]> git.openfabrics.org - ~shefty/rdma-dev.git/blob - fs/btrfs/free-space-cache.c
22ee0dc2e6b8a712900c68a8c44df4a453fb86f0
[~shefty/rdma-dev.git] / fs / btrfs / free-space-cache.c
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 "ctree.h"
24 #include "free-space-cache.h"
25 #include "transaction.h"
26 #include "disk-io.h"
27
28 #define BITS_PER_BITMAP         (PAGE_CACHE_SIZE * 8)
29 #define MAX_CACHE_BYTES_PER_GIG (32 * 1024)
30
31 static void recalculate_thresholds(struct btrfs_block_group_cache
32                                    *block_group);
33 static int link_free_space(struct btrfs_block_group_cache *block_group,
34                            struct btrfs_free_space *info);
35
36 struct inode *lookup_free_space_inode(struct btrfs_root *root,
37                                       struct btrfs_block_group_cache
38                                       *block_group, struct btrfs_path *path)
39 {
40         struct btrfs_key key;
41         struct btrfs_key location;
42         struct btrfs_disk_key disk_key;
43         struct btrfs_free_space_header *header;
44         struct extent_buffer *leaf;
45         struct inode *inode = NULL;
46         int ret;
47
48         spin_lock(&block_group->lock);
49         if (block_group->inode)
50                 inode = igrab(block_group->inode);
51         spin_unlock(&block_group->lock);
52         if (inode)
53                 return inode;
54
55         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
56         key.offset = block_group->key.objectid;
57         key.type = 0;
58
59         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
60         if (ret < 0)
61                 return ERR_PTR(ret);
62         if (ret > 0) {
63                 btrfs_release_path(root, path);
64                 return ERR_PTR(-ENOENT);
65         }
66
67         leaf = path->nodes[0];
68         header = btrfs_item_ptr(leaf, path->slots[0],
69                                 struct btrfs_free_space_header);
70         btrfs_free_space_key(leaf, header, &disk_key);
71         btrfs_disk_key_to_cpu(&location, &disk_key);
72         btrfs_release_path(root, path);
73
74         inode = btrfs_iget(root->fs_info->sb, &location, root, NULL);
75         if (!inode)
76                 return ERR_PTR(-ENOENT);
77         if (IS_ERR(inode))
78                 return inode;
79         if (is_bad_inode(inode)) {
80                 iput(inode);
81                 return ERR_PTR(-ENOENT);
82         }
83
84         spin_lock(&block_group->lock);
85         if (!root->fs_info->closing) {
86                 block_group->inode = igrab(inode);
87                 block_group->iref = 1;
88         }
89         spin_unlock(&block_group->lock);
90
91         return inode;
92 }
93
94 int create_free_space_inode(struct btrfs_root *root,
95                             struct btrfs_trans_handle *trans,
96                             struct btrfs_block_group_cache *block_group,
97                             struct btrfs_path *path)
98 {
99         struct btrfs_key key;
100         struct btrfs_disk_key disk_key;
101         struct btrfs_free_space_header *header;
102         struct btrfs_inode_item *inode_item;
103         struct extent_buffer *leaf;
104         u64 objectid;
105         int ret;
106
107         ret = btrfs_find_free_objectid(trans, root, 0, &objectid);
108         if (ret < 0)
109                 return ret;
110
111         ret = btrfs_insert_empty_inode(trans, root, path, objectid);
112         if (ret)
113                 return ret;
114
115         leaf = path->nodes[0];
116         inode_item = btrfs_item_ptr(leaf, path->slots[0],
117                                     struct btrfs_inode_item);
118         btrfs_item_key(leaf, &disk_key, path->slots[0]);
119         memset_extent_buffer(leaf, 0, (unsigned long)inode_item,
120                              sizeof(*inode_item));
121         btrfs_set_inode_generation(leaf, inode_item, trans->transid);
122         btrfs_set_inode_size(leaf, inode_item, 0);
123         btrfs_set_inode_nbytes(leaf, inode_item, 0);
124         btrfs_set_inode_uid(leaf, inode_item, 0);
125         btrfs_set_inode_gid(leaf, inode_item, 0);
126         btrfs_set_inode_mode(leaf, inode_item, S_IFREG | 0600);
127         btrfs_set_inode_flags(leaf, inode_item, BTRFS_INODE_NOCOMPRESS |
128                               BTRFS_INODE_PREALLOC | BTRFS_INODE_NODATASUM);
129         btrfs_set_inode_nlink(leaf, inode_item, 1);
130         btrfs_set_inode_transid(leaf, inode_item, trans->transid);
131         btrfs_set_inode_block_group(leaf, inode_item,
132                                     block_group->key.objectid);
133         btrfs_mark_buffer_dirty(leaf);
134         btrfs_release_path(root, path);
135
136         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
137         key.offset = block_group->key.objectid;
138         key.type = 0;
139
140         ret = btrfs_insert_empty_item(trans, root, path, &key,
141                                       sizeof(struct btrfs_free_space_header));
142         if (ret < 0) {
143                 btrfs_release_path(root, path);
144                 return ret;
145         }
146         leaf = path->nodes[0];
147         header = btrfs_item_ptr(leaf, path->slots[0],
148                                 struct btrfs_free_space_header);
149         memset_extent_buffer(leaf, 0, (unsigned long)header, sizeof(*header));
150         btrfs_set_free_space_key(leaf, header, &disk_key);
151         btrfs_mark_buffer_dirty(leaf);
152         btrfs_release_path(root, path);
153
154         return 0;
155 }
156
157 int btrfs_truncate_free_space_cache(struct btrfs_root *root,
158                                     struct btrfs_trans_handle *trans,
159                                     struct btrfs_path *path,
160                                     struct inode *inode)
161 {
162         loff_t oldsize;
163         int ret = 0;
164
165         trans->block_rsv = root->orphan_block_rsv;
166         ret = btrfs_block_rsv_check(trans, root,
167                                     root->orphan_block_rsv,
168                                     0, 5);
169         if (ret)
170                 return ret;
171
172         oldsize = i_size_read(inode);
173         btrfs_i_size_write(inode, 0);
174         truncate_pagecache(inode, oldsize, 0);
175
176         /*
177          * We don't need an orphan item because truncating the free space cache
178          * will never be split across transactions.
179          */
180         ret = btrfs_truncate_inode_items(trans, root, inode,
181                                          0, BTRFS_EXTENT_DATA_KEY);
182         if (ret) {
183                 WARN_ON(1);
184                 return ret;
185         }
186
187         return btrfs_update_inode(trans, root, inode);
188 }
189
190 static int readahead_cache(struct inode *inode)
191 {
192         struct file_ra_state *ra;
193         unsigned long last_index;
194
195         ra = kzalloc(sizeof(*ra), GFP_NOFS);
196         if (!ra)
197                 return -ENOMEM;
198
199         file_ra_state_init(ra, inode->i_mapping);
200         last_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;
201
202         page_cache_sync_readahead(inode->i_mapping, ra, NULL, 0, last_index);
203
204         kfree(ra);
205
206         return 0;
207 }
208
209 int load_free_space_cache(struct btrfs_fs_info *fs_info,
210                           struct btrfs_block_group_cache *block_group)
211 {
212         struct btrfs_root *root = fs_info->tree_root;
213         struct inode *inode;
214         struct btrfs_free_space_header *header;
215         struct extent_buffer *leaf;
216         struct page *page;
217         struct btrfs_path *path;
218         u32 *checksums = NULL, *crc;
219         char *disk_crcs = NULL;
220         struct btrfs_key key;
221         struct list_head bitmaps;
222         u64 num_entries;
223         u64 num_bitmaps;
224         u64 generation;
225         u32 cur_crc = ~(u32)0;
226         pgoff_t index = 0;
227         unsigned long first_page_offset;
228         int num_checksums;
229         int ret = 0;
230
231         /*
232          * If we're unmounting then just return, since this does a search on the
233          * normal root and not the commit root and we could deadlock.
234          */
235         smp_mb();
236         if (fs_info->closing)
237                 return 0;
238
239         /*
240          * If this block group has been marked to be cleared for one reason or
241          * another then we can't trust the on disk cache, so just return.
