?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
spi-topcliff-pch: Fix issue for transmitting over 4KByte
[zen-stable.git] / fs / btrfs / free-space-cache.c
blob710ea380c7edbdbeb4fbabc1266745f047606909
1 /*
2 * Copyright (C) 2008 Red Hat. 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/pagemap.h>
20 #include <linux/sched.h>
21 #include <linux/slab.h>
22 #include <linux/math64.h>
23 #include <linux/ratelimit.h>
24 #include "ctree.h"
25 #include "free-space-cache.h"
26 #include "transaction.h"
27 #include "disk-io.h"
28 #include "extent_io.h"
29 #include "inode-map.h"
30
31 #define BITS_PER_BITMAP (PAGE_CACHE_SIZE * 8)
32 #define MAX_CACHE_BYTES_PER_GIG (32 * 1024)
33
34 static int link_free_space(struct btrfs_free_space_ctl *ctl,
35 struct btrfs_free_space *info);
36
37 static struct inode *__lookup_free_space_inode(struct btrfs_root *root,
38 struct btrfs_path *path,
39 u64 offset)
40 {
41 struct btrfs_key key;
42 struct btrfs_key location;
43 struct btrfs_disk_key disk_key;
44 struct btrfs_free_space_header *header;
45 struct extent_buffer *leaf;
46 struct inode *inode = NULL;
47 int ret;
48
49 key.objectid = BTRFS_FREE_SPACE_OBJECTID;
50 key.offset = offset;
51 key.type = 0;
52
53 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
54 if (ret < 0)
55 return ERR_PTR(ret);
56 if (ret > 0) {
57 btrfs_release_path(path);
58 return ERR_PTR(-ENOENT);
59 }
60
61 leaf = path->nodes[0];
62 header = btrfs_item_ptr(leaf, path->slots[0],
63 struct btrfs_free_space_header);
64 btrfs_free_space_key(leaf, header, &disk_key);
65 btrfs_disk_key_to_cpu(&location, &disk_key);
66 btrfs_release_path(path);
67
68 inode = btrfs_iget(root->fs_info->sb, &location, root, NULL);
69 if (!inode)
70 return ERR_PTR(-ENOENT);
71 if (IS_ERR(inode))
72 return inode;
73 if (is_bad_inode(inode)) {
74 iput(inode);
75 return ERR_PTR(-ENOENT);
76 }
77
78 inode->i_mapping->flags &= ~__GFP_FS;
79
80 return inode;
81 }
82
83 struct inode *lookup_free_space_inode(struct btrfs_root *root,
84 struct btrfs_block_group_cache
85 *block_group, struct btrfs_path *path)
86 {
87 struct inode *inode = NULL;
88 u32 flags = BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
89
90 spin_lock(&block_group->lock);
91 if (block_group->inode)
92 inode = igrab(block_group->inode);
93 spin_unlock(&block_group->lock);
94 if (inode)
95 return inode;
96
97 inode = __lookup_free_space_inode(root, path,
98 block_group->key.objectid);
99 if (IS_ERR(inode))
100 return inode;
101
102 spin_lock(&block_group->lock);
103 if (!((BTRFS_I(inode)->flags & flags) == flags)) {
104 printk(KERN_INFO "Old style space inode found, converting.\n");
105 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM |
106 BTRFS_INODE_NODATACOW;
107 block_group->disk_cache_state = BTRFS_DC_CLEAR;
108 }
109
110 if (!block_group->iref) {
111 block_group->inode = igrab(inode);
112 block_group->iref = 1;
113 }
114 spin_unlock(&block_group->lock);
115
116 return inode;
117 }
118
119 int __create_free_space_inode(struct btrfs_root *root,
120 struct btrfs_trans_handle *trans,
121 struct btrfs_path *path, u64 ino, u64 offset)
122 {
123 struct btrfs_key key;
124 struct btrfs_disk_key disk_key;
125 struct btrfs_free_space_header *header;
126 struct btrfs_inode_item *inode_item;
127 struct extent_buffer *leaf;
128 u64 flags = BTRFS_INODE_NOCOMPRESS | BTRFS_INODE_PREALLOC;
129 int ret;
130
131 ret = btrfs_insert_empty_inode(trans, root, path, ino);
132 if (ret)
133 return ret;
134
135 /* We inline crc's for the free disk space cache */
136 if (ino != BTRFS_FREE_INO_OBJECTID)
137 flags |= BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
138
139 leaf = path->nodes[0];
140 inode_item = btrfs_item_ptr(leaf, path->slots[0],
141 struct btrfs_inode_item);
142 btrfs_item_key(leaf, &disk_key, path->slots[0]);
143 memset_extent_buffer(leaf, 0, (unsigned long)inode_item,
144 sizeof(*inode_item));
145 btrfs_set_inode_generation(leaf, inode_item, trans->transid);
146 btrfs_set_inode_size(leaf, inode_item, 0);
147 btrfs_set_inode_nbytes(leaf, inode_item, 0);
148 btrfs_set_inode_uid(leaf, inode_item, 0);
149 btrfs_set_inode_gid(leaf, inode_item, 0);
150 btrfs_set_inode_mode(leaf, inode_item, S_IFREG | 0600);
151 btrfs_set_inode_flags(leaf, inode_item, flags);
152 btrfs_set_inode_nlink(leaf, inode_item, 1);
153 btrfs_set_inode_transid(leaf, inode_item, trans->transid);
154 btrfs_set_inode_block_group(leaf, inode_item, offset);
155 btrfs_mark_buffer_dirty(leaf);
156 btrfs_release_path(path);
157
158 key.objectid = BTRFS_FREE_SPACE_OBJECTID;
159 key.offset = offset;
160 key.type = 0;
161
162 ret = btrfs_insert_empty_item(trans, root, path, &key,
163 sizeof(struct btrfs_free_space_header));
164 if (ret < 0) {
165 btrfs_release_path(path);
166 return ret;
167 }
168 leaf = path->nodes[0];
169 header = btrfs_item_ptr(leaf, path->slots[0],
170 struct btrfs_free_space_header);
171 memset_extent_buffer(leaf, 0, (unsigned long)header, sizeof(*header));
172 btrfs_set_free_space_key(leaf, header, &disk_key);
173 btrfs_mark_buffer_dirty(leaf);
174 btrfs_release_path(path);
175
176 return 0;
177 }
178
179 int create_free_space_inode(struct btrfs_root *root,
180 struct btrfs_trans_handle *trans,
181 struct btrfs_block_group_cache *block_group,
182 struct btrfs_path *path)
183 {
184 int ret;
185 u64 ino;
186
187 ret = btrfs_find_free_objectid(root, &ino);
188 if (ret < 0)
189 return ret;
190
191 return __create_free_space_inode(root, trans, path, ino,
192 block_group->key.objectid);
193 }
194
195 int btrfs_truncate_free_space_cache(struct btrfs_root *root,
196 struct btrfs_trans_handle *trans,
197 struct btrfs_path *path,
198 struct inode *inode)
199 {
200 struct btrfs_block_rsv *rsv;
201 u64 needed_bytes;
202 loff_t oldsize;
203 int ret = 0;
204
205 rsv = trans->block_rsv;
206 trans->block_rsv = &root->fs_info->global_block_rsv;
207
208 /* 1 for slack space, 1 for updating the inode */
209 needed_bytes = btrfs_calc_trunc_metadata_size(root, 1) +
210 btrfs_calc_trans_metadata_size(root, 1);
211
212 spin_lock(&trans->block_rsv->lock);
213 if (trans->block_rsv->reserved < needed_bytes) {
214 spin_unlock(&trans->block_rsv->lock);
215 trans->block_rsv = rsv;
216 return -ENOSPC;
217 }
218 spin_unlock(&trans->block_rsv->lock);
219
220 oldsize = i_size_read(inode);
221 btrfs_i_size_write(inode, 0);
222 truncate_pagecache(inode, oldsize, 0);
223
224 /*
225 * We don't need an orphan item because truncating the free space cache
226 * will never be split across transactions.
227 */
228 ret = btrfs_truncate_inode_items(trans, root, inode,
229 0, BTRFS_EXTENT_DATA_KEY);
230
231 if (ret) {
232 trans->block_rsv = rsv;
233 WARN_ON(1);
234 return ret;
235 }
236
237 ret = btrfs_update_inode(trans, root, inode);
238 trans->block_rsv = rsv;
239
240 return ret;
241 }
242
243 static int readahead_cache(struct inode *inode)
244 {
245 struct file_ra_state *ra;
246 unsigned long last_index;
247
248 ra = kzalloc(sizeof(*ra), GFP_NOFS);
249 if (!ra)
250 return -ENOMEM;
251
252 file_ra_state_init(ra, inode->i_mapping);
253 last_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;
254
255 page_cache_sync_readahead(inode->i_mapping, ra, NULL, 0, last_index);
256
257 kfree(ra);
258
259 return 0;
260 }
261
262 struct io_ctl {
263 void *cur, *orig;
264 struct page *page;
265 struct page **pages;
266 struct btrfs_root *root;
267 unsigned long size;
268 int index;
269 int num_pages;
270 unsigned check_crcs:1;
271 };
272
273 static int io_ctl_init(struct io_ctl *io_ctl, struct inode *inode,
274 struct btrfs_root *root)
275 {
276 memset(io_ctl, 0, sizeof(struct io_ctl));
277 io_ctl->num_pages = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
278 PAGE_CACHE_SHIFT;
279 io_ctl->pages = kzalloc(sizeof(struct page *) * io_ctl->num_pages,
280 GFP_NOFS);
281 if (!io_ctl->pages)
282 return -ENOMEM;
283 io_ctl->root = root;
284 if (btrfs_ino(inode) != BTRFS_FREE_INO_OBJECTID)
285 io_ctl->check_crcs = 1;
286 return 0;
287 }
288
289 static void io_ctl_free(struct io_ctl *io_ctl)
290 {
291 kfree(io_ctl->pages);
292 }
293
294 static void io_ctl_unmap_page(struct io_ctl *io_ctl)
295 {
296 if (io_ctl->cur) {
297 kunmap(io_ctl->page);
298 io_ctl->cur = NULL;
299 io_ctl->orig = NULL;
300 }
301 }
302
303 static void io_ctl_map_page(struct io_ctl *io_ctl, int clear)
304 {
305 WARN_ON(io_ctl->cur);
306 BUG_ON(io_ctl->index >= io_ctl->num_pages);
307 io_ctl->page = io_ctl->pages[io_ctl->index++];
308 io_ctl->cur = kmap(io_ctl->page);
309 io_ctl->orig = io_ctl->cur;
310 io_ctl->size = PAGE_CACHE_SIZE;
311 if (clear)
312 memset(io_ctl->cur, 0, PAGE_CACHE_SIZE);
313 }
314
315 static void io_ctl_drop_pages(struct io_ctl *io_ctl)
316 {
317 int i;
318
319 io_ctl_unmap_page(io_ctl);
320
321 for (i = 0; i < io_ctl->num_pages; i++) {
322 if (io_ctl->pages[i]) {
323 ClearPageChecked(io_ctl->pages[i]);
324 unlock_page(io_ctl->pages[i]);
325 page_cache_release(io_ctl->pages[i]);
326 }
327 }
328 }
329
330 static int io_ctl_prepare_pages(struct io_ctl *io_ctl, struct inode *inode,
331 int uptodate)
332 {
333 struct page *page;
334 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
335 int i;
336
337 for (i = 0; i < io_ctl->num_pages; i++) {
338 page = find_or_create_page(inode->i_mapping, i, mask);
339 if (!page) {
340 io_ctl_drop_pages(io_ctl);
341 return -ENOMEM;
342 }
343 io_ctl->pages[i] = page;
344 if (uptodate && !PageUptodate(page)) {
345 btrfs_readpage(NULL, page);
346 lock_page(page);
347 if (!PageUptodate(page)) {
348 printk(KERN_ERR "btrfs: error reading free "
349 "space cache\n");
350 io_ctl_drop_pages(io_ctl);
351 return -EIO;
352 }
353 }
354 }
355
356 for (i = 0; i < io_ctl->num_pages; i++) {
357 clear_page_dirty_for_io(io_ctl->pages[i]);
358 set_page_extent_mapped(io_ctl->pages[i]);
359 }
360
361 return 0;
362 }
363
364 static void io_ctl_set_generation(struct io_ctl *io_ctl, u64 generation)
365 {
366 u64 *val;
367
368 io_ctl_map_page(io_ctl, 1);
369
370 /*
371 * Skip the csum areas. If we don't check crcs then we just have a
372 * 64bit chunk at the front of the first page.
