Commit | Line | Data |
---|---|---|
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" | |
26 | ||
78fae27e | 27 | static int total_trans = 0; |
2c90e5d6 CM |
28 | extern struct kmem_cache *btrfs_trans_handle_cachep; |
29 | extern struct kmem_cache *btrfs_transaction_cachep; | |
30 | ||
08607c1b CM |
31 | static struct workqueue_struct *trans_wq; |
32 | ||
0f7d52f4 | 33 | #define BTRFS_ROOT_TRANS_TAG 0 |
6702ed49 | 34 | #define BTRFS_ROOT_DEFRAG_TAG 1 |
0f7d52f4 | 35 | |
79154b1b CM |
36 | static void put_transaction(struct btrfs_transaction *transaction) |
37 | { | |
2c90e5d6 | 38 | WARN_ON(transaction->use_count == 0); |
79154b1b | 39 | transaction->use_count--; |
78fae27e CM |
40 | if (transaction->use_count == 0) { |
41 | WARN_ON(total_trans == 0); | |
42 | total_trans--; | |
8fd17795 | 43 | list_del_init(&transaction->list); |
2c90e5d6 CM |
44 | memset(transaction, 0, sizeof(*transaction)); |
45 | kmem_cache_free(btrfs_transaction_cachep, transaction); | |
78fae27e | 46 | } |
79154b1b CM |
47 | } |
48 | ||
49 | static int join_transaction(struct btrfs_root *root) | |
50 | { | |
51 | struct btrfs_transaction *cur_trans; | |
52 | cur_trans = root->fs_info->running_transaction; | |
53 | if (!cur_trans) { | |
2c90e5d6 CM |
54 | cur_trans = kmem_cache_alloc(btrfs_transaction_cachep, |
55 | GFP_NOFS); | |
78fae27e | 56 | total_trans++; |
79154b1b | 57 | BUG_ON(!cur_trans); |
0f7d52f4 | 58 | root->fs_info->generation++; |
79154b1b | 59 | root->fs_info->running_transaction = cur_trans; |
e18e4809 | 60 | root->fs_info->last_alloc = 0; |
15ee9bc7 JB |
61 | cur_trans->num_writers = 1; |
62 | cur_trans->num_joined = 0; | |
0f7d52f4 | 63 | cur_trans->transid = root->fs_info->generation; |
79154b1b CM |
64 | init_waitqueue_head(&cur_trans->writer_wait); |
65 | init_waitqueue_head(&cur_trans->commit_wait); | |
66 | cur_trans->in_commit = 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); |
dc17ff8f | 72 | btrfs_ordered_inode_tree_init(&cur_trans->ordered_inode_tree); |
5f39d397 CM |
73 | extent_map_tree_init(&cur_trans->dirty_pages, |
74 | root->fs_info->btree_inode->i_mapping, | |
75 | GFP_NOFS); | |
15ee9bc7 JB |
76 | } else { |
77 | cur_trans->num_writers++; | |
78 | cur_trans->num_joined++; | |
79154b1b | 79 | } |
15ee9bc7 | 80 | |
79154b1b CM |
81 | return 0; |
82 | } | |
83 | ||
6702ed49 CM |
84 | static int record_root_in_trans(struct btrfs_root *root) |
85 | { | |
86 | u64 running_trans_id = root->fs_info->running_transaction->transid; | |
87 | if (root->ref_cows && root->last_trans < running_trans_id) { | |
88 | WARN_ON(root == root->fs_info->extent_root); | |
89 | if (root->root_item.refs != 0) { | |
90 | radix_tree_tag_set(&root->fs_info->fs_roots_radix, | |
91 | (unsigned long)root->root_key.objectid, | |
92 | BTRFS_ROOT_TRANS_TAG); | |
93 | radix_tree_tag_set(&root->fs_info->fs_roots_radix, | |
94 | (unsigned long)root->root_key.objectid, | |
95 | BTRFS_ROOT_DEFRAG_TAG); | |
96 | root->commit_root = root->node; | |
5f39d397 | 97 | extent_buffer_get(root->node); |
6702ed49 CM |
98 | } else { |
99 | WARN_ON(1); | |
100 | } | |
101 | root->last_trans = running_trans_id; | |
102 | } | |
103 | return 0; | |
104 | } | |
105 | ||
79154b1b CM |
106 | struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, |
107 | int num_blocks) | |
108 | { | |
2c90e5d6 CM |
109 | struct btrfs_trans_handle *h = |
110 | kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS); | |
79154b1b CM |
111 | int ret; |
112 | ||
113 | mutex_lock(&root->fs_info->trans_mutex); | |
114 | ret = join_transaction(root); | |
115 | BUG_ON(ret); | |
0f7d52f4 | 116 | |
6702ed49 CM |
117 | record_root_in_trans(root); |
118 | h->transid = root->fs_info->running_transaction->transid; | |
79154b1b CM |
119 | h->transaction = root->fs_info->running_transaction; |
120 | h->blocks_reserved = num_blocks; | |
