xref: /linux/fs/btrfs/free-space-cache.c (revision 6eb2fb3170549737207974c2c6ad34bcc2f3025e)
1 /*
2  * Copyright (C) 2008 Red Hat.  All rights reserved.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public
6  * License v2 as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful,
9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public
14  * License along with this program; if not, write to the
15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16  * Boston, MA 021110-1307, USA.
17  */
18 
19 #include <linux/pagemap.h>
20 #include <linux/sched.h>
21 #include <linux/slab.h>
22 #include <linux/math64.h>
23 #include <linux/ratelimit.h>
24 #include "ctree.h"
25 #include "free-space-cache.h"
26 #include "transaction.h"
27 #include "disk-io.h"
28 #include "extent_io.h"
29 #include "inode-map.h"
30 
31 #define BITS_PER_BITMAP		(PAGE_CACHE_SIZE * 8)
32 #define MAX_CACHE_BYTES_PER_GIG	(32 * 1024)
33 
34 static int link_free_space(struct btrfs_free_space_ctl *ctl,
35 			   struct btrfs_free_space *info);
36 static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
37 			      struct btrfs_free_space *info);
38 
39 static struct inode *__lookup_free_space_inode(struct btrfs_root *root,
40 					       struct btrfs_path *path,
41 					       u64 offset)
42 {
43 	struct btrfs_key key;
44 	struct btrfs_key location;
45 	struct btrfs_disk_key disk_key;
46 	struct btrfs_free_space_header *header;
47 	struct extent_buffer *leaf;
48 	struct inode *inode = NULL;
49 	int ret;
50 
51 	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
52 	key.offset = offset;
53 	key.type = 0;
54 
55 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
56 	if (ret < 0)
57 		return ERR_PTR(ret);
58 	if (ret > 0) {
59 		btrfs_release_path(path);
60 		return ERR_PTR(-ENOENT);
61 	}
62 
63 	leaf = path->nodes[0];
64 	header = btrfs_item_ptr(leaf, path->slots[0],
65 				struct btrfs_free_space_header);
66 	btrfs_free_space_key(leaf, header, &disk_key);
67 	btrfs_disk_key_to_cpu(&location, &disk_key);
68 	btrfs_release_path(path);
69 
70 	inode = btrfs_iget(root->fs_info->sb, &location, root, NULL);
71 	if (!inode)
72 		return ERR_PTR(-ENOENT);
73 	if (IS_ERR(inode))
74 		return inode;
75 	if (is_bad_inode(inode)) {
76 		iput(inode);
77 		return ERR_PTR(-ENOENT);
78 	}
79 
80 	mapping_set_gfp_mask(inode->i_mapping,
81 			mapping_gfp_mask(inode->i_mapping) & ~__GFP_FS);
82 
83 	return inode;
84 }
85 
86 struct inode *lookup_free_space_inode(struct btrfs_root *root,
87 				      struct btrfs_block_group_cache
88 				      *block_group, struct btrfs_path *path)
89 {
90 	struct inode *inode = NULL;
91 	u32 flags = BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
92 
93 	spin_lock(&block_group->lock);
94 	if (block_group->inode)
95 		inode = igrab(block_group->inode);
96 	spin_unlock(&block_group->lock);
97 	if (inode)
98 		return inode;
99 
100 	inode = __lookup_free_space_inode(root, path,
101 					  block_group->key.objectid);
102 	if (IS_ERR(inode))
103 		return inode;
104 
105 	spin_lock(&block_group->lock);
106 	if (!((BTRFS_I(inode)->flags & flags) == flags)) {
107 		btrfs_info(root->fs_info,
108 			"Old style space inode found, converting.");
109 		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM |
110 			BTRFS_INODE_NODATACOW;
111 		block_group->disk_cache_state = BTRFS_DC_CLEAR;
112 	}
113 
114 	if (!block_group->iref) {
115 		block_group->inode = igrab(inode);
116 		block_group->iref = 1;
117 	}
118 	spin_unlock(&block_group->lock);
119 
120 	return inode;
121 }
122 
123 static int __create_free_space_inode(struct btrfs_root *root,
124 				     struct btrfs_trans_handle *trans,
