blob: d81cda2e077c9b3c4685d26a55057826d0c06b2d [file] [log] [blame]
Chris Masone02119d2008-09-05 16:13:11 -04001/*
2 * Copyright (C) 2008 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/sched.h>
20#include "ctree.h"
21#include "transaction.h"
22#include "disk-io.h"
23#include "locking.h"
24#include "print-tree.h"
25#include "compat.h"
Christoph Hellwigb2950862008-12-02 09:54:17 -050026#include "tree-log.h"
Chris Masone02119d2008-09-05 16:13:11 -040027
28/* magic values for the inode_only field in btrfs_log_inode:
29 *
30 * LOG_INODE_ALL means to log everything
31 * LOG_INODE_EXISTS means to log just enough to recreate the inode
32 * during log replay
33 */
34#define LOG_INODE_ALL 0
35#define LOG_INODE_EXISTS 1
36
37/*
38 * stages for the tree walking. The first
39 * stage (0) is to only pin down the blocks we find
40 * the second stage (1) is to make sure that all the inodes
41 * we find in the log are created in the subvolume.
42 *
43 * The last stage is to deal with directories and links and extents
44 * and all the other fun semantics
45 */
46#define LOG_WALK_PIN_ONLY 0
47#define LOG_WALK_REPLAY_INODES 1
48#define LOG_WALK_REPLAY_ALL 2
49
50static int __btrfs_log_inode(struct btrfs_trans_handle *trans,
51 struct btrfs_root *root, struct inode *inode,
52 int inode_only);
Yan Zhengec051c02009-01-05 15:43:42 -050053static int link_to_fixup_dir(struct btrfs_trans_handle *trans,
54 struct btrfs_root *root,
55 struct btrfs_path *path, u64 objectid);
Chris Masone02119d2008-09-05 16:13:11 -040056
57/*
58 * tree logging is a special write ahead log used to make sure that
59 * fsyncs and O_SYNCs can happen without doing full tree commits.
60 *
61 * Full tree commits are expensive because they require commonly
62 * modified blocks to be recowed, creating many dirty pages in the
63 * extent tree an 4x-6x higher write load than ext3.
64 *
65 * Instead of doing a tree commit on every fsync, we use the
66 * key ranges and transaction ids to find items for a given file or directory
67 * that have changed in this transaction. Those items are copied into
68 * a special tree (one per subvolume root), that tree is written to disk
69 * and then the fsync is considered complete.
70 *
71 * After a crash, items are copied out of the log-tree back into the
72 * subvolume tree. Any file data extents found are recorded in the extent
73 * allocation tree, and the log-tree freed.
74 *
75 * The log tree is read three times, once to pin down all the extents it is
76 * using in ram and once, once to create all the inodes logged in the tree
77 * and once to do all the other items.
78 */
79
80/*
81 * btrfs_add_log_tree adds a new per-subvolume log tree into the
82 * tree of log tree roots. This must be called with a tree log transaction
83 * running (see start_log_trans).
84 */
Christoph Hellwigb2950862008-12-02 09:54:17 -050085static int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
Chris Masone02119d2008-09-05 16:13:11 -040086 struct btrfs_root *root)
87{
88 struct btrfs_key key;
89 struct btrfs_root_item root_item;
90 struct btrfs_inode_item *inode_item;
91 struct extent_buffer *leaf;
92 struct btrfs_root *new_root = root;
93 int ret;
94 u64 objectid = root->root_key.objectid;
95
Zheng Yan31840ae2008-09-23 13:14:14 -040096 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
Chris Masone02119d2008-09-05 16:13:11 -040097 BTRFS_TREE_LOG_OBJECTID,
Zheng Yan31840ae2008-09-23 13:14:14 -040098 trans->transid, 0, 0, 0);
Chris Masone02119d2008-09-05 16:13:11 -040099 if (IS_ERR(leaf)) {
100 ret = PTR_ERR(leaf);
101 return ret;
102 }
103
104 btrfs_set_header_nritems(leaf, 0);
105 btrfs_set_header_level(leaf, 0);
106 btrfs_set_header_bytenr(leaf, leaf->start);
107 btrfs_set_header_generation(leaf, trans->transid);
108 btrfs_set_header_owner(leaf, BTRFS_TREE_LOG_OBJECTID);
109
110 write_extent_buffer(leaf, root->fs_info->fsid,
111 (unsigned long)btrfs_header_fsid(leaf),
112 BTRFS_FSID_SIZE);
113 btrfs_mark_buffer_dirty(leaf);
114
115 inode_item = &root_item.inode;
116 memset(inode_item, 0, sizeof(*inode_item));
117 inode_item->generation = cpu_to_le64(1);
118 inode_item->size = cpu_to_le64(3);
119 inode_item->nlink = cpu_to_le32(1);
Yan Zhenga76a3cd2008-10-09 11:46:29 -0400120 inode_item->nbytes = cpu_to_le64(root->leafsize);
Chris Masone02119d2008-09-05 16:13:11 -0400121 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
122
123 btrfs_set_root_bytenr(&root_item, leaf->start);
Yan Zheng84234f32008-10-29 14:49:05 -0400124 btrfs_set_root_generation(&root_item, trans->transid);
Chris Masone02119d2008-09-05 16:13:11 -0400125 btrfs_set_root_level(&root_item, 0);
126 btrfs_set_root_refs(&root_item, 0);
127 btrfs_set_root_used(&root_item, 0);
128
129 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
130 root_item.drop_level = 0;
131
132 btrfs_tree_unlock(leaf);
133 free_extent_buffer(leaf);
134 leaf = NULL;
135
136 btrfs_set_root_dirid(&root_item, 0);
137
138 key.objectid = BTRFS_TREE_LOG_OBJECTID;
139 key.offset = objectid;
140 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
141 ret = btrfs_insert_root(trans, root->fs_info->log_root_tree, &key,
142 &root_item);
143 if (ret)
144 goto fail;
145
146 new_root = btrfs_read_fs_root_no_radix(root->fs_info->log_root_tree,
147 &key);
148 BUG_ON(!new_root);
149
150 WARN_ON(root->log_root);
151 root->log_root = new_root;
152
153 /*
154 * log trees do not get reference counted because they go away
155 * before a real commit is actually done. They do store pointers
156 * to file data extents, and those reference counts still get
157 * updated (along with back refs to the log tree).
158 */
159 new_root->ref_cows = 0;
160 new_root->last_trans = trans->transid;
Chris Masone293e972009-01-09 13:14:17 -0500161
162 /*
163 * we need to make sure the root block for this new tree
164 * is marked as dirty in the dirty_log_pages tree. This
165 * is how it gets flushed down to disk at tree log commit time.
166 *
167 * the tree logging mutex keeps others from coming in and changing
168 * the new_root->node, so we can safely access it here
169 */
170 set_extent_dirty(&new_root->dirty_log_pages, new_root->node->start,
171 new_root->node->start + new_root->node->len - 1,
172 GFP_NOFS);
173
Chris Masone02119d2008-09-05 16:13:11 -0400174fail:
175 return ret;
176}
177
178/*
179 * start a sub transaction and setup the log tree
180 * this increments the log tree writer count to make the people
181 * syncing the tree wait for us to finish
182 */
183static int start_log_trans(struct btrfs_trans_handle *trans,
184 struct btrfs_root *root)
185{
186 int ret;
187 mutex_lock(&root->fs_info->tree_log_mutex);
188 if (!root->fs_info->log_root_tree) {
189 ret = btrfs_init_log_root_tree(trans, root->fs_info);
190 BUG_ON(ret);
191 }
192 if (!root->log_root) {
193 ret = btrfs_add_log_tree(trans, root);
194 BUG_ON(ret);
195 }
196 atomic_inc(&root->fs_info->tree_log_writers);
197 root->fs_info->tree_log_batch++;
198 mutex_unlock(&root->fs_info->tree_log_mutex);
199 return 0;
200}
201
202/*
203 * returns 0 if there was a log transaction running and we were able
204 * to join, or returns -ENOENT if there were not transactions
205 * in progress
206 */
207static int join_running_log_trans(struct btrfs_root *root)
208{
209 int ret = -ENOENT;
210
211 smp_mb();
212 if (!root->log_root)
213 return -ENOENT;
214
215 mutex_lock(&root->fs_info->tree_log_mutex);
216 if (root->log_root) {
217 ret = 0;
218 atomic_inc(&root->fs_info->tree_log_writers);
219 root->fs_info->tree_log_batch++;
220 }
221 mutex_unlock(&root->fs_info->tree_log_mutex);
222 return ret;
223}
224
225/*
226 * indicate we're done making changes to the log tree
227 * and wake up anyone waiting to do a sync
228 */
229static int end_log_trans(struct btrfs_root *root)
230{
231 atomic_dec(&root->fs_info->tree_log_writers);
232 smp_mb();
233 if (waitqueue_active(&root->fs_info->tree_log_wait))
234 wake_up(&root->fs_info->tree_log_wait);
235 return 0;
236}
237
238
239/*
240 * the walk control struct is used to pass state down the chain when
241 * processing the log tree. The stage field tells us which part
242 * of the log tree processing we are currently doing. The others
243 * are state fields used for that specific part
244 */
245struct walk_control {
246 /* should we free the extent on disk when done? This is used
247 * at transaction commit time while freeing a log tree
248 */
249 int free;
250
251 /* should we write out the extent buffer? This is used
252 * while flushing the log tree to disk during a sync
253 */
254 int write;
255
256 /* should we wait for the extent buffer io to finish? Also used
257 * while flushing the log tree to disk for a sync
258 */
259 int wait;
260
261 /* pin only walk, we record which extents on disk belong to the
262 * log trees
263 */
264 int pin;
265
266 /* what stage of the replay code we're currently in */
267 int stage;
268
269 /* the root we are currently replaying */
270 struct btrfs_root *replay_dest;
271
272 /* the trans handle for the current replay */
273 struct btrfs_trans_handle *trans;
274
275 /* the function that gets used to process blocks we find in the
276 * tree. Note the extent_buffer might not be up to date when it is
277 * passed in, and it must be checked or read if you need the data
278 * inside it
279 */
280 int (*process_func)(struct btrfs_root *log, struct extent_buffer *eb,
281 struct walk_control *wc, u64 gen);
282};
283
284/*
285 * process_func used to pin down extents, write them or wait on them
286 */
287static int process_one_buffer(struct btrfs_root *log,
288 struct extent_buffer *eb,
289 struct walk_control *wc, u64 gen)
290{
291 if (wc->pin) {
Josef Bacik25179202008-10-29 14:49:05 -0400292 mutex_lock(&log->fs_info->pinned_mutex);
Chris Masone02119d2008-09-05 16:13:11 -0400293 btrfs_update_pinned_extents(log->fs_info->extent_root,
294 eb->start, eb->len, 1);
Josef Bacik25179202008-10-29 14:49:05 -0400295 mutex_unlock(&log->fs_info->pinned_mutex);
Chris Masone02119d2008-09-05 16:13:11 -0400296 }
297
298 if (btrfs_buffer_uptodate(eb, gen)) {
299 if (wc->write)
300 btrfs_write_tree_block(eb);
301 if (wc->wait)
302 btrfs_wait_tree_block_writeback(eb);
303 }
304 return 0;
305}
306
307/*
308 * Item overwrite used by replay and tree logging. eb, slot and key all refer
309 * to the src data we are copying out.
310 *
311 * root is the tree we are copying into, and path is a scratch
312 * path for use in this function (it should be released on entry and
313 * will be released on exit).
314 *
315 * If the key is already in the destination tree the existing item is
316 * overwritten. If the existing item isn't big enough, it is extended.
317 * If it is too large, it is truncated.
318 *
319 * If the key isn't in the destination yet, a new item is inserted.
320 */
321static noinline int overwrite_item(struct btrfs_trans_handle *trans,
322 struct btrfs_root *root,
323 struct btrfs_path *path,
324 struct extent_buffer *eb, int slot,
325 struct btrfs_key *key)
326{
327 int ret;
328 u32 item_size;
329 u64 saved_i_size = 0;
330 int save_old_i_size = 0;
331 unsigned long src_ptr;
332 unsigned long dst_ptr;
333 int overwrite_root = 0;
334
335 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
336 overwrite_root = 1;
337
338 item_size = btrfs_item_size_nr(eb, slot);
339 src_ptr = btrfs_item_ptr_offset(eb, slot);
340
341 /* look for the key in the destination tree */
342 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
343 if (ret == 0) {
344 char *src_copy;
345 char *dst_copy;
346 u32 dst_size = btrfs_item_size_nr(path->nodes[0],
347 path->slots[0]);
348 if (dst_size != item_size)
349 goto insert;
350
351 if (item_size == 0) {
352 btrfs_release_path(root, path);
353 return 0;
354 }
355 dst_copy = kmalloc(item_size, GFP_NOFS);
356 src_copy = kmalloc(item_size, GFP_NOFS);
357
358 read_extent_buffer(eb, src_copy, src_ptr, item_size);
359
360 dst_ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
361 read_extent_buffer(path->nodes[0], dst_copy, dst_ptr,
362 item_size);
363 ret = memcmp(dst_copy, src_copy, item_size);
364
365 kfree(dst_copy);
366 kfree(src_copy);
367 /*
368 * they have the same contents, just return, this saves
369 * us from cowing blocks in the destination tree and doing
370 * extra writes that may not have been done by a previous
371 * sync
372 */
373 if (ret == 0) {
374 btrfs_release_path(root, path);
375 return 0;
376 }
377
378 }
379insert:
380 btrfs_release_path(root, path);
381 /* try to insert the key into the destination tree */
382 ret = btrfs_insert_empty_item(trans, root, path,
383 key, item_size);
384
385 /* make sure any existing item is the correct size */
386 if (ret == -EEXIST) {
387 u32 found_size;
388 found_size = btrfs_item_size_nr(path->nodes[0],
389 path->slots[0]);
390 if (found_size > item_size) {
391 btrfs_truncate_item(trans, root, path, item_size, 1);
392 } else if (found_size < item_size) {
Yan Zheng87b29b22008-12-17 10:21:48 -0500393 ret = btrfs_extend_item(trans, root, path,
394 item_size - found_size);
Chris Masone02119d2008-09-05 16:13:11 -0400395 BUG_ON(ret);
396 }
397 } else if (ret) {
398 BUG();
399 }
400 dst_ptr = btrfs_item_ptr_offset(path->nodes[0],
401 path->slots[0]);
402
403 /* don't overwrite an existing inode if the generation number
404 * was logged as zero. This is done when the tree logging code
405 * is just logging an inode to make sure it exists after recovery.
