Commit | Line | Data |
---|---|---|
56bec294 CM |
1 | /* |
2 | * Copyright (C) 2009 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
19 | #include <linux/sched.h> | |
20 | #include <linux/sort.h> | |
56bec294 CM |
21 | #include "ctree.h" |
22 | #include "delayed-ref.h" | |
23 | #include "transaction.h" | |
24 | ||
25 | /* | |
26 | * delayed back reference update tracking. For subvolume trees | |
27 | * we queue up extent allocations and backref maintenance for | |
28 | * delayed processing. This avoids deep call chains where we | |
29 | * add extents in the middle of btrfs_search_slot, and it allows | |
30 | * us to buffer up frequently modified backrefs in an rb tree instead | |
31 | * of hammering updates on the extent allocation tree. | |
56bec294 CM |
32 | */ |
33 | ||
34 | /* | |
5d4f98a2 YZ |
35 | * compare two delayed tree backrefs with same bytenr and type |
36 | */ | |
37 | static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref2, | |
38 | struct btrfs_delayed_tree_ref *ref1) | |
39 | { | |
40 | if (ref1->node.type == BTRFS_TREE_BLOCK_REF_KEY) { | |
41 | if (ref1->root < ref2->root) | |
42 | return -1; | |
43 | if (ref1->root > ref2->root) | |
44 | return 1; | |
45 | } else { | |
46 | if (ref1->parent < ref2->parent) | |
47 | return -1; | |
48 | if (ref1->parent > ref2->parent) | |
49 | return 1; | |
50 | } | |
51 | return 0; | |
52 | } | |
53 | ||
54 | /* | |
55 | * compare two delayed data backrefs with same bytenr and type | |
56bec294 | 56 | */ |
5d4f98a2 YZ |
57 | static int comp_data_refs(struct btrfs_delayed_data_ref *ref2, |
58 | struct btrfs_delayed_data_ref *ref1) | |
56bec294 | 59 | { |
5d4f98a2 YZ |
60 | if (ref1->node.type == BTRFS_EXTENT_DATA_REF_KEY) { |
61 | if (ref1->root < ref2->root) | |
62 | return -1; | |
63 | if (ref1->root > ref2->root) | |
64 | return 1; | |
65 | if (ref1->objectid < ref2->objectid) | |
66 | return -1; | |
67 | if (ref1->objectid > ref2->objectid) | |
68 | return 1; | |
69 | if (ref1->offset < ref2->offset) | |
70 | return -1; | |
71 | if (ref1->offset > ref2->offset) | |
72 | return 1; | |
73 | } else { | |
74 | if (ref1->parent < ref2->parent) | |
75 | return -1; | |
76 | if (ref1->parent > ref2->parent) | |
77 | return 1; | |
78 | } | |
79 | return 0; | |
80 | } | |
81 | ||
82 | /* | |
83 | * entries in the rb tree are ordered by the byte number of the extent, | |
84 | * type of the delayed backrefs and content of delayed backrefs. | |
85 | */ | |
86 | static int comp_entry(struct btrfs_delayed_ref_node *ref2, | |
87 | struct btrfs_delayed_ref_node *ref1) | |
88 | { | |
89 | if (ref1->bytenr < ref2->bytenr) | |
56bec294 | 90 | return -1; |
5d4f98a2 | 91 | if (ref1->bytenr > ref2->bytenr) |
56bec294 | 92 | return 1; |
5d4f98a2 YZ |
93 | if (ref1->is_head && ref2->is_head) |
94 | return 0; | |
95 | if (ref2->is_head) | |
56bec294 | 96 | return -1; |
5d4f98a2 | 97 | if (ref1->is_head) |
56bec294 | 98 | return 1; |
5d4f98a2 YZ |
99 | if (ref1->type < ref2->type) |
100 | return -1; | |
101 | if (ref1->type > ref2->type) | |
102 | return 1; | |
103 | if (ref1->type == BTRFS_TREE_BLOCK_REF_KEY || | |
104 | ref1->type == BTRFS_SHARED_BLOCK_REF_KEY) { | |
105 | return comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref2), | |
106 | btrfs_delayed_node_to_tree_ref(ref1)); | |
107 | } else if (ref1->type == BTRFS_EXTENT_DATA_REF_KEY || | |
108 | ref1->type == BTRFS_SHARED_DATA_REF_KEY) { | |
109 | return comp_data_refs(btrfs_delayed_node_to_data_ref(ref2), | |
110 | btrfs_delayed_node_to_data_ref(ref1)); | |
111 | } | |
112 | BUG(); | |
56bec294 CM |
113 | return 0; |
114 | } | |
115 | ||
116 | /* | |
117 | * insert a new ref into the rbtree. This returns any existing refs | |
118 | * for the same (bytenr,parent) tuple, or NULL if the new node was properly | |
119 | * inserted. | |
120 | */ | |
121 | static struct btrfs_delayed_ref_node *tree_insert(struct rb_root *root, | |
56bec294 CM |
122 | struct rb_node *node) |
123 | { | |
124 | struct rb_node **p = &root->rb_node; | |
125 | struct rb_node *parent_node = NULL; | |
126 | struct btrfs_delayed_ref_node *entry; | |
5d4f98a2 | 127 | struct btrfs_delayed_ref_node *ins; |
