Btrfs: Fix verify_parent_transid
[linux-2.6-block.git] / fs / btrfs / transaction.c
CommitLineData
6cbd5570
CM
1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
79154b1b 19#include <linux/fs.h>
34088780 20#include <linux/sched.h>
d3c2fdcf 21#include <linux/writeback.h>
5f39d397 22#include <linux/pagemap.h>
79154b1b
CM
23#include "ctree.h"
24#include "disk-io.h"
25#include "transaction.h"
925baedd 26#include "locking.h"
31153d81 27#include "ref-cache.h"
79154b1b 28
78fae27e 29static int total_trans = 0;
2c90e5d6
CM
30extern struct kmem_cache *btrfs_trans_handle_cachep;
31extern struct kmem_cache *btrfs_transaction_cachep;
32
0f7d52f4
CM
33#define BTRFS_ROOT_TRANS_TAG 0
34
80b6794d 35static noinline void put_transaction(struct btrfs_transaction *transaction)
79154b1b 36{
2c90e5d6 37 WARN_ON(transaction->use_count == 0);
79154b1b 38 transaction->use_count--;
78fae27e
CM
39 if (transaction->use_count == 0) {
40 WARN_ON(total_trans == 0);
41 total_trans--;
8fd17795 42 list_del_init(&transaction->list);
2c90e5d6
CM
43 memset(transaction, 0, sizeof(*transaction));
44 kmem_cache_free(btrfs_transaction_cachep, transaction);
78fae27e 45 }
79154b1b
CM
46}
47
80b6794d 48static noinline int join_transaction(struct btrfs_root *root)
79154b1b
CM
49{
50 struct btrfs_transaction *cur_trans;
51 cur_trans = root->fs_info->running_transaction;
52 if (!cur_trans) {
2c90e5d6
CM
53 cur_trans = kmem_cache_alloc(btrfs_transaction_cachep,
54 GFP_NOFS);
78fae27e 55 total_trans++;
79154b1b 56 BUG_ON(!cur_trans);
0f7d52f4 57 root->fs_info->generation++;
e18e4809 58 root->fs_info->last_alloc = 0;
4529ba49 59 root->fs_info->last_data_alloc = 0;
15ee9bc7
JB
60 cur_trans->num_writers = 1;
61 cur_trans->num_joined = 0;
0f7d52f4 62 cur_trans->transid = root->fs_info->generation;
79154b1b
CM
63 init_waitqueue_head(&cur_trans->writer_wait);
64 init_waitqueue_head(&cur_trans->commit_wait);
65 cur_trans->in_commit = 0;
f9295749 66 cur_trans->blocked = 0;
d5719762 67 cur_trans->use_count = 1;
79154b1b 68 cur_trans->commit_done = 0;
08607c1b 69 cur_trans->start_time = get_seconds();
3063d29f 70 INIT_LIST_HEAD(&cur_trans->pending_snapshots);
8fd17795 71 list_add_tail(&cur_trans->list, &root->fs_info->trans_list);
d1310b2e 72 extent_io_tree_init(&cur_trans->dirty_pages,
5f39d397
CM
73 root->fs_info->btree_inode->i_mapping,
74 GFP_NOFS);
48ec2cf8
CM
75 spin_lock(&root->fs_info->new_trans_lock);
76 root->fs_info->running_transaction = cur_trans;
77 spin_unlock(&root->fs_info->new_trans_lock);
15ee9bc7
JB
78 } else {
79 cur_trans->num_writers++;
80 cur_trans->num_joined++;
79154b1b 81 }
15ee9bc7 82
79154b1b
CM
83 return 0;
84}
85
80b6794d 86static noinline int record_root_in_trans(struct btrfs_root *root)
6702ed49 87{
f321e491 88 struct btrfs_dirty_root *dirty;
6702ed49
CM
89 u64 running_trans_id = root->fs_info->running_transaction->transid;
90 if (root->ref_cows && root->last_trans < running_trans_id) {
91 WARN_ON(root == root->fs_info->extent_root);
92 if (root->root_item.refs != 0) {
93 radix_tree_tag_set(&root->fs_info->fs_roots_radix,
94 (unsigned long)root->root_key.objectid,
95 BTRFS_ROOT_TRANS_TAG);
31153d81
YZ
96
97 dirty = kmalloc(sizeof(*dirty), GFP_NOFS);
98 BUG_ON(!dirty);
99 dirty->root = kmalloc(sizeof(*dirty->root), GFP_NOFS);
100 BUG_ON(!dirty->root);
101