242          */
243         spin_lock(&block_group->lock);
244         if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
245                 spin_unlock(&block_group->lock);
246                 return 0;
247         }
248         spin_unlock(&block_group->lock);
249
250         INIT_LIST_HEAD(&bitmaps);
251
252         path = btrfs_alloc_path();
253         if (!path)
254                 return 0;
255
256         inode = lookup_free_space_inode(root, block_group, path);
257         if (IS_ERR(inode)) {
258                 btrfs_free_path(path);
259                 return 0;
260         }
261
262         /* Nothing in the space cache, goodbye */
263         if (!i_size_read(inode)) {
264                 btrfs_free_path(path);
265                 goto out;
266         }
267
268         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
269         key.offset = block_group->key.objectid;
270         key.type = 0;
271
272         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
273         if (ret) {
274                 btrfs_free_path(path);
275                 goto out;
276         }
277
278         leaf = path->nodes[0];
279         header = btrfs_item_ptr(leaf, path->slots[0],
280                                 struct btrfs_free_space_header);
281         num_entries = btrfs_free_space_entries(leaf, header);
282         num_bitmaps = btrfs_free_space_bitmaps(leaf, header);
283         generation = btrfs_free_space_generation(leaf, header);
284         btrfs_free_path(path);
285
286         if (BTRFS_I(inode)->generation != generation) {
287                 printk(KERN_ERR "btrfs: free space inode generation (%llu) did"
288                        " not match free space cache generation (%llu) for "
289                        "block group %llu\n",
290                        (unsigned long long)BTRFS_I(inode)->generation,
291                        (unsigned long long)generation,
292                        (unsigned long long)block_group->key.objectid);
293                 goto out;
294         }
295
296         if (!num_entries)
297                 goto out;
298
299         /* Setup everything for doing checksumming */
300         num_checksums = i_size_read(inode) / PAGE_CACHE_SIZE;
301         checksums = crc = kzalloc(sizeof(u32) * num_checksums, GFP_NOFS);
302         if (!checksums)
303                 goto out;
304         first_page_offset = (sizeof(u32) * num_checksums) + sizeof(u64);
305         disk_crcs = kzalloc(first_page_offset, GFP_NOFS);
306         if (!disk_crcs)
307                 goto out;
308
309         ret = readahead_cache(inode);
310         if (ret) {
311                 ret = 0;
312                 goto out;
313         }
314
315         while (1) {
316                 struct btrfs_free_space_entry *entry;
317                 struct btrfs_free_space *e;
318                 void *addr;
319                 unsigned long offset = 0;
320                 unsigned long start_offset = 0;
321                 int need_loop = 0;
322
323                 if (!num_entries && !num_bitmaps)
324                         break;
325
326                 if (index == 0) {
327                         start_offset = first_page_offset;
328                         offset = start_offset;
329                 }
330
331                 page = grab_cache_page(inode->i_mapping, index);
332                 if (!page) {
333                         ret = 0;
334                         goto free_cache;
335                 }
336
337                 if (!PageUptodate(page)) {
338                         btrfs_readpage(NULL, page);
339                         lock_page(page);
340                         if (!PageUptodate(page)) {
341                                 unlock_page(page);
342                                 page_cache_release(page);
343                                 printk(KERN_ERR "btrfs: error reading free "
344                                        "space cache: %llu\n",
345                                        (unsigned long long)
346                                        block_group->key.objectid);
347                                 goto free_cache;
348                         }
349                 }
350                 addr = kmap(page);
351
352                 if (index == 0) {
353                         u64 *gen;
354
355                         memcpy(disk_crcs, addr, first_page_offset);
356                         gen = addr + (sizeof(u32) * num_checksums);
357                         if (*gen != BTRFS_I(inode)->generation) {
358                                 printk(KERN_ERR "btrfs: space cache generation"
359                                        " (%llu) does not match inode (%llu) "
360                                        "for block group %llu\n",
361                                        (unsigned long long)*gen,
362                                        (unsigned long long)
363                                        BTRFS_I(inode)->generation,
364                                        (unsigned long long)
365                                        block_group->key.objectid);
366                                 kunmap(page);
367                                 unlock_page(page);
368                                 page_cache_release(page);
369                                 goto free_cache;
370                         }
371                         crc = (u32 *)disk_crcs;
372                 }
373                 entry = addr + start_offset;
374
375                 /* First lets check our crc before we do anything fun */
376                 cur_crc = ~(u32)0;
377                 cur_crc = btrfs_csum_data(root, addr + start_offset, cur_crc,
378                                           PAGE_CACHE_SIZE - start_offset);
379                 btrfs_csum_final(cur_crc, (char *)&cur_crc);
380                 if (cur_crc != *crc) {
381                         printk(KERN_ERR "btrfs: crc mismatch for page %lu in "
382                                "block group %llu\n", index,
383                                (unsigned long long)block_group->key.objectid);
384                         kunmap(page);
385                         unlock_page(page);
386                         page_cache_release(page);
387                         goto free_cache;
388                 }
389                 crc++;
390
391                 while (1) {
392                         if (!num_entries)
393                                 break;
394
395                         need_loop = 1;
396                         e = kzalloc(sizeof(struct btrfs_free_space), GFP_NOFS);
397                         if (!e) {
398                                 kunmap(page);
399                                 unlock_page(page);
400                                 page_cache_release(page);
401                                 goto free_cache;
402                         }
403
404                         e->offset = le64_to_cpu(entry->offset);
405                         e->bytes = le64_to_cpu(entry->bytes);
406                         if (!e->bytes) {
407                                 kunmap(page);
408                                 kfree(e);
409                                 unlock_page(page);
410                                 page_cache_release(page);
411                                 goto free_cache;
412                         }
413
414                         if (entry->type == BTRFS_FREE_SPACE_EXTENT) {
415                                 spin_lock(&block_group->tree_lock);
416                                 ret = link_free_space(block_group, e);
417                                 spin_unlock(&block_group->tree_lock);
418                                 BUG_ON(ret);
419                         } else {
420                                 e->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
421                                 if (!e->bitmap) {
422                                         kunmap(page);
423                                         kfree(e);
424                                         unlock_page(page);
425                                         page_cache_release(page);
426                                         goto free_cache;
427                                 }
428                                 spin_lock(&block_group->tree_lock);
429                                 ret = link_free_space(block_group, e);
430                                 block_group->total_bitmaps++;
431                                 recalculate_thresholds(block_group);
432                                 spin_unlock(&block_group->tree_lock);
433                                 list_add_tail(&e->list, &bitmaps);
434                         }
435
436                         num_entries--;
437                         offset += sizeof(struct btrfs_free_space_entry);
438                         if (offset + sizeof(struct btrfs_free_space_entry) >=
439                             PAGE_CACHE_SIZE)
440                                 break;
441                         entry++;
442                 }
443
444                 /*
445                  * We read an entry out of this page, we need to move on to the
446                  * next page.
447                  */
448                 if (need_loop) {
449                         kunmap(page);
450                         goto next;
451                 }
452
453                 /*
454                  * We add the bitmaps at the end of the entries in order that
455                  * the bitmap entries are added to the cache.
456                  */
457                 e = list_entry(bitmaps.next, struct btrfs_free_space, list);
458                 list_del_init(&e->list);
459                 memcpy(e->bitmap, addr, PAGE_CACHE_SIZE);
460                 kunmap(page);
461                 num_bitmaps--;
462 next:
463                 unlock_page(page);
464                 page_cache_release(page);
465                 index++;
466         }
467
468         ret = 1;
469 out:
470         kfree(checksums);
471         kfree(disk_crcs);
472         iput(inode);
473         return ret;
474
475 free_cache:
476         /* This cache is bogus, make sure it gets cleared */
477         spin_lock(&block_group->lock);
478         block_group->disk_cache_state = BTRFS_DC_CLEAR;
479         spin_unlock(&block_group->lock);
480         btrfs_remove_free_space_cache(block_group);
481         goto out;
482 }
483
484 int btrfs_write_out_cache(struct btrfs_root *root,
485                           struct btrfs_trans_handle *trans,
486                           struct btrfs_block_group_cache *block_group,
487                           struct btrfs_path *path)
488 {
489         struct btrfs_free_space_header *header;
490         struct extent_buffer *leaf;
491         struct inode *inode;
492         struct rb_node *node;
493         struct list_head *pos, *n;
494         struct page *page;
495         struct extent_state *cached_state = NULL;
496         struct list_head bitmap_list;
497         struct btrfs_key key;
498         u64 bytes = 0;
499         u32 *crc, *checksums;
500         pgoff_t index = 0, last_index = 0;
501         unsigned long first_page_offset;
502         int num_checksums;
503         int entries = 0;
504         int bitmaps = 0;
505         int ret = 0;
506
507         root = root->fs_info->tree_root;
508
509         INIT_LIST_HEAD(&bitmap_list);
510
511         spin_lock(&block_group->lock);
512         if (block_group->disk_cache_state < BTRFS_DC_SETUP) {
513                 spin_unlock(&block_group->lock);
514                 return 0;
515         }
516         spin_unlock(&block_group->lock);
517
518         inode = lookup_free_space_inode(root, block_group, path);
519         if (IS_ERR(inode))
520                 return 0;
521
522         if (!i_size_read(inode)) {
523                 iput(inode);
524                 return 0;
525         }
526
527         last_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;
528         filemap_write_and_wait(inode->i_mapping);
529         btrfs_wait_ordered_range(inode, inode->i_size &
530                                  ~(root->sectorsize - 1), (u64)-1);
531
532         /* We need a checksum per page. */
533         num_checksums = i_size_read(inode) / PAGE_CACHE_SIZE;
534         crc = checksums  = kzalloc(sizeof(u32) * num_checksums, GFP_NOFS);
535         if (!crc) {
536                 iput(inode);
537                 return 0;
538         }
539
540         /* Since the first page has all of our checksums and our generation we
541          * need to calculate the offset into the page that we can start writing
542          * our entries.