121 | h->blocks_used = 0; | |
31f3c99b | 122 | h->block_group = NULL; |
26b8003f CM |
123 | h->alloc_exclude_nr = 0; |
124 | h->alloc_exclude_start = 0; | |
79154b1b CM |
125 | root->fs_info->running_transaction->use_count++; |
126 | mutex_unlock(&root->fs_info->trans_mutex); | |
127 | return h; | |
128 | } | |
129 | ||
130 | int btrfs_end_transaction(struct btrfs_trans_handle *trans, | |
131 | struct btrfs_root *root) | |
132 | { | |
133 | struct btrfs_transaction *cur_trans; | |
d6e4a428 | 134 | |
79154b1b CM |
135 | mutex_lock(&root->fs_info->trans_mutex); |
136 | cur_trans = root->fs_info->running_transaction; | |
ccd467d6 | 137 | WARN_ON(cur_trans != trans->transaction); |
d5719762 | 138 | WARN_ON(cur_trans->num_writers < 1); |
ccd467d6 | 139 | cur_trans->num_writers--; |
79154b1b CM |
140 | if (waitqueue_active(&cur_trans->writer_wait)) |
141 | wake_up(&cur_trans->writer_wait); | |
79154b1b CM |
142 | put_transaction(cur_trans); |
143 | mutex_unlock(&root->fs_info->trans_mutex); | |
d6025579 | 144 | memset(trans, 0, sizeof(*trans)); |
2c90e5d6 | 145 | kmem_cache_free(btrfs_trans_handle_cachep, trans); |
79154b1b CM |
146 | return 0; |
147 | } | |
148 | ||
149 | ||
150 | int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, | |
151 | struct btrfs_root *root) | |
152 | { | |
7c4452b9 | 153 | int ret; |
7c4452b9 CM |
154 | int err; |
155 | int werr = 0; | |
5f39d397 | 156 | struct extent_map_tree *dirty_pages; |
7c4452b9 | 157 | struct page *page; |
7c4452b9 | 158 | struct inode *btree_inode = root->fs_info->btree_inode; |
5f39d397 CM |
159 | u64 start; |
160 | u64 end; | |
161 | unsigned long index; | |
7c4452b9 CM |
162 | |
163 | if (!trans || !trans->transaction) { | |
164 | return filemap_write_and_wait(btree_inode->i_mapping); | |
165 | } | |
166 | dirty_pages = &trans->transaction->dirty_pages; | |
167 | while(1) { | |
5f39d397 CM |
168 | ret = find_first_extent_bit(dirty_pages, 0, &start, &end, |
169 | EXTENT_DIRTY); | |
170 | if (ret) | |
7c4452b9 | 171 | break; |
5f39d397 CM |
172 | clear_extent_dirty(dirty_pages, start, end, GFP_NOFS); |
173 | while(start <= end) { | |
174 | index = start >> PAGE_CACHE_SHIFT; | |
35ebb934 | 175 | start = (u64)(index + 1) << PAGE_CACHE_SHIFT; |
5f39d397 | 176 | page = find_lock_page(btree_inode->i_mapping, index); |
7c4452b9 CM |
177 | if (!page) |
178 | continue; | |
6702ed49 CM |
179 | if (PageWriteback(page)) { |
180 | if (PageDirty(page)) | |
181 | wait_on_page_writeback(page); | |
182 | else { | |
183 | unlock_page(page); | |
184 | page_cache_release(page); | |
185 | continue; | |
186 | } | |
187 | } | |
7c4452b9 CM |
188 | err = write_one_page(page, 0); |
189 | if (err) | |
190 | werr = err; | |
191 | page_cache_release(page); | |
192 | } | |
193 | } | |
194 | err = filemap_fdatawait(btree_inode->i_mapping); | |
195 | if (err) | |
196 | werr = err; | |
197 | return werr; | |
79154b1b CM |
198 | } |
199 | ||
200 | int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, | |
201 | struct btrfs_root *root) | |
202 | { | |
203 | int ret; | |
204 | u64 old_extent_block; | |
205 | struct btrfs_fs_info *fs_info = root->fs_info; | |
206 | struct btrfs_root *tree_root = fs_info->tree_root; | |
207 | struct btrfs_root *extent_root = fs_info->extent_root; | |
79154b1b | 208 | |
9078a3e1 | 209 | btrfs_write_dirty_block_groups(trans, extent_root); |
79154b1b | 210 | while(1) { |
db94535d CM |
211 | old_extent_block = btrfs_root_bytenr(&extent_root->root_item); |
212 | if (old_extent_block == extent_root->node->start) | |
79154b1b | 213 | break; |
db94535d CM |
214 | btrfs_set_root_bytenr(&extent_root->root_item, |
215 | extent_root->node->start); | |
216 | btrfs_set_root_level(&extent_root->root_item, | |
217 | btrfs_header_level(extent_root->node)); | |
79154b1b CM |