102 dirty->latest_root = root;
103 INIT_LIST_HEAD(&dirty->list);
31153d81 104
925baedd 105 root->commit_root = btrfs_root_node(root);
017e5369 106 root->dirty_root = dirty;
31153d81
YZ
107
108 memcpy(dirty->root, root, sizeof(*root));
017e5369
CM
109 dirty->root->ref_tree = &root->ref_tree_struct;
110
31153d81
YZ
111 spin_lock_init(&dirty->root->node_lock);
112 mutex_init(&dirty->root->objectid_mutex);
113 dirty->root->node = root->commit_root;
114 dirty->root->commit_root = NULL;
6702ed49
CM
115 } else {
116 WARN_ON(1);
117 }
118 root->last_trans = running_trans_id;
119 }
120 return 0;
121}
122
f9295749
CM
123struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
124 int num_blocks, int join)
79154b1b 125{
2c90e5d6
CM
126 struct btrfs_trans_handle *h =
127 kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
f9295749 128 struct btrfs_transaction *cur_trans;
79154b1b
CM
129 int ret;
130
131 mutex_lock(&root->fs_info->trans_mutex);
f9295749
CM
132 cur_trans = root->fs_info->running_transaction;
133 if (cur_trans && cur_trans->blocked && !join) {
134 DEFINE_WAIT(wait);
135 cur_trans->use_count++;
136 while(1) {
137 prepare_to_wait(&root->fs_info->transaction_wait, &wait,
138 TASK_UNINTERRUPTIBLE);
139 if (cur_trans->blocked) {
140 mutex_unlock(&root->fs_info->trans_mutex);
141 schedule();
142 mutex_lock(&root->fs_info->trans_mutex);
143 finish_wait(&root->fs_info->transaction_wait,
144 &wait);
145 } else {
146 finish_wait(&root->fs_info->transaction_wait,
147 &wait);
148 break;
149 }
150 }
151 put_transaction(cur_trans);
152 }
79154b1b
CM
153 ret = join_transaction(root);
154 BUG_ON(ret);
0f7d52f4 155
6702ed49
CM
156 record_root_in_trans(root);
157 h->transid = root->fs_info->running_transaction->transid;
79154b1b
CM
158 h->transaction = root->fs_info->running_transaction;
159 h->blocks_reserved = num_blocks;
160 h->blocks_used = 0;
31f3c99b 161 h->block_group = NULL;
26b8003f
CM
162 h->alloc_exclude_nr = 0;
163 h->alloc_exclude_start = 0;
79154b1b
CM
164 root->fs_info->running_transaction->use_count++;
165 mutex_unlock(&root->fs_info->trans_mutex);
166 return h;
167}
168
f9295749
CM
169struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
170 int num_blocks)
171{
172 return start_transaction(root, num_blocks, 0);
173}
174struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root,
175 int num_blocks)
176{
177 return start_transaction(root, num_blocks, 1);
178}
179
89ce8a63
CM
180static noinline int wait_for_commit(struct btrfs_root *root,
181 struct btrfs_transaction *commit)
182{
183 DEFINE_WAIT(wait);
184 mutex_lock(&root->fs_info->trans_mutex);
185 while(!commit->commit_done) {
186 prepare_to_wait(&commit->commit_wait, &wait,
187 TASK_UNINTERRUPTIBLE);
188 if (commit->commit_done)
189 break;
190 mutex_unlock(&root->fs_info->trans_mutex);
191 schedule();
192 mutex_lock(&root->fs_info->trans_mutex);
193 }
194 mutex_unlock(&root->fs_info->trans_mutex);
195 finish_wait(&commit->commit_wait, &wait);
196 return 0;
197}
198
ab78c84d
CM
199void btrfs_throttle(struct btrfs_root *root)
200{
201 struct btrfs_fs_info *info = root->fs_info;
202
203harder:
204 if (atomic_read(&info->throttles)) {
205 DEFINE_WAIT(wait);
206 int thr;
207 int harder_count = 0;
208 thr = atomic_read(&info->throttle_gen);
209
210 do {
