree block, and either verify it or write it
255 : * into the csum field of the block.
256 : */
257 59210 : static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
258 : int verify)
259 : {
260 118420 : u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
261 : char *result = NULL;
262 : unsigned long len;
263 : unsigned long cur_len;
264 : unsigned long offset = BTRFS_CSUM_SIZE;
265 : char *kaddr;
266 : unsigned long map_start;
267 : unsigned long map_len;
268 : int err;
269 : u32 crc = ~(u32)0;
270 : unsigned long inline_result;
271 :
272 59210 : len = buf->len - offset;
273 244534 : while (len > 0) {
274 185324 : err = map_private_extent_buffer(buf, offset, 32,
275 : &kaddr, &map_start, &map_len);
276 185324 : if (err)
277 : return 1;
278 185324 : cur_len = min(len, map_len - (offset - map_start));
279 185324 : crc = btrfs_csum_data(kaddr + offset - map_start,
280 : crc, cur_len);
281 185324 : len -= cur_len;
282 185324 : offset += cur_len;
283 : }
284 59210 : if (csum_size > sizeof(inline_result)) {
285 0 : result = kzalloc(csum_size * sizeof(char), GFP_NOFS);
286 0 : if (!result)
287 : return 1;
288 : } else {
289 : result = (char *)&inline_result;
290 : }
291 :
292 : btrfs_csum_final(crc, result);
293 :
294 59210 : if (verify) {
295 2184 : if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
296 : u32 val;
297 0 : u32 found = 0;
298 0 : memcpy(&found, result, csum_size);
299 :
300 0 : read_extent_buffer(buf, &val, 0, csum_size);
301 0 : printk_ratelimited(KERN_INFO
302 : "BTRFS: %s checksum verify failed on %llu wanted %X found %X "
303 : "level %d\n",
304 : root->fs_info->sb->s_id, buf->start,
305 : val, found, btrfs_header_level(buf));
306 0 : if (result != (char *)&inline_result)
307 0 : kfree(result);
308 : return 1;
309 : }
310 : } else {
311 57026 : write_extent_buffer(buf, result, 0, csum_size);
312 : }
313 59210 : if (result != (char *)&inline_result)
314 0 : kfree(result);
315 : return 0;
316 : }
317 :
318 : /*
319 : * we can't consider a given block up to date unless the transid of the
320 : * block matches the transid in the parent node's pointer. This is how we
321 : * detect blocks that either didn't get written at all or got written
322 : * in the wrong place.
323 : */
324 3420814 : static int verify_parent_transid(struct extent_io_tree *io_tree,
325 2252182 : struct extent_buffer *eb, u64 parent_transid,
326 : int atomic)
327 : {
328 3420814 : struct extent_state *cached_state = NULL;
329 : int ret;
330 3420814 : bool need_lock = (current->journal_info ==
331 : (void *)BTRFS_SEND_TRANS_STUB);
332 :
333 5672996 : if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
334 : return 0;
335 :
336 0 : if (atomic)
337 : return -EAGAIN;
338 :
339 0 : if (need_lock) {
340 0 : btrfs_tree_read_lock(eb);
341 0 : btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
342 : }
343 :
344 0 : lock_extent_bits(io_tree, eb->start, eb->start + eb->len - 1,
345 : 0, &cached_state);
346 0 : if (extent_buffer_uptodate(eb) &&
347 : btrfs_header_generation(eb) == parent_transid) {
348 : ret = 0;
349 : goto out;
350 : }
351 0 : printk_ratelimited("parent transid verify failed on %llu wanted %llu "
352 : "found %llu\n",
353 : eb->start, parent_transid, btrfs_header_generation(eb));
354 : ret = 1;
355 :
356 : /*
357 : * Things reading via commit roots that don't have normal protection,
358 : * like send, can have a really old block in cache that may point at a
359 : * block that has been free'd and re-allocated. So don't clear uptodate
360 : * if we find an eb that is under IO (dirty/writeback) because we could
361 : * end up reading in the stale data and then writing it back out and
362 : * making everybody very sad.
363 : */
364 0 : if (!extent_buffer_under_io(eb))
365 0 : clear_extent_buffer_uptodate(eb);
366 : out:
367 0 : unlock_extent_cached(io_tree, eb->start, eb->start + eb->len - 1,
368 : &cached_state, GFP_NOFS);
369 0 : if (need_lock)
370 0 : btrfs_tree_read_unlock_blocking(eb);
371 0 : return ret;
372 : }
373 :
374 : /*
375 : * Return 0 if the superblock checksum type matches the checksum value of that
376 : * algorithm. Pass the raw disk superblock data.
377 : */
378 221 : static int btrfs_check_super_csum(char *raw_disk_sb)
379 : {
380 : struct btrfs_super_block *disk_sb =
381 : (struct btrfs_super_block *)raw_disk_sb;
382 : u16 csum_type = btrfs_super_csum_type(disk_sb);
383 : int ret = 0;
384 :
385 221 : if (csum_type == BTRFS_CSUM_TYPE_CRC32) {
386 : u32 crc = ~(u32)0;
387 : const int csum_size = sizeof(crc);
388 : char result[csum_size];
389 :
390 : /*
391 : * The super_block structure does not span the whole
392 : * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space
393 : * is filled with zeros and is included in the checkum.
394 : */
395 221 : crc = btrfs_csum_data(raw_disk_sb + BTRFS_CSUM_SIZE,
396 : crc, BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
397 : btrfs_csum_final(crc, result);
398 :
399 221 : if (memcmp(raw_disk_sb, result, csum_size))
400 : ret = 1;
401 :
402 221 : if (ret && btrfs_super_generation(disk_sb) < 10) {
403 0 : printk(KERN_WARNING
404 : "BTRFS: super block crcs don't match, older mkfs detected\n");
405 : ret = 0;
406 : }
407 : }
408 :
409 221 : if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
410 0 : printk(KERN_ERR "BTRFS: unsupported checksum algorithm %u\n",
411 : csum_type);
412 : ret = 1;
413 : }
414 :
415 221 : return ret;
416 : }
417 :
418 : /*
419 : * helper to read a given tree block, doing retries as required when
420 : * the checksums don't match and we have alternate mirrors to try.
421 : */
422 1193259 : static int btree_read_extent_buffer_pages(struct btrfs_root *root,
423 : struct extent_buffer *eb,
424 : u64 start, u64 parent_transid)
425 : {
426 : struct extent_io_tree *io_tree;
427 : int failed = 0;
428 : int ret;
429 : int num_copies = 0;
430 : int mirror_num = 0;
431 : int failed_mirror = 0;
432 :
433 : clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
434 1193260 : io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
435 : while (1) {
436 1193260 : ret = read_extent_buffer_pages(io_tree, eb, start,
437 : WAIT_COMPLETE,
438 : btree_get_extent, mirror_num);
439 1193260 : if (!ret) {
440 1193260 : if (!verify_parent_transid(io_tree, eb,
441 : parent_transid, 0))
442 : break;
443 : else
444 : ret = -EIO;
445 : }
446 :
447 : /*
448 : * This buffer's crc is fine, but its contents are corrupted, so
449 : * there is no reason to read the other copies, they won't be
450 : * any less wrong.
451 : */
452 0 : if (test_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags))
453 : break;
454 :
455 0 : num_copies = btrfs_num_copies(root->fs_info,
456 0 : eb->start, eb->len);
457 0 : if (num_copies == 1)
458 : break;
459 :
460 0 : if (!failed_mirror) {
461 : failed = 1;
462 0 : failed_mirror = eb->read_mirror;
463 : }
464 :
465 0 : mirror_num++;
466 0 : if (mirror_num == failed_mirror)
467 0 : mirror_num++;
468 :
469 0 : if (mirror_num > num_copies)
470 : break;
471 : }
472 :
473 1193260 : if (failed && !ret && failed_mirror)
474 0 : repair_eb_io_failure(root, eb, failed_mirror);
475 :
476 1193260 : return ret;
477 : }
478 :
479 : /*
480 : * checksum a dirty tree block before IO. This has extra checks to make sure
481 : * we only fill in the checksum field in the first page of a multi-page block
482 : */
483 :
484 177047 : static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
485 : {
486 177047 : u64 start = page_offset(page);
487 : u64 found_start;
488 : struct extent_buffer *eb;
489 :
490 177047 : eb = (struct extent_buffer *)page->private;
491 177047 : if (page != eb->pages[0])
492 : return 0;
493 : found_start = btrfs_header_bytenr(eb);
494 114052 : if (WARN_ON(found_start != start || !PageUptodate(page)))
495 : return 0;
496 57026 : csum_tree_block(root, eb, 0);
497 57026 : return 0;
498 : }
499 :
500 2184 : static int check_tree_block_fsid(struct btrfs_root *root,
501 : struct extent_buffer *eb)
502 : {
503 2184 : struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
504 : u8 fsid[BTRFS_UUID_SIZE];
505 : int ret = 1;
506 :
507 2184 : read_extent_buffer(eb, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
508 2184 : while (fs_devices) {
509 2184 : if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) {
510 : ret = 0;
511 : break;
512 : }
513 0 : fs_devices = fs_devices->seed;
514 : }
515 2184 : return ret;
516 : }
517 :
518 : #define CORRUPT(reason, eb, root, slot) \
519 : btrfs_crit(root->fs_info, "corrupt leaf, %s: block=%llu," \
520 : "root=%llu, slot=%d", reason, \
521 : btrfs_header_bytenr(eb), root->objectid, slot)
522 :
523 2111 : static noinline int check_leaf(struct btrfs_root *root,
524 2111 : struct extent_buffer *leaf)
525 : {
526 : struct btrfs_key key;
527 : struct btrfs_key leaf_key;
528 : u32 nritems = btrfs_header_nritems(leaf);
529 : int slot;
530 :
531 2111 : if (nritems == 0)
532 : return 0;
533 :
534 : /* Check the 0 item */
535 3810 : if (btrfs_item_offset_nr(leaf, 0) + btrfs_item_size_nr(leaf, 0) !=
536 1905 : BTRFS_LEAF_DATA_SIZE(root)) {
537 0 : CORRUPT("invalid item offset size pair", leaf, root, 0);
538 0 : return -EIO;
539 : }
540 :
541 : /*
542 : * Check to make sure each items keys are in the correct order and their
543 : * offsets make sense. We only have to loop through nritems-1 because
544 : * we check the current slot against the next slot, which verifies the
545 : * next slot's offset+size makes sense and that the current's slot
546 : * offset is correct.
547 : */
548 46295 : for (slot = 0; slot < nritems - 1; slot++) {
549 44394 : btrfs_item_key_to_cpu(leaf, &leaf_key, slot);
550 44429 : btrfs_item_key_to_cpu(leaf, &key, slot + 1);
551 :
552 : /* Make sure the keys are in the right order */
553 44490 : if (btrfs_comp_cpu_keys(&leaf_key, &key) >= 0) {
554 0 : CORRUPT("bad key order", leaf, root, slot);
555 0 : return -EIO;
556 : }
557 :
558 : /*
559 : * Make sure the offset and ends are right, remember that the
560 : * item data starts at the end of the leaf and grows towards the
561 : * front.
562 : */
563 44412 : if (btrfs_item_offset_nr(leaf, slot) !=
564 : btrfs_item_end_nr(leaf, slot + 1)) {
565 0 : CORRUPT("slot offset bad", leaf, root, slot);
566 0 : return -EIO;
567 : }
568 :
569 : /*
570 : * Check to make sure that we don't point outside of the leaf,
571 : * just incase all the items are consistent to eachother, but
572 : * all point outside of the leaf.
573 : */
574 88780 : if (btrfs_item_end_nr(leaf, slot) >
575 44390 : BTRFS_LEAF_DATA_SIZE(root)) {
576 0 : CORRUPT("slot end outside of leaf", leaf, root, slot);
577 0 : return -EIO;
578 : }
579 : }
580 :
581 : return 0;
582 : }
583 :
584 8277 : static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
585 : u64 phy_offset, struct page *page,
586 : u64 start, u64 end, int mirror)
587 : {
588 : u64 found_start;
589 : int found_level;
590 4368 : struct extent_buffer *eb;
591 10461 : struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
592 : int ret = 0;
593 : int reads_done;
594 :
595 8277 : if (!page->private)
596 : goto out;
597 :
598 8277 : eb = (struct extent_buffer *)page->private;
599 :
600 : /* the pending IO might have been the only thing that kept this buffer
601 : * in memory. Make sure we have a ref for all this other checks
602 : */
603 : extent_buffer_get(eb);
604 :
605 8277 : reads_done = atomic_dec_and_test(&eb->io_pages);
606 8276 : if (!reads_done)
607 : goto err;
608 :
609 2184 : eb->read_mirror = mirror;
610 2184 : if (test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
611 : ret = -EIO;
612 : goto err;
613 : }
614 :
615 : found_start = btrfs_header_bytenr(eb);
616 2184 : if (found_start != eb->start) {
617 0 : printk_ratelimited(KERN_INFO "BTRFS: bad tree block start "
618 : "%llu %llu\n",
619 : found_start, eb->start);
620 : ret = -EIO;
621 : goto err;
622 : }
623 2184 : if (check_tree_block_fsid(root, eb)) {
624 0 : printk_ratelimited(KERN_INFO "BTRFS: bad fsid on block %llu\n",
625 : eb->start);
626 : ret = -EIO;
627 : goto err;
628 : }
629 2184 : found_level = btrfs_header_level(eb);
630 2184 : if (found_level >= BTRFS_MAX_LEVEL) {
631 0 : btrfs_info(root->fs_info, "bad tree block level %d",
632 : (int)btrfs_header_level(eb));
633 : ret = -EIO;
634 0 : goto err;
635 : }
636 :
637 : btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb),
638 : eb, found_level);
639 :
640 2184 : ret = csum_tree_block(root, eb, 1);
641 2184 : if (ret) {
642 : ret = -EIO;
643 : goto err;
644 : }
645 :
646 : /*
647 : * If this is a leaf block and it is corrupt, set the corrupt bit so
648 : * that we don't try and read the other copies of this block, just
649 : * return -EIO.
650 : */
651 2184 : if (found_level == 0 && check_leaf(root, eb)) {
652 : set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
653 : ret = -EIO;
654 : }
655 :
656 2184 : if (!ret)
657 2184 : set_extent_buffer_uptodate(eb);
658 : err:
659 10460 : if (reads_done &&
660 2184 : test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
661 144 : btree_readahead_hook(root, eb, eb->start, ret);
662 :
663 8276 : if (ret) {
664 : /*
665 : * our io error hook is going to dec the io pages
666 : * again, we have to make sure it has something
667 : * to decrement
668 : */
669 : atomic_inc(&eb->io_pages);
670 0 : clear_extent_buffer_uptodate(eb);
671 : }
672 8276 : free_extent_buffer(eb);
673 : out:
674 8277 : return ret;
675 : }
676 :
677 0 : static int btree_io_failed_hook(struct page *page, int failed_mirror)
678 : {
679 : struct extent_buffer *eb;
680 0 : struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
681 :
682 0 : eb = (struct extent_buffer *)page->private;
683 : set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
684 0 : eb->read_mirror = failed_mirror;
685 0 : atomic_dec(&eb->io_pages);
686 0 : if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
687 0 : btree_readahead_hook(root, eb, eb->start, -EIO);
688 0 : return -EIO; /* we fixed nothing */
689 : }
690 :
691 28785 : static void end_workqueue_bio(struct bio *bio, int err)
692 : {
693 28785 : struct end_io_wq *end_io_wq = bio->bi_private;
694 : struct btrfs_fs_info *fs_info;
695 : struct btrfs_workqueue *wq;
696 : btrfs_work_func_t func;
697 :
698 28785 : fs_info = end_io_wq->info;
699 28785 : end_io_wq->error = err;
700 :
701 28785 : if (bio->bi_rw & REQ_WRITE) {
702 153 : if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA) {
703 0 : wq = fs_info->endio_meta_write_workers;
704 : func = btrfs_endio_meta_write_helper;
705 153 : } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE) {
706 0 : wq = fs_info->endio_freespace_worker;
707 : func = btrfs_freespace_write_helper;
708 153 : } else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
709 0 : wq = fs_info->endio_raid56_workers;
710 : func = btrfs_endio_raid56_helper;
711 : } else {
712 153 : wq = fs_info->endio_write_workers;
713 : func = btrfs_endio_write_helper;
714 : }
715 : } else {
716 28632 : if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56) {
717 10 : wq = fs_info->endio_raid56_workers;
718 : func = btrfs_endio_raid56_helper;
719 28622 : } else if (end_io_wq->metadata) {
720 2441 : wq = fs_info->endio_meta_workers;
721 : func = btrfs_endio_meta_helper;
722 : } else {
723 26181 : wq = fs_info->endio_workers;
724 : func = btrfs_endio_helper;
725 : }
726 : }
727 :
728 28785 : btrfs_init_work(&end_io_wq->work, func, end_workqueue_fn, NULL, NULL);
729 28785 : btrfs_queue_work(wq, &end_io_wq->work);
730 28785 : }
731 :
732 : /*
733 : * For the metadata arg you want
734 : *
735 : * 0 - if data
736 : * 1 - if normal metadta
737 : * 2 - if writing to the free space cache area
738 : * 3 - raid parity work
739 : */
740 28784 : int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
741 : int metadata)
742 : {
743 : struct end_io_wq *end_io_wq;
744 : end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
745 28785 : if (!end_io_wq)
746 : return -ENOMEM;
747 :
748 28785 : end_io_wq->private = bio->bi_private;
749 28785 : end_io_wq->end_io = bio->bi_end_io;
750 28785 : end_io_wq->info = info;
751 28785 : end_io_wq->error = 0;
752 28785 : end_io_wq->bio = bio;
753 28785 : end_io_wq->metadata = metadata;
754 :
755 28785 : bio->bi_private = end_io_wq;
756 28785 : bio->bi_end_io = end_workqueue_bio;
757 28785 : return 0;
758 : }
759 :
760 7328 : unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info)
761 : {
762 66875 : unsigned long limit = min_t(unsigned long,
763 : info->thread_pool_size,
764 : info->fs_devices->open_devices);
765 66875 : return 256 * limit;
766 : }
767 :
768 59546 : static void run_one_async_start(struct btrfs_work *work)
769 : {
770 : struct async_submit_bio *async;
771 : int ret;
772 :
773 : async = container_of(work, struct async_submit_bio, work);
774 59546 : ret = async->submit_bio_start(async->inode, async->rw, async->bio,
775 : async->mirror_num, async->bio_flags,
776 : async->bio_offset);
777 59528 : if (ret)
778 0 : async->error = ret;
779 59528 : }
780 :
781 59547 : static void run_one_async_done(struct btrfs_work *work)
782 : {
783 : struct btrfs_fs_info *fs_info;
784 : struct async_submit_bio *async;
785 : int limit;
786 :
787 : async = container_of(work, struct async_submit_bio, work);
788 59547 : fs_info = BTRFS_I(async->inode)->root->fs_info;
789 :
790 59547 : limit = btrfs_async_submit_limit(fs_info);
791 59547 : limit = limit * 2 / 3;
792 :
793 156222 : if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
794 : waitqueue_active(&fs_info->async_submit_wait))
795 470 : wake_up(&fs_info->async_submit_wait);
796 :
797 : /* If an error occured we just want to clean up the bio and move on */
798 59547 : if (async->error) {
799 0 : bio_endio(async->bio, async->error);
800 59547 : return;
801 : }
802 :
803 59547 : async->submit_bio_done(async->inode, async->rw, async->bio,
804 : async->mirror_num, async->bio_flags,
805 : async->bio_offset);
806 : }
807 :
808 59547 : static void run_one_async_free(struct btrfs_work *work)
809 : {
810 : struct async_submit_bio *async;
811 :
812 59547 : async = container_of(work, struct async_submit_bio, work);
813 59547 : kfree(async);
814 59547 : }
815 :
816 59547 : int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
817 : int rw, struct bio *bio, int mirror_num,
818 : unsigned long bio_flags,
819 : u64 bio_offset,
820 : extent_submit_bio_hook_t *submit_bio_start,
821 : extent_submit_bio_hook_t *submit_bio_done)
822 : {
823 : struct async_submit_bio *async;
824 :
825 : async = kmalloc(sizeof(*async), GFP_NOFS);
826 59546 : if (!async)
827 : return -ENOMEM;
828 :
829 59546 : async->inode = inode;
830 59546 : async->rw = rw;
831 59546 : async->bio = bio;
832 59546 : async->mirror_num = mirror_num;
833 59546 : async->submit_bio_start = submit_bio_start;
834 59546 : async->submit_bio_done = submit_bio_done;
835 :
836 59546 : btrfs_init_work(&async->work, btrfs_worker_helper, run_one_async_start,
837 : run_one_async_done, run_one_async_free);
838 :
839 59544 : async->bio_flags = bio_flags;
840 59544 : async->bio_offset = bio_offset;
841 :
842 59544 : async->error = 0;
843 :
844 59544 : atomic_inc(&fs_info->nr_async_submits);
845 :
846 59547 : if (rw & REQ_SYNC)
847 608 : btrfs_set_work_high_priority(&async->work);
848 :
849 59547 : btrfs_queue_work(fs_info->workers, &async->work);
850 :
851 119094 : while (atomic_read(&fs_info->async_submit_draining) &&
852 : atomic_read(&fs_info->nr_async_submits)) {
853 0 : wait_event(fs_info->async_submit_wait,
854 : (atomic_read(&fs_info->nr_async_submits) == 0));
855 : }
856 :
857 : return 0;
858 : }
859 :
860 27921 : static int btree_csum_one_bio(struct bio *bio)
861 : {
862 : struct bio_vec *bvec;
863 : struct btrfs_root *root;
864 : int i, ret = 0;
865 :
866 204968 : bio_for_each_segment_all(bvec, bio, i) {
867 177047 : root = BTRFS_I(bvec->bv_page->mapping->host)->root;
868 177047 : ret = csum_dirty_buffer(root, bvec->bv_page);
869 177047 : if (ret)
870 : break;
871 : }
872 :
873 27921 : return ret;
874 : }
875 :
876 0 : static int __btree_submit_bio_start(struct inode *inode, int rw,
877 0 : struct bio *bio, int mirror_num,
878 : unsigned long bio_flags,
879 : u64 bio_offset)
880 : {
881 : /*
882 : * when we're called for a write, we're already in the async
883 : * submission context. Just jump into btrfs_map_bio
884 : */
885 0 : return btree_csum_one_bio(bio);
886 : }
887 :
888 0 : static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
889 : int mirror_num, unsigned long bio_flags,
890 : u64 bio_offset)
891 : {
892 : int ret;
893 :
894 : /*
895 : * when we're called for a write, we're already in the async
896 : * submission context. Just jump into btrfs_map_bio
897 : */
898 0 : ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1);
899 0 : if (ret)
900 0 : bio_endio(bio, ret);
901 0 : return ret;
902 : }
903 :
904 : static int check_async_write(struct inode *inode, unsigned long bio_flags)
905 : {
906 30105 : if (bio_flags & EXTENT_BIO_TREE_LOG)
907 : return 0;
908 : #ifdef CONFIG_X86
909 27074 : if (cpu_has_xmm4_2)
910 : return 0;
911 : #endif
912 : return 1;
913 : }
914 :
915 58026 : static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
916 : int mirror_num, unsigned long bio_flags,
917 : u64 bio_offset)
918 : {
919 : int async = check_async_write(inode, bio_flags);
920 : int ret;
921 :
922 30105 : if (!(rw & REQ_WRITE)) {
923 : /*
924 : * called for a read, do the setup so that checksum validation
925 : * can happen in the async kernel threads
926 : */
927 2184 : ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info,
928 : bio, 1);
929 2184 : if (ret)
930 : goto out_w_error;
931 2184 : ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
932 : mirror_num, 0);
933 27921 : } else if (!async) {
934 27921 : ret = btree_csum_one_bio(bio);
935 27921 : if (ret)
936 : goto out_w_error;
937 27921 : ret = btrfs_map_bio(BTRFS_I(inode)->root, rw, bio,
938 : mirror_num, 0);
939 : } else {
940 : /*
941 : * kthread helpers are used to submit writes so that
942 : * checksumming can happen in parallel across all CPUs
943 : */
944 0 : ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
945 : inode, rw, bio, mirror_num, 0,
946 : bio_offset,
947 : __btree_submit_bio_start,
948 : __btree_submit_bio_done);
949 : }
950 :
951 30105 : if (ret) {
952 : out_w_error:
953 0 : bio_endio(bio, ret);
954 : }
955 30105 : return ret;
956 : }
957 :
958 : #ifdef CONFIG_MIGRATION
959 0 : static int btree_migratepage(struct address_space *mapping,
960 0 : struct page *newpage, struct page *page,
961 : enum migrate_mode mode)
962 : {
963 : /*
964 : * we can't safely write a btree page from here,
965 : * we haven't done the locking hook
966 : */
967 0 : if (PageDirty(page))
968 : return -EAGAIN;
969 : /*
970 : * Buffers may be managed in a filesystem specific way.