56bec294 CM |
128 | int cmp; |
129 | ||
5d4f98a2 | 130 | ins = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); |
56bec294 CM |
131 | while (*p) { |
132 | parent_node = *p; | |
133 | entry = rb_entry(parent_node, struct btrfs_delayed_ref_node, | |
134 | rb_node); | |
135 | ||
5d4f98a2 | 136 | cmp = comp_entry(entry, ins); |
56bec294 CM |
137 | if (cmp < 0) |
138 | p = &(*p)->rb_left; | |
139 | else if (cmp > 0) | |
140 | p = &(*p)->rb_right; | |
141 | else | |
142 | return entry; | |
143 | } | |
144 | ||
56bec294 CM |
145 | rb_link_node(node, parent_node, p); |
146 | rb_insert_color(node, root); | |
147 | return NULL; | |
148 | } | |
149 | ||
150 | /* | |
5d4f98a2 YZ |
151 | * find an head entry based on bytenr. This returns the delayed ref |
152 | * head if it was able to find one, or NULL if nothing was in that spot | |
56bec294 | 153 | */ |
5d4f98a2 YZ |
154 | static struct btrfs_delayed_ref_node *find_ref_head(struct rb_root *root, |
155 | u64 bytenr, | |
c3e69d58 | 156 | struct btrfs_delayed_ref_node **last) |
56bec294 CM |
157 | { |
158 | struct rb_node *n = root->rb_node; | |
159 | struct btrfs_delayed_ref_node *entry; | |
160 | int cmp; | |
161 | ||
162 | while (n) { | |
163 | entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node); | |
164 | WARN_ON(!entry->in_tree); | |
c3e69d58 CM |
165 | if (last) |
166 | *last = entry; | |
56bec294 | 167 | |
5d4f98a2 YZ |
168 | if (bytenr < entry->bytenr) |
169 | cmp = -1; | |
170 | else if (bytenr > entry->bytenr) | |
171 | cmp = 1; | |
172 | else if (!btrfs_delayed_ref_is_head(entry)) | |
173 | cmp = 1; | |
174 | else | |
175 | cmp = 0; | |
176 | ||
56bec294 CM |
177 | if (cmp < 0) |
178 | n = n->rb_left; | |
179 | else if (cmp > 0) | |
180 | n = n->rb_right; | |
181 | else | |
182 | return entry; | |
183 | } | |
184 | return NULL; | |
185 | } | |
186 | ||
c3e69d58 CM |
187 | int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans, |
188 | struct btrfs_delayed_ref_head *head) | |
56bec294 | 189 | { |
c3e69d58 CM |
190 | struct btrfs_delayed_ref_root *delayed_refs; |
191 | ||
192 | delayed_refs = &trans->transaction->delayed_refs; | |
193 | assert_spin_locked(&delayed_refs->lock); | |
194 | if (mutex_trylock(&head->mutex)) | |
195 | return 0; | |
196 | ||
197 | atomic_inc(&head->node.refs); | |
198 | spin_unlock(&delayed_refs->lock); | |
199 | ||
200 | mutex_lock(&head->mutex); | |
201 | spin_lock(&delayed_refs->lock); | |
202 | if (!head->node.in_tree) { | |
203 | mutex_unlock(&head->mutex); | |
204 | btrfs_put_delayed_ref(&head->node); | |
205 | return -EAGAIN; | |
206 | } | |
207 | btrfs_put_delayed_ref(&head->node); | |
208 | return 0; | |
209 | } | |
210 | ||
211 | int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans, | |
212 | struct list_head *cluster, u64 start) | |
213 | { | |
214 | int count = 0; | |
215 | struct btrfs_delayed_ref_root *delayed_refs; | |
56bec294 | 216 | struct rb_node *node; |
c3e69d58 | 217 | struct btrfs_delayed_ref_node *ref; |
56bec294 | 218 | struct btrfs_delayed_ref_head *head; |
56bec294 | 219 | |
c3e69d58 CM |
220 | delayed_refs = &trans->transaction->delayed_refs; |
221 | if (start == 0) { | |
222 | node = rb_first(&delayed_refs->root); | |
223 | } else { | |
224 | ref = NULL; | |
5d4f98a2 | 225 | find_ref_head(&delayed_refs->root, start, &ref); |
c3e69d58 CM |
226 | if (ref) { |
227 | struct btrfs_delayed_ref_node *tmp; | |
228 | ||
229 | node = rb_prev(&ref->rb_node); | |
230 | while (node) { | |
231 | tmp = rb_entry(node, | |
232 | struct btrfs_delayed_ref_node, | |
233 | rb_node); | |
234 | if (tmp->bytenr < start) | |
235 | break; | |
236 | ref = tmp; | |
237 | node = rb_prev(&ref->rb_node); | |
238 | } | |
239 | node = &ref->rb_node; | |
240 | } else | |
241 | node = rb_first(&delayed_refs->root); | |
242 | } | |
243 | again: | |
244 | while (node && count < 32) { | |
245 | ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node); | |
56bec294 CM |
246 | if (btrfs_delayed_ref_is_head(ref)) { |
247 | head = btrfs_delayed_node_to_head(ref); | |
c3e69d58 CM |
248 | if (list_empty(&head->cluster)) { |
249 | list_add_tail(&head->cluster, cluster); | |
