]>
Commit | Line | Data |
---|---|---|
dc17ff8f 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 | ||
19 | #include <linux/gfp.h> | |
20 | #include <linux/slab.h> | |
d6bfde87 | 21 | #include <linux/blkdev.h> |
f421950f CM |
22 | #include <linux/writeback.h> |
23 | #include <linux/pagevec.h> | |
dc17ff8f CM |
24 | #include "ctree.h" |
25 | #include "transaction.h" | |
26 | #include "btrfs_inode.h" | |
e6dcd2dc | 27 | #include "extent_io.h" |
dc17ff8f | 28 | |
dc17ff8f | 29 | |
e6dcd2dc | 30 | static u64 entry_end(struct btrfs_ordered_extent *entry) |
dc17ff8f | 31 | { |
e6dcd2dc CM |
32 | if (entry->file_offset + entry->len < entry->file_offset) |
33 | return (u64)-1; | |
34 | return entry->file_offset + entry->len; | |
dc17ff8f CM |
35 | } |
36 | ||
e6dcd2dc CM |
37 | static struct rb_node *tree_insert(struct rb_root *root, u64 file_offset, |
38 | struct rb_node *node) | |
dc17ff8f CM |
39 | { |
40 | struct rb_node ** p = &root->rb_node; | |
41 | struct rb_node * parent = NULL; | |
e6dcd2dc | 42 | struct btrfs_ordered_extent *entry; |
dc17ff8f CM |
43 | |
44 | while(*p) { | |
45 | parent = *p; | |
e6dcd2dc | 46 | entry = rb_entry(parent, struct btrfs_ordered_extent, rb_node); |
dc17ff8f | 47 | |
e6dcd2dc | 48 | if (file_offset < entry->file_offset) |
dc17ff8f | 49 | p = &(*p)->rb_left; |
e6dcd2dc | 50 | else if (file_offset >= entry_end(entry)) |
dc17ff8f CM |
51 | p = &(*p)->rb_right; |
52 | else | |
53 | return parent; | |
54 | } | |
55 | ||
56 | rb_link_node(node, parent, p); | |
57 | rb_insert_color(node, root); | |
58 | return NULL; | |
59 | } | |
60 | ||
e6dcd2dc CM |
61 | static struct rb_node *__tree_search(struct rb_root *root, u64 file_offset, |
62 | struct rb_node **prev_ret) | |
dc17ff8f CM |
63 | { |
64 | struct rb_node * n = root->rb_node; | |
65 | struct rb_node *prev = NULL; | |
e6dcd2dc CM |
66 | struct rb_node *test; |
67 | struct btrfs_ordered_extent *entry; | |
68 | struct btrfs_ordered_extent *prev_entry = NULL; | |
dc17ff8f CM |
69 | |
70 | while(n) { | |
e6dcd2dc | 71 | entry = rb_entry(n, struct btrfs_ordered_extent, rb_node); |
dc17ff8f CM |
72 | prev = n; |
73 | prev_entry = entry; | |
dc17ff8f | 74 | |
e6dcd2dc | 75 | if (file_offset < entry->file_offset) |
dc17ff8f | 76 | n = n->rb_left; |
e6dcd2dc | 77 | else if (file_offset >= entry_end(entry)) |
dc17ff8f CM |
78 | n = n->rb_right; |
79 | else | |
80 | return n; | |
81 | } | |
82 | if (!prev_ret) | |
83 | return NULL; | |
84 | ||
e6dcd2dc CM |
85 | while(prev && file_offset >= entry_end(prev_entry)) { |
86 | test = rb_next(prev); | |
87 | if (!test) | |
88 | break; | |
89 | prev_entry = rb_entry(test, struct btrfs_ordered_extent, | |
90 | rb_node); | |
91 | if (file_offset < entry_end(prev_entry)) | |
92 | break; | |
93 | ||
94 | prev = test; | |
95 | } | |
96 | if (prev) | |
97 | prev_entry = rb_entry(prev, struct btrfs_ordered_extent, | |
98 | rb_node); | |
99 | while(prev && file_offset < entry_end(prev_entry)) { | |
100 | test = rb_prev(prev); | |
101 | if (!test) | |
102 | break; | |
103 | prev_entry = rb_entry(test, struct btrfs_ordered_extent, | |
104 | rb_node); | |
105 | prev = test; | |
dc17ff8f CM |
106 | } |
107 | *prev_ret = prev; | |
108 | return NULL; | |
109 | } | |
110 | ||
e6dcd2dc CM |
111 | static int offset_in_entry(struct btrfs_ordered_extent *entry, u64 file_offset) |
112 | { | |