406 *
407 * Also, don't overwrite i_size on directories during replay.
408 * log replay inserts and removes directory items based on the
409 * state of the tree found in the subvolume, and i_size is modified
410 * as it goes
411 */
412 if (key->type == BTRFS_INODE_ITEM_KEY && ret == -EEXIST) {
413 struct btrfs_inode_item *src_item;
414 struct btrfs_inode_item *dst_item;
415
416 src_item = (struct btrfs_inode_item *)src_ptr;
417 dst_item = (struct btrfs_inode_item *)dst_ptr;
418
419 if (btrfs_inode_generation(eb, src_item) == 0)
420 goto no_copy;
421
422 if (overwrite_root &&
423 S_ISDIR(btrfs_inode_mode(eb, src_item)) &&
424 S_ISDIR(btrfs_inode_mode(path->nodes[0], dst_item))) {
425 save_old_i_size = 1;
426 saved_i_size = btrfs_inode_size(path->nodes[0],
427 dst_item);
428 }
429 }
430
431 copy_extent_buffer(path->nodes[0], eb, dst_ptr,
432 src_ptr, item_size);
433
434 if (save_old_i_size) {
435 struct btrfs_inode_item *dst_item;
436 dst_item = (struct btrfs_inode_item *)dst_ptr;
437 btrfs_set_inode_size(path->nodes[0], dst_item, saved_i_size);
438 }
439
440 /* make sure the generation is filled in */
441 if (key->type == BTRFS_INODE_ITEM_KEY) {
442 struct btrfs_inode_item *dst_item;
443 dst_item = (struct btrfs_inode_item *)dst_ptr;
444 if (btrfs_inode_generation(path->nodes[0], dst_item) == 0) {
445 btrfs_set_inode_generation(path->nodes[0], dst_item,
446 trans->transid);
447 }
448 }
449no_copy:
450 btrfs_mark_buffer_dirty(path->nodes[0]);
451 btrfs_release_path(root, path);
452 return 0;
453}
454
455/*
456 * simple helper to read an inode off the disk from a given root
457 * This can only be called for subvolume roots and not for the log
458 */
459static noinline struct inode *read_one_inode(struct btrfs_root *root,
460 u64 objectid)
461{
462 struct inode *inode;
463 inode = btrfs_iget_locked(root->fs_info->sb, objectid, root);
464 if (inode->i_state & I_NEW) {
465 BTRFS_I(inode)->root = root;
466 BTRFS_I(inode)->location.objectid = objectid;
467 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
468 BTRFS_I(inode)->location.offset = 0;
469 btrfs_read_locked_inode(inode);
470 unlock_new_inode(inode);
471
472 }
473 if (is_bad_inode(inode)) {
474 iput(inode);
475 inode = NULL;
476 }
477 return inode;
478}
479
480/* replays a single extent in 'eb' at 'slot' with 'key' into the
481 * subvolume 'root'. path is released on entry and should be released
482 * on exit.
483 *
484 * extents in the log tree have not been allocated out of the extent
485 * tree yet. So, this completes the allocation, taking a reference
486 * as required if the extent already exists or creating a new extent
487 * if it isn't in the extent allocation tree yet.
488 *
489 * The extent is inserted into the file, dropping any existing extents
490 * from the file that overlap the new one.
491 */
492static noinline int replay_one_extent(struct btrfs_trans_handle *trans,
493 struct btrfs_root *root,
494 struct btrfs_path *path,
495 struct extent_buffer *eb, int slot,
496 struct btrfs_key *key)
497{
498 int found_type;
499 u64 mask = root->sectorsize - 1;
500 u64 extent_end;
501 u64 alloc_hint;
502 u64 start = key->offset;
Yan Zheng07d400a2009-01-06 11:42:00 -0500503 u64 saved_nbytes;
Chris Masone02119d2008-09-05 16:13:11 -0400504 struct btrfs_file_extent_item *item;
505 struct inode *inode = NULL;
506 unsigned long size;
507 int ret = 0;
508
509 item = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
510 found_type = btrfs_file_extent_type(eb, item);
511
Yan Zhengd899e052008-10-30 14:25:28 -0400512 if (found_type == BTRFS_FILE_EXTENT_REG ||
513 found_type == BTRFS_FILE_EXTENT_PREALLOC)
Chris Masone02119d2008-09-05 16:13:11 -0400514 extent_end = start + btrfs_file_extent_num_bytes(eb, item);
515 else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
Chris Masonc8b97812008-10-29 14:49:59 -0400516 size = btrfs_file_extent_inline_len(eb, item);
Chris Masone02119d2008-09-05 16:13:11 -0400517 extent_end = (start + size + mask) & ~mask;
518 } else {
519 ret = 0;
520 goto out;
521 }
522
523 inode = read_one_inode(root, key->objectid);
524 if (!inode) {
525 ret = -EIO;
526 goto out;
527 }
528
529 /*
530 * first check to see if we already have this extent in the
531 * file. This must be done before the btrfs_drop_extents run
532 * so we don't try to drop this extent.
533 */
534 ret = btrfs_lookup_file_extent(trans, root, path, inode->i_ino,
535 start, 0);
536
Yan Zhengd899e052008-10-30 14:25:28 -0400537 if (ret == 0 &&
538 (found_type == BTRFS_FILE_EXTENT_REG ||
539 found_type == BTRFS_FILE_EXTENT_PREALLOC)) {
Chris Masone02119d2008-09-05 16:13:11 -0400540 struct btrfs_file_extent_item cmp1;
541 struct btrfs_file_extent_item cmp2;
542 struct btrfs_file_extent_item *existing;
543 struct extent_buffer *leaf;
544
545 leaf = path->nodes[0];
546 existing = btrfs_item_ptr(leaf, path->slots[0],
547 struct btrfs_file_extent_item);
548
549 read_extent_buffer(eb, &cmp1, (unsigned long)item,
550 sizeof(cmp1));
551 read_extent_buffer(leaf, &cmp2, (unsigned long)existing,
552 sizeof(cmp2));
553
554 /*
555 * we already have a pointer to this exact extent,
556 * we don't have to do anything
557 */
558 if (memcmp(&cmp1, &cmp2, sizeof(cmp1)) == 0) {
559 btrfs_release_path(root, path);
560 goto out;
561 }
562 }
563 btrfs_release_path(root, path);
564
Yan Zheng07d400a2009-01-06 11:42:00 -0500565 saved_nbytes = inode_get_bytes(inode);
Chris Masone02119d2008-09-05 16:13:11 -0400566 /* drop any overlapping extents */
567 ret = btrfs_drop_extents(trans, root, inode,
568 start, extent_end, start, &alloc_hint);
569 BUG_ON(ret);
570
Yan Zheng07d400a2009-01-06 11:42:00 -0500571 if (found_type == BTRFS_FILE_EXTENT_REG ||
572 found_type == BTRFS_FILE_EXTENT_PREALLOC) {
573 unsigned long dest_offset;
574 struct btrfs_key ins;
Chris Masone02119d2008-09-05 16:13:11 -0400575
Yan Zheng07d400a2009-01-06 11:42:00 -0500576 ret = btrfs_insert_empty_item(trans, root, path, key,
577 sizeof(*item));
578 BUG_ON(ret);
579 dest_offset = btrfs_item_ptr_offset(path->nodes[0],
580 path->slots[0]);
581 copy_extent_buffer(path->nodes[0], eb, dest_offset,
582 (unsigned long)item, sizeof(*item));
583
584 ins.objectid = btrfs_file_extent_disk_bytenr(eb, item);
585 ins.offset = btrfs_file_extent_disk_num_bytes(eb, item);
586 ins.type = BTRFS_EXTENT_ITEM_KEY;
587
588 if (ins.objectid > 0) {
589 u64 csum_start;
590 u64 csum_end;
591 LIST_HEAD(ordered_sums);
592 /*
593 * is this extent already allocated in the extent
594 * allocation tree? If so, just add a reference
595 */
596 ret = btrfs_lookup_extent(root, ins.objectid,
597 ins.offset);
598 if (ret == 0) {
599 ret = btrfs_inc_extent_ref(trans, root,
600 ins.objectid, ins.offset,
601 path->nodes[0]->start,
602 root->root_key.objectid,
603 trans->transid, key->objectid);
604 } else {
605 /*
606 * insert the extent pointer in the extent
607 * allocation tree
608 */
609 ret = btrfs_alloc_logged_extent(trans, root,
610 path->nodes[0]->start,
611 root->root_key.objectid,
612 trans->transid, key->objectid,
613 &ins);
614 BUG_ON(ret);
615 }
616 btrfs_release_path(root, path);
617
618 if (btrfs_file_extent_compression(eb, item)) {
619 csum_start = ins.objectid;
620 csum_end = csum_start + ins.offset;
621 } else {
622 csum_start = ins.objectid +
623 btrfs_file_extent_offset(eb, item);
624 csum_end = csum_start +
625 btrfs_file_extent_num_bytes(eb, item);
626 }
627
628 ret = btrfs_lookup_csums_range(root->log_root,
629 csum_start, csum_end - 1,
630 &ordered_sums);
631 BUG_ON(ret);
632 while (!list_empty(&ordered_sums)) {
633 struct btrfs_ordered_sum *sums;
634 sums = list_entry(ordered_sums.next,
635 struct btrfs_ordered_sum,
636 list);
637 ret = btrfs_csum_file_blocks(trans,
638 root->fs_info->csum_root,
639 sums);
640 BUG_ON(ret);
641 list_del(&sums->list);
642 kfree(sums);
643 }
644 } else {
645 btrfs_release_path(root, path);
646 }
647 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
648 /* inline extents are easy, we just overwrite them */
649 ret = overwrite_item(trans, root, path, eb, slot, key);
650 BUG_ON(ret);
651 }
652
653 inode_set_bytes(inode, saved_nbytes);
Chris Masone02119d2008-09-05 16:13:11 -0400654 btrfs_update_inode(trans, root, inode);
655out:
656 if (inode)
657 iput(inode);
658 return ret;
659}
660
661/*
662 * when cleaning up conflicts between the directory names in the
663 * subvolume, directory names in the log and directory names in the
664 * inode back references, we may have to unlink inodes from directories.
665 *
666 * This is a helper function to do the unlink of a specific directory
667 * item
668 */
669static noinline int drop_one_dir_item(struct btrfs_trans_handle *trans,
670 struct btrfs_root *root,
671 struct btrfs_path *path,
672 struct inode *dir,
673 struct btrfs_dir_item *di)
674{
675 struct inode *inode;
676 char *name;
677 int name_len;
678 struct extent_buffer *leaf;
679 struct btrfs_key location;
680 int ret;
681
682 leaf = path->nodes[0];
683
684 btrfs_dir_item_key_to_cpu(leaf, di, &location);
685 name_len = btrfs_dir_name_len(leaf, di);
686 name = kmalloc(name_len, GFP_NOFS);
687 read_extent_buffer(leaf, name, (unsigned long)(di + 1), name_len);
688 btrfs_release_path(root, path);
689
690 inode = read_one_inode(root, location.objectid);
691 BUG_ON(!inode);
692
Yan Zhengec051c02009-01-05 15:43:42 -0500693 ret = link_to_fixup_dir(trans, root, path, location.objectid);
694 BUG_ON(ret);
Chris Masone02119d2008-09-05 16:13:11 -0400695 ret = btrfs_unlink_inode(trans, root, dir, inode, name, name_len);
Yan Zhengec051c02009-01-05 15:43:42 -0500696 BUG_ON(ret);
Chris Masone02119d2008-09-05 16:13:11 -0400697 kfree(name);
698
699 iput(inode);
700 return ret;
701}
702
703/*
704 * helper function to see if a given name and sequence number found
705 * in an inode back reference are already in a directory and correctly
706 * point to this inode
707 */
708static noinline int inode_in_dir(struct btrfs_root *root,
709 struct btrfs_path *path,
710 u64 dirid, u64 objectid, u64 index,
711 const char *name, int name_len)
712{
713 struct btrfs_dir_item *di;
714 struct btrfs_key location;
715 int match = 0;
716
717 di = btrfs_lookup_dir_index_item(NULL, root, path, dirid,
718 index, name, name_len, 0);
719 if (di && !IS_ERR(di)) {
720 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
721 if (location.objectid != objectid)
722 goto out;
723 } else
724 goto out;
725 btrfs_release_path(root, path);
726
727 di = btrfs_lookup_dir_item(NULL, root, path, dirid, name, name_len, 0);
728 if (di && !IS_ERR(di)) {
729 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
730 if (location.objectid != objectid)
731 goto out;
732 } else
733 goto out;
734 match = 1;
735out:
736 btrfs_release_path(root, path);
737 return match;
738}
739
740/*
741 * helper function to check a log tree for a named back reference in
742 * an inode. This is used to decide if a back reference that is
743 * found in the subvolume conflicts with what we find in the log.
744 *
745 * inode backreferences may have multiple refs in a single item,
746 * during replay we process one reference at a time, and we don't
747 * want to delete valid links to a file from the subvolume if that
748 * link is also in the log.