971 : * We must have no buffers or drop them.
972 : */
973 0 : if (page_has_private(page) &&
974 0 : !try_to_release_page(page, GFP_KERNEL))
975 : return -EAGAIN;
976 0 : return migrate_page(mapping, newpage, page, mode);
977 : }
978 : #endif
979 :
980 :
981 25727 : static int btree_writepages(struct address_space *mapping,
982 : struct writeback_control *wbc)
983 : {
984 : struct btrfs_fs_info *fs_info;
985 : int ret;
986 :
987 25727 : if (wbc->sync_mode == WB_SYNC_NONE) {
988 :
989 4495 : if (wbc->for_kupdate)
990 : return 0;
991 :
992 4478 : fs_info = BTRFS_I(mapping->host)->root->fs_info;
993 : /* this is a bit racy, but that's ok */
994 4478 : ret = percpu_counter_compare(&fs_info->dirty_metadata_bytes,
995 : BTRFS_DIRTY_METADATA_THRESH);
996 4478 : if (ret < 0)
997 : return 0;
998 : }
999 21232 : return btree_write_cache_pages(mapping, wbc);
1000 : }
1001 :
1002 0 : static int btree_readpage(struct file *file, struct page *page)
1003 : {
1004 : struct extent_io_tree *tree;
1005 0 : tree = &BTRFS_I(page->mapping->host)->io_tree;
1006 0 : return extent_read_full_page(tree, page, btree_get_extent, 0);
1007 : }
1008 :
1009 13164 : static int btree_releasepage(struct page *page, gfp_t gfp_flags)
1010 : {
1011 26328 : if (PageWriteback(page) || PageDirty(page))
1012 : return 0;
1013 :
1014 13164 : return try_release_extent_buffer(page);
1015 : }
1016 :
1017 1335 : static void btree_invalidatepage(struct page *page, unsigned int offset,
1018 : unsigned int length)
1019 : {
1020 : struct extent_io_tree *tree;
1021 1335 : tree = &BTRFS_I(page->mapping->host)->io_tree;
1022 1335 : extent_invalidatepage(tree, page, offset);
1023 1335 : btree_releasepage(page, GFP_NOFS);
1024 1335 : if (PagePrivate(page)) {
1025 0 : btrfs_warn(BTRFS_I(page->mapping->host)->root->fs_info,
1026 : "page private not zero on page %llu",
1027 : (unsigned long long)page_offset(page));
1028 : ClearPagePrivate(page);
1029 0 : set_page_private(page, 0);
1030 0 : page_cache_release(page);
1031 : }
1032 1335 : }
1033 :
1034 4015739 : static int btree_set_page_dirty(struct page *page)
1035 : {
1036 : #ifdef DEBUG
1037 : struct extent_buffer *eb;
1038 :
1039 : BUG_ON(!PagePrivate(page));
1040 : eb = (struct extent_buffer *)page->private;
1041 : BUG_ON(!eb);
1042 : BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
1043 : BUG_ON(!atomic_read(&eb->refs));
1044 : btrfs_assert_tree_locked(eb);
1045 : #endif
1046 4015739 : return __set_page_dirty_nobuffers(page);
1047 : }
1048 :
1049 : static const struct address_space_operations btree_aops = {
1050 : .readpage = btree_readpage,
1051 : .writepages = btree_writepages,
1052 : .releasepage = btree_releasepage,
1053 : .invalidatepage = btree_invalidatepage,
1054 : #ifdef CONFIG_MIGRATION
1055 : .migratepage = btree_migratepage,
1056 : #endif
1057 : .set_page_dirty = btree_set_page_dirty,
1058 : };
1059 :
1060 646 : int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1061 : u64 parent_transid)
1062 : {
1063 : struct extent_buffer *buf = NULL;
1064 646 : struct inode *btree_inode = root->fs_info->btree_inode;
1065 : int ret = 0;
1066 :
1067 : buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1068 646 : if (!buf)
1069 : return 0;
1070 646 : read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
1071 : buf, 0, WAIT_NONE, btree_get_extent, 0);
1072 646 : free_extent_buffer(buf);
1073 646 : return ret;
1074 : }
1075 :
1076 15891 : int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
1077 : int mirror_num, struct extent_buffer **eb)
1078 : {
1079 : struct extent_buffer *buf = NULL;
1080 15891 : struct inode *btree_inode = root->fs_info->btree_inode;
1081 15891 : struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
1082 : int ret;
1083 :
1084 : buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1085 15887 : if (!buf)
1086 : return 0;
1087 :
1088 : set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
1089 :
1090 15890 : ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
1091 : btree_get_extent, mirror_num);
1092 15890 : if (ret) {
1093 0 : free_extent_buffer(buf);
1094 0 : return ret;
1095 : }
1096 :
1097 15890 : if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
1098 0 : free_extent_buffer(buf);
1099 0 : return -EIO;
1100 15890 : } else if (extent_buffer_uptodate(buf)) {
1101 15746 : *eb = buf;
1102 : } else {
1103 144 : free_extent_buffer(buf);
1104 : }
1105 : return 0;
1106 : }
1107 :
1108 2224513 : struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
1109 : u64 bytenr, u32 blocksize)
1110 : {
1111 2224513 : return find_extent_buffer(root->fs_info, bytenr);
1112 : }
1113 :
1114 59180 : struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
1115 : u64 bytenr, u32 blocksize)
1116 : {
1117 : #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1118 : if (unlikely(test_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state)))
1119 : return alloc_test_extent_buffer(root->fs_info, bytenr,
1120 : blocksize);
1121 : #endif
1122 1268784 : return alloc_extent_buffer(root->fs_info, bytenr, blocksize);
1123 : }
1124 :
1125 :
1126 0 : int btrfs_write_tree_block(struct extent_buffer *buf)
1127 : {
1128 0 : return filemap_fdatawrite_range(buf->pages[0]->mapping, buf->start,
1129 0 : buf->start + buf->len - 1);
1130 : }
1131 :
1132 505 : int btrfs_wait_tree_block_writeback(struct extent_buffer *buf)
1133 : {
1134 505 : return filemap_fdatawait_range(buf->pages[0]->mapping,
1135 505 : buf->start, buf->start + buf->len - 1);
1136 : }
1137 :
1138 1193067 : struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
1139 : u32 blocksize, u64 parent_transid)
1140 : {
1141 : struct extent_buffer *buf = NULL;
1142 : int ret;
1143 :
1144 : buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
1145 1193071 : if (!buf)
1146 : return NULL;
1147 :
1148 1193071 : ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
1149 1193072 : if (ret) {
1150 0 : free_extent_buffer(buf);
1151 0 : return NULL;
1152 : }
1153 : return buf;
1154 :
1155 : }
1156 :
1157 113383 : void clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1158 113383 : struct extent_buffer *buf)
1159 : {
1160 113383 : struct btrfs_fs_info *fs_info = root->fs_info;
1161 :
1162 113383 : if (btrfs_header_generation(buf) ==
1163 113383 : fs_info->running_transaction->transid) {
1164 73680 : btrfs_assert_tree_locked(buf);
1165 :
1166 147360 : if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) {
1167 4012 : __percpu_counter_add(&fs_info->dirty_metadata_bytes,
1168 2006 : -buf->len,
1169 : fs_info->dirty_metadata_batch);
1170 : /* ugh, clear_extent_buffer_dirty needs to lock the page */
1171 : btrfs_set_lock_blocking(buf);
1172 2006 : clear_extent_buffer_dirty(buf);
1173 : }
1174 : }
1175 113383 : }
1176 :
1177 688 : static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
1178 : {
1179 : struct btrfs_subvolume_writers *writers;
1180 : int ret;
1181 :
1182 : writers = kmalloc(sizeof(*writers), GFP_NOFS);
1183 688 : if (!writers)
1184 : return ERR_PTR(-ENOMEM);
1185 :
1186 688 : ret = percpu_counter_init(&writers->counter, 0);
1187 688 : if (ret < 0) {
1188 0 : kfree(writers);
1189 0 : return ERR_PTR(ret);
1190 : }
1191 :
1192 688 : init_waitqueue_head(&writers->wait);
1193 688 : return writers;
1194 : }
1195 :
1196 : static void
1197 : btrfs_free_subvolume_writers(struct btrfs_subvolume_writers *writers)
1198 : {
1199 688 : percpu_counter_destroy(&writers->counter);
1200 688 : kfree(writers);
1201 : }
1202 :
1203 3229 : static void __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
1204 : u32 stripesize, struct btrfs_root *root,
1205 : struct btrfs_fs_info *fs_info,
1206 : u64 objectid)
1207 : {
1208 3229 : root->node = NULL;
1209 3229 : root->commit_root = NULL;
1210 3229 : root->sectorsize = sectorsize;
1211 3229 : root->nodesize = nodesize;
1212 3229 : root->leafsize = leafsize;
1213 3229 : root->stripesize = stripesize;
1214 3229 : root->state = 0;
1215 3229 : root->orphan_cleanup_state = 0;
1216 :
1217 3229 : root->objectid = objectid;
1218 3229 : root->last_trans = 0;
1219 3229 : root->highest_objectid = 0;
1220 3229 : root->nr_delalloc_inodes = 0;
1221 3229 : root->nr_ordered_extents = 0;
1222 3229 : root->name = NULL;
1223 3229 : root->inode_tree = RB_ROOT;
1224 3229 : INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC);
1225 3229 : root->block_rsv = NULL;
1226 3229 : root->orphan_block_rsv = NULL;
1227 :
1228 3229 : INIT_LIST_HEAD(&root->dirty_list);
1229 3229 : INIT_LIST_HEAD(&root->root_list);
1230 3229 : INIT_LIST_HEAD(&root->delalloc_inodes);
1231 3229 : INIT_LIST_HEAD(&root->delalloc_root);
1232 3229 : INIT_LIST_HEAD(&root->ordered_extents);
1233 3229 : INIT_LIST_HEAD(&root->ordered_root);
1234 3229 : INIT_LIST_HEAD(&root->logged_list[0]);
1235 3229 : INIT_LIST_HEAD(&root->logged_list[1]);
1236 3229 : spin_lock_init(&root->orphan_lock);
1237 3229 : spin_lock_init(&root->inode_lock);
1238 3229 : spin_lock_init(&root->delalloc_lock);
1239 3229 : spin_lock_init(&root->ordered_extent_lock);
1240 3229 : spin_lock_init(&root->accounting_lock);
1241 3229 : spin_lock_init(&root->log_extents_lock[0]);
1242 3229 : spin_lock_init(&root->log_extents_lock[1]);
1243 3229 : mutex_init(&root->objectid_mutex);
1244 3229 : mutex_init(&root->log_mutex);
1245 3229 : mutex_init(&root->ordered_extent_mutex);
1246 3229 : mutex_init(&root->delalloc_mutex);
1247 3229 : init_waitqueue_head(&root->log_writer_wait);
1248 3229 : init_waitqueue_head(&root->log_commit_wait[0]);
1249 3229 : init_waitqueue_head(&root->log_commit_wait[1]);
1250 3229 : INIT_LIST_HEAD(&root->log_ctxs[0]);
1251 3229 : INIT_LIST_HEAD(&root->log_ctxs[1]);
1252 : atomic_set(&root->log_commit[0], 0);
1253 : atomic_set(&root->log_commit[1], 0);
1254 : atomic_set(&root->log_writers, 0);
1255 : atomic_set(&root->log_batch, 0);
1256 : atomic_set(&root->orphan_inodes, 0);
1257 : atomic_set(&root->refs, 1);
1258 : atomic_set(&root->will_be_snapshoted, 0);
1259 3229 : root->log_transid = 0;
1260 3229 : root->log_transid_committed = -1;
1261 3229 : root->last_log_commit = 0;
1262 3229 : if (fs_info)
1263 3229 : extent_io_tree_init(&root->dirty_log_pages,
1264 3229 : fs_info->btree_inode->i_mapping);
1265 :
1266 3229 : memset(&root->root_key, 0, sizeof(root->root_key));
1267 3229 : memset(&root->root_item, 0, sizeof(root->root_item));
1268 3229 : memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
1269 3229 : memset(&root->root_kobj, 0, sizeof(root->root_kobj));
1270 3229 : if (fs_info)
1271 3229 : root->defrag_trans_start = fs_info->generation;
1272 : else
1273 0 : root->defrag_trans_start = 0;
1274 : init_completion(&root->kobj_unregister);
1275 3229 : root->root_key.objectid = objectid;
1276 3229 : root->anon_dev = 0;
1277 :
1278 3229 : spin_lock_init(&root->root_item_lock);
1279 3229 : }
1280 :
1281 : static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info)
1282 : {
1283 3229 : struct btrfs_root *root = kzalloc(sizeof(*root), GFP_NOFS);
1284 3229 : if (root)
1285 3229 : root->fs_info = fs_info;
1286 : return root;
1287 : }
1288 :
1289 : #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
1290 : /* Should only be used by the testing infrastructure */
1291 : struct btrfs_root *btrfs_alloc_dummy_root(void)
1292 : {
1293 : struct btrfs_root *root;
1294 :
1295 : root = btrfs_alloc_root(NULL);
1296 : if (!root)
1297 : return ERR_PTR(-ENOMEM);
1298 : __setup_root(4096, 4096, 4096, 4096, root, NULL, 1);
1299 : set_bit(BTRFS_ROOT_DUMMY_ROOT, &root->state);
1300 : root->alloc_bytenr = 0;
1301 :
1302 : return root;
1303 : }
1304 : #endif
1305 :
1306 105 : struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans,
1307 : struct btrfs_fs_info *fs_info,
1308 : u64 objectid)
1309 : {
1310 315 : struct extent_buffer *leaf;
1311 105 : struct btrfs_root *tree_root = fs_info->tree_root;
1312 : struct btrfs_root *root;
1313 : struct btrfs_key key;
1314 : int ret = 0;
1315 : uuid_le uuid;
1316 :
1317 : root = btrfs_alloc_root(fs_info);
1318 105 : if (!root)
1319 : return ERR_PTR(-ENOMEM);
1320 :
1321 105 : __setup_root(tree_root->nodesize, tree_root->leafsize,
1322 : tree_root->sectorsize, tree_root->stripesize,
1323 : root, fs_info, objectid);
1324 105 : root->root_key.objectid = objectid;
1325 105 : root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1326 105 : root->root_key.offset = 0;
1327 :
1328 105 : leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
1329 : 0, objectid, NULL, 0, 0, 0);
1330 105 : if (IS_ERR(leaf)) {
1331 0 : ret = PTR_ERR(leaf);
1332 : leaf = NULL;
1333 0 : goto fail;
1334 : }
1335 :
1336 105 : memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1337 105 : btrfs_set_header_bytenr(leaf, leaf->start);
1338 105 : btrfs_set_header_generation(leaf, trans->transid);
1339 : btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1340 : btrfs_set_header_owner(leaf, objectid);
1341 105 : root->node = leaf;
1342 :
1343 105 : write_extent_buffer(leaf, fs_info->fsid, btrfs_header_fsid(),
1344 : BTRFS_FSID_SIZE);
1345 105 : write_extent_buffer(leaf, fs_info->chunk_tree_uuid,
1346 : btrfs_header_chunk_tree_uuid(leaf),
1347 : BTRFS_UUID_SIZE);
1348 105 : btrfs_mark_buffer_dirty(leaf);
1349 :
1350 105 : root->commit_root = btrfs_root_node(root);
1351 : set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state);
1352 :
1353 105 : root->root_item.flags = 0;
1354 105 : root->root_item.byte_limit = 0;
1355 105 : btrfs_set_root_bytenr(&root->root_item, leaf->start);
1356 105 : btrfs_set_root_generation(&root->root_item, trans->transid);
1357 : btrfs_set_root_level(&root->root_item, 0);
1358 : btrfs_set_root_refs(&root->root_item, 1);
1359 105 : btrfs_set_root_used(&root->root_item, leaf->len);
1360 : btrfs_set_root_last_snapshot(&root->root_item, 0);
1361 : btrfs_set_root_dirid(&root->root_item, 0);
1362 105 : uuid_le_gen(&uuid);
1363 105 : memcpy(root->root_item.uuid, uuid.b, BTRFS_UUID_SIZE);
1364 105 : root->root_item.drop_level = 0;
1365 :
1366 105 : key.objectid = objectid;
1367 105 : key.type = BTRFS_ROOT_ITEM_KEY;
1368 105 : key.offset = 0;
1369 105 : ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item);
1370 105 : if (ret)
1371 : goto fail;
1372 :
1373 105 : btrfs_tree_unlock(leaf);
1374 :
1375 105 : return root;
1376 :
1377 : fail:
1378 0 : if (leaf) {
1379 0 : btrfs_tree_unlock(leaf);
1380 0 : free_extent_buffer(root->commit_root);
1381 0 : free_extent_buffer(leaf);
1382 : }
1383 0 : kfree(root);
1384 :
1385 0 : return ERR_PTR(ret);
1386 : }
1387 :
1388 450 : static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans,
1389 : struct btrfs_fs_info *fs_info)
1390 : {
1391 : struct btrfs_root *root;
1392 450 : struct btrfs_root *tree_root = fs_info->tree_root;
1393 1350 : struct extent_buffer *leaf;
1394 :
1395 : root = btrfs_alloc_root(fs_info);
1396 450 : if (!root)
1397 : return ERR_PTR(-ENOMEM);
1398 :
1399 450 : __setup_root(tree_root->nodesize, tree_root->leafsize,
1400 : tree_root->sectorsize, tree_root->stripesize,
1401 : root, fs_info, BTRFS_TREE_LOG_OBJECTID);
1402 :
1403 450 : root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID;
1404 450 : root->root_key.type = BTRFS_ROOT_ITEM_KEY;
1405 450 : root->root_key.offset = BTRFS_TREE_LOG_OBJECTID;
1406 :
1407 : /*
1408 : * DON'T set REF_COWS for log trees
1409 : *
1410 : * log trees do not get reference counted because they go away
1411 : * before a real commit is actually done. They do store pointers
1412 : * to file data extents, and those reference counts still get
1413 : * updated (along with back refs to the log tree).