543          */
544         first_page_offset = (sizeof(u32) * num_checksums) + sizeof(u64);
545
546         node = rb_first(&block_group->free_space_offset);
547         if (!node)
548                 goto out_free;
549
550         /*
551          * Lock all pages first so we can lock the extent safely.
552          *
553          * NOTE: Because we hold the ref the entire time we're going to write to
554          * the page find_get_page should never fail, so we don't do a check
555          * after find_get_page at this point.  Just putting this here so people
556          * know and don't freak out.
557          */
558         while (index <= last_index) {
559                 page = grab_cache_page(inode->i_mapping, index);
560                 if (!page) {
561                         pgoff_t i = 0;
562
563                         while (i < index) {
564                                 page = find_get_page(inode->i_mapping, i);
565                                 unlock_page(page);
566                                 page_cache_release(page);
567                                 page_cache_release(page);
568                                 i++;
569                         }
570                         goto out_free;
571                 }
572                 index++;
573         }
574
575         index = 0;
576         lock_extent_bits(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
577                          0, &cached_state, GFP_NOFS);
578
579         /* Write out the extent entries */
580         do {
581                 struct btrfs_free_space_entry *entry;
582                 void *addr;
583                 unsigned long offset = 0;
584                 unsigned long start_offset = 0;
585
586                 if (index == 0) {
587                         start_offset = first_page_offset;
588                         offset = start_offset;
589                 }
590
591                 page = find_get_page(inode->i_mapping, index);
592
593                 addr = kmap(page);
594                 entry = addr + start_offset;
595
596                 memset(addr, 0, PAGE_CACHE_SIZE);
597                 while (1) {
598                         struct btrfs_free_space *e;
599
600                         e = rb_entry(node, struct btrfs_free_space, offset_index);
601                         entries++;
602
603                         entry->offset = cpu_to_le64(e->offset);
604                         entry->bytes = cpu_to_le64(e->bytes);
605                         if (e->bitmap) {
606                                 entry->type = BTRFS_FREE_SPACE_BITMAP;
607                                 list_add_tail(&e->list, &bitmap_list);
608                                 bitmaps++;
609                         } else {
610                                 entry->type = BTRFS_FREE_SPACE_EXTENT;
611                         }
612                         node = rb_next(node);
613                         if (!node)
614                                 break;
615                         offset += sizeof(struct btrfs_free_space_entry);
616                         if (offset + sizeof(struct btrfs_free_space_entry) >=
617                             PAGE_CACHE_SIZE)
618                                 break;
619                         entry++;
620                 }
621                 *crc = ~(u32)0;
622                 *crc = btrfs_csum_data(root, addr + start_offset, *crc,
623                                        PAGE_CACHE_SIZE - start_offset);
624                 kunmap(page);
625
626                 btrfs_csum_final(*crc, (char *)crc);
627                 crc++;
628
629                 bytes += PAGE_CACHE_SIZE;
630
631                 ClearPageChecked(page);
632                 set_page_extent_mapped(page);
633                 SetPageUptodate(page);
634                 set_page_dirty(page);
635
636                 /*
637                  * We need to release our reference we got for grab_cache_page,
638                  * except for the first page which will hold our checksums, we
639                  * do that below.
640                  */
641                 if (index != 0) {
642                         unlock_page(page);
643                         page_cache_release(page);
644                 }
645
646                 page_cache_release(page);
647
648                 index++;
649         } while (node);
650
651         /* Write out the bitmaps */
652         list_for_each_safe(pos, n, &bitmap_list) {
653                 void *addr;
654                 struct btrfs_free_space *entry =
655                         list_entry(pos, struct btrfs_free_space, list);
656
657                 page = find_get_page(inode->i_mapping, index);
658
659                 addr = kmap(page);
660                 memcpy(addr, entry->bitmap, PAGE_CACHE_SIZE);
661                 *crc = ~(u32)0;
662                 *crc = btrfs_csum_data(root, addr, *crc, PAGE_CACHE_SIZE);
663                 kunmap(page);
664                 btrfs_csum_final(*crc, (char *)crc);
665                 crc++;
666                 bytes += PAGE_CACHE_SIZE;
667
668                 ClearPageChecked(page);
669                 set_page_extent_mapped(page);
670                 SetPageUptodate(page);
671                 set_page_dirty(page);
672                 unlock_page(page);
673                 page_cache_release(page);
674                 page_cache_release(page);
675                 list_del_init(&entry->list);
676                 index++;
677         }
678
679         /* Zero out the rest of the pages just to make sure */
680         while (index <= last_index) {
681                 void *addr;
682
683                 page = find_get_page(inode->i_mapping, index);
684
685                 addr = kmap(page);
686                 memset(addr, 0, PAGE_CACHE_SIZE);
687                 kunmap(page);
688                 ClearPageChecked(page);
689                 set_page_extent_mapped(page);
690                 SetPageUptodate(page);
691                 set_page_dirty(page);
692                 unlock_page(page);
693                 page_cache_release(page);
694                 page_cache_release(page);
695                 bytes += PAGE_CACHE_SIZE;
696                 index++;
697         }
698
699         btrfs_set_extent_delalloc(inode, 0, bytes - 1, &cached_state);
700
701         /* Write the checksums and trans id to the first page */
702         {
703                 void *addr;
704                 u64 *gen;
705
706                 page = find_get_page(inode->i_mapping, 0);
707
708                 addr = kmap(page);
709                 memcpy(addr, checksums, sizeof(u32) * num_checksums);
710                 gen = addr + (sizeof(u32) * num_checksums);
711                 *gen = trans->transid;
712                 kunmap(page);
713                 ClearPageChecked(page);
714                 set_page_extent_mapped(page);
715                 SetPageUptodate(page);
716                 set_page_dirty(page);
717                 unlock_page(page);
718                 page_cache_release(page);
719                 page_cache_release(page);
720         }
721         BTRFS_I(inode)->generation = trans->transid;
722
723         unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
724                              i_size_read(inode) - 1, &cached_state, GFP_NOFS);
725
726         filemap_write_and_wait(inode->i_mapping);
727
728         key.objectid = BTRFS_FREE_SPACE_OBJECTID;
729         key.offset = block_group->key.objectid;
730         key.type = 0;
731
732         ret = btrfs_search_slot(trans, root, &key, path, 1, 1);
733         if (ret < 0) {
734                 ret = 0;
735                 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, bytes - 1,
736                                  EXTENT_DIRTY | EXTENT_DELALLOC |
737                                  EXTENT_DO_ACCOUNTING, 0, 0, NULL, GFP_NOFS);
738                 goto out_free;
739         }
740         leaf = path->nodes[0];
741         if (ret > 0) {
742                 struct btrfs_key found_key;
743                 BUG_ON(!path->slots[0]);
744                 path->slots[0]--;
745                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
746                 if (found_key.objectid != BTRFS_FREE_SPACE_OBJECTID ||
747                     found_key.offset != block_group->key.objectid) {
748                         ret = 0;
749                         clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, bytes - 1,
750                                          EXTENT_DIRTY | EXTENT_DELALLOC |
751                                          EXTENT_DO_ACCOUNTING, 0, 0, NULL,
752                                          GFP_NOFS);
753                         btrfs_release_path(root, path);
754                         goto out_free;
755                 }
756         }
757         header = btrfs_item_ptr(leaf, path->slots[0],
758                                 struct btrfs_free_space_header);
759         btrfs_set_free_space_entries(leaf, header, entries);
760         btrfs_set_free_space_bitmaps(leaf, header, bitmaps);
761         btrfs_set_free_space_generation(leaf, header, trans->transid);
762         btrfs_mark_buffer_dirty(leaf);
763         btrfs_release_path(root, path);
764
765         ret = 1;
766
767 out_free:
768         if (ret == 0) {
769                 invalidate_inode_pages2_range(inode->i_mapping, 0, index);
770                 spin_lock(&block_group->lock);
771                 block_group->disk_cache_state = BTRFS_DC_ERROR;
772                 spin_unlock(&block_group->lock);
773                 BTRFS_I(inode)->generation = 0;
774         }
775         kfree(checksums);
776         btrfs_update_inode(trans, root, inode);
777         iput(inode);
778         return ret;
779 }
780
781 static inline unsigned long offset_to_bit(u64 bitmap_start, u64 sectorsize,
782                                           u64 offset)
783 {
784         BUG_ON(offset < bitmap_start);
785         offset -= bitmap_start;
786         return (unsigned long)(div64_u64(offset, sectorsize));
787 }
788
789 static inline unsigned long bytes_to_bits(u64 bytes, u64 sectorsize)
790 {
791         return (unsigned long)(div64_u64(bytes, sectorsize));
792 }
793
794 static inline u64 offset_to_bitmap(struct btrfs_block_group_cache *block_group,
795                                    u64 offset)
796 {
797         u64 bitmap_start;
798         u64 bytes_per_bitmap;
799
800         bytes_per_bitmap = BITS_PER_BITMAP * block_group->sectorsize;
801         bitmap_start = offset - block_group->key.objectid;
802         bitmap_start = div64_u64(bitmap_start, bytes_per_bitmap);
803         bitmap_start *= bytes_per_bitmap;
804         bitmap_start += block_group->key.objectid;
805
806         return bitmap_start;
807 }
808
809 static int tree_insert_offset(struct rb_root *root, u64 offset,
810                               struct rb_node *node, int bitmap)
811 {
812         struct rb_node **p = &root->rb_node;
813         struct rb_node *parent = NULL;
814         struct btrfs_free_space *info;
815
816         while (*p) {
817                 parent = *p;
818                 info = rb_entry(parent, struct btrfs_free_space, offset_index);
819
820                 if (offset < info->offset) {
821                         p = &(*p)->rb_left;
822                 } else if (offset > info->offset) {
823                         p = &(*p)->rb_right;
824                 } else {
825                         /*
826                          * we could have a bitmap entry and an extent entry
827                          * share the same offset.  If this is the case, we want
828                          * the extent entry to always be found first if we do a
829                          * linear search through the tree, since we want to have
830                          * the quickest allocation time, and allocating from an
831                          * extent is faster than allocating from a bitmap.  So
832                          * if we're inserting a bitmap and we find an entry at
833                          * this offset, we want to go right, or after this entry
834                          * logically.  If we are inserting an extent and we've
835                          * found a bitmap, we want to go left, or before
836                          * logically.
