?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'xarray-4.20-rc7' of git://git.infradead.org/users/willy/linux-dax
[linux/fpc-iii.git] / fs / btrfs / free-space-cache.c
blob74aa552f47930699aa2a81134c3200bf2f5d5c5e
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (C) 2008 Red Hat. All rights reserved.
4 */
5
6 #include <linux/pagemap.h>
7 #include <linux/sched.h>
8 #include <linux/sched/signal.h>
9 #include <linux/slab.h>
10 #include <linux/math64.h>
11 #include <linux/ratelimit.h>
12 #include <linux/error-injection.h>
13 #include <linux/sched/mm.h>
14 #include "ctree.h"
15 #include "free-space-cache.h"
16 #include "transaction.h"
17 #include "disk-io.h"
18 #include "extent_io.h"
19 #include "inode-map.h"
20 #include "volumes.h"
21
22 #define BITS_PER_BITMAP (PAGE_SIZE * 8UL)
23 #define MAX_CACHE_BYTES_PER_GIG SZ_32K
24
25 struct btrfs_trim_range {
26 u64 start;
27 u64 bytes;
28 struct list_head list;
29 };
30
31 static int link_free_space(struct btrfs_free_space_ctl *ctl,
32 struct btrfs_free_space *info);
33 static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
34 struct btrfs_free_space *info);
35 static int btrfs_wait_cache_io_root(struct btrfs_root *root,
36 struct btrfs_trans_handle *trans,
37 struct btrfs_io_ctl *io_ctl,
38 struct btrfs_path *path);
39
40 static struct inode *__lookup_free_space_inode(struct btrfs_root *root,
41 struct btrfs_path *path,
42 u64 offset)
43 {
44 struct btrfs_fs_info *fs_info = root->fs_info;
45 struct btrfs_key key;
46 struct btrfs_key location;
47 struct btrfs_disk_key disk_key;
48 struct btrfs_free_space_header *header;
49 struct extent_buffer *leaf;
50 struct inode *inode = NULL;
51 unsigned nofs_flag;
52 int ret;
53
54 key.objectid = BTRFS_FREE_SPACE_OBJECTID;
55 key.offset = offset;
56 key.type = 0;
57
58 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
59 if (ret < 0)
60 return ERR_PTR(ret);
61 if (ret > 0) {
62 btrfs_release_path(path);
63 return ERR_PTR(-ENOENT);
64 }
65
66 leaf = path->nodes[0];
67 header = btrfs_item_ptr(leaf, path->slots[0],
68 struct btrfs_free_space_header);
69 btrfs_free_space_key(leaf, header, &disk_key);
70 btrfs_disk_key_to_cpu(&location, &disk_key);
71 btrfs_release_path(path);
72
73 /*
74 * We are often under a trans handle at this point, so we need to make
75 * sure NOFS is set to keep us from deadlocking.
76 */
77 nofs_flag = memalloc_nofs_save();
78 inode = btrfs_iget_path(fs_info->sb, &location, root, NULL, path);
79 btrfs_release_path(path);
80 memalloc_nofs_restore(nofs_flag);
81 if (IS_ERR(inode))
82 return inode;
83
84 mapping_set_gfp_mask(inode->i_mapping,
85 mapping_gfp_constraint(inode->i_mapping,
86 ~(__GFP_FS | __GFP_HIGHMEM)));
87
88 return inode;
89 }
90
91 struct inode *lookup_free_space_inode(struct btrfs_fs_info *fs_info,
92 struct btrfs_block_group_cache
93 *block_group, struct btrfs_path *path)
94 {
95 struct inode *inode = NULL;
96 u32 flags = BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
97
98 spin_lock(&block_group->lock);
99 if (block_group->inode)
100 inode = igrab(block_group->inode);
101 spin_unlock(&block_group->lock);
102 if (inode)
103 return inode;
104
105 inode = __lookup_free_space_inode(fs_info->tree_root, path,
106 block_group->key.objectid);
107 if (IS_ERR(inode))
108 return inode;
109
110 spin_lock(&block_group->lock);
111 if (!((BTRFS_I(inode)->flags & flags) == flags)) {
112 btrfs_info(fs_info, "Old style space inode found, converting.");
113 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM |
114 BTRFS_INODE_NODATACOW;
115 block_group->disk_cache_state = BTRFS_DC_CLEAR;
116 }
117
118 if (!block_group->iref) {
119 block_group->inode = igrab(inode);
120 block_group->iref = 1;
121 }
122 spin_unlock(&block_group->lock);
123
124 return inode;
125 }
126
127 static int __create_free_space_inode(struct btrfs_root *root,
128 struct btrfs_trans_handle *trans,
129 struct btrfs_path *path,
130 u64 ino, u64 offset)
131 {
132 struct btrfs_key key;
133 struct btrfs_disk_key disk_key;
134 struct btrfs_free_space_header *header;
135 struct btrfs_inode_item *inode_item;
136 struct extent_buffer *leaf;
137 u64 flags = BTRFS_INODE_NOCOMPRESS | BTRFS_INODE_PREALLOC;
138 int ret;
139
140 ret = btrfs_insert_empty_inode(trans, root, path, ino);
141 if (ret)
142 return ret;
143
144 /* We inline crc's for the free disk space cache */
145 if (ino != BTRFS_FREE_INO_OBJECTID)
146 flags |= BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
147
148 leaf = path->nodes[0];
149 inode_item = btrfs_item_ptr(leaf, path->slots[0],
150 struct btrfs_inode_item);
151 btrfs_item_key(leaf, &disk_key, path->slots[0]);
152 memzero_extent_buffer(leaf, (unsigned long)inode_item,
153 sizeof(*inode_item));
154 btrfs_set_inode_generation(leaf, inode_item, trans->transid);
155 btrfs_set_inode_size(leaf, inode_item, 0);
156 btrfs_set_inode_nbytes(leaf, inode_item, 0);
157 btrfs_set_inode_uid(leaf, inode_item, 0);
158 btrfs_set_inode_gid(leaf, inode_item, 0);
159 btrfs_set_inode_mode(leaf, inode_item, S_IFREG | 0600);
160 btrfs_set_inode_flags(leaf, inode_item, flags);
161 btrfs_set_inode_nlink(leaf, inode_item, 1);
162 btrfs_set_inode_transid(leaf, inode_item, trans->transid);
163 btrfs_set_inode_block_group(leaf, inode_item, offset);
164 btrfs_mark_buffer_dirty(leaf);
165 btrfs_release_path(path);
166
167 key.objectid = BTRFS_FREE_SPACE_OBJECTID;
168 key.offset = offset;
169 key.type = 0;
170 ret = btrfs_insert_empty_item(trans, root, path, &key,
171 sizeof(struct btrfs_free_space_header));
172 if (ret < 0) {
173 btrfs_release_path(path);
174 return ret;
175 }
176
177 leaf = path->nodes[0];
178 header = btrfs_item_ptr(leaf, path->slots[0],
179 struct btrfs_free_space_header);
180 memzero_extent_buffer(leaf, (unsigned long)header, sizeof(*header));
181 btrfs_set_free_space_key(leaf, header, &disk_key);
182 btrfs_mark_buffer_dirty(leaf);
183 btrfs_release_path(path);
184
185 return 0;
186 }
187
188 int create_free_space_inode(struct btrfs_fs_info *fs_info,
189 struct btrfs_trans_handle *trans,
190 struct btrfs_block_group_cache *block_group,
191 struct btrfs_path *path)
192 {
193 int ret;
194 u64 ino;
195
196 ret = btrfs_find_free_objectid(fs_info->tree_root, &ino);
197 if (ret < 0)
198 return ret;
199
200 return __create_free_space_inode(fs_info->tree_root, trans, path, ino,
201 block_group->key.objectid);
202 }
203
204 int btrfs_check_trunc_cache_free_space(struct btrfs_fs_info *fs_info,
205 struct btrfs_block_rsv *rsv)
206 {
207 u64 needed_bytes;
208 int ret;
209
210 /* 1 for slack space, 1 for updating the inode */
211 needed_bytes = btrfs_calc_trunc_metadata_size(fs_info, 1) +
212 btrfs_calc_trans_metadata_size(fs_info, 1);
213
214 spin_lock(&rsv->lock);
215 if (rsv->reserved < needed_bytes)
216 ret = -ENOSPC;
217 else
218 ret = 0;
219 spin_unlock(&rsv->lock);
220 return ret;
221 }
222
223 int btrfs_truncate_free_space_cache(struct btrfs_trans_handle *trans,
224 struct btrfs_block_group_cache *block_group,
225 struct inode *inode)
226 {
227 struct btrfs_root *root = BTRFS_I(inode)->root;
228 int ret = 0;
229 bool locked = false;
230
231 if (block_group) {
232 struct btrfs_path *path = btrfs_alloc_path();
233
234 if (!path) {
235 ret = -ENOMEM;
236 goto fail;
237 }
238 locked = true;
239 mutex_lock(&trans->transaction->cache_write_mutex);
240 if (!list_empty(&block_group->io_list)) {
241 list_del_init(&block_group->io_list);
242
243 btrfs_wait_cache_io(trans, block_group, path);
244 btrfs_put_block_group(block_group);
245 }
246
247 /*
248 * now that we've truncated the cache away, its no longer
249 * setup or written
250 */
251 spin_lock(&block_group->lock);
252 block_group->disk_cache_state = BTRFS_DC_CLEAR;
253 spin_unlock(&block_group->lock);
254 btrfs_free_path(path);
255 }
256
257 btrfs_i_size_write(BTRFS_I(inode), 0);
258 truncate_pagecache(inode, 0);
259
260 /*
261 * We skip the throttling logic for free space cache inodes, so we don't
262 * need to check for -EAGAIN.
263 */
264 ret = btrfs_truncate_inode_items(trans, root, inode,
265 0, BTRFS_EXTENT_DATA_KEY);
266 if (ret)
267 goto fail;
268
269 ret = btrfs_update_inode(trans, root, inode);
270
271 fail:
272 if (locked)
273 mutex_unlock(&trans->transaction->cache_write_mutex);
274 if (ret)
275 btrfs_abort_transaction(trans, ret);
276
277 return ret;
278 }
279
280 static void readahead_cache(struct inode *inode)
281 {
282 struct file_ra_state *ra;
283 unsigned long last_index;
284
285 ra = kzalloc(sizeof(*ra), GFP_NOFS);
286 if (!ra)
287 return;
288
289 file_ra_state_init(ra, inode->i_mapping);
290 last_index = (i_size_read(inode) - 1) >> PAGE_SHIFT;
291
292 page_cache_sync_readahead(inode->i_mapping, ra, NULL, 0, last_index);
293
294 kfree(ra);
295 }
296
297 static int io_ctl_init(struct btrfs_io_ctl *io_ctl, struct inode *inode,
298 int write)
299 {
300 int num_pages;
301 int check_crcs = 0;
302
303 num_pages = DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE);
304
305 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FREE_INO_OBJECTID)
306 check_crcs = 1;
307
308 /* Make sure we can fit our crcs and generation into the first page */
309 if (write && check_crcs &&
310 (num_pages * sizeof(u32) + sizeof(u64)) > PAGE_SIZE)
311 return -ENOSPC;
312
313 memset(io_ctl, 0, sizeof(struct btrfs_io_ctl));
314
315 io_ctl->pages = kcalloc(num_pages, sizeof(struct page *), GFP_NOFS);
316 if (!io_ctl->pages)
317 return -ENOMEM;
318
319 io_ctl->num_pages = num_pages;
320 io_ctl->fs_info = btrfs_sb(inode->i_sb);
321 io_ctl->check_crcs = check_crcs;
322 io_ctl->inode = inode;
323
324 return 0;
325 }
326 ALLOW_ERROR_INJECTION(io_ctl_init, ERRNO);
327
328 static void io_ctl_free(struct btrfs_io_ctl *io_ctl)
329 {
330 kfree(io_ctl->pages);
331 io_ctl->pages = NULL;
332 }
333
334 static void io_ctl_unmap_page(struct btrfs_io_ctl *io_ctl)
335 {
336 if (io_ctl->cur) {
337 io_ctl->cur = NULL;
338 io_ctl->orig = NULL;
339 }
340 }
341
342 static void io_ctl_map_page(struct btrfs_io_ctl *io_ctl, int clear)
343 {
344 ASSERT(io_ctl->index < io_ctl->num_pages);
345 io_ctl->page = io_ctl->pages[io_ctl->index++];
346 io_ctl->cur = page_address(io_ctl->page);
347 io_ctl->orig = io_ctl->cur;
348 io_ctl->size = PAGE_SIZE;
349 if (clear)
350 clear_page(io_ctl->cur);
351 }
352
353 static void io_ctl_drop_pages(struct btrfs_io_ctl *io_ctl)
354 {
355 int i;
356
357 io_ctl_unmap_page(io_ctl);
358
359 for (i = 0; i < io_ctl->num_pages; i++) {
360 if (io_ctl->pages[i]) {
361 ClearPageChecked(io_ctl->pages[i]);
362 unlock_page(io_ctl->pages[i]);
363 put_page(io_ctl->pages[i]);
364 }
365 }
366 }
367
368 static int io_ctl_prepare_pages(struct btrfs_io_ctl *io_ctl, struct inode *inode,
369 int uptodate)
370 {
371 struct page *page;
372 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
373 int i;
374
375 for (i = 0; i < io_ctl->num_pages; i++) {
376 page = find_or_create_page(inode->i_mapping, i, mask);
377 if (!page) {
378 io_ctl_drop_pages(io_ctl);
379 return -ENOMEM;
380 }
381 io_ctl->pages[i] = page;
382 if (uptodate && !PageUptodate(page)) {
383 btrfs_readpage(NULL, page);
384 lock_page(page);
385 if (!PageUptodate(page)) {
386 btrfs_err(BTRFS_I(inode)->root->fs_info,
387 "error reading free space cache");
388 io_ctl_drop_pages(io_ctl);
389 return -EIO;
390 }
391 }
392 }
393
394 for (i = 0; i < io_ctl->num_pages; i++) {
395 clear_page_dirty_for_io(io_ctl->pages[i]);
396 set_page_extent_mapped(io_ctl->pages[i]);
397 }
398
399 return 0;
400 }
401
402 static void io_ctl_set_generation(struct btrfs_io_ctl *io_ctl, u64 generation)
403 {
404 __le64 *val;
405
406 io_ctl_map_page(io_ctl, 1);
407
408 /*
409 * Skip the csum areas. If we don't check crcs then we just have a
410 * 64bit chunk at the front of the first page.
411 */
412 if (io_ctl->check_crcs) {
413 io_ctl->cur += (sizeof(u32) * io_ctl->num_pages);
414 io_ctl->size -= sizeof(u64) + (sizeof(u32) * io_ctl->num_pages);
415 } else {
416 io_ctl->cur += sizeof(u64);
417 io_ctl->size -= sizeof(u64) * 2;
418 }
419
420 val = io_ctl->cur;
421 *val = cpu_to_le64(generation);
422 io_ctl->cur += sizeof(u64);
423 }
424
425 static int io_ctl_check_generation(struct btrfs_io_ctl *io_ctl, u64 generation)
426 {
427 __le64 *gen;
428
429 /*
430 * Skip the crc area. If we don't check crcs then we just have a 64bit
431 * chunk at the front of the first page.