125 				     struct btrfs_path *path,
126 				     u64 ino, u64 offset)
127 {
128 	struct btrfs_key key;
129 	struct btrfs_disk_key disk_key;
130 	struct btrfs_free_space_header *header;
131 	struct btrfs_inode_item *inode_item;
132 	struct extent_buffer *leaf;
133 	u64 flags = BTRFS_INODE_NOCOMPRESS | BTRFS_INODE_PREALLOC;
134 	int ret;
135 
136 	ret = btrfs_insert_empty_inode(trans, root, path, ino);
137 	if (ret)
138 		return ret;
139 
140 	/* We inline crc's for the free disk space cache */
141 	if (ino != BTRFS_FREE_INO_OBJECTID)
142 		flags |= BTRFS_INODE_NODATASUM | BTRFS_INODE_NODATACOW;
143 
144 	leaf = path->nodes[0];
145 	inode_item = btrfs_item_ptr(leaf, path->slots[0],
146 				    struct btrfs_inode_item);
147 	btrfs_item_key(leaf, &disk_key, path->slots[0]);
148 	memset_extent_buffer(leaf, 0, (unsigned long)inode_item,
149 			     sizeof(*inode_item));
150 	btrfs_set_inode_generation(leaf, inode_item, trans->transid);
151 	btrfs_set_inode_size(leaf, inode_item, 0);
152 	btrfs_set_inode_nbytes(leaf, inode_item, 0);
153 	btrfs_set_inode_uid(leaf, inode_item, 0);
154 	btrfs_set_inode_gid(leaf, inode_item, 0);
155 	btrfs_set_inode_mode(leaf, inode_item, S_IFREG | 0600);
156 	btrfs_set_inode_flags(leaf, inode_item, flags);
157 	btrfs_set_inode_nlink(leaf, inode_item, 1);
158 	btrfs_set_inode_transid(leaf, inode_item, trans->transid);
159 	btrfs_set_inode_block_group(leaf, inode_item, offset);
160 	btrfs_mark_buffer_dirty(leaf);
161 	btrfs_release_path(path);
162 
163 	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
164 	key.offset = offset;
165 	key.type = 0;
166 
167 	ret = btrfs_insert_empty_item(trans, root, path, &key,
168 				      sizeof(struct btrfs_free_space_header));
169 	if (ret < 0) {
170 		btrfs_release_path(path);
171 		return ret;
172 	}
173 	leaf = path->nodes[0];
174 	header = btrfs_item_ptr(leaf, path->slots[0],
175 				struct btrfs_free_space_header);
176 	memset_extent_buffer(leaf, 0, (unsigned long)header, sizeof(*header));
177 	btrfs_set_free_space_key(leaf, header, &disk_key);
178 	btrfs_mark_buffer_dirty(leaf);
179 	btrfs_release_path(path);
180 
181 	return 0;
182 }
183 
184 int create_free_space_inode(struct btrfs_root *root,
185 			    struct btrfs_trans_handle *trans,
186 			    struct btrfs_block_group_cache *block_group,
187 			    struct btrfs_path *path)
188 {
189 	int ret;
190 	u64 ino;
191 
192 	ret = btrfs_find_free_objectid(root, &ino);
193 	if (ret < 0)
194 		return ret;
195 
196 	return __create_free_space_inode(root, trans, path, ino,
197 					 block_group->key.objectid);
198 }
199 
200 int btrfs_truncate_free_space_cache(struct btrfs_root *root,
201 				    struct btrfs_trans_handle *trans,
202 				    struct btrfs_path *path,
203 				    struct inode *inode)
204 {
205 	struct btrfs_block_rsv *rsv;
206 	u64 needed_bytes;
207 	loff_t oldsize;
208 	int ret = 0;
209 
210 	rsv = trans->block_rsv;
211 	trans->block_rsv = &root->fs_info->global_block_rsv;
212 
213 	/* 1 for slack space, 1 for updating the inode */
214 	needed_bytes = btrfs_calc_trunc_metadata_size(root, 1) +
215 		btrfs_calc_trans_metadata_size(root, 1);
216 
217 	spin_lock(&trans->block_rsv->lock);
218 	if (trans->block_rsv->reserved < needed_bytes) {
219 		spin_unlock(&trans->block_rsv->lock);
220 		trans->block_rsv = rsv;
221 		return -ENOSPC;
222 	}
223 	spin_unlock(&trans->block_rsv->lock);
224 
225 	oldsize = i_size_read(inode);
226 	btrfs_i_size_write(inode, 0);
227 	truncate_pagecache(inode, oldsize, 0);
228 
229 	/*
230 	 * We don't need an orphan item because truncating the free space cache
231 	 * will never be split across transactions.