211 prepare_to_wait(&info->transaction_throttle,
212 &wait, TASK_UNINTERRUPTIBLE);
213 if (!atomic_read(&info->throttles)) {
214 finish_wait(&info->transaction_throttle, &wait);
215 break;
216 }
217 schedule();
218 finish_wait(&info->transaction_throttle, &wait);
219 } while (thr == atomic_read(&info->throttle_gen));
220
221 if (harder_count < 5 &&
222 info->total_ref_cache_size > 5 * 1024 * 1024) {
223 harder_count++;
224 goto harder;
225 }
226
227 if (harder_count < 10 &&
228 info->total_ref_cache_size > 10 * 1024 * 1024) {
229 harder_count++;
230 goto harder;
231 }
232 }
233}
234
89ce8a63
CM
235static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
236 struct btrfs_root *root, int throttle)
79154b1b
CM
237{
238 struct btrfs_transaction *cur_trans;
ab78c84d 239 struct btrfs_fs_info *info = root->fs_info;
d6e4a428 240
ab78c84d
CM
241 mutex_lock(&info->trans_mutex);
242 cur_trans = info->running_transaction;
ccd467d6 243 WARN_ON(cur_trans != trans->transaction);
d5719762 244 WARN_ON(cur_trans->num_writers < 1);
ccd467d6 245 cur_trans->num_writers--;
89ce8a63 246
79154b1b
CM
247 if (waitqueue_active(&cur_trans->writer_wait))
248 wake_up(&cur_trans->writer_wait);
79154b1b 249 put_transaction(cur_trans);
ab78c84d 250 mutex_unlock(&info->trans_mutex);
d6025579 251 memset(trans, 0, sizeof(*trans));
2c90e5d6 252 kmem_cache_free(btrfs_trans_handle_cachep, trans);
ab78c84d
CM
253
254 if (throttle)
255 btrfs_throttle(root);
256
79154b1b
CM
257 return 0;
258}
259
89ce8a63
CM
260int btrfs_end_transaction(struct btrfs_trans_handle *trans,
261 struct btrfs_root *root)
262{
263 return __btrfs_end_transaction(trans, root, 0);
264}
265
266int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
267 struct btrfs_root *root)
268{
269 return __btrfs_end_transaction(trans, root, 1);
270}
271
79154b1b
CM
272
273int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
274 struct btrfs_root *root)
275{
7c4452b9 276 int ret;
7c4452b9
CM
277 int err;
278 int werr = 0;
d1310b2e 279 struct extent_io_tree *dirty_pages;
7c4452b9 280 struct page *page;
7c4452b9 281 struct inode *btree_inode = root->fs_info->btree_inode;
5f39d397
CM
282 u64 start;
283 u64 end;
284 unsigned long index;
7c4452b9
CM
285
286 if (!trans || !trans->transaction) {
287 return filemap_write_and_wait(btree_inode->i_mapping);
288 }
289 dirty_pages = &trans->transaction->dirty_pages;
290 while(1) {
5f39d397
CM
291 ret = find_first_extent_bit(dirty_pages, 0, &start, &end,
292 EXTENT_DIRTY);
293 if (ret)
7c4452b9 294 break;
5f39d397
CM
295 clear_extent_dirty(dirty_pages, start, end, GFP_NOFS);
296 while(start <= end) {
297 index = start >> PAGE_CACHE_SHIFT;
35ebb934 298 start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
5f39d397 299 page = find_lock_page(btree_inode->i_mapping, index);
7c4452b9
CM
300 if (!page)
301 continue;
6702ed49
CM
302 if (PageWriteback(page)) {
303 if (PageDirty(page))
304 wait_on_page_writeback(page);
305 else {
306 unlock_page(page);
307 page_cache_release(page);
308 continue;
309 }
310 }
7c4452b9
CM
311 err = write_one_page(page, 0);
312 if (err)
313 werr = err;
314 page_cache_release(page);
315 }
316 }
317 err = filemap_fdatawait(btree_inode->i_mapping);
318 if (err)
319 werr = err;
320 return werr;
79154b1b
CM
321}
322
0b86a832
CM
323static int update_cowonly_root(struct btrfs_trans_handle *trans,