250 | delayed_refs->run_delayed_start = | |
251 | head->node.bytenr; | |
252 | count++; | |
253 | ||
254 | WARN_ON(delayed_refs->num_heads_ready == 0); | |
255 | delayed_refs->num_heads_ready--; | |
256 | } else if (count) { | |
257 | /* the goal of the clustering is to find extents | |
258 | * that are likely to end up in the same extent | |
259 | * leaf on disk. So, we don't want them spread | |
260 | * all over the tree. Stop now if we've hit | |
261 | * a head that was already in use | |
262 | */ | |
56bec294 CM |
263 | break; |
264 | } | |
265 | } | |
c3e69d58 | 266 | node = rb_next(node); |
56bec294 | 267 | } |
c3e69d58 CM |
268 | if (count) { |
269 | return 0; | |
270 | } else if (start) { | |
271 | /* | |
272 | * we've gone to the end of the rbtree without finding any | |
273 | * clusters. start from the beginning and try again | |
274 | */ | |
275 | start = 0; | |
276 | node = rb_first(&delayed_refs->root); | |
277 | goto again; | |
278 | } | |
279 | return 1; | |
56bec294 CM |
280 | } |
281 | ||
282 | /* | |
283 | * This checks to see if there are any delayed refs in the | |
284 | * btree for a given bytenr. It returns one if it finds any | |
285 | * and zero otherwise. | |
286 | * | |
287 | * If it only finds a head node, it returns 0. | |
288 | * | |
289 | * The idea is to use this when deciding if you can safely delete an | |
290 | * extent from the extent allocation tree. There may be a pending | |
291 | * ref in the rbtree that adds or removes references, so as long as this | |
292 | * returns one you need to leave the BTRFS_EXTENT_ITEM in the extent | |
293 | * allocation tree. | |
294 | */ | |
295 | int btrfs_delayed_ref_pending(struct btrfs_trans_handle *trans, u64 bytenr) | |
296 | { | |
297 | struct btrfs_delayed_ref_node *ref; | |
298 | struct btrfs_delayed_ref_root *delayed_refs; | |
299 | struct rb_node *prev_node; | |
300 | int ret = 0; | |
301 | ||
302 | delayed_refs = &trans->transaction->delayed_refs; | |
303 | spin_lock(&delayed_refs->lock); | |
304 | ||
5d4f98a2 | 305 | ref = find_ref_head(&delayed_refs->root, bytenr, NULL); |
56bec294 CM |
306 | if (ref) { |
307 | prev_node = rb_prev(&ref->rb_node); | |
308 | if (!prev_node) | |
309 | goto out; | |
310 | ref = rb_entry(prev_node, struct btrfs_delayed_ref_node, | |
311 | rb_node); | |
312 | if (ref->bytenr == bytenr) | |
313 | ret = 1; | |
314 | } | |
315 | out: | |
316 | spin_unlock(&delayed_refs->lock); | |
317 | return ret; | |
318 | } | |
319 | ||
320 | /* | |
5d4f98a2 | 321 | * helper function to lookup reference count and flags of extent. |
56bec294 CM |
322 | * |
323 | * the head node for delayed ref is used to store the sum of all the | |
5d4f98a2 YZ |
324 | * reference count modifications queued up in the rbtree. the head |
325 | * node may also store the extent flags to set. This way you can check | |
326 | * to see what the reference count and extent flags would be if all of | |
327 | * the delayed refs are not processed. | |
56bec294 | 328 | */ |
5d4f98a2 YZ |
329 | int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans, |
330 | struct btrfs_root *root, u64 bytenr, | |
331 | u64 num_bytes, u64 *refs, u64 *flags) | |
56bec294 CM |
332 | { |
333 | struct btrfs_delayed_ref_node *ref; | |
334 | struct btrfs_delayed_ref_head *head; | |
335 | struct btrfs_delayed_ref_root *delayed_refs; | |
336 | struct btrfs_path *path; | |
56bec294 | 337 | struct btrfs_extent_item *ei; |
5d4f98a2 | 338 | struct extent_buffer *leaf; |
56bec294 | 339 | struct btrfs_key key; |
5d4f98a2 YZ |
340 | u32 item_size; |
341 | u64 num_refs; | |
342 | u64 extent_flags; | |
56bec294 CM |
343 | int ret; |
344 | ||
345 | path = btrfs_alloc_path(); | |
346 | if (!path) | |
347 | return -ENOMEM; | |
348 | ||
349 | key.objectid = bytenr; | |
350 | key.type = BTRFS_EXTENT_ITEM_KEY; | |
351 | key.offset = num_bytes; | |
352 | delayed_refs = &trans->transaction->delayed_refs; | |
353 | again: | |
354 | ret = btrfs_search_slot(trans, root->fs_info->extent_root, | |
355 | &key, path, 0, 0); | |
356 | if (ret < 0) | |
357 | goto out; | |
358 | ||
359 | if (ret == 0) { | |
360 | leaf = path->nodes[0]; | |
5d4f98a2 YZ |