432 */
433 if (io_ctl->check_crcs) {
434 io_ctl->cur += sizeof(u32) * io_ctl->num_pages;
435 io_ctl->size -= sizeof(u64) +
436 (sizeof(u32) * io_ctl->num_pages);
437 } else {
438 io_ctl->cur += sizeof(u64);
439 io_ctl->size -= sizeof(u64) * 2;
440 }
441
442 gen = io_ctl->cur;
443 if (le64_to_cpu(*gen) != generation) {
444 btrfs_err_rl(io_ctl->fs_info,
445 "space cache generation (%llu) does not match inode (%llu)",
446 *gen, generation);
447 io_ctl_unmap_page(io_ctl);
448 return -EIO;
449 }
450 io_ctl->cur += sizeof(u64);
451 return 0;
452 }
453
454 static void io_ctl_set_crc(struct btrfs_io_ctl *io_ctl, int index)
455 {
456 u32 *tmp;
457 u32 crc = ~(u32)0;
458 unsigned offset = 0;
459
460 if (!io_ctl->check_crcs) {
461 io_ctl_unmap_page(io_ctl);
462 return;
463 }
464
465 if (index == 0)
466 offset = sizeof(u32) * io_ctl->num_pages;
467
468 crc = btrfs_csum_data(io_ctl->orig + offset, crc,
469 PAGE_SIZE - offset);
470 btrfs_csum_final(crc, (u8 *)&crc);
471 io_ctl_unmap_page(io_ctl);
472 tmp = page_address(io_ctl->pages[0]);
473 tmp += index;
474 *tmp = crc;
475 }
476
477 static int io_ctl_check_crc(struct btrfs_io_ctl *io_ctl, int index)
478 {
479 u32 *tmp, val;
480 u32 crc = ~(u32)0;
481 unsigned offset = 0;
482
483 if (!io_ctl->check_crcs) {
484 io_ctl_map_page(io_ctl, 0);
485 return 0;
486 }
487
488 if (index == 0)
489 offset = sizeof(u32) * io_ctl->num_pages;
490
491 tmp = page_address(io_ctl->pages[0]);
492 tmp += index;
493 val = *tmp;
494
495 io_ctl_map_page(io_ctl, 0);
496 crc = btrfs_csum_data(io_ctl->orig + offset, crc,
497 PAGE_SIZE - offset);
498 btrfs_csum_final(crc, (u8 *)&crc);
499 if (val != crc) {
500 btrfs_err_rl(io_ctl->fs_info,
501 "csum mismatch on free space cache");
502 io_ctl_unmap_page(io_ctl);
503 return -EIO;
504 }
505
506 return 0;
507 }
508
509 static int io_ctl_add_entry(struct btrfs_io_ctl *io_ctl, u64 offset, u64 bytes,
510 void *bitmap)
511 {
512 struct btrfs_free_space_entry *entry;
513
514 if (!io_ctl->cur)
515 return -ENOSPC;
516
517 entry = io_ctl->cur;
518 entry->offset = cpu_to_le64(offset);
519 entry->bytes = cpu_to_le64(bytes);
520 entry->type = (bitmap) ? BTRFS_FREE_SPACE_BITMAP :
521 BTRFS_FREE_SPACE_EXTENT;
522 io_ctl->cur += sizeof(struct btrfs_free_space_entry);
523 io_ctl->size -= sizeof(struct btrfs_free_space_entry);
524
525 if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
526 return 0;
527
528 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
529
530 /* No more pages to map */
531 if (io_ctl->index >= io_ctl->num_pages)
532 return 0;
533
534 /* map the next page */
535 io_ctl_map_page(io_ctl, 1);
536 return 0;
537 }
538
539 static int io_ctl_add_bitmap(struct btrfs_io_ctl *io_ctl, void *bitmap)
540 {
541 if (!io_ctl->cur)
542 return -ENOSPC;
543
544 /*
545 * If we aren't at the start of the current page, unmap this one and
546 * map the next one if there is any left.
547 */
548 if (io_ctl->cur != io_ctl->orig) {
549 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
550 if (io_ctl->index >= io_ctl->num_pages)
551 return -ENOSPC;
552 io_ctl_map_page(io_ctl, 0);
553 }
554
555 copy_page(io_ctl->cur, bitmap);
556 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
557 if (io_ctl->index < io_ctl->num_pages)
558 io_ctl_map_page(io_ctl, 0);
559 return 0;
560 }
561
562 static void io_ctl_zero_remaining_pages(struct btrfs_io_ctl *io_ctl)
563 {
564 /*
565 * If we're not on the boundary we know we've modified the page and we
566 * need to crc the page.
567 */
568 if (io_ctl->cur != io_ctl->orig)
569 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
570 else
571 io_ctl_unmap_page(io_ctl);
572
573 while (io_ctl->index < io_ctl->num_pages) {
574 io_ctl_map_page(io_ctl, 1);
575 io_ctl_set_crc(io_ctl, io_ctl->index - 1);
576 }
577 }
578
579 static int io_ctl_read_entry(struct btrfs_io_ctl *io_ctl,
580 struct btrfs_free_space *entry, u8 *type)
581 {
582 struct btrfs_free_space_entry *e;
583 int ret;
584
585 if (!io_ctl->cur) {
586 ret = io_ctl_check_crc(io_ctl, io_ctl->index);
587 if (ret)
588 return ret;
589 }
590
591 e = io_ctl->cur;
592 entry->offset = le64_to_cpu(e->offset);
593 entry->bytes = le64_to_cpu(e->bytes);
594 *type = e->type;
595 io_ctl->cur += sizeof(struct btrfs_free_space_entry);
596 io_ctl->size -= sizeof(struct btrfs_free_space_entry);
597
598 if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
599 return 0;
600
601 io_ctl_unmap_page(io_ctl);
602
603 return 0;
604 }
605
606 static int io_ctl_read_bitmap(struct btrfs_io_ctl *io_ctl,
607 struct btrfs_free_space *entry)
608 {
609 int ret;
610
611 ret = io_ctl_check_crc(io_ctl, io_ctl->index);
612 if (ret)
613 return ret;
614
615 copy_page(entry->bitmap, io_ctl->cur);
616 io_ctl_unmap_page(io_ctl);
617
618 return 0;
619 }
620
621 /*
622 * Since we attach pinned extents after the fact we can have contiguous sections
623 * of free space that are split up in entries. This poses a problem with the
624 * tree logging stuff since it could have allocated across what appears to be 2
625 * entries since we would have merged the entries when adding the pinned extents
626 * back to the free space cache. So run through the space cache that we just
627 * loaded and merge contiguous entries. This will make the log replay stuff not
628 * blow up and it will make for nicer allocator behavior.
629 */
630 static void merge_space_tree(struct btrfs_free_space_ctl *ctl)
631 {
632 struct btrfs_free_space *e, *prev = NULL;
633 struct rb_node *n;
634
635 again:
636 spin_lock(&ctl->tree_lock);
637 for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
638 e = rb_entry(n, struct btrfs_free_space, offset_index);
639 if (!prev)
640 goto next;
641 if (e->bitmap || prev->bitmap)
642 goto next;
643 if (prev->offset + prev->bytes == e->offset) {
644 unlink_free_space(ctl, prev);
645 unlink_free_space(ctl, e);
646 prev->bytes += e->bytes;
647 kmem_cache_free(btrfs_free_space_cachep, e);
648 link_free_space(ctl, prev);
649 prev = NULL;
650 spin_unlock(&ctl->tree_lock);
651 goto again;
652 }
653 next:
654 prev = e;
655 }
656 spin_unlock(&ctl->tree_lock);
657 }
658
659 static int __load_free_space_cache(struct btrfs_root *root, struct inode *inode,
660 struct btrfs_free_space_ctl *ctl,
661 struct btrfs_path *path, u64 offset)
662 {
663 struct btrfs_fs_info *fs_info = root->fs_info;
664 struct btrfs_free_space_header *header;
665 struct extent_buffer *leaf;
666 struct btrfs_io_ctl io_ctl;
667 struct btrfs_key key;
668 struct btrfs_free_space *e, *n;
669 LIST_HEAD(bitmaps);
670 u64 num_entries;
671 u64 num_bitmaps;
672 u64 generation;
673 u8 type;
674 int ret = 0;
675
676 /* Nothing in the space cache, goodbye */
677 if (!i_size_read(inode))
678 return 0;
679
680 key.objectid = BTRFS_FREE_SPACE_OBJECTID;
681 key.offset = offset;
682 key.type = 0;
683
684 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
685 if (ret < 0)
686 return 0;
687 else if (ret > 0) {
688 btrfs_release_path(path);
689 return 0;
690 }
691
692 ret = -1;
693
694 leaf = path->nodes[0];
695 header = btrfs_item_ptr(leaf, path->slots[0],
696 struct btrfs_free_space_header);
697 num_entries = btrfs_free_space_entries(leaf, header);
698 num_bitmaps = btrfs_free_space_bitmaps(leaf, header);
699 generation = btrfs_free_space_generation(leaf, header);
700 btrfs_release_path(path);
701
702 if (!BTRFS_I(inode)->generation) {
703 btrfs_info(fs_info,
704 "the free space cache file (%llu) is invalid, skip it",
705 offset);
706 return 0;
707 }
708
709 if (BTRFS_I(inode)->generation != generation) {
710 btrfs_err(fs_info,
711 "free space inode generation (%llu) did not match free space cache generation (%llu)",
712 BTRFS_I(inode)->generation, generation);
713 return 0;
714 }
715
716 if (!num_entries)
717 return 0;
718
719 ret = io_ctl_init(&io_ctl, inode, 0);
720 if (ret)
721 return ret;
722
723 readahead_cache(inode);
724
725 ret = io_ctl_prepare_pages(&io_ctl, inode, 1);
726 if (ret)
727 goto out;
728
729 ret = io_ctl_check_crc(&io_ctl, 0);
730 if (ret)
731 goto free_cache;
732
733 ret = io_ctl_check_generation(&io_ctl, generation);
734 if (ret)
735 goto free_cache;
736
737 while (num_entries) {
738 e = kmem_cache_zalloc(btrfs_free_space_cachep,
739 GFP_NOFS);
740 if (!e)
741 goto free_cache;
742
743 ret = io_ctl_read_entry(&io_ctl, e, &type);
744 if (ret) {
745 kmem_cache_free(btrfs_free_space_cachep, e);
746 goto free_cache;
747 }
748
749 if (!e->bytes) {
750 kmem_cache_free(btrfs_free_space_cachep, e);
751 goto free_cache;
752 }
753
754 if (type == BTRFS_FREE_SPACE_EXTENT) {
755 spin_lock(&ctl->tree_lock);
756 ret = link_free_space(ctl, e);
757 spin_unlock(&ctl->tree_lock);
758 if (ret) {
759 btrfs_err(fs_info,
760 "Duplicate entries in free space cache, dumping");
761 kmem_cache_free(btrfs_free_space_cachep, e);
762 goto free_cache;
763 }
764 } else {
765 ASSERT(num_bitmaps);
766 num_bitmaps--;
767 e->bitmap = kzalloc(PAGE_SIZE, GFP_NOFS);
768 if (!e->bitmap) {
769 kmem_cache_free(
770 btrfs_free_space_cachep, e);
771 goto free_cache;
772 }
773 spin_lock(&ctl->tree_lock);
774 ret = link_free_space(ctl, e);
775 ctl->total_bitmaps++;
776 ctl->op->recalc_thresholds(ctl);
777 spin_unlock(&ctl->tree_lock);
778 if (ret) {
779 btrfs_err(fs_info,
780 "Duplicate entries in free space cache, dumping");
781 kmem_cache_free(btrfs_free_space_cachep, e);
782 goto free_cache;
783 }
784 list_add_tail(&e->list, &bitmaps);
785 }
786
787 num_entries--;
788 }
789
790 io_ctl_unmap_page(&io_ctl);
791
792 /*
793 * We add the bitmaps at the end of the entries in order that
794 * the bitmap entries are added to the cache.
795 */
796 list_for_each_entry_safe(e, n, &bitmaps, list) {
797 list_del_init(&e->list);
798 ret = io_ctl_read_bitmap(&io_ctl, e);
799 if (ret)
800 goto free_cache;
801 }
802
803 io_ctl_drop_pages(&io_ctl);
804 merge_space_tree(ctl);
805 ret = 1;
806 out:
807 io_ctl_free(&io_ctl);
808 return ret;
809 free_cache:
810 io_ctl_drop_pages(&io_ctl);
811 __btrfs_remove_free_space_cache(ctl);
812 goto out;
813 }
814
815 int load_free_space_cache(struct btrfs_fs_info *fs_info,
816 struct btrfs_block_group_cache *block_group)
817 {
818 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
819 struct inode *inode;
820 struct btrfs_path *path;
821 int ret = 0;
822 bool matched;
823 u64 used = btrfs_block_group_used(&block_group->item);
824
825 /*
826 * If this block group has been marked to be cleared for one reason or
827 * another then we can't trust the on disk cache, so just return.
828 */
829 spin_lock(&block_group->lock);
830 if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
831 spin_unlock(&block_group->lock);
832 return 0;
833 }
834 spin_unlock(&block_group->lock);
835
836 path = btrfs_alloc_path();
837 if (!path)
838 return 0;
839 path->search_commit_root = 1;
840 path->skip_locking = 1;
841
842 /*
843 * We must pass a path with search_commit_root set to btrfs_iget in
844 * order to avoid a deadlock when allocating extents for the tree root.
845 *
846 * When we are COWing an extent buffer from the tree root, when looking
847 * for a free extent, at extent-tree.c:find_free_extent(), we can find
848 * block group without its free space cache loaded. When we find one
849 * we must load its space cache which requires reading its free space
850 * cache's inode item from the root tree. If this inode item is located
851 * in the same leaf that we started COWing before, then we end up in
852 * deadlock on the extent buffer (trying to read lock it when we
853 * previously write locked it).
854 *
855 * It's safe to read the inode item using the commit root because
856 * block groups, once loaded, stay in memory forever (until they are
857 * removed) as well as their space caches once loaded. New block groups
858 * once created get their ->cached field set to BTRFS_CACHE_FINISHED so
859 * we will never try to read their inode item while the fs is mounted.
860 */
861 inode = lookup_free_space_inode(fs_info, block_group, path);
862 if (IS_ERR(inode)) {
863 btrfs_free_path(path);
864 return 0;
865 }
866
867 /* We may have converted the inode and made the cache invalid. */
868 spin_lock(&block_group->lock);
869 if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
870 spin_unlock(&block_group->lock);
871 btrfs_free_path(path);
872 goto out;
873 }
874 spin_unlock(&block_group->lock);
875
876 ret = __load_free_space_cache(fs_info->tree_root, inode, ctl,
877 path, block_group->key.objectid);
878 btrfs_free_path(path);
879 if (ret <= 0)
880 goto out;
881
882 spin_lock(&ctl->tree_lock);
883 matched = (ctl->free_space == (block_group->key.offset - used -
884 block_group->bytes_super));
885 spin_unlock(&ctl->tree_lock);
886
887 if (!matched) {
888 __btrfs_remove_free_space_cache(ctl);
889 btrfs_warn(fs_info,
890 "block group %llu has wrong amount of free space",
891 block_group->key.objectid);
892 ret = -1;
893 }
894 out:
895 if (ret < 0) {
896 /* This cache is bogus, make sure it gets cleared */
897 spin_lock(&block_group->lock);
898 block_group->disk_cache_state = BTRFS_DC_CLEAR;
899 spin_unlock(&block_group->lock);
900 ret = 0;
901
902 btrfs_warn(fs_info,
903 "failed to load free space cache for block group %llu, rebuilding it now",
904 block_group->key.objectid);
905 }
906
907 iput(inode);
908 return ret;
909 }
910
911 static noinline_for_stack
912 int write_cache_extent_entries(struct btrfs_io_ctl *io_ctl,
913 struct btrfs_free_space_ctl *ctl,
914 struct btrfs_block_group_cache *block_group,
915 int *entries, int *bitmaps,
916 struct list_head *bitmap_list)
917 {
918 int ret;
919 struct btrfs_free_cluster *cluster = NULL;
920 struct btrfs_free_cluster *cluster_locked = NULL;
921 struct rb_node *node = rb_first(&ctl->free_space_offset);
922 struct btrfs_trim_range *trim_entry;
923
924 /* Get the cluster for this block_group if it exists */
925 if (block_group && !list_empty(&block_group->cluster_list)) {
926 cluster = list_entry(block_group->cluster_list.next,
927 struct btrfs_free_cluster,
928 block_group_list);
929 }
930
931 if (!node && cluster) {
932 cluster_locked = cluster;
933 spin_lock(&cluster_locked->lock);
934 node = rb_first(&cluster->root);
935 cluster = NULL;
936 }
937
938 /* Write out the extent entries */
939 while (node) {
940 struct btrfs_free_space *e;
941
942 e = rb_entry(node, struct btrfs_free_space, offset_index);
943 *entries += 1;
944
945 ret = io_ctl_add_entry(io_ctl, e->offset, e->bytes,
946 e->bitmap);
947 if (ret)
948 goto fail;
949
950 if (e->bitmap) {
951 list_add_tail(&e->list, bitmap_list);
952 *bitmaps += 1;
953 }
954 node = rb_next(node);
955 if (!node && cluster) {
956 node = rb_first(&cluster->root);
957 cluster_locked = cluster;
958 spin_lock(&cluster_locked->lock);
959 cluster = NULL;
960 }
961 }
962 if (cluster_locked) {
963 spin_unlock(&cluster_locked->lock);
964 cluster_locked = NULL;
965 }
966
967 /*
968 * Make sure we don't miss any range that was removed from our rbtree
969 * because trimming is running. Otherwise after a umount+mount (or crash
970 * after committing the transaction) we would leak free space and get
971 * an inconsistent free space cache report from fsck.