1414 : */
1415 :
1416 450 : leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
1417 : BTRFS_TREE_LOG_OBJECTID, NULL,
1418 : 0, 0, 0);
1419 450 : if (IS_ERR(leaf)) {
1420 0 : kfree(root);
1421 0 : return ERR_CAST(leaf);
1422 : }
1423 :
1424 450 : memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
1425 450 : btrfs_set_header_bytenr(leaf, leaf->start);
1426 450 : btrfs_set_header_generation(leaf, trans->transid);
1427 : btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
1428 : btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
1429 450 : root->node = leaf;
1430 :
1431 450 : write_extent_buffer(root->node, root->fs_info->fsid,
1432 : btrfs_header_fsid(), BTRFS_FSID_SIZE);
1433 450 : btrfs_mark_buffer_dirty(root->node);
1434 450 : btrfs_tree_unlock(root->node);
1435 450 : return root;
1436 : }
1437 :
1438 225 : int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
1439 : struct btrfs_fs_info *fs_info)
1440 : {
1441 : struct btrfs_root *log_root;
1442 :
1443 225 : log_root = alloc_log_tree(trans, fs_info);
1444 225 : if (IS_ERR(log_root))
1445 0 : return PTR_ERR(log_root);
1446 225 : WARN_ON(fs_info->log_root_tree);
1447 225 : fs_info->log_root_tree = log_root;
1448 225 : return 0;
1449 : }
1450 :
1451 225 : int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
1452 : struct btrfs_root *root)
1453 : {
1454 : struct btrfs_root *log_root;
1455 : struct btrfs_inode_item *inode_item;
1456 :
1457 225 : log_root = alloc_log_tree(trans, root->fs_info);
1458 225 : if (IS_ERR(log_root))
1459 0 : return PTR_ERR(log_root);
1460 :
1461 225 : log_root->last_trans = trans->transid;
1462 225 : log_root->root_key.offset = root->root_key.objectid;
1463 :
1464 : inode_item = &log_root->root_item.inode;
1465 : btrfs_set_stack_inode_generation(inode_item, 1);
1466 : btrfs_set_stack_inode_size(inode_item, 3);
1467 : btrfs_set_stack_inode_nlink(inode_item, 1);
1468 225 : btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
1469 : btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1470 :
1471 225 : btrfs_set_root_node(&log_root->root_item, log_root->node);
1472 :
1473 225 : WARN_ON(root->log_root);
1474 225 : root->log_root = log_root;
1475 225 : root->log_transid = 0;
1476 225 : root->log_transid_committed = -1;
1477 225 : root->last_log_commit = 0;
1478 225 : return 0;
1479 : }
1480 :
1481 2232 : static struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root,
1482 : struct btrfs_key *key)
1483 : {
1484 1915 : struct btrfs_root *root;
1485 2232 : struct btrfs_fs_info *fs_info = tree_root->fs_info;
1486 : struct btrfs_path *path;
1487 : u64 generation;
1488 : u32 blocksize;
1489 : int ret;
1490 :
1491 2232 : path = btrfs_alloc_path();
1492 2232 : if (!path)
1493 : return ERR_PTR(-ENOMEM);
1494 :
1495 : root = btrfs_alloc_root(fs_info);
1496 2232 : if (!root) {
1497 : ret = -ENOMEM;
1498 : goto alloc_fail;
1499 : }
1500 :
1501 2232 : __setup_root(tree_root->nodesize, tree_root->leafsize,
1502 : tree_root->sectorsize, tree_root->stripesize,
1503 : root, fs_info, key->objectid);
1504 :
1505 4147 : ret = btrfs_find_root(tree_root, key, path,
1506 : &root->root_item, &root->root_key);
1507 2232 : if (ret) {
1508 317 : if (ret > 0)
1509 : ret = -ENOENT;
1510 : goto find_fail;
1511 : }
1512 :
1513 : generation = btrfs_root_generation(&root->root_item);
1514 : blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
1515 1915 : root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
1516 : blocksize, generation);
1517 1915 : if (!root->node) {
1518 : ret = -ENOMEM;
1519 : goto find_fail;
1520 1915 : } else if (!btrfs_buffer_uptodate(root->node, generation, 0)) {
1521 : ret = -EIO;
1522 : goto read_fail;
1523 : }
1524 1915 : root->commit_root = btrfs_root_node(root);
1525 : out:
1526 2232 : btrfs_free_path(path);
1527 2232 : return root;
1528 :
1529 : read_fail:
1530 0 : free_extent_buffer(root->node);
1531 : find_fail:
1532 317 : kfree(root);
1533 : alloc_fail:
1534 317 : root = ERR_PTR(ret);
1535 317 : goto out;
1536 : }
1537 :
1538 1127 : struct btrfs_root *btrfs_read_fs_root(struct btrfs_root *tree_root,
1539 : struct btrfs_key *location)
1540 : {
1541 : struct btrfs_root *root;
1542 :
1543 1127 : root = btrfs_read_tree_root(tree_root, location);
1544 1127 : if (IS_ERR(root))
1545 : return root;
1546 :
1547 1127 : if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
1548 : set_bit(BTRFS_ROOT_REF_COWS, &root->state);
1549 1127 : btrfs_check_and_init_root_item(&root->root_item);
1550 : }
1551 :
1552 : return root;
1553 : }
1554 :
1555 688 : int btrfs_init_fs_root(struct btrfs_root *root)
1556 : {
1557 : int ret;
1558 : struct btrfs_subvolume_writers *writers;
1559 :
1560 688 : root->free_ino_ctl = kzalloc(sizeof(*root->free_ino_ctl), GFP_NOFS);
1561 688 : root->free_ino_pinned = kzalloc(sizeof(*root->free_ino_pinned),
1562 : GFP_NOFS);
1563 688 : if (!root->free_ino_pinned || !root->free_ino_ctl) {
1564 : ret = -ENOMEM;
1565 : goto fail;
1566 : }
1567 :
1568 688 : writers = btrfs_alloc_subvolume_writers();
1569 688 : if (IS_ERR(writers)) {
1570 0 : ret = PTR_ERR(writers);
1571 0 : goto fail;
1572 : }
1573 688 : root->subv_writers = writers;
1574 :
1575 688 : btrfs_init_free_ino_ctl(root);
1576 688 : spin_lock_init(&root->cache_lock);
1577 688 : init_waitqueue_head(&root->cache_wait);
1578 :
1579 688 : ret = get_anon_bdev(&root->anon_dev);
1580 688 : if (ret)
1581 : goto free_writers;
1582 : return 0;
1583 :
1584 : free_writers:
1585 0 : btrfs_free_subvolume_writers(root->subv_writers);
1586 : fail:
1587 0 : kfree(root->free_ino_ctl);
1588 0 : kfree(root->free_ino_pinned);
1589 0 : return ret;
1590 : }
1591 :
1592 1187792 : static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info,
1593 : u64 root_id)
1594 : {
1595 : struct btrfs_root *root;
1596 :
1597 : spin_lock(&fs_info->fs_roots_radix_lock);
1598 1187792 : root = radix_tree_lookup(&fs_info->fs_roots_radix,
1599 : (unsigned long)root_id);
1600 : spin_unlock(&fs_info->fs_roots_radix_lock);
1601 1187792 : return root;
1602 : }
1603 :
1604 688 : int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info,
1605 : struct btrfs_root *root)
1606 : {
1607 : int ret;
1608 :
1609 688 : ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
1610 688 : if (ret)
1611 : return ret;
1612 :
1613 : spin_lock(&fs_info->fs_roots_radix_lock);
1614 688 : ret = radix_tree_insert(&fs_info->fs_roots_radix,
1615 688 : (unsigned long)root->root_key.objectid,
1616 : root);
1617 688 : if (ret == 0)
1618 : set_bit(BTRFS_ROOT_IN_RADIX, &root->state);
1619 : spin_unlock(&fs_info->fs_roots_radix_lock);
1620 : radix_tree_preload_end();
1621 :
1622 688 : return ret;
1623 : }
1624 :
1625 1191361 : struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info,
1626 : struct btrfs_key *location,
1627 : bool check_ref)
1628 : {
1629 : struct btrfs_root *root;
1630 : int ret;
1631 :
1632 1191361 : if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1633 3263 : return fs_info->tree_root;
1634 1188098 : if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
1635 139 : return fs_info->extent_root;
1636 1187959 : if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
1637 29 : return fs_info->chunk_root;
1638 1187930 : if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
1639 17 : return fs_info->dev_root;
1640 1187913 : if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
1641 53 : return fs_info->csum_root;
1642 1187860 : if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
1643 19 : return fs_info->quota_root ? fs_info->quota_root :
1644 : ERR_PTR(-ENOENT);
1645 1187841 : if (location->objectid == BTRFS_UUID_TREE_OBJECTID)
1646 49 : return fs_info->uuid_root ? fs_info->uuid_root :
1647 : ERR_PTR(-ENOENT);
1648 : again:
1649 1187792 : root = btrfs_lookup_fs_root(fs_info, location->objectid);
1650 1187792 : if (root) {
1651 2372977 : if (check_ref && btrfs_root_refs(&root->root_item) == 0)
1652 : return ERR_PTR(-ENOENT);
1653 1187176 : return root;
1654 : }
1655 :
1656 616 : root = btrfs_read_fs_root(fs_info->tree_root, location);
1657 616 : if (IS_ERR(root))
1658 : return root;
1659 :
1660 1038 : if (check_ref && btrfs_root_refs(&root->root_item) == 0) {
1661 : ret = -ENOENT;
1662 : goto fail;
1663 : }
1664 :
1665 616 : ret = btrfs_init_fs_root(root);
1666 616 : if (ret)
1667 : goto fail;
1668 :
1669 616 : ret = btrfs_find_item(fs_info->tree_root, NULL, BTRFS_ORPHAN_OBJECTID,
1670 : location->objectid, BTRFS_ORPHAN_ITEM_KEY, NULL);
1671 616 : if (ret < 0)
1672 : goto fail;
1673 616 : if (ret == 0)
1674 : set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state);
1675 :
1676 616 : ret = btrfs_insert_fs_root(fs_info, root);
1677 616 : if (ret) {
1678 0 : if (ret == -EEXIST) {
1679 0 : free_fs_root(root);
1680 0 : goto again;
1681 : }
1682 : goto fail;
1683 : }
1684 : return root;
1685 : fail:
1686 0 : free_fs_root(root);
1687 0 : return ERR_PTR(ret);
1688 : }
1689 :
1690 1251 : static int btrfs_congested_fn(void *congested_data, int bdi_bits)
1691 : {
1692 : struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data;
1693 : int ret = 0;
1694 : struct btrfs_device *device;
1695 : struct backing_dev_info *bdi;
1696 :
1697 : rcu_read_lock();
1698 2502 : list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
1699 1251 : if (!device->bdev)
1700 0 : continue;
1701 1251 : bdi = blk_get_backing_dev_info(device->bdev);
1702 2502 : if (bdi && bdi_congested(bdi, bdi_bits)) {
1703 : ret = 1;
1704 : break;
1705 : }
1706 : }
1707 : rcu_read_unlock();
1708 1251 : return ret;
1709 : }
1710 :
1711 : /*
1712 : * If this fails, caller must call bdi_destroy() to get rid of the
1713 : * bdi again.
1714 : */
1715 221 : static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
1716 : {
1717 : int err;
1718 :
1719 221 : bdi->capabilities = BDI_CAP_MAP_COPY;
1720 221 : err = bdi_setup_and_register(bdi, "btrfs", BDI_CAP_MAP_COPY);
1721 221 : if (err)
1722 : return err;
1723 :
1724 221 : bdi->ra_pages = default_backing_dev_info.ra_pages;
1725 221 : bdi->congested_fn = btrfs_congested_fn;
1726 221 : bdi->congested_data = info;
1727 221 : return 0;
1728 : }
1729 :
1730 : /*
1731 : * called by the kthread helper functions to finally call the bio end_io
1732 : * functions. This is where read checksum verification actually happens
1733 : */
1734 28785 : static void end_workqueue_fn(struct btrfs_work *work)
1735 : {
1736 : struct bio *bio;
1737 : struct end_io_wq *end_io_wq;
1738 : int error;
1739 :
1740 28785 : end_io_wq = container_of(work, struct end_io_wq, work);
1741 28785 : bio = end_io_wq->bio;
1742 :
1743 28785 : error = end_io_wq->error;
1744 28785 : bio->bi_private = end_io_wq->private;
1745 28785 : bio->bi_end_io = end_io_wq->end_io;
1746 28785 : kfree(end_io_wq);
1747 28784 : bio_endio_nodec(bio, error);
1748 28785 : }
1749 :
1750 221 : static int cleaner_kthread(void *arg)
1751 : {
1752 1024 : struct btrfs_root *root = arg;
1753 : int again;
1754 :
1755 : do {
1756 : again = 0;
1757 :
1758 : /* Make the cleaner go to sleep early. */
1759 729 : if (btrfs_need_cleaner_sleep(root))
1760 : goto sleep;
1761 :
1762 476 : if (!mutex_trylock(&root->fs_info->cleaner_mutex))
1763 : goto sleep;
1764 :
1765 : /*
1766 : * Avoid the problem that we change the status of the fs
1767 : * during the above check and trylock.
1768 : */
1769 295 : if (btrfs_need_cleaner_sleep(root)) {
1770 0 : mutex_unlock(&root->fs_info->cleaner_mutex);
1771 0 : goto sleep;
1772 : }
1773 :
1774 295 : btrfs_run_delayed_iputs(root);
1775 295 : again = btrfs_clean_one_deleted_snapshot(root);
1776 295 : mutex_unlock(&root->fs_info->cleaner_mutex);
1777 :
1778 : /*
1779 : * The defragger has dealt with the R/O remount and umount,
1780 : * needn't do anything special here.
1781 : */
1782 295 : btrfs_run_defrag_inodes(root->fs_info);
1783 : sleep:
1784 729 : if (!try_to_freeze() && !again) {
1785 1436 : set_current_state(TASK_INTERRUPTIBLE);
1786 718 : if (!kthread_should_stop())
1787 718 : schedule();
1788 718 : __set_current_state(TASK_RUNNING);
1789 : }
1790 729 : } while (!kthread_should_stop());
1791 221 : return 0;
1792 : }
1793 :
1794 221 : static int transaction_kthread(void *arg)
1795 : {
1796 : struct btrfs_root *root = arg;
1797 : struct btrfs_trans_handle *trans;
1798 : struct btrfs_transaction *cur;
1799 : u64 transid;
1800 : unsigned long now;
1801 : unsigned long delay;
1802 : bool cannot_commit;
1803 :
1804 : do {
1805 : cannot_commit = false;
1806 303 : delay = HZ * root->fs_info->commit_interval;
1807 303 : mutex_lock(&root->fs_info->transaction_kthread_mutex);
1808 :
1809 303 : spin_lock(&root->fs_info->trans_lock);
1810 303 : cur = root->fs_info->running_transaction;
1811 303 : if (!cur) {
1812 : spin_unlock(&root->fs_info->trans_lock);
1813 : goto sleep;
1814 : }
1815 :
1816 64 : now = get_seconds();
1817 125 : if (cur->state < TRANS_STATE_BLOCKED &&
1818 122 : (now < cur->start_time ||
1819 61 : now - cur->start_time < root->fs_info->commit_interval)) {
1820 46 : spin_unlock(&root->fs_info->trans_lock);
1821 : delay = HZ * 5;
1822 46 : goto sleep;
1823 : }
1824 18 : transid = cur->transid;
1825 18 : spin_unlock(&root->fs_info->trans_lock);
1826 :
1827 : /* If the file system is aborted, this will always fail. */
1828 18 : trans = btrfs_attach_transaction(root);
1829 18 : if (IS_ERR(trans)) {
1830 2 : if (PTR_ERR(trans) != -ENOENT)
1831 : cannot_commit = true;
1832 : goto sleep;
1833 : }
1834 16 : if (transid == trans->transid) {
1835 16 : btrfs_commit_transaction(trans, root);
1836 : } else {
1837 0 : btrfs_end_transaction(trans, root);
1838 : }
1839 : sleep:
1840 303 : wake_up_process(root->fs_info->cleaner_kthread);
1841 303 : mutex_unlock(&root->fs_info->transaction_kthread_mutex);
1842 :
1843 606 : if (unlikely(test_bit(BTRFS_FS_STATE_ERROR,
1844 : &root->fs_info->fs_state)))
1845 0 : btrfs_cleanup_transaction(root);
1846 303 : if (!try_to_freeze()) {
1847 606 : set_current_state(TASK_INTERRUPTIBLE);
1848 606 : if (!kthread_should_stop() &&
1849 304 : (!btrfs_transaction_blocked(root->fs_info) ||
1850 : cannot_commit))
1851 302 : schedule_timeout(delay);
1852 303 : __set_current_state(TASK_RUNNING);
1853 : }
1854 303 : } while (!kthread_should_stop());
1855 221 : return 0;
1856 : }
1857 :
1858 : /*
1859 : * this will find the highest generation in the array of
1860 : * root backups. The index of the highest array is returned,
1861 : * or -1 if we can't find anything.