324 struct btrfs_root *root)
79154b1b
CM
325{
326 int ret;
0b86a832
CM
327 u64 old_root_bytenr;
328 struct btrfs_root *tree_root = root->fs_info->tree_root;
79154b1b 329
0b86a832 330 btrfs_write_dirty_block_groups(trans, root);
79154b1b 331 while(1) {
0b86a832
CM
332 old_root_bytenr = btrfs_root_bytenr(&root->root_item);
333 if (old_root_bytenr == root->node->start)
79154b1b 334 break;
0b86a832
CM
335 btrfs_set_root_bytenr(&root->root_item,
336 root->node->start);
337 btrfs_set_root_level(&root->root_item,
338 btrfs_header_level(root->node));
79154b1b 339 ret = btrfs_update_root(trans, tree_root,
0b86a832
CM
340 &root->root_key,
341 &root->root_item);
79154b1b 342 BUG_ON(ret);
0b86a832
CM
343 btrfs_write_dirty_block_groups(trans, root);
344 }
345 return 0;
346}
347
348int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans,
349 struct btrfs_root *root)
350{
351 struct btrfs_fs_info *fs_info = root->fs_info;
352 struct list_head *next;
353
354 while(!list_empty(&fs_info->dirty_cowonly_roots)) {
355 next = fs_info->dirty_cowonly_roots.next;
356 list_del_init(next);
357 root = list_entry(next, struct btrfs_root, dirty_list);
358 update_cowonly_root(trans, root);
017e5369
CM
359 if (root->fs_info->closing)
360 btrfs_remove_leaf_refs(root);
79154b1b
CM
361 }
362 return 0;
363}
364
5ce14bbc
CM
365int btrfs_add_dead_root(struct btrfs_root *root,
366 struct btrfs_root *latest,
367 struct list_head *dead_list)
5eda7b5e 368{
f321e491 369 struct btrfs_dirty_root *dirty;
5eda7b5e
CM
370
371 dirty = kmalloc(sizeof(*dirty), GFP_NOFS);
372 if (!dirty)
373 return -ENOMEM;
5eda7b5e 374 dirty->root = root;
5ce14bbc 375 dirty->latest_root = latest;
5eda7b5e
CM
376 list_add(&dirty->list, dead_list);
377 return 0;
378}
379
80b6794d
CM
380static noinline int add_dirty_roots(struct btrfs_trans_handle *trans,
381 struct radix_tree_root *radix,
382 struct list_head *list)
0f7d52f4 383{
f321e491 384 struct btrfs_dirty_root *dirty;
0f7d52f4
CM
385 struct btrfs_root *gang[8];
386 struct btrfs_root *root;
387 int i;
388 int ret;
54aa1f4d 389 int err = 0;
5eda7b5e 390 u32 refs;
54aa1f4d 391
0f7d52f4
CM
392 while(1) {
393 ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0,
394 ARRAY_SIZE(gang),
395 BTRFS_ROOT_TRANS_TAG);
396 if (ret == 0)
397 break;
398 for (i = 0; i < ret; i++) {
399 root = gang[i];
2619ba1f
CM
400 radix_tree_tag_clear(radix,
401 (unsigned long)root->root_key.objectid,
402 BTRFS_ROOT_TRANS_TAG);
31153d81
YZ
403
404 BUG_ON(!root->ref_tree);
017e5369 405 dirty = root->dirty_root;
31153d81 406
0f7d52f4 407 if (root->commit_root == root->node) {
db94535d
CM
408 WARN_ON(root->node->start !=
409 btrfs_root_bytenr(&root->root_item));
31153d81 410
5f39d397 411 free_extent_buffer(root->commit_root);
0f7d52f4 412 root->commit_root = NULL;
31153d81
YZ
413
414 kfree(dirty->root);
415 kfree(dirty);
58176a96
JB
416
417 /* make sure to update the root on disk
418 * so we get any updates to the block used
419 * counts
420 */
421 err = btrfs_update_root(trans,
422 root->fs_info->tree_root,
423 &root->root_key,
424 &root->root_item);
0f7d52f4
CM
425 continue;
426 }
9f3a7427
CM
427
428 memset(&root->root_item.drop_progress, 0,
429 sizeof(struct btrfs_disk_key));
430 root->root_item.drop_level = 0;
0f7d52f4 431 root->commit_root = NULL;