972 */
973 list_for_each_entry(trim_entry, &ctl->trimming_ranges, list) {
974 ret = io_ctl_add_entry(io_ctl, trim_entry->start,
975 trim_entry->bytes, NULL);
976 if (ret)
977 goto fail;
978 *entries += 1;
979 }
980
981 return 0;
982 fail:
983 if (cluster_locked)
984 spin_unlock(&cluster_locked->lock);
985 return -ENOSPC;
986 }
987
988 static noinline_for_stack int
989 update_cache_item(struct btrfs_trans_handle *trans,
990 struct btrfs_root *root,
991 struct inode *inode,
992 struct btrfs_path *path, u64 offset,
993 int entries, int bitmaps)
994 {
995 struct btrfs_key key;
996 struct btrfs_free_space_header *header;
997 struct extent_buffer *leaf;
998 int ret;
999
1000 key.objectid = BTRFS_FREE_SPACE_OBJECTID;
1001 key.offset = offset;
1002 key.type = 0;
1003
1004 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
1005 if (ret < 0) {
1006 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
1007 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL);
1008 goto fail;
1009 }
1010 leaf = path->nodes[0];
1011 if (ret > 0) {
1012 struct btrfs_key found_key;
1013 ASSERT(path->slots[0]);
1014 path->slots[0]--;
1015 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1016 if (found_key.objectid != BTRFS_FREE_SPACE_OBJECTID ||
1017 found_key.offset != offset) {
1018 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0,
1019 inode->i_size - 1,
1020 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0,
1021 NULL);
1022 btrfs_release_path(path);
1023 goto fail;
1024 }
1025 }
1026
1027 BTRFS_I(inode)->generation = trans->transid;
1028 header = btrfs_item_ptr(leaf, path->slots[0],
1029 struct btrfs_free_space_header);
1030 btrfs_set_free_space_entries(leaf, header, entries);
1031 btrfs_set_free_space_bitmaps(leaf, header, bitmaps);
1032 btrfs_set_free_space_generation(leaf, header, trans->transid);
1033 btrfs_mark_buffer_dirty(leaf);
1034 btrfs_release_path(path);
1035
1036 return 0;
1037
1038 fail:
1039 return -1;
1040 }
1041
1042 static noinline_for_stack int
1043 write_pinned_extent_entries(struct btrfs_fs_info *fs_info,
1044 struct btrfs_block_group_cache *block_group,
1045 struct btrfs_io_ctl *io_ctl,
1046 int *entries)
1047 {
1048 u64 start, extent_start, extent_end, len;
1049 struct extent_io_tree *unpin = NULL;
1050 int ret;
1051
1052 if (!block_group)
1053 return 0;
1054
1055 /*
1056 * We want to add any pinned extents to our free space cache
1057 * so we don't leak the space
1058 *
1059 * We shouldn't have switched the pinned extents yet so this is the
1060 * right one
1061 */
1062 unpin = fs_info->pinned_extents;
1063
1064 start = block_group->key.objectid;
1065
1066 while (start < block_group->key.objectid + block_group->key.offset) {
1067 ret = find_first_extent_bit(unpin, start,
1068 &extent_start, &extent_end,
1069 EXTENT_DIRTY, NULL);
1070 if (ret)
1071 return 0;
1072
1073 /* This pinned extent is out of our range */
1074 if (extent_start >= block_group->key.objectid +
1075 block_group->key.offset)
1076 return 0;
1077
1078 extent_start = max(extent_start, start);
1079 extent_end = min(block_group->key.objectid +
1080 block_group->key.offset, extent_end + 1);
1081 len = extent_end - extent_start;
1082
1083 *entries += 1;
1084 ret = io_ctl_add_entry(io_ctl, extent_start, len, NULL);
1085 if (ret)
1086 return -ENOSPC;
1087
1088 start = extent_end;
1089 }
1090
1091 return 0;
1092 }
1093
1094 static noinline_for_stack int
1095 write_bitmap_entries(struct btrfs_io_ctl *io_ctl, struct list_head *bitmap_list)
1096 {
1097 struct btrfs_free_space *entry, *next;
1098 int ret;
1099
1100 /* Write out the bitmaps */
1101 list_for_each_entry_safe(entry, next, bitmap_list, list) {
1102 ret = io_ctl_add_bitmap(io_ctl, entry->bitmap);
1103 if (ret)
1104 return -ENOSPC;
1105 list_del_init(&entry->list);
1106 }
1107
1108 return 0;
1109 }
1110
1111 static int flush_dirty_cache(struct inode *inode)
1112 {
1113 int ret;
1114
1115 ret = btrfs_wait_ordered_range(inode, 0, (u64)-1);
1116 if (ret)
1117 clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
1118 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL);
1119
1120 return ret;
1121 }
1122
1123 static void noinline_for_stack
1124 cleanup_bitmap_list(struct list_head *bitmap_list)
1125 {
1126 struct btrfs_free_space *entry, *next;
1127
1128 list_for_each_entry_safe(entry, next, bitmap_list, list)
1129 list_del_init(&entry->list);
1130 }
1131
1132 static void noinline_for_stack
1133 cleanup_write_cache_enospc(struct inode *inode,
1134 struct btrfs_io_ctl *io_ctl,
1135 struct extent_state **cached_state)
1136 {
1137 io_ctl_drop_pages(io_ctl);
1138 unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
1139 i_size_read(inode) - 1, cached_state);
1140 }
1141
1142 static int __btrfs_wait_cache_io(struct btrfs_root *root,
1143 struct btrfs_trans_handle *trans,
1144 struct btrfs_block_group_cache *block_group,
1145 struct btrfs_io_ctl *io_ctl,
1146 struct btrfs_path *path, u64 offset)
1147 {
1148 int ret;
1149 struct inode *inode = io_ctl->inode;
1150
1151 if (!inode)
1152 return 0;
1153
1154 /* Flush the dirty pages in the cache file. */
1155 ret = flush_dirty_cache(inode);
1156 if (ret)
1157 goto out;
1158
1159 /* Update the cache item to tell everyone this cache file is valid. */
1160 ret = update_cache_item(trans, root, inode, path, offset,
1161 io_ctl->entries, io_ctl->bitmaps);
1162 out:
1163 io_ctl_free(io_ctl);
1164 if (ret) {
1165 invalidate_inode_pages2(inode->i_mapping);
1166 BTRFS_I(inode)->generation = 0;
1167 if (block_group) {
1168 #ifdef DEBUG
1169 btrfs_err(root->fs_info,
1170 "failed to write free space cache for block group %llu",
1171 block_group->key.objectid);
1172 #endif
1173 }
1174 }
1175 btrfs_update_inode(trans, root, inode);
1176
1177 if (block_group) {
1178 /* the dirty list is protected by the dirty_bgs_lock */
1179 spin_lock(&trans->transaction->dirty_bgs_lock);
1180
1181 /* the disk_cache_state is protected by the block group lock */
1182 spin_lock(&block_group->lock);
1183
1184 /*
1185 * only mark this as written if we didn't get put back on
1186 * the dirty list while waiting for IO. Otherwise our
1187 * cache state won't be right, and we won't get written again
1188 */
1189 if (!ret && list_empty(&block_group->dirty_list))
1190 block_group->disk_cache_state = BTRFS_DC_WRITTEN;
1191 else if (ret)
1192 block_group->disk_cache_state = BTRFS_DC_ERROR;
1193
1194 spin_unlock(&block_group->lock);
1195 spin_unlock(&trans->transaction->dirty_bgs_lock);
1196 io_ctl->inode = NULL;
1197 iput(inode);
1198 }
1199
1200 return ret;
1201
1202 }
1203
1204 static int btrfs_wait_cache_io_root(struct btrfs_root *root,
1205 struct btrfs_trans_handle *trans,
1206 struct btrfs_io_ctl *io_ctl,
1207 struct btrfs_path *path)
1208 {
1209 return __btrfs_wait_cache_io(root, trans, NULL, io_ctl, path, 0);
1210 }
1211
1212 int btrfs_wait_cache_io(struct btrfs_trans_handle *trans,
1213 struct btrfs_block_group_cache *block_group,
1214 struct btrfs_path *path)
1215 {
1216 return __btrfs_wait_cache_io(block_group->fs_info->tree_root, trans,
1217 block_group, &block_group->io_ctl,
1218 path, block_group->key.objectid);
1219 }
1220
1221 /**
1222 * __btrfs_write_out_cache - write out cached info to an inode
1223 * @root - the root the inode belongs to
1224 * @ctl - the free space cache we are going to write out
1225 * @block_group - the block_group for this cache if it belongs to a block_group
1226 * @trans - the trans handle
1227 *
1228 * This function writes out a free space cache struct to disk for quick recovery
1229 * on mount. This will return 0 if it was successful in writing the cache out,
1230 * or an errno if it was not.
1231 */
1232 static int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode,
1233 struct btrfs_free_space_ctl *ctl,
1234 struct btrfs_block_group_cache *block_group,
1235 struct btrfs_io_ctl *io_ctl,
1236 struct btrfs_trans_handle *trans)
1237 {
1238 struct btrfs_fs_info *fs_info = root->fs_info;
1239 struct extent_state *cached_state = NULL;
1240 LIST_HEAD(bitmap_list);
1241 int entries = 0;
1242 int bitmaps = 0;
1243 int ret;
1244 int must_iput = 0;
1245
1246 if (!i_size_read(inode))
1247 return -EIO;
1248
1249 WARN_ON(io_ctl->pages);
1250 ret = io_ctl_init(io_ctl, inode, 1);
1251 if (ret)
1252 return ret;
1253
1254 if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA)) {
1255 down_write(&block_group->data_rwsem);
1256 spin_lock(&block_group->lock);
1257 if (block_group->delalloc_bytes) {
1258 block_group->disk_cache_state = BTRFS_DC_WRITTEN;
1259 spin_unlock(&block_group->lock);
1260 up_write(&block_group->data_rwsem);
1261 BTRFS_I(inode)->generation = 0;
1262 ret = 0;
1263 must_iput = 1;
1264 goto out;
1265 }
1266 spin_unlock(&block_group->lock);
1267 }
1268
1269 /* Lock all pages first so we can lock the extent safely. */
1270 ret = io_ctl_prepare_pages(io_ctl, inode, 0);
1271 if (ret)
1272 goto out_unlock;
1273
1274 lock_extent_bits(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
1275 &cached_state);
1276
1277 io_ctl_set_generation(io_ctl, trans->transid);
1278
1279 mutex_lock(&ctl->cache_writeout_mutex);
1280 /* Write out the extent entries in the free space cache */
1281 spin_lock(&ctl->tree_lock);
1282 ret = write_cache_extent_entries(io_ctl, ctl,
1283 block_group, &entries, &bitmaps,
1284 &bitmap_list);
1285 if (ret)
1286 goto out_nospc_locked;
1287
1288 /*
1289 * Some spaces that are freed in the current transaction are pinned,
1290 * they will be added into free space cache after the transaction is
1291 * committed, we shouldn't lose them.
1292 *
1293 * If this changes while we are working we'll get added back to
1294 * the dirty list and redo it. No locking needed
1295 */
1296 ret = write_pinned_extent_entries(fs_info, block_group,
1297 io_ctl, &entries);
1298 if (ret)
1299 goto out_nospc_locked;
1300
1301 /*
1302 * At last, we write out all the bitmaps and keep cache_writeout_mutex
1303 * locked while doing it because a concurrent trim can be manipulating
1304 * or freeing the bitmap.
1305 */
1306 ret = write_bitmap_entries(io_ctl, &bitmap_list);
1307 spin_unlock(&ctl->tree_lock);
1308 mutex_unlock(&ctl->cache_writeout_mutex);
1309 if (ret)
1310 goto out_nospc;
1311
1312 /* Zero out the rest of the pages just to make sure */
1313 io_ctl_zero_remaining_pages(io_ctl);
1314
1315 /* Everything is written out, now we dirty the pages in the file. */
1316 ret = btrfs_dirty_pages(inode, io_ctl->pages, io_ctl->num_pages, 0,
1317 i_size_read(inode), &cached_state);
1318 if (ret)
1319 goto out_nospc;
1320
1321 if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
1322 up_write(&block_group->data_rwsem);
1323 /*
1324 * Release the pages and unlock the extent, we will flush
1325 * them out later
1326 */
1327 io_ctl_drop_pages(io_ctl);
1328
1329 unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
1330 i_size_read(inode) - 1, &cached_state);
1331
1332 /*
1333 * at this point the pages are under IO and we're happy,
1334 * The caller is responsible for waiting on them and updating the
1335 * the cache and the inode
1336 */
1337 io_ctl->entries = entries;
1338 io_ctl->bitmaps = bitmaps;
1339
1340 ret = btrfs_fdatawrite_range(inode, 0, (u64)-1);
1341 if (ret)
1342 goto out;
1343
1344 return 0;
1345
1346 out:
1347 io_ctl->inode = NULL;
1348 io_ctl_free(io_ctl);
1349 if (ret) {
1350 invalidate_inode_pages2(inode->i_mapping);
1351 BTRFS_I(inode)->generation = 0;
1352 }
1353 btrfs_update_inode(trans, root, inode);
1354 if (must_iput)
1355 iput(inode);
1356 return ret;
1357
1358 out_nospc_locked:
1359 cleanup_bitmap_list(&bitmap_list);
1360 spin_unlock(&ctl->tree_lock);
1361 mutex_unlock(&ctl->cache_writeout_mutex);
1362
1363 out_nospc:
1364 cleanup_write_cache_enospc(inode, io_ctl, &cached_state);
1365
1366 out_unlock:
1367 if (block_group && (block_group->flags & BTRFS_BLOCK_GROUP_DATA))
1368 up_write(&block_group->data_rwsem);
1369
1370 goto out;
1371 }
1372
1373 int btrfs_write_out_cache(struct btrfs_fs_info *fs_info,
1374 struct btrfs_trans_handle *trans,
1375 struct btrfs_block_group_cache *block_group,
1376 struct btrfs_path *path)
1377 {
1378 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1379 struct inode *inode;
1380 int ret = 0;
1381
1382 spin_lock(&block_group->lock);
1383 if (block_group->disk_cache_state < BTRFS_DC_SETUP) {
1384 spin_unlock(&block_group->lock);
1385 return 0;
1386 }
1387 spin_unlock(&block_group->lock);
1388
1389 inode = lookup_free_space_inode(fs_info, block_group, path);
1390 if (IS_ERR(inode))
1391 return 0;
1392
1393 ret = __btrfs_write_out_cache(fs_info->tree_root, inode, ctl,
1394 block_group, &block_group->io_ctl, trans);
1395 if (ret) {
1396 #ifdef DEBUG
1397 btrfs_err(fs_info,
1398 "failed to write free space cache for block group %llu",
1399 block_group->key.objectid);
1400 #endif
1401 spin_lock(&block_group->lock);
1402 block_group->disk_cache_state = BTRFS_DC_ERROR;
1403 spin_unlock(&block_group->lock);
1404
1405 block_group->io_ctl.inode = NULL;
1406 iput(inode);
1407 }
1408
1409 /*
1410 * if ret == 0 the caller is expected to call btrfs_wait_cache_io
1411 * to wait for IO and put the inode
1412 */
1413
1414 return ret;
1415 }
1416
1417 static inline unsigned long offset_to_bit(u64 bitmap_start, u32 unit,
1418 u64 offset)
1419 {
1420 ASSERT(offset >= bitmap_start);
1421 offset -= bitmap_start;
1422 return (unsigned long)(div_u64(offset, unit));
1423 }
1424
1425 static inline unsigned long bytes_to_bits(u64 bytes, u32 unit)
1426 {
1427 return (unsigned long)(div_u64(bytes, unit));
1428 }
1429
1430 static inline u64 offset_to_bitmap(struct btrfs_free_space_ctl *ctl,
1431 u64 offset)
1432 {
1433 u64 bitmap_start;
1434 u64 bytes_per_bitmap;
1435
1436 bytes_per_bitmap = BITS_PER_BITMAP * ctl->unit;
1437 bitmap_start = offset - ctl->start;
1438 bitmap_start = div64_u64(bitmap_start, bytes_per_bitmap);
1439 bitmap_start *= bytes_per_bitmap;
1440 bitmap_start += ctl->start;
1441
1442 return bitmap_start;
1443 }
1444
1445 static int tree_insert_offset(struct rb_root *root, u64 offset,
1446 struct rb_node *node, int bitmap)
1447 {
1448 struct rb_node **p = &root->rb_node;
1449 struct rb_node *parent = NULL;
1450 struct btrfs_free_space *info;
1451
1452 while (*p) {
1453 parent = *p;
1454 info = rb_entry(parent, struct btrfs_free_space, offset_index);
1455
1456 if (offset < info->offset) {
1457 p = &(*p)->rb_left;
1458 } else if (offset > info->offset) {
1459 p = &(*p)->rb_right;
1460 } else {
1461 /*
1462 * we could have a bitmap entry and an extent entry
1463 * share the same offset. If this is the case, we want
1464 * the extent entry to always be found first if we do a
1465 * linear search through the tree, since we want to have
1466 * the quickest allocation time, and allocating from an
1467 * extent is faster than allocating from a bitmap. So
1468 * if we're inserting a bitmap and we find an entry at
1469 * this offset, we want to go right, or after this entry
1470 * logically. If we are inserting an extent and we've
1471 * found a bitmap, we want to go left, or before
1472 * logically.