1862 : *
1863 : * We check to make sure the array is valid by comparing the
1864 : * generation of the latest root in the array with the generation
1865 : * in the super block. If they don't match we pitch it.
1866 : */
1867 : static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
1868 : {
1869 : u64 cur;
1870 : int newest_index = -1;
1871 : struct btrfs_root_backup *root_backup;
1872 : int i;
1873 :
1874 884 : for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
1875 884 : root_backup = info->super_copy->super_roots + i;
1876 : cur = btrfs_backup_tree_root_gen(root_backup);
1877 884 : if (cur == newest_gen)
1878 : newest_index = i;
1879 : }
1880 :
1881 : /* check to see if we actually wrapped around */
1882 221 : if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
1883 27 : root_backup = info->super_copy->super_roots;
1884 : cur = btrfs_backup_tree_root_gen(root_backup);
1885 27 : if (cur == newest_gen)
1886 : newest_index = 0;
1887 : }
1888 : return newest_index;
1889 : }
1890 :
1891 :
1892 : /*
1893 : * find the oldest backup so we know where to store new entries
1894 : * in the backup array. This will set the backup_root_index
1895 : * field in the fs_info struct
1896 : */
1897 221 : static void find_oldest_super_backup(struct btrfs_fs_info *info,
1898 : u64 newest_gen)
1899 : {
1900 : int newest_index = -1;
1901 :
1902 : newest_index = find_newest_super_backup(info, newest_gen);
1903 : /* if there was garbage in there, just move along */
1904 221 : if (newest_index == -1) {
1905 99 : info->backup_root_index = 0;
1906 : } else {
1907 122 : info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
1908 : }
1909 221 : }
1910 :
1911 : /*
1912 : * copy all the root pointers into the super backup array.
1913 : * this will bump the backup pointer by one when it is
1914 : * done
1915 : */
1916 3575 : static void backup_super_roots(struct btrfs_fs_info *info)
1917 : {
1918 : int next_backup;
1919 : struct btrfs_root_backup *root_backup;
1920 : int last_backup;
1921 :
1922 3575 : next_backup = info->backup_root_index;
1923 3575 : last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
1924 : BTRFS_NUM_BACKUP_ROOTS;
1925 :
1926 : /*
1927 : * just overwrite the last backup if we're at the same generation
1928 : * this happens only at umount
1929 : */
1930 3575 : root_backup = info->super_for_commit->super_roots + last_backup;
1931 3575 : if (btrfs_backup_tree_root_gen(root_backup) ==
1932 10725 : btrfs_header_generation(info->tree_root->node))
1933 : next_backup = last_backup;
1934 :
1935 3575 : root_backup = info->super_for_commit->super_roots + next_backup;
1936 :
1937 : /*
1938 : * make sure all of our padding and empty slots get zero filled
1939 : * regardless of which ones we use today
1940 : */
1941 3575 : memset(root_backup, 0, sizeof(*root_backup));
1942 :
1943 3575 : info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
1944 :
1945 3575 : btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
1946 3575 : btrfs_set_backup_tree_root_gen(root_backup,
1947 3575 : btrfs_header_generation(info->tree_root->node));
1948 :
1949 : btrfs_set_backup_tree_root_level(root_backup,
1950 3575 : btrfs_header_level(info->tree_root->node));
1951 :
1952 10725 : btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
1953 3575 : btrfs_set_backup_chunk_root_gen(root_backup,
1954 3575 : btrfs_header_generation(info->chunk_root->node));
1955 : btrfs_set_backup_chunk_root_level(root_backup,
1956 3575 : btrfs_header_level(info->chunk_root->node));
1957 :
1958 10725 : btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
1959 3575 : btrfs_set_backup_extent_root_gen(root_backup,
1960 3575 : btrfs_header_generation(info->extent_root->node));
1961 : btrfs_set_backup_extent_root_level(root_backup,
1962 3575 : btrfs_header_level(info->extent_root->node));
1963 :
1964 : /*
1965 : * we might commit during log recovery, which happens before we set
1966 : * the fs_root. Make sure it is valid before we fill it in.
1967 : */
1968 10725 : if (info->fs_root && info->fs_root->node) {
1969 3575 : btrfs_set_backup_fs_root(root_backup,
1970 : info->fs_root->node->start);
1971 3575 : btrfs_set_backup_fs_root_gen(root_backup,
1972 3575 : btrfs_header_generation(info->fs_root->node));
1973 : btrfs_set_backup_fs_root_level(root_backup,
1974 3575 : btrfs_header_level(info->fs_root->node));
1975 : }
1976 :
1977 10725 : btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
1978 3575 : btrfs_set_backup_dev_root_gen(root_backup,
1979 3575 : btrfs_header_generation(info->dev_root->node));
1980 : btrfs_set_backup_dev_root_level(root_backup,
1981 3575 : btrfs_header_level(info->dev_root->node));
1982 :
1983 10725 : btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
1984 3575 : btrfs_set_backup_csum_root_gen(root_backup,
1985 3575 : btrfs_header_generation(info->csum_root->node));
1986 : btrfs_set_backup_csum_root_level(root_backup,
1987 3575 : btrfs_header_level(info->csum_root->node));
1988 :
1989 3575 : btrfs_set_backup_total_bytes(root_backup,
1990 : btrfs_super_total_bytes(info->super_copy));
1991 3575 : btrfs_set_backup_bytes_used(root_backup,
1992 : btrfs_super_bytes_used(info->super_copy));
1993 3575 : btrfs_set_backup_num_devices(root_backup,
1994 : btrfs_super_num_devices(info->super_copy));
1995 :
1996 : /*
1997 : * if we don't copy this out to the super_copy, it won't get remembered
1998 : * for the next commit
1999 : */
2000 3575 : memcpy(&info->super_copy->super_roots,
2001 3575 : &info->super_for_commit->super_roots,
2002 : sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
2003 3575 : }
2004 :
2005 : /*
2006 : * this copies info out of the root backup array and back into
2007 : * the in-memory super block. It is meant to help iterate through
2008 : * the array, so you send it the number of backups you've already
2009 : * tried and the last backup index you used.
2010 : *
2011 : * this returns -1 when it has tried all the backups
2012 : */
2013 0 : static noinline int next_root_backup(struct btrfs_fs_info *info,
2014 : struct btrfs_super_block *super,
2015 : int *num_backups_tried, int *backup_index)
2016 : {
2017 0 : struct btrfs_root_backup *root_backup;
2018 : int newest = *backup_index;
2019 :
2020 0 : if (*num_backups_tried == 0) {
2021 : u64 gen = btrfs_super_generation(super);
2022 :
2023 : newest = find_newest_super_backup(info, gen);
2024 0 : if (newest == -1)
2025 : return -1;
2026 :
2027 0 : *backup_index = newest;
2028 0 : *num_backups_tried = 1;
2029 0 : } else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
2030 : /* we've tried all the backups, all done */
2031 : return -1;
2032 : } else {
2033 : /* jump to the next oldest backup */
2034 0 : newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
2035 : BTRFS_NUM_BACKUP_ROOTS;
2036 0 : *backup_index = newest;
2037 0 : *num_backups_tried += 1;
2038 : }
2039 0 : root_backup = super->super_roots + newest;
2040 :
2041 : btrfs_set_super_generation(super,
2042 : btrfs_backup_tree_root_gen(root_backup));
2043 : btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
2044 : btrfs_set_super_root_level(super,
2045 : btrfs_backup_tree_root_level(root_backup));
2046 : btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
2047 :
2048 : /*
2049 : * fixme: the total bytes and num_devices need to match or we should
2050 : * need a fsck
2051 : */
2052 : btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
2053 : btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
2054 0 : return 0;
2055 : }
2056 :
2057 : /* helper to cleanup workers */
2058 221 : static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info)
2059 : {
2060 221 : btrfs_destroy_workqueue(fs_info->fixup_workers);
2061 221 : btrfs_destroy_workqueue(fs_info->delalloc_workers);
2062 221 : btrfs_destroy_workqueue(fs_info->workers);
2063 221 : btrfs_destroy_workqueue(fs_info->endio_workers);
2064 221 : btrfs_destroy_workqueue(fs_info->endio_meta_workers);
2065 221 : btrfs_destroy_workqueue(fs_info->endio_raid56_workers);
2066 221 : btrfs_destroy_workqueue(fs_info->rmw_workers);
2067 221 : btrfs_destroy_workqueue(fs_info->endio_meta_write_workers);
2068 221 : btrfs_destroy_workqueue(fs_info->endio_write_workers);
2069 221 : btrfs_destroy_workqueue(fs_info->endio_freespace_worker);
2070 221 : btrfs_destroy_workqueue(fs_info->submit_workers);
2071 221 : btrfs_destroy_workqueue(fs_info->delayed_workers);
2072 221 : btrfs_destroy_workqueue(fs_info->caching_workers);
2073 221 : btrfs_destroy_workqueue(fs_info->readahead_workers);
2074 221 : btrfs_destroy_workqueue(fs_info->flush_workers);
2075 221 : btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers);
2076 221 : btrfs_destroy_workqueue(fs_info->extent_workers);
2077 221 : }
2078 :
2079 1547 : static void free_root_extent_buffers(struct btrfs_root *root)
2080 : {
2081 1547 : if (root) {
2082 1335 : free_extent_buffer(root->node);
2083 1335 : free_extent_buffer(root->commit_root);
2084 1335 : root->node = NULL;
2085 1335 : root->commit_root = NULL;
2086 : }
2087 1547 : }
2088 :
2089 : /* helper to cleanup tree roots */
2090 221 : static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
2091 : {
2092 221 : free_root_extent_buffers(info->tree_root);
2093 :
2094 221 : free_root_extent_buffers(info->dev_root);
2095 221 : free_root_extent_buffers(info->extent_root);
2096 221 : free_root_extent_buffers(info->csum_root);
2097 221 : free_root_extent_buffers(info->quota_root);
2098 221 : free_root_extent_buffers(info->uuid_root);
2099 221 : if (chunk_root)
2100 221 : free_root_extent_buffers(info->chunk_root);
2101 221 : }
2102 :
2103 221 : void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info)
2104 : {
2105 : int ret;
2106 : struct btrfs_root *gang[8];
2107 : int i;
2108 :
2109 851 : while (!list_empty(&fs_info->dead_roots)) {
2110 94 : gang[0] = list_entry(fs_info->dead_roots.next,
2111 : struct btrfs_root, root_list);
2112 94 : list_del(&gang[0]->root_list);
2113 :
2114 188 : if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) {
2115 94 : btrfs_drop_and_free_fs_root(fs_info, gang[0]);
2116 : } else {
2117 0 : free_extent_buffer(gang[0]->node);
2118 0 : free_extent_buffer(gang[0]->commit_root);
2119 0 : btrfs_put_fs_root(gang[0]);
2120 : }
2121 : }
2122 :
2123 : while (1) {
2124 450 : ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
2125 : (void **)gang, 0,
2126 : ARRAY_SIZE(gang));
2127 450 : if (!ret)
2128 : break;
2129 583 : for (i = 0; i < ret; i++)
2130 583 : btrfs_drop_and_free_fs_root(fs_info, gang[i]);
2131 : }
2132 :
2133 221 : if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2134 0 : btrfs_free_log_root_tree(NULL, fs_info);
2135 0 : btrfs_destroy_pinned_extent(fs_info->tree_root,
2136 : fs_info->pinned_extents);
2137 : }
2138 221 : }
2139 :
2140 221 : int open_ctree(struct super_block *sb,
2141 : struct btrfs_fs_devices *fs_devices,
2142 : char *options)
2143 : {
2144 : u32 sectorsize;
2145 : u32 nodesize;
2146 : u32 leafsize;
2147 : u32 blocksize;
2148 : u32 stripesize;
2149 : u64 generation;
2150 : u64 features;
2151 : struct btrfs_key location;
2152 : struct buffer_head *bh;
2153 442 : struct btrfs_super_block *disk_super;
2154 : struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2155 442 : struct btrfs_root *tree_root;
2156 : struct btrfs_root *extent_root;
2157 : struct btrfs_root *csum_root;
2158 : struct btrfs_root *chunk_root;
2159 : struct btrfs_root *dev_root;
2160 : struct btrfs_root *quota_root;
2161 : struct btrfs_root *uuid_root;
2162 : struct btrfs_root *log_tree_root;
2163 : int ret;
2164 : int err = -EINVAL;
2165 221 : int num_backups_tried = 0;
2166 221 : int backup_index = 0;
2167 : int max_active;
2168 : int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND;
2169 : bool create_uuid_tree;
2170 : bool check_uuid_tree;
2171 :
2172 221 : tree_root = fs_info->tree_root = btrfs_alloc_root(fs_info);
2173 221 : chunk_root = fs_info->chunk_root = btrfs_alloc_root(fs_info);
2174 221 : if (!tree_root || !chunk_root) {
2175 : err = -ENOMEM;
2176 : goto fail;
2177 : }
2178 :
2179 221 : ret = init_srcu_struct(&fs_info->subvol_srcu);
2180 221 : if (ret) {
2181 : err = ret;
2182 : goto fail;
2183 : }
2184 :
2185 221 : ret = setup_bdi(fs_info, &fs_info->bdi);
2186 221 : if (ret) {
2187 : err = ret;
2188 : goto fail_srcu;
2189 : }
2190 :
2191 221 : ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
2192 221 : if (ret) {
2193 : err = ret;
2194 : goto fail_bdi;
2195 : }
2196 442 : fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
2197 442 : (1 + ilog2(nr_cpu_ids));
2198 :
2199 221 : ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
2200 221 : if (ret) {
2201 : err = ret;
2202 : goto fail_dirty_metadata_bytes;
2203 : }
2204 :
2205 221 : ret = percpu_counter_init(&fs_info->bio_counter, 0);
2206 221 : if (ret) {
2207 : err = ret;
2208 : goto fail_delalloc_bytes;
2209 : }
2210 :
2211 221 : fs_info->btree_inode = new_inode(sb);
2212 221 : if (!fs_info->btree_inode) {
2213 : err = -ENOMEM;
2214 : goto fail_bio_counter;
2215 : }
2216 :
2217 221 : mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
2218 :
2219 221 : INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC);
2220 221 : INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC);
2221 221 : INIT_LIST_HEAD(&fs_info->trans_list);
2222 221 : INIT_LIST_HEAD(&fs_info->dead_roots);
2223 221 : INIT_LIST_HEAD(&fs_info->delayed_iputs);
2224 221 : INIT_LIST_HEAD(&fs_info->delalloc_roots);
2225 221 : INIT_LIST_HEAD(&fs_info->caching_block_groups);
2226 221 : spin_lock_init(&fs_info->delalloc_root_lock);
2227 221 : spin_lock_init(&fs_info->trans_lock);
2228 221 : spin_lock_init(&fs_info->fs_roots_radix_lock);
2229 221 : spin_lock_init(&fs_info->delayed_iput_lock);
2230 221 : spin_lock_init(&fs_info->defrag_inodes_lock);
2231 221 : spin_lock_init(&fs_info->free_chunk_lock);
2232 221 : spin_lock_init(&fs_info->tree_mod_seq_lock);
2233 221 : spin_lock_init(&fs_info->super_lock);
2234 221 : spin_lock_init(&fs_info->qgroup_op_lock);
2235 221 : spin_lock_init(&fs_info->buffer_lock);
2236 221 : rwlock_init(&fs_info->tree_mod_log_lock);
2237 221 : mutex_init(&fs_info->reloc_mutex);
2238 221 : mutex_init(&fs_info->delalloc_root_mutex);
2239 221 : seqlock_init(&fs_info->profiles_lock);
2240 :
2241 : init_completion(&fs_info->kobj_unregister);
2242 221 : INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
2243 221 : INIT_LIST_HEAD(&fs_info->space_info);
2244 221 : INIT_LIST_HEAD(&fs_info->tree_mod_seq_list);
2245 221 : btrfs_mapping_init(&fs_info->mapping_tree);
2246 221 : btrfs_init_block_rsv(&fs_info->global_block_rsv,
2247 : BTRFS_BLOCK_RSV_GLOBAL);
2248 221 : btrfs_init_block_rsv(&fs_info->delalloc_block_rsv,
2249 : BTRFS_BLOCK_RSV_DELALLOC);
2250 221 : btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS);
2251 221 : btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK);
2252 221 : btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY);
2253 221 : btrfs_init_block_rsv(&fs_info->delayed_block_rsv,
2254 : BTRFS_BLOCK_RSV_DELOPS);
2255 : atomic_set(&fs_info->nr_async_submits, 0);
2256 : atomic_set(&fs_info->async_delalloc_pages, 0);
2257 : atomic_set(&fs_info->async_submit_draining, 0);
2258 : atomic_set(&fs_info->nr_async_bios, 0);
2259 : atomic_set(&fs_info->defrag_running, 0);
2260 : atomic_set(&fs_info->qgroup_op_seq, 0);
2261 : atomic64_set(&fs_info->tree_mod_seq, 0);
2262 221 : fs_info->sb = sb;
2263 221 : fs_info->max_inline = 8192 * 1024;
2264 221 : fs_info->metadata_ratio = 0;
2265 221 : fs_info->defrag_inodes = RB_ROOT;
2266 221 : fs_info->free_chunk_space = 0;
2267 221 : fs_info->tree_mod_log = RB_ROOT;
2268 221 : fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2269 221 : fs_info->avg_delayed_ref_runtime = div64_u64(NSEC_PER_SEC, 64);
2270 : /* readahead state */
2271 221 : INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
2272 221 : spin_lock_init(&fs_info->reada_lock);
2273 :
2274 442 : fs_info->thread_pool_size = min_t(unsigned long,
2275 : num_online_cpus() + 2, 8);
2276 :
2277 221 : INIT_LIST_HEAD(&fs_info->ordered_roots);
2278 221 : spin_lock_init(&fs_info->ordered_root_lock);
2279 221 : fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root),
2280 : GFP_NOFS);
2281 221 : if (!fs_info->delayed_root) {
2282 : err = -ENOMEM;
2283 : goto fail_iput;
2284 : }
2285 221 : btrfs_init_delayed_root(fs_info->delayed_root);
2286 :
2287 221 : mutex_init(&fs_info->scrub_lock);
2288 : atomic_set(&fs_info->scrubs_running, 0);
2289 : atomic_set(&fs_info->scrub_pause_req, 0);
2290 : atomic_set(&fs_info->scrubs_paused, 0);
2291 : atomic_set(&fs_info->scrub_cancel_req, 0);
2292 221 : init_waitqueue_head(&fs_info->replace_wait);
2293 221 : init_waitqueue_head(&fs_info->scrub_pause_wait);
2294 221 : fs_info->scrub_workers_refcnt = 0;
2295 : #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2296 : fs_info->check_integrity_print_mask = 0;
2297 : #endif
2298 :
2299 221 : spin_lock_init(&fs_info->balance_lock);
2300 221 : mutex_init(&fs_info->balance_mutex);
2301 : atomic_set(&fs_info->balance_running, 0);
2302 : atomic_set(&fs_info->balance_pause_req, 0);
2303 : atomic_set(&fs_info->balance_cancel_req, 0);
2304 221 : fs_info->balance_ctl = NULL;
2305 221 : init_waitqueue_head(&fs_info->balance_wait_q);
2306 221 : btrfs_init_async_reclaim_work(&fs_info->async_reclaim_work);
2307 :
2308 221 : sb->s_blocksize = 4096;
2309 221 : sb->s_blocksize_bits = blksize_bits(4096);
2310 221 : sb->s_bdi = &fs_info->bdi;
2311 :
2312 221 : fs_info->btree_inode->i_ino = BTRFS_BTREE_INODE_OBJECTID;
2313 221 : set_nlink(fs_info->btree_inode, 1);
2314 : /*
2315 : * we set the i_size on the btree inode to the max possible int.