218 | ret = btrfs_update_root(trans, tree_root, |
219 | &extent_root->root_key, | |
220 | &extent_root->root_item); | |
221 | BUG_ON(ret); | |
9078a3e1 | 222 | btrfs_write_dirty_block_groups(trans, extent_root); |
79154b1b CM |
223 | } |
224 | return 0; | |
225 | } | |
226 | ||
227 | static int wait_for_commit(struct btrfs_root *root, | |
228 | struct btrfs_transaction *commit) | |
229 | { | |
230 | DEFINE_WAIT(wait); | |
ccd467d6 | 231 | mutex_lock(&root->fs_info->trans_mutex); |
79154b1b CM |
232 | while(!commit->commit_done) { |
233 | prepare_to_wait(&commit->commit_wait, &wait, | |
234 | TASK_UNINTERRUPTIBLE); | |
235 | if (commit->commit_done) | |
236 | break; | |
237 | mutex_unlock(&root->fs_info->trans_mutex); | |
238 | schedule(); | |
239 | mutex_lock(&root->fs_info->trans_mutex); | |
240 | } | |
ccd467d6 | 241 | mutex_unlock(&root->fs_info->trans_mutex); |
79154b1b CM |
242 | finish_wait(&commit->commit_wait, &wait); |
243 | return 0; | |
244 | } | |
245 | ||
0f7d52f4 CM |
246 | struct dirty_root { |
247 | struct list_head list; | |
0f7d52f4 | 248 | struct btrfs_root *root; |
58176a96 | 249 | struct btrfs_root *latest_root; |
0f7d52f4 CM |
250 | }; |
251 | ||
5ce14bbc CM |
252 | int btrfs_add_dead_root(struct btrfs_root *root, |
253 | struct btrfs_root *latest, | |
254 | struct list_head *dead_list) | |
5eda7b5e CM |
255 | { |
256 | struct dirty_root *dirty; | |
257 | ||
258 | dirty = kmalloc(sizeof(*dirty), GFP_NOFS); | |
259 | if (!dirty) | |
260 | return -ENOMEM; | |
5eda7b5e | 261 | dirty->root = root; |
5ce14bbc | 262 | dirty->latest_root = latest; |
5eda7b5e CM |
263 | list_add(&dirty->list, dead_list); |
264 | return 0; | |
265 | } | |
266 | ||
35b7e476 CM |
267 | static int add_dirty_roots(struct btrfs_trans_handle *trans, |
268 | struct radix_tree_root *radix, | |
269 | struct list_head *list) | |
0f7d52f4 CM |
270 | { |
271 | struct dirty_root *dirty; | |
272 | struct btrfs_root *gang[8]; | |
273 | struct btrfs_root *root; | |
274 | int i; | |
275 | int ret; | |
54aa1f4d | 276 | int err = 0; |
5eda7b5e | 277 | u32 refs; |
54aa1f4d | 278 | |
0f7d52f4 CM |
279 | while(1) { |
280 | ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0, | |
281 | ARRAY_SIZE(gang), | |
282 | BTRFS_ROOT_TRANS_TAG); | |
283 | if (ret == 0) | |
284 | break; | |
285 | for (i = 0; i < ret; i++) { | |
286 | root = gang[i]; | |
2619ba1f CM |
287 | radix_tree_tag_clear(radix, |
288 | (unsigned long)root->root_key.objectid, | |
289 | BTRFS_ROOT_TRANS_TAG); | |
0f7d52f4 | 290 | if (root->commit_root == root->node) { |
db94535d CM |
291 | WARN_ON(root->node->start != |
292 | btrfs_root_bytenr(&root->root_item)); | |
5f39d397 | 293 | free_extent_buffer(root->commit_root); |
0f7d52f4 | 294 | root->commit_root = NULL; |
58176a96 JB |
295 | |
296 | /* make sure to update the root on disk | |
297 | * so we get any updates to the block used | |
298 | * counts | |
299 | */ | |
300 | err = btrfs_update_root(trans, | |
301 | root->fs_info->tree_root, | |
302 | &root->root_key, | |
303 | &root->root_item); | |
0f7d52f4 CM |
304 | continue; |
305 | } | |
306 | dirty = kmalloc(sizeof(*dirty), GFP_NOFS); | |
307 | BUG_ON(!dirty); | |
9f3a7427 CM |
308 | dirty->root = kmalloc(sizeof(*dirty->root), GFP_NOFS); |
309 | BUG_ON(!dirty->root); | |
310 | ||
311 | memset(&root->root_item.drop_progress, 0, | |
312 | sizeof(struct btrfs_disk_key)); | |
313 | root->root_item.drop_level = 0; | |
314 | ||
315 | memcpy(dirty->root, root, sizeof(*root)); | |
316 | dirty->root->node = root->commit_root; | |
58176a96 | 317 | dirty->latest_root = root; |
0f7d52f4 | 318 | root->commit_root = NULL; |
5eda7b5e | 319 | |
0f7d52f4 | 320 | root->root_key.offset = root->fs_info->generation; |
db94535d CM |
321 | btrfs_set_root_bytenr(&root->root_item, |
322 | root->node->start); | |
323 | btrfs_set_root_level(&root->root_item, | |