1473 */
1474 if (bitmap) {
1475 if (info->bitmap) {
1476 WARN_ON_ONCE(1);
1477 return -EEXIST;
1478 }
1479 p = &(*p)->rb_right;
1480 } else {
1481 if (!info->bitmap) {
1482 WARN_ON_ONCE(1);
1483 return -EEXIST;
1484 }
1485 p = &(*p)->rb_left;
1486 }
1487 }
1488 }
1489
1490 rb_link_node(node, parent, p);
1491 rb_insert_color(node, root);
1492
1493 return 0;
1494 }
1495
1496 /*
1497 * searches the tree for the given offset.
1498 *
1499 * fuzzy - If this is set, then we are trying to make an allocation, and we just
1500 * want a section that has at least bytes size and comes at or after the given
1501 * offset.
1502 */
1503 static struct btrfs_free_space *
1504 tree_search_offset(struct btrfs_free_space_ctl *ctl,
1505 u64 offset, int bitmap_only, int fuzzy)
1506 {
1507 struct rb_node *n = ctl->free_space_offset.rb_node;
1508 struct btrfs_free_space *entry, *prev = NULL;
1509
1510 /* find entry that is closest to the 'offset' */
1511 while (1) {
1512 if (!n) {
1513 entry = NULL;
1514 break;
1515 }
1516
1517 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1518 prev = entry;
1519
1520 if (offset < entry->offset)
1521 n = n->rb_left;
1522 else if (offset > entry->offset)
1523 n = n->rb_right;
1524 else
1525 break;
1526 }
1527
1528 if (bitmap_only) {
1529 if (!entry)
1530 return NULL;
1531 if (entry->bitmap)
1532 return entry;
1533
1534 /*
1535 * bitmap entry and extent entry may share same offset,
1536 * in that case, bitmap entry comes after extent entry.
1537 */
1538 n = rb_next(n);
1539 if (!n)
1540 return NULL;
1541 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1542 if (entry->offset != offset)
1543 return NULL;
1544
1545 WARN_ON(!entry->bitmap);
1546 return entry;
1547 } else if (entry) {
1548 if (entry->bitmap) {
1549 /*
1550 * if previous extent entry covers the offset,
1551 * we should return it instead of the bitmap entry
1552 */
1553 n = rb_prev(&entry->offset_index);
1554 if (n) {
1555 prev = rb_entry(n, struct btrfs_free_space,
1556 offset_index);
1557 if (!prev->bitmap &&
1558 prev->offset + prev->bytes > offset)
1559 entry = prev;
1560 }
1561 }
1562 return entry;
1563 }
1564
1565 if (!prev)
1566 return NULL;
1567
1568 /* find last entry before the 'offset' */
1569 entry = prev;
1570 if (entry->offset > offset) {
1571 n = rb_prev(&entry->offset_index);
1572 if (n) {
1573 entry = rb_entry(n, struct btrfs_free_space,
1574 offset_index);
1575 ASSERT(entry->offset <= offset);
1576 } else {
1577 if (fuzzy)
1578 return entry;
1579 else
1580 return NULL;
1581 }
1582 }
1583
1584 if (entry->bitmap) {
1585 n = rb_prev(&entry->offset_index);
1586 if (n) {
1587 prev = rb_entry(n, struct btrfs_free_space,
1588 offset_index);
1589 if (!prev->bitmap &&
1590 prev->offset + prev->bytes > offset)
1591 return prev;
1592 }
1593 if (entry->offset + BITS_PER_BITMAP * ctl->unit > offset)
1594 return entry;
1595 } else if (entry->offset + entry->bytes > offset)
1596 return entry;
1597
1598 if (!fuzzy)
1599 return NULL;
1600
1601 while (1) {
1602 if (entry->bitmap) {
1603 if (entry->offset + BITS_PER_BITMAP *
1604 ctl->unit > offset)
1605 break;
1606 } else {
1607 if (entry->offset + entry->bytes > offset)
1608 break;
1609 }
1610
1611 n = rb_next(&entry->offset_index);
1612 if (!n)
1613 return NULL;
1614 entry = rb_entry(n, struct btrfs_free_space, offset_index);
1615 }
1616 return entry;
1617 }
1618
1619 static inline void
1620 __unlink_free_space(struct btrfs_free_space_ctl *ctl,
1621 struct btrfs_free_space *info)
1622 {
1623 rb_erase(&info->offset_index, &ctl->free_space_offset);
1624 ctl->free_extents--;
1625 }
1626
1627 static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
1628 struct btrfs_free_space *info)
1629 {
1630 __unlink_free_space(ctl, info);
1631 ctl->free_space -= info->bytes;
1632 }
1633
1634 static int link_free_space(struct btrfs_free_space_ctl *ctl,
1635 struct btrfs_free_space *info)
1636 {
1637 int ret = 0;
1638
1639 ASSERT(info->bytes || info->bitmap);
1640 ret = tree_insert_offset(&ctl->free_space_offset, info->offset,
1641 &info->offset_index, (info->bitmap != NULL));
1642 if (ret)
1643 return ret;
1644
1645 ctl->free_space += info->bytes;
1646 ctl->free_extents++;
1647 return ret;
1648 }
1649
1650 static void recalculate_thresholds(struct btrfs_free_space_ctl *ctl)
1651 {
1652 struct btrfs_block_group_cache *block_group = ctl->private;
1653 u64 max_bytes;
1654 u64 bitmap_bytes;
1655 u64 extent_bytes;
1656 u64 size = block_group->key.offset;
1657 u64 bytes_per_bg = BITS_PER_BITMAP * ctl->unit;
1658 u64 max_bitmaps = div64_u64(size + bytes_per_bg - 1, bytes_per_bg);
1659
1660 max_bitmaps = max_t(u64, max_bitmaps, 1);
1661
1662 ASSERT(ctl->total_bitmaps <= max_bitmaps);
1663
1664 /*
1665 * The goal is to keep the total amount of memory used per 1gb of space
1666 * at or below 32k, so we need to adjust how much memory we allow to be
1667 * used by extent based free space tracking
1668 */
1669 if (size < SZ_1G)
1670 max_bytes = MAX_CACHE_BYTES_PER_GIG;
1671 else
1672 max_bytes = MAX_CACHE_BYTES_PER_GIG * div_u64(size, SZ_1G);
1673
1674 /*
1675 * we want to account for 1 more bitmap than what we have so we can make
1676 * sure we don't go over our overall goal of MAX_CACHE_BYTES_PER_GIG as
1677 * we add more bitmaps.
1678 */
1679 bitmap_bytes = (ctl->total_bitmaps + 1) * ctl->unit;
1680
1681 if (bitmap_bytes >= max_bytes) {
1682 ctl->extents_thresh = 0;
1683 return;
1684 }
1685
1686 /*
1687 * we want the extent entry threshold to always be at most 1/2 the max
1688 * bytes we can have, or whatever is less than that.
1689 */
1690 extent_bytes = max_bytes - bitmap_bytes;
1691 extent_bytes = min_t(u64, extent_bytes, max_bytes >> 1);
1692
1693 ctl->extents_thresh =
1694 div_u64(extent_bytes, sizeof(struct btrfs_free_space));
1695 }
1696
1697 static inline void __bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
1698 struct btrfs_free_space *info,
1699 u64 offset, u64 bytes)
1700 {
1701 unsigned long start, count;
1702
1703 start = offset_to_bit(info->offset, ctl->unit, offset);
1704 count = bytes_to_bits(bytes, ctl->unit);
1705 ASSERT(start + count <= BITS_PER_BITMAP);
1706
1707 bitmap_clear(info->bitmap, start, count);
1708
1709 info->bytes -= bytes;
1710 if (info->max_extent_size > ctl->unit)
1711 info->max_extent_size = 0;
1712 }
1713
1714 static void bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
1715 struct btrfs_free_space *info, u64 offset,
1716 u64 bytes)
1717 {
1718 __bitmap_clear_bits(ctl, info, offset, bytes);
1719 ctl->free_space -= bytes;
1720 }
1721
1722 static void bitmap_set_bits(struct btrfs_free_space_ctl *ctl,
1723 struct btrfs_free_space *info, u64 offset,
1724 u64 bytes)
1725 {
1726 unsigned long start, count;
1727
1728 start = offset_to_bit(info->offset, ctl->unit, offset);
1729 count = bytes_to_bits(bytes, ctl->unit);
1730 ASSERT(start + count <= BITS_PER_BITMAP);
1731
1732 bitmap_set(info->bitmap, start, count);
1733
1734 info->bytes += bytes;
1735 ctl->free_space += bytes;
1736 }
1737
1738 /*
1739 * If we can not find suitable extent, we will use bytes to record
1740 * the size of the max extent.
1741 */
1742 static int search_bitmap(struct btrfs_free_space_ctl *ctl,
1743 struct btrfs_free_space *bitmap_info, u64 *offset,
1744 u64 *bytes, bool for_alloc)
1745 {
1746 unsigned long found_bits = 0;
1747 unsigned long max_bits = 0;
1748 unsigned long bits, i;
1749 unsigned long next_zero;
1750 unsigned long extent_bits;
1751
1752 /*
1753 * Skip searching the bitmap if we don't have a contiguous section that
1754 * is large enough for this allocation.
1755 */
1756 if (for_alloc &&
1757 bitmap_info->max_extent_size &&
1758 bitmap_info->max_extent_size < *bytes) {
1759 *bytes = bitmap_info->max_extent_size;
1760 return -1;
1761 }
1762
1763 i = offset_to_bit(bitmap_info->offset, ctl->unit,
1764 max_t(u64, *offset, bitmap_info->offset));
1765 bits = bytes_to_bits(*bytes, ctl->unit);
1766
1767 for_each_set_bit_from(i, bitmap_info->bitmap, BITS_PER_BITMAP) {
1768 if (for_alloc && bits == 1) {
1769 found_bits = 1;
1770 break;
1771 }
1772 next_zero = find_next_zero_bit(bitmap_info->bitmap,
1773 BITS_PER_BITMAP, i);
1774 extent_bits = next_zero - i;
1775 if (extent_bits >= bits) {
1776 found_bits = extent_bits;
1777 break;
1778 } else if (extent_bits > max_bits) {
1779 max_bits = extent_bits;
1780 }
1781 i = next_zero;
1782 }
1783
1784 if (found_bits) {
1785 *offset = (u64)(i * ctl->unit) + bitmap_info->offset;
1786 *bytes = (u64)(found_bits) * ctl->unit;
1787 return 0;
1788 }
1789
1790 *bytes = (u64)(max_bits) * ctl->unit;
1791 bitmap_info->max_extent_size = *bytes;
1792 return -1;
1793 }
1794
1795 static inline u64 get_max_extent_size(struct btrfs_free_space *entry)
1796 {
1797 if (entry->bitmap)
1798 return entry->max_extent_size;
1799 return entry->bytes;
1800 }
1801
1802 /* Cache the size of the max extent in bytes */
1803 static struct btrfs_free_space *
1804 find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes,
1805 unsigned long align, u64 *max_extent_size)
1806 {
1807 struct btrfs_free_space *entry;
1808 struct rb_node *node;
1809 u64 tmp;
1810 u64 align_off;
1811 int ret;
1812
1813 if (!ctl->free_space_offset.rb_node)
1814 goto out;
1815
1816 entry = tree_search_offset(ctl, offset_to_bitmap(ctl, *offset), 0, 1);
1817 if (!entry)
1818 goto out;
1819
1820 for (node = &entry->offset_index; node; node = rb_next(node)) {
1821 entry = rb_entry(node, struct btrfs_free_space, offset_index);
1822 if (entry->bytes < *bytes) {
1823 *max_extent_size = max(get_max_extent_size(entry),
1824 *max_extent_size);
1825 continue;
1826 }
1827
1828 /* make sure the space returned is big enough
1829 * to match our requested alignment
1830 */
1831 if (*bytes >= align) {
1832 tmp = entry->offset - ctl->start + align - 1;
1833 tmp = div64_u64(tmp, align);
1834 tmp = tmp * align + ctl->start;
1835 align_off = tmp - entry->offset;
1836 } else {
1837 align_off = 0;
1838 tmp = entry->offset;
1839 }
1840
1841 if (entry->bytes < *bytes + align_off) {
1842 *max_extent_size = max(get_max_extent_size(entry),
1843 *max_extent_size);
1844 continue;
1845 }
1846
1847 if (entry->bitmap) {
1848 u64 size = *bytes;
1849
1850 ret = search_bitmap(ctl, entry, &tmp, &size, true);
1851 if (!ret) {
1852 *offset = tmp;
1853 *bytes = size;
1854 return entry;
1855 } else {
1856 *max_extent_size =
1857 max(get_max_extent_size(entry),
1858 *max_extent_size);
1859 }
1860 continue;
1861 }
1862
1863 *offset = tmp;
1864 *bytes = entry->bytes - align_off;
1865 return entry;
1866 }
1867 out:
1868 return NULL;
1869 }
1870
1871 static void add_new_bitmap(struct btrfs_free_space_ctl *ctl,
1872 struct btrfs_free_space *info, u64 offset)
1873 {
1874 info->offset = offset_to_bitmap(ctl, offset);
1875 info->bytes = 0;
1876 INIT_LIST_HEAD(&info->list);
1877 link_free_space(ctl, info);
1878 ctl->total_bitmaps++;
1879
1880 ctl->op->recalc_thresholds(ctl);
1881 }
1882
1883 static void free_bitmap(struct btrfs_free_space_ctl *ctl,
1884 struct btrfs_free_space *bitmap_info)
1885 {
1886 unlink_free_space(ctl, bitmap_info);
1887 kfree(bitmap_info->bitmap);
1888 kmem_cache_free(btrfs_free_space_cachep, bitmap_info);
1889 ctl->total_bitmaps--;
1890 ctl->op->recalc_thresholds(ctl);
1891 }
1892
1893 static noinline int remove_from_bitmap(struct btrfs_free_space_ctl *ctl,
1894 struct btrfs_free_space *bitmap_info,
1895 u64 *offset, u64 *bytes)
1896 {
1897 u64 end;
1898 u64 search_start, search_bytes;
1899 int ret;
1900
1901 again:
1902 end = bitmap_info->offset + (u64)(BITS_PER_BITMAP * ctl->unit) - 1;
1903
1904 /*
1905 * We need to search for bits in this bitmap. We could only cover some
1906 * of the extent in this bitmap thanks to how we add space, so we need
1907 * to search for as much as it as we can and clear that amount, and then
1908 * go searching for the next bit.