361 | item_size = btrfs_item_size_nr(leaf, path->slots[0]); |
362 | if (item_size >= sizeof(*ei)) { | |
363 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
364 | struct btrfs_extent_item); | |
365 | num_refs = btrfs_extent_refs(leaf, ei); | |
366 | extent_flags = btrfs_extent_flags(leaf, ei); | |
367 | } else { | |
368 | #ifdef BTRFS_COMPAT_EXTENT_TREE_V0 | |
369 | struct btrfs_extent_item_v0 *ei0; | |
370 | BUG_ON(item_size != sizeof(*ei0)); | |
371 | ei0 = btrfs_item_ptr(leaf, path->slots[0], | |
372 | struct btrfs_extent_item_v0); | |
373 | num_refs = btrfs_extent_refs_v0(leaf, ei0); | |
374 | /* FIXME: this isn't correct for data */ | |
375 | extent_flags = BTRFS_BLOCK_FLAG_FULL_BACKREF; | |
376 | #else | |
377 | BUG(); | |
378 | #endif | |
379 | } | |
380 | BUG_ON(num_refs == 0); | |
56bec294 CM |
381 | } else { |
382 | num_refs = 0; | |
5d4f98a2 | 383 | extent_flags = 0; |
56bec294 CM |
384 | ret = 0; |
385 | } | |
386 | ||
387 | spin_lock(&delayed_refs->lock); | |
5d4f98a2 | 388 | ref = find_ref_head(&delayed_refs->root, bytenr, NULL); |
56bec294 CM |
389 | if (ref) { |
390 | head = btrfs_delayed_node_to_head(ref); | |
5d4f98a2 YZ |
391 | if (!mutex_trylock(&head->mutex)) { |
392 | atomic_inc(&ref->refs); | |
393 | spin_unlock(&delayed_refs->lock); | |
56bec294 | 394 | |
5d4f98a2 | 395 | btrfs_release_path(root->fs_info->extent_root, path); |
56bec294 | 396 | |
5d4f98a2 YZ |
397 | mutex_lock(&head->mutex); |
398 | mutex_unlock(&head->mutex); | |
399 | btrfs_put_delayed_ref(ref); | |
400 | goto again; | |
401 | } | |
402 | if (head->extent_op && head->extent_op->update_flags) | |
403 | extent_flags |= head->extent_op->flags_to_set; | |
404 | else | |
405 | BUG_ON(num_refs == 0); | |
56bec294 | 406 | |
5d4f98a2 | 407 | num_refs += ref->ref_mod; |
56bec294 | 408 | mutex_unlock(&head->mutex); |
56bec294 | 409 | } |
5d4f98a2 YZ |
410 | WARN_ON(num_refs == 0); |
411 | if (refs) | |
412 | *refs = num_refs; | |
413 | if (flags) | |
414 | *flags = extent_flags; | |
56bec294 CM |
415 | out: |
416 | spin_unlock(&delayed_refs->lock); | |
417 | btrfs_free_path(path); | |
418 | return ret; | |
419 | } | |
420 | ||
421 | /* | |
422 | * helper function to update an extent delayed ref in the | |
423 | * rbtree. existing and update must both have the same | |
424 | * bytenr and parent | |
425 | * | |
426 | * This may free existing if the update cancels out whatever | |
427 | * operation it was doing. | |
428 | */ | |
429 | static noinline void | |
430 | update_existing_ref(struct btrfs_trans_handle *trans, | |
431 | struct btrfs_delayed_ref_root *delayed_refs, | |
432 | struct btrfs_delayed_ref_node *existing, | |
433 | struct btrfs_delayed_ref_node *update) | |
434 | { | |
5d4f98a2 | 435 | if (update->action != existing->action) { |
56bec294 CM |
436 | /* |
437 | * this is effectively undoing either an add or a | |
438 | * drop. We decrement the ref_mod, and if it goes | |
439 | * down to zero we just delete the entry without | |
440 | * every changing the extent allocation tree. | |
441 | */ | |
442 | existing->ref_mod--; | |
443 | if (existing->ref_mod == 0) { | |
444 | rb_erase(&existing->rb_node, | |
445 | &delayed_refs->root); | |
446 | existing->in_tree = 0; | |
447 | btrfs_put_delayed_ref(existing); | |
448 | delayed_refs->num_entries--; | |
449 | if (trans->delayed_ref_updates) | |
450 | trans->delayed_ref_updates--; | |
5d4f98a2 YZ |
451 | } else { |
452 | WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY || | |
453 | existing->type == BTRFS_SHARED_BLOCK_REF_KEY); | |
56bec294 CM |
454 | } |
455 | } else { | |
5d4f98a2 YZ |
456 | WARN_ON(existing->type == BTRFS_TREE_BLOCK_REF_KEY || |
457 | existing->type == BTRFS_SHARED_BLOCK_REF_KEY); | |
56bec294 CM |
458 | /* |
459 | * the action on the existing ref matches | |
460 | * the action on the ref we're trying to add. | |
461 | * Bump the ref_mod by one so the backref that | |
462 | * is eventually added/removed has the correct | |
463 | * reference count | |
464 | */ | |
465 | existing->ref_mod += update->ref_mod; | |
466 | } | |
467 | } | |
468 | ||
469 | /* | |
470 | * helper function to update the accounting in the head ref | |