113 | if (file_offset < entry->file_offset || | |
114 | entry->file_offset + entry->len <= file_offset) | |
115 | return 0; | |
116 | return 1; | |
117 | } | |
118 | ||
119 | static inline struct rb_node *tree_search(struct btrfs_ordered_inode_tree *tree, | |
120 | u64 file_offset) | |
dc17ff8f | 121 | { |
e6dcd2dc | 122 | struct rb_root *root = &tree->tree; |
dc17ff8f CM |
123 | struct rb_node *prev; |
124 | struct rb_node *ret; | |
e6dcd2dc CM |
125 | struct btrfs_ordered_extent *entry; |
126 | ||
127 | if (tree->last) { | |
128 | entry = rb_entry(tree->last, struct btrfs_ordered_extent, | |
129 | rb_node); | |
130 | if (offset_in_entry(entry, file_offset)) | |
131 | return tree->last; | |
132 | } | |
133 | ret = __tree_search(root, file_offset, &prev); | |
dc17ff8f | 134 | if (!ret) |
e6dcd2dc CM |
135 | ret = prev; |
136 | if (ret) | |
137 | tree->last = ret; | |
dc17ff8f CM |
138 | return ret; |
139 | } | |
140 | ||
eb84ae03 CM |
141 | /* allocate and add a new ordered_extent into the per-inode tree. |
142 | * file_offset is the logical offset in the file | |
143 | * | |
144 | * start is the disk block number of an extent already reserved in the | |
145 | * extent allocation tree | |
146 | * | |
147 | * len is the length of the extent | |
148 | * | |
149 | * This also sets the EXTENT_ORDERED bit on the range in the inode. | |
150 | * | |
151 | * The tree is given a single reference on the ordered extent that was | |
152 | * inserted. | |
153 | */ | |
e6dcd2dc CM |
154 | int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, |
155 | u64 start, u64 len) | |
dc17ff8f | 156 | { |
dc17ff8f | 157 | struct btrfs_ordered_inode_tree *tree; |
e6dcd2dc CM |
158 | struct rb_node *node; |
159 | struct btrfs_ordered_extent *entry; | |
dc17ff8f | 160 | |
e6dcd2dc CM |
161 | tree = &BTRFS_I(inode)->ordered_tree; |
162 | entry = kzalloc(sizeof(*entry), GFP_NOFS); | |
dc17ff8f CM |
163 | if (!entry) |
164 | return -ENOMEM; | |
165 | ||
e6dcd2dc CM |
166 | mutex_lock(&tree->mutex); |
167 | entry->file_offset = file_offset; | |
168 | entry->start = start; | |
169 | entry->len = len; | |
e6dcd2dc CM |
170 | /* one ref for the tree */ |
171 | atomic_set(&entry->refs, 1); | |
172 | init_waitqueue_head(&entry->wait); | |
173 | INIT_LIST_HEAD(&entry->list); | |
dc17ff8f | 174 | |
e6dcd2dc CM |
175 | node = tree_insert(&tree->tree, file_offset, |
176 | &entry->rb_node); | |
177 | if (node) { | |
178 | entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); | |
179 | atomic_inc(&entry->refs); | |
180 | } | |
181 | set_extent_ordered(&BTRFS_I(inode)->io_tree, file_offset, | |
182 | entry_end(entry) - 1, GFP_NOFS); | |
1b1e2135 | 183 | |
e6dcd2dc CM |
184 | mutex_unlock(&tree->mutex); |
185 | BUG_ON(node); | |
dc17ff8f CM |
186 | return 0; |
187 | } | |
188 | ||
eb84ae03 CM |
189 | /* |
190 | * Add a struct btrfs_ordered_sum into the list of checksums to be inserted | |
3edf7d33 CM |
191 | * when an ordered extent is finished. If the list covers more than one |
192 | * ordered extent, it is split across multiples. | |
eb84ae03 | 193 | */ |
3edf7d33 CM |
194 | int btrfs_add_ordered_sum(struct inode *inode, |
195 | struct btrfs_ordered_extent *entry, | |
196 | struct btrfs_ordered_sum *sum) | |
dc17ff8f | 197 | { |
e6dcd2dc | 198 | struct btrfs_ordered_inode_tree *tree; |
dc17ff8f | 199 | |
e6dcd2dc CM |
200 | tree = &BTRFS_I(inode)->ordered_tree; |