837                          */
838                         if (bitmap) {
839                                 WARN_ON(info->bitmap);
840                                 p = &(*p)->rb_right;
841                         } else {
842                                 WARN_ON(!info->bitmap);
843                                 p = &(*p)->rb_left;
844                         }
845                 }
846         }
847
848         rb_link_node(node, parent, p);
849         rb_insert_color(node, root);
850
851         return 0;
852 }
853
854 /*
855  * searches the tree for the given offset.
856  *
857  * fuzzy - If this is set, then we are trying to make an allocation, and we just
858  * want a section that has at least bytes size and comes at or after the given
859  * offset.
860  */
861 static struct btrfs_free_space *
862 tree_search_offset(struct btrfs_block_group_cache *block_group,
863                    u64 offset, int bitmap_only, int fuzzy)
864 {
865         struct rb_node *n = block_group->free_space_offset.rb_node;
866         struct btrfs_free_space *entry, *prev = NULL;
867
868         /* find entry that is closest to the 'offset' */
869         while (1) {
870                 if (!n) {
871                         entry = NULL;
872                         break;
873                 }
874
875                 entry = rb_entry(n, struct btrfs_free_space, offset_index);
876                 prev = entry;
877
878                 if (offset < entry->offset)
879                         n = n->rb_left;
880                 else if (offset > entry->offset)
881                         n = n->rb_right;
882                 else
883                         break;
884         }
885
886         if (bitmap_only) {
887                 if (!entry)
888                         return NULL;
889                 if (entry->bitmap)
890                         return entry;
891
892                 /*
893                  * bitmap entry and extent entry may share same offset,
894                  * in that case, bitmap entry comes after extent entry.
895                  */
896                 n = rb_next(n);
897                 if (!n)
898                         return NULL;
899                 entry = rb_entry(n, struct btrfs_free_space, offset_index);
900                 if (entry->offset != offset)
901                         return NULL;
902
903                 WARN_ON(!entry->bitmap);
904                 return entry;
905         } else if (entry) {
906                 if (entry->bitmap) {
907                         /*
908                          * if previous extent entry covers the offset,
909                          * we should return it instead of the bitmap entry
910                          */
911                         n = &entry->offset_index;
912                         while (1) {
913                                 n = rb_prev(n);
914                                 if (!n)
915                                         break;
916                                 prev = rb_entry(n, struct btrfs_free_space,
917                                                 offset_index);
918                                 if (!prev->bitmap) {
919                                         if (prev->offset + prev->bytes > offset)
920                                                 entry = prev;
921                                         break;
922                                 }
923                         }
924                 }
925                 return entry;
926         }
927
928         if (!prev)
929                 return NULL;
930
931         /* find last entry before the 'offset' */
932         entry = prev;
933         if (entry->offset > offset) {
934                 n = rb_prev(&entry->offset_index);
935                 if (n) {
936                         entry = rb_entry(n, struct btrfs_free_space,
937                                         offset_index);
938                         BUG_ON(entry->offset > offset);
939                 } else {
940                         if (fuzzy)
941                                 return entry;
942                         else
943                                 return NULL;
944                 }
945         }
946
947         if (entry->bitmap) {
948                 n = &entry->offset_index;
949                 while (1) {
950                         n = rb_prev(n);
951                         if (!n)
952                                 break;
953                         prev = rb_entry(n, struct btrfs_free_space,
954                                         offset_index);
955                         if (!prev->bitmap) {
956                                 if (prev->offset + prev->bytes > offset)
957                                         return prev;
958                                 break;
959                         }
960                 }
961                 if (entry->offset + BITS_PER_BITMAP *
962                     block_group->sectorsize > offset)
963                         return entry;
964         } else if (entry->offset + entry->bytes > offset)
965                 return entry;
966
967         if (!fuzzy)
968                 return NULL;
969
970         while (1) {
971                 if (entry->bitmap) {
972                         if (entry->offset + BITS_PER_BITMAP *
973                             block_group->sectorsize > offset)
974                                 break;
975                 } else {
976                         if (entry->offset + entry->bytes > offset)
977                                 break;
978                 }
979
980                 n = rb_next(&entry->offset_index);
981                 if (!n)
982                         return NULL;
983                 entry = rb_entry(n, struct btrfs_free_space, offset_index);
984         }
985         return entry;
986 }
987
988 static void unlink_free_space(struct btrfs_block_group_cache *block_group,
989                               struct btrfs_free_space *info)
990 {
991         rb_erase(&info->offset_index, &block_group->free_space_offset);
992         block_group->free_extents--;
993         block_group->free_space -= info->bytes;
994 }
995
996 static int link_free_space(struct btrfs_block_group_cache *block_group,
997                            struct btrfs_free_space *info)
998 {
999         int ret = 0;
1000
1001         BUG_ON(!info->bitmap && !info->bytes);
1002         ret = tree_insert_offset(&block_group->free_space_offset, info->offset,
1003                                  &info->offset_index, (info->bitmap != NULL));
1004         if (ret)
1005                 return ret;
1006
1007         block_group->free_space += info->bytes;
1008         block_group->free_extents++;
1009         return ret;
1010 }
1011
1012 static void recalculate_thresholds(struct btrfs_block_group_cache *block_group)
1013 {
1014         u64 max_bytes;
1015         u64 bitmap_bytes;
1016         u64 extent_bytes;
1017
1018         /*
1019          * The goal is to keep the total amount of memory used per 1gb of space
1020          * at or below 32k, so we need to adjust how much memory we allow to be
1021          * used by extent based free space tracking
1022          */
1023         max_bytes = MAX_CACHE_BYTES_PER_GIG *
1024                 (div64_u64(block_group->key.offset, 1024 * 1024 * 1024));
1025
1026         /*
1027          * we want to account for 1 more bitmap than what we have so we can make
1028          * sure we don't go over our overall goal of MAX_CACHE_BYTES_PER_GIG as
1029          * we add more bitmaps.
1030          */
1031         bitmap_bytes = (block_group->total_bitmaps + 1) * PAGE_CACHE_SIZE;
1032
1033         if (bitmap_bytes >= max_bytes) {
1034                 block_group->extents_thresh = 0;
1035                 return;
1036         }
1037
1038         /*
1039          * we want the extent entry threshold to always be at most 1/2 the maxw
1040          * bytes we can have, or whatever is less than that.