749 */
750static noinline int backref_in_log(struct btrfs_root *log,
751 struct btrfs_key *key,
752 char *name, int namelen)
753{
754 struct btrfs_path *path;
755 struct btrfs_inode_ref *ref;
756 unsigned long ptr;
757 unsigned long ptr_end;
758 unsigned long name_ptr;
759 int found_name_len;
760 int item_size;
761 int ret;
762 int match = 0;
763
764 path = btrfs_alloc_path();
765 ret = btrfs_search_slot(NULL, log, key, path, 0, 0);
766 if (ret != 0)
767 goto out;
768
769 item_size = btrfs_item_size_nr(path->nodes[0], path->slots[0]);
770 ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
771 ptr_end = ptr + item_size;
772 while (ptr < ptr_end) {
773 ref = (struct btrfs_inode_ref *)ptr;
774 found_name_len = btrfs_inode_ref_name_len(path->nodes[0], ref);
775 if (found_name_len == namelen) {
776 name_ptr = (unsigned long)(ref + 1);
777 ret = memcmp_extent_buffer(path->nodes[0], name,
778 name_ptr, namelen);
779 if (ret == 0) {
780 match = 1;
781 goto out;
782 }
783 }
784 ptr = (unsigned long)(ref + 1) + found_name_len;
785 }
786out:
787 btrfs_free_path(path);
788 return match;
789}
790
791
792/*
793 * replay one inode back reference item found in the log tree.
794 * eb, slot and key refer to the buffer and key found in the log tree.
795 * root is the destination we are replaying into, and path is for temp
796 * use by this function. (it should be released on return).
797 */
798static noinline int add_inode_ref(struct btrfs_trans_handle *trans,
799 struct btrfs_root *root,
800 struct btrfs_root *log,
801 struct btrfs_path *path,
802 struct extent_buffer *eb, int slot,
803 struct btrfs_key *key)
804{
805 struct inode *dir;
806 int ret;
807 struct btrfs_key location;
808 struct btrfs_inode_ref *ref;
809 struct btrfs_dir_item *di;
810 struct inode *inode;
811 char *name;
812 int namelen;
813 unsigned long ref_ptr;
814 unsigned long ref_end;
815
816 location.objectid = key->objectid;
817 location.type = BTRFS_INODE_ITEM_KEY;
818 location.offset = 0;
819
820 /*
821 * it is possible that we didn't log all the parent directories
822 * for a given inode. If we don't find the dir, just don't
823 * copy the back ref in. The link count fixup code will take
824 * care of the rest
825 */
826 dir = read_one_inode(root, key->offset);
827 if (!dir)
828 return -ENOENT;
829
830 inode = read_one_inode(root, key->objectid);
831 BUG_ON(!dir);
832
833 ref_ptr = btrfs_item_ptr_offset(eb, slot);
834 ref_end = ref_ptr + btrfs_item_size_nr(eb, slot);
835
836again:
837 ref = (struct btrfs_inode_ref *)ref_ptr;
838
839 namelen = btrfs_inode_ref_name_len(eb, ref);
840 name = kmalloc(namelen, GFP_NOFS);
841 BUG_ON(!name);
842
843 read_extent_buffer(eb, name, (unsigned long)(ref + 1), namelen);
844
845 /* if we already have a perfect match, we're done */
846 if (inode_in_dir(root, path, dir->i_ino, inode->i_ino,
847 btrfs_inode_ref_index(eb, ref),
848 name, namelen)) {
849 goto out;
850 }
851
852 /*
853 * look for a conflicting back reference in the metadata.
854 * if we find one we have to unlink that name of the file
855 * before we add our new link. Later on, we overwrite any
856 * existing back reference, and we don't want to create
857 * dangling pointers in the directory.
858 */
859conflict_again:
860 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
861 if (ret == 0) {
862 char *victim_name;
863 int victim_name_len;
864 struct btrfs_inode_ref *victim_ref;
865 unsigned long ptr;
866 unsigned long ptr_end;
867 struct extent_buffer *leaf = path->nodes[0];
868
869 /* are we trying to overwrite a back ref for the root directory
870 * if so, just jump out, we're done
871 */
872 if (key->objectid == key->offset)
873 goto out_nowrite;
874
875 /* check all the names in this back reference to see
876 * if they are in the log. if so, we allow them to stay
877 * otherwise they must be unlinked as a conflict
878 */
879 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
880 ptr_end = ptr + btrfs_item_size_nr(leaf, path->slots[0]);
Chris Masond3977122009-01-05 21:25:51 -0500881 while (ptr < ptr_end) {
Chris Masone02119d2008-09-05 16:13:11 -0400882 victim_ref = (struct btrfs_inode_ref *)ptr;
883 victim_name_len = btrfs_inode_ref_name_len(leaf,
884 victim_ref);
885 victim_name = kmalloc(victim_name_len, GFP_NOFS);
886 BUG_ON(!victim_name);
887
888 read_extent_buffer(leaf, victim_name,
889 (unsigned long)(victim_ref + 1),
890 victim_name_len);
891
892 if (!backref_in_log(log, key, victim_name,
893 victim_name_len)) {
894 btrfs_inc_nlink(inode);
895 btrfs_release_path(root, path);
896 ret = btrfs_unlink_inode(trans, root, dir,
897 inode, victim_name,
898 victim_name_len);
899 kfree(victim_name);
900 btrfs_release_path(root, path);
901 goto conflict_again;
902 }
903 kfree(victim_name);
904 ptr = (unsigned long)(victim_ref + 1) + victim_name_len;
905 }
906 BUG_ON(ret);
907 }
908 btrfs_release_path(root, path);
909
910 /* look for a conflicting sequence number */
911 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
912 btrfs_inode_ref_index(eb, ref),
913 name, namelen, 0);
914 if (di && !IS_ERR(di)) {
915 ret = drop_one_dir_item(trans, root, path, dir, di);
916 BUG_ON(ret);
917 }
918 btrfs_release_path(root, path);
919
920
921 /* look for a conflicting name */
922 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
923 name, namelen, 0);
924 if (di && !IS_ERR(di)) {
925 ret = drop_one_dir_item(trans, root, path, dir, di);
926 BUG_ON(ret);
927 }
928 btrfs_release_path(root, path);
929
930 /* insert our name */
931 ret = btrfs_add_link(trans, dir, inode, name, namelen, 0,
932 btrfs_inode_ref_index(eb, ref));
933 BUG_ON(ret);
934
935 btrfs_update_inode(trans, root, inode);
936
937out:
938 ref_ptr = (unsigned long)(ref + 1) + namelen;
939 kfree(name);
940 if (ref_ptr < ref_end)
941 goto again;
942
943 /* finally write the back reference in the inode */
944 ret = overwrite_item(trans, root, path, eb, slot, key);
945 BUG_ON(ret);
946
947out_nowrite:
948 btrfs_release_path(root, path);
949 iput(dir);
950 iput(inode);
951 return 0;
952}
953
954/*
Chris Masone02119d2008-09-05 16:13:11 -0400955 * There are a few corners where the link count of the file can't
956 * be properly maintained during replay. So, instead of adding
957 * lots of complexity to the log code, we just scan the backrefs
958 * for any file that has been through replay.
959 *
960 * The scan will update the link count on the inode to reflect the
961 * number of back refs found. If it goes down to zero, the iput
962 * will free the inode.
963 */
964static noinline int fixup_inode_link_count(struct btrfs_trans_handle *trans,
965 struct btrfs_root *root,
966 struct inode *inode)
967{
968 struct btrfs_path *path;
969 int ret;
970 struct btrfs_key key;
971 u64 nlink = 0;
972 unsigned long ptr;
973 unsigned long ptr_end;
974 int name_len;
975
976 key.objectid = inode->i_ino;
977 key.type = BTRFS_INODE_REF_KEY;
978 key.offset = (u64)-1;
979
980 path = btrfs_alloc_path();
981
Chris Masond3977122009-01-05 21:25:51 -0500982 while (1) {
Chris Masone02119d2008-09-05 16:13:11 -0400983 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
984 if (ret < 0)
985 break;
986 if (ret > 0) {
987 if (path->slots[0] == 0)
988 break;
989 path->slots[0]--;
990 }
991 btrfs_item_key_to_cpu(path->nodes[0], &key,
992 path->slots[0]);
993 if (key.objectid != inode->i_ino ||
994 key.type != BTRFS_INODE_REF_KEY)
995 break;
996 ptr = btrfs_item_ptr_offset(path->nodes[0], path->slots[0]);
997 ptr_end = ptr + btrfs_item_size_nr(path->nodes[0],
998 path->slots[0]);
Chris Masond3977122009-01-05 21:25:51 -0500999 while (ptr < ptr_end) {
Chris Masone02119d2008-09-05 16:13:11 -04001000 struct btrfs_inode_ref *ref;
1001
1002 ref = (struct btrfs_inode_ref *)ptr;
1003 name_len = btrfs_inode_ref_name_len(path->nodes[0],
1004 ref);
1005 ptr = (unsigned long)(ref + 1) + name_len;
1006 nlink++;
1007 }
1008
1009 if (key.offset == 0)
1010 break;
1011 key.offset--;
1012 btrfs_release_path(root, path);
1013 }
1014 btrfs_free_path(path);
1015 if (nlink != inode->i_nlink) {
1016 inode->i_nlink = nlink;
1017 btrfs_update_inode(trans, root, inode);
1018 }
Chris Mason8d5bf1c2008-09-11 15:51:21 -04001019 BTRFS_I(inode)->index_cnt = (u64)-1;
Chris Masone02119d2008-09-05 16:13:11 -04001020
1021 return 0;
1022}
1023
1024static noinline int fixup_inode_link_counts(struct btrfs_trans_handle *trans,
1025 struct btrfs_root *root,
1026 struct btrfs_path *path)
1027{
1028 int ret;
1029 struct btrfs_key key;
1030 struct inode *inode;
1031
1032 key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID;
1033 key.type = BTRFS_ORPHAN_ITEM_KEY;
1034 key.offset = (u64)-1;
Chris Masond3977122009-01-05 21:25:51 -05001035 while (1) {
Chris Masone02119d2008-09-05 16:13:11 -04001036 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1037 if (ret < 0)
1038 break;
1039
1040 if (ret == 1) {
1041 if (path->slots[0] == 0)
1042 break;
1043 path->slots[0]--;
1044 }
1045
1046 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1047 if (key.objectid != BTRFS_TREE_LOG_FIXUP_OBJECTID ||
1048 key.type != BTRFS_ORPHAN_ITEM_KEY)
1049 break;
1050
1051 ret = btrfs_del_item(trans, root, path);
1052 BUG_ON(ret);
1053
1054 btrfs_release_path(root, path);
1055 inode = read_one_inode(root, key.offset);
1056 BUG_ON(!inode);
1057
1058 ret = fixup_inode_link_count(trans, root, inode);
1059 BUG_ON(ret);
1060
1061 iput(inode);
1062
1063 if (key.offset == 0)
1064 break;
1065 key.offset--;
1066 }
1067 btrfs_release_path(root, path);
1068 return 0;
1069}
1070
1071
1072/*
1073 * record a given inode in the fixup dir so we can check its link
1074 * count when replay is done. The link count is incremented here
1075 * so the inode won't go away until we check it
1076 */
1077static noinline int link_to_fixup_dir(struct btrfs_trans_handle *trans,
1078 struct btrfs_root *root,
1079 struct btrfs_path *path,
1080 u64 objectid)
1081{
1082 struct btrfs_key key;
1083 int ret = 0;
1084 struct inode *inode;
1085
1086 inode = read_one_inode(root, objectid);
1087 BUG_ON(!inode);
1088
1089 key.objectid = BTRFS_TREE_LOG_FIXUP_OBJECTID;
1090 btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY);
1091 key.offset = objectid;
1092
1093 ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
1094
1095 btrfs_release_path(root, path);
1096 if (ret == 0) {
1097 btrfs_inc_nlink(inode);
1098 btrfs_update_inode(trans, root, inode);
1099 } else if (ret == -EEXIST) {
1100 ret = 0;
1101 } else {
1102 BUG();
1103 }
1104 iput(inode);
1105
1106 return ret;
1107}
1108
1109/*
1110 * when replaying the log for a directory, we only insert names
1111 * for inodes that actually exist. This means an fsync on a directory
1112 * does not implicitly fsync all the new files in it
1113 */
1114static noinline int insert_one_name(struct btrfs_trans_handle *trans,
1115 struct btrfs_root *root,
1116 struct btrfs_path *path,
1117 u64 dirid, u64 index,
1118 char *name, int name_len, u8 type,
1119 struct btrfs_key *location)
1120{
1121 struct inode *inode;
1122 struct inode *dir;
1123 int ret;
1124
1125 inode = read_one_inode(root, location->objectid);
1126 if (!inode)
1127 return -ENOENT;
1128
1129 dir = read_one_inode(root, dirid);
1130 if (!dir) {
1131 iput(inode);
1132 return -EIO;
1133 }
1134 ret = btrfs_add_link(trans, dir, inode, name, name_len, 1, index);
1135
1136 /* FIXME, put inode into FIXUP list */
1137
1138 iput(inode);
1139 iput(dir);
1140 return ret;
1141}
1142
1143/*
1144 * take a single entry in a log directory item and replay it into
1145 * the subvolume.
1146 *
1147 * if a conflicting item exists in the subdirectory already,
1148 * the inode it points to is unlinked and put into the link count
1149 * fix up tree.
1150 *
1151 * If a name from the log points to a file or directory that does
1152 * not exist in the FS, it is skipped. fsyncs on directories
1153 * do not force down inodes inside that directory, just changes to the
1154 * names or unlinks in a directory.