201 | mutex_lock(&tree->mutex); | |
e6dcd2dc CM |
202 | list_add_tail(&sum->list, &entry->list); |
203 | mutex_unlock(&tree->mutex); | |
204 | return 0; | |
dc17ff8f CM |
205 | } |
206 | ||
eb84ae03 CM |
207 | /* |
208 | * this is used to account for finished IO across a given range | |
209 | * of the file. The IO should not span ordered extents. If | |
210 | * a given ordered_extent is completely done, 1 is returned, otherwise | |
211 | * 0. | |
212 | * | |
213 | * test_and_set_bit on a flag in the struct btrfs_ordered_extent is used | |
214 | * to make sure this function only returns 1 once for a given ordered extent. | |
215 | */ | |
e6dcd2dc CM |
216 | int btrfs_dec_test_ordered_pending(struct inode *inode, |
217 | u64 file_offset, u64 io_size) | |
dc17ff8f | 218 | { |
e6dcd2dc | 219 | struct btrfs_ordered_inode_tree *tree; |
dc17ff8f | 220 | struct rb_node *node; |
e6dcd2dc CM |
221 | struct btrfs_ordered_extent *entry; |
222 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
223 | int ret; | |
224 | ||
225 | tree = &BTRFS_I(inode)->ordered_tree; | |
226 | mutex_lock(&tree->mutex); | |
227 | clear_extent_ordered(io_tree, file_offset, file_offset + io_size - 1, | |
228 | GFP_NOFS); | |
229 | node = tree_search(tree, file_offset); | |
dc17ff8f | 230 | if (!node) { |
e6dcd2dc CM |
231 | ret = 1; |
232 | goto out; | |
dc17ff8f CM |
233 | } |
234 | ||
e6dcd2dc CM |
235 | entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); |
236 | if (!offset_in_entry(entry, file_offset)) { | |
237 | ret = 1; | |
238 | goto out; | |
dc17ff8f | 239 | } |
e6dcd2dc CM |
240 | |
241 | ret = test_range_bit(io_tree, entry->file_offset, | |
242 | entry->file_offset + entry->len - 1, | |
243 | EXTENT_ORDERED, 0); | |
e6dcd2dc CM |
244 | if (ret == 0) |
245 | ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags); | |
246 | out: | |
247 | mutex_unlock(&tree->mutex); | |
248 | return ret == 0; | |
249 | } | |
dc17ff8f | 250 | |
eb84ae03 CM |
251 | /* |
252 | * used to drop a reference on an ordered extent. This will free | |
253 | * the extent if the last reference is dropped | |
254 | */ | |
e6dcd2dc CM |
255 | int btrfs_put_ordered_extent(struct btrfs_ordered_extent *entry) |
256 | { | |
ba1da2f4 CM |
257 | struct list_head *cur; |
258 | struct btrfs_ordered_sum *sum; | |
259 | ||
260 | if (atomic_dec_and_test(&entry->refs)) { | |
261 | while(!list_empty(&entry->list)) { | |
262 | cur = entry->list.next; | |
263 | sum = list_entry(cur, struct btrfs_ordered_sum, list); | |
264 | list_del(&sum->list); | |
265 | kfree(sum); | |
266 | } | |
e6dcd2dc | 267 | kfree(entry); |
ba1da2f4 | 268 | } |
e6dcd2dc | 269 | return 0; |
dc17ff8f | 270 | } |
cee36a03 | 271 | |
eb84ae03 CM |
272 | /* |
273 | * remove an ordered extent from the tree. No references are dropped | |
274 | * but, anyone waiting on this extent is woken up. | |
275 | */ | |
e6dcd2dc CM |
276 | int btrfs_remove_ordered_extent(struct inode *inode, |
277 | struct btrfs_ordered_extent *entry) | |
cee36a03 | 278 | { |
e6dcd2dc | 279 | struct btrfs_ordered_inode_tree *tree; |
cee36a03 | 280 | struct rb_node *node; |
cee36a03 | 281 | |
e6dcd2dc CM |
282 | tree = &BTRFS_I(inode)->ordered_tree; |
283 | mutex_lock(&tree->mutex); | |
284 | node = &entry->rb_node; | |
cee36a03 | 285 | rb_erase(node, &tree->tree); |
e6dcd2dc CM |
286 | tree->last = NULL; |