232 	 */
233 	ret = btrfs_truncate_inode_items(trans, root, inode,
234 					 0, BTRFS_EXTENT_DATA_KEY);
235 
236 	if (ret) {
237 		trans->block_rsv = rsv;
238 		btrfs_abort_transaction(trans, root, ret);
239 		return ret;
240 	}
241 
242 	ret = btrfs_update_inode(trans, root, inode);
243 	if (ret)
244 		btrfs_abort_transaction(trans, root, ret);
245 	trans->block_rsv = rsv;
246 
247 	return ret;
248 }
249 
250 static int readahead_cache(struct inode *inode)
251 {
252 	struct file_ra_state *ra;
253 	unsigned long last_index;
254 
255 	ra = kzalloc(sizeof(*ra), GFP_NOFS);
256 	if (!ra)
257 		return -ENOMEM;
258 
259 	file_ra_state_init(ra, inode->i_mapping);
260 	last_index = (i_size_read(inode) - 1) >> PAGE_CACHE_SHIFT;
261 
262 	page_cache_sync_readahead(inode->i_mapping, ra, NULL, 0, last_index);
263 
264 	kfree(ra);
265 
266 	return 0;
267 }
268 
269 struct io_ctl {
270 	void *cur, *orig;
271 	struct page *page;
272 	struct page **pages;
273 	struct btrfs_root *root;
274 	unsigned long size;
275 	int index;
276 	int num_pages;
277 	unsigned check_crcs:1;
278 };
279 
280 static int io_ctl_init(struct io_ctl *io_ctl, struct inode *inode,
281 		       struct btrfs_root *root)
282 {
283 	memset(io_ctl, 0, sizeof(struct io_ctl));
284 	io_ctl->num_pages = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
285 		PAGE_CACHE_SHIFT;
286 	io_ctl->pages = kzalloc(sizeof(struct page *) * io_ctl->num_pages,
287 				GFP_NOFS);
288 	if (!io_ctl->pages)
289 		return -ENOMEM;
290 	io_ctl->root = root;
291 	if (btrfs_ino(inode) != BTRFS_FREE_INO_OBJECTID)
292 		io_ctl->check_crcs = 1;
293 	return 0;
294 }
295 
296 static void io_ctl_free(struct io_ctl *io_ctl)
297 {
298 	kfree(io_ctl->pages);
299 }
300 
301 static void io_ctl_unmap_page(struct io_ctl *io_ctl)
302 {
303 	if (io_ctl->cur) {
304 		kunmap(io_ctl->page);
305 		io_ctl->cur = NULL;
306 		io_ctl->orig = NULL;
307 	}
308 }
309 
310 static void io_ctl_map_page(struct io_ctl *io_ctl, int clear)
311 {
312 	BUG_ON(io_ctl->index >= io_ctl->num_pages);
313 	io_ctl->page = io_ctl->pages[io_ctl->index++];
314 	io_ctl->cur = kmap(io_ctl->page);
315 	io_ctl->orig = io_ctl->cur;
316 	io_ctl->size = PAGE_CACHE_SIZE;
317 	if (clear)
318 		memset(io_ctl->cur, 0, PAGE_CACHE_SIZE);
319 }
320 
321 static void io_ctl_drop_pages(struct io_ctl *io_ctl)
322 {
323 	int i;
324 
325 	io_ctl_unmap_page(io_ctl);
326 
327 	for (i = 0; i < io_ctl->num_pages; i++) {
328 		if (io_ctl->pages[i]) {
329 			ClearPageChecked(io_ctl->pages[i]);
330 			unlock_page(io_ctl->pages[i]);
331 			page_cache_release(io_ctl->pages[i]);
332 		}
333 	}
334 }
335 
336 static int io_ctl_prepare_pages(struct io_ctl *io_ctl, struct inode *inode,
337 				int uptodate)