1155 */
1156static noinline int replay_one_name(struct btrfs_trans_handle *trans,
1157 struct btrfs_root *root,
1158 struct btrfs_path *path,
1159 struct extent_buffer *eb,
1160 struct btrfs_dir_item *di,
1161 struct btrfs_key *key)
1162{
1163 char *name;
1164 int name_len;
1165 struct btrfs_dir_item *dst_di;
1166 struct btrfs_key found_key;
1167 struct btrfs_key log_key;
1168 struct inode *dir;
Chris Masone02119d2008-09-05 16:13:11 -04001169 u8 log_type;
Chris Mason4bef0842008-09-08 11:18:08 -04001170 int exists;
Chris Masone02119d2008-09-05 16:13:11 -04001171 int ret;
1172
1173 dir = read_one_inode(root, key->objectid);
1174 BUG_ON(!dir);
1175
1176 name_len = btrfs_dir_name_len(eb, di);
1177 name = kmalloc(name_len, GFP_NOFS);
1178 log_type = btrfs_dir_type(eb, di);
1179 read_extent_buffer(eb, name, (unsigned long)(di + 1),
1180 name_len);
1181
1182 btrfs_dir_item_key_to_cpu(eb, di, &log_key);
Chris Mason4bef0842008-09-08 11:18:08 -04001183 exists = btrfs_lookup_inode(trans, root, path, &log_key, 0);
1184 if (exists == 0)
1185 exists = 1;
1186 else
1187 exists = 0;
1188 btrfs_release_path(root, path);
1189
Chris Masone02119d2008-09-05 16:13:11 -04001190 if (key->type == BTRFS_DIR_ITEM_KEY) {
1191 dst_di = btrfs_lookup_dir_item(trans, root, path, key->objectid,
1192 name, name_len, 1);
Chris Masond3977122009-01-05 21:25:51 -05001193 } else if (key->type == BTRFS_DIR_INDEX_KEY) {
Chris Masone02119d2008-09-05 16:13:11 -04001194 dst_di = btrfs_lookup_dir_index_item(trans, root, path,
1195 key->objectid,
1196 key->offset, name,
1197 name_len, 1);
1198 } else {
1199 BUG();
1200 }
1201 if (!dst_di || IS_ERR(dst_di)) {
1202 /* we need a sequence number to insert, so we only
1203 * do inserts for the BTRFS_DIR_INDEX_KEY types
1204 */
1205 if (key->type != BTRFS_DIR_INDEX_KEY)
1206 goto out;
1207 goto insert;
1208 }
1209
1210 btrfs_dir_item_key_to_cpu(path->nodes[0], dst_di, &found_key);
1211 /* the existing item matches the logged item */
1212 if (found_key.objectid == log_key.objectid &&
1213 found_key.type == log_key.type &&
1214 found_key.offset == log_key.offset &&
1215 btrfs_dir_type(path->nodes[0], dst_di) == log_type) {
1216 goto out;
1217 }
1218
1219 /*
1220 * don't drop the conflicting directory entry if the inode
1221 * for the new entry doesn't exist
1222 */
Chris Mason4bef0842008-09-08 11:18:08 -04001223 if (!exists)
Chris Masone02119d2008-09-05 16:13:11 -04001224 goto out;
1225
Chris Masone02119d2008-09-05 16:13:11 -04001226 ret = drop_one_dir_item(trans, root, path, dir, dst_di);
1227 BUG_ON(ret);
1228
1229 if (key->type == BTRFS_DIR_INDEX_KEY)
1230 goto insert;
1231out:
1232 btrfs_release_path(root, path);
1233 kfree(name);
1234 iput(dir);
1235 return 0;
1236
1237insert:
1238 btrfs_release_path(root, path);
1239 ret = insert_one_name(trans, root, path, key->objectid, key->offset,
1240 name, name_len, log_type, &log_key);
1241
1242 if (ret && ret != -ENOENT)
1243 BUG();
1244 goto out;
1245}
1246
1247/*
1248 * find all the names in a directory item and reconcile them into
1249 * the subvolume. Only BTRFS_DIR_ITEM_KEY types will have more than
1250 * one name in a directory item, but the same code gets used for
1251 * both directory index types
1252 */
1253static noinline int replay_one_dir_item(struct btrfs_trans_handle *trans,
1254 struct btrfs_root *root,
1255 struct btrfs_path *path,
1256 struct extent_buffer *eb, int slot,
1257 struct btrfs_key *key)
1258{
1259 int ret;
1260 u32 item_size = btrfs_item_size_nr(eb, slot);
1261 struct btrfs_dir_item *di;
1262 int name_len;
1263 unsigned long ptr;
1264 unsigned long ptr_end;
1265
1266 ptr = btrfs_item_ptr_offset(eb, slot);
1267 ptr_end = ptr + item_size;
Chris Masond3977122009-01-05 21:25:51 -05001268 while (ptr < ptr_end) {
Chris Masone02119d2008-09-05 16:13:11 -04001269 di = (struct btrfs_dir_item *)ptr;
1270 name_len = btrfs_dir_name_len(eb, di);
1271 ret = replay_one_name(trans, root, path, eb, di, key);
1272 BUG_ON(ret);
1273 ptr = (unsigned long)(di + 1);
1274 ptr += name_len;
1275 }
1276 return 0;
1277}
1278
1279/*
1280 * directory replay has two parts. There are the standard directory
1281 * items in the log copied from the subvolume, and range items
1282 * created in the log while the subvolume was logged.
1283 *
1284 * The range items tell us which parts of the key space the log
1285 * is authoritative for. During replay, if a key in the subvolume
1286 * directory is in a logged range item, but not actually in the log
1287 * that means it was deleted from the directory before the fsync
1288 * and should be removed.
1289 */
1290static noinline int find_dir_range(struct btrfs_root *root,
1291 struct btrfs_path *path,
1292 u64 dirid, int key_type,
1293 u64 *start_ret, u64 *end_ret)
1294{
1295 struct btrfs_key key;
1296 u64 found_end;
1297 struct btrfs_dir_log_item *item;
1298 int ret;
1299 int nritems;
1300
1301 if (*start_ret == (u64)-1)
1302 return 1;
1303
1304 key.objectid = dirid;
1305 key.type = key_type;
1306 key.offset = *start_ret;
1307
1308 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1309 if (ret < 0)
1310 goto out;
1311 if (ret > 0) {
1312 if (path->slots[0] == 0)
1313 goto out;
1314 path->slots[0]--;
1315 }
1316 if (ret != 0)
1317 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1318
1319 if (key.type != key_type || key.objectid != dirid) {
1320 ret = 1;
1321 goto next;
1322 }
1323 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1324 struct btrfs_dir_log_item);
1325 found_end = btrfs_dir_log_end(path->nodes[0], item);
1326
1327 if (*start_ret >= key.offset && *start_ret <= found_end) {
1328 ret = 0;
1329 *start_ret = key.offset;
1330 *end_ret = found_end;
1331 goto out;
1332 }
1333 ret = 1;
1334next:
1335 /* check the next slot in the tree to see if it is a valid item */
1336 nritems = btrfs_header_nritems(path->nodes[0]);
1337 if (path->slots[0] >= nritems) {
1338 ret = btrfs_next_leaf(root, path);
1339 if (ret)
1340 goto out;
1341 } else {
1342 path->slots[0]++;
1343 }
1344
1345 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1346
1347 if (key.type != key_type || key.objectid != dirid) {
1348 ret = 1;
1349 goto out;
1350 }
1351 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1352 struct btrfs_dir_log_item);
1353 found_end = btrfs_dir_log_end(path->nodes[0], item);
1354 *start_ret = key.offset;
1355 *end_ret = found_end;
1356 ret = 0;
1357out:
1358 btrfs_release_path(root, path);
1359 return ret;
1360}
1361
1362/*
1363 * this looks for a given directory item in the log. If the directory
1364 * item is not in the log, the item is removed and the inode it points
1365 * to is unlinked
1366 */
1367static noinline int check_item_in_log(struct btrfs_trans_handle *trans,
1368 struct btrfs_root *root,
1369 struct btrfs_root *log,
1370 struct btrfs_path *path,
1371 struct btrfs_path *log_path,
1372 struct inode *dir,
1373 struct btrfs_key *dir_key)
1374{
1375 int ret;
1376 struct extent_buffer *eb;
1377 int slot;
1378 u32 item_size;
1379 struct btrfs_dir_item *di;
1380 struct btrfs_dir_item *log_di;
1381 int name_len;
1382 unsigned long ptr;
1383 unsigned long ptr_end;
1384 char *name;
1385 struct inode *inode;
1386 struct btrfs_key location;
1387
1388again:
1389 eb = path->nodes[0];
1390 slot = path->slots[0];
1391 item_size = btrfs_item_size_nr(eb, slot);
1392 ptr = btrfs_item_ptr_offset(eb, slot);
1393 ptr_end = ptr + item_size;
Chris Masond3977122009-01-05 21:25:51 -05001394 while (ptr < ptr_end) {
Chris Masone02119d2008-09-05 16:13:11 -04001395 di = (struct btrfs_dir_item *)ptr;
1396 name_len = btrfs_dir_name_len(eb, di);
1397 name = kmalloc(name_len, GFP_NOFS);
1398 if (!name) {
1399 ret = -ENOMEM;
1400 goto out;
1401 }
1402 read_extent_buffer(eb, name, (unsigned long)(di + 1),
1403 name_len);
1404 log_di = NULL;
1405 if (dir_key->type == BTRFS_DIR_ITEM_KEY) {
1406 log_di = btrfs_lookup_dir_item(trans, log, log_path,
1407 dir_key->objectid,
1408 name, name_len, 0);
1409 } else if (dir_key->type == BTRFS_DIR_INDEX_KEY) {
1410 log_di = btrfs_lookup_dir_index_item(trans, log,
1411 log_path,
1412 dir_key->objectid,
1413 dir_key->offset,
1414 name, name_len, 0);
1415 }
1416 if (!log_di || IS_ERR(log_di)) {
1417 btrfs_dir_item_key_to_cpu(eb, di, &location);
1418 btrfs_release_path(root, path);
1419 btrfs_release_path(log, log_path);
1420 inode = read_one_inode(root, location.objectid);
1421 BUG_ON(!inode);
1422
1423 ret = link_to_fixup_dir(trans, root,
1424 path, location.objectid);
1425 BUG_ON(ret);
1426 btrfs_inc_nlink(inode);
1427 ret = btrfs_unlink_inode(trans, root, dir, inode,
1428 name, name_len);
1429 BUG_ON(ret);
1430 kfree(name);
1431 iput(inode);
1432
1433 /* there might still be more names under this key
1434 * check and repeat if required
1435 */
1436 ret = btrfs_search_slot(NULL, root, dir_key, path,
1437 0, 0);
1438 if (ret == 0)
1439 goto again;
1440 ret = 0;
1441 goto out;
1442 }
1443 btrfs_release_path(log, log_path);
1444 kfree(name);
1445
1446 ptr = (unsigned long)(di + 1);
1447 ptr += name_len;
1448 }
1449 ret = 0;
1450out:
1451 btrfs_release_path(root, path);
1452 btrfs_release_path(log, log_path);
1453 return ret;
1454}
1455
1456/*
1457 * deletion replay happens before we copy any new directory items
1458 * out of the log or out of backreferences from inodes. It
1459 * scans the log to find ranges of keys that log is authoritative for,
1460 * and then scans the directory to find items in those ranges that are
1461 * not present in the log.
1462 *
1463 * Anything we don't find in the log is unlinked and removed from the
1464 * directory.
1465 */
1466static noinline int replay_dir_deletes(struct btrfs_trans_handle *trans,
1467 struct btrfs_root *root,
1468 struct btrfs_root *log,
1469 struct btrfs_path *path,
1470 u64 dirid)
1471{
1472 u64 range_start;
1473 u64 range_end;
1474 int key_type = BTRFS_DIR_LOG_ITEM_KEY;
1475 int ret = 0;
1476 struct btrfs_key dir_key;
1477 struct btrfs_key found_key;
1478 struct btrfs_path *log_path;
1479 struct inode *dir;
1480
1481 dir_key.objectid = dirid;
1482 dir_key.type = BTRFS_DIR_ITEM_KEY;
1483 log_path = btrfs_alloc_path();
1484 if (!log_path)
1485 return -ENOMEM;
1486
1487 dir = read_one_inode(root, dirid);
1488 /* it isn't an error if the inode isn't there, that can happen
1489 * because we replay the deletes before we copy in the inode item
1490 * from the log
1491 */
1492 if (!dir) {
1493 btrfs_free_path(log_path);
1494 return 0;
1495 }
1496again:
1497 range_start = 0;
1498 range_end = 0;
Chris Masond3977122009-01-05 21:25:51 -05001499 while (1) {
Chris Masone02119d2008-09-05 16:13:11 -04001500 ret = find_dir_range(log, path, dirid, key_type,
1501 &range_start, &range_end);
1502 if (ret != 0)
1503 break;
1504
1505 dir_key.offset = range_start;
Chris Masond3977122009-01-05 21:25:51 -05001506 while (1) {
Chris Masone02119d2008-09-05 16:13:11 -04001507 int nritems;
1508 ret = btrfs_search_slot(NULL, root, &dir_key, path,
1509 0, 0);
1510 if (ret < 0)
1511 goto out;
1512
1513 nritems = btrfs_header_nritems(path->nodes[0]);
1514 if (path->slots[0] >= nritems) {
1515 ret = btrfs_next_leaf(root, path);
1516 if (ret)
1517 break;
1518 }
1519 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
1520 path->slots[0]);
1521 if (found_key.objectid != dirid ||
1522 found_key.type != dir_key.type)
1523 goto next_type;
1524
1525 if (found_key.offset > range_end)
1526 break;
1527
1528 ret = check_item_in_log(trans, root, log, path,
1529 log_path, dir, &found_key);
1530 BUG_ON(ret);
1531 if (found_key.offset == (u64)-1)
1532 break;
1533 dir_key.offset = found_key.offset + 1;
1534 }
1535 btrfs_release_path(root, path);
1536 if (range_end == (u64)-1)
1537 break;
1538 range_start = range_end + 1;
1539 }
1540
1541next_type:
1542 ret = 0;
1543 if (key_type == BTRFS_DIR_LOG_ITEM_KEY) {
1544 key_type = BTRFS_DIR_LOG_INDEX_KEY;
1545 dir_key.type = BTRFS_DIR_INDEX_KEY;
1546 btrfs_release_path(root, path);
1547 goto again;
1548 }
1549out:
1550 btrfs_release_path(root, path);
1551 btrfs_free_path(log_path);
1552 iput(dir);
1553 return ret;
1554}
1555
1556/*
1557 * the process_func used to replay items from the log tree. This
1558 * gets called in two different stages. The first stage just looks
1559 * for inodes and makes sure they are all copied into the subvolume.