373 */
374 if (io_ctl->check_crcs) {
375 io_ctl->cur += (sizeof(u32) * io_ctl->num_pages);
376 io_ctl->size -= sizeof(u64) + (sizeof(u32) * io_ctl->num_pages);
377 } else {
378 io_ctl->cur += sizeof(u64);
379 io_ctl->size -= sizeof(u64) * 2;
380 }
381
382 val = io_ctl->cur;
383 *val = cpu_to_le64(generation);
384 io_ctl->cur += sizeof(u64);
385 }
386
387 static int io_ctl_check_generation(struct io_ctl *io_ctl, u64 generation)
388 {
389 u64 *gen;
390
391 /*
392 * Skip the crc area. If we don't check crcs then we just have a 64bit
393 * chunk at the front of the first page.
394 */
395 if (io_ctl->check_crcs) {
396 io_ctl->cur += sizeof(u32) * io_ctl->num_pages;
397 io_ctl->size -= sizeof(u64) +
398 (sizeof(u32) * io_ctl->num_pages);
399 } else {
400 io_ctl->cur += sizeof(u64);
401 io_ctl->size -= sizeof(u64) * 2;
402 }
403
404 gen = io_ctl->cur;
405 if (le64_to_cpu(*gen) != generation) {
406 printk_ratelimited(KERN_ERR "btrfs: space cache generation "
407 "(%Lu) does not match inode (%Lu)\n", *gen,
408 generation);
409 io_ctl_unmap_page(io_ctl);
410 return -EIO;
411 }
412 io_ctl->cur += sizeof(u64);
413 return 0;
414 }
415
416 static void io_ctl_set_crc(struct io_ctl *io_ctl, int index)
417 {
418 u32 *tmp;
419 u32 crc = ~(u32)0;
420 unsigned offset = 0;
421
422 if (!io_ctl->check_crcs) {
423 io_ctl_unmap_page(io_ctl);
424 return;
425 }
426
427 if (index == 0)
428 offset = sizeof(u32) * io_ctl->num_pages;
429
430 crc = btrfs_csum_data(io_ctl->root, io_ctl->orig + offset, crc,
431 PAGE_CACHE_SIZE - offset);
432 btrfs_csum_final(crc, (char *)&crc);
433 io_ctl_unmap_page(io_ctl);
434 tmp = kmap(io_ctl->pages[0]);
435 tmp += index;
436 *tmp = crc;
437 kunmap(io_ctl->pages[0]);
438 }
439
440 static int io_ctl_check_crc(struct io_ctl *io_ctl, int index)
441 {
442 u32 *tmp, val;
443 u32 crc = ~(u32)0;
444 unsigned offset = 0;
445
446 if (!io_ctl->check_crcs) {
447 io_ctl_map_page(io_ctl, 0);
448 return 0;
449 }
450
451 if (index == 0)
452 offset = sizeof(u32) * io_ctl->num_pages;
453
454 tmp = kmap(io_ctl->pages[0]);
455 tmp += index;
456 val = *tmp;
457 kunmap(io_ctl->pages[0]);
458
459 io_ctl_map_page(io_ctl, 0);
460 crc = btrfs_csum_data(io_ctl->root, io_ctl->orig + offset, crc,
461 PAGE_CACHE_SIZE - offset);
462 btrfs_csum_final(crc, (char *)&crc);
463 if (val != crc) {
464 printk_ratelimited(KERN_ERR "btrfs: csum mismatch on free "
465 "space cache\n");
466 io_ctl_unmap_page(io_ctl);
467 return -EIO;
468 }
469
470 return 0;
471 }
472
473 static int io_ctl_add_entry(struct io_ctl *io_ctl, u64 offset, u64 bytes,
474 void *bitmap)
475 {
476 struct btrfs_free_space_entry *entry;
477
478 if (!io_ctl->cur)
479 return -ENOSPC;
480
481 entry = io_ctl->cur;
482 entry->offset = cpu_to_le64(offset);
483 entry->bytes = cpu_to_le64(bytes);
484 entry->type = (bitmap) ? BTRFS_FREE_SPACE_BITMAP :
485 BTRFS_FREE_SPACE_EXTENT;
486 io_ctl->cur += sizeof(struct btrfs_free_space_entry);
487 io_ctl->size -= sizeof(struct btrfs_free_space_entry);
488
489 if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
490 return 0;
491
492 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
493
494 /* No more pages to map */
495 if (io_ctl->index >= io_ctl->num_pages)
496 return 0;
497
498 /* map the next page */
499 io_ctl_map_page(io_ctl, 1);
500 return 0;
501 }
502
503 static int io_ctl_add_bitmap(struct io_ctl *io_ctl, void *bitmap)
504 {
505 if (!io_ctl->cur)
506 return -ENOSPC;
507
508 /*
509 * If we aren't at the start of the current page, unmap this one and
510 * map the next one if there is any left.
511 */
512 if (io_ctl->cur != io_ctl->orig) {
513 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
514 if (io_ctl->index >= io_ctl->num_pages)
515 return -ENOSPC;
516 io_ctl_map_page(io_ctl, 0);
517 }
518
519 memcpy(io_ctl->cur, bitmap, PAGE_CACHE_SIZE);
520 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
521 if (io_ctl->index < io_ctl->num_pages)
522 io_ctl_map_page(io_ctl, 0);
523 return 0;
524 }
525
526 static void io_ctl_zero_remaining_pages(struct io_ctl *io_ctl)
527 {
528 /*
529 * If we're not on the boundary we know we've modified the page and we
530 * need to crc the page.
531 */
532 if (io_ctl->cur != io_ctl->orig)
533 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
534 else
535 io_ctl_unmap_page(io_ctl);
536
537 while (io_ctl->index < io_ctl->num_pages) {
538 io_ctl_map_page(io_ctl, 1);
539 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
540 }
541 }
542
543 static int io_ctl_read_entry(struct io_ctl *io_ctl,
544 struct btrfs_free_space *entry, u8 *type)
545 {
546 struct btrfs_free_space_entry *e;
547 int ret;
548
549 if (!io_ctl->cur) {
550 ret = io_ctl_check_crc(io_ctl, io_ctl->index);
551 if (ret)
552 return ret;
553 }
554
555 e = io_ctl->cur;
556 entry->offset = le64_to_cpu(e->offset);
557 entry->bytes = le64_to_cpu(e->bytes);
558 *type = e->type;
559 io_ctl->cur += sizeof(struct btrfs_free_space_entry);
560 io_ctl->size -= sizeof(struct btrfs_free_space_entry);
561
562 if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
563 return 0;
564
565 io_ctl_unmap_page(io_ctl);
566
567 return 0;
568 }
569
570 static int io_ctl_read_bitmap(struct io_ctl *io_ctl,
571 struct btrfs_free_space *entry)
572 {
573 int ret;
574
575 ret = io_ctl_check_crc(io_ctl, io_ctl->index);
576 if (ret)
577 return ret;
578
579 memcpy(entry->bitmap, io_ctl->cur, PAGE_CACHE_SIZE);
580 io_ctl_unmap_page(io_ctl);
581
582 return 0;
583 }
584
585 int __load_free_space_cache(struct btrfs_root *root, struct inode *inode,
586 struct btrfs_free_space_ctl *ctl,
587 struct btrfs_path *path, u64 offset)
588 {
589 struct btrfs_free_space_header *header;
590 struct extent_buffer *leaf;
591 struct io_ctl io_ctl;
592 struct btrfs_key key;
593 struct btrfs_free_space *e, *n;
594 struct list_head bitmaps;
595 u64 num_entries;
596 u64 num_bitmaps;
597 u64 generation;
598 u8 type;
599 int ret = 0;
600
601 INIT_LIST_HEAD(&bitmaps);
602
603 /* Nothing in the space cache, goodbye */
604 if (!i_size_read(inode))
605 return 0;
606
607 key.objectid = BTRFS_FREE_SPACE_OBJECTID;
608 key.offset = offset;
609 key.type = 0;
610
611 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
612 if (ret < 0)
613 return 0;
614 else if (ret > 0) {
615 btrfs_release_path(path);
616 return 0;
617 }
618
619 ret = -1;
620
621 leaf = path->nodes[0];
622 header = btrfs_item_ptr(leaf, path->slots[0],
623 struct btrfs_free_space_header);
624 num_entries = btrfs_free_space_entries(leaf, header);
625 num_bitmaps = btrfs_free_space_bitmaps(leaf, header);
626 generation = btrfs_free_space_generation(leaf, header);
627 btrfs_release_path(path);
628
629 if (BTRFS_I(inode)->generation != generation) {
630 printk(KERN_ERR "btrfs: free space inode generation (%llu) did"
631 " not match free space cache generation (%llu)\n",
632 (unsigned long long)BTRFS_I(inode)->generation,
633 (unsigned long long)generation);
634 return 0;
635 }
636
637 if (!num_entries)
638 return 0;
639
640 ret = io_ctl_init(&io_ctl, inode, root);
641 if (ret)
642 return ret;
643
644 ret = readahead_cache(inode);
645 if (ret)
646 goto out;
647
648 ret = io_ctl_prepare_pages(&io_ctl, inode, 1);
649 if (ret)
650 goto out;
651
652 ret = io_ctl_check_crc(&io_ctl, 0);
653 if (ret)
654 goto free_cache;
655
656 ret = io_ctl_check_generation(&io_ctl, generation);
657 if (ret)
658 goto free_cache;
659
660 while (num_entries) {
661 e = kmem_cache_zalloc(btrfs_free_space_cachep,
662 GFP_NOFS);
663 if (!e)
664 goto free_cache;
665
666 ret = io_ctl_read_entry(&io_ctl, e, &type);
667 if (ret) {
668 kmem_cache_free(btrfs_free_space_cachep, e);
669 goto free_cache;
670 }
671
672 if (!e->bytes) {
673 kmem_cache_free(btrfs_free_space_cachep, e);
674 goto free_cache;
675 }
676
677 if (type == BTRFS_FREE_SPACE_EXTENT) {
678 spin_lock(&ctl->tree_lock);
679 ret = link_free_space(ctl, e);
680 spin_unlock(&ctl->tree_lock);
681 if (ret) {
682 printk(KERN_ERR "Duplicate entries in "
683 "free space cache, dumping\n");
684 kmem_cache_free(btrfs_free_space_cachep, e);
685 goto free_cache;
686 }
687 } else {
688 BUG_ON(!num_bitmaps);
689 num_bitmaps--;
690 e->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
691 if (!e->bitmap) {
692 kmem_cache_free(
693 btrfs_free_space_cachep, e);
694 goto free_cache;
695 }
696 spin_lock(&ctl->tree_lock);
697 ret = link_free_space(ctl, e);
698 ctl->total_bitmaps++;
699 ctl->op->recalc_thresholds(ctl);
700 spin_unlock(&ctl->tree_lock);
701 if (ret) {
702 printk(KERN_ERR "Duplicate entries in "
703 "free space cache, dumping\n");
704 kmem_cache_free(btrfs_free_space_cachep, e);
705 goto free_cache;
706 }
707 list_add_tail(&e->list, &bitmaps);
708 }
709
710 num_entries--;
711 }
712
713 io_ctl_unmap_page(&io_ctl);
714
715 /*
716 * We add the bitmaps at the end of the entries in order that
717 * the bitmap entries are added to the cache.
718 */
719 list_for_each_entry_safe(e, n, &bitmaps, list) {
720 list_del_init(&e->list);
721 ret = io_ctl_read_bitmap(&io_ctl, e);
722 if (ret)
723 goto free_cache;
724 }
725
726 io_ctl_drop_pages(&io_ctl);
727 ret = 1;
728 out:
729 io_ctl_free(&io_ctl);
730 return ret;
731 free_cache:
732 io_ctl_drop_pages(&io_ctl);
733 __btrfs_remove_free_space_cache(ctl);
734 goto out;
735 }
736
737 int load_free_space_cache(struct btrfs_fs_info *fs_info,
738 struct btrfs_block_group_cache *block_group)
739 {
740 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
741 struct btrfs_root *root = fs_info->tree_root;
742 struct inode *inode;
743 struct btrfs_path *path;
744 int ret = 0;
745 bool matched;
746 u64 used = btrfs_block_group_used(&block_group->item);
747
748 /*
749 * If we're unmounting then just return, since this does a search on the
750 * normal root and not the commit root and we could deadlock.