338 {
339 	struct page *page;
340 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
341 	int i;
342 
343 	for (i = 0; i < io_ctl->num_pages; i++) {
344 		page = find_or_create_page(inode->i_mapping, i, mask);
345 		if (!page) {
346 			io_ctl_drop_pages(io_ctl);
347 			return -ENOMEM;
348 		}
349 		io_ctl->pages[i] = page;
350 		if (uptodate && !PageUptodate(page)) {
351 			btrfs_readpage(NULL, page);
352 			lock_page(page);
353 			if (!PageUptodate(page)) {
354 				printk(KERN_ERR "btrfs: error reading free "
355 				       "space cache\n");
356 				io_ctl_drop_pages(io_ctl);
357 				return -EIO;
358 			}
359 		}
360 	}
361 
362 	for (i = 0; i < io_ctl->num_pages; i++) {
363 		clear_page_dirty_for_io(io_ctl->pages[i]);
364 		set_page_extent_mapped(io_ctl->pages[i]);
365 	}
366 
367 	return 0;
368 }
369 
370 static void io_ctl_set_generation(struct io_ctl *io_ctl, u64 generation)
371 {
372 	__le64 *val;
373 
374 	io_ctl_map_page(io_ctl, 1);
375 
376 	/*
377 	 * Skip the csum areas.  If we don't check crcs then we just have a
378 	 * 64bit chunk at the front of the first page.
379 	 */
380 	if (io_ctl->check_crcs) {
381 		io_ctl->cur += (sizeof(u32) * io_ctl->num_pages);
382 		io_ctl->size -= sizeof(u64) + (sizeof(u32) * io_ctl->num_pages);
383 	} else {
384 		io_ctl->cur += sizeof(u64);
385 		io_ctl->size -= sizeof(u64) * 2;
386 	}
387 
388 	val = io_ctl->cur;
389 	*val = cpu_to_le64(generation);
390 	io_ctl->cur += sizeof(u64);
391 }
392 
393 static int io_ctl_check_generation(struct io_ctl *io_ctl, u64 generation)
394 {
395 	__le64 *gen;
396 
397 	/*
398 	 * Skip the crc area.  If we don't check crcs then we just have a 64bit
399 	 * chunk at the front of the first page.
400 	 */
401 	if (io_ctl->check_crcs) {
402 		io_ctl->cur += sizeof(u32) * io_ctl->num_pages;
403 		io_ctl->size -= sizeof(u64) +
404 			(sizeof(u32) * io_ctl->num_pages);
405 	} else {
406 		io_ctl->cur += sizeof(u64);
407 		io_ctl->size -= sizeof(u64) * 2;
408 	}
409 
410 	gen = io_ctl->cur;
411 	if (le64_to_cpu(*gen) != generation) {
412 		printk_ratelimited(KERN_ERR "btrfs: space cache generation "
413 				   "(%Lu) does not match inode (%Lu)\n", *gen,
414 				   generation);
415 		io_ctl_unmap_page(io_ctl);
416 		return -EIO;
417 	}
418 	io_ctl->cur += sizeof(u64);
419 	return 0;
420 }
421 
422 static void io_ctl_set_crc(struct io_ctl *io_ctl, int index)
423 {
424 	u32 *tmp;
425 	u32 crc = ~(u32)0;
426 	unsigned offset = 0;
427 
428 	if (!io_ctl->check_crcs) {
429 		io_ctl_unmap_page(io_ctl);
430 		return;
431 	}
432 
433 	if (index == 0)
434 		offset = sizeof(u32) * io_ctl->num_pages;
435 
436 	crc = btrfs_csum_data(io_ctl->orig + offset, crc,
437 			      PAGE_CACHE_SIZE - offset);