1041          */
1042         extent_bytes = max_bytes - bitmap_bytes;
1043         extent_bytes = min_t(u64, extent_bytes, div64_u64(max_bytes, 2));
1044
1045         block_group->extents_thresh =
1046                 div64_u64(extent_bytes, (sizeof(struct btrfs_free_space)));
1047 }
1048
1049 static void bitmap_clear_bits(struct btrfs_block_group_cache *block_group,
1050                               struct btrfs_free_space *info, u64 offset,
1051                               u64 bytes)
1052 {
1053         unsigned long start, end;
1054         unsigned long i;
1055
1056         start = offset_to_bit(info->offset, block_group->sectorsize, offset);
1057         end = start + bytes_to_bits(bytes, block_group->sectorsize);
1058         BUG_ON(end > BITS_PER_BITMAP);
1059
1060         for (i = start; i < end; i++)
1061                 clear_bit(i, info->bitmap);
1062
1063         info->bytes -= bytes;
1064         block_group->free_space -= bytes;
1065 }
1066
1067 static void bitmap_set_bits(struct btrfs_block_group_cache *block_group,
1068                             struct btrfs_free_space *info, u64 offset,
1069                             u64 bytes)
1070 {
1071         unsigned long start, end;
1072         unsigned long i;
1073
1074         start = offset_to_bit(info->offset, block_group->sectorsize, offset);
1075         end = start + bytes_to_bits(bytes, block_group->sectorsize);
1076         BUG_ON(end > BITS_PER_BITMAP);
1077
1078         for (i = start; i < end; i++)
1079                 set_bit(i, info->bitmap);
1080
1081         info->bytes += bytes;
1082         block_group->free_space += bytes;
1083 }
1084
1085 static int search_bitmap(struct btrfs_block_group_cache *block_group,
1086                          struct btrfs_free_space *bitmap_info, u64 *offset,
1087                          u64 *bytes)
1088 {
1089         unsigned long found_bits = 0;
1090         unsigned long bits, i;
1091         unsigned long next_zero;
1092
1093         i = offset_to_bit(bitmap_info->offset, block_group->sectorsize,
1094                           max_t(u64, *offset, bitmap_info->offset));
1095         bits = bytes_to_bits(*bytes, block_group->sectorsize);
1096
1097         for (i = find_next_bit(bitmap_info->bitmap, BITS_PER_BITMAP, i);
1098              i < BITS_PER_BITMAP;
1099              i = find_next_bit(bitmap_info->bitmap, BITS_PER_BITMAP, i + 1)) {
1100                 next_zero = find_next_zero_bit(bitmap_info->bitmap,
1101                                                BITS_PER_BITMAP, i);
1102                 if ((next_zero - i) >= bits) {
1103                         found_bits = next_zero - i;
1104                         break;
1105                 }
1106                 i = next_zero;
1107         }
1108
1109         if (found_bits) {
1110                 *offset = (u64)(i * block_group->sectorsize) +
1111                         bitmap_info->offset;
1112                 *bytes = (u64)(found_bits) * block_group->sectorsize;
1113                 return 0;
1114         }
1115
1116         return -1;
1117 }
1118
1119 static struct btrfs_free_space *find_free_space(struct btrfs_block_group_cache
1120                                                 *block_group, u64 *offset,
1121                                                 u64 *bytes, int debug)
1122 {
1123         struct btrfs_free_space *entry;
1124         struct rb_node *node;
1125         int ret;
1126
1127         if (!block_group->free_space_offset.rb_node)
1128                 return NULL;
1129
1130         entry = tree_search_offset(block_group,
1131                                    offset_to_bitmap(block_group, *offset),
1132                                    0, 1);
1133         if (!entry)
1134                 return NULL;
1135
1136         for (node = &entry->offset_index; node; node = rb_next(node)) {
1137                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
1138                 if (entry->bytes < *bytes)
1139                         continue;
1140
1141                 if (entry->bitmap) {
1142                         ret = search_bitmap(block_group, entry, offset, bytes);
1143                         if (!ret)
1144                                 return entry;
1145                         continue;
1146                 }
1147
1148                 *offset = entry->offset;
1149                 *bytes = entry->bytes;
1150                 return entry;
1151         }
1152
1153         return NULL;
1154 }
1155
1156 static void add_new_bitmap(struct btrfs_block_group_cache *block_group,
1157                            struct btrfs_free_space *info, u64 offset)
1158 {
1159         u64 bytes_per_bg = BITS_PER_BITMAP * block_group->sectorsize;
1160         int max_bitmaps = (int)div64_u64(block_group->key.offset +
1161                                          bytes_per_bg - 1, bytes_per_bg);
1162         BUG_ON(block_group->total_bitmaps >= max_bitmaps);
1163
1164         info->offset = offset_to_bitmap(block_group, offset);
1165         info->bytes = 0;
1166         link_free_space(block_group, info);
1167         block_group->total_bitmaps++;
1168
1169         recalculate_thresholds(block_group);
1170 }
1171
1172 static noinline int remove_from_bitmap(struct btrfs_block_group_cache *block_group,
1173                               struct btrfs_free_space *bitmap_info,
1174                               u64 *offset, u64 *bytes)
1175 {
1176         u64 end;
1177         u64 search_start, search_bytes;
1178         int ret;
1179
1180 again:
1181         end = bitmap_info->offset +
1182                 (u64)(BITS_PER_BITMAP * block_group->sectorsize) - 1;
1183
1184         /*
1185          * XXX - this can go away after a few releases.
1186          *
1187          * since the only user of btrfs_remove_free_space is the tree logging
1188          * stuff, and the only way to test that is under crash conditions, we
1189          * want to have this debug stuff here just in case somethings not
1190          * working.  Search the bitmap for the space we are trying to use to
1191          * make sure its actually there.  If its not there then we need to stop
1192          * because something has gone wrong.
1193          */
1194         search_start = *offset;
1195         search_bytes = *bytes;
1196         ret = search_bitmap(block_group, bitmap_info, &search_start,
1197                             &search_bytes);
1198         BUG_ON(ret < 0 || search_start != *offset);
1199
1200         if (*offset > bitmap_info->offset && *offset + *bytes > end) {
1201                 bitmap_clear_bits(block_group, bitmap_info, *offset,
1202                                   end - *offset + 1);
1203                 *bytes -= end - *offset + 1;
1204                 *offset = end + 1;
1205         } else if (*offset >= bitmap_info->offset && *offset + *bytes <= end) {
1206                 bitmap_clear_bits(block_group, bitmap_info, *offset, *bytes);
1207                 *bytes = 0;
1208         }
1209
1210         if (*bytes) {
1211                 struct rb_node *next = rb_next(&bitmap_info->offset_index);
1212                 if (!bitmap_info->bytes) {
1213                         unlink_free_space(block_group, bitmap_info);
1214                         kfree(bitmap_info->bitmap);
1215                         kfree(bitmap_info);
1216                         block_group->total_bitmaps--;
1217                         recalculate_thresholds(block_group);
1218                 }
1219
1220                 /*
1221                  * no entry after this bitmap, but we still have bytes to
1222                  * remove, so something has gone wrong.
1223                  */
1224                 if (!next)
1225                         return -EINVAL;
1226
1227                 bitmap_info = rb_entry(next, struct btrfs_free_space,
1228                                        offset_index);
1229
1230                 /*
1231                  * if the next entry isn't a bitmap we need to return to let the
1232                  * extent stuff do its work.
1233                  */
1234                 if (!bitmap_info->bitmap)
1235                         return -EAGAIN;
1236
1237                 /*
1238                  * Ok the next item is a bitmap, but it may not actually hold
1239                  * the information for the rest of this free space stuff, so
1240                  * look for it, and if we don't find it return so we can try
1241                  * everything over again.