751 */
752 if (btrfs_fs_closing(fs_info))
753 return 0;
754
755 /*
756 * If this block group has been marked to be cleared for one reason or
757 * another then we can't trust the on disk cache, so just return.
758 */
759 spin_lock(&block_group->lock);
760 if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
761 spin_unlock(&block_group->lock);
762 return 0;
763 }
764 spin_unlock(&block_group->lock);
765
766 path = btrfs_alloc_path();
767 if (!path)
768 return 0;
769
770 inode = lookup_free_space_inode(root, block_group, path);
771 if (IS_ERR(inode)) {
772 btrfs_free_path(path);
773 return 0;
774 }
775
776 /* We may have converted the inode and made the cache invalid. */
777 spin_lock(&block_group->lock);
778 if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
779 spin_unlock(&block_group->lock);
780 btrfs_free_path(path);
781 goto out;
782 }
783 spin_unlock(&block_group->lock);
784
785 ret = __load_free_space_cache(fs_info->tree_root, inode, ctl,
786 path, block_group->key.objectid);
787 btrfs_free_path(path);
788 if (ret <= 0)
789 goto out;
790
791 spin_lock(&ctl->tree_lock);
792 matched = (ctl->free_space == (block_group->key.offset - used -
793 block_group->bytes_super));
794 spin_unlock(&ctl->tree_lock);
795
796 if (!matched) {
797 __btrfs_remove_free_space_cache(ctl);
798 printk(KERN_ERR "block group %llu has an wrong amount of free "
799 "space\n", block_group->key.objectid);
800 ret = -1;
801 }
802 out:
803 if (ret < 0) {
804 /* This cache is bogus, make sure it gets cleared */
805 spin_lock(&block_group->lock);
806 block_group->disk_cache_state = BTRFS_DC_CLEAR;
807 spin_unlock(&block_group->lock);
808 ret = 0;
809
810 printk(KERN_ERR "btrfs: failed to load free space cache "
811 "for block group %llu\n", block_group->key.objectid);
812 }
813
814 iput(inode);
815 return ret;
816 }
817
818 /**
819 * __btrfs_write_out_cache - write out cached info to an inode
820 * @root - the root the inode belongs to
821 * @ctl - the free space cache we are going to write out
822 * @block_group - the block_group for this cache if it belongs to a block_group
823 * @trans - the trans handle
824 * @path - the path to use
825 * @offset - the offset for the key we'll insert
826 *
827 * This function writes out a free space cache struct to disk for quick recovery
828 * on mount. This will return 0 if it was successfull in writing the cache out,
829 * and -1 if it was not.
830 */
831 int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode,
832 struct btrfs_free_space_ctl *ctl,
833 struct btrfs_block_group_cache *block_group,
834 struct btrfs_trans_handle *trans,
835 struct btrfs_path *path, u64 offset)
836 {
837 struct btrfs_free_space_header *header;
838 struct extent_buffer *leaf;
839 struct rb_node *node;
840 struct list_head *pos, *n;
841 struct extent_state *cached_state = NULL;
842 struct btrfs_free_cluster *cluster = NULL;
843 struct extent_io_tree *unpin = NULL;
844 struct io_ctl io_ctl;
845 struct list_head bitmap_list;
846 struct btrfs_key key;
847 u64 start, extent_start, extent_end, len;
848 int entries = 0;
849 int bitmaps = 0;
850 int ret;
851 int err = -1;
852
853 INIT_LIST_HEAD(&bitmap_list);
854
855 if (!i_size_read(inode))
856 return -1;
857
858 ret = io_ctl_init(&io_ctl, inode, root);
859 if (ret)
860 return -1;
861
862 /* Get the cluster for this block_group if it exists */
863 if (block_group && !list_empty(&block_group->cluster_list))
864 cluster = list_entry(block_group->cluster_list.next,
865 struct btrfs_free_cluster,
866 block_group_list);
867
868 /* Lock all pages first so we can lock the extent safely. */
869 io_ctl_prepare_pages(&io_ctl, inode, 0);
870
871 lock_extent_bits(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
872 0, &cached_state, GFP_NOFS);
873
874 node = rb_first(&ctl->free_space_offset);
875 if (!node && cluster) {
876 node = rb_first(&cluster->root);
877 cluster = NULL;
878 }
879
880 /* Make sure we can fit our crcs into the first page */
881 if (io_ctl.check_crcs &&
882 (io_ctl.num_pages * sizeof(u32)) >= PAGE_CACHE_SIZE) {
883 WARN_ON(1);
884 goto out_nospc;
885 }
886
887 io_ctl_set_generation(&io_ctl, trans->transid);
888
889 /* Write out the extent entries */
890 while (node) {
891 struct btrfs_free_space *e;
892
893 e = rb_entry(node, struct btrfs_free_space, offset_index);
894 entries++;
895
896 ret = io_ctl_add_entry(&io_ctl, e->offset, e->bytes,
897 e->bitmap);
898 if (ret)
899 goto out_nospc;
900
901 if (e->bitmap) {
902 list_add_tail(&e->list, &bitmap_list);
903 bitmaps++;
904 }
905 node = rb_next(node);
906 if (!node && cluster) {
907 node = rb_first(&cluster->root);
908 cluster = NULL;
909 }
910 }
911
912 /*
913 * We want to add any pinned extents to our free space cache
914 * so we don't leak the space
915 */
916
917 /*
918 * We shouldn't have switched the pinned extents yet so this is the
919 * right one
920 */
921 unpin = root->fs_info->pinned_extents;
922
923 if (block_group)
924 start = block_group->key.objectid;
925
926 while (block_group && (start < block_group->key.objectid +
927 block_group->key.offset)) {
928 ret = find_first_extent_bit(unpin, start,
929 &extent_start, &extent_end,
930 EXTENT_DIRTY);
931 if (ret) {
932 ret = 0;
933 break;
934 }
935
936 /* This pinned extent is out of our range */
937 if (extent_start >= block_group->key.objectid +
938 block_group->key.offset)
939 break;
940
941 extent_start = max(extent_start, start);
942 extent_end = min(block_group->key.objectid +
943 block_group->key.offset, extent_end + 1);
944 len = extent_end - extent_start;
945
946 entries++;
947 ret = io_ctl_add_entry(&io_ctl, extent_start, len, NULL);
948 if (ret)
949 goto out_nospc;
950
951 start = extent_end;
952 }
953
954 /* Write out the bitmaps */
955 list_for_each_safe(pos, n, &bitmap_list) {
956 struct btrfs_free_space *entry =
957 list_entry(pos, struct btrfs_free_space, list);
958
959 ret = io_ctl_add_bitmap(&io_ctl, entry->bitmap);
960 if (ret)
961 goto out_nospc;
962 list_del_init(&entry->list);
963 }
964
965 /* Zero out the rest of the pages just to make sure */
966 io_ctl_zero_remaining_pages(&io_ctl);
967
968 ret = btrfs_dirty_pages(root, inode, io_ctl.pages, io_ctl.num_pages,
969 0, i_size_read(inode), &cached_state);
970 io_ctl_drop_pages(&io_ctl);
971 unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
972 i_size_read(inode) - 1, &cached_state, GFP_NOFS);
973
974 if (ret)
975 goto out;
976
977
978 ret = filemap_write_and_wait(inode->i_mapping);
979 if (ret)
980 goto out;
981
982 key.objectid = BTRFS_FREE_SPACE_OBJECTID;
983 key.offset = offset;
984 key.type = 0;
985
986 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
987 if (ret < 0) {
988 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
989 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL,
990 GFP_NOFS);
991 goto out;
992 }
993 leaf = path->nodes[0];
994 if (ret > 0) {
995 struct btrfs_key found_key;
996 BUG_ON(!path->slots[0]);
997 path->slots[0]--;
998 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
999 if (found_key.objectid != BTRFS_FREE_SPACE_OBJECTID ||
1000 found_key.offset != offset) {
1001 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0,
1002 inode->i_size - 1,
1003 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0,
1004 NULL, GFP_NOFS);
1005 btrfs_release_path(path);
1006 goto out;
1007 }
1008 }
1009
1010 BTRFS_I(inode)->generation = trans->transid;
1011 header = btrfs_item_ptr(leaf, path->slots[0],
1012 struct btrfs_free_space_header);
1013 btrfs_set_free_space_entries(leaf, header, entries);
1014 btrfs_set_free_space_bitmaps(leaf, header, bitmaps);
1015 btrfs_set_free_space_generation(leaf, header, trans->transid);
1016 btrfs_mark_buffer_dirty(leaf);
1017 btrfs_release_path(path);
1018
1019 err = 0;
1020 out:
1021 io_ctl_free(&io_ctl);
1022 if (err) {
1023 invalidate_inode_pages2(inode->i_mapping);
1024 BTRFS_I(inode)->generation = 0;
1025 }
1026 btrfs_update_inode(trans, root, inode);
1027 return err;
1028
1029 out_nospc:
1030 list_for_each_safe(pos, n, &bitmap_list) {
1031 struct btrfs_free_space *entry =
1032 list_entry(pos, struct btrfs_free_space, list);
1033 list_del_init(&entry->list);
1034 }
1035 io_ctl_drop_pages(&io_ctl);
1036 unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
1037 i_size_read(inode) - 1, &cached_state, GFP_NOFS);
1038 goto out;
1039 }
1040
1041 int btrfs_write_out_cache(struct btrfs_root *root,
1042 struct btrfs_trans_handle *trans,
1043 struct btrfs_block_group_cache *block_group,
1044 struct btrfs_path *path)
1045 {
1046 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1047 struct inode *inode;
1048 int ret = 0;
1049
1050 root = root->fs_info->tree_root;
1051
1052 spin_lock(&block_group->lock);
1053 if (block_group->disk_cache_state < BTRFS_DC_SETUP) {
1054 spin_unlock(&block_group->lock);
1055 return 0;
1056 }
1057 spin_unlock(&block_group->lock);
1058
1059 inode = lookup_free_space_inode(root, block_group, path);
1060 if (IS_ERR(inode))
1061 return 0;
1062
1063 ret = __btrfs_write_out_cache(root, inode, ctl, block_group, trans,
1064 path, block_group->key.objectid);
1065 if (ret) {
1066 spin_lock(&block_group->lock);
1067 block_group->disk_cache_state = BTRFS_DC_ERROR;
1068 spin_unlock(&block_group->lock);
1069 ret = 0;
1070 #ifdef DEBUG
1071 printk(KERN_ERR "btrfs: failed to write free space cace "
1072 "for block group %llu\n", block_group->key.objectid);
1073 #endif
1074 }
1075
1076 iput(inode);
1077 return ret;
1078 }
1079
1080 static inline unsigned long offset_to_bit(u64 bitmap_start, u32 unit,
1081 u64 offset)
1082 {
1083 BUG_ON(offset < bitmap_start);
1084 offset -= bitmap_start;
1085 return (unsigned long)(div_u64(offset, unit));
1086 }
1087
1088 static inline unsigned long bytes_to_bits(u64 bytes, u32 unit)
1089 {
1090 return (unsigned long)(div_u64(bytes, unit));
1091 }
1092
1093 static inline u64 offset_to_bitmap(struct btrfs_free_space_ctl *ctl,
1094 u64 offset)
1095 {
1096 u64 bitmap_start;
1097 u64 bytes_per_bitmap;
1098
1099 bytes_per_bitmap = BITS_PER_BITMAP * ctl->unit;
1100 bitmap_start = offset - ctl->start;
1101 bitmap_start = div64_u64(bitmap_start, bytes_per_bitmap);
1102 bitmap_start *= bytes_per_bitmap;
1103 bitmap_start += ctl->start;
1104
1105 return bitmap_start;
1106 }
1107
1108 static int tree_insert_offset(struct rb_root *root, u64 offset,
1109 struct rb_node *node, int bitmap)
1110 {
1111 struct rb_node **p = &root->rb_node;
1112 struct rb_node *parent = NULL;
1113 struct btrfs_free_space *info;
1114
1115 while (*p) {
1116 parent = *p;
1117 info = rb_entry(parent, struct btrfs_free_space, offset_index);
1118
1119 if (offset < info->offset) {
1120 p = &(*p)->rb_left;
1121 } else if (offset > info->offset) {
1122 p = &(*p)->rb_right;
1123 } else {
1124 /*
1125 * we could have a bitmap entry and an extent entry
1126 * share the same offset. If this is the case, we want
1127 * the extent entry to always be found first if we do a
1128 * linear search through the tree, since we want to have
1129 * the quickest allocation time, and allocating from an
1130 * extent is faster than allocating from a bitmap. So
1131 * if we're inserting a bitmap and we find an entry at
1132 * this offset, we want to go right, or after this entry
1133 * logically. If we are inserting an extent and we've
1134 * found a bitmap, we want to go left, or before
1135 * logically.