1909 */
1910 search_start = *offset;
1911 search_bytes = ctl->unit;
1912 search_bytes = min(search_bytes, end - search_start + 1);
1913 ret = search_bitmap(ctl, bitmap_info, &search_start, &search_bytes,
1914 false);
1915 if (ret < 0 || search_start != *offset)
1916 return -EINVAL;
1917
1918 /* We may have found more bits than what we need */
1919 search_bytes = min(search_bytes, *bytes);
1920
1921 /* Cannot clear past the end of the bitmap */
1922 search_bytes = min(search_bytes, end - search_start + 1);
1923
1924 bitmap_clear_bits(ctl, bitmap_info, search_start, search_bytes);
1925 *offset += search_bytes;
1926 *bytes -= search_bytes;
1927
1928 if (*bytes) {
1929 struct rb_node *next = rb_next(&bitmap_info->offset_index);
1930 if (!bitmap_info->bytes)
1931 free_bitmap(ctl, bitmap_info);
1932
1933 /*
1934 * no entry after this bitmap, but we still have bytes to
1935 * remove, so something has gone wrong.
1936 */
1937 if (!next)
1938 return -EINVAL;
1939
1940 bitmap_info = rb_entry(next, struct btrfs_free_space,
1941 offset_index);
1942
1943 /*
1944 * if the next entry isn't a bitmap we need to return to let the
1945 * extent stuff do its work.
1946 */
1947 if (!bitmap_info->bitmap)
1948 return -EAGAIN;
1949
1950 /*
1951 * Ok the next item is a bitmap, but it may not actually hold
1952 * the information for the rest of this free space stuff, so
1953 * look for it, and if we don't find it return so we can try
1954 * everything over again.
1955 */
1956 search_start = *offset;
1957 search_bytes = ctl->unit;
1958 ret = search_bitmap(ctl, bitmap_info, &search_start,
1959 &search_bytes, false);
1960 if (ret < 0 || search_start != *offset)
1961 return -EAGAIN;
1962
1963 goto again;
1964 } else if (!bitmap_info->bytes)
1965 free_bitmap(ctl, bitmap_info);
1966
1967 return 0;
1968 }
1969
1970 static u64 add_bytes_to_bitmap(struct btrfs_free_space_ctl *ctl,
1971 struct btrfs_free_space *info, u64 offset,
1972 u64 bytes)
1973 {
1974 u64 bytes_to_set = 0;
1975 u64 end;
1976
1977 end = info->offset + (u64)(BITS_PER_BITMAP * ctl->unit);
1978
1979 bytes_to_set = min(end - offset, bytes);
1980
1981 bitmap_set_bits(ctl, info, offset, bytes_to_set);
1982
1983 /*
1984 * We set some bytes, we have no idea what the max extent size is
1985 * anymore.
1986 */
1987 info->max_extent_size = 0;
1988
1989 return bytes_to_set;
1990
1991 }
1992
1993 static bool use_bitmap(struct btrfs_free_space_ctl *ctl,
1994 struct btrfs_free_space *info)
1995 {
1996 struct btrfs_block_group_cache *block_group = ctl->private;
1997 struct btrfs_fs_info *fs_info = block_group->fs_info;
1998 bool forced = false;
1999
2000 #ifdef CONFIG_BTRFS_DEBUG
2001 if (btrfs_should_fragment_free_space(block_group))
2002 forced = true;
2003 #endif
2004
2005 /*
2006 * If we are below the extents threshold then we can add this as an
2007 * extent, and don't have to deal with the bitmap
2008 */
2009 if (!forced && ctl->free_extents < ctl->extents_thresh) {
2010 /*
2011 * If this block group has some small extents we don't want to
2012 * use up all of our free slots in the cache with them, we want
2013 * to reserve them to larger extents, however if we have plenty
2014 * of cache left then go ahead an dadd them, no sense in adding
2015 * the overhead of a bitmap if we don't have to.
2016 */
2017 if (info->bytes <= fs_info->sectorsize * 4) {
2018 if (ctl->free_extents * 2 <= ctl->extents_thresh)
2019 return false;
2020 } else {
2021 return false;
2022 }
2023 }
2024
2025 /*
2026 * The original block groups from mkfs can be really small, like 8
2027 * megabytes, so don't bother with a bitmap for those entries. However
2028 * some block groups can be smaller than what a bitmap would cover but
2029 * are still large enough that they could overflow the 32k memory limit,
2030 * so allow those block groups to still be allowed to have a bitmap
2031 * entry.
2032 */
2033 if (((BITS_PER_BITMAP * ctl->unit) >> 1) > block_group->key.offset)
2034 return false;
2035
2036 return true;
2037 }
2038
2039 static const struct btrfs_free_space_op free_space_op = {
2040 .recalc_thresholds = recalculate_thresholds,
2041 .use_bitmap = use_bitmap,
2042 };
2043
2044 static int insert_into_bitmap(struct btrfs_free_space_ctl *ctl,
2045 struct btrfs_free_space *info)
2046 {
2047 struct btrfs_free_space *bitmap_info;
2048 struct btrfs_block_group_cache *block_group = NULL;
2049 int added = 0;
2050 u64 bytes, offset, bytes_added;
2051 int ret;
2052
2053 bytes = info->bytes;
2054 offset = info->offset;
2055
2056 if (!ctl->op->use_bitmap(ctl, info))
2057 return 0;
2058
2059 if (ctl->op == &free_space_op)
2060 block_group = ctl->private;
2061 again:
2062 /*
2063 * Since we link bitmaps right into the cluster we need to see if we
2064 * have a cluster here, and if so and it has our bitmap we need to add
2065 * the free space to that bitmap.
2066 */
2067 if (block_group && !list_empty(&block_group->cluster_list)) {
2068 struct btrfs_free_cluster *cluster;
2069 struct rb_node *node;
2070 struct btrfs_free_space *entry;
2071
2072 cluster = list_entry(block_group->cluster_list.next,
2073 struct btrfs_free_cluster,
2074 block_group_list);
2075 spin_lock(&cluster->lock);
2076 node = rb_first(&cluster->root);
2077 if (!node) {
2078 spin_unlock(&cluster->lock);
2079 goto no_cluster_bitmap;
2080 }
2081
2082 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2083 if (!entry->bitmap) {
2084 spin_unlock(&cluster->lock);
2085 goto no_cluster_bitmap;
2086 }
2087
2088 if (entry->offset == offset_to_bitmap(ctl, offset)) {
2089 bytes_added = add_bytes_to_bitmap(ctl, entry,
2090 offset, bytes);
2091 bytes -= bytes_added;
2092 offset += bytes_added;
2093 }
2094 spin_unlock(&cluster->lock);
2095 if (!bytes) {
2096 ret = 1;
2097 goto out;
2098 }
2099 }
2100
2101 no_cluster_bitmap:
2102 bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
2103 1, 0);
2104 if (!bitmap_info) {
2105 ASSERT(added == 0);
2106 goto new_bitmap;
2107 }
2108
2109 bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes);
2110 bytes -= bytes_added;
2111 offset += bytes_added;
2112 added = 0;
2113
2114 if (!bytes) {
2115 ret = 1;
2116 goto out;
2117 } else
2118 goto again;
2119
2120 new_bitmap:
2121 if (info && info->bitmap) {
2122 add_new_bitmap(ctl, info, offset);
2123 added = 1;
2124 info = NULL;
2125 goto again;
2126 } else {
2127 spin_unlock(&ctl->tree_lock);
2128
2129 /* no pre-allocated info, allocate a new one */
2130 if (!info) {
2131 info = kmem_cache_zalloc(btrfs_free_space_cachep,
2132 GFP_NOFS);
2133 if (!info) {
2134 spin_lock(&ctl->tree_lock);
2135 ret = -ENOMEM;
2136 goto out;
2137 }
2138 }
2139
2140 /* allocate the bitmap */
2141 info->bitmap = kzalloc(PAGE_SIZE, GFP_NOFS);
2142 spin_lock(&ctl->tree_lock);
2143 if (!info->bitmap) {
2144 ret = -ENOMEM;
2145 goto out;
2146 }
2147 goto again;
2148 }
2149
2150 out:
2151 if (info) {
2152 kfree(info->bitmap);
2153 kmem_cache_free(btrfs_free_space_cachep, info);
2154 }
2155
2156 return ret;
2157 }
2158
2159 static bool try_merge_free_space(struct btrfs_free_space_ctl *ctl,
2160 struct btrfs_free_space *info, bool update_stat)
2161 {
2162 struct btrfs_free_space *left_info;
2163 struct btrfs_free_space *right_info;
2164 bool merged = false;
2165 u64 offset = info->offset;
2166 u64 bytes = info->bytes;
2167
2168 /*
2169 * first we want to see if there is free space adjacent to the range we
2170 * are adding, if there is remove that struct and add a new one to
2171 * cover the entire range
2172 */
2173 right_info = tree_search_offset(ctl, offset + bytes, 0, 0);
2174 if (right_info && rb_prev(&right_info->offset_index))
2175 left_info = rb_entry(rb_prev(&right_info->offset_index),
2176 struct btrfs_free_space, offset_index);
2177 else
2178 left_info = tree_search_offset(ctl, offset - 1, 0, 0);
2179
2180 if (right_info && !right_info->bitmap) {
2181 if (update_stat)
2182 unlink_free_space(ctl, right_info);
2183 else
2184 __unlink_free_space(ctl, right_info);
2185 info->bytes += right_info->bytes;
2186 kmem_cache_free(btrfs_free_space_cachep, right_info);
2187 merged = true;
2188 }
2189
2190 if (left_info && !left_info->bitmap &&
2191 left_info->offset + left_info->bytes == offset) {
2192 if (update_stat)
2193 unlink_free_space(ctl, left_info);
2194 else
2195 __unlink_free_space(ctl, left_info);
2196 info->offset = left_info->offset;
2197 info->bytes += left_info->bytes;
2198 kmem_cache_free(btrfs_free_space_cachep, left_info);
2199 merged = true;
2200 }
2201
2202 return merged;
2203 }
2204
2205 static bool steal_from_bitmap_to_end(struct btrfs_free_space_ctl *ctl,
2206 struct btrfs_free_space *info,
2207 bool update_stat)
2208 {
2209 struct btrfs_free_space *bitmap;
2210 unsigned long i;
2211 unsigned long j;
2212 const u64 end = info->offset + info->bytes;
2213 const u64 bitmap_offset = offset_to_bitmap(ctl, end);
2214 u64 bytes;
2215
2216 bitmap = tree_search_offset(ctl, bitmap_offset, 1, 0);
2217 if (!bitmap)
2218 return false;
2219
2220 i = offset_to_bit(bitmap->offset, ctl->unit, end);
2221 j = find_next_zero_bit(bitmap->bitmap, BITS_PER_BITMAP, i);
2222 if (j == i)
2223 return false;
2224 bytes = (j - i) * ctl->unit;
2225 info->bytes += bytes;
2226
2227 if (update_stat)
2228 bitmap_clear_bits(ctl, bitmap, end, bytes);
2229 else
2230 __bitmap_clear_bits(ctl, bitmap, end, bytes);
2231
2232 if (!bitmap->bytes)
2233 free_bitmap(ctl, bitmap);
2234
2235 return true;
2236 }
2237
2238 static bool steal_from_bitmap_to_front(struct btrfs_free_space_ctl *ctl,
2239 struct btrfs_free_space *info,
2240 bool update_stat)
2241 {
2242 struct btrfs_free_space *bitmap;
2243 u64 bitmap_offset;
2244 unsigned long i;
2245 unsigned long j;
2246 unsigned long prev_j;
2247 u64 bytes;
2248
2249 bitmap_offset = offset_to_bitmap(ctl, info->offset);
2250 /* If we're on a boundary, try the previous logical bitmap. */
2251 if (bitmap_offset == info->offset) {
2252 if (info->offset == 0)
2253 return false;
2254 bitmap_offset = offset_to_bitmap(ctl, info->offset - 1);
2255 }
2256
2257 bitmap = tree_search_offset(ctl, bitmap_offset, 1, 0);
2258 if (!bitmap)
2259 return false;
2260
2261 i = offset_to_bit(bitmap->offset, ctl->unit, info->offset) - 1;
2262 j = 0;
2263 prev_j = (unsigned long)-1;
2264 for_each_clear_bit_from(j, bitmap->bitmap, BITS_PER_BITMAP) {
2265 if (j > i)
2266 break;
2267 prev_j = j;
2268 }
2269 if (prev_j == i)
2270 return false;
2271
2272 if (prev_j == (unsigned long)-1)
2273 bytes = (i + 1) * ctl->unit;
2274 else
2275 bytes = (i - prev_j) * ctl->unit;
2276
2277 info->offset -= bytes;
2278 info->bytes += bytes;
2279
2280 if (update_stat)
2281 bitmap_clear_bits(ctl, bitmap, info->offset, bytes);
2282 else
2283 __bitmap_clear_bits(ctl, bitmap, info->offset, bytes);
2284
2285 if (!bitmap->bytes)
2286 free_bitmap(ctl, bitmap);
2287
2288 return true;
2289 }
2290
2291 /*
2292 * We prefer always to allocate from extent entries, both for clustered and
2293 * non-clustered allocation requests. So when attempting to add a new extent
2294 * entry, try to see if there's adjacent free space in bitmap entries, and if
2295 * there is, migrate that space from the bitmaps to the extent.
2296 * Like this we get better chances of satisfying space allocation requests
2297 * because we attempt to satisfy them based on a single cache entry, and never
2298 * on 2 or more entries - even if the entries represent a contiguous free space
2299 * region (e.g. 1 extent entry + 1 bitmap entry starting where the extent entry
2300 * ends).
2301 */
2302 static void steal_from_bitmap(struct btrfs_free_space_ctl *ctl,
2303 struct btrfs_free_space *info,
2304 bool update_stat)
2305 {
2306 /*
2307 * Only work with disconnected entries, as we can change their offset,
2308 * and must be extent entries.