324 | btrfs_header_level(root->node)); | |
0f7d52f4 CM |
325 | err = btrfs_insert_root(trans, root->fs_info->tree_root, |
326 | &root->root_key, | |
327 | &root->root_item); | |
54aa1f4d CM |
328 | if (err) |
329 | break; | |
9f3a7427 CM |
330 | |
331 | refs = btrfs_root_refs(&dirty->root->root_item); | |
332 | btrfs_set_root_refs(&dirty->root->root_item, refs - 1); | |
5eda7b5e | 333 | err = btrfs_update_root(trans, root->fs_info->tree_root, |
9f3a7427 CM |
334 | &dirty->root->root_key, |
335 | &dirty->root->root_item); | |
5eda7b5e CM |
336 | |
337 | BUG_ON(err); | |
9f3a7427 | 338 | if (refs == 1) { |
5eda7b5e | 339 | list_add(&dirty->list, list); |
9f3a7427 CM |
340 | } else { |
341 | WARN_ON(1); | |
342 | kfree(dirty->root); | |
5eda7b5e | 343 | kfree(dirty); |
9f3a7427 | 344 | } |
0f7d52f4 CM |
345 | } |
346 | } | |
54aa1f4d | 347 | return err; |
0f7d52f4 CM |
348 | } |
349 | ||
e9d0b13b CM |
350 | int btrfs_defrag_root(struct btrfs_root *root, int cacheonly) |
351 | { | |
352 | struct btrfs_fs_info *info = root->fs_info; | |
353 | int ret; | |
354 | struct btrfs_trans_handle *trans; | |
d3c2fdcf | 355 | unsigned long nr; |
e9d0b13b CM |
356 | |
357 | if (root->defrag_running) | |
358 | return 0; | |
e9d0b13b | 359 | trans = btrfs_start_transaction(root, 1); |
6b80053d | 360 | while (1) { |
e9d0b13b CM |
361 | root->defrag_running = 1; |
362 | ret = btrfs_defrag_leaves(trans, root, cacheonly); | |
d3c2fdcf | 363 | nr = trans->blocks_used; |
e9d0b13b CM |
364 | btrfs_end_transaction(trans, root); |
365 | mutex_unlock(&info->fs_mutex); | |
d3c2fdcf | 366 | btrfs_btree_balance_dirty(info->tree_root, nr); |
e9d0b13b CM |
367 | cond_resched(); |
368 | ||
369 | mutex_lock(&info->fs_mutex); | |
370 | trans = btrfs_start_transaction(root, 1); | |
371 | if (ret != -EAGAIN) | |
372 | break; | |
373 | } | |
374 | root->defrag_running = 0; | |
375 | radix_tree_tag_clear(&info->fs_roots_radix, | |
376 | (unsigned long)root->root_key.objectid, | |
377 | BTRFS_ROOT_DEFRAG_TAG); | |
378 | btrfs_end_transaction(trans, root); | |
379 | return 0; | |
380 | } | |
381 | ||
6702ed49 CM |
382 | int btrfs_defrag_dirty_roots(struct btrfs_fs_info *info) |
383 | { | |
384 | struct btrfs_root *gang[1]; | |
385 | struct btrfs_root *root; | |
6702ed49 CM |
386 | int i; |
387 | int ret; | |
388 | int err = 0; | |
389 | u64 last = 0; | |
390 | ||
6702ed49 CM |
391 | while(1) { |
392 | ret = radix_tree_gang_lookup_tag(&info->fs_roots_radix, | |
393 | (void **)gang, last, | |
394 | ARRAY_SIZE(gang), | |
395 | BTRFS_ROOT_DEFRAG_TAG); | |
396 | if (ret == 0) | |
397 | break; | |
398 | for (i = 0; i < ret; i++) { | |
399 | root = gang[i]; | |
400 | last = root->root_key.objectid + 1; | |
f510cfec | 401 | btrfs_defrag_root(root, 1); |
6702ed49 CM |
402 | } |
403 | } | |
6b80053d | 404 | btrfs_defrag_root(info->extent_root, 1); |
6702ed49 CM |
405 | return err; |
406 | } | |
407 | ||
35b7e476 CM |
408 | static int drop_dirty_roots(struct btrfs_root *tree_root, |
409 | struct list_head *list) | |
0f7d52f4 CM |
410 | { |
411 | struct dirty_root *dirty; | |
412 | struct btrfs_trans_handle *trans; | |
d3c2fdcf | 413 | unsigned long nr; |
db94535d CM |
414 | u64 num_bytes; |
415 | u64 bytes_used; | |
54aa1f4d | 416 | int ret = 0; |
9f3a7427 CM |
417 | int err; |
418 | ||
0f7d52f4 | 419 | while(!list_empty(list)) { |
58176a96 JB |
420 | struct btrfs_root *root; |
421 | ||
facda1e7 | 422 | mutex_lock(&tree_root->fs_info->fs_mutex); |
0f7d52f4 CM |
423 | dirty = list_entry(list->next, struct dirty_root, list); |
424 | list_del_init(&dirty->list); | |
5eda7b5e | 425 | |
db94535d | 426 | num_bytes = btrfs_root_used(&dirty->root->root_item); |
58176a96 | 427 | root = dirty->latest_root; |
e2008b61 | 428 | root->fs_info->throttles++; |
58176a96 | 429 | |
9f3a7427 CM |
430 | while(1) { |
