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 CM |
19 | #include <linux/module.h> |
20 | #include <linux/fs.h> | |
34088780 | 21 | #include <linux/sched.h> |
79154b1b CM |
22 | #include "ctree.h" |
23 | #include "disk-io.h" | |
24 | #include "transaction.h" | |
25 | ||
78fae27e | 26 | static int total_trans = 0; |
2c90e5d6 CM |
27 | extern struct kmem_cache *btrfs_trans_handle_cachep; |
28 | extern struct kmem_cache *btrfs_transaction_cachep; | |
29 | ||
08607c1b CM |
30 | static struct workqueue_struct *trans_wq; |
31 | ||
0f7d52f4 CM |
32 | #define BTRFS_ROOT_TRANS_TAG 0 |
33 | ||
79154b1b CM |
34 | static void put_transaction(struct btrfs_transaction *transaction) |
35 | { | |
2c90e5d6 | 36 | WARN_ON(transaction->use_count == 0); |
79154b1b | 37 | transaction->use_count--; |
78fae27e CM |
38 | if (transaction->use_count == 0) { |
39 | WARN_ON(total_trans == 0); | |
40 | total_trans--; | |
8fd17795 | 41 | list_del_init(&transaction->list); |
2c90e5d6 CM |
42 | memset(transaction, 0, sizeof(*transaction)); |
43 | kmem_cache_free(btrfs_transaction_cachep, transaction); | |
78fae27e | 44 | } |
79154b1b CM |
45 | } |
46 | ||
47 | static int join_transaction(struct btrfs_root *root) | |
48 | { | |
49 | struct btrfs_transaction *cur_trans; | |
50 | cur_trans = root->fs_info->running_transaction; | |
51 | if (!cur_trans) { | |
2c90e5d6 CM |
52 | cur_trans = kmem_cache_alloc(btrfs_transaction_cachep, |
53 | GFP_NOFS); | |
78fae27e | 54 | total_trans++; |
79154b1b | 55 | BUG_ON(!cur_trans); |
0f7d52f4 | 56 | root->fs_info->generation++; |
79154b1b CM |
57 | root->fs_info->running_transaction = cur_trans; |
58 | cur_trans->num_writers = 0; | |
0f7d52f4 | 59 | cur_trans->transid = root->fs_info->generation; |
79154b1b CM |
60 | init_waitqueue_head(&cur_trans->writer_wait); |
61 | init_waitqueue_head(&cur_trans->commit_wait); | |
62 | cur_trans->in_commit = 0; | |
d5719762 | 63 | cur_trans->use_count = 1; |
79154b1b | 64 | cur_trans->commit_done = 0; |
08607c1b | 65 | cur_trans->start_time = get_seconds(); |
8fd17795 | 66 | list_add_tail(&cur_trans->list, &root->fs_info->trans_list); |
7c4452b9 | 67 | init_bit_radix(&cur_trans->dirty_pages); |
79154b1b CM |
68 | } |
69 | cur_trans->num_writers++; | |
70 | return 0; | |
71 | } | |
72 | ||
73 | struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root, | |
74 | int num_blocks) | |
75 | { | |
2c90e5d6 CM |
76 | struct btrfs_trans_handle *h = |
77 | kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS); | |
79154b1b | 78 | int ret; |
0f7d52f4 | 79 | u64 running_trans_id; |
79154b1b CM |
80 | |
81 | mutex_lock(&root->fs_info->trans_mutex); | |
82 | ret = join_transaction(root); | |
83 | BUG_ON(ret); | |
0f7d52f4 CM |
84 | running_trans_id = root->fs_info->running_transaction->transid; |
85 | ||
86 | if (root != root->fs_info->tree_root && root->last_trans < | |
87 | running_trans_id) { | |
5eda7b5e CM |
88 | if (root->root_item.refs != 0) { |
89 | radix_tree_tag_set(&root->fs_info->fs_roots_radix, | |