1242                  */
1243                 search_start = *offset;
1244                 search_bytes = *bytes;
1245                 ret = search_bitmap(block_group, bitmap_info, &search_start,
1246                                     &search_bytes);
1247                 if (ret < 0 || search_start != *offset)
1248                         return -EAGAIN;
1249
1250                 goto again;
1251         } else if (!bitmap_info->bytes) {
1252                 unlink_free_space(block_group, bitmap_info);
1253                 kfree(bitmap_info->bitmap);
1254                 kfree(bitmap_info);
1255                 block_group->total_bitmaps--;
1256                 recalculate_thresholds(block_group);
1257         }
1258
1259         return 0;
1260 }
1261
1262 static int insert_into_bitmap(struct btrfs_block_group_cache *block_group,
1263                               struct btrfs_free_space *info)
1264 {
1265         struct btrfs_free_space *bitmap_info;
1266         int added = 0;
1267         u64 bytes, offset, end;
1268         int ret;
1269
1270         /*
1271          * If we are below the extents threshold then we can add this as an
1272          * extent, and don't have to deal with the bitmap
1273          */
1274         if (block_group->free_extents < block_group->extents_thresh &&
1275             info->bytes > block_group->sectorsize * 4)
1276                 return 0;
1277
1278         /*
1279          * some block groups are so tiny they can't be enveloped by a bitmap, so
1280          * don't even bother to create a bitmap for this
1281          */
1282         if (BITS_PER_BITMAP * block_group->sectorsize >
1283             block_group->key.offset)
1284                 return 0;
1285
1286         bytes = info->bytes;
1287         offset = info->offset;
1288
1289 again:
1290         bitmap_info = tree_search_offset(block_group,
1291                                          offset_to_bitmap(block_group, offset),
1292                                          1, 0);
1293         if (!bitmap_info) {
1294                 BUG_ON(added);
1295                 goto new_bitmap;
1296         }
1297
1298         end = bitmap_info->offset +
1299                 (u64)(BITS_PER_BITMAP * block_group->sectorsize);
1300
1301         if (offset >= bitmap_info->offset && offset + bytes > end) {
1302                 bitmap_set_bits(block_group, bitmap_info, offset,
1303                                 end - offset);
1304                 bytes -= end - offset;
1305                 offset = end;
1306                 added = 0;
1307         } else if (offset >= bitmap_info->offset && offset + bytes <= end) {
1308                 bitmap_set_bits(block_group, bitmap_info, offset, bytes);
1309                 bytes = 0;
1310         } else {
1311                 BUG();
1312         }
1313
1314         if (!bytes) {
1315                 ret = 1;
1316                 goto out;
1317         } else
1318                 goto again;
1319
1320 new_bitmap:
1321         if (info && info->bitmap) {
1322                 add_new_bitmap(block_group, info, offset);
1323                 added = 1;
1324                 info = NULL;
1325                 goto again;
1326         } else {
1327                 spin_unlock(&block_group->tree_lock);
1328
1329                 /* no pre-allocated info, allocate a new one */
1330                 if (!info) {
1331                         info = kzalloc(sizeof(struct btrfs_free_space),
1332                                        GFP_NOFS);
1333                         if (!info) {
1334                                 spin_lock(&block_group->tree_lock);
1335                                 ret = -ENOMEM;
1336                                 goto out;
1337                         }
1338                 }
1339
1340                 /* allocate the bitmap */
1341                 info->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
1342                 spin_lock(&block_group->tree_lock);
1343                 if (!info->bitmap) {
1344                         ret = -ENOMEM;
1345                         goto out;
1346                 }
1347                 goto again;
1348         }
1349
1350 out:
1351         if (info) {
1352                 if (info->bitmap)
1353                         kfree(info->bitmap);
1354                 kfree(info);
1355         }
1356
1357         return ret;
1358 }
1359
1360 int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
1361                          u64 offset, u64 bytes)
1362 {
1363         struct btrfs_free_space *right_info = NULL;
1364         struct btrfs_free_space *left_info = NULL;
1365         struct btrfs_free_space *info = NULL;
1366         int ret = 0;
1367
1368         info = kzalloc(sizeof(struct btrfs_free_space), GFP_NOFS);
1369         if (!info)
1370                 return -ENOMEM;
1371
1372         info->offset = offset;
1373         info->bytes = bytes;
1374
1375         spin_lock(&block_group->tree_lock);
1376
1377         /*
1378          * first we want to see if there is free space adjacent to the range we
1379          * are adding, if there is remove that struct and add a new one to
1380          * cover the entire range
1381          */
1382         right_info = tree_search_offset(block_group, offset + bytes, 0, 0);
1383         if (right_info && rb_prev(&right_info->offset_index))
1384                 left_info = rb_entry(rb_prev(&right_info->offset_index),
1385                                      struct btrfs_free_space, offset_index);
1386         else
1387                 left_info = tree_search_offset(block_group, offset - 1, 0, 0);
1388
1389         /*
1390          * If there was no extent directly to the left or right of this new
1391          * extent then we know we're going to have to allocate a new extent, so
1392          * before we do that see if we need to drop this into a bitmap
1393          */
1394         if ((!left_info || left_info->bitmap) &&
1395             (!right_info || right_info->bitmap)) {
1396                 ret = insert_into_bitmap(block_group, info);
1397
1398                 if (ret < 0) {
1399                         goto out;
1400                 } else if (ret) {
1401                         ret = 0;
1402                         goto out;
1403                 }
1404         }
1405
1406         if (right_info && !right_info->bitmap) {
1407                 unlink_free_space(block_group, right_info);
1408                 info->bytes += right_info->bytes;
1409                 kfree(right_info);
1410         }
1411
1412         if (left_info && !left_info->bitmap &&
1413             left_info->offset + left_info->bytes == offset) {
1414                 unlink_free_space(block_group, left_info);
1415                 info->offset = left_info->offset;
1416                 info->bytes += left_info->bytes;
1417                 kfree(left_info);
1418         }
1419
1420         ret = link_free_space(block_group, info);
1421         if (ret)
1422                 kfree(info);
1423 out:
1424         spin_unlock(&block_group->tree_lock);
1425
1426         if (ret) {
1427                 printk(KERN_CRIT "btrfs: unable to add free space :%d\n", ret);
1428                 BUG_ON(ret == -EEXIST);
1429         }
1430
1431         return ret;
1432 }
1433
1434 int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
1435                             u64 offset, u64 bytes)
1436 {
1437         struct btrfs_free_space *info;
1438         struct btrfs_free_space *next_info = NULL;
1439         int ret = 0;
1440
1441         spin_lock(&block_group->tree_lock);
1442
1443 again:
1444         info = tree_search_offset(block_group, offset, 0, 0);
1445         if (!info) {
1446                 /*
1447                  * oops didn't find an extent that matched the space we wanted
1448                  * to remove, look for a bitmap instead
1449                  */
1450                 info = tree_search_offset(block_group,
1451                                           offset_to_bitmap(block_group, offset),
1452                                           1, 0);
1453                 if (!info) {
1454                         WARN_ON(1);
1455                         goto out_lock;
1456                 }
1457         }
1458
1459         if (info->bytes < bytes && rb_next(&info->offset_index)) {
1460                 u64 end;
1461                 next_info = rb_entry(rb_next(&info->offset_index),
1462                                              struct btrfs_free_space,
1463                                              offset_index);
1464
1465                 if (next_info->bitmap)
1466                         end = next_info->offset + BITS_PER_BITMAP *
1467                                 block_group->sectorsize - 1;
1468                 else
1469                         end = next_info->offset + next_info->bytes;
1470
1471                 if (next_info->bytes < bytes ||
1472                     next_info->offset > offset || offset > end) {
1473                         printk(KERN_CRIT "Found free space at %llu, size %llu,"
1474                               " trying to use %llu\n",
1475                               (unsigned long long)info->offset,
1476                               (unsigned long long)info->bytes,
1477                               (unsigned long long)bytes);
1478                         WARN_ON(1);
1479                         ret = -EINVAL;
1480                         goto out_lock;
1481                 }
1482
1483                 info = next_info;
1484         }
1485
1486         if (info->bytes == bytes) {
1487                 unlink_free_space(block_group, info);
1488                 if (info->bitmap) {
1489                         kfree(info->bitmap);
1490                         block_group->total_bitmaps--;
1491                 }
1492                 kfree(info);
1493                 goto out_lock;
1494         }
1495
1496         if (!info->bitmap && info->offset == offset) {
1497                 unlink_free_space(block_group, info);
1498                 info->offset += bytes;
1499                 info->bytes -= bytes;
1500                 link_free_space(block_group, info);
1501                 goto out_lock;
1502         }
1503
1504         if (!info->bitmap && info->offset <= offset &&
1505             info->offset + info->bytes >= offset + bytes) {
1506                 u64 old_start = info->offset;
1507                 /*
1508                  * we're freeing space in the middle of the info,
1509                  * this can happen during tree log replay
1510                  *
1511                  * first unlink the old info and then
1512                  * insert it again after the hole we're creating
1513                  */
1514                 unlink_free_space(block_group, info);
1515                 if (offset + bytes < info->offset + info->bytes) {
1516                         u64 old_end = info->offset + info->bytes;
1517
1518                         info->offset = offset + bytes;
1519                         info->bytes = old_end - info->offset;
1520                         ret = link_free_space(block_group, info);
1521                         WARN_ON(ret);
1522                         if (ret)
1523                                 goto out_lock;
1524                 } else {
1525                         /* the hole we're creating ends at the end
1526                          * of the info struct, just free the info
1527                          */
1528                         kfree(info);
1529                 }
1530                 spin_unlock(&block_group->tree_lock);
1531
1532                 /* step two, insert a new info struct to cover
1533                  * anything before the hole
1534                  */
1535                 ret = btrfs_add_free_space(block_group, old_start,
1536                                            offset - old_start);
1537                 WARN_ON(ret);
1538                 goto out;
1539         }
1540
1541         ret = remove_from_bitmap(block_group, info, &offset, &bytes);
1542         if (ret == -EAGAIN)
1543                 goto again;
1544         BUG_ON(ret);
1545 out_lock:
1546         spin_unlock(&block_group->tree_lock);
1547 out:
1548         return ret;
1549 }
1550
1551 void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
1552                            u64 bytes)
1553 {
1554         struct btrfs_free_space *info;
1555         struct rb_node *n;
1556         int count = 0;
1557
1558         for (n = rb_first(&block_group->free_space_offset); n; n = rb_next(n)) {
1559                 info = rb_entry(n, struct btrfs_free_space, offset_index);
1560                 if (info->bytes >= bytes)
1561                         count++;
1562                 printk(KERN_CRIT "entry offset %llu, bytes %llu, bitmap %s\n",
1563                        (unsigned long long)info->offset,
1564                        (unsigned long long)info->bytes,
1565                        (info->bitmap) ? "yes" : "no");
1566         }
1567         printk(KERN_INFO "block group has cluster?: %s\n",
1568                list_empty(&block_group->cluster_list) ? "no" : "yes");
1569         printk(KERN_INFO "%d blocks of free space at or bigger than bytes is"
1570                "\n", count);
1571 }
1572
1573 u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group)