431 | trans = btrfs_start_transaction(tree_root, 1); | |
432 | ret = btrfs_drop_snapshot(trans, dirty->root); | |
433 | if (ret != -EAGAIN) { | |
434 | break; | |
435 | } | |
58176a96 | 436 | |
9f3a7427 CM |
437 | err = btrfs_update_root(trans, |
438 | tree_root, | |
439 | &dirty->root->root_key, | |
440 | &dirty->root->root_item); | |
441 | if (err) | |
442 | ret = err; | |
d3c2fdcf | 443 | nr = trans->blocks_used; |
9f3a7427 CM |
444 | ret = btrfs_end_transaction(trans, tree_root); |
445 | BUG_ON(ret); | |
f4468e94 | 446 | mutex_unlock(&tree_root->fs_info->fs_mutex); |
d3c2fdcf | 447 | btrfs_btree_balance_dirty(tree_root, nr); |
4dc11904 | 448 | cond_resched(); |
f4468e94 | 449 | mutex_lock(&tree_root->fs_info->fs_mutex); |
9f3a7427 | 450 | } |
0f7d52f4 | 451 | BUG_ON(ret); |
e2008b61 | 452 | root->fs_info->throttles--; |
58176a96 | 453 | |
db94535d CM |
454 | num_bytes -= btrfs_root_used(&dirty->root->root_item); |
455 | bytes_used = btrfs_root_used(&root->root_item); | |
456 | if (num_bytes) { | |
58176a96 | 457 | record_root_in_trans(root); |
5f39d397 | 458 | btrfs_set_root_used(&root->root_item, |
db94535d | 459 | bytes_used - num_bytes); |
58176a96 | 460 | } |
9f3a7427 | 461 | ret = btrfs_del_root(trans, tree_root, &dirty->root->root_key); |
58176a96 JB |
462 | if (ret) { |
463 | BUG(); | |
54aa1f4d | 464 | break; |
58176a96 | 465 | } |
d3c2fdcf | 466 | nr = trans->blocks_used; |
0f7d52f4 CM |
467 | ret = btrfs_end_transaction(trans, tree_root); |
468 | BUG_ON(ret); | |
5eda7b5e | 469 | |
f510cfec | 470 | free_extent_buffer(dirty->root->node); |
9f3a7427 | 471 | kfree(dirty->root); |
0f7d52f4 | 472 | kfree(dirty); |
facda1e7 | 473 | mutex_unlock(&tree_root->fs_info->fs_mutex); |
d3c2fdcf CM |
474 | |
475 | btrfs_btree_balance_dirty(tree_root, nr); | |
4dc11904 | 476 | cond_resched(); |
0f7d52f4 | 477 | } |
54aa1f4d | 478 | return ret; |
0f7d52f4 CM |
479 | } |
480 | ||
dc17ff8f CM |
481 | int btrfs_write_ordered_inodes(struct btrfs_trans_handle *trans, |
482 | struct btrfs_root *root) | |
483 | { | |
484 | struct btrfs_transaction *cur_trans = trans->transaction; | |
485 | struct inode *inode; | |
486 | u64 root_objectid = 0; | |
487 | u64 objectid = 0; | |
dc17ff8f CM |
488 | int ret; |
489 | ||
e2008b61 | 490 | root->fs_info->throttles++; |
dc17ff8f CM |
491 | while(1) { |
492 | ret = btrfs_find_first_ordered_inode( | |
493 | &cur_trans->ordered_inode_tree, | |
4d5e74bc | 494 | &root_objectid, &objectid, &inode); |
dc17ff8f CM |
495 | if (!ret) |
496 | break; | |
497 | ||
498 | mutex_unlock(&root->fs_info->trans_mutex); | |
499 | mutex_unlock(&root->fs_info->fs_mutex); | |
4d5e74bc CM |
500 | |
501 | if (S_ISREG(inode->i_mode)) | |
502 | filemap_fdatawrite(inode->i_mapping); | |
503 | iput(inode); | |
504 | ||
dc17ff8f CM |
505 | mutex_lock(&root->fs_info->fs_mutex); |
506 | mutex_lock(&root->fs_info->trans_mutex); | |
507 | } | |
508 | while(1) { | |
509 | root_objectid = 0; | |
510 | objectid = 0; | |
511 | ret = btrfs_find_del_first_ordered_inode( | |
512 | &cur_trans->ordered_inode_tree, | |
4d5e74bc | 513 | &root_objectid, &objectid, &inode); |
dc17ff8f CM |
514 | if (!ret) |
515 | break; | |
516 | mutex_unlock(&root->fs_info->trans_mutex); | |
517 | mutex_unlock(&root->fs_info->fs_mutex); | |
4d5e74bc CM |
518 | |
519 | if (S_ISREG(inode->i_mode)) | |
520 | filemap_write_and_wait(inode->i_mapping); | |
521 | atomic_dec(&inode->i_count); | |
522 | iput(inode); | |
523 | ||
dc17ff8f CM |
524 | mutex_lock(&root->fs_info->fs_mutex); |
525 | mutex_lock(&root->fs_info->trans_mutex); | |
526 | } | |
e2008b61 | 527 | root->fs_info->throttles--; |
3063d29f CM |
528 | return 0; |
529 | } | |
530 | ||
531 | static int create_pending_snapshot(struct btrfs_trans_handle *trans, | |
532 | struct btrfs_fs_info *fs_info, | |
533 | struct btrfs_pending_snapshot *pending) | |