2309 */
2310 ASSERT(!info->bitmap);
2311 ASSERT(RB_EMPTY_NODE(&info->offset_index));
2312
2313 if (ctl->total_bitmaps > 0) {
2314 bool stole_end;
2315 bool stole_front = false;
2316
2317 stole_end = steal_from_bitmap_to_end(ctl, info, update_stat);
2318 if (ctl->total_bitmaps > 0)
2319 stole_front = steal_from_bitmap_to_front(ctl, info,
2320 update_stat);
2321
2322 if (stole_end || stole_front)
2323 try_merge_free_space(ctl, info, update_stat);
2324 }
2325 }
2326
2327 int __btrfs_add_free_space(struct btrfs_fs_info *fs_info,
2328 struct btrfs_free_space_ctl *ctl,
2329 u64 offset, u64 bytes)
2330 {
2331 struct btrfs_free_space *info;
2332 int ret = 0;
2333
2334 info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
2335 if (!info)
2336 return -ENOMEM;
2337
2338 info->offset = offset;
2339 info->bytes = bytes;
2340 RB_CLEAR_NODE(&info->offset_index);
2341
2342 spin_lock(&ctl->tree_lock);
2343
2344 if (try_merge_free_space(ctl, info, true))
2345 goto link;
2346
2347 /*
2348 * There was no extent directly to the left or right of this new
2349 * extent then we know we're going to have to allocate a new extent, so
2350 * before we do that see if we need to drop this into a bitmap
2351 */
2352 ret = insert_into_bitmap(ctl, info);
2353 if (ret < 0) {
2354 goto out;
2355 } else if (ret) {
2356 ret = 0;
2357 goto out;
2358 }
2359 link:
2360 /*
2361 * Only steal free space from adjacent bitmaps if we're sure we're not
2362 * going to add the new free space to existing bitmap entries - because
2363 * that would mean unnecessary work that would be reverted. Therefore
2364 * attempt to steal space from bitmaps if we're adding an extent entry.
2365 */
2366 steal_from_bitmap(ctl, info, true);
2367
2368 ret = link_free_space(ctl, info);
2369 if (ret)
2370 kmem_cache_free(btrfs_free_space_cachep, info);
2371 out:
2372 spin_unlock(&ctl->tree_lock);
2373
2374 if (ret) {
2375 btrfs_crit(fs_info, "unable to add free space :%d", ret);
2376 ASSERT(ret != -EEXIST);
2377 }
2378
2379 return ret;
2380 }
2381
2382 int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
2383 u64 offset, u64 bytes)
2384 {
2385 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2386 struct btrfs_free_space *info;
2387 int ret;
2388 bool re_search = false;
2389
2390 spin_lock(&ctl->tree_lock);
2391
2392 again:
2393 ret = 0;
2394 if (!bytes)
2395 goto out_lock;
2396
2397 info = tree_search_offset(ctl, offset, 0, 0);
2398 if (!info) {
2399 /*
2400 * oops didn't find an extent that matched the space we wanted
2401 * to remove, look for a bitmap instead
2402 */
2403 info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
2404 1, 0);
2405 if (!info) {
2406 /*
2407 * If we found a partial bit of our free space in a
2408 * bitmap but then couldn't find the other part this may
2409 * be a problem, so WARN about it.
2410 */
2411 WARN_ON(re_search);
2412 goto out_lock;
2413 }
2414 }
2415
2416 re_search = false;
2417 if (!info->bitmap) {
2418 unlink_free_space(ctl, info);
2419 if (offset == info->offset) {
2420 u64 to_free = min(bytes, info->bytes);
2421
2422 info->bytes -= to_free;
2423 info->offset += to_free;
2424 if (info->bytes) {
2425 ret = link_free_space(ctl, info);
2426 WARN_ON(ret);
2427 } else {
2428 kmem_cache_free(btrfs_free_space_cachep, info);
2429 }
2430
2431 offset += to_free;
2432 bytes -= to_free;
2433 goto again;
2434 } else {
2435 u64 old_end = info->bytes + info->offset;
2436
2437 info->bytes = offset - info->offset;
2438 ret = link_free_space(ctl, info);
2439 WARN_ON(ret);
2440 if (ret)
2441 goto out_lock;
2442
2443 /* Not enough bytes in this entry to satisfy us */
2444 if (old_end < offset + bytes) {
2445 bytes -= old_end - offset;
2446 offset = old_end;
2447 goto again;
2448 } else if (old_end == offset + bytes) {
2449 /* all done */
2450 goto out_lock;
2451 }
2452 spin_unlock(&ctl->tree_lock);
2453
2454 ret = btrfs_add_free_space(block_group, offset + bytes,
2455 old_end - (offset + bytes));
2456 WARN_ON(ret);
2457 goto out;
2458 }
2459 }
2460
2461 ret = remove_from_bitmap(ctl, info, &offset, &bytes);
2462 if (ret == -EAGAIN) {
2463 re_search = true;
2464 goto again;
2465 }
2466 out_lock:
2467 spin_unlock(&ctl->tree_lock);
2468 out:
2469 return ret;
2470 }
2471
2472 void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
2473 u64 bytes)
2474 {
2475 struct btrfs_fs_info *fs_info = block_group->fs_info;
2476 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2477 struct btrfs_free_space *info;
2478 struct rb_node *n;
2479 int count = 0;
2480
2481 spin_lock(&ctl->tree_lock);
2482 for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
2483 info = rb_entry(n, struct btrfs_free_space, offset_index);
2484 if (info->bytes >= bytes && !block_group->ro)
2485 count++;
2486 btrfs_crit(fs_info, "entry offset %llu, bytes %llu, bitmap %s",
2487 info->offset, info->bytes,
2488 (info->bitmap) ? "yes" : "no");
2489 }
2490 spin_unlock(&ctl->tree_lock);
2491 btrfs_info(fs_info, "block group has cluster?: %s",
2492 list_empty(&block_group->cluster_list) ? "no" : "yes");
2493 btrfs_info(fs_info,
2494 "%d blocks of free space at or bigger than bytes is", count);
2495 }
2496
2497 void btrfs_init_free_space_ctl(struct btrfs_block_group_cache *block_group)
2498 {
2499 struct btrfs_fs_info *fs_info = block_group->fs_info;
2500 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2501
2502 spin_lock_init(&ctl->tree_lock);
2503 ctl->unit = fs_info->sectorsize;
2504 ctl->start = block_group->key.objectid;
2505 ctl->private = block_group;
2506 ctl->op = &free_space_op;
2507 INIT_LIST_HEAD(&ctl->trimming_ranges);
2508 mutex_init(&ctl->cache_writeout_mutex);
2509
2510 /*
2511 * we only want to have 32k of ram per block group for keeping
2512 * track of free space, and if we pass 1/2 of that we want to
2513 * start converting things over to using bitmaps
2514 */
2515 ctl->extents_thresh = (SZ_32K / 2) / sizeof(struct btrfs_free_space);
2516 }
2517
2518 /*
2519 * for a given cluster, put all of its extents back into the free
2520 * space cache. If the block group passed doesn't match the block group
2521 * pointed to by the cluster, someone else raced in and freed the
2522 * cluster already. In that case, we just return without changing anything
2523 */
2524 static int
2525 __btrfs_return_cluster_to_free_space(
2526 struct btrfs_block_group_cache *block_group,
2527 struct btrfs_free_cluster *cluster)
2528 {
2529 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2530 struct btrfs_free_space *entry;
2531 struct rb_node *node;
2532
2533 spin_lock(&cluster->lock);
2534 if (cluster->block_group != block_group)
2535 goto out;
2536
2537 cluster->block_group = NULL;
2538 cluster->window_start = 0;
2539 list_del_init(&cluster->block_group_list);
2540
2541 node = rb_first(&cluster->root);
2542 while (node) {
2543 bool bitmap;
2544
2545 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2546 node = rb_next(&entry->offset_index);
2547 rb_erase(&entry->offset_index, &cluster->root);
2548 RB_CLEAR_NODE(&entry->offset_index);
2549
2550 bitmap = (entry->bitmap != NULL);
2551 if (!bitmap) {
2552 try_merge_free_space(ctl, entry, false);
2553 steal_from_bitmap(ctl, entry, false);
2554 }
2555 tree_insert_offset(&ctl->free_space_offset,
2556 entry->offset, &entry->offset_index, bitmap);
2557 }
2558 cluster->root = RB_ROOT;
2559
2560 out:
2561 spin_unlock(&cluster->lock);
2562 btrfs_put_block_group(block_group);
2563 return 0;
2564 }
2565
2566 static void __btrfs_remove_free_space_cache_locked(
2567 struct btrfs_free_space_ctl *ctl)
2568 {
2569 struct btrfs_free_space *info;
2570 struct rb_node *node;
2571
2572 while ((node = rb_last(&ctl->free_space_offset)) != NULL) {
2573 info = rb_entry(node, struct btrfs_free_space, offset_index);
2574 if (!info->bitmap) {
2575 unlink_free_space(ctl, info);
2576 kmem_cache_free(btrfs_free_space_cachep, info);
2577 } else {
2578 free_bitmap(ctl, info);
2579 }
2580
2581 cond_resched_lock(&ctl->tree_lock);
2582 }
2583 }
2584
2585 void __btrfs_remove_free_space_cache(struct btrfs_free_space_ctl *ctl)
2586 {
2587 spin_lock(&ctl->tree_lock);
2588 __btrfs_remove_free_space_cache_locked(ctl);
2589 spin_unlock(&ctl->tree_lock);
2590 }
2591
2592 void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
2593 {
2594 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2595 struct btrfs_free_cluster *cluster;
2596 struct list_head *head;
2597
2598 spin_lock(&ctl->tree_lock);
2599 while ((head = block_group->cluster_list.next) !=
2600 &block_group->cluster_list) {
2601 cluster = list_entry(head, struct btrfs_free_cluster,
2602 block_group_list);
2603
2604 WARN_ON(cluster->block_group != block_group);
2605 __btrfs_return_cluster_to_free_space(block_group, cluster);
2606
2607 cond_resched_lock(&ctl->tree_lock);
2608 }
2609 __btrfs_remove_free_space_cache_locked(ctl);
2610 spin_unlock(&ctl->tree_lock);
2611
2612 }
2613
2614 u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
2615 u64 offset, u64 bytes, u64 empty_size,
2616 u64 *max_extent_size)
2617 {
2618 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2619 struct btrfs_free_space *entry = NULL;
2620 u64 bytes_search = bytes + empty_size;
2621 u64 ret = 0;
2622 u64 align_gap = 0;
2623 u64 align_gap_len = 0;
2624
2625 spin_lock(&ctl->tree_lock);
2626 entry = find_free_space(ctl, &offset, &bytes_search,
2627 block_group->full_stripe_len, max_extent_size);
2628 if (!entry)
2629 goto out;
2630
2631 ret = offset;
2632 if (entry->bitmap) {
2633 bitmap_clear_bits(ctl, entry, offset, bytes);
2634 if (!entry->bytes)
2635 free_bitmap(ctl, entry);
2636 } else {
2637 unlink_free_space(ctl, entry);
2638 align_gap_len = offset - entry->offset;
2639 align_gap = entry->offset;
2640
2641 entry->offset = offset + bytes;
2642 WARN_ON(entry->bytes < bytes + align_gap_len);
2643
2644 entry->bytes -= bytes + align_gap_len;
2645 if (!entry->bytes)
2646 kmem_cache_free(btrfs_free_space_cachep, entry);
2647 else
2648 link_free_space(ctl, entry);
2649 }
2650 out:
2651 spin_unlock(&ctl->tree_lock);
2652
2653 if (align_gap_len)
2654 __btrfs_add_free_space(block_group->fs_info, ctl,
2655 align_gap, align_gap_len);
2656 return ret;
2657 }
2658
2659 /*
2660 * given a cluster, put all of its extents back into the free space
2661 * cache. If a block group is passed, this function will only free
2662 * a cluster that belongs to the passed block group.
2663 *
2664 * Otherwise, it'll get a reference on the block group pointed to by the
2665 * cluster and remove the cluster from it.
2666 */
2667 int btrfs_return_cluster_to_free_space(
2668 struct btrfs_block_group_cache *block_group,
2669 struct btrfs_free_cluster *cluster)
2670 {
2671 struct btrfs_free_space_ctl *ctl;
2672 int ret;
2673
2674 /* first, get a safe pointer to the block group */
2675 spin_lock(&cluster->lock);
2676 if (!block_group) {
2677 block_group = cluster->block_group;
2678 if (!block_group) {
2679 spin_unlock(&cluster->lock);
2680 return 0;
2681 }
2682 } else if (cluster->block_group != block_group) {
2683 /* someone else has already freed it don't redo their work */
2684 spin_unlock(&cluster->lock);
2685 return 0;
2686 }
2687 atomic_inc(&block_group->count);
2688 spin_unlock(&cluster->lock);
2689
2690 ctl = block_group->free_space_ctl;
2691
2692 /* now return any extents the cluster had on it */
2693 spin_lock(&ctl->tree_lock);
2694 ret = __btrfs_return_cluster_to_free_space(block_group, cluster);
2695 spin_unlock(&ctl->tree_lock);
2696
2697 /* finally drop our ref */
2698 btrfs_put_block_group(block_group);
2699 return ret;
2700 }
2701
2702 static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
2703 struct btrfs_free_cluster *cluster,
2704 struct btrfs_free_space *entry,
2705 u64 bytes, u64 min_start,
2706 u64 *max_extent_size)
2707 {
2708 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2709 int err;
2710 u64 search_start = cluster->window_start;
2711 u64 search_bytes = bytes;
2712 u64 ret = 0;
2713
2714 search_start = min_start;
2715 search_bytes = bytes;
2716
2717 err = search_bitmap(ctl, entry, &search_start, &search_bytes, true);
2718 if (err) {
2719 *max_extent_size = max(get_max_extent_size(entry),
2720 *max_extent_size);
2721 return 0;
2722 }
2723
2724 ret = search_start;
2725 __bitmap_clear_bits(ctl, entry, ret, bytes);
2726
2727 return ret;
2728 }
2729
2730 /*
2731 * given a cluster, try to allocate 'bytes' from it, returns 0
2732 * if it couldn't find anything suitably large, or a logical disk offset
2733 * if things worked out
2734 */
2735 u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
2736 struct btrfs_free_cluster *cluster, u64 bytes,
2737 u64 min_start, u64 *max_extent_size)
2738 {
2739 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2740 struct btrfs_free_space *entry = NULL;
2741 struct rb_node *node;
2742 u64 ret = 0;
2743
2744 spin_lock(&cluster->lock);
2745 if (bytes > cluster->max_size)
2746 goto out;
2747
2748 if (cluster->block_group != block_group)
2749 goto out;
2750
2751 node = rb_first(&cluster->root);
2752 if (!node)
2753 goto out;
2754
2755 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2756 while (1) {
2757 if (entry->bytes < bytes)
2758 *max_extent_size = max(get_max_extent_size(entry),
2759 *max_extent_size);
2760
2761 if (entry->bytes < bytes ||
2762 (!entry->bitmap && entry->offset < min_start)) {
2763 node = rb_next(&entry->offset_index);
2764 if (!node)
2765 break;
2766 entry = rb_entry(node, struct btrfs_free_space,
2767 offset_index);
2768 continue;
2769 }
2770
2771 if (entry->bitmap) {
2772 ret = btrfs_alloc_from_bitmap(block_group,
2773 cluster, entry, bytes,
2774 cluster->window_start,
2775 max_extent_size);
2776 if (ret == 0) {
2777 node = rb_next(&entry->offset_index);
2778 if (!node)
2779 break;
2780 entry = rb_entry(node, struct btrfs_free_space,
2781 offset_index);
2782 continue;
2783 }
2784 cluster->window_start += bytes;
2785 } else {
2786 ret = entry->offset;
2787
2788 entry->offset += bytes;
2789 entry->bytes -= bytes;
2790 }
2791
2792 if (entry->bytes == 0)
2793 rb_erase(&entry->offset_index, &cluster->root);
2794 break;
2795 }
2796 out:
2797 spin_unlock(&cluster->lock);
2798
2799 if (!ret)
2800 return 0;
2801
2802 spin_lock(&ctl->tree_lock);
2803
2804 ctl->free_space -= bytes;
2805 if (entry->bytes == 0) {
2806 ctl->free_extents--;
2807 if (entry->bitmap) {
2808 kfree(entry->bitmap);
2809 ctl->total_bitmaps--;
2810 ctl->op->recalc_thresholds(ctl);
2811 }
2812 kmem_cache_free(btrfs_free_space_cachep, entry);
2813 }
2814
2815 spin_unlock(&ctl->tree_lock);
2816
2817 return ret;
2818 }
2819
2820 static int btrfs_bitmap_cluster(struct btrfs_block_group_cache *block_group,
2821 struct btrfs_free_space *entry,
2822 struct btrfs_free_cluster *cluster,
2823 u64 offset, u64 bytes,
2824 u64 cont1_bytes, u64 min_bytes)
2825 {
2826 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2827 unsigned long next_zero;
2828 unsigned long i;
2829 unsigned long want_bits;
2830 unsigned long min_bits;
2831 unsigned long found_bits;
2832 unsigned long max_bits = 0;
2833 unsigned long start = 0;
2834 unsigned long total_found = 0;
2835 int ret;
2836
2837 i = offset_to_bit(entry->offset, ctl->unit,
2838 max_t(u64, offset, entry->offset));
2839 want_bits = bytes_to_bits(bytes, ctl->unit);
2840 min_bits = bytes_to_bits(min_bytes, ctl->unit);
2841
2842 /*
2843 * Don't bother looking for a cluster in this bitmap if it's heavily
2844 * fragmented.