0f7d52f4 432 root->root_key.offset = root->fs_info->generation;
db94535d
CM
433 btrfs_set_root_bytenr(&root->root_item,
434 root->node->start);
435 btrfs_set_root_level(&root->root_item,
436 btrfs_header_level(root->node));
0f7d52f4
CM
437 err = btrfs_insert_root(trans, root->fs_info->tree_root,
438 &root->root_key,
439 &root->root_item);
54aa1f4d
CM
440 if (err)
441 break;
9f3a7427
CM
442
443 refs = btrfs_root_refs(&dirty->root->root_item);
444 btrfs_set_root_refs(&dirty->root->root_item, refs - 1);
5eda7b5e 445 err = btrfs_update_root(trans, root->fs_info->tree_root,
9f3a7427
CM
446 &dirty->root->root_key,
447 &dirty->root->root_item);
5eda7b5e
CM
448
449 BUG_ON(err);
9f3a7427 450 if (refs == 1) {
5eda7b5e 451 list_add(&dirty->list, list);
9f3a7427
CM
452 } else {
453 WARN_ON(1);
31153d81 454 free_extent_buffer(dirty->root->node);
9f3a7427 455 kfree(dirty->root);
5eda7b5e 456 kfree(dirty);
9f3a7427 457 }
0f7d52f4
CM
458 }
459 }
54aa1f4d 460 return err;
0f7d52f4
CM
461}
462
e9d0b13b
CM
463int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
464{
465 struct btrfs_fs_info *info = root->fs_info;
466 int ret;
467 struct btrfs_trans_handle *trans;
d3c2fdcf 468 unsigned long nr;
e9d0b13b 469
a2135011 470 smp_mb();
e9d0b13b
CM
471 if (root->defrag_running)
472 return 0;
e9d0b13b 473 trans = btrfs_start_transaction(root, 1);
6b80053d 474 while (1) {
e9d0b13b
CM
475 root->defrag_running = 1;
476 ret = btrfs_defrag_leaves(trans, root, cacheonly);
d3c2fdcf 477 nr = trans->blocks_used;
e9d0b13b 478 btrfs_end_transaction(trans, root);
d3c2fdcf 479 btrfs_btree_balance_dirty(info->tree_root, nr);
e9d0b13b
CM
480 cond_resched();
481
e9d0b13b 482 trans = btrfs_start_transaction(root, 1);
3f157a2f 483 if (root->fs_info->closing || ret != -EAGAIN)
e9d0b13b
CM
484 break;
485 }
486 root->defrag_running = 0;
a2135011 487 smp_mb();
e9d0b13b
CM
488 btrfs_end_transaction(trans, root);
489 return 0;
490}
491
80b6794d
CM
492static noinline int drop_dirty_roots(struct btrfs_root *tree_root,
493 struct list_head *list)
0f7d52f4 494{
f321e491 495 struct btrfs_dirty_root *dirty;
0f7d52f4 496 struct btrfs_trans_handle *trans;
d3c2fdcf 497 unsigned long nr;
db94535d
CM
498 u64 num_bytes;
499 u64 bytes_used;
54aa1f4d 500 int ret = 0;
9f3a7427
CM
501 int err;
502
0f7d52f4 503 while(!list_empty(list)) {
58176a96
JB
504 struct btrfs_root *root;
505
f321e491 506 dirty = list_entry(list->prev, struct btrfs_dirty_root, list);
0f7d52f4 507 list_del_init(&dirty->list);
5eda7b5e 508
db94535d 509 num_bytes = btrfs_root_used(&dirty->root->root_item);
58176a96 510 root = dirty->latest_root;
a2135011 511 atomic_inc(&root->fs_info->throttles);
58176a96 512
a2135011 513 mutex_lock(&root->fs_info->drop_mutex);
9f3a7427
CM
514 while(1) {
515 trans = btrfs_start_transaction(tree_root, 1);
516 ret = btrfs_drop_snapshot(trans, dirty->root);
517 if (ret != -EAGAIN) {
518 break;
519 }
58176a96 520
9f3a7427
CM
521 err = btrfs_update_root(trans,
522 tree_root,
523 &dirty->root->root_key,
524 &dirty->root->root_item);
525 if (err)
526 ret = err;
d3c2fdcf 527 nr = trans->blocks_used;
017e5369 528 ret = btrfs_end_transaction(trans, tree_root);
9f3a7427 529 BUG_ON(ret);
a2135011
CM
530
531 mutex_unlock(&root->fs_info->drop_mutex);
d3c2fdcf 532 btrfs_btree_balance_dirty(tree_root, nr);