287 | set_bit(BTRFS_ORDERED_COMPLETE, &entry->flags); | |
288 | mutex_unlock(&tree->mutex); | |
289 | wake_up(&entry->wait); | |
290 | return 0; | |
cee36a03 CM |
291 | } |
292 | ||
eb84ae03 CM |
293 | /* |
294 | * Used to start IO or wait for a given ordered extent to finish. | |
295 | * | |
296 | * If wait is one, this effectively waits on page writeback for all the pages | |
297 | * in the extent, and it waits on the io completion code to insert | |
298 | * metadata into the btree corresponding to the extent | |
299 | */ | |
300 | void btrfs_start_ordered_extent(struct inode *inode, | |
301 | struct btrfs_ordered_extent *entry, | |
302 | int wait) | |
e6dcd2dc CM |
303 | { |
304 | u64 start = entry->file_offset; | |
305 | u64 end = start + entry->len - 1; | |
e1b81e67 | 306 | |
eb84ae03 CM |
307 | /* |
308 | * pages in the range can be dirty, clean or writeback. We | |
309 | * start IO on any dirty ones so the wait doesn't stall waiting | |
310 | * for pdflush to find them | |
311 | */ | |
f421950f | 312 | btrfs_fdatawrite_range(inode->i_mapping, start, end, WB_SYNC_NONE); |
e6dcd2dc CM |
313 | if (wait) |
314 | wait_event(entry->wait, test_bit(BTRFS_ORDERED_COMPLETE, | |
315 | &entry->flags)); | |
316 | } | |
cee36a03 | 317 | |
eb84ae03 CM |
318 | /* |
319 | * Used to wait on ordered extents across a large range of bytes. | |
320 | */ | |
e6dcd2dc CM |
321 | void btrfs_wait_ordered_range(struct inode *inode, u64 start, u64 len) |
322 | { | |
323 | u64 end; | |
e5a2217e CM |
324 | u64 orig_end; |
325 | u64 wait_end; | |
e6dcd2dc | 326 | struct btrfs_ordered_extent *ordered; |
e5a2217e CM |
327 | |
328 | if (start + len < start) { | |
f421950f | 329 | orig_end = INT_LIMIT(loff_t); |
e5a2217e CM |
330 | } else { |
331 | orig_end = start + len - 1; | |
f421950f CM |
332 | if (orig_end > INT_LIMIT(loff_t)) |
333 | orig_end = INT_LIMIT(loff_t); | |
e5a2217e | 334 | } |
f421950f | 335 | wait_end = orig_end; |
4a096752 | 336 | again: |
e5a2217e CM |
337 | /* start IO across the range first to instantiate any delalloc |
338 | * extents | |
339 | */ | |
f421950f CM |
340 | btrfs_fdatawrite_range(inode->i_mapping, start, orig_end, WB_SYNC_NONE); |
341 | ||
342 | btrfs_wait_on_page_writeback_range(inode->i_mapping, | |
343 | start >> PAGE_CACHE_SHIFT, | |
344 | orig_end >> PAGE_CACHE_SHIFT); | |
e5a2217e | 345 | |
f421950f | 346 | end = orig_end; |
e6dcd2dc CM |
347 | while(1) { |
348 | ordered = btrfs_lookup_first_ordered_extent(inode, end); | |
349 | if (!ordered) { | |
350 | break; | |
351 | } | |
e5a2217e | 352 | if (ordered->file_offset > orig_end) { |
e6dcd2dc CM |
353 | btrfs_put_ordered_extent(ordered); |
354 | break; | |
355 | } | |
356 | if (ordered->file_offset + ordered->len < start) { | |
357 | btrfs_put_ordered_extent(ordered); | |
358 | break; | |
359 | } | |
e5a2217e | 360 | btrfs_start_ordered_extent(inode, ordered, 1); |
e6dcd2dc CM |
361 | end = ordered->file_offset; |
362 | btrfs_put_ordered_extent(ordered); | |
e5a2217e | 363 | if (end == 0 || end == start) |
e6dcd2dc CM |
364 | break; |
365 | end--; | |
366 | } | |
4a096752 CM |
367 | if (test_range_bit(&BTRFS_I(inode)->io_tree, start, orig_end, |
368 | EXTENT_ORDERED | EXTENT_DELALLOC, 0)) { | |
369 | printk("inode %lu still ordered or delalloc after wait " | |
370 | "%llu %llu\n", inode->i_ino, | |
371 | (unsigned long long)start, | |