90 | (unsigned long)root->root_key.objectid, | |
91 | BTRFS_ROOT_TRANS_TAG); | |
92 | root->commit_root = root->node; | |
93 | get_bh(root->node); | |
94 | } else { | |
95 | WARN_ON(1); | |
96 | } | |
0f7d52f4 CM |
97 | } |
98 | root->last_trans = running_trans_id; | |
99 | h->transid = running_trans_id; | |
79154b1b CM |
100 | h->transaction = root->fs_info->running_transaction; |
101 | h->blocks_reserved = num_blocks; | |
102 | h->blocks_used = 0; | |
31f3c99b | 103 | h->block_group = NULL; |
79154b1b CM |
104 | root->fs_info->running_transaction->use_count++; |
105 | mutex_unlock(&root->fs_info->trans_mutex); | |
106 | return h; | |
107 | } | |
108 | ||
109 | int btrfs_end_transaction(struct btrfs_trans_handle *trans, | |
110 | struct btrfs_root *root) | |
111 | { | |
112 | struct btrfs_transaction *cur_trans; | |
d6e4a428 | 113 | |
79154b1b CM |
114 | mutex_lock(&root->fs_info->trans_mutex); |
115 | cur_trans = root->fs_info->running_transaction; | |
d5719762 | 116 | WARN_ON(cur_trans->num_writers < 1); |
79154b1b CM |
117 | if (waitqueue_active(&cur_trans->writer_wait)) |
118 | wake_up(&cur_trans->writer_wait); | |
119 | cur_trans->num_writers--; | |
120 | put_transaction(cur_trans); | |
121 | mutex_unlock(&root->fs_info->trans_mutex); | |
d6025579 | 122 | memset(trans, 0, sizeof(*trans)); |
2c90e5d6 | 123 | kmem_cache_free(btrfs_trans_handle_cachep, trans); |
79154b1b CM |
124 | return 0; |
125 | } | |
126 | ||
127 | ||
128 | int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans, | |
129 | struct btrfs_root *root) | |
130 | { | |
7c4452b9 CM |
131 | unsigned long gang[16]; |
132 | int ret; | |
133 | int i; | |
134 | int err; | |
135 | int werr = 0; | |
136 | struct page *page; | |
137 | struct radix_tree_root *dirty_pages; | |
138 | struct inode *btree_inode = root->fs_info->btree_inode; | |
139 | ||
140 | if (!trans || !trans->transaction) { | |
141 | return filemap_write_and_wait(btree_inode->i_mapping); | |
142 | } | |
143 | dirty_pages = &trans->transaction->dirty_pages; | |
144 | while(1) { | |
e37c9e69 CM |
145 | ret = find_first_radix_bit(dirty_pages, gang, |
146 | 0, ARRAY_SIZE(gang)); | |
7c4452b9 CM |
147 | if (!ret) |
148 | break; | |
149 | for (i = 0; i < ret; i++) { | |
150 | /* FIXME EIO */ | |
151 | clear_radix_bit(dirty_pages, gang[i]); | |
152 | page = find_lock_page(btree_inode->i_mapping, | |
153 | gang[i]); | |
154 | if (!page) | |
155 | continue; | |
156 | err = write_one_page(page, 0); | |
157 | if (err) | |
158 | werr = err; | |
159 | page_cache_release(page); | |
160 | } | |
161 | } | |
162 | err = filemap_fdatawait(btree_inode->i_mapping); | |
163 | if (err) | |
164 | werr = err; | |
165 | return werr; | |
79154b1b CM |
166 | } |
167 | ||
168 | int btrfs_commit_tree_roots(struct btrfs_trans_handle *trans, | |
169 | struct btrfs_root *root) | |
170 | { | |
171 | int ret; | |
172 | u64 old_extent_block; | |
173 | struct btrfs_fs_info *fs_info = root->fs_info; | |
174 | struct btrfs_root *tree_root = fs_info->tree_root; | |
175 | struct btrfs_root *extent_root = fs_info->extent_root; | |