2316 : * the real end of the address space is determined by all of
2317 : * the devices in the system
2318 : */
2319 221 : fs_info->btree_inode->i_size = OFFSET_MAX;
2320 221 : fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
2321 221 : fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi;
2322 :
2323 221 : RB_CLEAR_NODE(&BTRFS_I(fs_info->btree_inode)->rb_node);
2324 221 : extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree,
2325 221 : fs_info->btree_inode->i_mapping);
2326 221 : BTRFS_I(fs_info->btree_inode)->io_tree.track_uptodate = 0;
2327 221 : extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree);
2328 :
2329 221 : BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops;
2330 :
2331 221 : BTRFS_I(fs_info->btree_inode)->root = tree_root;
2332 221 : memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
2333 : sizeof(struct btrfs_key));
2334 : set_bit(BTRFS_INODE_DUMMY,
2335 221 : &BTRFS_I(fs_info->btree_inode)->runtime_flags);
2336 221 : btrfs_insert_inode_hash(fs_info->btree_inode);
2337 :
2338 221 : spin_lock_init(&fs_info->block_group_cache_lock);
2339 221 : fs_info->block_group_cache_tree = RB_ROOT;
2340 221 : fs_info->first_logical_byte = (u64)-1;
2341 :
2342 221 : extent_io_tree_init(&fs_info->freed_extents[0],
2343 221 : fs_info->btree_inode->i_mapping);
2344 221 : extent_io_tree_init(&fs_info->freed_extents[1],
2345 221 : fs_info->btree_inode->i_mapping);
2346 221 : fs_info->pinned_extents = &fs_info->freed_extents[0];
2347 221 : fs_info->do_barriers = 1;
2348 :
2349 :
2350 221 : mutex_init(&fs_info->ordered_operations_mutex);
2351 221 : mutex_init(&fs_info->ordered_extent_flush_mutex);
2352 221 : mutex_init(&fs_info->tree_log_mutex);
2353 221 : mutex_init(&fs_info->chunk_mutex);
2354 221 : mutex_init(&fs_info->transaction_kthread_mutex);
2355 221 : mutex_init(&fs_info->cleaner_mutex);
2356 221 : mutex_init(&fs_info->volume_mutex);
2357 221 : init_rwsem(&fs_info->commit_root_sem);
2358 221 : init_rwsem(&fs_info->cleanup_work_sem);
2359 221 : init_rwsem(&fs_info->subvol_sem);
2360 : sema_init(&fs_info->uuid_tree_rescan_sem, 1);
2361 221 : fs_info->dev_replace.lock_owner = 0;
2362 : atomic_set(&fs_info->dev_replace.nesting_level, 0);
2363 221 : mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount);
2364 221 : mutex_init(&fs_info->dev_replace.lock_management_lock);
2365 221 : mutex_init(&fs_info->dev_replace.lock);
2366 :
2367 221 : spin_lock_init(&fs_info->qgroup_lock);
2368 221 : mutex_init(&fs_info->qgroup_ioctl_lock);
2369 221 : fs_info->qgroup_tree = RB_ROOT;
2370 221 : fs_info->qgroup_op_tree = RB_ROOT;
2371 221 : INIT_LIST_HEAD(&fs_info->dirty_qgroups);
2372 221 : fs_info->qgroup_seq = 1;
2373 221 : fs_info->quota_enabled = 0;
2374 221 : fs_info->pending_quota_state = 0;
2375 221 : fs_info->qgroup_ulist = NULL;
2376 221 : mutex_init(&fs_info->qgroup_rescan_lock);
2377 :
2378 221 : btrfs_init_free_cluster(&fs_info->meta_alloc_cluster);
2379 221 : btrfs_init_free_cluster(&fs_info->data_alloc_cluster);
2380 :
2381 221 : init_waitqueue_head(&fs_info->transaction_throttle);
2382 221 : init_waitqueue_head(&fs_info->transaction_wait);
2383 221 : init_waitqueue_head(&fs_info->transaction_blocked_wait);
2384 221 : init_waitqueue_head(&fs_info->async_submit_wait);
2385 :
2386 221 : ret = btrfs_alloc_stripe_hash_table(fs_info);
2387 221 : if (ret) {
2388 : err = ret;
2389 : goto fail_alloc;
2390 : }
2391 :
2392 221 : __setup_root(4096, 4096, 4096, 4096, tree_root,
2393 : fs_info, BTRFS_ROOT_TREE_OBJECTID);
2394 :
2395 221 : invalidate_bdev(fs_devices->latest_bdev);
2396 :
2397 : /*
2398 : * Read super block and check the signature bytes only
2399 : */
2400 221 : bh = btrfs_read_dev_super(fs_devices->latest_bdev);
2401 221 : if (!bh) {
2402 : err = -EINVAL;
2403 : goto fail_alloc;
2404 : }
2405 :
2406 : /*
2407 : * We want to check superblock checksum, the type is stored inside.
2408 : * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k).
2409 : */
2410 221 : if (btrfs_check_super_csum(bh->b_data)) {
2411 0 : printk(KERN_ERR "BTRFS: superblock checksum mismatch\n");
2412 : err = -EINVAL;
2413 0 : goto fail_alloc;
2414 : }
2415 :
2416 : /*
2417 : * super_copy is zeroed at allocation time and we never touch the
2418 : * following bytes up to INFO_SIZE, the checksum is calculated from
2419 : * the whole block of INFO_SIZE
2420 : */
2421 221 : memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
2422 221 : memcpy(fs_info->super_for_commit, fs_info->super_copy,
2423 : sizeof(*fs_info->super_for_commit));
2424 : brelse(bh);
2425 :
2426 221 : memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
2427 :
2428 : ret = btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
2429 : if (ret) {
2430 : printk(KERN_ERR "BTRFS: superblock contains fatal errors\n");
2431 : err = -EINVAL;
2432 : goto fail_alloc;
2433 : }
2434 :
2435 221 : disk_super = fs_info->super_copy;
2436 221 : if (!btrfs_super_root(disk_super))
2437 : goto fail_alloc;
2438 :
2439 : /* check FS state, whether FS is broken. */
2440 221 : if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR)
2441 : set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
2442 :
2443 : /*
2444 : * run through our array of backup supers and setup
2445 : * our ring pointer to the oldest one
2446 : */
2447 : generation = btrfs_super_generation(disk_super);
2448 221 : find_oldest_super_backup(fs_info, generation);
2449 :
2450 : /*
2451 : * In the long term, we'll store the compression type in the super
2452 : * block, and it'll be used for per file compression control.
2453 : */
2454 221 : fs_info->compress_type = BTRFS_COMPRESS_ZLIB;
2455 :
2456 221 : ret = btrfs_parse_options(tree_root, options);
2457 221 : if (ret) {
2458 : err = ret;
2459 : goto fail_alloc;
2460 : }
2461 :
2462 221 : features = btrfs_super_incompat_flags(disk_super) &
2463 : ~BTRFS_FEATURE_INCOMPAT_SUPP;
2464 221 : if (features) {
2465 0 : printk(KERN_ERR "BTRFS: couldn't mount because of "
2466 : "unsupported optional features (%Lx).\n",
2467 : features);
2468 : err = -EINVAL;
2469 0 : goto fail_alloc;
2470 : }
2471 :
2472 221 : if (btrfs_super_leafsize(disk_super) !=
2473 : btrfs_super_nodesize(disk_super)) {
2474 0 : printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2475 : "blocksizes don't match. node %d leaf %d\n",
2476 : btrfs_super_nodesize(disk_super),
2477 : btrfs_super_leafsize(disk_super));
2478 : err = -EINVAL;
2479 0 : goto fail_alloc;
2480 : }
2481 221 : if (btrfs_super_leafsize(disk_super) > BTRFS_MAX_METADATA_BLOCKSIZE) {
2482 0 : printk(KERN_ERR "BTRFS: couldn't mount because metadata "
2483 : "blocksize (%d) was too large\n",
2484 : btrfs_super_leafsize(disk_super));
2485 : err = -EINVAL;
2486 0 : goto fail_alloc;
2487 : }
2488 :
2489 : features = btrfs_super_incompat_flags(disk_super);
2490 221 : features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
2491 221 : if (tree_root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
2492 4 : features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
2493 :
2494 221 : if (features & BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA)
2495 0 : printk(KERN_ERR "BTRFS: has skinny extents\n");
2496 :
2497 : /*
2498 : * flag our filesystem as having big metadata blocks if
2499 : * they are bigger than the page size
2500 : */
2501 221 : if (btrfs_super_leafsize(disk_super) > PAGE_CACHE_SIZE) {
2502 214 : if (!(features & BTRFS_FEATURE_INCOMPAT_BIG_METADATA))
2503 95 : printk(KERN_INFO "BTRFS: flagging fs with big metadata feature\n");
2504 214 : features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA;
2505 : }
2506 :
2507 : nodesize = btrfs_super_nodesize(disk_super);
2508 : leafsize = btrfs_super_leafsize(disk_super);
2509 : sectorsize = btrfs_super_sectorsize(disk_super);
2510 : stripesize = btrfs_super_stripesize(disk_super);
2511 442 : fs_info->dirty_metadata_batch = leafsize * (1 + ilog2(nr_cpu_ids));
2512 442 : fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids));
2513 :
2514 : /*
2515 : * mixed block groups end up with duplicate but slightly offset
2516 : * extent buffers for the same range. It leads to corruptions
2517 : */
2518 221 : if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) &&
2519 : (sectorsize != leafsize)) {
2520 0 : printk(KERN_WARNING "BTRFS: unequal leaf/node/sector sizes "
2521 : "are not allowed for mixed block groups on %s\n",
2522 0 : sb->s_id);
2523 0 : goto fail_alloc;
2524 : }
2525 :
2526 : /*
2527 : * Needn't use the lock because there is no other task which will
2528 : * update the flag.
2529 : */
2530 : btrfs_set_super_incompat_flags(disk_super, features);
2531 :
2532 : features = btrfs_super_compat_ro_flags(disk_super) &
2533 : ~BTRFS_FEATURE_COMPAT_RO_SUPP;
2534 221 : if (!(sb->s_flags & MS_RDONLY) && features) {
2535 0 : printk(KERN_ERR "BTRFS: couldn't mount RDWR because of "
2536 : "unsupported option features (%Lx).\n",
2537 : features);
2538 : err = -EINVAL;
2539 0 : goto fail_alloc;
2540 : }
2541 :
2542 221 : max_active = fs_info->thread_pool_size;
2543 :
2544 221 : fs_info->workers =
2545 221 : btrfs_alloc_workqueue("worker", flags | WQ_HIGHPRI,
2546 : max_active, 16);
2547 :
2548 221 : fs_info->delalloc_workers =
2549 221 : btrfs_alloc_workqueue("delalloc", flags, max_active, 2);
2550 :
2551 221 : fs_info->flush_workers =
2552 221 : btrfs_alloc_workqueue("flush_delalloc", flags, max_active, 0);
2553 :
2554 221 : fs_info->caching_workers =
2555 221 : btrfs_alloc_workqueue("cache", flags, max_active, 0);
2556 :
2557 : /*
2558 : * a higher idle thresh on the submit workers makes it much more
2559 : * likely that bios will be send down in a sane order to the
2560 : * devices
2561 : */
2562 221 : fs_info->submit_workers =
2563 221 : btrfs_alloc_workqueue("submit", flags,
2564 221 : min_t(u64, fs_devices->num_devices,
2565 : max_active), 64);
2566 :
2567 221 : fs_info->fixup_workers =
2568 221 : btrfs_alloc_workqueue("fixup", flags, 1, 0);
2569 :
2570 : /*
2571 : * endios are largely parallel and should have a very
2572 : * low idle thresh
2573 : */
2574 221 : fs_info->endio_workers =
2575 221 : btrfs_alloc_workqueue("endio", flags, max_active, 4);
2576 221 : fs_info->endio_meta_workers =
2577 221 : btrfs_alloc_workqueue("endio-meta", flags, max_active, 4);
2578 221 : fs_info->endio_meta_write_workers =
2579 221 : btrfs_alloc_workqueue("endio-meta-write", flags, max_active, 2);
2580 221 : fs_info->endio_raid56_workers =
2581 221 : btrfs_alloc_workqueue("endio-raid56", flags, max_active, 4);
2582 221 : fs_info->rmw_workers =
2583 221 : btrfs_alloc_workqueue("rmw", flags, max_active, 2);
2584 221 : fs_info->endio_write_workers =
2585 221 : btrfs_alloc_workqueue("endio-write", flags, max_active, 2);
2586 221 : fs_info->endio_freespace_worker =
2587 221 : btrfs_alloc_workqueue("freespace-write", flags, max_active, 0);
2588 221 : fs_info->delayed_workers =
2589 221 : btrfs_alloc_workqueue("delayed-meta", flags, max_active, 0);
2590 221 : fs_info->readahead_workers =
2591 221 : btrfs_alloc_workqueue("readahead", flags, max_active, 2);
2592 221 : fs_info->qgroup_rescan_workers =
2593 221 : btrfs_alloc_workqueue("qgroup-rescan", flags, 1, 0);
2594 221 : fs_info->extent_workers =
2595 221 : btrfs_alloc_workqueue("extent-refs", flags,
2596 221 : min_t(u64, fs_devices->num_devices,
2597 : max_active), 8);
2598 :
2599 663 : if (!(fs_info->workers && fs_info->delalloc_workers &&
2600 663 : fs_info->submit_workers && fs_info->flush_workers &&
2601 663 : fs_info->endio_workers && fs_info->endio_meta_workers &&
2602 442 : fs_info->endio_meta_write_workers &&
2603 663 : fs_info->endio_write_workers && fs_info->endio_raid56_workers &&
2604 663 : fs_info->endio_freespace_worker && fs_info->rmw_workers &&
2605 663 : fs_info->caching_workers && fs_info->readahead_workers &&
2606 442 : fs_info->fixup_workers && fs_info->delayed_workers &&
2607 221 : fs_info->fixup_workers && fs_info->extent_workers &&
2608 221 : fs_info->qgroup_rescan_workers)) {
2609 : err = -ENOMEM;
2610 : goto fail_sb_buffer;
2611 : }
2612 :
2613 442 : fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
2614 221 : fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
2615 : 4 * 1024 * 1024 / PAGE_CACHE_SIZE);
2616 :
2617 221 : tree_root->nodesize = nodesize;
2618 221 : tree_root->leafsize = leafsize;
2619 221 : tree_root->sectorsize = sectorsize;
2620 221 : tree_root->stripesize = stripesize;
2621 :
2622 221 : sb->s_blocksize = sectorsize;
2623 221 : sb->s_blocksize_bits = blksize_bits(sectorsize);
2624 :
2625 221 : if (btrfs_super_magic(disk_super) != BTRFS_MAGIC) {
2626 0 : printk(KERN_INFO "BTRFS: valid FS not found on %s\n", sb->s_id);
2627 0 : goto fail_sb_buffer;
2628 : }
2629 :
2630 221 : if (sectorsize != PAGE_SIZE) {
2631 0 : printk(KERN_WARNING "BTRFS: Incompatible sector size(%lu) "
2632 0 : "found on %s\n", (unsigned long)sectorsize, sb->s_id);
2633 0 : goto fail_sb_buffer;
2634 : }
2635 :
2636 221 : mutex_lock(&fs_info->chunk_mutex);
2637 221 : ret = btrfs_read_sys_array(tree_root);
2638 221 : mutex_unlock(&fs_info->chunk_mutex);
2639 221 : if (ret) {
2640 0 : printk(KERN_WARNING "BTRFS: failed to read the system "
2641 0 : "array on %s\n", sb->s_id);
2642 0 : goto fail_sb_buffer;
2643 : }
2644 :
2645 : blocksize = btrfs_level_size(tree_root,
2646 : btrfs_super_chunk_root_level(disk_super));
2647 : generation = btrfs_super_chunk_root_generation(disk_super);
2648 :
2649 221 : __setup_root(nodesize, leafsize, sectorsize, stripesize,
2650 : chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
2651 :
2652 221 : chunk_root->node = read_tree_block(chunk_root,
2653 : btrfs_super_chunk_root(disk_super),
2654 : blocksize, generation);
2655 442 : if (!chunk_root->node ||
2656 : !test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
2657 0 : printk(KERN_WARNING "BTRFS: failed to read chunk root on %s\n",
2658 0 : sb->s_id);
2659 0 : goto fail_tree_roots;
2660 : }
2661 221 : btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
2662 221 : chunk_root->commit_root = btrfs_root_node(chunk_root);
2663 :
2664 221 : read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid,
2665 : btrfs_header_chunk_tree_uuid(chunk_root->node), BTRFS_UUID_SIZE);
2666 :
2667 221 : ret = btrfs_read_chunk_tree(chunk_root);
2668 221 : if (ret) {
2669 0 : printk(KERN_WARNING "BTRFS: failed to read chunk tree on %s\n",
2670 0 : sb->s_id);
2671 0 : goto fail_tree_roots;
2672 : }
2673 :
2674 : /*
2675 : * keep the device that is marked to be the target device for the
2676 : * dev_replace procedure
2677 : */
2678 221 : btrfs_close_extra_devices(fs_info, fs_devices, 0);
2679 :
2680 221 : if (!fs_devices->latest_bdev) {
2681 0 : printk(KERN_CRIT "BTRFS: failed to read devices on %s\n",
2682 0 : sb->s_id);
2683 0 : goto fail_tree_roots;
2684 : }
2685 :
2686 : retry_root_backup:
2687 : blocksize = btrfs_level_size(tree_root,
2688 : btrfs_super_root_level(disk_super));
2689 : generation = btrfs_super_generation(disk_super);
2690 :
2691 221 : tree_root->node = read_tree_block(tree_root,
2692 : btrfs_super_root(disk_super),
2693 : blocksize, generation);
2694 442 : if (!tree_root->node ||
2695 : !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
2696 0 : printk(KERN_WARNING "BTRFS: failed to read tree root on %s\n",
2697 0 : sb->s_id);
2698 :
2699 0 : goto recovery_tree_root;
2700 : }
2701 :
2702 221 : btrfs_set_root_node(&tree_root->root_item, tree_root->node);
2703 221 : tree_root->commit_root = btrfs_root_node(tree_root);
2704 : btrfs_set_root_refs(&tree_root->root_item, 1);
2705 :
2706 221 : location.objectid = BTRFS_EXTENT_TREE_OBJECTID;
2707 221 : location.type = BTRFS_ROOT_ITEM_KEY;
2708 221 : location.offset = 0;
2709 :
2710 221 : extent_root = btrfs_read_tree_root(tree_root, &location);
2711 221 : if (IS_ERR(extent_root)) {
2712 : ret = PTR_ERR(extent_root);
2713 : goto recovery_tree_root;