438 	btrfs_csum_final(crc, (char *)&crc);
439 	io_ctl_unmap_page(io_ctl);
440 	tmp = kmap(io_ctl->pages[0]);
441 	tmp += index;
442 	*tmp = crc;
443 	kunmap(io_ctl->pages[0]);
444 }
445 
446 static int io_ctl_check_crc(struct io_ctl *io_ctl, int index)
447 {
448 	u32 *tmp, val;
449 	u32 crc = ~(u32)0;
450 	unsigned offset = 0;
451 
452 	if (!io_ctl->check_crcs) {
453 		io_ctl_map_page(io_ctl, 0);
454 		return 0;
455 	}
456 
457 	if (index == 0)
458 		offset = sizeof(u32) * io_ctl->num_pages;
459 
460 	tmp = kmap(io_ctl->pages[0]);
461 	tmp += index;
462 	val = *tmp;
463 	kunmap(io_ctl->pages[0]);
464 
465 	io_ctl_map_page(io_ctl, 0);
466 	crc = btrfs_csum_data(io_ctl->orig + offset, crc,
467 			      PAGE_CACHE_SIZE - offset);
468 	btrfs_csum_final(crc, (char *)&crc);
469 	if (val != crc) {
470 		printk_ratelimited(KERN_ERR "btrfs: csum mismatch on free "
471 				   "space cache\n");
472 		io_ctl_unmap_page(io_ctl);
473 		return -EIO;
474 	}
475 
476 	return 0;
477 }
478 
479 static int io_ctl_add_entry(struct io_ctl *io_ctl, u64 offset, u64 bytes,
480 			    void *bitmap)
481 {
482 	struct btrfs_free_space_entry *entry;
483 
484 	if (!io_ctl->cur)
485 		return -ENOSPC;
486 
487 	entry = io_ctl->cur;
488 	entry->offset = cpu_to_le64(offset);
489 	entry->bytes = cpu_to_le64(bytes);
490 	entry->type = (bitmap) ? BTRFS_FREE_SPACE_BITMAP :
491 		BTRFS_FREE_SPACE_EXTENT;
492 	io_ctl->cur += sizeof(struct btrfs_free_space_entry);
493 	io_ctl->size -= sizeof(struct btrfs_free_space_entry);
494 
495 	if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
496 		return 0;
497 
498 	io_ctl_set_crc(io_ctl, io_ctl->index - 1);
499 
500 	/* No more pages to map */
501 	if (io_ctl->index >= io_ctl->num_pages)
502 		return 0;
503 
504 	/* map the next page */
505 	io_ctl_map_page(io_ctl, 1);
506 	return 0;
507 }
508 
509 static int io_ctl_add_bitmap(struct io_ctl *io_ctl, void *bitmap)
510 {
511 	if (!io_ctl->cur)
512 		return -ENOSPC;
513 
514 	/*
515 	 * If we aren't at the start of the current page, unmap this one and
516 	 * map the next one if there is any left.
517 	 */
518 	if (io_ctl->cur != io_ctl->orig) {
519 		io_ctl_set_crc(io_ctl, io_ctl->index - 1);
520 		if (io_ctl->index >= io_ctl->num_pages)
521 			return -ENOSPC;
522 		io_ctl_map_page(io_ctl, 0);
523 	}
524 
525 	memcpy(io_ctl->cur, bitmap, PAGE_CACHE_SIZE);
526 	io_ctl_set_crc(io_ctl, io_ctl->index - 1);
527 	if (io_ctl->index < io_ctl->num_pages)
528 		io_ctl_map_page(io_ctl, 0);
529 	return 0;
530 }
531 
532 static void io_ctl_zero_remaining_pages(struct io_ctl *io_ctl)
533 {
534 	/*
535 	 * If we're not on the boundary we know we've modified the page and we
536 	 * need to crc the page.