471 | * existing and update must have the same bytenr | |
472 | */ | |
473 | static noinline void | |
474 | update_existing_head_ref(struct btrfs_delayed_ref_node *existing, | |
475 | struct btrfs_delayed_ref_node *update) | |
476 | { | |
477 | struct btrfs_delayed_ref_head *existing_ref; | |
478 | struct btrfs_delayed_ref_head *ref; | |
479 | ||
480 | existing_ref = btrfs_delayed_node_to_head(existing); | |
481 | ref = btrfs_delayed_node_to_head(update); | |
5d4f98a2 | 482 | BUG_ON(existing_ref->is_data != ref->is_data); |
56bec294 CM |
483 | |
484 | if (ref->must_insert_reserved) { | |
485 | /* if the extent was freed and then | |
486 | * reallocated before the delayed ref | |
487 | * entries were processed, we can end up | |
488 | * with an existing head ref without | |
489 | * the must_insert_reserved flag set. | |
490 | * Set it again here | |
491 | */ | |
492 | existing_ref->must_insert_reserved = ref->must_insert_reserved; | |
493 | ||
494 | /* | |
495 | * update the num_bytes so we make sure the accounting | |
496 | * is done correctly | |
497 | */ | |
498 | existing->num_bytes = update->num_bytes; | |
499 | ||
500 | } | |
501 | ||
5d4f98a2 YZ |
502 | if (ref->extent_op) { |
503 | if (!existing_ref->extent_op) { | |
504 | existing_ref->extent_op = ref->extent_op; | |
505 | } else { | |
506 | if (ref->extent_op->update_key) { | |
507 | memcpy(&existing_ref->extent_op->key, | |
508 | &ref->extent_op->key, | |
509 | sizeof(ref->extent_op->key)); | |
510 | existing_ref->extent_op->update_key = 1; | |
511 | } | |
512 | if (ref->extent_op->update_flags) { | |
513 | existing_ref->extent_op->flags_to_set |= | |
514 | ref->extent_op->flags_to_set; | |
515 | existing_ref->extent_op->update_flags = 1; | |
516 | } | |
517 | kfree(ref->extent_op); | |
518 | } | |
519 | } | |
56bec294 CM |
520 | /* |
521 | * update the reference mod on the head to reflect this new operation | |
522 | */ | |
523 | existing->ref_mod += update->ref_mod; | |
524 | } | |
525 | ||
526 | /* | |
5d4f98a2 | 527 | * helper function to actually insert a head node into the rbtree. |
56bec294 | 528 | * this does all the dirty work in terms of maintaining the correct |
5d4f98a2 | 529 | * overall modification count. |
56bec294 | 530 | */ |
5d4f98a2 YZ |
531 | static noinline int add_delayed_ref_head(struct btrfs_trans_handle *trans, |
532 | struct btrfs_delayed_ref_node *ref, | |
533 | u64 bytenr, u64 num_bytes, | |
534 | int action, int is_data) | |
56bec294 CM |
535 | { |
536 | struct btrfs_delayed_ref_node *existing; | |
c3e69d58 | 537 | struct btrfs_delayed_ref_head *head_ref = NULL; |
56bec294 CM |
538 | struct btrfs_delayed_ref_root *delayed_refs; |
539 | int count_mod = 1; | |
540 | int must_insert_reserved = 0; | |
541 | ||
542 | /* | |
543 | * the head node stores the sum of all the mods, so dropping a ref | |
544 | * should drop the sum in the head node by one. | |
545 | */ | |
5d4f98a2 YZ |
546 | if (action == BTRFS_UPDATE_DELAYED_HEAD) |
547 | count_mod = 0; | |
548 | else if (action == BTRFS_DROP_DELAYED_REF) | |
549 | count_mod = -1; | |
56bec294 CM |
550 | |
551 | /* | |
552 | * BTRFS_ADD_DELAYED_EXTENT means that we need to update | |
553 | * the reserved accounting when the extent is finally added, or | |
554 | * if a later modification deletes the delayed ref without ever | |
555 | * inserting the extent into the extent allocation tree. | |
556 | * ref->must_insert_reserved is the flag used to record | |
557 | * that accounting mods are required. | |
558 | * | |
559 | * Once we record must_insert_reserved, switch the action to | |
560 | * BTRFS_ADD_DELAYED_REF because other special casing is not required. | |
561 | */ | |
5d4f98a2 | 562 | if (action == BTRFS_ADD_DELAYED_EXTENT) |
56bec294 | 563 | must_insert_reserved = 1; |
5d4f98a2 | 564 | else |
56bec294 | 565 | must_insert_reserved = 0; |
56bec294 CM |
566 | |
567 | delayed_refs = &trans->transaction->delayed_refs; | |
568 | ||
569 | /* first set the basic ref node struct up */ | |
570 | atomic_set(&ref->refs, 1); | |
571 | ref->bytenr = bytenr; | |
5d4f98a2 | 572 | ref->num_bytes = num_bytes; |