534 | { | |
535 | struct btrfs_key key; | |
536 | struct btrfs_root_item new_root_item; | |
537 | struct btrfs_root *tree_root = fs_info->tree_root; | |
538 | struct btrfs_root *root = pending->root; | |
539 | struct extent_buffer *tmp; | |
540 | int ret; | |
541 | u64 objectid; | |
542 | ||
543 | ret = btrfs_find_free_objectid(trans, tree_root, 0, &objectid); | |
544 | if (ret) | |
545 | goto fail; | |
546 | ||
547 | memcpy(&new_root_item, &root->root_item, sizeof(new_root_item)); | |
548 | ||
549 | key.objectid = objectid; | |
550 | key.offset = 1; | |
551 | btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY); | |
552 | ||
553 | extent_buffer_get(root->node); | |
554 | btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp); | |
555 | free_extent_buffer(tmp); | |
556 | ||
557 | btrfs_copy_root(trans, root, root->node, &tmp, objectid); | |
558 | ||
559 | btrfs_set_root_bytenr(&new_root_item, tmp->start); | |
560 | btrfs_set_root_level(&new_root_item, btrfs_header_level(tmp)); | |
561 | ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key, | |
562 | &new_root_item); | |
563 | free_extent_buffer(tmp); | |
564 | if (ret) | |
565 | goto fail; | |
566 | ||
567 | /* | |
568 | * insert the directory item | |
569 | */ | |
570 | key.offset = (u64)-1; | |
571 | ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root, | |
572 | pending->name, strlen(pending->name), | |
573 | root->fs_info->sb->s_root->d_inode->i_ino, | |
574 | &key, BTRFS_FT_DIR); | |
575 | ||
576 | if (ret) | |
577 | goto fail; | |
578 | ||
579 | ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root, | |
580 | pending->name, strlen(pending->name), objectid, | |
581 | root->fs_info->sb->s_root->d_inode->i_ino); | |
582 | fail: | |
583 | return ret; | |
584 | } | |
585 | ||
586 | static int create_pending_snapshots(struct btrfs_trans_handle *trans, | |
587 | struct btrfs_fs_info *fs_info) | |
588 | { | |
589 | struct btrfs_pending_snapshot *pending; | |
590 | struct list_head *head = &trans->transaction->pending_snapshots; | |
591 | int ret; | |
592 | ||
593 | while(!list_empty(head)) { | |
594 | pending = list_entry(head->next, | |
595 | struct btrfs_pending_snapshot, list); | |
596 | ret = create_pending_snapshot(trans, fs_info, pending); | |
597 | BUG_ON(ret); | |
598 | list_del(&pending->list); | |
599 | kfree(pending->name); | |
600 | kfree(pending); | |
601 | } | |
dc17ff8f CM |
602 | return 0; |
603 | } | |
604 | ||
79154b1b CM |
605 | int btrfs_commit_transaction(struct btrfs_trans_handle *trans, |
606 | struct btrfs_root *root) | |
607 | { | |
15ee9bc7 JB |
608 | unsigned long joined = 0; |
609 | unsigned long timeout = 1; | |
79154b1b | 610 | struct btrfs_transaction *cur_trans; |
8fd17795 | 611 | struct btrfs_transaction *prev_trans = NULL; |
0f7d52f4 | 612 | struct list_head dirty_fs_roots; |
4313b399 | 613 | struct extent_map_tree *pinned_copy; |
79154b1b | 614 | DEFINE_WAIT(wait); |
15ee9bc7 | 615 | int ret; |
79154b1b | 616 | |
0f7d52f4 | 617 | INIT_LIST_HEAD(&dirty_fs_roots); |
d6e4a428 | 618 | |
79154b1b CM |
619 | mutex_lock(&root->fs_info->trans_mutex); |
620 | if (trans->transaction->in_commit) { | |
621 | cur_trans = trans->transaction; | |
622 | trans->transaction->use_count++; | |
ccd467d6 | 623 | mutex_unlock(&root->fs_info->trans_mutex); |
79154b1b | 624 | btrfs_end_transaction(trans, root); |
ccd467d6 CM |
625 | |
626 | mutex_unlock(&root->fs_info->fs_mutex); | |
79154b1b CM |
627 | ret = wait_for_commit(root, cur_trans); |
628 | BUG_ON(ret); | |
15ee9bc7 JB |
629 | |
630 | mutex_lock(&root->fs_info->trans_mutex); | |
79154b1b | 631 | put_transaction(cur_trans); |
15ee9bc7 JB |
632 | mutex_unlock(&root->fs_info->trans_mutex); |
633 | ||
ccd467d6 | 634 | mutex_lock(&root->fs_info->fs_mutex); |
79154b1b CM |
635 | return 0; |
636 | } | |
4313b399 CM |