1560 *
1561 * The second stage copies all the other item types from the log into
1562 * the subvolume. The two stage approach is slower, but gets rid of
1563 * lots of complexity around inodes referencing other inodes that exist
1564 * only in the log (references come from either directory items or inode
1565 * back refs).
1566 */
1567static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
1568 struct walk_control *wc, u64 gen)
1569{
1570 int nritems;
1571 struct btrfs_path *path;
1572 struct btrfs_root *root = wc->replay_dest;
1573 struct btrfs_key key;
1574 u32 item_size;
1575 int level;
1576 int i;
1577 int ret;
1578
1579 btrfs_read_buffer(eb, gen);
1580
1581 level = btrfs_header_level(eb);
1582
1583 if (level != 0)
1584 return 0;
1585
1586 path = btrfs_alloc_path();
1587 BUG_ON(!path);
1588
1589 nritems = btrfs_header_nritems(eb);
1590 for (i = 0; i < nritems; i++) {
1591 btrfs_item_key_to_cpu(eb, &key, i);
1592 item_size = btrfs_item_size_nr(eb, i);
1593
1594 /* inode keys are done during the first stage */
1595 if (key.type == BTRFS_INODE_ITEM_KEY &&
1596 wc->stage == LOG_WALK_REPLAY_INODES) {
1597 struct inode *inode;
1598 struct btrfs_inode_item *inode_item;
1599 u32 mode;
1600
1601 inode_item = btrfs_item_ptr(eb, i,
1602 struct btrfs_inode_item);
1603 mode = btrfs_inode_mode(eb, inode_item);
1604 if (S_ISDIR(mode)) {
1605 ret = replay_dir_deletes(wc->trans,
1606 root, log, path, key.objectid);
1607 BUG_ON(ret);
1608 }
1609 ret = overwrite_item(wc->trans, root, path,
1610 eb, i, &key);
1611 BUG_ON(ret);
1612
1613 /* for regular files, truncate away
1614 * extents past the new EOF
1615 */
1616 if (S_ISREG(mode)) {
1617 inode = read_one_inode(root,
1618 key.objectid);
1619 BUG_ON(!inode);
1620
1621 ret = btrfs_truncate_inode_items(wc->trans,
1622 root, inode, inode->i_size,
1623 BTRFS_EXTENT_DATA_KEY);
1624 BUG_ON(ret);
1625 iput(inode);
1626 }
1627 ret = link_to_fixup_dir(wc->trans, root,
1628 path, key.objectid);
1629 BUG_ON(ret);
1630 }
1631 if (wc->stage < LOG_WALK_REPLAY_ALL)
1632 continue;
1633
1634 /* these keys are simply copied */
1635 if (key.type == BTRFS_XATTR_ITEM_KEY) {
1636 ret = overwrite_item(wc->trans, root, path,
1637 eb, i, &key);
1638 BUG_ON(ret);
1639 } else if (key.type == BTRFS_INODE_REF_KEY) {
1640 ret = add_inode_ref(wc->trans, root, log, path,
1641 eb, i, &key);
1642 BUG_ON(ret && ret != -ENOENT);
1643 } else if (key.type == BTRFS_EXTENT_DATA_KEY) {
1644 ret = replay_one_extent(wc->trans, root, path,
1645 eb, i, &key);
1646 BUG_ON(ret);
Chris Masone02119d2008-09-05 16:13:11 -04001647 } else if (key.type == BTRFS_DIR_ITEM_KEY ||
1648 key.type == BTRFS_DIR_INDEX_KEY) {
1649 ret = replay_one_dir_item(wc->trans, root, path,
1650 eb, i, &key);
1651 BUG_ON(ret);
1652 }
1653 }
1654 btrfs_free_path(path);
1655 return 0;
1656}
1657
Chris Masond3977122009-01-05 21:25:51 -05001658static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
Chris Masone02119d2008-09-05 16:13:11 -04001659 struct btrfs_root *root,
1660 struct btrfs_path *path, int *level,
1661 struct walk_control *wc)
1662{
1663 u64 root_owner;
1664 u64 root_gen;
1665 u64 bytenr;
1666 u64 ptr_gen;
1667 struct extent_buffer *next;
1668 struct extent_buffer *cur;
1669 struct extent_buffer *parent;
1670 u32 blocksize;
1671 int ret = 0;
1672
1673 WARN_ON(*level < 0);
1674 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1675
Chris Masond3977122009-01-05 21:25:51 -05001676 while (*level > 0) {
Chris Masone02119d2008-09-05 16:13:11 -04001677 WARN_ON(*level < 0);
1678 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1679 cur = path->nodes[*level];
1680
1681 if (btrfs_header_level(cur) != *level)
1682 WARN_ON(1);
1683
1684 if (path->slots[*level] >=
1685 btrfs_header_nritems(cur))
1686 break;
1687
1688 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
1689 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
1690 blocksize = btrfs_level_size(root, *level - 1);
1691
1692 parent = path->nodes[*level];
1693 root_owner = btrfs_header_owner(parent);
1694 root_gen = btrfs_header_generation(parent);
1695
1696 next = btrfs_find_create_tree_block(root, bytenr, blocksize);
1697
1698 wc->process_func(root, next, wc, ptr_gen);
1699
1700 if (*level == 1) {
1701 path->slots[*level]++;
1702 if (wc->free) {
1703 btrfs_read_buffer(next, ptr_gen);
1704
1705 btrfs_tree_lock(next);
1706 clean_tree_block(trans, root, next);
1707 btrfs_wait_tree_block_writeback(next);
1708 btrfs_tree_unlock(next);
1709
1710 ret = btrfs_drop_leaf_ref(trans, root, next);
1711 BUG_ON(ret);
1712
1713 WARN_ON(root_owner !=
1714 BTRFS_TREE_LOG_OBJECTID);
Chris Masond00aff02008-09-11 15:54:42 -04001715 ret = btrfs_free_reserved_extent(root,
1716 bytenr, blocksize);
Chris Masone02119d2008-09-05 16:13:11 -04001717 BUG_ON(ret);
1718 }
1719 free_extent_buffer(next);
1720 continue;
1721 }
1722 btrfs_read_buffer(next, ptr_gen);
1723
1724 WARN_ON(*level <= 0);
1725 if (path->nodes[*level-1])
1726 free_extent_buffer(path->nodes[*level-1]);
1727 path->nodes[*level-1] = next;
1728 *level = btrfs_header_level(next);
1729 path->slots[*level] = 0;
1730 cond_resched();
1731 }
1732 WARN_ON(*level < 0);
1733 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1734
Chris Masond3977122009-01-05 21:25:51 -05001735 if (path->nodes[*level] == root->node)
Chris Masone02119d2008-09-05 16:13:11 -04001736 parent = path->nodes[*level];
Chris Masond3977122009-01-05 21:25:51 -05001737 else
Chris Masone02119d2008-09-05 16:13:11 -04001738 parent = path->nodes[*level + 1];
Chris Masond3977122009-01-05 21:25:51 -05001739
Chris Masone02119d2008-09-05 16:13:11 -04001740 bytenr = path->nodes[*level]->start;
1741
1742 blocksize = btrfs_level_size(root, *level);
1743 root_owner = btrfs_header_owner(parent);
1744 root_gen = btrfs_header_generation(parent);
1745
1746 wc->process_func(root, path->nodes[*level], wc,
1747 btrfs_header_generation(path->nodes[*level]));
1748
1749 if (wc->free) {
1750 next = path->nodes[*level];
1751 btrfs_tree_lock(next);
1752 clean_tree_block(trans, root, next);
1753 btrfs_wait_tree_block_writeback(next);
1754 btrfs_tree_unlock(next);
1755
1756 if (*level == 0) {
1757 ret = btrfs_drop_leaf_ref(trans, root, next);
1758 BUG_ON(ret);
1759 }
1760 WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID);
Chris Masond00aff02008-09-11 15:54:42 -04001761 ret = btrfs_free_reserved_extent(root, bytenr, blocksize);
Chris Masone02119d2008-09-05 16:13:11 -04001762 BUG_ON(ret);
1763 }
1764 free_extent_buffer(path->nodes[*level]);
1765 path->nodes[*level] = NULL;
1766 *level += 1;
1767
1768 cond_resched();
1769 return 0;
1770}
1771
Chris Masond3977122009-01-05 21:25:51 -05001772static noinline int walk_up_log_tree(struct btrfs_trans_handle *trans,
Chris Masone02119d2008-09-05 16:13:11 -04001773 struct btrfs_root *root,
1774 struct btrfs_path *path, int *level,
1775 struct walk_control *wc)
1776{
1777 u64 root_owner;
1778 u64 root_gen;
1779 int i;
1780 int slot;
1781 int ret;
1782
Chris Masond3977122009-01-05 21:25:51 -05001783 for (i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
Chris Masone02119d2008-09-05 16:13:11 -04001784 slot = path->slots[i];
1785 if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
1786 struct extent_buffer *node;
1787 node = path->nodes[i];
1788 path->slots[i]++;
1789 *level = i;
1790 WARN_ON(*level == 0);
1791 return 0;
1792 } else {
Zheng Yan31840ae2008-09-23 13:14:14 -04001793 struct extent_buffer *parent;
1794 if (path->nodes[*level] == root->node)
1795 parent = path->nodes[*level];
1796 else
1797 parent = path->nodes[*level + 1];
1798
1799 root_owner = btrfs_header_owner(parent);
1800 root_gen = btrfs_header_generation(parent);
Chris Masone02119d2008-09-05 16:13:11 -04001801 wc->process_func(root, path->nodes[*level], wc,
1802 btrfs_header_generation(path->nodes[*level]));
1803 if (wc->free) {
1804 struct extent_buffer *next;
1805
1806 next = path->nodes[*level];
1807
1808 btrfs_tree_lock(next);
1809 clean_tree_block(trans, root, next);
1810 btrfs_wait_tree_block_writeback(next);
1811 btrfs_tree_unlock(next);
1812
1813 if (*level == 0) {
1814 ret = btrfs_drop_leaf_ref(trans, root,
1815 next);
1816 BUG_ON(ret);
1817 }
1818
1819 WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID);
Chris Masond00aff02008-09-11 15:54:42 -04001820 ret = btrfs_free_reserved_extent(root,
Chris Masone02119d2008-09-05 16:13:11 -04001821 path->nodes[*level]->start,
Chris Masond00aff02008-09-11 15:54:42 -04001822 path->nodes[*level]->len);
Chris Masone02119d2008-09-05 16:13:11 -04001823 BUG_ON(ret);
1824 }
1825 free_extent_buffer(path->nodes[*level]);
1826 path->nodes[*level] = NULL;
1827 *level = i + 1;
1828 }
1829 }
1830 return 1;
1831}
1832
1833/*
1834 * drop the reference count on the tree rooted at 'snap'. This traverses
1835 * the tree freeing any blocks that have a ref count of zero after being
1836 * decremented.