4dc11904 533 cond_resched();
a2135011 534 mutex_lock(&root->fs_info->drop_mutex);
9f3a7427 535 }
0f7d52f4 536 BUG_ON(ret);
a2135011 537 atomic_dec(&root->fs_info->throttles);
017e5369 538 wake_up(&root->fs_info->transaction_throttle);
58176a96 539
a2135011 540 mutex_lock(&root->fs_info->alloc_mutex);
db94535d
CM
541 num_bytes -= btrfs_root_used(&dirty->root->root_item);
542 bytes_used = btrfs_root_used(&root->root_item);
543 if (num_bytes) {
58176a96 544 record_root_in_trans(root);
5f39d397 545 btrfs_set_root_used(&root->root_item,
db94535d 546 bytes_used - num_bytes);
58176a96 547 }
a2135011
CM
548 mutex_unlock(&root->fs_info->alloc_mutex);
549
9f3a7427 550 ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key);
58176a96
JB
551 if (ret) {
552 BUG();
54aa1f4d 553 break;
58176a96 554 }
a2135011
CM
555 mutex_unlock(&root->fs_info->drop_mutex);
556
d3c2fdcf 557 nr = trans->blocks_used;
0f7d52f4
CM
558 ret = btrfs_end_transaction(trans, tree_root);
559 BUG_ON(ret);
5eda7b5e 560
f510cfec 561 free_extent_buffer(dirty->root->node);
9f3a7427 562 kfree(dirty->root);
0f7d52f4 563 kfree(dirty);
d3c2fdcf
CM
564
565 btrfs_btree_balance_dirty(tree_root, nr);
4dc11904 566 cond_resched();
0f7d52f4 567 }
54aa1f4d 568 return ret;
0f7d52f4
CM
569}
570
80b6794d 571static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
3063d29f
CM
572 struct btrfs_fs_info *fs_info,
573 struct btrfs_pending_snapshot *pending)
574{
575 struct btrfs_key key;
80b6794d 576 struct btrfs_root_item *new_root_item;
3063d29f
CM
577 struct btrfs_root *tree_root = fs_info->tree_root;
578 struct btrfs_root *root = pending->root;
579 struct extent_buffer *tmp;
925baedd 580 struct extent_buffer *old;
3063d29f 581 int ret;
3b96362c 582 int namelen;
3063d29f
CM
583 u64 objectid;
584
80b6794d
CM
585 new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS);
586 if (!new_root_item) {
587 ret = -ENOMEM;
588 goto fail;
589 }
3063d29f
CM
590 ret = btrfs_find_free_objectid(trans, tree_root, 0, &objectid);
591 if (ret)
592 goto fail;
593
80b6794d 594 memcpy(new_root_item, &root->root_item, sizeof(*new_root_item));
3063d29f
CM
595
596 key.objectid = objectid;
597 key.offset = 1;
598 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
599
925baedd
CM
600 old = btrfs_lock_root_node(root);
601 btrfs_cow_block(trans, root, old, NULL, 0, &old);
3063d29f 602
925baedd
CM
603 btrfs_copy_root(trans, root, old, &tmp, objectid);
604 btrfs_tree_unlock(old);
605 free_extent_buffer(old);
3063d29f 606
80b6794d
CM
607 btrfs_set_root_bytenr(new_root_item, tmp->start);
608 btrfs_set_root_level(new_root_item, btrfs_header_level(tmp));
3063d29f 609 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
80b6794d 610 new_root_item);
925baedd 611 btrfs_tree_unlock(tmp);
3063d29f
CM
612 free_extent_buffer(tmp);
613 if (ret)
614 goto fail;
615
616 /*
617 * insert the directory item
618 */
619 key.offset = (u64)-1;
3b96362c 620 namelen = strlen(pending->name);
3063d29f 621 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
3b96362c 622 pending->name, namelen,
3063d29f 623 root->fs_info->sb->s_root->d_inode->i_ino,
aec7477b 624 &key, BTRFS_FT_DIR, 0);
3063d29f
CM
625
626 if (ret)
627 goto fail;
628
629 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
630 pending->name, strlen(pending->name), objectid,