2845 */
2846 if (entry->max_extent_size &&
2847 entry->max_extent_size < cont1_bytes)
2848 return -ENOSPC;
2849 again:
2850 found_bits = 0;
2851 for_each_set_bit_from(i, entry->bitmap, BITS_PER_BITMAP) {
2852 next_zero = find_next_zero_bit(entry->bitmap,
2853 BITS_PER_BITMAP, i);
2854 if (next_zero - i >= min_bits) {
2855 found_bits = next_zero - i;
2856 if (found_bits > max_bits)
2857 max_bits = found_bits;
2858 break;
2859 }
2860 if (next_zero - i > max_bits)
2861 max_bits = next_zero - i;
2862 i = next_zero;
2863 }
2864
2865 if (!found_bits) {
2866 entry->max_extent_size = (u64)max_bits * ctl->unit;
2867 return -ENOSPC;
2868 }
2869
2870 if (!total_found) {
2871 start = i;
2872 cluster->max_size = 0;
2873 }
2874
2875 total_found += found_bits;
2876
2877 if (cluster->max_size < found_bits * ctl->unit)
2878 cluster->max_size = found_bits * ctl->unit;
2879
2880 if (total_found < want_bits || cluster->max_size < cont1_bytes) {
2881 i = next_zero + 1;
2882 goto again;
2883 }
2884
2885 cluster->window_start = start * ctl->unit + entry->offset;
2886 rb_erase(&entry->offset_index, &ctl->free_space_offset);
2887 ret = tree_insert_offset(&cluster->root, entry->offset,
2888 &entry->offset_index, 1);
2889 ASSERT(!ret); /* -EEXIST; Logic error */
2890
2891 trace_btrfs_setup_cluster(block_group, cluster,
2892 total_found * ctl->unit, 1);
2893 return 0;
2894 }
2895
2896 /*
2897 * This searches the block group for just extents to fill the cluster with.
2898 * Try to find a cluster with at least bytes total bytes, at least one
2899 * extent of cont1_bytes, and other clusters of at least min_bytes.
2900 */
2901 static noinline int
2902 setup_cluster_no_bitmap(struct btrfs_block_group_cache *block_group,
2903 struct btrfs_free_cluster *cluster,
2904 struct list_head *bitmaps, u64 offset, u64 bytes,
2905 u64 cont1_bytes, u64 min_bytes)
2906 {
2907 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2908 struct btrfs_free_space *first = NULL;
2909 struct btrfs_free_space *entry = NULL;
2910 struct btrfs_free_space *last;
2911 struct rb_node *node;
2912 u64 window_free;
2913 u64 max_extent;
2914 u64 total_size = 0;
2915
2916 entry = tree_search_offset(ctl, offset, 0, 1);
2917 if (!entry)
2918 return -ENOSPC;
2919
2920 /*
2921 * We don't want bitmaps, so just move along until we find a normal
2922 * extent entry.
2923 */
2924 while (entry->bitmap || entry->bytes < min_bytes) {
2925 if (entry->bitmap && list_empty(&entry->list))
2926 list_add_tail(&entry->list, bitmaps);
2927 node = rb_next(&entry->offset_index);
2928 if (!node)
2929 return -ENOSPC;
2930 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2931 }
2932
2933 window_free = entry->bytes;
2934 max_extent = entry->bytes;
2935 first = entry;
2936 last = entry;
2937
2938 for (node = rb_next(&entry->offset_index); node;
2939 node = rb_next(&entry->offset_index)) {
2940 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2941
2942 if (entry->bitmap) {
2943 if (list_empty(&entry->list))
2944 list_add_tail(&entry->list, bitmaps);
2945 continue;
2946 }
2947
2948 if (entry->bytes < min_bytes)
2949 continue;
2950
2951 last = entry;
2952 window_free += entry->bytes;
2953 if (entry->bytes > max_extent)
2954 max_extent = entry->bytes;
2955 }
2956
2957 if (window_free < bytes || max_extent < cont1_bytes)
2958 return -ENOSPC;
2959
2960 cluster->window_start = first->offset;
2961
2962 node = &first->offset_index;
2963
2964 /*
2965 * now we've found our entries, pull them out of the free space
2966 * cache and put them into the cluster rbtree
2967 */
2968 do {
2969 int ret;
2970
2971 entry = rb_entry(node, struct btrfs_free_space, offset_index);
2972 node = rb_next(&entry->offset_index);
2973 if (entry->bitmap || entry->bytes < min_bytes)
2974 continue;
2975
2976 rb_erase(&entry->offset_index, &ctl->free_space_offset);
2977 ret = tree_insert_offset(&cluster->root, entry->offset,
2978 &entry->offset_index, 0);
2979 total_size += entry->bytes;
2980 ASSERT(!ret); /* -EEXIST; Logic error */
2981 } while (node && entry != last);
2982
2983 cluster->max_size = max_extent;
2984 trace_btrfs_setup_cluster(block_group, cluster, total_size, 0);
2985 return 0;
2986 }
2987
2988 /*
2989 * This specifically looks for bitmaps that may work in the cluster, we assume
2990 * that we have already failed to find extents that will work.
2991 */
2992 static noinline int
2993 setup_cluster_bitmap(struct btrfs_block_group_cache *block_group,
2994 struct btrfs_free_cluster *cluster,
2995 struct list_head *bitmaps, u64 offset, u64 bytes,
2996 u64 cont1_bytes, u64 min_bytes)
2997 {
2998 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2999 struct btrfs_free_space *entry = NULL;
3000 int ret = -ENOSPC;
3001 u64 bitmap_offset = offset_to_bitmap(ctl, offset);
3002
3003 if (ctl->total_bitmaps == 0)
3004 return -ENOSPC;
3005
3006 /*
3007 * The bitmap that covers offset won't be in the list unless offset
3008 * is just its start offset.
3009 */
3010 if (!list_empty(bitmaps))
3011 entry = list_first_entry(bitmaps, struct btrfs_free_space, list);
3012
3013 if (!entry || entry->offset != bitmap_offset) {
3014 entry = tree_search_offset(ctl, bitmap_offset, 1, 0);
3015 if (entry && list_empty(&entry->list))
3016 list_add(&entry->list, bitmaps);
3017 }
3018
3019 list_for_each_entry(entry, bitmaps, list) {
3020 if (entry->bytes < bytes)
3021 continue;
3022 ret = btrfs_bitmap_cluster(block_group, entry, cluster, offset,
3023 bytes, cont1_bytes, min_bytes);
3024 if (!ret)
3025 return 0;
3026 }
3027
3028 /*
3029 * The bitmaps list has all the bitmaps that record free space
3030 * starting after offset, so no more search is required.
3031 */
3032 return -ENOSPC;
3033 }
3034
3035 /*
3036 * here we try to find a cluster of blocks in a block group. The goal
3037 * is to find at least bytes+empty_size.
3038 * We might not find them all in one contiguous area.
3039 *
3040 * returns zero and sets up cluster if things worked out, otherwise
3041 * it returns -enospc
3042 */
3043 int btrfs_find_space_cluster(struct btrfs_fs_info *fs_info,
3044 struct btrfs_block_group_cache *block_group,
3045 struct btrfs_free_cluster *cluster,
3046 u64 offset, u64 bytes, u64 empty_size)
3047 {
3048 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3049 struct btrfs_free_space *entry, *tmp;
3050 LIST_HEAD(bitmaps);
3051 u64 min_bytes;
3052 u64 cont1_bytes;
3053 int ret;
3054
3055 /*
3056 * Choose the minimum extent size we'll require for this
3057 * cluster. For SSD_SPREAD, don't allow any fragmentation.
3058 * For metadata, allow allocates with smaller extents. For
3059 * data, keep it dense.
3060 */
3061 if (btrfs_test_opt(fs_info, SSD_SPREAD)) {
3062 cont1_bytes = min_bytes = bytes + empty_size;
3063 } else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
3064 cont1_bytes = bytes;
3065 min_bytes = fs_info->sectorsize;
3066 } else {
3067 cont1_bytes = max(bytes, (bytes + empty_size) >> 2);
3068 min_bytes = fs_info->sectorsize;
3069 }
3070
3071 spin_lock(&ctl->tree_lock);
3072
3073 /*
3074 * If we know we don't have enough space to make a cluster don't even
3075 * bother doing all the work to try and find one.
3076 */
3077 if (ctl->free_space < bytes) {
3078 spin_unlock(&ctl->tree_lock);
3079 return -ENOSPC;
3080 }
3081
3082 spin_lock(&cluster->lock);
3083
3084 /* someone already found a cluster, hooray */
3085 if (cluster->block_group) {
3086 ret = 0;
3087 goto out;
3088 }
3089
3090 trace_btrfs_find_cluster(block_group, offset, bytes, empty_size,
3091 min_bytes);
3092
3093 ret = setup_cluster_no_bitmap(block_group, cluster, &bitmaps, offset,
3094 bytes + empty_size,
3095 cont1_bytes, min_bytes);
3096 if (ret)
3097 ret = setup_cluster_bitmap(block_group, cluster, &bitmaps,
3098 offset, bytes + empty_size,
3099 cont1_bytes, min_bytes);
3100
3101 /* Clear our temporary list */
3102 list_for_each_entry_safe(entry, tmp, &bitmaps, list)
3103 list_del_init(&entry->list);
3104
3105 if (!ret) {
3106 atomic_inc(&block_group->count);
3107 list_add_tail(&cluster->block_group_list,
3108 &block_group->cluster_list);
3109 cluster->block_group = block_group;
3110 } else {
3111 trace_btrfs_failed_cluster_setup(block_group);
3112 }
3113 out:
3114 spin_unlock(&cluster->lock);
3115 spin_unlock(&ctl->tree_lock);
3116
3117 return ret;
3118 }
3119
3120 /*
3121 * simple code to zero out a cluster
3122 */
3123 void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster)
3124 {
3125 spin_lock_init(&cluster->lock);
3126 spin_lock_init(&cluster->refill_lock);
3127 cluster->root = RB_ROOT;
3128 cluster->max_size = 0;
3129 cluster->fragmented = false;
3130 INIT_LIST_HEAD(&cluster->block_group_list);
3131 cluster->block_group = NULL;
3132 }
3133
3134 static int do_trimming(struct btrfs_block_group_cache *block_group,
3135 u64 *total_trimmed, u64 start, u64 bytes,
3136 u64 reserved_start, u64 reserved_bytes,
3137 struct btrfs_trim_range *trim_entry)
3138 {
3139 struct btrfs_space_info *space_info = block_group->space_info;
3140 struct btrfs_fs_info *fs_info = block_group->fs_info;
3141 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3142 int ret;
3143 int update = 0;
3144 u64 trimmed = 0;
3145
3146 spin_lock(&space_info->lock);
3147 spin_lock(&block_group->lock);
3148 if (!block_group->ro) {
3149 block_group->reserved += reserved_bytes;
3150 space_info->bytes_reserved += reserved_bytes;
3151 update = 1;
3152 }
3153 spin_unlock(&block_group->lock);
3154 spin_unlock(&space_info->lock);
3155
3156 ret = btrfs_discard_extent(fs_info, start, bytes, &trimmed);
3157 if (!ret)
3158 *total_trimmed += trimmed;
3159
3160 mutex_lock(&ctl->cache_writeout_mutex);
3161 btrfs_add_free_space(block_group, reserved_start, reserved_bytes);
3162 list_del(&trim_entry->list);
3163 mutex_unlock(&ctl->cache_writeout_mutex);
3164
3165 if (update) {
3166 spin_lock(&space_info->lock);
3167 spin_lock(&block_group->lock);
3168 if (block_group->ro)
3169 space_info->bytes_readonly += reserved_bytes;
3170 block_group->reserved -= reserved_bytes;
3171 space_info->bytes_reserved -= reserved_bytes;
3172 spin_unlock(&space_info->lock);
3173 spin_unlock(&block_group->lock);
3174 }
3175
3176 return ret;
3177 }
3178
3179 static int trim_no_bitmap(struct btrfs_block_group_cache *block_group,
3180 u64 *total_trimmed, u64 start, u64 end, u64 minlen)
3181 {
3182 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3183 struct btrfs_free_space *entry;
3184 struct rb_node *node;
3185 int ret = 0;
3186 u64 extent_start;
3187 u64 extent_bytes;
3188 u64 bytes;
3189
3190 while (start < end) {
3191 struct btrfs_trim_range trim_entry;
3192
3193 mutex_lock(&ctl->cache_writeout_mutex);
3194 spin_lock(&ctl->tree_lock);
3195
3196 if (ctl->free_space < minlen) {
3197 spin_unlock(&ctl->tree_lock);
3198 mutex_unlock(&ctl->cache_writeout_mutex);
3199 break;
3200 }
3201
3202 entry = tree_search_offset(ctl, start, 0, 1);
3203 if (!entry) {
3204 spin_unlock(&ctl->tree_lock);
3205 mutex_unlock(&ctl->cache_writeout_mutex);
3206 break;
3207 }
3208
3209 /* skip bitmaps */
3210 while (entry->bitmap) {
3211 node = rb_next(&entry->offset_index);
3212 if (!node) {
3213 spin_unlock(&ctl->tree_lock);
3214 mutex_unlock(&ctl->cache_writeout_mutex);
3215 goto out;
3216 }
3217 entry = rb_entry(node, struct btrfs_free_space,
3218 offset_index);
3219 }
3220
3221 if (entry->offset >= end) {
3222 spin_unlock(&ctl->tree_lock);
3223 mutex_unlock(&ctl->cache_writeout_mutex);
3224 break;
3225 }
3226
3227 extent_start = entry->offset;
3228 extent_bytes = entry->bytes;
3229 start = max(start, extent_start);
3230 bytes = min(extent_start + extent_bytes, end) - start;
3231 if (bytes < minlen) {
3232 spin_unlock(&ctl->tree_lock);
3233 mutex_unlock(&ctl->cache_writeout_mutex);
3234 goto next;
3235 }
3236
3237 unlink_free_space(ctl, entry);
3238 kmem_cache_free(btrfs_free_space_cachep, entry);
3239
3240 spin_unlock(&ctl->tree_lock);
3241 trim_entry.start = extent_start;
3242 trim_entry.bytes = extent_bytes;
3243 list_add_tail(&trim_entry.list, &ctl->trimming_ranges);
3244 mutex_unlock(&ctl->cache_writeout_mutex);
3245
3246 ret = do_trimming(block_group, total_trimmed, start, bytes,
3247 extent_start, extent_bytes, &trim_entry);
3248 if (ret)
3249 break;
3250 next:
3251 start += bytes;
3252
3253 if (fatal_signal_pending(current)) {
3254 ret = -ERESTARTSYS;
3255 break;
3256 }
3257
3258 cond_resched();
3259 }
3260 out:
3261 return ret;
3262 }
3263
3264 static int trim_bitmaps(struct btrfs_block_group_cache *block_group,
3265 u64 *total_trimmed, u64 start, u64 end, u64 minlen)
3266 {
3267 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3268 struct btrfs_free_space *entry;
3269 int ret = 0;
3270 int ret2;
3271 u64 bytes;
3272 u64 offset = offset_to_bitmap(ctl, start);
3273
3274 while (offset < end) {
3275 bool next_bitmap = false;
3276 struct btrfs_trim_range trim_entry;
3277
3278 mutex_lock(&ctl->cache_writeout_mutex);
3279 spin_lock(&ctl->tree_lock);
3280
3281 if (ctl->free_space < minlen) {
3282 spin_unlock(&ctl->tree_lock);
3283 mutex_unlock(&ctl->cache_writeout_mutex);
3284 break;
3285 }
3286
3287 entry = tree_search_offset(ctl, offset, 1, 0);
3288 if (!entry) {
3289 spin_unlock(&ctl->tree_lock);
3290 mutex_unlock(&ctl->cache_writeout_mutex);
3291 next_bitmap = true;
3292 goto next;
3293 }
3294
3295 bytes = minlen;
3296 ret2 = search_bitmap(ctl, entry, &start, &bytes, false);
3297 if (ret2 || start >= end) {
3298 spin_unlock(&ctl->tree_lock);
3299 mutex_unlock(&ctl->cache_writeout_mutex);
3300 next_bitmap = true;
3301 goto next;
3302 }
3303
3304 bytes = min(bytes, end - start);
3305 if (bytes < minlen) {
3306 spin_unlock(&ctl->tree_lock);
3307 mutex_unlock(&ctl->cache_writeout_mutex);
3308 goto next;
3309 }
3310
3311 bitmap_clear_bits(ctl, entry, start, bytes);
3312 if (entry->bytes == 0)
3313 free_bitmap(ctl, entry);
3314
3315 spin_unlock(&ctl->tree_lock);
3316 trim_entry.start = start;
3317 trim_entry.bytes = bytes;
3318 list_add_tail(&trim_entry.list, &ctl->trimming_ranges);
3319 mutex_unlock(&ctl->cache_writeout_mutex);
3320
3321 ret = do_trimming(block_group, total_trimmed, start, bytes,
3322 start, bytes, &trim_entry);
3323 if (ret)
3324 break;
3325 next:
3326 if (next_bitmap) {
3327 offset += BITS_PER_BITMAP * ctl->unit;
3328 } else {
3329 start += bytes;
3330 if (start >= offset + BITS_PER_BITMAP * ctl->unit)
3331 offset += BITS_PER_BITMAP * ctl->unit;
3332 }
3333
3334 if (fatal_signal_pending(current)) {
3335 ret = -ERESTARTSYS;
3336 break;
3337 }
3338
3339 cond_resched();
3340 }
3341
3342 return ret;
3343 }
3344
3345 void btrfs_get_block_group_trimming(struct btrfs_block_group_cache *cache)
3346 {
3347 atomic_inc(&cache->trimming);
3348 }
3349
3350 void btrfs_put_block_group_trimming(struct btrfs_block_group_cache *block_group)
3351 {
3352 struct btrfs_fs_info *fs_info = block_group->fs_info;
3353 struct extent_map_tree *em_tree;
3354 struct extent_map *em;
3355 bool cleanup;
3356
3357 spin_lock(&block_group->lock);
3358 cleanup = (atomic_dec_and_test(&block_group->trimming) &&
3359 block_group->removed);
3360 spin_unlock(&block_group->lock);
3361
3362 if (cleanup) {
3363 mutex_lock(&fs_info->chunk_mutex);
3364 em_tree = &fs_info->mapping_tree.map_tree;
3365 write_lock(&em_tree->lock);
3366 em = lookup_extent_mapping(em_tree, block_group->key.objectid,
3367 1);
3368 BUG_ON(!em); /* logic error, can't happen */
3369 /*
3370 * remove_extent_mapping() will delete us from the pinned_chunks
3371 * list, which is protected by the chunk mutex.