1136 */
1137 if (bitmap) {
1138 if (info->bitmap) {
1139 WARN_ON_ONCE(1);
1140 return -EEXIST;
1141 }
1142 p = &(*p)->rb_right;
1143 } else {
1144 if (!info->bitmap) {
1145 WARN_ON_ONCE(1);
1146 return -EEXIST;
1147 }
1148 p = &(*p)->rb_left;
1149 }
1150 }
1151 }
1152
1153 rb_link_node(node, parent, p);
1154 rb_insert_color(node, root);
1155
1156 return 0;
1157 }
1158
1159 /*
1160 * searches the tree for the given offset.
1161 *
1162 * fuzzy - If this is set, then we are trying to make an allocation, and we just
1163 * want a section that has at least bytes size and comes at or after the given
1164 * offset.
1165 */
1166 static struct btrfs_free_space *
1167 tree_search_offset(struct btrfs_free_space_ctl *ctl,
1168 u64 offset, int bitmap_only, int fuzzy)
1169 {
1170 struct rb_node *n = ctl->free_space_offset.rb_node;
1171 struct btrfs_free_space *entry, *prev = NULL;
1172
1173 /* find entry that is closest to the 'offset' */
1174 while (1) {
1175 if (!n) {
1176 entry = NULL;
1177 break;
1178 }
1179
1180 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1181 prev = entry;
1182
1183 if (offset < entry->offset)
1184 n = n->rb_left;
1185 else if (offset > entry->offset)
1186 n = n->rb_right;
1187 else
1188 break;
1189 }
1190
1191 if (bitmap_only) {
1192 if (!entry)
1193 return NULL;
1194 if (entry->bitmap)
1195 return entry;
1196
1197 /*
1198 * bitmap entry and extent entry may share same offset,
1199 * in that case, bitmap entry comes after extent entry.
1200 */
1201 n = rb_next(n);
1202 if (!n)
1203 return NULL;
1204 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1205 if (entry->offset != offset)
1206 return NULL;
1207
1208 WARN_ON(!entry->bitmap);
1209 return entry;
1210 } else if (entry) {
1211 if (entry->bitmap) {
1212 /*
1213 * if previous extent entry covers the offset,
1214 * we should return it instead of the bitmap entry
1215 */
1216 n = &entry->offset_index;
1217 while (1) {
1218 n = rb_prev(n);
1219 if (!n)
1220 break;
1221 prev = rb_entry(n, struct btrfs_free_space,
1222 offset_index);
1223 if (!prev->bitmap) {
1224 if (prev->offset + prev->bytes > offset)
1225 entry = prev;
1226 break;
1227 }
1228 }
1229 }
1230 return entry;
1231 }
1232
1233 if (!prev)
1234 return NULL;
1235
1236 /* find last entry before the 'offset' */
1237 entry = prev;
1238 if (entry->offset > offset) {
1239 n = rb_prev(&entry->offset_index);
1240 if (n) {
1241 entry = rb_entry(n, struct btrfs_free_space,
1242 offset_index);
1243 BUG_ON(entry->offset > offset);
1244 } else {
1245 if (fuzzy)
1246 return entry;
1247 else
1248 return NULL;
1249 }
1250 }
1251
1252 if (entry->bitmap) {
1253 n = &entry->offset_index;
1254 while (1) {
1255 n = rb_prev(n);
1256 if (!n)
1257 break;
1258 prev = rb_entry(n, struct btrfs_free_space,
1259 offset_index);
1260 if (!prev->bitmap) {
1261 if (prev->offset + prev->bytes > offset)
1262 return prev;
1263 break;
1264 }
1265 }
1266 if (entry->offset + BITS_PER_BITMAP * ctl->unit > offset)
1267 return entry;
1268 } else if (entry->offset + entry->bytes > offset)
1269 return entry;
1270
1271 if (!fuzzy)
1272 return NULL;
1273
1274 while (1) {
1275 if (entry->bitmap) {
1276 if (entry->offset + BITS_PER_BITMAP *
1277 ctl->unit > offset)
1278 break;
1279 } else {
1280 if (entry->offset + entry->bytes > offset)
1281 break;
1282 }
1283
1284 n = rb_next(&entry->offset_index);
1285 if (!n)
1286 return NULL;
1287 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1288 }
1289 return entry;
1290 }
1291
1292 static inline void
1293 __unlink_free_space(struct btrfs_free_space_ctl *ctl,
1294 struct btrfs_free_space *info)
1295 {
1296 rb_erase(&info->offset_index, &ctl->free_space_offset);
1297 ctl->free_extents--;
1298 }
1299
1300 static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
1301 struct btrfs_free_space *info)
1302 {
1303 __unlink_free_space(ctl, info);
1304 ctl->free_space -= info->bytes;
1305 }
1306
1307 static int link_free_space(struct btrfs_free_space_ctl *ctl,
1308 struct btrfs_free_space *info)
1309 {
1310 int ret = 0;
1311
1312 BUG_ON(!info->bitmap && !info->bytes);
1313 ret = tree_insert_offset(&ctl->free_space_offset, info->offset,
1314 &info->offset_index, (info->bitmap != NULL));
1315 if (ret)
1316 return ret;
1317
1318 ctl->free_space += info->bytes;
1319 ctl->free_extents++;
1320 return ret;
1321 }
1322
1323 static void recalculate_thresholds(struct btrfs_free_space_ctl *ctl)
1324 {
1325 struct btrfs_block_group_cache *block_group = ctl->private;
1326 u64 max_bytes;
1327 u64 bitmap_bytes;
1328 u64 extent_bytes;
1329 u64 size = block_group->key.offset;
1330 u64 bytes_per_bg = BITS_PER_BITMAP * block_group->sectorsize;
1331 int max_bitmaps = div64_u64(size + bytes_per_bg - 1, bytes_per_bg);
1332
1333 BUG_ON(ctl->total_bitmaps > max_bitmaps);
1334
1335 /*
1336 * The goal is to keep the total amount of memory used per 1gb of space
1337 * at or below 32k, so we need to adjust how much memory we allow to be
1338 * used by extent based free space tracking
1339 */
1340 if (size < 1024 * 1024 * 1024)
1341 max_bytes = MAX_CACHE_BYTES_PER_GIG;
1342 else
1343 max_bytes = MAX_CACHE_BYTES_PER_GIG *
1344 div64_u64(size, 1024 * 1024 * 1024);
1345
1346 /*
1347 * we want to account for 1 more bitmap than what we have so we can make
1348 * sure we don't go over our overall goal of MAX_CACHE_BYTES_PER_GIG as
1349 * we add more bitmaps.
1350 */
1351 bitmap_bytes = (ctl->total_bitmaps + 1) * PAGE_CACHE_SIZE;
1352
1353 if (bitmap_bytes >= max_bytes) {
1354 ctl->extents_thresh = 0;
1355 return;
1356 }
1357
1358 /*
1359 * we want the extent entry threshold to always be at most 1/2 the maxw
1360 * bytes we can have, or whatever is less than that.
1361 */
1362 extent_bytes = max_bytes - bitmap_bytes;
1363 extent_bytes = min_t(u64, extent_bytes, div64_u64(max_bytes, 2));
1364
1365 ctl->extents_thresh =
1366 div64_u64(extent_bytes, (sizeof(struct btrfs_free_space)));
1367 }
1368
1369 static inline void __bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
1370 struct btrfs_free_space *info,
1371 u64 offset, u64 bytes)
1372 {
1373 unsigned long start, count;
1374
1375 start = offset_to_bit(info->offset, ctl->unit, offset);
1376 count = bytes_to_bits(bytes, ctl->unit);
1377 BUG_ON(start + count > BITS_PER_BITMAP);
1378
1379 bitmap_clear(info->bitmap, start, count);
1380
1381 info->bytes -= bytes;
1382 }
1383
1384 static void bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
1385 struct btrfs_free_space *info, u64 offset,
1386 u64 bytes)
1387 {
1388 __bitmap_clear_bits(ctl, info, offset, bytes);
1389 ctl->free_space -= bytes;
1390 }
1391
1392 static void bitmap_set_bits(struct btrfs_free_space_ctl *ctl,
1393 struct btrfs_free_space *info, u64 offset,
1394 u64 bytes)
1395 {
1396 unsigned long start, count;
1397
1398 start = offset_to_bit(info->offset, ctl->unit, offset);
1399 count = bytes_to_bits(bytes, ctl->unit);
1400 BUG_ON(start + count > BITS_PER_BITMAP);
1401
1402 bitmap_set(info->bitmap, start, count);
1403
1404 info->bytes += bytes;
1405 ctl->free_space += bytes;
1406 }
1407
1408 static int search_bitmap(struct btrfs_free_space_ctl *ctl,
1409 struct btrfs_free_space *bitmap_info, u64 *offset,
1410 u64 *bytes)
1411 {
1412 unsigned long found_bits = 0;
1413 unsigned long bits, i;
1414 unsigned long next_zero;
1415
1416 i = offset_to_bit(bitmap_info->offset, ctl->unit,
1417 max_t(u64, *offset, bitmap_info->offset));
1418 bits = bytes_to_bits(*bytes, ctl->unit);
1419
1420 for (i = find_next_bit(bitmap_info->bitmap, BITS_PER_BITMAP, i);
1421 i < BITS_PER_BITMAP;
1422 i = find_next_bit(bitmap_info->bitmap, BITS_PER_BITMAP, i + 1)) {
1423 next_zero = find_next_zero_bit(bitmap_info->bitmap,
1424 BITS_PER_BITMAP, i);
1425 if ((next_zero - i) >= bits) {
1426 found_bits = next_zero - i;
1427 break;
1428 }
1429 i = next_zero;
1430 }
1431
1432 if (found_bits) {
1433 *offset = (u64)(i * ctl->unit) + bitmap_info->offset;
1434 *bytes = (u64)(found_bits) * ctl->unit;
1435 return 0;
1436 }
1437
1438 return -1;
1439 }
1440
1441 static struct btrfs_free_space *
1442 find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes)
1443 {
1444 struct btrfs_free_space *entry;
1445 struct rb_node *node;
1446 int ret;
1447
1448 if (!ctl->free_space_offset.rb_node)
1449 return NULL;
1450
1451 entry = tree_search_offset(ctl, offset_to_bitmap(ctl, *offset), 0, 1);
1452 if (!entry)
1453 return NULL;
1454
1455 for (node = &entry->offset_index; node; node = rb_next(node)) {
1456 entry = rb_entry(node, struct btrfs_free_space, offset_index);
1457 if (entry->bytes < *bytes)
1458 continue;
1459
1460 if (entry->bitmap) {
1461 ret = search_bitmap(ctl, entry, offset, bytes);
1462 if (!ret)
1463 return entry;
1464 continue;
1465 }
1466
1467 *offset = entry->offset;
1468 *bytes = entry->bytes;
1469 return entry;
1470 }
1471
1472 return NULL;
1473 }
1474
1475 static void add_new_bitmap(struct btrfs_free_space_ctl *ctl,
1476 struct btrfs_free_space *info, u64 offset)
1477 {
1478 info->offset = offset_to_bitmap(ctl, offset);
1479 info->bytes = 0;
1480 INIT_LIST_HEAD(&info->list);
1481 link_free_space(ctl, info);
1482 ctl->total_bitmaps++;
1483
1484 ctl->op->recalc_thresholds(ctl);
1485 }
1486
1487 static void free_bitmap(struct btrfs_free_space_ctl *ctl,
1488 struct btrfs_free_space *bitmap_info)
1489 {
1490 unlink_free_space(ctl, bitmap_info);
1491 kfree(bitmap_info->bitmap);
1492 kmem_cache_free(btrfs_free_space_cachep, bitmap_info);
1493 ctl->total_bitmaps--;
1494 ctl->op->recalc_thresholds(ctl);
1495 }
1496
1497 static noinline int remove_from_bitmap(struct btrfs_free_space_ctl *ctl,
1498 struct btrfs_free_space *bitmap_info,
1499 u64 *offset, u64 *bytes)
1500 {
1501 u64 end;
1502 u64 search_start, search_bytes;
1503 int ret;
1504
1505 again:
1506 end = bitmap_info->offset + (u64)(BITS_PER_BITMAP * ctl->unit) - 1;
1507
1508 /*
1509 * XXX - this can go away after a few releases.