1837 */
1838static int walk_log_tree(struct btrfs_trans_handle *trans,
1839 struct btrfs_root *log, struct walk_control *wc)
1840{
1841 int ret = 0;
1842 int wret;
1843 int level;
1844 struct btrfs_path *path;
1845 int i;
1846 int orig_level;
1847
1848 path = btrfs_alloc_path();
1849 BUG_ON(!path);
1850
1851 level = btrfs_header_level(log->node);
1852 orig_level = level;
1853 path->nodes[level] = log->node;
1854 extent_buffer_get(log->node);
1855 path->slots[level] = 0;
1856
Chris Masond3977122009-01-05 21:25:51 -05001857 while (1) {
Chris Masone02119d2008-09-05 16:13:11 -04001858 wret = walk_down_log_tree(trans, log, path, &level, wc);
1859 if (wret > 0)
1860 break;
1861 if (wret < 0)
1862 ret = wret;
1863
1864 wret = walk_up_log_tree(trans, log, path, &level, wc);
1865 if (wret > 0)
1866 break;
1867 if (wret < 0)
1868 ret = wret;
1869 }
1870
1871 /* was the root node processed? if not, catch it here */
1872 if (path->nodes[orig_level]) {
1873 wc->process_func(log, path->nodes[orig_level], wc,
1874 btrfs_header_generation(path->nodes[orig_level]));
1875 if (wc->free) {
1876 struct extent_buffer *next;
1877
1878 next = path->nodes[orig_level];
1879
1880 btrfs_tree_lock(next);
1881 clean_tree_block(trans, log, next);
1882 btrfs_wait_tree_block_writeback(next);
1883 btrfs_tree_unlock(next);
1884
1885 if (orig_level == 0) {
1886 ret = btrfs_drop_leaf_ref(trans, log,
1887 next);
1888 BUG_ON(ret);
1889 }
1890 WARN_ON(log->root_key.objectid !=
1891 BTRFS_TREE_LOG_OBJECTID);
Chris Masond00aff02008-09-11 15:54:42 -04001892 ret = btrfs_free_reserved_extent(log, next->start,
1893 next->len);
Chris Masone02119d2008-09-05 16:13:11 -04001894 BUG_ON(ret);
1895 }
1896 }
1897
1898 for (i = 0; i <= orig_level; i++) {
1899 if (path->nodes[i]) {
1900 free_extent_buffer(path->nodes[i]);
1901 path->nodes[i] = NULL;
1902 }
1903 }
1904 btrfs_free_path(path);
1905 if (wc->free)
1906 free_extent_buffer(log->node);
1907 return ret;
1908}
1909
Christoph Hellwigb2950862008-12-02 09:54:17 -05001910static int wait_log_commit(struct btrfs_root *log)
Chris Masone02119d2008-09-05 16:13:11 -04001911{
1912 DEFINE_WAIT(wait);
1913 u64 transid = log->fs_info->tree_log_transid;
1914
1915 do {
1916 prepare_to_wait(&log->fs_info->tree_log_wait, &wait,
1917 TASK_UNINTERRUPTIBLE);
1918 mutex_unlock(&log->fs_info->tree_log_mutex);
1919 if (atomic_read(&log->fs_info->tree_log_commit))
1920 schedule();
1921 finish_wait(&log->fs_info->tree_log_wait, &wait);
1922 mutex_lock(&log->fs_info->tree_log_mutex);
Chris Masond3977122009-01-05 21:25:51 -05001923 } while (transid == log->fs_info->tree_log_transid &&
Chris Masone02119d2008-09-05 16:13:11 -04001924 atomic_read(&log->fs_info->tree_log_commit));
1925 return 0;
1926}
1927
1928/*
1929 * btrfs_sync_log does sends a given tree log down to the disk and
1930 * updates the super blocks to record it. When this call is done,
1931 * you know that any inodes previously logged are safely on disk
1932 */
1933int btrfs_sync_log(struct btrfs_trans_handle *trans,
1934 struct btrfs_root *root)
1935{
1936 int ret;
1937 unsigned long batch;
1938 struct btrfs_root *log = root->log_root;
Chris Masone02119d2008-09-05 16:13:11 -04001939
1940 mutex_lock(&log->fs_info->tree_log_mutex);
1941 if (atomic_read(&log->fs_info->tree_log_commit)) {
1942 wait_log_commit(log);
1943 goto out;
1944 }
1945 atomic_set(&log->fs_info->tree_log_commit, 1);
1946
Chris Masond3977122009-01-05 21:25:51 -05001947 while (1) {
Chris Mason49eb7e42008-09-11 15:53:12 -04001948 batch = log->fs_info->tree_log_batch;
Chris Masone02119d2008-09-05 16:13:11 -04001949 mutex_unlock(&log->fs_info->tree_log_mutex);
1950 schedule_timeout_uninterruptible(1);
1951 mutex_lock(&log->fs_info->tree_log_mutex);
Chris Masone02119d2008-09-05 16:13:11 -04001952
Chris Masond3977122009-01-05 21:25:51 -05001953 while (atomic_read(&log->fs_info->tree_log_writers)) {
Chris Masone02119d2008-09-05 16:13:11 -04001954 DEFINE_WAIT(wait);
1955 prepare_to_wait(&log->fs_info->tree_log_wait, &wait,
1956 TASK_UNINTERRUPTIBLE);
Chris Masone02119d2008-09-05 16:13:11 -04001957 mutex_unlock(&log->fs_info->tree_log_mutex);
1958 if (atomic_read(&log->fs_info->tree_log_writers))
1959 schedule();
1960 mutex_lock(&log->fs_info->tree_log_mutex);
1961 finish_wait(&log->fs_info->tree_log_wait, &wait);
1962 }
1963 if (batch == log->fs_info->tree_log_batch)
1964 break;
1965 }
Chris Masond0c803c2008-09-11 16:17:57 -04001966
1967 ret = btrfs_write_and_wait_marked_extents(log, &log->dirty_log_pages);
Chris Masone02119d2008-09-05 16:13:11 -04001968 BUG_ON(ret);
Chris Masond0c803c2008-09-11 16:17:57 -04001969 ret = btrfs_write_and_wait_marked_extents(root->fs_info->log_root_tree,
1970 &root->fs_info->log_root_tree->dirty_log_pages);
Chris Masone02119d2008-09-05 16:13:11 -04001971 BUG_ON(ret);
1972
1973 btrfs_set_super_log_root(&root->fs_info->super_for_commit,
1974 log->fs_info->log_root_tree->node->start);
1975 btrfs_set_super_log_root_level(&root->fs_info->super_for_commit,
1976 btrfs_header_level(log->fs_info->log_root_tree->node));
1977
Yan Zhenga512bbf2008-12-08 16:46:26 -05001978 write_ctree_super(trans, log->fs_info->tree_root, 2);
Chris Masone02119d2008-09-05 16:13:11 -04001979 log->fs_info->tree_log_transid++;
1980 log->fs_info->tree_log_batch = 0;
1981 atomic_set(&log->fs_info->tree_log_commit, 0);
1982 smp_mb();
1983 if (waitqueue_active(&log->fs_info->tree_log_wait))
1984 wake_up(&log->fs_info->tree_log_wait);
1985out:
1986 mutex_unlock(&log->fs_info->tree_log_mutex);
1987 return 0;
Chris Masone02119d2008-09-05 16:13:11 -04001988}
1989
Chris Mason3a5f1d42008-09-11 15:53:37 -04001990/* * free all the extents used by the tree log. This should be called
Chris Masone02119d2008-09-05 16:13:11 -04001991 * at commit time of the full transaction
1992 */
1993int btrfs_free_log(struct btrfs_trans_handle *trans, struct btrfs_root *root)
1994{
1995 int ret;
1996 struct btrfs_root *log;
1997 struct key;
Chris Masond0c803c2008-09-11 16:17:57 -04001998 u64 start;
1999 u64 end;
Chris Masone02119d2008-09-05 16:13:11 -04002000 struct walk_control wc = {
2001 .free = 1,
2002 .process_func = process_one_buffer
2003 };
2004
Yan Zheng07d400a2009-01-06 11:42:00 -05002005 if (!root->log_root || root->fs_info->log_root_recovering)
Chris Masone02119d2008-09-05 16:13:11 -04002006 return 0;
2007
2008 log = root->log_root;
2009 ret = walk_log_tree(trans, log, &wc);
2010 BUG_ON(ret);
2011
Chris Masond3977122009-01-05 21:25:51 -05002012 while (1) {
Chris Masond0c803c2008-09-11 16:17:57 -04002013 ret = find_first_extent_bit(&log->dirty_log_pages,
2014 0, &start, &end, EXTENT_DIRTY);
2015 if (ret)
2016 break;
2017
2018 clear_extent_dirty(&log->dirty_log_pages,
2019 start, end, GFP_NOFS);
2020 }
2021
Chris Masone02119d2008-09-05 16:13:11 -04002022 log = root->log_root;
2023 ret = btrfs_del_root(trans, root->fs_info->log_root_tree,
2024 &log->root_key);
2025 BUG_ON(ret);
2026 root->log_root = NULL;
2027 kfree(root->log_root);
2028 return 0;
2029}
2030
2031/*
2032 * helper function to update the item for a given subvolumes log root
2033 * in the tree of log roots
2034 */
2035static int update_log_root(struct btrfs_trans_handle *trans,
2036 struct btrfs_root *log)
2037{
2038 u64 bytenr = btrfs_root_bytenr(&log->root_item);
2039 int ret;
2040
2041 if (log->node->start == bytenr)
2042 return 0;
2043
2044 btrfs_set_root_bytenr(&log->root_item, log->node->start);
Yan Zheng84234f32008-10-29 14:49:05 -04002045 btrfs_set_root_generation(&log->root_item, trans->transid);
Chris Masone02119d2008-09-05 16:13:11 -04002046 btrfs_set_root_level(&log->root_item, btrfs_header_level(log->node));
2047 ret = btrfs_update_root(trans, log->fs_info->log_root_tree,
2048 &log->root_key, &log->root_item);
2049 BUG_ON(ret);
2050 return ret;
2051}
2052
2053/*
2054 * If both a file and directory are logged, and unlinks or renames are
2055 * mixed in, we have a few interesting corners:
2056 *
2057 * create file X in dir Y
2058 * link file X to X.link in dir Y
2059 * fsync file X
2060 * unlink file X but leave X.link
2061 * fsync dir Y
2062 *
2063 * After a crash we would expect only X.link to exist. But file X
2064 * didn't get fsync'd again so the log has back refs for X and X.link.
2065 *
2066 * We solve this by removing directory entries and inode backrefs from the
2067 * log when a file that was logged in the current transaction is
2068 * unlinked. Any later fsync will include the updated log entries, and
2069 * we'll be able to reconstruct the proper directory items from backrefs.
2070 *
2071 * This optimizations allows us to avoid relogging the entire inode
2072 * or the entire directory.
2073 */
2074int btrfs_del_dir_entries_in_log(struct btrfs_trans_handle *trans,
2075 struct btrfs_root *root,
2076 const char *name, int name_len,
2077 struct inode *dir, u64 index)
2078{
2079 struct btrfs_root *log;
2080 struct btrfs_dir_item *di;
2081 struct btrfs_path *path;
2082 int ret;
2083 int bytes_del = 0;
2084
Chris Mason3a5f1d42008-09-11 15:53:37 -04002085 if (BTRFS_I(dir)->logged_trans < trans->transid)
2086 return 0;
2087
Chris Masone02119d2008-09-05 16:13:11 -04002088 ret = join_running_log_trans(root);
2089 if (ret)
2090 return 0;
2091
2092 mutex_lock(&BTRFS_I(dir)->log_mutex);
2093
2094 log = root->log_root;
2095 path = btrfs_alloc_path();
2096 di = btrfs_lookup_dir_item(trans, log, path, dir->i_ino,
2097 name, name_len, -1);
2098 if (di && !IS_ERR(di)) {
2099 ret = btrfs_delete_one_dir_name(trans, log, path, di);
2100 bytes_del += name_len;
2101 BUG_ON(ret);
2102 }
2103 btrfs_release_path(log, path);
2104 di = btrfs_lookup_dir_index_item(trans, log, path, dir->i_ino,
2105 index, name, name_len, -1);
2106 if (di && !IS_ERR(di)) {
2107 ret = btrfs_delete_one_dir_name(trans, log, path, di);
2108 bytes_del += name_len;
2109 BUG_ON(ret);
2110 }
2111
2112 /* update the directory size in the log to reflect the names
2113 * we have removed
2114 */
2115 if (bytes_del) {
2116 struct btrfs_key key;
2117
2118 key.objectid = dir->i_ino;
2119 key.offset = 0;
2120 key.type = BTRFS_INODE_ITEM_KEY;
2121 btrfs_release_path(log, path);
2122
2123 ret = btrfs_search_slot(trans, log, &key, path, 0, 1);
2124 if (ret == 0) {
2125 struct btrfs_inode_item *item;
2126 u64 i_size;
2127
2128 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2129 struct btrfs_inode_item);
2130 i_size = btrfs_inode_size(path->nodes[0], item);
2131 if (i_size > bytes_del)
2132 i_size -= bytes_del;
2133 else
2134 i_size = 0;
2135 btrfs_set_inode_size(path->nodes[0], item, i_size);
2136 btrfs_mark_buffer_dirty(path->nodes[0]);
2137 } else
2138 ret = 0;
2139 btrfs_release_path(log, path);
2140 }
2141
2142 btrfs_free_path(path);
2143 mutex_unlock(&BTRFS_I(dir)->log_mutex);
2144 end_log_trans(root);
2145
2146 return 0;
2147}
2148
2149/* see comments for btrfs_del_dir_entries_in_log */
2150int btrfs_del_inode_ref_in_log(struct btrfs_trans_handle *trans,
2151 struct btrfs_root *root,
2152 const char *name, int name_len,
2153 struct inode *inode, u64 dirid)
2154{
2155 struct btrfs_root *log;
2156 u64 index;
2157 int ret;
2158
Chris Mason3a5f1d42008-09-11 15:53:37 -04002159 if (BTRFS_I(inode)->logged_trans < trans->transid)
2160 return 0;
2161
Chris Masone02119d2008-09-05 16:13:11 -04002162 ret = join_running_log_trans(root);
2163 if (ret)
2164 return 0;
2165 log = root->log_root;
2166 mutex_lock(&BTRFS_I(inode)->log_mutex);
2167
2168 ret = btrfs_del_inode_ref(trans, log, name, name_len, inode->i_ino,
2169 dirid, &index);
2170 mutex_unlock(&BTRFS_I(inode)->log_mutex);
2171 end_log_trans(root);
2172
Chris Masone02119d2008-09-05 16:13:11 -04002173 return ret;
2174}
2175
2176/*
2177 * creates a range item in the log for 'dirid'. first_offset and
2178 * last_offset tell us which parts of the key space the log should
2179 * be considered authoritative for.
2180 */
2181static noinline int insert_dir_log_key(struct btrfs_trans_handle *trans,
2182 struct btrfs_root *log,
2183 struct btrfs_path *path,
2184 int key_type, u64 dirid,
2185 u64 first_offset, u64 last_offset)
2186{
2187 int ret;
2188 struct btrfs_key key;
2189 struct btrfs_dir_log_item *item;
2190
2191 key.objectid = dirid;
2192 key.offset = first_offset;
2193 if (key_type == BTRFS_DIR_ITEM_KEY)
2194 key.type = BTRFS_DIR_LOG_ITEM_KEY;
2195 else
2196 key.type = BTRFS_DIR_LOG_INDEX_KEY;
2197 ret = btrfs_insert_empty_item(trans, log, path, &key, sizeof(*item));
2198 BUG_ON(ret);
2199
2200 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2201 struct btrfs_dir_log_item);
2202 btrfs_set_dir_log_end(path->nodes[0], item, last_offset);
2203 btrfs_mark_buffer_dirty(path->nodes[0]);
2204 btrfs_release_path(log, path);
2205 return 0;
2206}
2207
2208/*
2209 * log all the items included in the current transaction for a given
2210 * directory. This also creates the range items in the log tree required
2211 * to replay anything deleted before the fsync
2212 */
2213static noinline int log_dir_items(struct btrfs_trans_handle *trans,
2214 struct btrfs_root *root, struct inode *inode,
2215 struct btrfs_path *path,
2216 struct btrfs_path *dst_path, int key_type,
2217 u64 min_offset, u64 *last_offset_ret)
2218{
2219 struct btrfs_key min_key;
2220 struct btrfs_key max_key;
2221 struct btrfs_root *log = root->log_root;
2222 struct extent_buffer *src;
2223 int ret;
2224 int i;
2225 int nritems;
2226 u64 first_offset = min_offset;
2227 u64 last_offset = (u64)-1;
2228
2229 log = root->log_root;
2230 max_key.objectid = inode->i_ino;
2231 max_key.offset = (u64)-1;
2232 max_key.type = key_type;
2233
2234 min_key.objectid = inode->i_ino;
2235 min_key.type = key_type;
2236 min_key.offset = min_offset;
2237
2238 path->keep_locks = 1;
2239
2240 ret = btrfs_search_forward(root, &min_key, &max_key,
2241 path, 0, trans->transid);
2242
2243 /*
2244 * we didn't find anything from this transaction, see if there
2245 * is anything at all
2246 */
2247 if (ret != 0 || min_key.objectid != inode->i_ino ||
2248 min_key.type != key_type) {
2249 min_key.objectid = inode->i_ino;
2250 min_key.type = key_type;
2251 min_key.offset = (u64)-1;
2252 btrfs_release_path(root, path);
2253 ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
2254 if (ret < 0) {
2255 btrfs_release_path(root, path);
2256 return ret;
2257 }
2258 ret = btrfs_previous_item(root, path, inode->i_ino, key_type);
2259
2260 /* if ret == 0 there are items for this type,
2261 * create a range to tell us the last key of this type.