3372 */
3373 remove_extent_mapping(em_tree, em);
3374 write_unlock(&em_tree->lock);
3375 mutex_unlock(&fs_info->chunk_mutex);
3376
3377 /* once for us and once for the tree */
3378 free_extent_map(em);
3379 free_extent_map(em);
3380
3381 /*
3382 * We've left one free space entry and other tasks trimming
3383 * this block group have left 1 entry each one. Free them.
3384 */
3385 __btrfs_remove_free_space_cache(block_group->free_space_ctl);
3386 }
3387 }
3388
3389 int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group,
3390 u64 *trimmed, u64 start, u64 end, u64 minlen)
3391 {
3392 int ret;
3393
3394 *trimmed = 0;
3395
3396 spin_lock(&block_group->lock);
3397 if (block_group->removed) {
3398 spin_unlock(&block_group->lock);
3399 return 0;
3400 }
3401 btrfs_get_block_group_trimming(block_group);
3402 spin_unlock(&block_group->lock);
3403
3404 ret = trim_no_bitmap(block_group, trimmed, start, end, minlen);
3405 if (ret)
3406 goto out;
3407
3408 ret = trim_bitmaps(block_group, trimmed, start, end, minlen);
3409 out:
3410 btrfs_put_block_group_trimming(block_group);
3411 return ret;
3412 }
3413
3414 /*
3415 * Find the left-most item in the cache tree, and then return the
3416 * smallest inode number in the item.
3417 *
3418 * Note: the returned inode number may not be the smallest one in
3419 * the tree, if the left-most item is a bitmap.
3420 */
3421 u64 btrfs_find_ino_for_alloc(struct btrfs_root *fs_root)
3422 {
3423 struct btrfs_free_space_ctl *ctl = fs_root->free_ino_ctl;
3424 struct btrfs_free_space *entry = NULL;
3425 u64 ino = 0;
3426
3427 spin_lock(&ctl->tree_lock);
3428
3429 if (RB_EMPTY_ROOT(&ctl->free_space_offset))
3430 goto out;
3431
3432 entry = rb_entry(rb_first(&ctl->free_space_offset),
3433 struct btrfs_free_space, offset_index);
3434
3435 if (!entry->bitmap) {
3436 ino = entry->offset;
3437
3438 unlink_free_space(ctl, entry);
3439 entry->offset++;
3440 entry->bytes--;
3441 if (!entry->bytes)
3442 kmem_cache_free(btrfs_free_space_cachep, entry);
3443 else
3444 link_free_space(ctl, entry);
3445 } else {
3446 u64 offset = 0;
3447 u64 count = 1;
3448 int ret;
3449
3450 ret = search_bitmap(ctl, entry, &offset, &count, true);
3451 /* Logic error; Should be empty if it can't find anything */
3452 ASSERT(!ret);
3453
3454 ino = offset;
3455 bitmap_clear_bits(ctl, entry, offset, 1);
3456 if (entry->bytes == 0)
3457 free_bitmap(ctl, entry);
3458 }
3459 out:
3460 spin_unlock(&ctl->tree_lock);
3461
3462 return ino;
3463 }
3464
3465 struct inode *lookup_free_ino_inode(struct btrfs_root *root,
3466 struct btrfs_path *path)
3467 {
3468 struct inode *inode = NULL;
3469
3470 spin_lock(&root->ino_cache_lock);
3471 if (root->ino_cache_inode)
3472 inode = igrab(root->ino_cache_inode);
3473 spin_unlock(&root->ino_cache_lock);
3474 if (inode)
3475 return inode;
3476
3477 inode = __lookup_free_space_inode(root, path, 0);
3478 if (IS_ERR(inode))
3479 return inode;
3480
3481 spin_lock(&root->ino_cache_lock);
3482 if (!btrfs_fs_closing(root->fs_info))
3483 root->ino_cache_inode = igrab(inode);
3484 spin_unlock(&root->ino_cache_lock);
3485
3486 return inode;
3487 }
3488
3489 int create_free_ino_inode(struct btrfs_root *root,
3490 struct btrfs_trans_handle *trans,
3491 struct btrfs_path *path)
3492 {
3493 return __create_free_space_inode(root, trans, path,
3494 BTRFS_FREE_INO_OBJECTID, 0);
3495 }
3496
3497 int load_free_ino_cache(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
3498 {
3499 struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
3500 struct btrfs_path *path;
3501 struct inode *inode;
3502 int ret = 0;
3503 u64 root_gen = btrfs_root_generation(&root->root_item);
3504
3505 if (!btrfs_test_opt(fs_info, INODE_MAP_CACHE))
3506 return 0;
3507
3508 /*
3509 * If we're unmounting then just return, since this does a search on the
3510 * normal root and not the commit root and we could deadlock.
3511 */
3512 if (btrfs_fs_closing(fs_info))
3513 return 0;
3514
3515 path = btrfs_alloc_path();
3516 if (!path)
3517 return 0;
3518
3519 inode = lookup_free_ino_inode(root, path);
3520 if (IS_ERR(inode))
3521 goto out;
3522
3523 if (root_gen != BTRFS_I(inode)->generation)
3524 goto out_put;
3525
3526 ret = __load_free_space_cache(root, inode, ctl, path, 0);
3527
3528 if (ret < 0)
3529 btrfs_err(fs_info,
3530 "failed to load free ino cache for root %llu",
3531 root->root_key.objectid);
3532 out_put:
3533 iput(inode);
3534 out:
3535 btrfs_free_path(path);
3536 return ret;
3537 }
3538
3539 int btrfs_write_out_ino_cache(struct btrfs_root *root,
3540 struct btrfs_trans_handle *trans,
3541 struct btrfs_path *path,
3542 struct inode *inode)
3543 {
3544 struct btrfs_fs_info *fs_info = root->fs_info;
3545 struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
3546 int ret;
3547 struct btrfs_io_ctl io_ctl;
3548 bool release_metadata = true;
3549
3550 if (!btrfs_test_opt(fs_info, INODE_MAP_CACHE))
3551 return 0;
3552
3553 memset(&io_ctl, 0, sizeof(io_ctl));
3554 ret = __btrfs_write_out_cache(root, inode, ctl, NULL, &io_ctl, trans);
3555 if (!ret) {
3556 /*
3557 * At this point writepages() didn't error out, so our metadata
3558 * reservation is released when the writeback finishes, at
3559 * inode.c:btrfs_finish_ordered_io(), regardless of it finishing
3560 * with or without an error.
3561 */
3562 release_metadata = false;
3563 ret = btrfs_wait_cache_io_root(root, trans, &io_ctl, path);
3564 }
3565
3566 if (ret) {
3567 if (release_metadata)
3568 btrfs_delalloc_release_metadata(BTRFS_I(inode),
3569 inode->i_size, true);
3570 #ifdef DEBUG
3571 btrfs_err(fs_info,
3572 "failed to write free ino cache for root %llu",
3573 root->root_key.objectid);
3574 #endif
3575 }
3576
3577 return ret;
3578 }
3579
3580 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
3581 /*
3582 * Use this if you need to make a bitmap or extent entry specifically, it
3583 * doesn't do any of the merging that add_free_space does, this acts a lot like
3584 * how the free space cache loading stuff works, so you can get really weird
3585 * configurations.
3586 */
3587 int test_add_free_space_entry(struct btrfs_block_group_cache *cache,
3588 u64 offset, u64 bytes, bool bitmap)
3589 {
3590 struct btrfs_free_space_ctl *ctl = cache->free_space_ctl;
3591 struct btrfs_free_space *info = NULL, *bitmap_info;
3592 void *map = NULL;
3593 u64 bytes_added;
3594 int ret;
3595
3596 again:
3597 if (!info) {
3598 info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
3599 if (!info)
3600 return -ENOMEM;
3601 }
3602
3603 if (!bitmap) {
3604 spin_lock(&ctl->tree_lock);
3605 info->offset = offset;
3606 info->bytes = bytes;
3607 info->max_extent_size = 0;
3608 ret = link_free_space(ctl, info);
3609 spin_unlock(&ctl->tree_lock);
3610 if (ret)
3611 kmem_cache_free(btrfs_free_space_cachep, info);
3612 return ret;
3613 }
3614
3615 if (!map) {
3616 map = kzalloc(PAGE_SIZE, GFP_NOFS);
3617 if (!map) {
3618 kmem_cache_free(btrfs_free_space_cachep, info);
3619 return -ENOMEM;
3620 }
3621 }
3622
3623 spin_lock(&ctl->tree_lock);
3624 bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
3625 1, 0);
3626 if (!bitmap_info) {
3627 info->bitmap = map;
3628 map = NULL;
3629 add_new_bitmap(ctl, info, offset);
3630 bitmap_info = info;
3631 info = NULL;
3632 }
3633
3634 bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes);
3635
3636 bytes -= bytes_added;
3637 offset += bytes_added;
3638 spin_unlock(&ctl->tree_lock);
3639
3640 if (bytes)
3641 goto again;
3642
3643 if (info)
3644 kmem_cache_free(btrfs_free_space_cachep, info);
3645 kfree(map);
3646 return 0;
3647 }
3648
3649 /*
3650 * Checks to see if the given range is in the free space cache. This is really
3651 * just used to check the absence of space, so if there is free space in the
3652 * range at all we will return 1.
3653 */
3654 int test_check_exists(struct btrfs_block_group_cache *cache,
3655 u64 offset, u64 bytes)
3656 {
3657 struct btrfs_free_space_ctl *ctl = cache->free_space_ctl;
3658 struct btrfs_free_space *info;
3659 int ret = 0;
3660
3661 spin_lock(&ctl->tree_lock);
3662 info = tree_search_offset(ctl, offset, 0, 0);
3663 if (!info) {
3664 info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
3665 1, 0);
3666 if (!info)
3667 goto out;
3668 }
3669
3670 have_info:
3671 if (info->bitmap) {
3672 u64 bit_off, bit_bytes;
3673 struct rb_node *n;
3674 struct btrfs_free_space *tmp;
3675
3676 bit_off = offset;
3677 bit_bytes = ctl->unit;
3678 ret = search_bitmap(ctl, info, &bit_off, &bit_bytes, false);
3679 if (!ret) {
3680 if (bit_off == offset) {
3681 ret = 1;
3682 goto out;
3683 } else if (bit_off > offset &&
3684 offset + bytes > bit_off) {
3685 ret = 1;
3686 goto out;
3687 }
3688 }
3689
3690 n = rb_prev(&info->offset_index);
3691 while (n) {
3692 tmp = rb_entry(n, struct btrfs_free_space,
3693 offset_index);
3694 if (tmp->offset + tmp->bytes < offset)
3695 break;
3696 if (offset + bytes < tmp->offset) {
3697 n = rb_prev(&tmp->offset_index);
3698 continue;
3699 }
3700 info = tmp;
3701 goto have_info;
3702 }
3703
3704 n = rb_next(&info->offset_index);
3705 while (n) {
3706 tmp = rb_entry(n, struct btrfs_free_space,
3707 offset_index);
3708 if (offset + bytes < tmp->offset)
3709 break;
3710 if (tmp->offset + tmp->bytes < offset) {
3711 n = rb_next(&tmp->offset_index);
3712 continue;
3713 }
3714 info = tmp;
3715 goto have_info;
3716 }
3717
3718 ret = 0;
3719 goto out;
3720 }
3721
3722 if (info->offset == offset) {
3723 ret = 1;
3724 goto out;
3725 }
3726
3727 if (offset > info->offset && offset < info->offset + info->bytes)
3728 ret = 1;
3729 out:
3730 spin_unlock(&ctl->tree_lock);
3731 return ret;
3732 }
3733 #endif /* CONFIG_BTRFS_FS_RUN_SANITY_TESTS */
Linux with added FPC-III support