2714 : }
2715 : set_bit(BTRFS_ROOT_TRACK_DIRTY, &extent_root->state);
2716 221 : fs_info->extent_root = extent_root;
2717 :
2718 221 : location.objectid = BTRFS_DEV_TREE_OBJECTID;
2719 221 : dev_root = btrfs_read_tree_root(tree_root, &location);
2720 221 : if (IS_ERR(dev_root)) {
2721 : ret = PTR_ERR(dev_root);
2722 : goto recovery_tree_root;
2723 : }
2724 : set_bit(BTRFS_ROOT_TRACK_DIRTY, &dev_root->state);
2725 221 : fs_info->dev_root = dev_root;
2726 221 : btrfs_init_devices_late(fs_info);
2727 :
2728 221 : location.objectid = BTRFS_CSUM_TREE_OBJECTID;
2729 221 : csum_root = btrfs_read_tree_root(tree_root, &location);
2730 221 : if (IS_ERR(csum_root)) {
2731 : ret = PTR_ERR(csum_root);
2732 : goto recovery_tree_root;
2733 : }
2734 : set_bit(BTRFS_ROOT_TRACK_DIRTY, &csum_root->state);
2735 221 : fs_info->csum_root = csum_root;
2736 :
2737 221 : location.objectid = BTRFS_QUOTA_TREE_OBJECTID;
2738 221 : quota_root = btrfs_read_tree_root(tree_root, &location);
2739 221 : if (!IS_ERR(quota_root)) {
2740 : set_bit(BTRFS_ROOT_TRACK_DIRTY, "a_root->state);
2741 3 : fs_info->quota_enabled = 1;
2742 3 : fs_info->pending_quota_state = 1;
2743 3 : fs_info->quota_root = quota_root;
2744 : }
2745 :
2746 221 : location.objectid = BTRFS_UUID_TREE_OBJECTID;
2747 221 : uuid_root = btrfs_read_tree_root(tree_root, &location);
2748 221 : if (IS_ERR(uuid_root)) {
2749 99 : ret = PTR_ERR(uuid_root);
2750 99 : if (ret != -ENOENT)
2751 : goto recovery_tree_root;
2752 : create_uuid_tree = true;
2753 : check_uuid_tree = false;
2754 : } else {
2755 : set_bit(BTRFS_ROOT_TRACK_DIRTY, &uuid_root->state);
2756 122 : fs_info->uuid_root = uuid_root;
2757 : create_uuid_tree = false;
2758 122 : check_uuid_tree =
2759 : generation != btrfs_super_uuid_tree_generation(disk_super);
2760 : }
2761 :
2762 221 : fs_info->generation = generation;
2763 221 : fs_info->last_trans_committed = generation;
2764 :
2765 221 : ret = btrfs_recover_balance(fs_info);
2766 221 : if (ret) {
2767 0 : printk(KERN_WARNING "BTRFS: failed to recover balance\n");
2768 0 : goto fail_block_groups;
2769 : }
2770 :
2771 221 : ret = btrfs_init_dev_stats(fs_info);
2772 221 : if (ret) {
2773 0 : printk(KERN_ERR "BTRFS: failed to init dev_stats: %d\n",
2774 : ret);
2775 0 : goto fail_block_groups;
2776 : }
2777 :
2778 221 : ret = btrfs_init_dev_replace(fs_info);
2779 221 : if (ret) {
2780 0 : pr_err("BTRFS: failed to init dev_replace: %d\n", ret);
2781 0 : goto fail_block_groups;
2782 : }
2783 :
2784 221 : btrfs_close_extra_devices(fs_info, fs_devices, 1);
2785 :
2786 221 : ret = btrfs_sysfs_add_one(fs_info);
2787 221 : if (ret) {
2788 0 : pr_err("BTRFS: failed to init sysfs interface: %d\n", ret);
2789 0 : goto fail_block_groups;
2790 : }
2791 :
2792 221 : ret = btrfs_init_space_info(fs_info);
2793 221 : if (ret) {
2794 0 : printk(KERN_ERR "BTRFS: Failed to initial space info: %d\n", ret);
2795 0 : goto fail_sysfs;
2796 : }
2797 :
2798 221 : ret = btrfs_read_block_groups(extent_root);
2799 221 : if (ret) {
2800 0 : printk(KERN_ERR "BTRFS: Failed to read block groups: %d\n", ret);
2801 0 : goto fail_sysfs;
2802 : }
2803 221 : fs_info->num_tolerated_disk_barrier_failures =
2804 221 : btrfs_calc_num_tolerated_disk_barrier_failures(fs_info);
2805 221 : if (fs_info->fs_devices->missing_devices >
2806 0 : fs_info->num_tolerated_disk_barrier_failures &&
2807 0 : !(sb->s_flags & MS_RDONLY)) {
2808 0 : printk(KERN_WARNING "BTRFS: "
2809 : "too many missing devices, writeable mount is not allowed\n");
2810 0 : goto fail_sysfs;
2811 : }
2812 :
2813 442 : fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
2814 : "btrfs-cleaner");
2815 221 : if (IS_ERR(fs_info->cleaner_kthread))
2816 : goto fail_sysfs;
2817 :
2818 442 : fs_info->transaction_kthread = kthread_run(transaction_kthread,
2819 : tree_root,
2820 : "btrfs-transaction");
2821 221 : if (IS_ERR(fs_info->transaction_kthread))
2822 : goto fail_cleaner;
2823 :
2824 221 : if (!btrfs_test_opt(tree_root, SSD) &&
2825 221 : !btrfs_test_opt(tree_root, NOSSD) &&
2826 221 : !fs_info->fs_devices->rotating) {
2827 0 : printk(KERN_INFO "BTRFS: detected SSD devices, enabling SSD "
2828 : "mode\n");
2829 0 : btrfs_set_opt(fs_info->mount_opt, SSD);
2830 : }
2831 :
2832 : /* Set the real inode map cache flag */
2833 221 : if (btrfs_test_opt(tree_root, CHANGE_INODE_CACHE))
2834 0 : btrfs_set_opt(tree_root->fs_info->mount_opt, INODE_MAP_CACHE);
2835 :
2836 : #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2837 : if (btrfs_test_opt(tree_root, CHECK_INTEGRITY)) {
2838 : ret = btrfsic_mount(tree_root, fs_devices,
2839 : btrfs_test_opt(tree_root,
2840 : CHECK_INTEGRITY_INCLUDING_EXTENT_DATA) ?
2841 : 1 : 0,
2842 : fs_info->check_integrity_print_mask);
2843 : if (ret)
2844 : printk(KERN_WARNING "BTRFS: failed to initialize"
2845 : " integrity check module %s\n", sb->s_id);
2846 : }
2847 : #endif
2848 221 : ret = btrfs_read_qgroup_config(fs_info);
2849 221 : if (ret)
2850 : goto fail_trans_kthread;
2851 :
2852 : /* do not make disk changes in broken FS */
2853 221 : if (btrfs_super_log_root(disk_super) != 0) {
2854 : u64 bytenr = btrfs_super_log_root(disk_super);
2855 :
2856 0 : if (fs_devices->rw_devices == 0) {
2857 0 : printk(KERN_WARNING "BTRFS: log replay required "
2858 : "on RO media\n");
2859 : err = -EIO;
2860 0 : goto fail_qgroup;
2861 : }
2862 : blocksize =
2863 : btrfs_level_size(tree_root,
2864 : btrfs_super_log_root_level(disk_super));
2865 :
2866 : log_tree_root = btrfs_alloc_root(fs_info);
2867 0 : if (!log_tree_root) {
2868 : err = -ENOMEM;
2869 : goto fail_qgroup;
2870 : }
2871 :
2872 0 : __setup_root(nodesize, leafsize, sectorsize, stripesize,
2873 : log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
2874 :
2875 0 : log_tree_root->node = read_tree_block(tree_root, bytenr,
2876 : blocksize,
2877 : generation + 1);
2878 0 : if (!log_tree_root->node ||
2879 0 : !extent_buffer_uptodate(log_tree_root->node)) {
2880 0 : printk(KERN_ERR "BTRFS: failed to read log tree\n");
2881 0 : free_extent_buffer(log_tree_root->node);
2882 0 : kfree(log_tree_root);
2883 0 : goto fail_qgroup;
2884 : }
2885 : /* returns with log_tree_root freed on success */
2886 0 : ret = btrfs_recover_log_trees(log_tree_root);
2887 0 : if (ret) {
2888 0 : btrfs_error(tree_root->fs_info, ret,
2889 : "Failed to recover log tree");
2890 0 : free_extent_buffer(log_tree_root->node);
2891 0 : kfree(log_tree_root);
2892 0 : goto fail_qgroup;
2893 : }
2894 :
2895 0 : if (sb->s_flags & MS_RDONLY) {
2896 0 : ret = btrfs_commit_super(tree_root);
2897 0 : if (ret)
2898 : goto fail_qgroup;
2899 : }
2900 : }
2901 :
2902 221 : ret = btrfs_find_orphan_roots(tree_root);
2903 221 : if (ret)
2904 : goto fail_qgroup;
2905 :
2906 221 : if (!(sb->s_flags & MS_RDONLY)) {
2907 192 : ret = btrfs_cleanup_fs_roots(fs_info);
2908 192 : if (ret)
2909 : goto fail_qgroup;
2910 :
2911 192 : mutex_lock(&fs_info->cleaner_mutex);
2912 192 : ret = btrfs_recover_relocation(tree_root);
2913 192 : mutex_unlock(&fs_info->cleaner_mutex);
2914 192 : if (ret < 0) {
2915 0 : printk(KERN_WARNING
2916 : "BTRFS: failed to recover relocation\n");
2917 : err = -EINVAL;
2918 0 : goto fail_qgroup;
2919 : }
2920 : }
2921 :
2922 221 : location.objectid = BTRFS_FS_TREE_OBJECTID;
2923 221 : location.type = BTRFS_ROOT_ITEM_KEY;
2924 221 : location.offset = 0;
2925 :
2926 221 : fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location);
2927 221 : if (IS_ERR(fs_info->fs_root)) {
2928 0 : err = PTR_ERR(fs_info->fs_root);
2929 0 : goto fail_qgroup;
2930 : }
2931 :
2932 221 : if (sb->s_flags & MS_RDONLY)
2933 : return 0;
2934 :
2935 192 : down_read(&fs_info->cleanup_work_sem);
2936 384 : if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) ||
2937 192 : (ret = btrfs_orphan_cleanup(fs_info->tree_root))) {
2938 0 : up_read(&fs_info->cleanup_work_sem);
2939 0 : close_ctree(tree_root);
2940 0 : return ret;
2941 : }
2942 192 : up_read(&fs_info->cleanup_work_sem);
2943 :
2944 192 : ret = btrfs_resume_balance_async(fs_info);
2945 192 : if (ret) {
2946 0 : printk(KERN_WARNING "BTRFS: failed to resume balance\n");
2947 0 : close_ctree(tree_root);
2948 0 : return ret;
2949 : }
2950 :
2951 192 : ret = btrfs_resume_dev_replace_async(fs_info);
2952 192 : if (ret) {
2953 0 : pr_warn("BTRFS: failed to resume dev_replace\n");
2954 0 : close_ctree(tree_root);
2955 0 : return ret;
2956 : }
2957 :
2958 192 : btrfs_qgroup_rescan_resume(fs_info);
2959 :
2960 192 : if (create_uuid_tree) {
2961 97 : pr_info("BTRFS: creating UUID tree\n");
2962 97 : ret = btrfs_create_uuid_tree(fs_info);
2963 97 : if (ret) {
2964 0 : pr_warn("BTRFS: failed to create the UUID tree %d\n",
2965 : ret);
2966 0 : close_ctree(tree_root);
2967 0 : return ret;
2968 : }
2969 190 : } else if (check_uuid_tree ||
2970 95 : btrfs_test_opt(tree_root, RESCAN_UUID_TREE)) {
2971 0 : pr_info("BTRFS: checking UUID tree\n");
2972 0 : ret = btrfs_check_uuid_tree(fs_info);
2973 0 : if (ret) {
2974 0 : pr_warn("BTRFS: failed to check the UUID tree %d\n",
2975 : ret);
2976 0 : close_ctree(tree_root);
2977 0 : return ret;
2978 : }
2979 : } else {
2980 95 : fs_info->update_uuid_tree_gen = 1;
2981 : }
2982 :
2983 : return 0;
2984 :
2985 : fail_qgroup:
2986 0 : btrfs_free_qgroup_config(fs_info);
2987 : fail_trans_kthread:
2988 0 : kthread_stop(fs_info->transaction_kthread);
2989 0 : btrfs_cleanup_transaction(fs_info->tree_root);
2990 0 : btrfs_free_fs_roots(fs_info);
2991 : fail_cleaner:
2992 0 : kthread_stop(fs_info->cleaner_kthread);
2993 :
2994 : /*
2995 : * make sure we're done with the btree inode before we stop our
2996 : * kthreads
2997 : */
2998 0 : filemap_write_and_wait(fs_info->btree_inode->i_mapping);
2999 :
3000 : fail_sysfs:
3001 0 : btrfs_sysfs_remove_one(fs_info);
3002 :
3003 : fail_block_groups:
3004 0 : btrfs_put_block_group_cache(fs_info);
3005 0 : btrfs_free_block_groups(fs_info);
3006 :
3007 : fail_tree_roots:
3008 0 : free_root_pointers(fs_info, 1);
3009 0 : invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
3010 :
3011 : fail_sb_buffer:
3012 0 : btrfs_stop_all_workers(fs_info);
3013 : fail_alloc:
3014 : fail_iput:
3015 0 : btrfs_mapping_tree_free(&fs_info->mapping_tree);
3016 :
3017 0 : iput(fs_info->btree_inode);
3018 : fail_bio_counter:
3019 0 : percpu_counter_destroy(&fs_info->bio_counter);
3020 : fail_delalloc_bytes:
3021 0 : percpu_counter_destroy(&fs_info->delalloc_bytes);
3022 : fail_dirty_metadata_bytes:
3023 0 : percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
3024 : fail_bdi:
3025 0 : bdi_destroy(&fs_info->bdi);
3026 : fail_srcu:
3027 0 : cleanup_srcu_struct(&fs_info->subvol_srcu);
3028 : fail:
3029 0 : btrfs_free_stripe_hash_table(fs_info);
3030 0 : btrfs_close_devices(fs_info->fs_devices);
3031 0 : return err;
3032 :
3033 : recovery_tree_root:
3034 0 : if (!btrfs_test_opt(tree_root, RECOVERY))
3035 : goto fail_tree_roots;
3036 :
3037 0 : free_root_pointers(fs_info, 0);
3038 :
3039 : /* don't use the log in recovery mode, it won't be valid */
3040 : btrfs_set_super_log_root(disk_super, 0);
3041 :
3042 : /* we can't trust the free space cache either */
3043 0 : btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
3044 :
3045 0 : ret = next_root_backup(fs_info, fs_info->super_copy,
3046 : &num_backups_tried, &backup_index);
3047 0 : if (ret == -1)
3048 : goto fail_block_groups;
3049 : goto retry_root_backup;
3050 : }
3051 :
3052 5835 : static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
3053 : {
3054 5835 : if (uptodate) {
3055 : set_buffer_uptodate(bh);
3056 : } else {
3057 0 : struct btrfs_device *device = (struct btrfs_device *)
3058 : bh->b_private;
3059 :
3060 0 : printk_ratelimited_in_rcu(KERN_WARNING "BTRFS: lost page write due to "
3061 : "I/O error on %s\n",
3062 : rcu_str_deref(device->name));
3063 : /* note, we dont' set_buffer_write_io_error because we have
3064 : * our own ways of dealing with the IO errors
3065 : */
3066 : clear_buffer_uptodate(bh);
3067 0 : btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_WRITE_ERRS);
3068 : }
3069 5835 : unlock_buffer(bh);
3070 : put_bh(bh);
3071 5835 : }
3072 :
3073 485 : struct buffer_head *btrfs_read_dev_super(struct block_device *bdev)
3074 : {
3075 : struct buffer_head *bh;
3076 : struct buffer_head *latest = NULL;
3077 : struct btrfs_super_block *super;
3078 : int i;
3079 : u64 transid = 0;
3080 : u64 bytenr;
3081 :
3082 : /* we would like to check all the supers, but that would make
3083 : * a btrfs mount succeed after a mkfs from a different FS.
3084 : * So, we need to add a special mount option to scan for
3085 : * later supers, using BTRFS_SUPER_MIRROR_MAX instead
3086 : */
3087 970 : for (i = 0; i < 1; i++) {
3088 : bytenr = btrfs_sb_offset(i);
3089 970 : if (bytenr + BTRFS_SUPER_INFO_SIZE >=
3090 485 : i_size_read(bdev->bd_inode))
3091 : break;
3092 485 : bh = __bread(bdev, bytenr / 4096,
3093 : BTRFS_SUPER_INFO_SIZE);
3094 485 : if (!bh)
3095 0 : continue;
3096 :
3097 485 : super = (struct btrfs_super_block *)bh->b_data;
3098 970 : if (btrfs_super_bytenr(super) != bytenr ||
3099 : btrfs_super_magic(super) != BTRFS_MAGIC) {
3100 : brelse(bh);
3101 0 : continue;
3102 : }
3103 :
3104 485 : if (!latest || btrfs_super_generation(super) > transid) {
3105 : brelse(latest);
3106 : latest = bh;
3107 485 : transid = btrfs_super_generation(super);
3108 : } else {
3109 : brelse(bh);
3110 : }
3111 : }
3112 485 : return latest;
3113 : }
3114 :
3115 : /*
3116 : * this should be called twice, once with wait == 0 and
3117 : * once with wait == 1. When wait == 0 is done, all the buffer heads
3118 : * we write are pinned.
3119 : *
3120 : * They are released when wait == 1 is done.
3121 : * max_mirrors must be the same for both runs, and it indicates how
3122 : * many supers on this one device should be written.
3123 : *
3124 : * max_mirrors == 0 means to write them all.
3125 : */
3126 7312 : static int write_dev_supers(struct btrfs_device *device,
3127 : struct btrfs_super_block *sb,
3128 : int do_barriers, int wait, int max_mirrors)
3129 : {
3130 : struct buffer_head *bh;
3131 : int i;
3132 : int ret;
3133 : int errors = 0;
3134 : u32 crc;
3135 : u64 bytenr;
3136 :
3137 7312 : if (max_mirrors == 0)
3138 : max_mirrors = BTRFS_SUPER_MIRROR_MAX;
3139 :
3140 11670 : for (i = 0; i < max_mirrors; i++) {
3141 : bytenr = btrfs_sb_offset(i);
3142 16028 : if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
3143 : break;
3144 :
3145 11670 : if (wait) {
3146 5835 : bh = __find_get_block(device->bdev, bytenr / 4096,
3147 : BTRFS_SUPER_INFO_SIZE);
3148 5835 : if (!bh) {
3149 0 : errors++;
3150 0 : continue;
3151 : }
3152 5835 : wait_on_buffer(bh);
3153 5835 : if (!buffer_uptodate(bh))
3154 0 : errors++;
3155 :
3156 : /* drop our reference */
3157 : brelse(bh);
3158 :
3159 : /* drop the reference from the wait == 0 run */
3160 : brelse(bh);
3161 5835 : continue;
3162 : } else {
3163 : btrfs_set_super_bytenr(sb, bytenr);
3164 :
3165 : crc = ~(u32)0;
3166 5835 : crc = btrfs_csum_data((char *)sb +
3167 : BTRFS_CSUM_SIZE, crc,
3168 : BTRFS_SUPER_INFO_SIZE -
3169 : BTRFS_CSUM_SIZE);
3170 : btrfs_csum_final(crc, sb->csum);
3171 :
3172 : /*
3173 : * one reference for us, and we leave it for the
3174 : * caller
3175 : */
3176 5835 : bh = __getblk(device->bdev, bytenr / 4096,
3177 : BTRFS_SUPER_INFO_SIZE);
3178 5835 : if (!bh) {
3179 0 : printk(KERN_ERR "BTRFS: couldn't get super "
3180 : "buffer head for bytenr %Lu\n", bytenr);
3181 0 : errors++;
3182 0 : continue;
3183 : }
3184 :
3185 5835 : memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE);
3186 :
3187 : /* one reference for submit_bh */
3188 : get_bh(bh);
3189 :
3190 : set_buffer_uptodate(bh);
3191 5835 : lock_buffer(bh);
3192 5835 : bh->b_end_io = btrfs_end_buffer_write_sync;
3193 5835 : bh->b_private = device;
3194 : }
3195 :
3196 : /*
3197 : * we fua the first super. The others we allow
3198 : * to go down lazy.