372 | (unsigned long long)orig_end); | |
373 | goto again; | |
374 | } | |
cee36a03 CM |
375 | } |
376 | ||
eb84ae03 CM |
377 | /* |
378 | * find an ordered extent corresponding to file_offset. return NULL if | |
379 | * nothing is found, otherwise take a reference on the extent and return it | |
380 | */ | |
e6dcd2dc CM |
381 | struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode, |
382 | u64 file_offset) | |
383 | { | |
384 | struct btrfs_ordered_inode_tree *tree; | |
385 | struct rb_node *node; | |
386 | struct btrfs_ordered_extent *entry = NULL; | |
387 | ||
388 | tree = &BTRFS_I(inode)->ordered_tree; | |
389 | mutex_lock(&tree->mutex); | |
390 | node = tree_search(tree, file_offset); | |
391 | if (!node) | |
392 | goto out; | |
393 | ||
394 | entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); | |
395 | if (!offset_in_entry(entry, file_offset)) | |
396 | entry = NULL; | |
397 | if (entry) | |
398 | atomic_inc(&entry->refs); | |
399 | out: | |
400 | mutex_unlock(&tree->mutex); | |
401 | return entry; | |
402 | } | |
403 | ||
eb84ae03 CM |
404 | /* |
405 | * lookup and return any extent before 'file_offset'. NULL is returned | |
406 | * if none is found | |
407 | */ | |
e6dcd2dc CM |
408 | struct btrfs_ordered_extent * |
409 | btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset) | |
410 | { | |
411 | struct btrfs_ordered_inode_tree *tree; | |
412 | struct rb_node *node; | |
413 | struct btrfs_ordered_extent *entry = NULL; | |
414 | ||
415 | tree = &BTRFS_I(inode)->ordered_tree; | |
416 | mutex_lock(&tree->mutex); | |
417 | node = tree_search(tree, file_offset); | |
418 | if (!node) | |
419 | goto out; | |
420 | ||
421 | entry = rb_entry(node, struct btrfs_ordered_extent, rb_node); | |
422 | atomic_inc(&entry->refs); | |
423 | out: | |
424 | mutex_unlock(&tree->mutex); | |
425 | return entry; | |
81d7ed29 | 426 | } |
dbe674a9 | 427 | |
eb84ae03 CM |
428 | /* |
429 | * After an extent is done, call this to conditionally update the on disk | |
430 | * i_size. i_size is updated to cover any fully written part of the file. | |
431 | */ | |
dbe674a9 CM |
432 | int btrfs_ordered_update_i_size(struct inode *inode, |
433 | struct btrfs_ordered_extent *ordered) | |
434 | { | |
435 | struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree; | |
436 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
437 | u64 disk_i_size; | |
438 | u64 new_i_size; | |
439 | u64 i_size_test; | |
440 | struct rb_node *node; | |
441 | struct btrfs_ordered_extent *test; | |
442 | ||
443 | mutex_lock(&tree->mutex); | |
444 | disk_i_size = BTRFS_I(inode)->disk_i_size; | |
445 | ||
446 | /* | |
447 | * if the disk i_size is already at the inode->i_size, or | |
448 | * this ordered extent is inside the disk i_size, we're done | |
449 | */ | |
450 | if (disk_i_size >= inode->i_size || | |
451 | ordered->file_offset + ordered->len <= disk_i_size) { | |
452 | goto out; | |
453 | } | |
454 | ||
455 | /* | |
456 | * we can't update the disk_isize if there are delalloc bytes | |
457 | * between disk_i_size and this ordered extent | |
458 | */ | |
459 | if (test_range_bit(io_tree, disk_i_size, | |
460 | ordered->file_offset + ordered->len - 1, | |
461 | EXTENT_DELALLOC, 0)) { | |
462 | goto out; | |
463 | } | |
464 | /* | |
465 | * walk backward from this ordered extent to disk_i_size. | |
466 | * if we find an ordered extent then we can't update disk i_size | |