79154b1b | 176 | |
9078a3e1 | 177 | btrfs_write_dirty_block_groups(trans, extent_root); |
79154b1b CM |
178 | while(1) { |
179 | old_extent_block = btrfs_root_blocknr(&extent_root->root_item); | |
7eccb903 | 180 | if (old_extent_block == bh_blocknr(extent_root->node)) |
79154b1b CM |
181 | break; |
182 | btrfs_set_root_blocknr(&extent_root->root_item, | |
7eccb903 | 183 | bh_blocknr(extent_root->node)); |
79154b1b CM |
184 | ret = btrfs_update_root(trans, tree_root, |
185 | &extent_root->root_key, | |
186 | &extent_root->root_item); | |
187 | BUG_ON(ret); | |
9078a3e1 | 188 | btrfs_write_dirty_block_groups(trans, extent_root); |
79154b1b CM |
189 | } |
190 | return 0; | |
191 | } | |
192 | ||
193 | static int wait_for_commit(struct btrfs_root *root, | |
194 | struct btrfs_transaction *commit) | |
195 | { | |
196 | DEFINE_WAIT(wait); | |
79154b1b CM |
197 | while(!commit->commit_done) { |
198 | prepare_to_wait(&commit->commit_wait, &wait, | |
199 | TASK_UNINTERRUPTIBLE); | |
200 | if (commit->commit_done) | |
201 | break; | |
202 | mutex_unlock(&root->fs_info->trans_mutex); | |
203 | schedule(); | |
204 | mutex_lock(&root->fs_info->trans_mutex); | |
205 | } | |
206 | finish_wait(&commit->commit_wait, &wait); | |
207 | return 0; | |
208 | } | |
209 | ||
0f7d52f4 CM |
210 | struct dirty_root { |
211 | struct list_head list; | |
212 | struct btrfs_key snap_key; | |
213 | struct buffer_head *commit_root; | |
214 | struct btrfs_root *root; | |
5eda7b5e | 215 | int free_on_drop; |
0f7d52f4 CM |
216 | }; |
217 | ||
5eda7b5e CM |
218 | int btrfs_add_dead_root(struct btrfs_root *root, struct list_head *dead_list) |
219 | { | |
220 | struct dirty_root *dirty; | |
221 | ||
222 | dirty = kmalloc(sizeof(*dirty), GFP_NOFS); | |
223 | if (!dirty) | |
224 | return -ENOMEM; | |
225 | memcpy(&dirty->snap_key, &root->root_key, sizeof(root->root_key)); | |
226 | dirty->commit_root = root->node; | |
227 | dirty->root = root; | |
228 | dirty->free_on_drop = 1; | |
229 | list_add(&dirty->list, dead_list); | |
230 | return 0; | |
231 | } | |
232 | ||
35b7e476 CM |
233 | static int add_dirty_roots(struct btrfs_trans_handle *trans, |
234 | struct radix_tree_root *radix, | |
235 | struct list_head *list) | |
0f7d52f4 CM |
236 | { |
237 | struct dirty_root *dirty; | |
238 | struct btrfs_root *gang[8]; | |
239 | struct btrfs_root *root; | |
5eda7b5e | 240 | struct btrfs_root_item tmp_item; |
0f7d52f4 CM |
241 | int i; |
242 | int ret; | |
54aa1f4d | 243 | int err = 0; |
5eda7b5e | 244 | u32 refs; |
54aa1f4d | 245 | |
0f7d52f4 CM |
246 | while(1) { |
247 | ret = radix_tree_gang_lookup_tag(radix, (void **)gang, 0, | |
248 | ARRAY_SIZE(gang), | |
249 | BTRFS_ROOT_TRANS_TAG); | |
250 | if (ret == 0) | |
251 | break; | |
252 | for (i = 0; i < ret; i++) { | |
253 | root = gang[i]; | |
2619ba1f CM |
254 | radix_tree_tag_clear(radix, |
255 | (unsigned long)root->root_key.objectid, | |
256 | BTRFS_ROOT_TRANS_TAG); | |
0f7d52f4 | 257 | if (root->commit_root == root->node) { |