3199 : */
3200 5835 : if (i == 0)
3201 3656 : ret = btrfsic_submit_bh(WRITE_FUA, bh);
3202 : else
3203 2179 : ret = btrfsic_submit_bh(WRITE_SYNC, bh);
3204 5835 : if (ret)
3205 0 : errors++;
3206 : }
3207 7312 : return errors < i ? 0 : -1;
3208 : }
3209 :
3210 : /*
3211 : * endio for the write_dev_flush, this will wake anyone waiting
3212 : * for the barrier when it is done
3213 : */
3214 3656 : static void btrfs_end_empty_barrier(struct bio *bio, int err)
3215 : {
3216 3656 : if (err) {
3217 0 : if (err == -EOPNOTSUPP)
3218 : set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
3219 : clear_bit(BIO_UPTODATE, &bio->bi_flags);
3220 : }
3221 3656 : if (bio->bi_private)
3222 3656 : complete(bio->bi_private);
3223 3656 : bio_put(bio);
3224 3656 : }
3225 :
3226 : /*
3227 : * trigger flushes for one the devices. If you pass wait == 0, the flushes are
3228 : * sent down. With wait == 1, it waits for the previous flush.
3229 : *
3230 : * any device where the flush fails with eopnotsupp are flagged as not-barrier
3231 : * capable
3232 : */
3233 7312 : static int write_dev_flush(struct btrfs_device *device, int wait)
3234 : {
3235 : struct bio *bio;
3236 : int ret = 0;
3237 :
3238 7312 : if (device->nobarriers)
3239 : return 0;
3240 :
3241 7312 : if (wait) {
3242 3656 : bio = device->flush_bio;
3243 3656 : if (!bio)
3244 : return 0;
3245 :
3246 3656 : wait_for_completion(&device->flush_wait);
3247 :
3248 3656 : if (bio_flagged(bio, BIO_EOPNOTSUPP)) {
3249 0 : printk_in_rcu("BTRFS: disabling barriers on dev %s\n",
3250 : rcu_str_deref(device->name));
3251 0 : device->nobarriers = 1;
3252 3656 : } else if (!bio_flagged(bio, BIO_UPTODATE)) {
3253 : ret = -EIO;
3254 0 : btrfs_dev_stat_inc_and_print(device,
3255 : BTRFS_DEV_STAT_FLUSH_ERRS);
3256 : }
3257 :
3258 : /* drop the reference from the wait == 0 run */
3259 3656 : bio_put(bio);
3260 3656 : device->flush_bio = NULL;
3261 :
3262 3656 : return ret;
3263 : }
3264 :
3265 : /*
3266 : * one reference for us, and we leave it for the
3267 : * caller
3268 : */
3269 3656 : device->flush_bio = NULL;
3270 3656 : bio = btrfs_io_bio_alloc(GFP_NOFS, 0);
3271 3656 : if (!bio)
3272 : return -ENOMEM;
3273 :
3274 3656 : bio->bi_end_io = btrfs_end_empty_barrier;
3275 3656 : bio->bi_bdev = device->bdev;
3276 : init_completion(&device->flush_wait);
3277 3656 : bio->bi_private = &device->flush_wait;
3278 3656 : device->flush_bio = bio;
3279 :
3280 3656 : bio_get(bio);
3281 3656 : btrfsic_submit_bio(WRITE_FLUSH, bio);
3282 :
3283 3656 : return 0;
3284 : }
3285 :
3286 : /*
3287 : * send an empty flush down to each device in parallel,
3288 : * then wait for them
3289 : */
3290 3575 : static int barrier_all_devices(struct btrfs_fs_info *info)
3291 : {
3292 : struct list_head *head;
3293 : struct btrfs_device *dev;
3294 : int errors_send = 0;
3295 : int errors_wait = 0;
3296 : int ret;
3297 :
3298 : /* send down all the barriers */
3299 3575 : head = &info->fs_devices->devices;
3300 7231 : list_for_each_entry_rcu(dev, head, dev_list) {
3301 3656 : if (dev->missing)
3302 0 : continue;
3303 3656 : if (!dev->bdev) {
3304 0 : errors_send++;
3305 0 : continue;
3306 : }
3307 3656 : if (!dev->in_fs_metadata || !dev->writeable)
3308 0 : continue;
3309 :
3310 3656 : ret = write_dev_flush(dev, 0);
3311 3656 : if (ret)
3312 0 : errors_send++;
3313 : }
3314 :
3315 : /* wait for all the barriers */
3316 7231 : list_for_each_entry_rcu(dev, head, dev_list) {
3317 3656 : if (dev->missing)
3318 0 : continue;
3319 3656 : if (!dev->bdev) {
3320 0 : errors_wait++;
3321 0 : continue;
3322 : }
3323 3656 : if (!dev->in_fs_metadata || !dev->writeable)
3324 0 : continue;
3325 :
3326 3656 : ret = write_dev_flush(dev, 1);
3327 3656 : if (ret)
3328 0 : errors_wait++;
3329 : }
3330 3575 : if (errors_send > info->num_tolerated_disk_barrier_failures ||
3331 : errors_wait > info->num_tolerated_disk_barrier_failures)
3332 : return -EIO;
3333 : return 0;
3334 : }
3335 :
3336 221 : int btrfs_calc_num_tolerated_disk_barrier_failures(
3337 : struct btrfs_fs_info *fs_info)
3338 : {
3339 : struct btrfs_ioctl_space_info space;
3340 : struct btrfs_space_info *sinfo;
3341 221 : u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3342 : BTRFS_BLOCK_GROUP_SYSTEM,
3343 : BTRFS_BLOCK_GROUP_METADATA,
3344 : BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3345 : int num_types = 4;
3346 : int i;
3347 : int c;
3348 221 : int num_tolerated_disk_barrier_failures =
3349 221 : (int)fs_info->fs_devices->num_devices;
3350 :
3351 1105 : for (i = 0; i < num_types; i++) {
3352 : struct btrfs_space_info *tmp;
3353 :
3354 : sinfo = NULL;
3355 : rcu_read_lock();
3356 2204 : list_for_each_entry_rcu(tmp, &fs_info->space_info, list) {
3357 1977 : if (tmp->flags == types[i]) {
3358 : sinfo = tmp;
3359 : break;
3360 : }
3361 : }
3362 : rcu_read_unlock();
3363 :
3364 884 : if (!sinfo)
3365 227 : continue;
3366 :
3367 657 : down_read(&sinfo->groups_sem);
3368 5256 : for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3369 9198 : if (!list_empty(&sinfo->block_groups[c])) {
3370 : u64 flags;
3371 :
3372 1090 : btrfs_get_block_group_info(
3373 : &sinfo->block_groups[c], &space);
3374 2180 : if (space.total_bytes == 0 ||
3375 1090 : space.used_bytes == 0)
3376 527 : continue;
3377 563 : flags = space.flags;
3378 : /*
3379 : * return
3380 : * 0: if dup, single or RAID0 is configured for
3381 : * any of metadata, system or data, else
3382 : * 1: if RAID5 is configured, or if RAID1 or
3383 : * RAID10 is configured and only two mirrors
3384 : * are used, else
3385 : * 2: if RAID6 is configured, else
3386 : * num_mirrors - 1: if RAID1 or RAID10 is
3387 : * configured and more than
3388 : * 2 mirrors are used.
3389 : */
3390 796 : if (num_tolerated_disk_barrier_failures > 0 &&
3391 233 : ((flags & (BTRFS_BLOCK_GROUP_DUP |
3392 141 : BTRFS_BLOCK_GROUP_RAID0)) ||
3393 141 : ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
3394 : == 0)))
3395 : num_tolerated_disk_barrier_failures = 0;
3396 351 : else if (num_tolerated_disk_barrier_failures > 1) {
3397 10 : if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
3398 : BTRFS_BLOCK_GROUP_RAID5 |
3399 : BTRFS_BLOCK_GROUP_RAID10)) {
3400 : num_tolerated_disk_barrier_failures = 1;
3401 2 : } else if (flags &
3402 : BTRFS_BLOCK_GROUP_RAID6) {
3403 : num_tolerated_disk_barrier_failures = 2;
3404 : }
3405 : }
3406 : }
3407 : }
3408 657 : up_read(&sinfo->groups_sem);
3409 : }
3410 :
3411 221 : return num_tolerated_disk_barrier_failures;
3412 : }
3413 :
3414 3575 : static int write_all_supers(struct btrfs_root *root, int max_mirrors)
3415 : {
3416 : struct list_head *head;
3417 : struct btrfs_device *dev;
3418 : struct btrfs_super_block *sb;
3419 : struct btrfs_dev_item *dev_item;
3420 : int ret;
3421 : int do_barriers;
3422 : int max_errors;
3423 : int total_errors = 0;
3424 : u64 flags;
3425 :
3426 3575 : do_barriers = !btrfs_test_opt(root, NOBARRIER);
3427 3575 : backup_super_roots(root->fs_info);
3428 :
3429 3575 : sb = root->fs_info->super_for_commit;
3430 : dev_item = &sb->dev_item;
3431 :
3432 3575 : mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
3433 3575 : head = &root->fs_info->fs_devices->devices;
3434 7150 : max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
3435 :
3436 3575 : if (do_barriers) {
3437 3575 : ret = barrier_all_devices(root->fs_info);
3438 3575 : if (ret) {
3439 0 : mutex_unlock(
3440 0 : &root->fs_info->fs_devices->device_list_mutex);
3441 0 : btrfs_error(root->fs_info, ret,
3442 : "errors while submitting device barriers.");
3443 : return ret;
3444 : }
3445 : }
3446 :
3447 7231 : list_for_each_entry_rcu(dev, head, dev_list) {
3448 3656 : if (!dev->bdev) {
3449 0 : total_errors++;
3450 0 : continue;
3451 : }
3452 3656 : if (!dev->in_fs_metadata || !dev->writeable)
3453 0 : continue;
3454 :
3455 : btrfs_set_stack_device_generation(dev_item, 0);
3456 3656 : btrfs_set_stack_device_type(dev_item, dev->type);
3457 3656 : btrfs_set_stack_device_id(dev_item, dev->devid);
3458 3656 : btrfs_set_stack_device_total_bytes(dev_item,
3459 : dev->disk_total_bytes);
3460 3656 : btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
3461 3656 : btrfs_set_stack_device_io_align(dev_item, dev->io_align);
3462 3656 : btrfs_set_stack_device_io_width(dev_item, dev->io_width);
3463 3656 : btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
3464 3656 : memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
3465 3656 : memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
3466 :
3467 : flags = btrfs_super_flags(sb);
3468 3656 : btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
3469 :
3470 3656 : ret = write_dev_supers(dev, sb, do_barriers, 0, max_mirrors);
3471 3656 : if (ret)
3472 0 : total_errors++;
3473 : }
3474 3575 : if (total_errors > max_errors) {
3475 0 : btrfs_err(root->fs_info, "%d errors while writing supers",
3476 : total_errors);
3477 0 : mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
3478 :
3479 : /* FUA is masked off if unsupported and can't be the reason */
3480 0 : btrfs_error(root->fs_info, -EIO,
3481 : "%d errors while writing supers", total_errors);
3482 : return -EIO;
3483 : }
3484 :
3485 : total_errors = 0;
3486 7231 : list_for_each_entry_rcu(dev, head, dev_list) {
3487 3656 : if (!dev->bdev)
3488 0 : continue;
3489 3656 : if (!dev->in_fs_metadata || !dev->writeable)
3490 0 : continue;
3491 :
3492 3656 : ret = write_dev_supers(dev, sb, do_barriers, 1, max_mirrors);
3493 3656 : if (ret)
3494 0 : total_errors++;
3495 : }
3496 3575 : mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
3497 3575 : if (total_errors > max_errors) {
3498 0 : btrfs_error(root->fs_info, -EIO,
3499 : "%d errors while writing supers", total_errors);
3500 : return -EIO;
3501 : }
3502 : return 0;
3503 : }
3504 :
3505 3575 : int write_ctree_super(struct btrfs_trans_handle *trans,
3506 : struct btrfs_root *root, int max_mirrors)
3507 : {
3508 3575 : return write_all_supers(root, max_mirrors);
3509 : }
3510 :
3511 : /* Drop a fs root from the radix tree and free it. */
3512 688 : void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info,
3513 : struct btrfs_root *root)
3514 : {
3515 : spin_lock(&fs_info->fs_roots_radix_lock);
3516 688 : radix_tree_delete(&fs_info->fs_roots_radix,
3517 688 : (unsigned long)root->root_key.objectid);
3518 : spin_unlock(&fs_info->fs_roots_radix_lock);
3519 :
3520 688 : if (btrfs_root_refs(&root->root_item) == 0)
3521 105 : synchronize_srcu(&fs_info->subvol_srcu);
3522 :
3523 688 : if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3524 0 : btrfs_free_log(NULL, root);
3525 :
3526 688 : if (root->free_ino_pinned)
3527 688 : __btrfs_remove_free_space_cache(root->free_ino_pinned);
3528 688 : if (root->free_ino_ctl)
3529 688 : __btrfs_remove_free_space_cache(root->free_ino_ctl);
3530 688 : free_fs_root(root);
3531 688 : }
3532 :
3533 688 : static void free_fs_root(struct btrfs_root *root)
3534 : {
3535 688 : iput(root->cache_inode);
3536 688 : WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree));
3537 688 : btrfs_free_block_rsv(root, root->orphan_block_rsv);
3538 688 : root->orphan_block_rsv = NULL;
3539 688 : if (root->anon_dev)
3540 688 : free_anon_bdev(root->anon_dev);
3541 688 : if (root->subv_writers)
3542 : btrfs_free_subvolume_writers(root->subv_writers);
3543 688 : free_extent_buffer(root->node);
3544 688 : free_extent_buffer(root->commit_root);
3545 688 : kfree(root->free_ino_ctl);
3546 688 : kfree(root->free_ino_pinned);
3547 688 : kfree(root->name);
3548 688 : btrfs_put_fs_root(root);
3549 688 : }
3550 :
3551 0 : void btrfs_free_fs_root(struct btrfs_root *root)
3552 : {
3553 0 : free_fs_root(root);
3554 0 : }
3555 :
3556 194 : int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info)
3557 : {
3558 : u64 root_objectid = 0;
3559 : struct btrfs_root *gang[8];
3560 : int i = 0;
3561 : int err = 0;
3562 : unsigned int ret = 0;
3563 : int index;
3564 :
3565 : while (1) {
3566 210 : index = srcu_read_lock(&fs_info->subvol_srcu);
3567 210 : ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
3568 : (void **)gang, root_objectid,
3569 : ARRAY_SIZE(gang));
3570 210 : if (!ret) {
3571 : srcu_read_unlock(&fs_info->subvol_srcu, index);
3572 : break;
3573 : }
3574 16 : root_objectid = gang[ret - 1]->root_key.objectid + 1;
3575 :
3576 89 : for (i = 0; i < ret; i++) {
3577 : /* Avoid to grab roots in dead_roots */
3578 146 : if (btrfs_root_refs(&gang[i]->root_item) == 0) {
3579 71 : gang[i] = NULL;
3580 71 : continue;
3581 : }
3582 : /* grab all the search result for later use */
3583 2 : gang[i] = btrfs_grab_fs_root(gang[i]);
3584 : }
3585 : srcu_read_unlock(&fs_info->subvol_srcu, index);
3586 :
3587 89 : for (i = 0; i < ret; i++) {
3588 73 : if (!gang[i])
3589 71 : continue;
3590 2 : root_objectid = gang[i]->root_key.objectid;
3591 2 : err = btrfs_orphan_cleanup(gang[i]);
3592 2 : if (err)
3593 : break;
3594 2 : btrfs_put_fs_root(gang[i]);
3595 : }
3596 16 : root_objectid++;
3597 16 : }
3598 :
3599 : /* release the uncleaned roots due to error */
3600 0 : for (; i < ret; i++) {
3601 0 : if (gang[i])
3602 0 : btrfs_put_fs_root(gang[i]);
3603 : }
3604 194 : return err;
3605 : }
3606 :
3607 194 : int btrfs_commit_super(struct btrfs_root *root)
3608 : {
3609 : struct btrfs_trans_handle *trans;
3610 :
3611 194 : mutex_lock(&root->fs_info->cleaner_mutex);
3612 194 : btrfs_run_delayed_iputs(root);
3613 194 : mutex_unlock(&root->fs_info->cleaner_mutex);
3614 194 : wake_up_process(root->fs_info->cleaner_kthread);
3615 :
3616 : /* wait until ongoing cleanup work done */
3617 194 : down_write(&root->fs_info->cleanup_work_sem);
3618 194 : up_write(&root->fs_info->cleanup_work_sem);
3619 :
3620 194 : trans = btrfs_join_transaction(root);
3621 194 : if (IS_ERR(trans))
3622 0 : return PTR_ERR(trans);
3623 194 : return btrfs_commit_transaction(trans, root);
3624 : }
3625 :
3626 221 : int close_ctree(struct btrfs_root *root)
3627 : {
3628 221 : struct btrfs_fs_info *fs_info = root->fs_info;
3629 : int ret;
3630 :
3631 221 : fs_info->closing = 1;
3632 221 : smp_mb();
3633 :
3634 : /* wait for the uuid_scan task to finish */
3635 221 : down(&fs_info->uuid_tree_rescan_sem);
3636 : /* avoid complains from lockdep et al., set sem back to initial state */
3637 221 : up(&fs_info->uuid_tree_rescan_sem);
3638 :
3639 : /* pause restriper - we want to resume on mount */
3640 221 : btrfs_pause_balance(fs_info);
3641 :
3642 221 : btrfs_dev_replace_suspend_for_unmount(fs_info);
3643 :
3644 221 : btrfs_scrub_cancel(fs_info);
3645 :
3646 : /* wait for any defraggers to finish */
3647 221 : wait_event(fs_info->transaction_wait,
3648 : (atomic_read(&fs_info->defrag_running) == 0));
3649 :
3650 : /* clear out the rbtree of defraggable inodes */
3651 221 : btrfs_cleanup_defrag_inodes(fs_info);
3652 :
3653 221 : cancel_work_sync(&fs_info->async_reclaim_work);
3654 :
3655 221 : if (!(fs_info->sb->s_flags & MS_RDONLY)) {
3656 193 : ret = btrfs_commit_super(root);
3657 193 : if (ret)
3658 0 : btrfs_err(root->fs_info, "commit super ret %d", ret);
3659 : }
3660 :
3661 221 : if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state))
3662 0 : btrfs_error_commit_super(root);
3663 :
3664 221 : kthread_stop(fs_info->transaction_kthread);
3665 221 : kthread_stop(fs_info->cleaner_kthread);
3666 :
3667 221 : fs_info->closing = 2;
3668 221 : smp_mb();
3669 :
3670 221 : btrfs_free_qgroup_config(root->fs_info);
3671 :
3672 442 : if (percpu_counter_sum(&fs_info->delalloc_bytes)) {
3673 0 : btrfs_info(root->fs_info, "at unmount delalloc count %lld",
3674 : percpu_counter_sum(&fs_info->delalloc_bytes));
3675 : }
3676 :
3677 221 : btrfs_sysfs_remove_one(fs_info);
3678 :
3679 221 : btrfs_free_fs_roots(fs_info);
3680 :
3681 221 : btrfs_put_block_group_cache(fs_info);
3682 :
3683 221 : btrfs_free_block_groups(fs_info);
3684 :
3685 : /*
3686 : * we must make sure there is not any read request to
3687 : * submit after we stopping all workers.