aec7477b 631 root->fs_info->sb->s_root->d_inode->i_ino, 0);
3b96362c
SW
632
633 /* Invalidate existing dcache entry for new snapshot. */
634 btrfs_invalidate_dcache_root(root, pending->name, namelen);
635
3063d29f 636fail:
80b6794d 637 kfree(new_root_item);
3063d29f
CM
638 return ret;
639}
640
80b6794d
CM
641static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans,
642 struct btrfs_fs_info *fs_info)
3063d29f
CM
643{
644 struct btrfs_pending_snapshot *pending;
645 struct list_head *head = &trans->transaction->pending_snapshots;
646 int ret;
647
648 while(!list_empty(head)) {
649 pending = list_entry(head->next,
650 struct btrfs_pending_snapshot, list);
651 ret = create_pending_snapshot(trans, fs_info, pending);
652 BUG_ON(ret);
653 list_del(&pending->list);
654 kfree(pending->name);
655 kfree(pending);
656 }
dc17ff8f
CM
657 return 0;
658}
659
79154b1b
CM
660int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
661 struct btrfs_root *root)
662{
15ee9bc7
JB
663 unsigned long joined = 0;
664 unsigned long timeout = 1;
79154b1b 665 struct btrfs_transaction *cur_trans;
8fd17795 666 struct btrfs_transaction *prev_trans = NULL;
0b86a832 667 struct btrfs_root *chunk_root = root->fs_info->chunk_root;
0f7d52f4 668 struct list_head dirty_fs_roots;
d1310b2e 669 struct extent_io_tree *pinned_copy;
79154b1b 670 DEFINE_WAIT(wait);
15ee9bc7 671 int ret;
79154b1b 672
0f7d52f4 673 INIT_LIST_HEAD(&dirty_fs_roots);
d6e4a428 674
79154b1b
CM
675 mutex_lock(&root->fs_info->trans_mutex);
676 if (trans->transaction->in_commit) {
677 cur_trans = trans->transaction;
678 trans->transaction->use_count++;
ccd467d6 679 mutex_unlock(&root->fs_info->trans_mutex);
79154b1b 680 btrfs_end_transaction(trans, root);
ccd467d6 681
79154b1b
CM
682 ret = wait_for_commit(root, cur_trans);
683 BUG_ON(ret);
15ee9bc7
JB
684
685 mutex_lock(&root->fs_info->trans_mutex);
79154b1b 686 put_transaction(cur_trans);
15ee9bc7
JB
687 mutex_unlock(&root->fs_info->trans_mutex);
688
79154b1b
CM
689 return 0;
690 }
4313b399
CM
691
692 pinned_copy = kmalloc(sizeof(*pinned_copy), GFP_NOFS);
693 if (!pinned_copy)
694 return -ENOMEM;
695
d1310b2e 696 extent_io_tree_init(pinned_copy,
4313b399
CM
697 root->fs_info->btree_inode->i_mapping, GFP_NOFS);
698
2c90e5d6 699 trans->transaction->in_commit = 1;
f9295749 700 trans->transaction->blocked = 1;
ccd467d6
CM
701 cur_trans = trans->transaction;
702 if (cur_trans->list.prev != &root->fs_info->trans_list) {
703 prev_trans = list_entry(cur_trans->list.prev,
704 struct btrfs_transaction, list);
705 if (!prev_trans->commit_done) {
706 prev_trans->use_count++;
ccd467d6
CM
707 mutex_unlock(&root->fs_info->trans_mutex);
708
709 wait_for_commit(root, prev_trans);
ccd467d6 710
ccd467d6 711 mutex_lock(&root->fs_info->trans_mutex);
15ee9bc7 712 put_transaction(prev_trans);
ccd467d6
CM
713 }
714 }
15ee9bc7
JB
715
716 do {
717 joined = cur_trans->num_joined;
2c90e5d6 718 WARN_ON(cur_trans != trans->transaction);
15ee9bc7 719 prepare_to_wait(&cur_trans->writer_wait, &wait,
79154b1b 720 TASK_UNINTERRUPTIBLE);
15ee9bc7
JB
721
722 if (cur_trans->num_writers > 1)
723 timeout = MAX_SCHEDULE_TIMEOUT;
724 else
725 timeout = 1;
726
79154b1b 727 mutex_unlock(&root->fs_info->trans_mutex);
15ee9bc7
JB
728
729 schedule_timeout(timeout);
730
79154b1b 731 mutex_lock(&root->fs_info->trans_mutex);