7eccb903 | 258 | WARN_ON(bh_blocknr(root->node) != |
0f7d52f4 CM |
259 | btrfs_root_blocknr(&root->root_item)); |
260 | brelse(root->commit_root); | |
261 | root->commit_root = NULL; | |
262 | continue; | |
263 | } | |
264 | dirty = kmalloc(sizeof(*dirty), GFP_NOFS); | |
265 | BUG_ON(!dirty); | |
266 | memcpy(&dirty->snap_key, &root->root_key, | |
267 | sizeof(root->root_key)); | |
268 | dirty->commit_root = root->commit_root; | |
269 | root->commit_root = NULL; | |
270 | dirty->root = root; | |
5eda7b5e CM |
271 | dirty->free_on_drop = 0; |
272 | memcpy(&tmp_item, &root->root_item, sizeof(tmp_item)); | |
273 | ||
0f7d52f4 CM |
274 | root->root_key.offset = root->fs_info->generation; |
275 | btrfs_set_root_blocknr(&root->root_item, | |
7eccb903 | 276 | bh_blocknr(root->node)); |
0f7d52f4 CM |
277 | err = btrfs_insert_root(trans, root->fs_info->tree_root, |
278 | &root->root_key, | |
279 | &root->root_item); | |
54aa1f4d CM |
280 | if (err) |
281 | break; | |
5eda7b5e CM |
282 | |
283 | refs = btrfs_root_refs(&tmp_item); | |
284 | btrfs_set_root_refs(&tmp_item, refs - 1); | |
285 | err = btrfs_update_root(trans, root->fs_info->tree_root, | |
286 | &dirty->snap_key, | |
287 | &tmp_item); | |
288 | ||
289 | BUG_ON(err); | |
290 | if (refs == 1) | |
291 | list_add(&dirty->list, list); | |
292 | else | |
293 | kfree(dirty); | |
0f7d52f4 CM |
294 | } |
295 | } | |
54aa1f4d | 296 | return err; |
0f7d52f4 CM |
297 | } |
298 | ||
35b7e476 CM |
299 | static int drop_dirty_roots(struct btrfs_root *tree_root, |
300 | struct list_head *list) | |
0f7d52f4 CM |
301 | { |
302 | struct dirty_root *dirty; | |
303 | struct btrfs_trans_handle *trans; | |
54aa1f4d | 304 | int ret = 0; |
0f7d52f4 | 305 | while(!list_empty(list)) { |
facda1e7 | 306 | mutex_lock(&tree_root->fs_info->fs_mutex); |
0f7d52f4 CM |
307 | dirty = list_entry(list->next, struct dirty_root, list); |
308 | list_del_init(&dirty->list); | |
5eda7b5e | 309 | |
0f7d52f4 | 310 | trans = btrfs_start_transaction(tree_root, 1); |
5eda7b5e | 311 | printk("deleting root %Lu %u %Lu\n", dirty->snap_key.objectid, dirty->snap_key.flags, dirty->snap_key.offset); |
0f7d52f4 CM |
312 | ret = btrfs_drop_snapshot(trans, dirty->root, |
313 | dirty->commit_root); | |
314 | BUG_ON(ret); | |
0f7d52f4 | 315 | ret = btrfs_del_root(trans, tree_root, &dirty->snap_key); |
54aa1f4d CM |
316 | if (ret) |
317 | break; | |
0f7d52f4 CM |
318 | ret = btrfs_end_transaction(trans, tree_root); |
319 | BUG_ON(ret); | |
5eda7b5e CM |
320 | |
321 | if (dirty->free_on_drop) | |
322 | kfree(dirty->root); | |
0f7d52f4 | 323 | kfree(dirty); |
facda1e7 | 324 | mutex_unlock(&tree_root->fs_info->fs_mutex); |
8c2383c3 | 325 | btrfs_btree_balance_dirty(tree_root); |
0f7d52f4 | 326 | } |
54aa1f4d | 327 | return ret; |
0f7d52f4 CM |
328 | } |
329 | ||
79154b1b CM |
330 | int btrfs_commit_transaction(struct btrfs_trans_handle *trans, |
331 | struct btrfs_root *root) | |
332 | { | |
333 | int ret = 0; | |
79154b1b | 334 | struct btrfs_transaction *cur_trans; |
8fd17795 | 335 | struct btrfs_transaction *prev_trans = NULL; |