3688 : */
3689 221 : invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
3690 221 : btrfs_stop_all_workers(fs_info);
3691 :
3692 221 : free_root_pointers(fs_info, 1);
3693 :
3694 221 : iput(fs_info->btree_inode);
3695 :
3696 : #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3697 : if (btrfs_test_opt(root, CHECK_INTEGRITY))
3698 : btrfsic_unmount(root, fs_info->fs_devices);
3699 : #endif
3700 :
3701 221 : btrfs_close_devices(fs_info->fs_devices);
3702 221 : btrfs_mapping_tree_free(&fs_info->mapping_tree);
3703 :
3704 221 : percpu_counter_destroy(&fs_info->dirty_metadata_bytes);
3705 221 : percpu_counter_destroy(&fs_info->delalloc_bytes);
3706 221 : percpu_counter_destroy(&fs_info->bio_counter);
3707 221 : bdi_destroy(&fs_info->bdi);
3708 221 : cleanup_srcu_struct(&fs_info->subvol_srcu);
3709 :
3710 221 : btrfs_free_stripe_hash_table(fs_info);
3711 :
3712 221 : btrfs_free_block_rsv(root, root->orphan_block_rsv);
3713 221 : root->orphan_block_rsv = NULL;
3714 :
3715 221 : return 0;
3716 : }
3717 :
3718 2227634 : int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid,
3719 : int atomic)
3720 : {
3721 : int ret;
3722 2227634 : struct inode *btree_inode = buf->pages[0]->mapping->host;
3723 :
3724 2227634 : ret = extent_buffer_uptodate(buf);
3725 2227610 : if (!ret)
3726 : return ret;
3727 :
3728 2227609 : ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf,
3729 : parent_transid, atomic);
3730 2227515 : if (ret == -EAGAIN)
3731 : return ret;
3732 2227522 : return !ret;
3733 : }
3734 :
3735 59105 : int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
3736 : {
3737 59105 : return set_extent_buffer_uptodate(buf);
3738 : }
3739 :
3740 1165630 : void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
3741 : {
3742 : struct btrfs_root *root;
3743 : u64 transid = btrfs_header_generation(buf);
3744 : int was_dirty;
3745 :
3746 : #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3747 : /*
3748 : * This is a fast path so only do this check if we have sanity tests
3749 : * enabled. Normal people shouldn't be marking dummy buffers as dirty
3750 : * outside of the sanity tests.
3751 : */
3752 : if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &buf->bflags)))
3753 : return;
3754 : #endif
3755 1165630 : root = BTRFS_I(buf->pages[0]->mapping->host)->root;
3756 1165630 : btrfs_assert_tree_locked(buf);
3757 1165619 : if (transid != root->fs_info->generation)
3758 0 : WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, "
3759 : "found %llu running %llu\n",
3760 : buf->start, transid, root->fs_info->generation);
3761 1165619 : was_dirty = set_extent_buffer_dirty(buf);
3762 1166240 : if (!was_dirty)
3763 117770 : __percpu_counter_add(&root->fs_info->dirty_metadata_bytes,
3764 58885 : buf->len,
3765 58885 : root->fs_info->dirty_metadata_batch);
3766 : #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
3767 : if (btrfs_header_level(buf) == 0 && check_leaf(root, buf)) {
3768 : btrfs_print_leaf(root, buf);
3769 : ASSERT(0);
3770 : }
3771 : #endif
3772 1166240 : }
3773 :
3774 124607 : static void __btrfs_btree_balance_dirty(struct btrfs_root *root,
3775 : int flush_delayed)
3776 : {
3777 : /*
3778 : * looks as though older kernels can get into trouble with
3779 : * this code, they end up stuck in balance_dirty_pages forever
3780 : */
3781 : int ret;
3782 :
3783 124607 : if (current->flags & PF_MEMALLOC)
3784 : return;
3785 :
3786 124607 : if (flush_delayed)
3787 99366 : btrfs_balance_delayed_items(root);
3788 :
3789 124606 : ret = percpu_counter_compare(&root->fs_info->dirty_metadata_bytes,
3790 : BTRFS_DIRTY_METADATA_THRESH);
3791 124604 : if (ret > 0) {
3792 0 : balance_dirty_pages_ratelimited(
3793 0 : root->fs_info->btree_inode->i_mapping);
3794 : }
3795 : return;
3796 : }
3797 :
3798 99366 : void btrfs_btree_balance_dirty(struct btrfs_root *root)
3799 : {
3800 99366 : __btrfs_btree_balance_dirty(root, 1);
3801 99365 : }
3802 :
3803 25243 : void btrfs_btree_balance_dirty_nodelay(struct btrfs_root *root)
3804 : {
3805 25243 : __btrfs_btree_balance_dirty(root, 0);
3806 25239 : }
3807 :
3808 188 : int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
3809 : {
3810 188 : struct btrfs_root *root = BTRFS_I(buf->pages[0]->mapping->host)->root;
3811 188 : return btree_read_extent_buffer_pages(root, buf, 0, parent_transid);
3812 : }
3813 :
3814 : static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
3815 : int read_only)
3816 : {
3817 : /*
3818 : * Placeholder for checks
3819 : */
3820 : return 0;
3821 : }
3822 :
3823 0 : static void btrfs_error_commit_super(struct btrfs_root *root)
3824 : {
3825 0 : mutex_lock(&root->fs_info->cleaner_mutex);
3826 0 : btrfs_run_delayed_iputs(root);
3827 0 : mutex_unlock(&root->fs_info->cleaner_mutex);
3828 :
3829 0 : down_write(&root->fs_info->cleanup_work_sem);
3830 0 : up_write(&root->fs_info->cleanup_work_sem);
3831 :
3832 : /* cleanup FS via transaction */
3833 0 : btrfs_cleanup_transaction(root);
3834 0 : }
3835 :
3836 0 : static void btrfs_destroy_ordered_extents(struct btrfs_root *root)
3837 : {
3838 : struct btrfs_ordered_extent *ordered;
3839 :
3840 : spin_lock(&root->ordered_extent_lock);
3841 : /*
3842 : * This will just short circuit the ordered completion stuff which will
3843 : * make sure the ordered extent gets properly cleaned up.
3844 : */
3845 0 : list_for_each_entry(ordered, &root->ordered_extents,
3846 : root_extent_list)
3847 : set_bit(BTRFS_ORDERED_IOERR, &ordered->flags);
3848 : spin_unlock(&root->ordered_extent_lock);
3849 0 : }
3850 :
3851 0 : static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info)
3852 : {
3853 : struct btrfs_root *root;
3854 : struct list_head splice;
3855 :
3856 : INIT_LIST_HEAD(&splice);
3857 :
3858 : spin_lock(&fs_info->ordered_root_lock);
3859 0 : list_splice_init(&fs_info->ordered_roots, &splice);
3860 0 : while (!list_empty(&splice)) {
3861 0 : root = list_first_entry(&splice, struct btrfs_root,
3862 : ordered_root);
3863 0 : list_move_tail(&root->ordered_root,
3864 : &fs_info->ordered_roots);
3865 :
3866 : spin_unlock(&fs_info->ordered_root_lock);
3867 0 : btrfs_destroy_ordered_extents(root);
3868 :
3869 0 : cond_resched();
3870 : spin_lock(&fs_info->ordered_root_lock);
3871 : }
3872 : spin_unlock(&fs_info->ordered_root_lock);
3873 0 : }
3874 :
3875 0 : static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans,
3876 : struct btrfs_root *root)
3877 : {
3878 : struct rb_node *node;
3879 : struct btrfs_delayed_ref_root *delayed_refs;
3880 : struct btrfs_delayed_ref_node *ref;
3881 : int ret = 0;
3882 :
3883 : delayed_refs = &trans->delayed_refs;
3884 :
3885 : spin_lock(&delayed_refs->lock);
3886 0 : if (atomic_read(&delayed_refs->num_entries) == 0) {
3887 : spin_unlock(&delayed_refs->lock);
3888 0 : btrfs_info(root->fs_info, "delayed_refs has NO entry");
3889 0 : return ret;
3890 : }
3891 :
3892 0 : while ((node = rb_first(&delayed_refs->href_root)) != NULL) {
3893 : struct btrfs_delayed_ref_head *head;
3894 : bool pin_bytes = false;
3895 :
3896 : head = rb_entry(node, struct btrfs_delayed_ref_head,
3897 : href_node);
3898 0 : if (!mutex_trylock(&head->mutex)) {
3899 0 : atomic_inc(&head->node.refs);
3900 : spin_unlock(&delayed_refs->lock);
3901 :
3902 0 : mutex_lock(&head->mutex);
3903 0 : mutex_unlock(&head->mutex);
3904 0 : btrfs_put_delayed_ref(&head->node);
3905 : spin_lock(&delayed_refs->lock);
3906 0 : continue;
3907 : }
3908 : spin_lock(&head->lock);
3909 0 : while ((node = rb_first(&head->ref_root)) != NULL) {
3910 : ref = rb_entry(node, struct btrfs_delayed_ref_node,
3911 : rb_node);
3912 0 : ref->in_tree = 0;
3913 0 : rb_erase(&ref->rb_node, &head->ref_root);
3914 0 : atomic_dec(&delayed_refs->num_entries);
3915 0 : btrfs_put_delayed_ref(ref);
3916 : }
3917 0 : if (head->must_insert_reserved)
3918 : pin_bytes = true;
3919 0 : btrfs_free_delayed_extent_op(head->extent_op);
3920 0 : delayed_refs->num_heads--;
3921 0 : if (head->processing == 0)
3922 0 : delayed_refs->num_heads_ready--;
3923 0 : atomic_dec(&delayed_refs->num_entries);
3924 0 : head->node.in_tree = 0;
3925 0 : rb_erase(&head->href_node, &delayed_refs->href_root);
3926 : spin_unlock(&head->lock);
3927 : spin_unlock(&delayed_refs->lock);
3928 0 : mutex_unlock(&head->mutex);
3929 :
3930 0 : if (pin_bytes)
3931 0 : btrfs_pin_extent(root, head->node.bytenr,
3932 : head->node.num_bytes, 1);
3933 0 : btrfs_put_delayed_ref(&head->node);
3934 0 : cond_resched();
3935 : spin_lock(&delayed_refs->lock);
3936 : }
3937 :
3938 : spin_unlock(&delayed_refs->lock);
3939 :
3940 0 : return ret;
3941 : }
3942 :
3943 0 : static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root)
3944 : {
3945 : struct btrfs_inode *btrfs_inode;
3946 : struct list_head splice;
3947 :
3948 : INIT_LIST_HEAD(&splice);
3949 :
3950 : spin_lock(&root->delalloc_lock);
3951 0 : list_splice_init(&root->delalloc_inodes, &splice);
3952 :
3953 0 : while (!list_empty(&splice)) {
3954 : btrfs_inode = list_first_entry(&splice, struct btrfs_inode,
3955 : delalloc_inodes);
3956 :
3957 0 : list_del_init(&btrfs_inode->delalloc_inodes);
3958 : clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
3959 : &btrfs_inode->runtime_flags);
3960 : spin_unlock(&root->delalloc_lock);
3961 :
3962 0 : btrfs_invalidate_inodes(btrfs_inode->root);
3963 :
3964 : spin_lock(&root->delalloc_lock);
3965 : }
3966 :
3967 : spin_unlock(&root->delalloc_lock);
3968 0 : }
3969 :
3970 0 : static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info)
3971 : {
3972 : struct btrfs_root *root;
3973 : struct list_head splice;
3974 :
3975 : INIT_LIST_HEAD(&splice);
3976 :
3977 : spin_lock(&fs_info->delalloc_root_lock);
3978 0 : list_splice_init(&fs_info->delalloc_roots, &splice);
3979 0 : while (!list_empty(&splice)) {
3980 0 : root = list_first_entry(&splice, struct btrfs_root,
3981 : delalloc_root);
3982 0 : list_del_init(&root->delalloc_root);
3983 0 : root = btrfs_grab_fs_root(root);
3984 0 : BUG_ON(!root);
3985 : spin_unlock(&fs_info->delalloc_root_lock);
3986 :
3987 0 : btrfs_destroy_delalloc_inodes(root);
3988 0 : btrfs_put_fs_root(root);
3989 :
3990 : spin_lock(&fs_info->delalloc_root_lock);
3991 : }
3992 : spin_unlock(&fs_info->delalloc_root_lock);
3993 0 : }
3994 :
3995 0 : static int btrfs_destroy_marked_extents(struct btrfs_root *root,
3996 : struct extent_io_tree *dirty_pages,
3997 : int mark)
3998 : {
3999 : int ret;
4000 : struct extent_buffer *eb;
4001 0 : u64 start = 0;
4002 : u64 end;
4003 :
4004 : while (1) {
4005 0 : ret = find_first_extent_bit(dirty_pages, start, &start, &end,
4006 : mark, NULL);
4007 0 : if (ret)
4008 : break;
4009 :
4010 0 : clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
4011 0 : while (start <= end) {
4012 : eb = btrfs_find_tree_block(root, start,
4013 : root->leafsize);
4014 0 : start += root->leafsize;
4015 0 : if (!eb)
4016 0 : continue;
4017 0 : wait_on_extent_buffer_writeback(eb);
4018 :
4019 0 : if (test_and_clear_bit(EXTENT_BUFFER_DIRTY,
4020 0 : &eb->bflags))
4021 0 : clear_extent_buffer_dirty(eb);
4022 0 : free_extent_buffer_stale(eb);
4023 : }
4024 : }
4025 :
4026 0 : return ret;
4027 : }
4028 :
4029 0 : static int btrfs_destroy_pinned_extent(struct btrfs_root *root,
4030 : struct extent_io_tree *pinned_extents)
4031 : {
4032 : struct extent_io_tree *unpin;
4033 : u64 start;
4034 : u64 end;
4035 : int ret;
4036 : bool loop = true;
4037 :
4038 : unpin = pinned_extents;
4039 : again:
4040 : while (1) {
4041 0 : ret = find_first_extent_bit(unpin, 0, &start, &end,
4042 : EXTENT_DIRTY, NULL);
4043 0 : if (ret)
4044 : break;
4045 :
4046 : /* opt_discard */
4047 0 : if (btrfs_test_opt(root, DISCARD))
4048 0 : ret = btrfs_error_discard_extent(root, start,
4049 0 : end + 1 - start,
4050 : NULL);
4051 :
4052 0 : clear_extent_dirty(unpin, start, end, GFP_NOFS);
4053 0 : btrfs_error_unpin_extent_range(root, start, end);
4054 0 : cond_resched();
4055 0 : }
4056 :
4057 0 : if (loop) {
4058 0 : if (unpin == &root->fs_info->freed_extents[0])
4059 0 : unpin = &root->fs_info->freed_extents[1];
4060 : else
4061 : unpin = &root->fs_info->freed_extents[0];
4062 : loop = false;
4063 : goto again;
4064 : }
4065 :
4066 0 : return 0;
4067 : }
4068 :
4069 0 : void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans,
4070 : struct btrfs_root *root)
4071 : {
4072 0 : btrfs_destroy_delayed_refs(cur_trans, root);
4073 :
4074 0 : cur_trans->state = TRANS_STATE_COMMIT_START;
4075 0 : wake_up(&root->fs_info->transaction_blocked_wait);
4076 :
4077 0 : cur_trans->state = TRANS_STATE_UNBLOCKED;
4078 0 : wake_up(&root->fs_info->transaction_wait);
4079 :
4080 0 : btrfs_destroy_delayed_inodes(root);
4081 0 : btrfs_assert_delayed_root_empty(root);
4082 :
4083 0 : btrfs_destroy_marked_extents(root, &cur_trans->dirty_pages,
4084 : EXTENT_DIRTY);
4085 0 : btrfs_destroy_pinned_extent(root,
4086 0 : root->fs_info->pinned_extents);
4087 :
4088 0 : cur_trans->state =TRANS_STATE_COMPLETED;
4089 0 : wake_up(&cur_trans->commit_wait);
4090 :
4091 : /*
4092 : memset(cur_trans, 0, sizeof(*cur_trans));
4093 : kmem_cache_free(btrfs_transaction_cachep, cur_trans);
4094 : */
4095 0 : }
4096 :
4097 0 : static int btrfs_cleanup_transaction(struct btrfs_root *root)
4098 : {
4099 : struct btrfs_transaction *t;
4100 :
4101 0 : mutex_lock(&root->fs_info->transaction_kthread_mutex);
4102 :
4103 0 : spin_lock(&root->fs_info->trans_lock);
4104 0 : while (!list_empty(&root->fs_info->trans_list)) {
4105 0 : t = list_first_entry(&root->fs_info->trans_list,
4106 : struct btrfs_transaction, list);
4107 0 : if (t->state >= TRANS_STATE_COMMIT_START) {
4108 0 : atomic_inc(&t->use_count);
4109 0 : spin_unlock(&root->fs_info->trans_lock);
4110 0 : btrfs_wait_for_commit(root, t->transid);
4111 0 : btrfs_put_transaction(t);
4112 0 : spin_lock(&root->fs_info->trans_lock);
4113 0 : continue;
4114 : }
4115 0 : if (t == root->fs_info->running_transaction) {
4116 0 : t->state = TRANS_STATE_COMMIT_DOING;
4117 0 : spin_unlock(&root->fs_info->trans_lock);
4118 : /*
4119 : * We wait for 0 num_writers since we don't hold a trans
4120 : * handle open currently for this transaction.
4121 : */
4122 0 : wait_event(t->writer_wait,
4123 : atomic_read(&t->num_writers) == 0);
4124 : } else {
4125 : spin_unlock(&root->fs_info->trans_lock);
4126 : }
4127 0 : btrfs_cleanup_one_transaction(t, root);
4128 :
4129 0 : spin_lock(&root->fs_info->trans_lock);
4130 0 : if (t == root->fs_info->running_transaction)
4131 0 : root->fs_info->running_transaction = NULL;
4132 0 : list_del_init(&t->list);
4133 0 : spin_unlock(&root->fs_info->trans_lock);
4134 :
4135 0 : btrfs_put_transaction(t);
4136 0 : trace_btrfs_transaction_commit(root);
4137 0 : spin_lock(&root->fs_info->trans_lock);
4138 : }
4139 : spin_unlock(&root->fs_info->trans_lock);
4140 0 : btrfs_destroy_all_ordered_extents(root->fs_info);
4141 0 : btrfs_destroy_delayed_inodes(root);
4142 0 : btrfs_assert_delayed_root_empty(root);
4143 0 : btrfs_destroy_pinned_extent(root, root->fs_info->pinned_extents);
4144 0 : btrfs_destroy_all_delalloc_inodes(root->fs_info);
4145 0 : mutex_unlock(&root->fs_info->transaction_kthread_mutex);
4146 :
4147 0 : return 0;
4148 : }
4149 :
4150 : static struct extent_io_ops btree_extent_io_ops = {
4151 : .readpage_end_io_hook = btree_readpage_end_io_hook,
4152 : .readpage_io_failed_hook = btree_io_failed_hook,
4153 : .submit_bio_hook = btree_submit_bio_hook,
4154 : /* note we're sharing with inode.c for the merge bio hook */
4155 : .merge_bio_hook = btrfs_merge_bio_hook,
4156 : };
|