15ee9bc7
JB
732 finish_wait(&cur_trans->writer_wait, &wait);
733 } while (cur_trans->num_writers > 1 ||
734 (cur_trans->num_joined != joined));
735
3063d29f
CM
736 ret = create_pending_snapshots(trans, root->fs_info);
737 BUG_ON(ret);
738
2c90e5d6 739 WARN_ON(cur_trans != trans->transaction);
dc17ff8f 740
54aa1f4d
CM
741 ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix,
742 &dirty_fs_roots);
743 BUG_ON(ret);
744
79154b1b
CM
745 ret = btrfs_commit_tree_roots(trans, root);
746 BUG_ON(ret);
54aa1f4d 747
78fae27e 748 cur_trans = root->fs_info->running_transaction;
cee36a03 749 spin_lock(&root->fs_info->new_trans_lock);
78fae27e 750 root->fs_info->running_transaction = NULL;
cee36a03 751 spin_unlock(&root->fs_info->new_trans_lock);
4b52dff6
CM
752 btrfs_set_super_generation(&root->fs_info->super_copy,
753 cur_trans->transid);
754 btrfs_set_super_root(&root->fs_info->super_copy,
db94535d
CM
755 root->fs_info->tree_root->node->start);
756 btrfs_set_super_root_level(&root->fs_info->super_copy,
757 btrfs_header_level(root->fs_info->tree_root->node));
5f39d397 758
0b86a832
CM
759 btrfs_set_super_chunk_root(&root->fs_info->super_copy,
760 chunk_root->node->start);
761 btrfs_set_super_chunk_root_level(&root->fs_info->super_copy,
762 btrfs_header_level(chunk_root->node));
a061fc8d
CM
763 memcpy(&root->fs_info->super_for_commit, &root->fs_info->super_copy,
764 sizeof(root->fs_info->super_copy));
ccd467d6 765
4313b399 766 btrfs_copy_pinned(root, pinned_copy);
ccd467d6 767
f9295749 768 trans->transaction->blocked = 0;
e6dcd2dc 769 wake_up(&root->fs_info->transaction_throttle);
f9295749 770 wake_up(&root->fs_info->transaction_wait);
e6dcd2dc 771
78fae27e 772 mutex_unlock(&root->fs_info->trans_mutex);
79154b1b
CM
773 ret = btrfs_write_and_wait_transaction(trans, root);
774 BUG_ON(ret);
79154b1b 775 write_ctree_super(trans, root);
4313b399 776
4313b399 777 btrfs_finish_extent_commit(trans, root, pinned_copy);
78fae27e 778 mutex_lock(&root->fs_info->trans_mutex);
4313b399
CM
779
780 kfree(pinned_copy);
781
2c90e5d6 782 cur_trans->commit_done = 1;
15ee9bc7 783 root->fs_info->last_trans_committed = cur_trans->transid;
2c90e5d6 784 wake_up(&cur_trans->commit_wait);
78fae27e 785 put_transaction(cur_trans);
79154b1b 786 put_transaction(cur_trans);
58176a96 787
facda1e7
CM
788 if (root->fs_info->closing)
789 list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots);
790 else
791 list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots);
58176a96 792
78fae27e 793 mutex_unlock(&root->fs_info->trans_mutex);
2c90e5d6 794 kmem_cache_free(btrfs_trans_handle_cachep, trans);
79154b1b 795
facda1e7 796 if (root->fs_info->closing) {
facda1e7 797 drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots);
facda1e7 798 }
79154b1b
CM
799 return ret;
800}
801
e9d0b13b
CM
802int btrfs_clean_old_snapshots(struct btrfs_root *root)
803{
804 struct list_head dirty_roots;
805 INIT_LIST_HEAD(&dirty_roots);
a74a4b97 806again:
e9d0b13b
CM
807 mutex_lock(&root->fs_info->trans_mutex);
808 list_splice_init(&root->fs_info->dead_roots, &dirty_roots);
809 mutex_unlock(&root->fs_info->trans_mutex);
810
811 if (!list_empty(&dirty_roots)) {
812 drop_dirty_roots(root, &dirty_roots);
a74a4b97 813 goto again;
e9d0b13b
CM
814 }
815 return 0;
816}
08607c1b 817