637 | |
638 | pinned_copy = kmalloc(sizeof(*pinned_copy), GFP_NOFS); | |
639 | if (!pinned_copy) | |
640 | return -ENOMEM; | |
641 | ||
642 | extent_map_tree_init(pinned_copy, | |
643 | root->fs_info->btree_inode->i_mapping, GFP_NOFS); | |
644 | ||
2c90e5d6 | 645 | trans->transaction->in_commit = 1; |
ccd467d6 CM |
646 | cur_trans = trans->transaction; |
647 | if (cur_trans->list.prev != &root->fs_info->trans_list) { | |
648 | prev_trans = list_entry(cur_trans->list.prev, | |
649 | struct btrfs_transaction, list); | |
650 | if (!prev_trans->commit_done) { | |
651 | prev_trans->use_count++; | |
652 | mutex_unlock(&root->fs_info->fs_mutex); | |
653 | mutex_unlock(&root->fs_info->trans_mutex); | |
654 | ||
655 | wait_for_commit(root, prev_trans); | |
ccd467d6 CM |
656 | |
657 | mutex_lock(&root->fs_info->fs_mutex); | |
658 | mutex_lock(&root->fs_info->trans_mutex); | |
15ee9bc7 | 659 | put_transaction(prev_trans); |
ccd467d6 CM |
660 | } |
661 | } | |
15ee9bc7 JB |
662 | |
663 | do { | |
664 | joined = cur_trans->num_joined; | |
2c90e5d6 | 665 | WARN_ON(cur_trans != trans->transaction); |
15ee9bc7 | 666 | prepare_to_wait(&cur_trans->writer_wait, &wait, |
79154b1b | 667 | TASK_UNINTERRUPTIBLE); |
15ee9bc7 JB |
668 | |
669 | if (cur_trans->num_writers > 1) | |
670 | timeout = MAX_SCHEDULE_TIMEOUT; | |
671 | else | |
672 | timeout = 1; | |
673 | ||
ccd467d6 | 674 | mutex_unlock(&root->fs_info->fs_mutex); |
79154b1b | 675 | mutex_unlock(&root->fs_info->trans_mutex); |
15ee9bc7 JB |
676 | |
677 | schedule_timeout(timeout); | |
678 | ||
ccd467d6 | 679 | mutex_lock(&root->fs_info->fs_mutex); |
79154b1b | 680 | mutex_lock(&root->fs_info->trans_mutex); |
15ee9bc7 | 681 | finish_wait(&cur_trans->writer_wait, &wait); |
dc17ff8f CM |
682 | ret = btrfs_write_ordered_inodes(trans, root); |
683 | ||
15ee9bc7 JB |
684 | } while (cur_trans->num_writers > 1 || |
685 | (cur_trans->num_joined != joined)); | |
686 | ||
3063d29f CM |
687 | ret = create_pending_snapshots(trans, root->fs_info); |
688 | BUG_ON(ret); | |
689 | ||
2c90e5d6 | 690 | WARN_ON(cur_trans != trans->transaction); |
dc17ff8f | 691 | |
54aa1f4d CM |
692 | ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix, |
693 | &dirty_fs_roots); | |
694 | BUG_ON(ret); | |
695 | ||
79154b1b CM |
696 | ret = btrfs_commit_tree_roots(trans, root); |
697 | BUG_ON(ret); | |
54aa1f4d | 698 | |
78fae27e | 699 | cur_trans = root->fs_info->running_transaction; |
cee36a03 | 700 | spin_lock(&root->fs_info->new_trans_lock); |
78fae27e | 701 | root->fs_info->running_transaction = NULL; |
cee36a03 | 702 | spin_unlock(&root->fs_info->new_trans_lock); |
4b52dff6 CM |
703 | btrfs_set_super_generation(&root->fs_info->super_copy, |
704 | cur_trans->transid); | |
705 | btrfs_set_super_root(&root->fs_info->super_copy, | |
db94535d CM |
706 | root->fs_info->tree_root->node->start); |
707 | btrfs_set_super_root_level(&root->fs_info->super_copy, | |
708 | btrfs_header_level(root->fs_info->tree_root->node)); | |
5f39d397 CM |
709 | |
710 | write_extent_buffer(root->fs_info->sb_buffer, | |
711 | &root->fs_info->super_copy, 0, | |
712 | sizeof(root->fs_info->super_copy)); | |
ccd467d6 | 713 | |
4313b399 | 714 | btrfs_copy_pinned(root, pinned_copy); |
ccd467d6 | 715 | |
78fae27e | 716 | mutex_unlock(&root->fs_info->trans_mutex); |
8fd17795 | 717 | mutex_unlock(&root->fs_info->fs_mutex); |
79154b1b CM |
718 | ret = btrfs_write_and_wait_transaction(trans, root); |
719 | BUG_ON(ret); | |
79154b1b | 720 | write_ctree_super(trans, root); |
4313b399 | 721 | |
8fd17795 | 722 | mutex_lock(&root->fs_info->fs_mutex); |
4313b399 | 723 | btrfs_finish_extent_commit(trans, root, pinned_copy); |
78fae27e | 724 | mutex_lock(&root->fs_info->trans_mutex); |
4313b399 CM |
725 | |
726 | kfree(pinned_copy); | |
727 | ||
2c90e5d6 | 728 | cur_trans->commit_done = 1; |