1510 *
1511 * since the only user of btrfs_remove_free_space is the tree logging
1512 * stuff, and the only way to test that is under crash conditions, we
1513 * want to have this debug stuff here just in case somethings not
1514 * working. Search the bitmap for the space we are trying to use to
1515 * make sure its actually there. If its not there then we need to stop
1516 * because something has gone wrong.
1517 */
1518 search_start = *offset;
1519 search_bytes = *bytes;
1520 search_bytes = min(search_bytes, end - search_start + 1);
1521 ret = search_bitmap(ctl, bitmap_info, &search_start, &search_bytes);
1522 BUG_ON(ret < 0 || search_start != *offset);
1523
1524 if (*offset > bitmap_info->offset && *offset + *bytes > end) {
1525 bitmap_clear_bits(ctl, bitmap_info, *offset, end - *offset + 1);
1526 *bytes -= end - *offset + 1;
1527 *offset = end + 1;
1528 } else if (*offset >= bitmap_info->offset && *offset + *bytes <= end) {
1529 bitmap_clear_bits(ctl, bitmap_info, *offset, *bytes);
1530 *bytes = 0;
1531 }
1532
1533 if (*bytes) {
1534 struct rb_node *next = rb_next(&bitmap_info->offset_index);
1535 if (!bitmap_info->bytes)
1536 free_bitmap(ctl, bitmap_info);
1537
1538 /*
1539 * no entry after this bitmap, but we still have bytes to
1540 * remove, so something has gone wrong.
1541 */
1542 if (!next)
1543 return -EINVAL;
1544
1545 bitmap_info = rb_entry(next, struct btrfs_free_space,
1546 offset_index);
1547
1548 /*
1549 * if the next entry isn't a bitmap we need to return to let the
1550 * extent stuff do its work.
1551 */
1552 if (!bitmap_info->bitmap)
1553 return -EAGAIN;
1554
1555 /*
1556 * Ok the next item is a bitmap, but it may not actually hold
1557 * the information for the rest of this free space stuff, so
1558 * look for it, and if we don't find it return so we can try
1559 * everything over again.
1560 */
1561 search_start = *offset;
1562 search_bytes = *bytes;
1563 ret = search_bitmap(ctl, bitmap_info, &search_start,
1564 &search_bytes);
1565 if (ret < 0 || search_start != *offset)
1566 return -EAGAIN;
1567
1568 goto again;
1569 } else if (!bitmap_info->bytes)
1570 free_bitmap(ctl, bitmap_info);
1571
1572 return 0;
1573 }
1574
1575 static u64 add_bytes_to_bitmap(struct btrfs_free_space_ctl *ctl,
1576 struct btrfs_free_space *info, u64 offset,
1577 u64 bytes)
1578 {
1579 u64 bytes_to_set = 0;
1580 u64 end;
1581
1582 end = info->offset + (u64)(BITS_PER_BITMAP * ctl->unit);
1583
1584 bytes_to_set = min(end - offset, bytes);
1585
1586 bitmap_set_bits(ctl, info, offset, bytes_to_set);
1587
1588 return bytes_to_set;
1589
1590 }
1591
1592 static bool use_bitmap(struct btrfs_free_space_ctl *ctl,
1593 struct btrfs_free_space *info)
1594 {
1595 struct btrfs_block_group_cache *block_group = ctl->private;
1596
1597 /*
1598 * If we are below the extents threshold then we can add this as an
1599 * extent, and don't have to deal with the bitmap
1600 */
1601 if (ctl->free_extents < ctl->extents_thresh) {
1602 /*
1603 * If this block group has some small extents we don't want to
1604 * use up all of our free slots in the cache with them, we want
1605 * to reserve them to larger extents, however if we have plent
1606 * of cache left then go ahead an dadd them, no sense in adding
1607 * the overhead of a bitmap if we don't have to.
1608 */
1609 if (info->bytes <= block_group->sectorsize * 4) {
1610 if (ctl->free_extents * 2 <= ctl->extents_thresh)
1611 return false;
1612 } else {
1613 return false;
1614 }
1615 }
1616
1617 /*
1618 * some block groups are so tiny they can't be enveloped by a bitmap, so
1619 * don't even bother to create a bitmap for this
1620 */
1621 if (BITS_PER_BITMAP * block_group->sectorsize >
1622 block_group->key.offset)
1623 return false;
1624
1625 return true;
1626 }
1627
1628 static struct btrfs_free_space_op free_space_op = {
1629 .recalc_thresholds = recalculate_thresholds,
1630 .use_bitmap = use_bitmap,
1631 };
1632
1633 static int insert_into_bitmap(struct btrfs_free_space_ctl *ctl,
1634 struct btrfs_free_space *info)
1635 {
1636 struct btrfs_free_space *bitmap_info;
1637 struct btrfs_block_group_cache *block_group = NULL;
1638 int added = 0;
1639 u64 bytes, offset, bytes_added;
1640 int ret;
1641
1642 bytes = info->bytes;
1643 offset = info->offset;
1644
1645 if (!ctl->op->use_bitmap(ctl, info))
1646 return 0;
1647
1648 if (ctl->op == &free_space_op)
1649 block_group = ctl->private;
1650 again:
1651 /*
1652 * Since we link bitmaps right into the cluster we need to see if we
1653 * have a cluster here, and if so and it has our bitmap we need to add
1654 * the free space to that bitmap.
1655 */
1656 if (block_group && !list_empty(&block_group->cluster_list)) {
1657 struct btrfs_free_cluster *cluster;
1658 struct rb_node *node;
1659 struct btrfs_free_space *entry;
1660
1661 cluster = list_entry(block_group->cluster_list.next,
1662 struct btrfs_free_cluster,
1663 block_group_list);
1664 spin_lock(&cluster->lock);
1665 node = rb_first(&cluster->root);
1666 if (!node) {
1667 spin_unlock(&cluster->lock);
1668 goto no_cluster_bitmap;
1669 }
1670
1671 entry = rb_entry(node, struct btrfs_free_space, offset_index);
1672 if (!entry->bitmap) {
1673 spin_unlock(&cluster->lock);
1674 goto no_cluster_bitmap;
1675 }
1676
1677 if (entry->offset == offset_to_bitmap(ctl, offset)) {
1678 bytes_added = add_bytes_to_bitmap(ctl, entry,
1679 offset, bytes);
1680 bytes -= bytes_added;
1681 offset += bytes_added;
1682 }
1683 spin_unlock(&cluster->lock);
1684 if (!bytes) {
1685 ret = 1;
1686 goto out;
1687 }
1688 }
1689
1690 no_cluster_bitmap:
1691 bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
1692 1, 0);
1693 if (!bitmap_info) {
1694 BUG_ON(added);
1695 goto new_bitmap;
1696 }
1697
1698 bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes);
1699 bytes -= bytes_added;
1700 offset += bytes_added;
1701 added = 0;
1702
1703 if (!bytes) {
1704 ret = 1;
1705 goto out;
1706 } else
1707 goto again;
1708
1709 new_bitmap:
1710 if (info && info->bitmap) {
1711 add_new_bitmap(ctl, info, offset);
1712 added = 1;
1713 info = NULL;
1714 goto again;
1715 } else {
1716 spin_unlock(&ctl->tree_lock);
1717
1718 /* no pre-allocated info, allocate a new one */
1719 if (!info) {
1720 info = kmem_cache_zalloc(btrfs_free_space_cachep,
1721 GFP_NOFS);
1722 if (!info) {
1723 spin_lock(&ctl->tree_lock);
1724 ret = -ENOMEM;
1725 goto out;
1726 }
1727 }
1728
1729 /* allocate the bitmap */
1730 info->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
1731 spin_lock(&ctl->tree_lock);
1732 if (!info->bitmap) {
1733 ret = -ENOMEM;
1734 goto out;
1735 }
1736 goto again;
1737 }
1738
1739 out:
1740 if (info) {
1741 if (info->bitmap)
1742 kfree(info->bitmap);
1743 kmem_cache_free(btrfs_free_space_cachep, info);
1744 }
1745
1746 return ret;
1747 }
1748
1749 static bool try_merge_free_space(struct btrfs_free_space_ctl *ctl,
1750 struct btrfs_free_space *info, bool update_stat)
1751 {
1752 struct btrfs_free_space *left_info;
1753 struct btrfs_free_space *right_info;
1754 bool merged = false;
1755 u64 offset = info->offset;
1756 u64 bytes = info->bytes;
1757
1758 /*
1759 * first we want to see if there is free space adjacent to the range we
1760 * are adding, if there is remove that struct and add a new one to
1761 * cover the entire range
1762 */
1763 right_info = tree_search_offset(ctl, offset + bytes, 0, 0);
1764 if (right_info && rb_prev(&right_info->offset_index))
1765 left_info = rb_entry(rb_prev(&right_info->offset_index),
1766 struct btrfs_free_space, offset_index);
1767 else
1768 left_info = tree_search_offset(ctl, offset - 1, 0, 0);
1769
1770 if (right_info && !right_info->bitmap) {
1771 if (update_stat)
1772 unlink_free_space(ctl, right_info);
1773 else
1774 __unlink_free_space(ctl, right_info);
1775 info->bytes += right_info->bytes;
1776 kmem_cache_free(btrfs_free_space_cachep, right_info);
1777 merged = true;
1778 }
1779
1780 if (left_info && !left_info->bitmap &&
1781 left_info->offset + left_info->bytes == offset) {
1782 if (update_stat)
1783 unlink_free_space(ctl, left_info);
1784 else
1785 __unlink_free_space(ctl, left_info);
1786 info->offset = left_info->offset;
1787 info->bytes += left_info->bytes;
1788 kmem_cache_free(btrfs_free_space_cachep, left_info);
1789 merged = true;
1790 }
1791
1792 return merged;
1793 }
1794
1795 int __btrfs_add_free_space(struct btrfs_free_space_ctl *ctl,
1796 u64 offset, u64 bytes)
1797 {
1798 struct btrfs_free_space *info;
1799 int ret = 0;
1800
1801 info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
1802 if (!info)
1803 return -ENOMEM;
1804
1805 info->offset = offset;
1806 info->bytes = bytes;
1807
1808 spin_lock(&ctl->tree_lock);
1809
1810 if (try_merge_free_space(ctl, info, true))
1811 goto link;
1812
1813 /*
1814 * There was no extent directly to the left or right of this new
1815 * extent then we know we're going to have to allocate a new extent, so
1816 * before we do that see if we need to drop this into a bitmap
1817 */
1818 ret = insert_into_bitmap(ctl, info);
1819 if (ret < 0) {
1820 goto out;
1821 } else if (ret) {
1822 ret = 0;
1823 goto out;
1824 }
1825 link:
1826 ret = link_free_space(ctl, info);
1827 if (ret)
1828 kmem_cache_free(btrfs_free_space_cachep, info);
1829 out:
1830 spin_unlock(&ctl->tree_lock);
1831
1832 if (ret) {
1833 printk(KERN_CRIT "btrfs: unable to add free space :%d\n", ret);
1834 BUG_ON(ret == -EEXIST);
1835 }
1836
1837 return ret;
1838 }
1839
1840 int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
1841 u64 offset, u64 bytes)
1842 {
1843 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1844 struct btrfs_free_space *info;
1845 struct btrfs_free_space *next_info = NULL;
1846 int ret = 0;
1847
1848 spin_lock(&ctl->tree_lock);
1849
1850 again:
1851 info = tree_search_offset(ctl, offset, 0, 0);
1852 if (!info) {
1853 /*
1854 * oops didn't find an extent that matched the space we wanted
1855 * to remove, look for a bitmap instead
1856 */
1857 info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
1858 1, 0);
1859 if (!info) {
1860 /* the tree logging code might be calling us before we
1861 * have fully loaded the free space rbtree for this
1862 * block group. So it is possible the entry won't
1863 * be in the rbtree yet at all. The caching code
1864 * will make sure not to put it in the rbtree if
1865 * the logging code has pinned it.