56bec294 | 573 | ref->ref_mod = count_mod; |
5d4f98a2 YZ |
574 | ref->type = 0; |
575 | ref->action = 0; | |
576 | ref->is_head = 1; | |
56bec294 | 577 | ref->in_tree = 1; |
5d4f98a2 YZ |
578 | |
579 | head_ref = btrfs_delayed_node_to_head(ref); | |
580 | head_ref->must_insert_reserved = must_insert_reserved; | |
581 | head_ref->is_data = is_data; | |
582 | ||
583 | INIT_LIST_HEAD(&head_ref->cluster); | |
584 | mutex_init(&head_ref->mutex); | |
585 | ||
586 | existing = tree_insert(&delayed_refs->root, &ref->rb_node); | |
587 | ||
588 | if (existing) { | |
589 | update_existing_head_ref(existing, ref); | |
590 | /* | |
591 | * we've updated the existing ref, free the newly | |
592 | * allocated ref | |
593 | */ | |
594 | kfree(ref); | |
595 | } else { | |
596 | delayed_refs->num_heads++; | |
597 | delayed_refs->num_heads_ready++; | |
598 | delayed_refs->num_entries++; | |
599 | trans->delayed_ref_updates++; | |
600 | } | |
601 | return 0; | |
602 | } | |
603 | ||
604 | /* | |
605 | * helper to insert a delayed tree ref into the rbtree. | |
606 | */ | |
607 | static noinline int add_delayed_tree_ref(struct btrfs_trans_handle *trans, | |
608 | struct btrfs_delayed_ref_node *ref, | |
609 | u64 bytenr, u64 num_bytes, u64 parent, | |
610 | u64 ref_root, int level, int action) | |
611 | { | |
612 | struct btrfs_delayed_ref_node *existing; | |
613 | struct btrfs_delayed_tree_ref *full_ref; | |
614 | struct btrfs_delayed_ref_root *delayed_refs; | |
615 | ||
616 | if (action == BTRFS_ADD_DELAYED_EXTENT) | |
617 | action = BTRFS_ADD_DELAYED_REF; | |
618 | ||
619 | delayed_refs = &trans->transaction->delayed_refs; | |
620 | ||
621 | /* first set the basic ref node struct up */ | |
622 | atomic_set(&ref->refs, 1); | |
623 | ref->bytenr = bytenr; | |
56bec294 | 624 | ref->num_bytes = num_bytes; |
5d4f98a2 YZ |
625 | ref->ref_mod = 1; |
626 | ref->action = action; | |
627 | ref->is_head = 0; | |
628 | ref->in_tree = 1; | |
56bec294 | 629 | |
5d4f98a2 YZ |
630 | full_ref = btrfs_delayed_node_to_tree_ref(ref); |
631 | if (parent) { | |
632 | full_ref->parent = parent; | |
633 | ref->type = BTRFS_SHARED_BLOCK_REF_KEY; | |
56bec294 | 634 | } else { |
56bec294 | 635 | full_ref->root = ref_root; |
5d4f98a2 | 636 | ref->type = BTRFS_TREE_BLOCK_REF_KEY; |
56bec294 | 637 | } |
5d4f98a2 | 638 | full_ref->level = level; |
56bec294 | 639 | |
5d4f98a2 | 640 | existing = tree_insert(&delayed_refs->root, &ref->rb_node); |
56bec294 CM |
641 | |
642 | if (existing) { | |
5d4f98a2 YZ |
643 | update_existing_ref(trans, delayed_refs, existing, ref); |
644 | /* | |
645 | * we've updated the existing ref, free the newly | |
646 | * allocated ref | |
647 | */ | |
648 | kfree(ref); | |
649 | } else { | |
650 | delayed_refs->num_entries++; | |
651 | trans->delayed_ref_updates++; | |
652 | } | |
653 | return 0; | |
654 | } | |
655 | ||
656 | /* | |
657 | * helper to insert a delayed data ref into the rbtree. | |
658 | */ | |
659 | static noinline int add_delayed_data_ref(struct btrfs_trans_handle *trans, | |
660 | struct btrfs_delayed_ref_node *ref, | |
661 | u64 bytenr, u64 num_bytes, u64 parent, | |
662 | u64 ref_root, u64 owner, u64 offset, | |
663 | int action) | |
664 | { | |
665 | struct btrfs_delayed_ref_node *existing; | |
666 | struct btrfs_delayed_data_ref *full_ref; | |
667 | struct btrfs_delayed_ref_root *delayed_refs; | |
668 | ||
669 | if (action == BTRFS_ADD_DELAYED_EXTENT) | |
670 | action = BTRFS_ADD_DELAYED_REF; | |
671 | ||
672 | delayed_refs = &trans->transaction->delayed_refs; | |
673 | ||
674 | /* first set the basic ref node struct up */ | |
675 | atomic_set(&ref->refs, 1); | |
676 | ref->bytenr = bytenr; | |
677 | ref->num_bytes = num_bytes; | |
678 | ref->ref_mod = 1; | |
679 | ref->action = action; | |
680 | ref->is_head = 0; | |
681 | ref->in_tree = 1; | |
682 | ||
683 | full_ref = btrfs_delayed_node_to_data_ref(ref); | |
684 | if (parent) { | |
685 | full_ref->parent = parent; | |
686 | ref->type = BTRFS_SHARED_DATA_REF_KEY; | |
687 | } else { | |
688 | full_ref->root = ref_root; | |
689 | ref->type = BTRFS_EXTENT_DATA_REF_KEY; | |
690 | } | |
691 | full_ref->objectid = owner; | |