2262 * otherwise, there are no items in this directory after
2263 * *min_offset, and we create a range to indicate that.
2264 */
2265 if (ret == 0) {
2266 struct btrfs_key tmp;
2267 btrfs_item_key_to_cpu(path->nodes[0], &tmp,
2268 path->slots[0]);
Chris Masond3977122009-01-05 21:25:51 -05002269 if (key_type == tmp.type)
Chris Masone02119d2008-09-05 16:13:11 -04002270 first_offset = max(min_offset, tmp.offset) + 1;
Chris Masone02119d2008-09-05 16:13:11 -04002271 }
2272 goto done;
2273 }
2274
2275 /* go backward to find any previous key */
2276 ret = btrfs_previous_item(root, path, inode->i_ino, key_type);
2277 if (ret == 0) {
2278 struct btrfs_key tmp;
2279 btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]);
2280 if (key_type == tmp.type) {
2281 first_offset = tmp.offset;
2282 ret = overwrite_item(trans, log, dst_path,
2283 path->nodes[0], path->slots[0],
2284 &tmp);
2285 }
2286 }
2287 btrfs_release_path(root, path);
2288
2289 /* find the first key from this transaction again */
2290 ret = btrfs_search_slot(NULL, root, &min_key, path, 0, 0);
2291 if (ret != 0) {
2292 WARN_ON(1);
2293 goto done;
2294 }
2295
2296 /*
2297 * we have a block from this transaction, log every item in it
2298 * from our directory
2299 */
Chris Masond3977122009-01-05 21:25:51 -05002300 while (1) {
Chris Masone02119d2008-09-05 16:13:11 -04002301 struct btrfs_key tmp;
2302 src = path->nodes[0];
2303 nritems = btrfs_header_nritems(src);
2304 for (i = path->slots[0]; i < nritems; i++) {
2305 btrfs_item_key_to_cpu(src, &min_key, i);
2306
2307 if (min_key.objectid != inode->i_ino ||
2308 min_key.type != key_type)
2309 goto done;
2310 ret = overwrite_item(trans, log, dst_path, src, i,
2311 &min_key);
2312 BUG_ON(ret);
2313 }
2314 path->slots[0] = nritems;
2315
2316 /*
2317 * look ahead to the next item and see if it is also
2318 * from this directory and from this transaction
2319 */
2320 ret = btrfs_next_leaf(root, path);
2321 if (ret == 1) {
2322 last_offset = (u64)-1;
2323 goto done;
2324 }
2325 btrfs_item_key_to_cpu(path->nodes[0], &tmp, path->slots[0]);
2326 if (tmp.objectid != inode->i_ino || tmp.type != key_type) {
2327 last_offset = (u64)-1;
2328 goto done;
2329 }
2330 if (btrfs_header_generation(path->nodes[0]) != trans->transid) {
2331 ret = overwrite_item(trans, log, dst_path,
2332 path->nodes[0], path->slots[0],
2333 &tmp);
2334
2335 BUG_ON(ret);
2336 last_offset = tmp.offset;
2337 goto done;
2338 }
2339 }
2340done:
2341 *last_offset_ret = last_offset;
2342 btrfs_release_path(root, path);
2343 btrfs_release_path(log, dst_path);
2344
2345 /* insert the log range keys to indicate where the log is valid */
2346 ret = insert_dir_log_key(trans, log, path, key_type, inode->i_ino,
2347 first_offset, last_offset);
2348 BUG_ON(ret);
2349 return 0;
2350}
2351
2352/*
2353 * logging directories is very similar to logging inodes, We find all the items
2354 * from the current transaction and write them to the log.
2355 *
2356 * The recovery code scans the directory in the subvolume, and if it finds a
2357 * key in the range logged that is not present in the log tree, then it means
2358 * that dir entry was unlinked during the transaction.
2359 *
2360 * In order for that scan to work, we must include one key smaller than
2361 * the smallest logged by this transaction and one key larger than the largest
2362 * key logged by this transaction.
2363 */
2364static noinline int log_directory_changes(struct btrfs_trans_handle *trans,
2365 struct btrfs_root *root, struct inode *inode,
2366 struct btrfs_path *path,
2367 struct btrfs_path *dst_path)
2368{
2369 u64 min_key;
2370 u64 max_key;
2371 int ret;
2372 int key_type = BTRFS_DIR_ITEM_KEY;
2373
2374again:
2375 min_key = 0;
2376 max_key = 0;
Chris Masond3977122009-01-05 21:25:51 -05002377 while (1) {
Chris Masone02119d2008-09-05 16:13:11 -04002378 ret = log_dir_items(trans, root, inode, path,
2379 dst_path, key_type, min_key,
2380 &max_key);
2381 BUG_ON(ret);
2382 if (max_key == (u64)-1)
2383 break;
2384 min_key = max_key + 1;
2385 }
2386
2387 if (key_type == BTRFS_DIR_ITEM_KEY) {
2388 key_type = BTRFS_DIR_INDEX_KEY;
2389 goto again;
2390 }
2391 return 0;
2392}
2393
2394/*
2395 * a helper function to drop items from the log before we relog an
2396 * inode. max_key_type indicates the highest item type to remove.
2397 * This cannot be run for file data extents because it does not
2398 * free the extents they point to.
2399 */
2400static int drop_objectid_items(struct btrfs_trans_handle *trans,
2401 struct btrfs_root *log,
2402 struct btrfs_path *path,
2403 u64 objectid, int max_key_type)
2404{
2405 int ret;
2406 struct btrfs_key key;
2407 struct btrfs_key found_key;
2408
2409 key.objectid = objectid;
2410 key.type = max_key_type;
2411 key.offset = (u64)-1;
2412
Chris Masond3977122009-01-05 21:25:51 -05002413 while (1) {
Chris Masone02119d2008-09-05 16:13:11 -04002414 ret = btrfs_search_slot(trans, log, &key, path, -1, 1);
2415
2416 if (ret != 1)
2417 break;
2418
2419 if (path->slots[0] == 0)
2420 break;
2421
2422 path->slots[0]--;
2423 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
2424 path->slots[0]);
2425
2426 if (found_key.objectid != objectid)
2427 break;
2428
2429 ret = btrfs_del_item(trans, log, path);
2430 BUG_ON(ret);
2431 btrfs_release_path(log, path);
2432 }
2433 btrfs_release_path(log, path);
2434 return 0;
2435}
2436
Chris Mason31ff1cd2008-09-11 16:17:57 -04002437static noinline int copy_items(struct btrfs_trans_handle *trans,
2438 struct btrfs_root *log,
2439 struct btrfs_path *dst_path,
2440 struct extent_buffer *src,
2441 int start_slot, int nr, int inode_only)
2442{
2443 unsigned long src_offset;
2444 unsigned long dst_offset;
2445 struct btrfs_file_extent_item *extent;
2446 struct btrfs_inode_item *inode_item;
2447 int ret;
2448 struct btrfs_key *ins_keys;
2449 u32 *ins_sizes;
2450 char *ins_data;
2451 int i;
Chris Masond20f7042008-12-08 16:58:54 -05002452 struct list_head ordered_sums;
2453
2454 INIT_LIST_HEAD(&ordered_sums);
Chris Mason31ff1cd2008-09-11 16:17:57 -04002455
2456 ins_data = kmalloc(nr * sizeof(struct btrfs_key) +
2457 nr * sizeof(u32), GFP_NOFS);
2458 ins_sizes = (u32 *)ins_data;
2459 ins_keys = (struct btrfs_key *)(ins_data + nr * sizeof(u32));
2460
2461 for (i = 0; i < nr; i++) {
2462 ins_sizes[i] = btrfs_item_size_nr(src, i + start_slot);
2463 btrfs_item_key_to_cpu(src, ins_keys + i, i + start_slot);
2464 }
2465 ret = btrfs_insert_empty_items(trans, log, dst_path,
2466 ins_keys, ins_sizes, nr);
2467 BUG_ON(ret);
2468
2469 for (i = 0; i < nr; i++) {
2470 dst_offset = btrfs_item_ptr_offset(dst_path->nodes[0],
2471 dst_path->slots[0]);
2472
2473 src_offset = btrfs_item_ptr_offset(src, start_slot + i);
2474
2475 copy_extent_buffer(dst_path->nodes[0], src, dst_offset,
2476 src_offset, ins_sizes[i]);
2477
2478 if (inode_only == LOG_INODE_EXISTS &&
2479 ins_keys[i].type == BTRFS_INODE_ITEM_KEY) {
2480 inode_item = btrfs_item_ptr(dst_path->nodes[0],
2481 dst_path->slots[0],
2482 struct btrfs_inode_item);
2483 btrfs_set_inode_size(dst_path->nodes[0], inode_item, 0);
2484
2485 /* set the generation to zero so the recover code
2486 * can tell the difference between an logging
2487 * just to say 'this inode exists' and a logging
2488 * to say 'update this inode with these values'
2489 */
2490 btrfs_set_inode_generation(dst_path->nodes[0],
2491 inode_item, 0);
2492 }
2493 /* take a reference on file data extents so that truncates
2494 * or deletes of this inode don't have to relog the inode
2495 * again
2496 */
2497 if (btrfs_key_type(ins_keys + i) == BTRFS_EXTENT_DATA_KEY) {
2498 int found_type;
2499 extent = btrfs_item_ptr(src, start_slot + i,
2500 struct btrfs_file_extent_item);
2501
2502 found_type = btrfs_file_extent_type(src, extent);
Yan Zhengd899e052008-10-30 14:25:28 -04002503 if (found_type == BTRFS_FILE_EXTENT_REG ||
2504 found_type == BTRFS_FILE_EXTENT_PREALLOC) {
Chris Mason31ff1cd2008-09-11 16:17:57 -04002505 u64 ds = btrfs_file_extent_disk_bytenr(src,
2506 extent);
2507 u64 dl = btrfs_file_extent_disk_num_bytes(src,
2508 extent);
Chris Masond20f7042008-12-08 16:58:54 -05002509 u64 cs = btrfs_file_extent_offset(src, extent);
2510 u64 cl = btrfs_file_extent_num_bytes(src,
2511 extent);;
Chris Mason580afd72008-12-08 19:15:39 -05002512 if (btrfs_file_extent_compression(src,
2513 extent)) {
2514 cs = 0;
2515 cl = dl;
2516 }
Chris Mason31ff1cd2008-09-11 16:17:57 -04002517 /* ds == 0 is a hole */
2518 if (ds != 0) {
2519 ret = btrfs_inc_extent_ref(trans, log,
2520 ds, dl,
Zheng Yan31840ae2008-09-23 13:14:14 -04002521 dst_path->nodes[0]->start,
Chris Mason31ff1cd2008-09-11 16:17:57 -04002522 BTRFS_TREE_LOG_OBJECTID,
Zheng Yan31840ae2008-09-23 13:14:14 -04002523 trans->transid,
Yan Zheng3bb1a1b2008-10-09 11:46:24 -04002524 ins_keys[i].objectid);
Chris Mason31ff1cd2008-09-11 16:17:57 -04002525 BUG_ON(ret);
Yan Zheng07d400a2009-01-06 11:42:00 -05002526 ret = btrfs_lookup_csums_range(
2527 log->fs_info->csum_root,
2528 ds + cs, ds + cs + cl - 1,
2529 &ordered_sums);
Chris Masond20f7042008-12-08 16:58:54 -05002530 BUG_ON(ret);
Chris Mason31ff1cd2008-09-11 16:17:57 -04002531 }
2532 }
2533 }
2534 dst_path->slots[0]++;
2535 }
2536
2537 btrfs_mark_buffer_dirty(dst_path->nodes[0]);
2538 btrfs_release_path(log, dst_path);
2539 kfree(ins_data);
Chris Masond20f7042008-12-08 16:58:54 -05002540
2541 /*
2542 * we have to do this after the loop above to avoid changing the
2543 * log tree while trying to change the log tree.
2544 */
Chris Masond3977122009-01-05 21:25:51 -05002545 while (!list_empty(&ordered_sums)) {
Chris Masond20f7042008-12-08 16:58:54 -05002546 struct btrfs_ordered_sum *sums = list_entry(ordered_sums.next,
2547 struct btrfs_ordered_sum,
2548 list);
2549 ret = btrfs_csum_file_blocks(trans, log, sums);
2550 BUG_ON(ret);
2551 list_del(&sums->list);
2552 kfree(sums);
2553 }
Chris Mason31ff1cd2008-09-11 16:17:57 -04002554 return 0;
2555}
2556
Chris Masone02119d2008-09-05 16:13:11 -04002557/* log a single inode in the tree log.
2558 * At least one parent directory for this inode must exist in the tree
2559 * or be logged already.
2560 *
2561 * Any items from this inode changed by the current transaction are copied
2562 * to the log tree. An extra reference is taken on any extents in this
2563 * file, allowing us to avoid a whole pile of corner cases around logging
2564 * blocks that have been removed from the tree.
2565 *
2566 * See LOG_INODE_ALL and related defines for a description of what inode_only
2567 * does.
2568 *
2569 * This handles both files and directories.