1574 {
1575         struct btrfs_free_space *info;
1576         struct rb_node *n;
1577         u64 ret = 0;
1578
1579         for (n = rb_first(&block_group->free_space_offset); n;
1580              n = rb_next(n)) {
1581                 info = rb_entry(n, struct btrfs_free_space, offset_index);
1582                 ret += info->bytes;
1583         }
1584
1585         return ret;
1586 }
1587
1588 /*
1589  * for a given cluster, put all of its extents back into the free
1590  * space cache.  If the block group passed doesn't match the block group
1591  * pointed to by the cluster, someone else raced in and freed the
1592  * cluster already.  In that case, we just return without changing anything
1593  */
1594 static int
1595 __btrfs_return_cluster_to_free_space(
1596                              struct btrfs_block_group_cache *block_group,
1597                              struct btrfs_free_cluster *cluster)
1598 {
1599         struct btrfs_free_space *entry;
1600         struct rb_node *node;
1601         bool bitmap;
1602
1603         spin_lock(&cluster->lock);
1604         if (cluster->block_group != block_group)
1605                 goto out;
1606
1607         bitmap = cluster->points_to_bitmap;
1608         cluster->block_group = NULL;
1609         cluster->window_start = 0;
1610         list_del_init(&cluster->block_group_list);
1611         cluster->points_to_bitmap = false;
1612
1613         if (bitmap)
1614                 goto out;
1615
1616         node = rb_first(&cluster->root);
1617         while (node) {
1618                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
1619                 node = rb_next(&entry->offset_index);
1620                 rb_erase(&entry->offset_index, &cluster->root);
1621                 BUG_ON(entry->bitmap);
1622                 tree_insert_offset(&block_group->free_space_offset,
1623                                    entry->offset, &entry->offset_index, 0);
1624         }
1625         cluster->root = RB_ROOT;
1626
1627 out:
1628         spin_unlock(&cluster->lock);
1629         btrfs_put_block_group(block_group);
1630         return 0;
1631 }
1632
1633 void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
1634 {
1635         struct btrfs_free_space *info;
1636         struct rb_node *node;
1637         struct btrfs_free_cluster *cluster;
1638         struct list_head *head;
1639
1640         spin_lock(&block_group->tree_lock);
1641         while ((head = block_group->cluster_list.next) !=
1642                &block_group->cluster_list) {
1643                 cluster = list_entry(head, struct btrfs_free_cluster,
1644                                      block_group_list);
1645
1646                 WARN_ON(cluster->block_group != block_group);
1647                 __btrfs_return_cluster_to_free_space(block_group, cluster);
1648                 if (need_resched()) {
1649                         spin_unlock(&block_group->tree_lock);
1650                         cond_resched();
1651                         spin_lock(&block_group->tree_lock);
1652                 }
1653         }
1654
1655         while ((node = rb_last(&block_group->free_space_offset)) != NULL) {
1656                 info = rb_entry(node, struct btrfs_free_space, offset_index);
1657                 unlink_free_space(block_group, info);
1658                 if (info->bitmap)
1659                         kfree(info->bitmap);
1660                 kfree(info);
1661                 if (need_resched()) {
1662                         spin_unlock(&block_group->tree_lock);
1663                         cond_resched();
1664                         spin_lock(&block_group->tree_lock);
1665                 }
1666         }
1667
1668         spin_unlock(&block_group->tree_lock);
1669 }
1670
1671 u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
1672                                u64 offset, u64 bytes, u64 empty_size)
1673 {
1674         struct btrfs_free_space *entry = NULL;
1675         u64 bytes_search = bytes + empty_size;
1676         u64 ret = 0;
1677
1678         spin_lock(&block_group->tree_lock);
1679         entry = find_free_space(block_group, &offset, &bytes_search, 0);
1680         if (!entry)
1681                 goto out;
1682
1683         ret = offset;
1684         if (entry->bitmap) {
1685                 bitmap_clear_bits(block_group, entry, offset, bytes);
1686                 if (!entry->bytes) {
1687                         unlink_free_space(block_group, entry);
1688                         kfree(entry->bitmap);
1689                         kfree(entry);
1690                         block_group->total_bitmaps--;
1691                         recalculate_thresholds(block_group);
1692                 }
1693         } else {
1694                 unlink_free_space(block_group, entry);
1695                 entry->offset += bytes;
1696                 entry->bytes -= bytes;
1697                 if (!entry->bytes)
1698                         kfree(entry);
1699                 else
1700                         link_free_space(block_group, entry);
1701         }
1702
1703 out:
1704         spin_unlock(&block_group->tree_lock);
1705
1706         return ret;
1707 }
1708
1709 /*
1710  * given a cluster, put all of its extents back into the free space
1711  * cache.  If a block group is passed, this function will only free
1712  * a cluster that belongs to the passed block group.
1713  *
1714  * Otherwise, it'll get a reference on the block group pointed to by the
1715  * cluster and remove the cluster from it.
1716  */
1717 int btrfs_return_cluster_to_free_space(
1718                                struct btrfs_block_group_cache *block_group,
1719                                struct btrfs_free_cluster *cluster)
1720 {
1721         int ret;
1722
1723         /* first, get a safe pointer to the block group */
1724         spin_lock(&cluster->lock);
1725         if (!block_group) {
1726                 block_group = cluster->block_group;
1727                 if (!block_group) {
1728                         spin_unlock(&cluster->lock);
1729                         return 0;
1730                 }
1731         } else if (cluster->block_group != block_group) {
1732                 /* someone else has already freed it don't redo their work */
1733                 spin_unlock(&cluster->lock);
1734                 return 0;
1735         }
1736         atomic_inc(&block_group->count);
1737         spin_unlock(&cluster->lock);
1738
1739         /* now return any extents the cluster had on it */
1740         spin_lock(&block_group->tree_lock);
1741         ret = __btrfs_return_cluster_to_free_space(block_group, cluster);
1742         spin_unlock(&block_group->tree_lock);
1743
1744         /* finally drop our ref */
1745         btrfs_put_block_group(block_group);
1746         return ret;
1747 }
1748
1749 static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
1750                                    struct btrfs_free_cluster *cluster,
1751                                    u64 bytes, u64 min_start)
1752 {
1753         struct btrfs_free_space *entry;
1754         int err;
1755         u64 search_start = cluster->window_start;
1756         u64 search_bytes = bytes;
1757         u64 ret = 0;
1758
1759         spin_lock(&block_group->tree_lock);
1760         spin_lock(&cluster->lock);
1761
1762         if (!cluster->points_to_bitmap)
1763                 goto out;
1764
1765         if (cluster->block_group != block_group)
1766                 goto out;
1767
1768         /*
1769          * search_start is the beginning of the bitmap, but at some point it may
1770          * be a good idea to point to the actual start of the free area in the
1771          * bitmap, so do the offset_to_bitmap trick anyway, and set bitmap_only
1772          * to 1 to make sure we get the bitmap entry
1773          */
1774         entry = tree_search_offset(block_group,
1775                                    offset_to_bitmap(block_group, search_start),
1776                                    1, 0);
1777         if (!entry || !entry->bitmap)
1778                 goto out;
1779
1780         search_start = min_start;
1781         search_bytes = bytes;
1782
1783         err = search_bitmap(block_group, entry, &search_start,
1784                             &search_bytes);
1785         if (err)
1786                 goto out;
1787
1788         ret = search_start;
1789         bitmap_clear_bits(block_group, entry, ret, bytes);
1790 out:
1791         spin_unlock(&cluster->lock);
1792         spin_unlock(&block_group->tree_lock);
1793
1794         return ret;
1795 }
1796
1797 /*
1798  * given a cluster, try to allocate 'bytes' from it, returns 0
1799  * if it couldn't find anything suitably large, or a logical disk offset
1800  * if things worked out
1801  */
1802 u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
1803                              struct btrfs_free_cluster *cluster, u64 bytes,
1804                              u64 min_start)
1805 {
1806         struct btrfs_free_space *entry = NULL;
1807         struct rb_node *node;
1808         u64 ret = 0;
1809
1810         if (cluster->points_to_bitmap)
1811                 return btrfs_alloc_from_bitmap(block_group, cluster, bytes,
1812                                                min_start);
1813
1814         spin_lock(&cluster->lock);
1815         if (bytes > cluster->max_size)
1816                 goto out;
1817
1818         if (cluster->block_group != block_group)
1819                 goto out;
1820
1821         node = rb_first(&cluster->root);
1822         if (!node)
1823                 goto out;
1824
1825         entry = rb_entry(node, struct btrfs_free_space, offset_index);
1826
1827         while(1) {
1828                 if (entry->bytes < bytes || entry->offset < min_start) {
1829                         struct rb_node *node;
1830
1831                         node = rb_next(&entry->offset_index);
1832                         if (!node)
1833                                 break;
1834                         entry = rb_entry(node, struct btrfs_free_space,
1835                                          offset_index);
1836                         continue;
1837                 }
1838                 ret = entry->offset;
1839
1840                 entry->offset += bytes;
1841                 entry->bytes -= bytes;
1842
1843                 if (entry->bytes == 0) {
1844                         rb_erase(&entry->offset_index, &cluster->root);
1845                         kfree(entry);
1846                 }
1847                 break;
1848         }
1849 out:
1850         spin_unlock(&cluster->lock);
1851
1852         return ret;
1853 }
1854
1855 static int btrfs_bitmap_cluster(struct btrfs_block_group_cache *block_group,
1856                                 struct btrfs_free_space *entry,
1857                                 struct btrfs_free_cluster *cluster,
1858                                 u64 offset, u64 bytes, u64 min_bytes)
1859 {
1860         unsigned long next_zero;
1861         unsigned long i;
1862         unsigned long search_bits;
1863         unsigned long total_bits;
1864         unsigned long found_bits;
1865         unsigned long start = 0;
1866         unsigned long total_found = 0;
1867         bool found = false;
1868
1869         i = offset_to_bit(entry->offset, block_group->sectorsize,
1870                           max_t(u64, offset, entry->offset));
1871         search_bits = bytes_to_bits(min_bytes, block_group->sectorsize);
1872         total_bits = bytes_to_bits(bytes, block_group->sectorsize);
1873
1874 again:
1875         found_bits = 0;
1876         for (i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, i);
1877              i < BITS_PER_BITMAP;
1878              i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, i + 1)) {
1879                 next_zero = find_next_zero_bit(entry->bitmap,
1880                                                BITS_PER_BITMAP, i);
1881                 if (next_zero - i >= search_bits) {
1882                         found_bits = next_zero - i;
1883                         break;
1884                 }
1885                 i = next_zero;
1886         }
1887
1888         if (!found_bits)
1889                 return -1;
1890
1891         if (!found) {
1892                 start = i;
1893                 found = true;
1894         }
1895
1896         total_found += found_bits;
1897
1898         if (cluster->max_size < found_bits * block_group->sectorsize)
1899                 cluster->max_size = found_bits * block_group->sectorsize;
1900
1901         if (total_found < total_bits) {
1902                 i = find_next_bit(entry->bitmap, BITS_PER_BITMAP, next_zero);
1903                 if (i - start > total_bits * 2) {
1904                         total_found = 0;
1905                         cluster->max_size = 0;
1906                         found = false;
1907                 }
1908                 goto again;
1909         }
1910
1911         cluster->window_start = start * block_group->sectorsize +
1912                 entry->offset;
1913         cluster->points_to_bitmap = true;
1914
1915         return 0;
1916 }
1917
1918 /*
1919  * here we try to find a cluster of blocks in a block group.  The goal
1920  * is to find at least bytes free and up to empty_size + bytes free.