692 | full_ref->offset = offset; | |
56bec294 | 693 | |
5d4f98a2 YZ |
694 | existing = tree_insert(&delayed_refs->root, &ref->rb_node); |
695 | ||
696 | if (existing) { | |
697 | update_existing_ref(trans, delayed_refs, existing, ref); | |
56bec294 CM |
698 | /* |
699 | * we've updated the existing ref, free the newly | |
700 | * allocated ref | |
701 | */ | |
702 | kfree(ref); | |
703 | } else { | |
704 | delayed_refs->num_entries++; | |
705 | trans->delayed_ref_updates++; | |
706 | } | |
707 | return 0; | |
708 | } | |
709 | ||
710 | /* | |
5d4f98a2 | 711 | * add a delayed tree ref. This does all of the accounting required |
56bec294 CM |
712 | * to make sure the delayed ref is eventually processed before this |
713 | * transaction commits. | |
714 | */ | |
5d4f98a2 YZ |
715 | int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans, |
716 | u64 bytenr, u64 num_bytes, u64 parent, | |
717 | u64 ref_root, int level, int action, | |
718 | struct btrfs_delayed_extent_op *extent_op) | |
56bec294 | 719 | { |
5d4f98a2 | 720 | struct btrfs_delayed_tree_ref *ref; |
56bec294 CM |
721 | struct btrfs_delayed_ref_head *head_ref; |
722 | struct btrfs_delayed_ref_root *delayed_refs; | |
723 | int ret; | |
724 | ||
5d4f98a2 | 725 | BUG_ON(extent_op && extent_op->is_data); |
56bec294 CM |
726 | ref = kmalloc(sizeof(*ref), GFP_NOFS); |
727 | if (!ref) | |
728 | return -ENOMEM; | |
729 | ||
5d4f98a2 YZ |
730 | head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS); |
731 | if (!head_ref) { | |
732 | kfree(ref); | |
733 | return -ENOMEM; | |
734 | } | |
735 | ||
736 | head_ref->extent_op = extent_op; | |
737 | ||
738 | delayed_refs = &trans->transaction->delayed_refs; | |
739 | spin_lock(&delayed_refs->lock); | |
740 | ||
56bec294 | 741 | /* |
5d4f98a2 YZ |
742 | * insert both the head node and the new ref without dropping |
743 | * the spin lock | |
56bec294 | 744 | */ |
5d4f98a2 YZ |
745 | ret = add_delayed_ref_head(trans, &head_ref->node, bytenr, num_bytes, |
746 | action, 0); | |
747 | BUG_ON(ret); | |
748 | ||
749 | ret = add_delayed_tree_ref(trans, &ref->node, bytenr, num_bytes, | |
750 | parent, ref_root, level, action); | |
751 | BUG_ON(ret); | |
752 | spin_unlock(&delayed_refs->lock); | |
753 | return 0; | |
754 | } | |
755 | ||
756 | /* | |
757 | * add a delayed data ref. it's similar to btrfs_add_delayed_tree_ref. | |
758 | */ | |
759 | int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans, | |
760 | u64 bytenr, u64 num_bytes, | |
761 | u64 parent, u64 ref_root, | |
762 | u64 owner, u64 offset, int action, | |
763 | struct btrfs_delayed_extent_op *extent_op) | |
764 | { | |
765 | struct btrfs_delayed_data_ref *ref; | |
766 | struct btrfs_delayed_ref_head *head_ref; | |
767 | struct btrfs_delayed_ref_root *delayed_refs; | |
768 | int ret; | |
769 | ||
770 | BUG_ON(extent_op && !extent_op->is_data); | |
771 | ref = kmalloc(sizeof(*ref), GFP_NOFS); | |
772 | if (!ref) | |
773 | return -ENOMEM; | |
56bec294 CM |
774 | |
775 | head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS); | |
776 | if (!head_ref) { | |
777 | kfree(ref); | |
778 | return -ENOMEM; | |
779 | } | |
5d4f98a2 YZ |
780 | |
781 | head_ref->extent_op = extent_op; | |
782 | ||
56bec294 CM |
783 | delayed_refs = &trans->transaction->delayed_refs; |
784 | spin_lock(&delayed_refs->lock); | |
785 | ||
786 | /* | |
787 | * insert both the head node and the new ref without dropping | |
788 | * the spin lock | |
789 | */ | |
5d4f98a2 YZ |
790 | ret = add_delayed_ref_head(trans, &head_ref->node, bytenr, num_bytes, |
791 | action, 1); | |
56bec294 CM |
792 | BUG_ON(ret); |
793 | ||
5d4f98a2 YZ |
794 | ret = add_delayed_data_ref(trans, &ref->node, bytenr, num_bytes, |
795 | parent, ref_root, owner, offset, action); | |
796 | BUG_ON(ret); | |
797 | spin_unlock(&delayed_refs->lock); | |
798 | return 0; | |
799 | } | |
800 | ||
801 | int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans, | |
802 | u64 bytenr, u64 num_bytes, | |
803 | struct btrfs_delayed_extent_op *extent_op) | |
804 | { | |
805 | struct btrfs_delayed_ref_head *head_ref; | |
806 | struct btrfs_delayed_ref_root *delayed_refs; | |
807 | int ret; | |