0f7d52f4 | 336 | struct list_head dirty_fs_roots; |
79154b1b CM |
337 | DEFINE_WAIT(wait); |
338 | ||
0f7d52f4 | 339 | INIT_LIST_HEAD(&dirty_fs_roots); |
d6e4a428 | 340 | |
79154b1b CM |
341 | mutex_lock(&root->fs_info->trans_mutex); |
342 | if (trans->transaction->in_commit) { | |
343 | cur_trans = trans->transaction; | |
344 | trans->transaction->use_count++; | |
345 | btrfs_end_transaction(trans, root); | |
346 | ret = wait_for_commit(root, cur_trans); | |
347 | BUG_ON(ret); | |
348 | put_transaction(cur_trans); | |
349 | mutex_unlock(&root->fs_info->trans_mutex); | |
350 | return 0; | |
351 | } | |
2c90e5d6 CM |
352 | cur_trans = trans->transaction; |
353 | trans->transaction->in_commit = 1; | |
79154b1b | 354 | while (trans->transaction->num_writers > 1) { |
2c90e5d6 | 355 | WARN_ON(cur_trans != trans->transaction); |
79154b1b CM |
356 | prepare_to_wait(&trans->transaction->writer_wait, &wait, |
357 | TASK_UNINTERRUPTIBLE); | |
358 | if (trans->transaction->num_writers <= 1) | |
359 | break; | |
360 | mutex_unlock(&root->fs_info->trans_mutex); | |
361 | schedule(); | |
362 | mutex_lock(&root->fs_info->trans_mutex); | |
2c90e5d6 | 363 | finish_wait(&trans->transaction->writer_wait, &wait); |
79154b1b CM |
364 | } |
365 | finish_wait(&trans->transaction->writer_wait, &wait); | |
2c90e5d6 | 366 | WARN_ON(cur_trans != trans->transaction); |
54aa1f4d CM |
367 | ret = add_dirty_roots(trans, &root->fs_info->fs_roots_radix, |
368 | &dirty_fs_roots); | |
369 | BUG_ON(ret); | |
370 | ||
79154b1b CM |
371 | ret = btrfs_commit_tree_roots(trans, root); |
372 | BUG_ON(ret); | |
54aa1f4d | 373 | |
78fae27e CM |
374 | cur_trans = root->fs_info->running_transaction; |
375 | root->fs_info->running_transaction = NULL; | |
8fd17795 CM |
376 | if (cur_trans->list.prev != &root->fs_info->trans_list) { |
377 | prev_trans = list_entry(cur_trans->list.prev, | |
378 | struct btrfs_transaction, list); | |
379 | if (prev_trans->commit_done) | |
380 | prev_trans = NULL; | |
381 | else | |
382 | prev_trans->use_count++; | |
383 | } | |
78fae27e | 384 | mutex_unlock(&root->fs_info->trans_mutex); |
8fd17795 | 385 | mutex_unlock(&root->fs_info->fs_mutex); |
79154b1b | 386 | ret = btrfs_write_and_wait_transaction(trans, root); |
8fd17795 CM |
387 | if (prev_trans) { |
388 | mutex_lock(&root->fs_info->trans_mutex); | |
389 | wait_for_commit(root, prev_trans); | |
390 | put_transaction(prev_trans); | |
391 | mutex_unlock(&root->fs_info->trans_mutex); | |
392 | } | |
393 | btrfs_set_super_generation(root->fs_info->disk_super, | |
394 | cur_trans->transid); | |
79154b1b | 395 | BUG_ON(ret); |
79154b1b | 396 | write_ctree_super(trans, root); |
8fd17795 CM |
397 | |
398 | mutex_lock(&root->fs_info->fs_mutex); | |
78fae27e CM |
399 | btrfs_finish_extent_commit(trans, root); |
400 | mutex_lock(&root->fs_info->trans_mutex); | |
2c90e5d6 CM |
401 | cur_trans->commit_done = 1; |
402 | wake_up(&cur_trans->commit_wait); | |
78fae27e | 403 | put_transaction(cur_trans); |