467 | * yet | |
468 | */ | |
ba1da2f4 | 469 | node = &ordered->rb_node; |
dbe674a9 | 470 | while(1) { |
ba1da2f4 | 471 | node = rb_prev(node); |
dbe674a9 CM |
472 | if (!node) |
473 | break; | |
474 | test = rb_entry(node, struct btrfs_ordered_extent, rb_node); | |
475 | if (test->file_offset + test->len <= disk_i_size) | |
476 | break; | |
477 | if (test->file_offset >= inode->i_size) | |
478 | break; | |
479 | if (test->file_offset >= disk_i_size) | |
480 | goto out; | |
481 | } | |
482 | new_i_size = min_t(u64, entry_end(ordered), i_size_read(inode)); | |
483 | ||
484 | /* | |
485 | * at this point, we know we can safely update i_size to at least | |
486 | * the offset from this ordered extent. But, we need to | |
487 | * walk forward and see if ios from higher up in the file have | |
488 | * finished. | |
489 | */ | |
490 | node = rb_next(&ordered->rb_node); | |
491 | i_size_test = 0; | |
492 | if (node) { | |
493 | /* | |
494 | * do we have an area where IO might have finished | |
495 | * between our ordered extent and the next one. | |
496 | */ | |
497 | test = rb_entry(node, struct btrfs_ordered_extent, rb_node); | |
498 | if (test->file_offset > entry_end(ordered)) { | |
499 | i_size_test = test->file_offset - 1; | |
500 | } | |
501 | } else { | |
502 | i_size_test = i_size_read(inode); | |
503 | } | |
504 | ||
505 | /* | |
506 | * i_size_test is the end of a region after this ordered | |
507 | * extent where there are no ordered extents. As long as there | |
508 | * are no delalloc bytes in this area, it is safe to update | |
509 | * disk_i_size to the end of the region. | |
510 | */ | |
511 | if (i_size_test > entry_end(ordered) && | |
512 | !test_range_bit(io_tree, entry_end(ordered), i_size_test, | |
513 | EXTENT_DELALLOC, 0)) { | |
514 | new_i_size = min_t(u64, i_size_test, i_size_read(inode)); | |
515 | } | |
516 | BTRFS_I(inode)->disk_i_size = new_i_size; | |
517 | out: | |
518 | mutex_unlock(&tree->mutex); | |
519 | return 0; | |
520 | } | |
ba1da2f4 | 521 | |
eb84ae03 CM |
522 | /* |
523 | * search the ordered extents for one corresponding to 'offset' and | |
524 | * try to find a checksum. This is used because we allow pages to | |
525 | * be reclaimed before their checksum is actually put into the btree | |
526 | */ | |
ba1da2f4 CM |
527 | int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) |
528 | { | |
529 | struct btrfs_ordered_sum *ordered_sum; | |
530 | struct btrfs_sector_sum *sector_sums; | |
531 | struct btrfs_ordered_extent *ordered; | |
532 | struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree; | |
533 | struct list_head *cur; | |
3edf7d33 CM |
534 | unsigned long num_sectors; |
535 | unsigned long i; | |
536 | u32 sectorsize = BTRFS_I(inode)->root->sectorsize; | |
ba1da2f4 | 537 | int ret = 1; |
ba1da2f4 CM |
538 | |
539 | ordered = btrfs_lookup_ordered_extent(inode, offset); | |
540 | if (!ordered) | |
541 | return 1; | |
542 | ||
543 | mutex_lock(&tree->mutex); | |
544 | list_for_each_prev(cur, &ordered->list) { | |
545 | ordered_sum = list_entry(cur, struct btrfs_ordered_sum, list); | |
3edf7d33 CM |
546 | if (offset >= ordered_sum->file_offset) { |
547 | num_sectors = ordered_sum->len / sectorsize; | |
ed98b56a | 548 | sector_sums = ordered_sum->sums; |
3edf7d33 CM |
549 | for (i = 0; i < num_sectors; i++) { |