1866 */
1867 goto out_lock;
1868 }
1869 }
1870
1871 if (info->bytes < bytes && rb_next(&info->offset_index)) {
1872 u64 end;
1873 next_info = rb_entry(rb_next(&info->offset_index),
1874 struct btrfs_free_space,
1875 offset_index);
1876
1877 if (next_info->bitmap)
1878 end = next_info->offset +
1879 BITS_PER_BITMAP * ctl->unit - 1;
1880 else
1881 end = next_info->offset + next_info->bytes;
1882
1883 if (next_info->bytes < bytes ||
1884 next_info->offset > offset || offset > end) {
1885 printk(KERN_CRIT "Found free space at %llu, size %llu,"
1886 " trying to use %llu\n",
1887 (unsigned long long)info->offset,
1888 (unsigned long long)info->bytes,
1889 (unsigned long long)bytes);
1890 WARN_ON(1);
1891 ret = -EINVAL;
1892 goto out_lock;
1893 }
1894
1895 info = next_info;
1896 }
1897
1898 if (info->bytes == bytes) {
1899 unlink_free_space(ctl, info);
1900 if (info->bitmap) {
1901 kfree(info->bitmap);
1902 ctl->total_bitmaps--;
1903 }
1904 kmem_cache_free(btrfs_free_space_cachep, info);
1905 ret = 0;
1906 goto out_lock;
1907 }
1908
1909 if (!info->bitmap && info->offset == offset) {
1910 unlink_free_space(ctl, info);
1911 info->offset += bytes;
1912 info->bytes -= bytes;
1913 ret = link_free_space(ctl, info);
1914 WARN_ON(ret);
1915 goto out_lock;
1916 }
1917
1918 if (!info->bitmap && info->offset <= offset &&
1919 info->offset + info->bytes >= offset + bytes) {
1920 u64 old_start = info->offset;
1921 /*
1922 * we're freeing space in the middle of the info,
1923 * this can happen during tree log replay
1924 *
1925 * first unlink the old info and then
1926 * insert it again after the hole we're creating
1927 */
1928 unlink_free_space(ctl, info);
1929 if (offset + bytes < info->offset + info->bytes) {
1930 u64 old_end = info->offset + info->bytes;
1931
1932 info->offset = offset + bytes;
1933 info->bytes = old_end - info->offset;
1934 ret = link_free_space(ctl, info);
1935 WARN_ON(ret);
1936 if (ret)
1937 goto out_lock;
1938 } else {
1939 /* the hole we're creating ends at the end
1940 * of the info struct, just free the info
1941 */
1942 kmem_cache_free(btrfs_free_space_cachep, info);
1943 }
1944 spin_unlock(&ctl->tree_lock);
1945
1946 /* step two, insert a new info struct to cover
1947 * anything before the hole
1948 */
1949 ret = btrfs_add_free_space(block_group, old_start,
1950 offset - old_start);
1951 WARN_ON(ret);
1952 goto out;
1953 }
1954
1955 ret = remove_from_bitmap(ctl, info, &offset, &bytes);
1956 if (ret == -EAGAIN)
1957 goto again;
1958 BUG_ON(ret);
1959 out_lock:
1960 spin_unlock(&ctl->tree_lock);
1961 out:
1962 return ret;
1963 }
1964
1965 void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
1966 u64 bytes)
1967 {
1968 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1969 struct btrfs_free_space *info;
1970 struct rb_node *n;
1971 int count = 0;
1972
1973 for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
1974 info = rb_entry(n, struct btrfs_free_space, offset_index);
1975 if (info->bytes >= bytes)
1976 count++;
1977 printk(KERN_CRIT "entry offset %llu, bytes %llu, bitmap %s\n",
1978 (unsigned long long)info->offset,
1979 (unsigned long long)info->bytes,
1980 (info->bitmap) ? "yes" : "no");
1981 }
1982 printk(KERN_INFO "block group has cluster?: %s\n",
1983 list_empty(&block_group->cluster_list) ? "no" : "yes");
1984 printk(KERN_INFO "%d blocks of free space at or bigger than bytes is"
1985 "\n", count);
1986 }
1987
1988 void btrfs_init_free_space_ctl(struct btrfs_block_group_cache *block_group)
1989 {
1990 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1991
1992 spin_lock_init(&ctl->tree_lock);
1993 ctl->unit = block_group->sectorsize;
1994 ctl->start = block_group->key.objectid;
1995 ctl->private = block_group;
1996 ctl->op = &free_space_op;
1997
1998 /*
1999 * we only want to have 32k of ram per block group for keeping
2000 * track of free space, and if we pass 1/2 of that we want to
2001 * start converting things over to using bitmaps
2002 */
2003 ctl->extents_thresh = ((1024 * 32) / 2) /
2004 sizeof(struct btrfs_free_space);
2005 }
2006
2007 /*
2008 * for a given cluster, put all of its extents back into the free
2009 * space cache. If the block group passed doesn't match the block group
2010 * pointed to by the cluster, someone else raced in and freed the
2011 * cluster already. In that case, we just return without changing anything
2012 */
2013 static int
2014 __btrfs_return_cluster_to_free_space(
2015 struct btrfs_block_group_cache *block_group,
2016 struct btrfs_free_cluster *cluster)
2017 {
2018 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2019 struct btrfs_free_space *entry;
2020 struct rb_node *node;
2021
2022 spin_lock(&cluster->lock);
2023 if (cluster->block_group != block_group)
2024 goto out;
2025
2026 cluster->block_group = NULL;
2027 cluster->window_start = 0;
2028 list_del_init(&cluster->block_group_list);
2029
2030 node = rb_first(&cluster->root);
2031 while (node) {
2032 bool bitmap;
2033
2034 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2035 node = rb_next(&entry->offset_index);
2036 rb_erase(&entry->offset_index, &cluster->root);
2037
2038 bitmap = (entry->bitmap != NULL);
2039 if (!bitmap)
2040 try_merge_free_space(ctl, entry, false);
2041 tree_insert_offset(&ctl->free_space_offset,
2042 entry->offset, &entry->offset_index, bitmap);
2043 }
2044 cluster->root = RB_ROOT;
2045
2046 out:
2047 spin_unlock(&cluster->lock);
2048 btrfs_put_block_group(block_group);
2049 return 0;
2050 }
2051
2052 void __btrfs_remove_free_space_cache_locked(struct btrfs_free_space_ctl *ctl)
2053 {
2054 struct btrfs_free_space *info;
2055 struct rb_node *node;
2056
2057 while ((node = rb_last(&ctl->free_space_offset)) != NULL) {
2058 info = rb_entry(node, struct btrfs_free_space, offset_index);
2059 if (!info->bitmap) {
2060 unlink_free_space(ctl, info);
2061 kmem_cache_free(btrfs_free_space_cachep, info);
2062 } else {
2063 free_bitmap(ctl, info);
2064 }
2065 if (need_resched()) {
2066 spin_unlock(&ctl->tree_lock);
2067 cond_resched();
2068 spin_lock(&ctl->tree_lock);
2069 }
2070 }
2071 }
2072
2073 void __btrfs_remove_free_space_cache(struct btrfs_free_space_ctl *ctl)
2074 {
2075 spin_lock(&ctl->tree_lock);
2076 __btrfs_remove_free_space_cache_locked(ctl);
2077 spin_unlock(&ctl->tree_lock);
2078 }
2079
2080 void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
2081 {
2082 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2083 struct btrfs_free_cluster *cluster;
2084 struct list_head *head;
2085
2086 spin_lock(&ctl->tree_lock);
2087 while ((head = block_group->cluster_list.next) !=
2088 &block_group->cluster_list) {
2089 cluster = list_entry(head, struct btrfs_free_cluster,
2090 block_group_list);
2091
2092 WARN_ON(cluster->block_group != block_group);
2093 __btrfs_return_cluster_to_free_space(block_group, cluster);
2094 if (need_resched()) {
2095 spin_unlock(&ctl->tree_lock);
2096 cond_resched();
2097 spin_lock(&ctl->tree_lock);
2098 }
2099 }
2100 __btrfs_remove_free_space_cache_locked(ctl);
2101 spin_unlock(&ctl->tree_lock);
2102
2103 }
2104
2105 u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
2106 u64 offset, u64 bytes, u64 empty_size)
2107 {
2108 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2109 struct btrfs_free_space *entry = NULL;
2110 u64 bytes_search = bytes + empty_size;
2111 u64 ret = 0;
2112
2113 spin_lock(&ctl->tree_lock);
2114 entry = find_free_space(ctl, &offset, &bytes_search);
2115 if (!entry)
2116 goto out;
2117
2118 ret = offset;
2119 if (entry->bitmap) {
2120 bitmap_clear_bits(ctl, entry, offset, bytes);
2121 if (!entry->bytes)
2122 free_bitmap(ctl, entry);
2123 } else {
2124 unlink_free_space(ctl, entry);
2125 entry->offset += bytes;
2126 entry->bytes -= bytes;
2127 if (!entry->bytes)
2128 kmem_cache_free(btrfs_free_space_cachep, entry);
2129 else
2130 link_free_space(ctl, entry);
2131 }
2132
2133 out:
2134 spin_unlock(&ctl->tree_lock);
2135
2136 return ret;
2137 }
2138
2139 /*
2140 * given a cluster, put all of its extents back into the free space
2141 * cache. If a block group is passed, this function will only free
2142 * a cluster that belongs to the passed block group.
2143 *
2144 * Otherwise, it'll get a reference on the block group pointed to by the
2145 * cluster and remove the cluster from it.