79154b1b | 404 | put_transaction(cur_trans); |
facda1e7 CM |
405 | if (root->fs_info->closing) |
406 | list_splice_init(&root->fs_info->dead_roots, &dirty_fs_roots); | |
407 | else | |
408 | list_splice_init(&dirty_fs_roots, &root->fs_info->dead_roots); | |
78fae27e | 409 | mutex_unlock(&root->fs_info->trans_mutex); |
2c90e5d6 | 410 | kmem_cache_free(btrfs_trans_handle_cachep, trans); |
79154b1b | 411 | |
facda1e7 CM |
412 | if (root->fs_info->closing) { |
413 | mutex_unlock(&root->fs_info->fs_mutex); | |
414 | drop_dirty_roots(root->fs_info->tree_root, &dirty_fs_roots); | |
415 | mutex_lock(&root->fs_info->fs_mutex); | |
416 | } | |
79154b1b CM |
417 | return ret; |
418 | } | |
419 | ||
08607c1b CM |
420 | void btrfs_transaction_cleaner(struct work_struct *work) |
421 | { | |
422 | struct btrfs_fs_info *fs_info = container_of(work, | |
423 | struct btrfs_fs_info, | |
424 | trans_work.work); | |
425 | ||
426 | struct btrfs_root *root = fs_info->tree_root; | |
427 | struct btrfs_transaction *cur; | |
428 | struct btrfs_trans_handle *trans; | |
facda1e7 | 429 | struct list_head dirty_roots; |
08607c1b CM |
430 | unsigned long now; |
431 | unsigned long delay = HZ * 30; | |
432 | int ret; | |
433 | ||
facda1e7 | 434 | INIT_LIST_HEAD(&dirty_roots); |
08607c1b CM |
435 | mutex_lock(&root->fs_info->fs_mutex); |
436 | mutex_lock(&root->fs_info->trans_mutex); | |
437 | cur = root->fs_info->running_transaction; | |
438 | if (!cur) { | |
439 | mutex_unlock(&root->fs_info->trans_mutex); | |
440 | goto out; | |
441 | } | |
442 | now = get_seconds(); | |
443 | if (now < cur->start_time || now - cur->start_time < 30) { | |
444 | mutex_unlock(&root->fs_info->trans_mutex); | |
445 | delay = HZ * 5; | |
446 | goto out; | |
447 | } | |
448 | mutex_unlock(&root->fs_info->trans_mutex); | |
08607c1b CM |
449 | trans = btrfs_start_transaction(root, 1); |
450 | ret = btrfs_commit_transaction(trans, root); | |
451 | out: | |
452 | mutex_unlock(&root->fs_info->fs_mutex); | |
ad693af6 CM |
453 | |
454 | mutex_lock(&root->fs_info->trans_mutex); | |
455 | list_splice_init(&root->fs_info->dead_roots, &dirty_roots); | |
456 | mutex_unlock(&root->fs_info->trans_mutex); | |
457 | ||
458 | if (!list_empty(&dirty_roots)) { | |
459 | drop_dirty_roots(root, &dirty_roots); | |
460 | } | |
08607c1b CM |
461 | btrfs_transaction_queue_work(root, delay); |
462 | } | |
463 | ||
464 | void btrfs_transaction_queue_work(struct btrfs_root *root, int delay) | |
465 | { | |
466 | queue_delayed_work(trans_wq, &root->fs_info->trans_work, delay); | |
467 | } | |
468 | ||
469 | void btrfs_transaction_flush_work(struct btrfs_root *root) | |
470 | { | |
471 | cancel_rearming_delayed_workqueue(trans_wq, &root->fs_info->trans_work); | |
472 | flush_workqueue(trans_wq); | |
473 | } | |
474 | ||
475 | void __init btrfs_init_transaction_sys(void) | |
476 | { | |
477 | trans_wq = create_workqueue("btrfs"); | |
478 | } | |
479 | ||
480 | void __exit btrfs_exit_transaction_sys(void) | |
481 | { | |
482 | destroy_workqueue(trans_wq); | |
483 | } | |
484 |