2570 */
2571static int __btrfs_log_inode(struct btrfs_trans_handle *trans,
2572 struct btrfs_root *root, struct inode *inode,
2573 int inode_only)
2574{
2575 struct btrfs_path *path;
2576 struct btrfs_path *dst_path;
2577 struct btrfs_key min_key;
2578 struct btrfs_key max_key;
2579 struct btrfs_root *log = root->log_root;
Chris Mason31ff1cd2008-09-11 16:17:57 -04002580 struct extent_buffer *src = NULL;
Chris Masone02119d2008-09-05 16:13:11 -04002581 u32 size;
2582 int ret;
Chris Mason3a5f1d42008-09-11 15:53:37 -04002583 int nritems;
Chris Mason31ff1cd2008-09-11 16:17:57 -04002584 int ins_start_slot = 0;
2585 int ins_nr;
Chris Masone02119d2008-09-05 16:13:11 -04002586
2587 log = root->log_root;
2588
2589 path = btrfs_alloc_path();
2590 dst_path = btrfs_alloc_path();
2591
2592 min_key.objectid = inode->i_ino;
2593 min_key.type = BTRFS_INODE_ITEM_KEY;
2594 min_key.offset = 0;
2595
2596 max_key.objectid = inode->i_ino;
2597 if (inode_only == LOG_INODE_EXISTS || S_ISDIR(inode->i_mode))
2598 max_key.type = BTRFS_XATTR_ITEM_KEY;
2599 else
2600 max_key.type = (u8)-1;
2601 max_key.offset = (u64)-1;
2602
2603 /*
2604 * if this inode has already been logged and we're in inode_only
2605 * mode, we don't want to delete the things that have already
2606 * been written to the log.
2607 *
2608 * But, if the inode has been through an inode_only log,
2609 * the logged_trans field is not set. This allows us to catch
2610 * any new names for this inode in the backrefs by logging it
2611 * again
2612 */
2613 if (inode_only == LOG_INODE_EXISTS &&
2614 BTRFS_I(inode)->logged_trans == trans->transid) {
2615 btrfs_free_path(path);
2616 btrfs_free_path(dst_path);
2617 goto out;
2618 }
2619 mutex_lock(&BTRFS_I(inode)->log_mutex);
2620
2621 /*
2622 * a brute force approach to making sure we get the most uptodate
2623 * copies of everything.
2624 */
2625 if (S_ISDIR(inode->i_mode)) {
2626 int max_key_type = BTRFS_DIR_LOG_INDEX_KEY;
2627
2628 if (inode_only == LOG_INODE_EXISTS)
2629 max_key_type = BTRFS_XATTR_ITEM_KEY;
2630 ret = drop_objectid_items(trans, log, path,
2631 inode->i_ino, max_key_type);
2632 } else {
2633 ret = btrfs_truncate_inode_items(trans, log, inode, 0, 0);
2634 }
2635 BUG_ON(ret);
2636 path->keep_locks = 1;
2637
Chris Masond3977122009-01-05 21:25:51 -05002638 while (1) {
Chris Mason31ff1cd2008-09-11 16:17:57 -04002639 ins_nr = 0;
Chris Masone02119d2008-09-05 16:13:11 -04002640 ret = btrfs_search_forward(root, &min_key, &max_key,
2641 path, 0, trans->transid);
2642 if (ret != 0)
2643 break;
Chris Mason3a5f1d42008-09-11 15:53:37 -04002644again:
Chris Mason31ff1cd2008-09-11 16:17:57 -04002645 /* note, ins_nr might be > 0 here, cleanup outside the loop */
Chris Masone02119d2008-09-05 16:13:11 -04002646 if (min_key.objectid != inode->i_ino)
2647 break;
2648 if (min_key.type > max_key.type)
2649 break;
Chris Mason31ff1cd2008-09-11 16:17:57 -04002650
Chris Masone02119d2008-09-05 16:13:11 -04002651 src = path->nodes[0];
2652 size = btrfs_item_size_nr(src, path->slots[0]);
Chris Mason31ff1cd2008-09-11 16:17:57 -04002653 if (ins_nr && ins_start_slot + ins_nr == path->slots[0]) {
2654 ins_nr++;
2655 goto next_slot;
2656 } else if (!ins_nr) {
2657 ins_start_slot = path->slots[0];
2658 ins_nr = 1;
2659 goto next_slot;
Chris Masone02119d2008-09-05 16:13:11 -04002660 }
2661
Chris Mason31ff1cd2008-09-11 16:17:57 -04002662 ret = copy_items(trans, log, dst_path, src, ins_start_slot,
2663 ins_nr, inode_only);
2664 BUG_ON(ret);
2665 ins_nr = 1;
2666 ins_start_slot = path->slots[0];
2667next_slot:
Chris Masone02119d2008-09-05 16:13:11 -04002668
Chris Mason3a5f1d42008-09-11 15:53:37 -04002669 nritems = btrfs_header_nritems(path->nodes[0]);
2670 path->slots[0]++;
2671 if (path->slots[0] < nritems) {
2672 btrfs_item_key_to_cpu(path->nodes[0], &min_key,
2673 path->slots[0]);
2674 goto again;
2675 }
Chris Mason31ff1cd2008-09-11 16:17:57 -04002676 if (ins_nr) {
2677 ret = copy_items(trans, log, dst_path, src,
2678 ins_start_slot,
2679 ins_nr, inode_only);
2680 BUG_ON(ret);
2681 ins_nr = 0;
2682 }
Chris Mason3a5f1d42008-09-11 15:53:37 -04002683 btrfs_release_path(root, path);
2684
Chris Masone02119d2008-09-05 16:13:11 -04002685 if (min_key.offset < (u64)-1)
2686 min_key.offset++;
2687 else if (min_key.type < (u8)-1)
2688 min_key.type++;
2689 else if (min_key.objectid < (u64)-1)
2690 min_key.objectid++;
2691 else
2692 break;
2693 }
Chris Mason31ff1cd2008-09-11 16:17:57 -04002694 if (ins_nr) {
2695 ret = copy_items(trans, log, dst_path, src,
2696 ins_start_slot,
2697 ins_nr, inode_only);
2698 BUG_ON(ret);
2699 ins_nr = 0;
2700 }
2701 WARN_ON(ins_nr);
Chris Mason9623f9a2008-09-11 17:42:42 -04002702 if (inode_only == LOG_INODE_ALL && S_ISDIR(inode->i_mode)) {
Chris Masone02119d2008-09-05 16:13:11 -04002703 btrfs_release_path(root, path);
2704 btrfs_release_path(log, dst_path);
Chris Mason49eb7e42008-09-11 15:53:12 -04002705 BTRFS_I(inode)->log_dirty_trans = 0;
Chris Masone02119d2008-09-05 16:13:11 -04002706 ret = log_directory_changes(trans, root, inode, path, dst_path);
2707 BUG_ON(ret);
2708 }
Chris Mason3a5f1d42008-09-11 15:53:37 -04002709 BTRFS_I(inode)->logged_trans = trans->transid;
Chris Masone02119d2008-09-05 16:13:11 -04002710 mutex_unlock(&BTRFS_I(inode)->log_mutex);
2711
2712 btrfs_free_path(path);
2713 btrfs_free_path(dst_path);
2714
2715 mutex_lock(&root->fs_info->tree_log_mutex);
2716 ret = update_log_root(trans, log);
2717 BUG_ON(ret);
2718 mutex_unlock(&root->fs_info->tree_log_mutex);
2719out:
2720 return 0;
2721}
2722
2723int btrfs_log_inode(struct btrfs_trans_handle *trans,
2724 struct btrfs_root *root, struct inode *inode,
2725 int inode_only)
2726{
2727 int ret;
2728
2729 start_log_trans(trans, root);
2730 ret = __btrfs_log_inode(trans, root, inode, inode_only);
2731 end_log_trans(root);
2732 return ret;
2733}
2734
2735/*
2736 * helper function around btrfs_log_inode to make sure newly created
2737 * parent directories also end up in the log. A minimal inode and backref
2738 * only logging is done of any parent directories that are older than
2739 * the last committed transaction
2740 */
2741int btrfs_log_dentry(struct btrfs_trans_handle *trans,
2742 struct btrfs_root *root, struct dentry *dentry)
2743{
2744 int inode_only = LOG_INODE_ALL;
2745 struct super_block *sb;
2746 int ret;
2747
2748 start_log_trans(trans, root);
2749 sb = dentry->d_inode->i_sb;
Chris Masond3977122009-01-05 21:25:51 -05002750 while (1) {
Chris Masone02119d2008-09-05 16:13:11 -04002751 ret = __btrfs_log_inode(trans, root, dentry->d_inode,
2752 inode_only);
2753 BUG_ON(ret);
2754 inode_only = LOG_INODE_EXISTS;
2755
2756 dentry = dentry->d_parent;
2757 if (!dentry || !dentry->d_inode || sb != dentry->d_inode->i_sb)
2758 break;
2759
2760 if (BTRFS_I(dentry->d_inode)->generation <=
2761 root->fs_info->last_trans_committed)
2762 break;
2763 }
2764 end_log_trans(root);
2765 return 0;
2766}
2767
2768/*
2769 * it is not safe to log dentry if the chunk root has added new
2770 * chunks. This returns 0 if the dentry was logged, and 1 otherwise.
2771 * If this returns 1, you must commit the transaction to safely get your
2772 * data on disk.
2773 */
2774int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans,
2775 struct btrfs_root *root, struct dentry *dentry)
2776{
2777 u64 gen;
2778 gen = root->fs_info->last_trans_new_blockgroup;
2779 if (gen > root->fs_info->last_trans_committed)
2780 return 1;
2781 else
2782 return btrfs_log_dentry(trans, root, dentry);
2783}
2784
2785/*
2786 * should be called during mount to recover any replay any log trees
2787 * from the FS
2788 */
2789int btrfs_recover_log_trees(struct btrfs_root *log_root_tree)
2790{
2791 int ret;
2792 struct btrfs_path *path;
2793 struct btrfs_trans_handle *trans;
2794 struct btrfs_key key;
2795 struct btrfs_key found_key;
2796 struct btrfs_key tmp_key;
2797 struct btrfs_root *log;
2798 struct btrfs_fs_info *fs_info = log_root_tree->fs_info;
Chris Mason8d5bf1c2008-09-11 15:51:21 -04002799 u64 highest_inode;
Chris Masone02119d2008-09-05 16:13:11 -04002800 struct walk_control wc = {
2801 .process_func = process_one_buffer,
2802 .stage = 0,
2803 };
2804
2805 fs_info->log_root_recovering = 1;
2806 path = btrfs_alloc_path();
2807 BUG_ON(!path);
2808
2809 trans = btrfs_start_transaction(fs_info->tree_root, 1);
2810
2811 wc.trans = trans;
2812 wc.pin = 1;
2813
2814 walk_log_tree(trans, log_root_tree, &wc);
2815
2816again:
2817 key.objectid = BTRFS_TREE_LOG_OBJECTID;
2818 key.offset = (u64)-1;
2819 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2820
Chris Masond3977122009-01-05 21:25:51 -05002821 while (1) {
Chris Masone02119d2008-09-05 16:13:11 -04002822 ret = btrfs_search_slot(NULL, log_root_tree, &key, path, 0, 0);
2823 if (ret < 0)
2824 break;
2825 if (ret > 0) {
2826 if (path->slots[0] == 0)
2827 break;
2828 path->slots[0]--;
2829 }
2830 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
2831 path->slots[0]);
2832 btrfs_release_path(log_root_tree, path);
2833 if (found_key.objectid != BTRFS_TREE_LOG_OBJECTID)
2834 break;
2835
2836 log = btrfs_read_fs_root_no_radix(log_root_tree,
2837 &found_key);
2838 BUG_ON(!log);
2839
2840
2841 tmp_key.objectid = found_key.offset;
2842 tmp_key.type = BTRFS_ROOT_ITEM_KEY;
2843 tmp_key.offset = (u64)-1;
2844
2845 wc.replay_dest = btrfs_read_fs_root_no_name(fs_info, &tmp_key);
Chris Masone02119d2008-09-05 16:13:11 -04002846 BUG_ON(!wc.replay_dest);
2847
Yan Zheng07d400a2009-01-06 11:42:00 -05002848 wc.replay_dest->log_root = log;
Chris Masone02119d2008-09-05 16:13:11 -04002849 btrfs_record_root_in_trans(wc.replay_dest);
2850 ret = walk_log_tree(trans, log, &wc);
2851 BUG_ON(ret);
2852
2853 if (wc.stage == LOG_WALK_REPLAY_ALL) {
2854 ret = fixup_inode_link_counts(trans, wc.replay_dest,
2855 path);
2856 BUG_ON(ret);
2857 }
Chris Mason8d5bf1c2008-09-11 15:51:21 -04002858 ret = btrfs_find_highest_inode(wc.replay_dest, &highest_inode);
2859 if (ret == 0) {
2860 wc.replay_dest->highest_inode = highest_inode;
2861 wc.replay_dest->last_inode_alloc = highest_inode;
2862 }
Chris Masone02119d2008-09-05 16:13:11 -04002863
2864 key.offset = found_key.offset - 1;
Yan Zheng07d400a2009-01-06 11:42:00 -05002865 wc.replay_dest->log_root = NULL;
Chris Masone02119d2008-09-05 16:13:11 -04002866 free_extent_buffer(log->node);
2867 kfree(log);
2868
2869 if (found_key.offset == 0)
2870 break;
2871 }
2872 btrfs_release_path(log_root_tree, path);
2873
2874 /* step one is to pin it all, step two is to replay just inodes */
2875 if (wc.pin) {
2876 wc.pin = 0;
2877 wc.process_func = replay_one_buffer;
2878 wc.stage = LOG_WALK_REPLAY_INODES;
2879 goto again;
2880 }
2881 /* step three is to replay everything */
2882 if (wc.stage < LOG_WALK_REPLAY_ALL) {
2883 wc.stage++;
2884 goto again;
2885 }
2886
2887 btrfs_free_path(path);
2888
2889 free_extent_buffer(log_root_tree->node);
2890 log_root_tree->log_root = NULL;
2891 fs_info->log_root_recovering = 0;
2892
2893 /* step 4: commit the transaction, which also unpins the blocks */
2894 btrfs_commit_transaction(trans, fs_info->tree_root);
2895
2896 kfree(log_root_tree);
2897 return 0;
2898}