15ee9bc7 | 729 | root->fs_info->last_trans_committed = cur_trans->transid; |
2c90e5d6 | 730 | wake_up(&cur_trans->commit_wait); |
78fae27e | 731 | put_transaction(cur_trans); |
79154b1b | 732 | put_transaction(cur_trans); |
58176a96 | 733 | |
facda1e7 CM |
734 | if (root->fs_info->closing) |
735 | list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots); | |
736 | else | |
737 | list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots); | |
58176a96 | 738 | |
78fae27e | 739 | mutex_unlock(&root->fs_info->trans_mutex); |
2c90e5d6 | 740 | kmem_cache_free(btrfs_trans_handle_cachep, trans); |
79154b1b | 741 | |
facda1e7 CM |
742 | if (root->fs_info->closing) { |
743 | mutex_unlock(&root->fs_info->fs_mutex); | |
744 | drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots); | |
745 | mutex_lock(&root->fs_info->fs_mutex); | |
746 | } | |
79154b1b CM |
747 | return ret; |
748 | } | |
749 | ||
e9d0b13b CM |
750 | int btrfs_clean_old_snapshots(struct btrfs_root *root) |
751 | { | |
752 | struct list_head dirty_roots; | |
753 | INIT_LIST_HEAD(&dirty_roots); | |
754 | ||
755 | mutex_lock(&root->fs_info->trans_mutex); | |
756 | list_splice_init(&root->fs_info->dead_roots, &dirty_roots); | |
757 | mutex_unlock(&root->fs_info->trans_mutex); | |
758 | ||
759 | if (!list_empty(&dirty_roots)) { | |
760 | drop_dirty_roots(root, &dirty_roots); | |
761 | } | |
762 | return 0; | |
763 | } | |
6da6abae CM |
764 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
765 | void btrfs_transaction_cleaner(void *p) | |
766 | #else | |
08607c1b | 767 | void btrfs_transaction_cleaner(struct work_struct *work) |
6da6abae | 768 | #endif |
08607c1b | 769 | { |
6da6abae CM |
770 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
771 | struct btrfs_fs_info *fs_info = p; | |
772 | #else | |
08607c1b CM |
773 | struct btrfs_fs_info *fs_info = container_of(work, |
774 | struct btrfs_fs_info, | |
775 | trans_work.work); | |
776 | ||
6da6abae | 777 | #endif |
08607c1b CM |
778 | struct btrfs_root *root = fs_info->tree_root; |
779 | struct btrfs_transaction *cur; | |
780 | struct btrfs_trans_handle *trans; | |
781 | unsigned long now; | |
782 | unsigned long delay = HZ * 30; | |
783 | int ret; | |
784 | ||
08607c1b CM |
785 | mutex_lock(&root->fs_info->fs_mutex); |
786 | mutex_lock(&root->fs_info->trans_mutex); | |
787 | cur = root->fs_info->running_transaction; | |
788 | if (!cur) { | |
789 | mutex_unlock(&root->fs_info->trans_mutex); | |
790 | goto out; | |
791 | } | |
792 | now = get_seconds(); | |
793 | if (now < cur->start_time || now - cur->start_time < 30) { | |
794 | mutex_unlock(&root->fs_info->trans_mutex); | |
795 | delay = HZ * 5; | |
796 | goto out; | |
797 | } | |
798 | mutex_unlock(&root->fs_info->trans_mutex); | |
6702ed49 | 799 | btrfs_defrag_dirty_roots(root->fs_info); |
08607c1b CM |
800 | trans = btrfs_start_transaction(root, 1); |
801 | ret = btrfs_commit_transaction(trans, root); | |
802 | out: | |
803 | mutex_unlock(&root->fs_info->fs_mutex); | |
e9d0b13b | 804 | btrfs_clean_old_snapshots(root); |
08607c1b CM |
805 | btrfs_transaction_queue_work(root, delay); |
806 | } | |
807 | ||
808 | void btrfs_transaction_queue_work(struct btrfs_root *root, int delay) | |
809 | { | |
810 | queue_delayed_work(trans_wq, &root->fs_info->trans_work, delay); | |
811 | } | |
812 | ||
813 | void btrfs_transaction_flush_work(struct btrfs_root *root) | |
814 | { | |
815 | cancel_rearming_delayed_workqueue(trans_wq, &root->fs_info->trans_work); | |
816 | flush_workqueue(trans_wq); | |
817 | } | |
818 | ||
819 | void __init btrfs_init_transaction_sys(void) | |
820 | { | |
821 | trans_wq = create_workqueue("btrfs"); | |
822 | } | |
823 | ||
17636e03 | 824 | void btrfs_exit_transaction_sys(void) |
08607c1b CM |
825 | { |
826 | destroy_workqueue(trans_wq); | |
827 | } | |
828 |