2146 */
2147 int btrfs_return_cluster_to_free_space(
2148 struct btrfs_block_group_cache *block_group,
2149 struct btrfs_free_cluster *cluster)
2150 {
2151 struct btrfs_free_space_ctl *ctl;
2152 int ret;
2153
2154 /* first, get a safe pointer to the block group */
2155 spin_lock(&cluster->lock);
2156 if (!block_group) {
2157 block_group = cluster->block_group;
2158 if (!block_group) {
2159 spin_unlock(&cluster->lock);
2160 return 0;
2161 }
2162 } else if (cluster->block_group != block_group) {
2163 /* someone else has already freed it don't redo their work */
2164 spin_unlock(&cluster->lock);
2165 return 0;
2166 }
2167 atomic_inc(&block_group->count);
2168 spin_unlock(&cluster->lock);
2169
2170 ctl = block_group->free_space_ctl;
2171
2172 /* now return any extents the cluster had on it */
2173 spin_lock(&ctl->tree_lock);
2174 ret = __btrfs_return_cluster_to_free_space(block_group, cluster);
2175 spin_unlock(&ctl->tree_lock);
2176
2177 /* finally drop our ref */
2178 btrfs_put_block_group(block_group);
2179 return ret;
2180 }
2181
2182 static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
2183 struct btrfs_free_cluster *cluster,
2184 struct btrfs_free_space *entry,
2185 u64 bytes, u64 min_start)
2186 {
2187 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2188 int err;
2189 u64 search_start = cluster->window_start;
2190 u64 search_bytes = bytes;
2191 u64 ret = 0;
2192
2193 search_start = min_start;
2194 search_bytes = bytes;
2195
2196 err = search_bitmap(ctl, entry, &search_start, &search_bytes);
2197 if (err)
2198 return 0;
2199
2200 ret = search_start;
2201 __bitmap_clear_bits(ctl, entry, ret, bytes);
2202
2203 return ret;
2204 }
2205
2206 /*
2207 * given a cluster, try to allocate 'bytes' from it, returns 0
2208 * if it couldn't find anything suitably large, or a logical disk offset
2209 * if things worked out
2210 */
2211 u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
2212 struct btrfs_free_cluster *cluster, u64 bytes,
2213 u64 min_start)
2214 {
2215 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2216 struct btrfs_free_space *entry = NULL;
2217 struct rb_node *node;
2218 u64 ret = 0;
2219
2220 spin_lock(&cluster->lock);
2221 if (bytes > cluster->max_size)
2222 goto out;
2223
2224 if (cluster->block_group != block_group)
2225 goto out;
2226
2227 node = rb_first(&cluster->root);
2228 if (!node)
2229 goto out;
2230
2231 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2232 while(1) {
2233 if (entry->bytes < bytes ||
2234 (!entry->bitmap && entry->offset < min_start)) {
2235 node = rb_next(&entry->offset_index);
2236 if (!node)
2237 break;
2238 entry = rb_entry(node, struct btrfs_free_space,
2239 offset_index);
2240 continue;
2241 }
2242
2243 if (entry->bitmap) {
2244 ret = btrfs_alloc_from_bitmap(block_group,
2245 cluster, entry, bytes,
2246 cluster->window_start);
2247 if (ret == 0) {
2248 node = rb_next(&entry->offset_index);
2249 if (!node)
2250 break;
2251 entry = rb_entry(node, struct btrfs_free_space,
2252 offset_index);
2253 continue;
2254 }
2255 cluster->window_start += bytes;
2256 } else {
2257 ret = entry->offset;
2258
2259 entry->offset += bytes;
2260 entry->bytes -= bytes;
2261 }
2262
2263 if (entry->bytes == 0)
2264 rb_erase(&entry->offset_index, &cluster->root);
2265 break;
2266 }
2267 out:
2268 spin_unlock(&cluster->lock);
2269
2270 if (!ret)
2271 return 0;
2272
2273 spin_lock(&ctl->tree_lock);
2274
2275 ctl->free_space -= bytes;
2276 if (entry->bytes == 0) {
2277 ctl->free_extents--;
2278 if (entry->bitmap) {
2279 kfree(entry->bitmap);
2280 ctl->total_bitmaps--;
2281 ctl->op->recalc_thresholds(ctl);
2282 }
2283 kmem_cache_free(btrfs_free_space_cachep, entry);
2284 }
2285
2286 spin_unlock(&ctl->tree_lock);
2287
2288 return ret;
2289 }
2290
2291 static int btrfs_bitmap_cluster(struct btrfs_block_group_cache *block_group,
2292 struct btrfs_free_space *entry,
2293 struct btrfs_free_cluster *cluster,
2294 u64 offset, u64 bytes,
2295 u64 cont1_bytes, u64 min_bytes)
2296 {
2297 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2298 unsigned long next_zero;
2299 unsigned long i;
2300 unsigned long want_bits;
2301 unsigned long min_bits;
2302 unsigned long found_bits;
2303 unsigned long start = 0;
2304 unsigned long total_found = 0;
2305 int ret;
2306
2307 i = offset_to_bit(entry->offset, block_group->sectorsize,
2308 max_t(u64, offset, entry->offset));
2309 want_bits = bytes_to_bits(bytes, block_group->sectorsize);
2310 min_bits = bytes_to_bits(min_bytes, block_group->sectorsize);
2311
2312 again:
2313 found_bits = 0;
2314 for (i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, i);
2315 i < BITS_PER_BITMAP;
2316 i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, i + 1)) {
2317 next_zero = find_next_zero_bit(entry->bitmap,
2318 BITS_PER_BITMAP, i);
2319 if (next_zero - i >= min_bits) {
2320 found_bits = next_zero - i;
2321 break;
2322 }
2323 i = next_zero;
2324 }
2325
2326 if (!found_bits)
2327 return -ENOSPC;
2328
2329 if (!total_found) {
2330 start = i;
2331 cluster->max_size = 0;
2332 }
2333
2334 total_found += found_bits;
2335
2336 if (cluster->max_size < found_bits * block_group->sectorsize)
2337 cluster->max_size = found_bits * block_group->sectorsize;
2338
2339 if (total_found < want_bits || cluster->max_size < cont1_bytes) {
2340 i = next_zero + 1;
2341 goto again;
2342 }
2343
2344 cluster->window_start = start * block_group->sectorsize +
2345 entry->offset;
2346 rb_erase(&entry->offset_index, &ctl->free_space_offset);
2347 ret = tree_insert_offset(&cluster->root, entry->offset,
2348 &entry->offset_index, 1);
2349 BUG_ON(ret);
2350
2351 trace_btrfs_setup_cluster(block_group, cluster,
2352 total_found * block_group->sectorsize, 1);
2353 return 0;
2354 }
2355
2356 /*
2357 * This searches the block group for just extents to fill the cluster with.
2358 * Try to find a cluster with at least bytes total bytes, at least one
2359 * extent of cont1_bytes, and other clusters of at least min_bytes.
2360 */
2361 static noinline int
2362 setup_cluster_no_bitmap(struct btrfs_block_group_cache *block_group,
2363 struct btrfs_free_cluster *cluster,
2364 struct list_head *bitmaps, u64 offset, u64 bytes,
2365 u64 cont1_bytes, u64 min_bytes)
2366 {
2367 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2368 struct btrfs_free_space *first = NULL;
2369 struct btrfs_free_space *entry = NULL;
2370 struct btrfs_free_space *last;
2371 struct rb_node *node;
2372 u64 window_start;
2373 u64 window_free;
2374 u64 max_extent;
2375 u64 total_size = 0;
2376
2377 entry = tree_search_offset(ctl, offset, 0, 1);
2378 if (!entry)
2379 return -ENOSPC;
2380
2381 /*
2382 * We don't want bitmaps, so just move along until we find a normal
2383 * extent entry.
2384 */
2385 while (entry->bitmap || entry->bytes < min_bytes) {
2386 if (entry->bitmap && list_empty(&entry->list))
2387 list_add_tail(&entry->list, bitmaps);
2388 node = rb_next(&entry->offset_index);
2389 if (!node)
2390 return -ENOSPC;
2391 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2392 }
2393
2394 window_start = entry->offset;
2395 window_free = entry->bytes;
2396 max_extent = entry->bytes;
2397 first = entry;
2398 last = entry;
2399
2400 for (node = rb_next(&entry->offset_index); node;
2401 node = rb_next(&entry->offset_index)) {
2402 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2403
2404 if (entry->bitmap) {
2405 if (list_empty(&entry->list))
2406 list_add_tail(&entry->list, bitmaps);
2407 continue;
2408 }
2409
2410 if (entry->bytes < min_bytes)
2411 continue;
2412
2413 last = entry;
2414 window_free += entry->bytes;
2415 if (entry->bytes > max_extent)
2416 max_extent = entry->bytes;
2417 }
2418
2419 if (window_free < bytes || max_extent < cont1_bytes)
2420 return -ENOSPC;
2421
2422 cluster->window_start = first->offset;
2423
2424 node = &first->offset_index;
2425
2426 /*
2427 * now we've found our entries, pull them out of the free space
2428 * cache and put them into the cluster rbtree
2429 */
2430 do {
2431 int ret;
2432
2433 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2434 node = rb_next(&entry->offset_index);
2435 if (entry->bitmap || entry->bytes < min_bytes)
2436 continue;
2437
2438 rb_erase(&entry->offset_index, &ctl->free_space_offset);
2439 ret = tree_insert_offset(&cluster->root, entry->offset,
2440 &entry->offset_index, 0);
2441 total_size += entry->bytes;
2442 BUG_ON(ret);
2443 } while (node && entry != last);
2444
2445 cluster->max_size = max_extent;
2446 trace_btrfs_setup_cluster(block_group, cluster, total_size, 0);
2447 return 0;
2448 }
2449
2450 /*
2451 * This specifically looks for bitmaps that may work in the cluster, we assume
2452 * that we have already failed to find extents that will work.
2453 */
2454 static noinline int
2455 setup_cluster_bitmap(struct btrfs_block_group_cache *block_group,
2456 struct btrfs_free_cluster *cluster,
2457 struct list_head *bitmaps, u64 offset, u64 bytes,
2458 u64 cont1_bytes, u64 min_bytes)
2459 {
2460 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2461 struct btrfs_free_space *entry;
2462 int ret = -ENOSPC;
2463 u64 bitmap_offset = offset_to_bitmap(ctl, offset);
2464
2465 if (ctl->total_bitmaps == 0)
2466 return -ENOSPC;
2467
2468 /*
2469 * The bitmap that covers offset won't be in the list unless offset
2470 * is just its start offset.
2471 */
2472 entry = list_first_entry(bitmaps, struct btrfs_free_space, list);
2473 if (entry->offset != bitmap_offset) {
2474 entry = tree_search_offset(ctl, bitmap_offset, 1, 0);
2475 if (entry && list_empty(&entry->list))
2476 list_add(&entry->list, bitmaps);
2477 }
2478
2479 list_for_each_entry(entry, bitmaps, list) {
2480 if (entry->bytes < bytes)
2481 continue;
2482 ret = btrfs_bitmap_cluster(block_group, entry, cluster, offset,
2483 bytes, cont1_bytes, min_bytes);
2484 if (!ret)
2485 return 0;
2486 }
2487
2488 /*
2489 * The bitmaps list has all the bitmaps that record free space
2490 * starting after offset, so no more search is required.
2491 */
2492 return -ENOSPC;
2493 }
2494
2495 /*
2496 * here we try to find a cluster of blocks in a block group. The goal
2497 * is to find at least bytes+empty_size.
2498 * We might not find them all in one contiguous area.
2499 *
2500 * returns zero and sets up cluster if things worked out, otherwise
2501 * it returns -enospc
2502 */
2503 int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
2504 struct btrfs_root *root,
2505 struct btrfs_block_group_cache *block_group,
2506 struct btrfs_free_cluster *cluster,
2507 u64 offset, u64 bytes, u64 empty_size)
2508 {
2509 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2510 struct btrfs_free_space *entry, *tmp;
2511 LIST_HEAD(bitmaps);
2512 u64 min_bytes;
2513 u64 cont1_bytes;
2514 int ret;
2515
2516 /*
2517 * Choose the minimum extent size we'll require for this
2518 * cluster. For SSD_SPREAD, don't allow any fragmentation.
2519 * For metadata, allow allocates with smaller extents. For
2520 * data, keep it dense.
2521 */
2522 if (btrfs_test_opt(root, SSD_SPREAD)) {
2523 cont1_bytes = min_bytes = bytes + empty_size;
2524 } else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
2525 cont1_bytes = bytes;
2526 min_bytes = block_group->sectorsize;
2527 } else {
2528 cont1_bytes = max(bytes, (bytes + empty_size) >> 2);
2529 min_bytes = block_group->sectorsize;
2530 }
2531
2532 spin_lock(&ctl->tree_lock);
2533
2534 /*
2535 * If we know we don't have enough space to make a cluster don't even
2536 * bother doing all the work to try and find one.