550 | if (sector_sums[i].offset == offset) { | |
3edf7d33 CM |
551 | *sum = sector_sums[i].sum; |
552 | ret = 0; | |
553 | goto out; | |
554 | } | |
555 | } | |
ba1da2f4 CM |
556 | } |
557 | } | |
558 | out: | |
559 | mutex_unlock(&tree->mutex); | |
89642229 | 560 | btrfs_put_ordered_extent(ordered); |
ba1da2f4 CM |
561 | return ret; |
562 | } | |
563 | ||
f421950f CM |
564 | |
565 | /** | |
566 | * taken from mm/filemap.c because it isn't exported | |
567 | * | |
568 | * __filemap_fdatawrite_range - start writeback on mapping dirty pages in range | |
569 | * @mapping: address space structure to write | |
570 | * @start: offset in bytes where the range starts | |
571 | * @end: offset in bytes where the range ends (inclusive) | |
572 | * @sync_mode: enable synchronous operation | |
573 | * | |
574 | * Start writeback against all of a mapping's dirty pages that lie | |
575 | * within the byte offsets <start, end> inclusive. | |
576 | * | |
577 | * If sync_mode is WB_SYNC_ALL then this is a "data integrity" operation, as | |
578 | * opposed to a regular memory cleansing writeback. The difference between | |
579 | * these two operations is that if a dirty page/buffer is encountered, it must | |
580 | * be waited upon, and not just skipped over. | |
581 | */ | |
582 | int btrfs_fdatawrite_range(struct address_space *mapping, loff_t start, | |
583 | loff_t end, int sync_mode) | |
584 | { | |
585 | struct writeback_control wbc = { | |
586 | .sync_mode = sync_mode, | |
587 | .nr_to_write = mapping->nrpages * 2, | |
588 | .range_start = start, | |
589 | .range_end = end, | |
590 | .for_writepages = 1, | |
591 | }; | |
592 | return btrfs_writepages(mapping, &wbc); | |
593 | } | |
594 | ||
595 | /** | |
596 | * taken from mm/filemap.c because it isn't exported | |
597 | * | |
598 | * wait_on_page_writeback_range - wait for writeback to complete | |
599 | * @mapping: target address_space | |
600 | * @start: beginning page index | |
601 | * @end: ending page index | |
602 | * | |
603 | * Wait for writeback to complete against pages indexed by start->end | |
604 | * inclusive | |
605 | */ | |
606 | int btrfs_wait_on_page_writeback_range(struct address_space *mapping, | |
607 | pgoff_t start, pgoff_t end) | |
608 | { | |
609 | struct pagevec pvec; | |
610 | int nr_pages; | |
611 | int ret = 0; | |
612 | pgoff_t index; | |
613 | ||
614 | if (end < start) | |
615 | return 0; | |
616 | ||
617 | pagevec_init(&pvec, 0); | |
618 | index = start; | |
619 | while ((index <= end) && | |
620 | (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, | |
621 | PAGECACHE_TAG_WRITEBACK, | |
622 | min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1)) != 0) { | |
623 | unsigned i; | |
624 | ||
625 | for (i = 0; i < nr_pages; i++) { | |
626 | struct page *page = pvec.pages[i]; | |
627 | ||
628 | /* until radix tree lookup accepts end_index */ | |
629 | if (page->index > end) | |
630 | continue; | |
631 | ||
632 | wait_on_page_writeback(page); | |
633 | if (PageError(page)) | |
634 | ret = -EIO; | |
635 | } | |
636 | pagevec_release(&pvec); | |
637 | cond_resched(); | |
638 | } | |
639 | ||
640 | /* Check for outstanding write errors */ | |
641 | if (test_and_clear_bit(AS_ENOSPC, &mapping->flags)) | |
642 | ret = -ENOSPC; | |
643 | if (test_and_clear_bit(AS_EIO, &mapping->flags)) | |
644 | ret = -EIO; | |
645 | ||
646 | return ret; | |
647 | } |