808 | ||
809 | head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS); | |
810 | if (!head_ref) | |
811 | return -ENOMEM; | |
812 | ||
813 | head_ref->extent_op = extent_op; | |
814 | ||
815 | delayed_refs = &trans->transaction->delayed_refs; | |
816 | spin_lock(&delayed_refs->lock); | |
817 | ||
818 | ret = add_delayed_ref_head(trans, &head_ref->node, bytenr, | |
819 | num_bytes, BTRFS_UPDATE_DELAYED_HEAD, | |
820 | extent_op->is_data); | |
56bec294 | 821 | BUG_ON(ret); |
5d4f98a2 | 822 | |
56bec294 CM |
823 | spin_unlock(&delayed_refs->lock); |
824 | return 0; | |
825 | } | |
826 | ||
1887be66 CM |
827 | /* |
828 | * this does a simple search for the head node for a given extent. | |
829 | * It must be called with the delayed ref spinlock held, and it returns | |
830 | * the head node if any where found, or NULL if not. | |
831 | */ | |
832 | struct btrfs_delayed_ref_head * | |
833 | btrfs_find_delayed_ref_head(struct btrfs_trans_handle *trans, u64 bytenr) | |
834 | { | |
835 | struct btrfs_delayed_ref_node *ref; | |
836 | struct btrfs_delayed_ref_root *delayed_refs; | |
837 | ||
838 | delayed_refs = &trans->transaction->delayed_refs; | |
5d4f98a2 | 839 | ref = find_ref_head(&delayed_refs->root, bytenr, NULL); |
1887be66 CM |
840 | if (ref) |
841 | return btrfs_delayed_node_to_head(ref); | |
842 | return NULL; | |
843 | } | |
844 | ||
56bec294 CM |
845 | /* |
846 | * add a delayed ref to the tree. This does all of the accounting required | |
847 | * to make sure the delayed ref is eventually processed before this | |
848 | * transaction commits. | |
849 | * | |
850 | * The main point of this call is to add and remove a backreference in a single | |
851 | * shot, taking the lock only once, and only searching for the head node once. | |
852 | * | |
853 | * It is the same as doing a ref add and delete in two separate calls. | |
854 | */ | |
5d4f98a2 | 855 | #if 0 |
56bec294 CM |
856 | int btrfs_update_delayed_ref(struct btrfs_trans_handle *trans, |
857 | u64 bytenr, u64 num_bytes, u64 orig_parent, | |
858 | u64 parent, u64 orig_ref_root, u64 ref_root, | |
859 | u64 orig_ref_generation, u64 ref_generation, | |
860 | u64 owner_objectid, int pin) | |
861 | { | |
862 | struct btrfs_delayed_ref *ref; | |
863 | struct btrfs_delayed_ref *old_ref; | |
864 | struct btrfs_delayed_ref_head *head_ref; | |
865 | struct btrfs_delayed_ref_root *delayed_refs; | |
866 | int ret; | |
867 | ||
868 | ref = kmalloc(sizeof(*ref), GFP_NOFS); | |
869 | if (!ref) | |
870 | return -ENOMEM; | |
871 | ||
872 | old_ref = kmalloc(sizeof(*old_ref), GFP_NOFS); | |
873 | if (!old_ref) { | |
874 | kfree(ref); | |
875 | return -ENOMEM; | |
876 | } | |
877 | ||
878 | /* | |
879 | * the parent = 0 case comes from cases where we don't actually | |
880 | * know the parent yet. It will get updated later via a add/drop | |
881 | * pair. | |
882 | */ | |
883 | if (parent == 0) | |
884 | parent = bytenr; | |
885 | if (orig_parent == 0) | |
886 | orig_parent = bytenr; | |
887 | ||
888 | head_ref = kmalloc(sizeof(*head_ref), GFP_NOFS); | |
889 | if (!head_ref) { | |
890 | kfree(ref); | |
891 | kfree(old_ref); | |
892 | return -ENOMEM; | |
893 | } | |
894 | delayed_refs = &trans->transaction->delayed_refs; | |
895 | spin_lock(&delayed_refs->lock); | |
896 | ||
897 | /* | |
898 | * insert both the head node and the new ref without dropping | |
899 | * the spin lock | |
900 | */ | |
901 | ret = __btrfs_add_delayed_ref(trans, &head_ref->node, bytenr, num_bytes, | |
902 | (u64)-1, 0, 0, 0, | |
1a81af4d | 903 | BTRFS_UPDATE_DELAYED_HEAD, 0); |
56bec294 CM |
904 | BUG_ON(ret); |
905 | ||
906 | ret = __btrfs_add_delayed_ref(trans, &ref->node, bytenr, num_bytes, | |
907 | parent, ref_root, ref_generation, | |
908 | owner_objectid, BTRFS_ADD_DELAYED_REF, 0); | |
909 | BUG_ON(ret); | |
910 | ||
911 | ret = __btrfs_add_delayed_ref(trans, &old_ref->node, bytenr, num_bytes, | |
912 | orig_parent, orig_ref_root, | |
913 | orig_ref_generation, owner_objectid, | |
914 | BTRFS_DROP_DELAYED_REF, pin); | |
915 | BUG_ON(ret); | |
916 | spin_unlock(&delayed_refs->lock); | |
917 | return 0; | |
918 | } | |
5d4f98a2 | 919 | #endif |