1921  * We might not find them all in one contiguous area.
1922  *
1923  * returns zero and sets up cluster if things worked out, otherwise
1924  * it returns -enospc
1925  */
1926 int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
1927                              struct btrfs_root *root,
1928                              struct btrfs_block_group_cache *block_group,
1929                              struct btrfs_free_cluster *cluster,
1930                              u64 offset, u64 bytes, u64 empty_size)
1931 {
1932         struct btrfs_free_space *entry = NULL;
1933         struct rb_node *node;
1934         struct btrfs_free_space *next;
1935         struct btrfs_free_space *last = NULL;
1936         u64 min_bytes;
1937         u64 window_start;
1938         u64 window_free;
1939         u64 max_extent = 0;
1940         bool found_bitmap = false;
1941         int ret;
1942
1943         /* for metadata, allow allocates with more holes */
1944         if (btrfs_test_opt(root, SSD_SPREAD)) {
1945                 min_bytes = bytes + empty_size;
1946         } else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
1947                 /*
1948                  * we want to do larger allocations when we are
1949                  * flushing out the delayed refs, it helps prevent
1950                  * making more work as we go along.
1951                  */
1952                 if (trans->transaction->delayed_refs.flushing)
1953                         min_bytes = max(bytes, (bytes + empty_size) >> 1);
1954                 else
1955                         min_bytes = max(bytes, (bytes + empty_size) >> 4);
1956         } else
1957                 min_bytes = max(bytes, (bytes + empty_size) >> 2);
1958
1959         spin_lock(&block_group->tree_lock);
1960         spin_lock(&cluster->lock);
1961
1962         /* someone already found a cluster, hooray */
1963         if (cluster->block_group) {
1964                 ret = 0;
1965                 goto out;
1966         }
1967 again:
1968         entry = tree_search_offset(block_group, offset, found_bitmap, 1);
1969         if (!entry) {
1970                 ret = -ENOSPC;
1971                 goto out;
1972         }
1973
1974         /*
1975          * If found_bitmap is true, we exhausted our search for extent entries,
1976          * and we just want to search all of the bitmaps that we can find, and
1977          * ignore any extent entries we find.
1978          */
1979         while (entry->bitmap || found_bitmap ||
1980                (!entry->bitmap && entry->bytes < min_bytes)) {
1981                 struct rb_node *node = rb_next(&entry->offset_index);
1982
1983                 if (entry->bitmap && entry->bytes > bytes + empty_size) {
1984                         ret = btrfs_bitmap_cluster(block_group, entry, cluster,
1985                                                    offset, bytes + empty_size,
1986                                                    min_bytes);
1987                         if (!ret)
1988                                 goto got_it;
1989                 }
1990
1991                 if (!node) {
1992                         ret = -ENOSPC;
1993                         goto out;
1994                 }
1995                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
1996         }
1997
1998         /*
1999          * We already searched all the extent entries from the passed in offset
2000          * to the end and didn't find enough space for the cluster, and we also
2001          * didn't find any bitmaps that met our criteria, just go ahead and exit
2002          */
2003         if (found_bitmap) {
2004                 ret = -ENOSPC;
2005                 goto out;
2006         }
2007
2008         cluster->points_to_bitmap = false;
2009         window_start = entry->offset;
2010         window_free = entry->bytes;
2011         last = entry;
2012         max_extent = entry->bytes;
2013
2014         while (1) {
2015                 /* out window is just right, lets fill it */
2016                 if (window_free >= bytes + empty_size)
2017                         break;
2018
2019                 node = rb_next(&last->offset_index);
2020                 if (!node) {
2021                         if (found_bitmap)
2022                                 goto again;
2023                         ret = -ENOSPC;
2024                         goto out;
2025                 }
2026                 next = rb_entry(node, struct btrfs_free_space, offset_index);
2027
2028                 /*
2029                  * we found a bitmap, so if this search doesn't result in a
2030                  * cluster, we know to go and search again for the bitmaps and
2031                  * start looking for space there
2032                  */
2033                 if (next->bitmap) {
2034                         if (!found_bitmap)
2035                                 offset = next->offset;
2036                         found_bitmap = true;
2037                         last = next;
2038                         continue;
2039                 }
2040
2041                 /*
2042                  * we haven't filled the empty size and the window is
2043                  * very large.  reset and try again
2044                  */
2045                 if (next->offset - (last->offset + last->bytes) > 128 * 1024 ||
2046                     next->offset - window_start > (bytes + empty_size) * 2) {
2047                         entry = next;
2048                         window_start = entry->offset;
2049                         window_free = entry->bytes;
2050                         last = entry;
2051                         max_extent = entry->bytes;
2052                 } else {
2053                         last = next;
2054                         window_free += next->bytes;
2055                         if (entry->bytes > max_extent)
2056                                 max_extent = entry->bytes;
2057                 }
2058         }
2059
2060         cluster->window_start = entry->offset;
2061
2062         /*
2063          * now we've found our entries, pull them out of the free space
2064          * cache and put them into the cluster rbtree
2065          *
2066          * The cluster includes an rbtree, but only uses the offset index
2067          * of each free space cache entry.
2068          */
2069         while (1) {
2070                 node = rb_next(&entry->offset_index);
2071                 if (entry->bitmap && node) {
2072                         entry = rb_entry(node, struct btrfs_free_space,
2073                                          offset_index);
2074                         continue;
2075                 } else if (entry->bitmap && !node) {
2076                         break;
2077                 }
2078
2079                 rb_erase(&entry->offset_index, &block_group->free_space_offset);
2080                 ret = tree_insert_offset(&cluster->root, entry->offset,
2081                                          &entry->offset_index, 0);
2082                 BUG_ON(ret);
2083
2084                 if (!node || entry == last)
2085                         break;
2086
2087                 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2088         }
2089
2090         cluster->max_size = max_extent;
2091 got_it:
2092         ret = 0;
2093         atomic_inc(&block_group->count);
2094         list_add_tail(&cluster->block_group_list, &block_group->cluster_list);
2095         cluster->block_group = block_group;
2096 out:
2097         spin_unlock(&cluster->lock);
2098         spin_unlock(&block_group->tree_lock);
2099
2100         return ret;
2101 }
2102
2103 /*
2104  * simple code to zero out a cluster
2105  */
2106 void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster)
2107 {
2108         spin_lock_init(&cluster->lock);
2109         spin_lock_init(&cluster->refill_lock);
2110         cluster->root = RB_ROOT;
2111         cluster->max_size = 0;
2112         cluster->points_to_bitmap = false;
2113         INIT_LIST_HEAD(&cluster->block_group_list);
2114         cluster->block_group = NULL;
2115 }
2116