2537 */
2538 if (ctl->free_space < bytes) {
2539 spin_unlock(&ctl->tree_lock);
2540 return -ENOSPC;
2541 }
2542
2543 spin_lock(&cluster->lock);
2544
2545 /* someone already found a cluster, hooray */
2546 if (cluster->block_group) {
2547 ret = 0;
2548 goto out;
2549 }
2550
2551 trace_btrfs_find_cluster(block_group, offset, bytes, empty_size,
2552 min_bytes);
2553
2554 INIT_LIST_HEAD(&bitmaps);
2555 ret = setup_cluster_no_bitmap(block_group, cluster, &bitmaps, offset,
2556 bytes + empty_size,
2557 cont1_bytes, min_bytes);
2558 if (ret)
2559 ret = setup_cluster_bitmap(block_group, cluster, &bitmaps,
2560 offset, bytes + empty_size,
2561 cont1_bytes, min_bytes);
2562
2563 /* Clear our temporary list */
2564 list_for_each_entry_safe(entry, tmp, &bitmaps, list)
2565 list_del_init(&entry->list);
2566
2567 if (!ret) {
2568 atomic_inc(&block_group->count);
2569 list_add_tail(&cluster->block_group_list,
2570 &block_group->cluster_list);
2571 cluster->block_group = block_group;
2572 } else {
2573 trace_btrfs_failed_cluster_setup(block_group);
2574 }
2575 out:
2576 spin_unlock(&cluster->lock);
2577 spin_unlock(&ctl->tree_lock);
2578
2579 return ret;
2580 }
2581
2582 /*
2583 * simple code to zero out a cluster
2584 */
2585 void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster)
2586 {
2587 spin_lock_init(&cluster->lock);
2588 spin_lock_init(&cluster->refill_lock);
2589 cluster->root = RB_ROOT;
2590 cluster->max_size = 0;
2591 INIT_LIST_HEAD(&cluster->block_group_list);
2592 cluster->block_group = NULL;
2593 }
2594
2595 static int do_trimming(struct btrfs_block_group_cache *block_group,
2596 u64 *total_trimmed, u64 start, u64 bytes,
2597 u64 reserved_start, u64 reserved_bytes)
2598 {
2599 struct btrfs_space_info *space_info = block_group->space_info;
2600 struct btrfs_fs_info *fs_info = block_group->fs_info;
2601 int ret;
2602 int update = 0;
2603 u64 trimmed = 0;
2604
2605 spin_lock(&space_info->lock);
2606 spin_lock(&block_group->lock);
2607 if (!block_group->ro) {
2608 block_group->reserved += reserved_bytes;
2609 space_info->bytes_reserved += reserved_bytes;
2610 update = 1;
2611 }
2612 spin_unlock(&block_group->lock);
2613 spin_unlock(&space_info->lock);
2614
2615 ret = btrfs_error_discard_extent(fs_info->extent_root,
2616 start, bytes, &trimmed);
2617 if (!ret)
2618 *total_trimmed += trimmed;
2619
2620 btrfs_add_free_space(block_group, reserved_start, reserved_bytes);
2621
2622 if (update) {
2623 spin_lock(&space_info->lock);
2624 spin_lock(&block_group->lock);
2625 if (block_group->ro)
2626 space_info->bytes_readonly += reserved_bytes;
2627 block_group->reserved -= reserved_bytes;
2628 space_info->bytes_reserved -= reserved_bytes;
2629 spin_unlock(&space_info->lock);
2630 spin_unlock(&block_group->lock);
2631 }
2632
2633 return ret;
2634 }
2635
2636 static int trim_no_bitmap(struct btrfs_block_group_cache *block_group,
2637 u64 *total_trimmed, u64 start, u64 end, u64 minlen)
2638 {
2639 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2640 struct btrfs_free_space *entry;
2641 struct rb_node *node;
2642 int ret = 0;
2643 u64 extent_start;
2644 u64 extent_bytes;
2645 u64 bytes;
2646
2647 while (start < end) {
2648 spin_lock(&ctl->tree_lock);
2649
2650 if (ctl->free_space < minlen) {
2651 spin_unlock(&ctl->tree_lock);
2652 break;
2653 }
2654
2655 entry = tree_search_offset(ctl, start, 0, 1);
2656 if (!entry) {
2657 spin_unlock(&ctl->tree_lock);
2658 break;
2659 }
2660
2661 /* skip bitmaps */
2662 while (entry->bitmap) {
2663 node = rb_next(&entry->offset_index);
2664 if (!node) {
2665 spin_unlock(&ctl->tree_lock);
2666 goto out;
2667 }
2668 entry = rb_entry(node, struct btrfs_free_space,
2669 offset_index);
2670 }
2671
2672 if (entry->offset >= end) {
2673 spin_unlock(&ctl->tree_lock);
2674 break;
2675 }
2676
2677 extent_start = entry->offset;
2678 extent_bytes = entry->bytes;
2679 start = max(start, extent_start);
2680 bytes = min(extent_start + extent_bytes, end) - start;
2681 if (bytes < minlen) {
2682 spin_unlock(&ctl->tree_lock);
2683 goto next;
2684 }
2685
2686 unlink_free_space(ctl, entry);
2687 kmem_cache_free(btrfs_free_space_cachep, entry);
2688
2689 spin_unlock(&ctl->tree_lock);
2690
2691 ret = do_trimming(block_group, total_trimmed, start, bytes,
2692 extent_start, extent_bytes);
2693 if (ret)
2694 break;
2695 next:
2696 start += bytes;
2697
2698 if (fatal_signal_pending(current)) {
2699 ret = -ERESTARTSYS;
2700 break;
2701 }
2702
2703 cond_resched();
2704 }
2705 out:
2706 return ret;
2707 }
2708
2709 static int trim_bitmaps(struct btrfs_block_group_cache *block_group,
2710 u64 *total_trimmed, u64 start, u64 end, u64 minlen)
2711 {
2712 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2713 struct btrfs_free_space *entry;
2714 int ret = 0;
2715 int ret2;
2716 u64 bytes;
2717 u64 offset = offset_to_bitmap(ctl, start);
2718
2719 while (offset < end) {
2720 bool next_bitmap = false;
2721
2722 spin_lock(&ctl->tree_lock);
2723
2724 if (ctl->free_space < minlen) {
2725 spin_unlock(&ctl->tree_lock);
2726 break;
2727 }
2728
2729 entry = tree_search_offset(ctl, offset, 1, 0);
2730 if (!entry) {
2731 spin_unlock(&ctl->tree_lock);
2732 next_bitmap = true;
2733 goto next;
2734 }
2735
2736 bytes = minlen;
2737 ret2 = search_bitmap(ctl, entry, &start, &bytes);
2738 if (ret2 || start >= end) {
2739 spin_unlock(&ctl->tree_lock);
2740 next_bitmap = true;
2741 goto next;
2742 }
2743
2744 bytes = min(bytes, end - start);
2745 if (bytes < minlen) {
2746 spin_unlock(&ctl->tree_lock);
2747 goto next;
2748 }
2749
2750 bitmap_clear_bits(ctl, entry, start, bytes);
2751 if (entry->bytes == 0)
2752 free_bitmap(ctl, entry);
2753
2754 spin_unlock(&ctl->tree_lock);
2755
2756 ret = do_trimming(block_group, total_trimmed, start, bytes,
2757 start, bytes);
2758 if (ret)
2759 break;
2760 next:
2761 if (next_bitmap) {
2762 offset += BITS_PER_BITMAP * ctl->unit;
2763 } else {
2764 start += bytes;
2765 if (start >= offset + BITS_PER_BITMAP * ctl->unit)
2766 offset += BITS_PER_BITMAP * ctl->unit;
2767 }
2768
2769 if (fatal_signal_pending(current)) {
2770 ret = -ERESTARTSYS;
2771 break;
2772 }
2773
2774 cond_resched();
2775 }
2776
2777 return ret;
2778 }
2779
2780 int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group,
2781 u64 *trimmed, u64 start, u64 end, u64 minlen)
2782 {
2783 int ret;
2784
2785 *trimmed = 0;
2786
2787 ret = trim_no_bitmap(block_group, trimmed, start, end, minlen);
2788 if (ret)
2789 return ret;
2790
2791 ret = trim_bitmaps(block_group, trimmed, start, end, minlen);
2792
2793 return ret;
2794 }
2795
2796 /*
2797 * Find the left-most item in the cache tree, and then return the
2798 * smallest inode number in the item.
2799 *
2800 * Note: the returned inode number may not be the smallest one in
2801 * the tree, if the left-most item is a bitmap.
2802 */
2803 u64 btrfs_find_ino_for_alloc(struct btrfs_root *fs_root)
2804 {
2805 struct btrfs_free_space_ctl *ctl = fs_root->free_ino_ctl;
2806 struct btrfs_free_space *entry = NULL;
2807 u64 ino = 0;
2808
2809 spin_lock(&ctl->tree_lock);
2810
2811 if (RB_EMPTY_ROOT(&ctl->free_space_offset))
2812 goto out;
2813
2814 entry = rb_entry(rb_first(&ctl->free_space_offset),
2815 struct btrfs_free_space, offset_index);
2816
2817 if (!entry->bitmap) {
2818 ino = entry->offset;
2819
2820 unlink_free_space(ctl, entry);
2821 entry->offset++;
2822 entry->bytes--;
2823 if (!entry->bytes)
2824 kmem_cache_free(btrfs_free_space_cachep, entry);
2825 else
2826 link_free_space(ctl, entry);
2827 } else {
2828 u64 offset = 0;
2829 u64 count = 1;
2830 int ret;
2831
2832 ret = search_bitmap(ctl, entry, &offset, &count);
2833 BUG_ON(ret);
2834
2835 ino = offset;
2836 bitmap_clear_bits(ctl, entry, offset, 1);
2837 if (entry->bytes == 0)
2838 free_bitmap(ctl, entry);
2839 }
2840 out:
2841 spin_unlock(&ctl->tree_lock);
2842
2843 return ino;
2844 }
2845
2846 struct inode *lookup_free_ino_inode(struct btrfs_root *root,
2847 struct btrfs_path *path)
2848 {
2849 struct inode *inode = NULL;
2850
2851 spin_lock(&root->cache_lock);
2852 if (root->cache_inode)
2853 inode = igrab(root->cache_inode);
2854 spin_unlock(&root->cache_lock);
2855 if (inode)
2856 return inode;
2857
2858 inode = __lookup_free_space_inode(root, path, 0);
2859 if (IS_ERR(inode))
2860 return inode;
2861
2862 spin_lock(&root->cache_lock);
2863 if (!btrfs_fs_closing(root->fs_info))
2864 root->cache_inode = igrab(inode);
2865 spin_unlock(&root->cache_lock);
2866
2867 return inode;
2868 }
2869
2870 int create_free_ino_inode(struct btrfs_root *root,
2871 struct btrfs_trans_handle *trans,
2872 struct btrfs_path *path)
2873 {
2874 return __create_free_space_inode(root, trans, path,
2875 BTRFS_FREE_INO_OBJECTID, 0);
2876 }
2877
2878 int load_free_ino_cache(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
2879 {
2880 struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
2881 struct btrfs_path *path;
2882 struct inode *inode;
2883 int ret = 0;
2884 u64 root_gen = btrfs_root_generation(&root->root_item);
2885
2886 if (!btrfs_test_opt(root, INODE_MAP_CACHE))
2887 return 0;
2888
2889 /*
2890 * If we're unmounting then just return, since this does a search on the
2891 * normal root and not the commit root and we could deadlock.
2892 */
2893 if (btrfs_fs_closing(fs_info))
2894 return 0;
2895
2896 path = btrfs_alloc_path();
2897 if (!path)
2898 return 0;
2899
2900 inode = lookup_free_ino_inode(root, path);
2901 if (IS_ERR(inode))
2902 goto out;
2903
2904 if (root_gen != BTRFS_I(inode)->generation)
2905 goto out_put;
2906
2907 ret = __load_free_space_cache(root, inode, ctl, path, 0);
2908
2909 if (ret < 0)
2910 printk(KERN_ERR "btrfs: failed to load free ino cache for "
2911 "root %llu\n", root->root_key.objectid);
2912 out_put:
2913 iput(inode);
2914 out:
2915 btrfs_free_path(path);
2916 return ret;
2917 }
2918
2919 int btrfs_write_out_ino_cache(struct btrfs_root *root,
2920 struct btrfs_trans_handle *trans,
2921 struct btrfs_path *path)
2922 {
2923 struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
2924 struct inode *inode;
2925 int ret;
2926
2927 if (!btrfs_test_opt(root, INODE_MAP_CACHE))
2928 return 0;
2929
2930 inode = lookup_free_ino_inode(root, path);
2931 if (IS_ERR(inode))
2932 return 0;
2933
2934 ret = __btrfs_write_out_cache(root, inode, ctl, NULL, trans, path, 0);
2935 if (ret) {
2936 btrfs_delalloc_release_metadata(inode, inode->i_size);
2937 #ifdef DEBUG
2938 printk(KERN_ERR "btrfs: failed to write free ino cache "
2939 "for root %llu\n", root->root_key.objectid);
2940 #endif
2941 }
2942
2943 iput(inode);
2944 return ret;
2945 }