537 	 */
538 	if (io_ctl->cur != io_ctl->orig)
539 		io_ctl_set_crc(io_ctl, io_ctl->index - 1);
540 	else
541 		io_ctl_unmap_page(io_ctl);
542 
543 	while (io_ctl->index < io_ctl->num_pages) {
544 		io_ctl_map_page(io_ctl, 1);
545 		io_ctl_set_crc(io_ctl, io_ctl->index - 1);
546 	}
547 }
548 
549 static int io_ctl_read_entry(struct io_ctl *io_ctl,
550 			    struct btrfs_free_space *entry, u8 *type)
551 {
552 	struct btrfs_free_space_entry *e;
553 	int ret;
554 
555 	if (!io_ctl->cur) {
556 		ret = io_ctl_check_crc(io_ctl, io_ctl->index);
557 		if (ret)
558 			return ret;
559 	}
560 
561 	e = io_ctl->cur;
562 	entry->offset = le64_to_cpu(e->offset);
563 	entry->bytes = le64_to_cpu(e->bytes);
564 	*type = e->type;
565 	io_ctl->cur += sizeof(struct btrfs_free_space_entry);
566 	io_ctl->size -= sizeof(struct btrfs_free_space_entry);
567 
568 	if (io_ctl->size >= sizeof(struct btrfs_free_space_entry))
569 		return 0;
570 
571 	io_ctl_unmap_page(io_ctl);
572 
573 	return 0;
574 }
575 
576 static int io_ctl_read_bitmap(struct io_ctl *io_ctl,
577 			      struct btrfs_free_space *entry)
578 {
579 	int ret;
580 
581 	ret = io_ctl_check_crc(io_ctl, io_ctl->index);
582 	if (ret)
583 		return ret;
584 
585 	memcpy(entry->bitmap, io_ctl->cur, PAGE_CACHE_SIZE);
586 	io_ctl_unmap_page(io_ctl);
587 
588 	return 0;
589 }
590 
591 /*
592  * Since we attach pinned extents after the fact we can have contiguous sections
593  * of free space that are split up in entries.  This poses a problem with the
594  * tree logging stuff since it could have allocated across what appears to be 2
595  * entries since we would have merged the entries when adding the pinned extents
596  * back to the free space cache.  So run through the space cache that we just
597  * loaded and merge contiguous entries.  This will make the log replay stuff not
598  * blow up and it will make for nicer allocator behavior.
599  */
600 static void merge_space_tree(struct btrfs_free_space_ctl *ctl)
601 {
602 	struct btrfs_free_space *e, *prev = NULL;
603 	struct rb_node *n;
604 
605 again:
606 	spin_lock(&ctl->tree_lock);
607 	for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
608 		e = rb_entry(n, struct btrfs_free_space, offset_index);
609 		if (!prev)
610 			goto next;
611 		if (e->bitmap || prev->bitmap)
612 			goto next;
613 		if (prev->offset + prev->bytes == e->offset) {
614 			unlink_free_space(ctl, prev);
615 			unlink_free_space(ctl, e);
616 			prev->bytes += e->bytes;
617 			kmem_cache_free(btrfs_free_space_cachep, e);
618 			link_free_space(ctl, prev);
619 			prev = NULL;
620 			spin_unlock(&ctl->tree_lock);
621 			goto again;
622 		}
623 next:
624 		prev = e;
625 	}
626 	spin_unlock(&ctl->tree_lock);
627 }
628 
629 static int __load_free_space_cache(struct btrfs_root *root, struct inode *inode,
630 				   struct btrfs_free_space_ctl *ctl,
631 				   struct btrfs_path *path, u64 offset)
632 {
633 	struct btrfs_free_space_header *header;
634 	struct extent_buffer *leaf;
635 	struct io_ctl io_ctl;
636 	struct btrfs_key key;
637 	struct btrfs_free_space *e, *n;
638 	struct list_head bitmaps;
639 	u64 num_entries;
640 	u64 num_bitmaps;
641 	u64 generation;
642 	u8 type;
643 	int ret = 0;
644 
645 	INIT_LIST_HEAD(&bitmaps);
646 
647 	/* Nothing in the space cache, goodbye */
648 	if (!i_size_read(inode))
649 		return 0;
650 
651 	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
652 	key.offset = offset;
653 	key.type = 0;
654 
655 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
656 	if (ret < 0)
657 		return 0;
658 	else if (ret > 0) {
659 		btrfs_release_path(path);
660 		return 0;
661 	}
662 
663 	ret = -1;
664 
665 	leaf = path->nodes[0];
666 	header = btrfs_item_ptr(leaf, path->slots[0],
667 				struct btrfs_free_space_header);
668 	num_entries = btrfs_free_space_entries(leaf, header);
669 	num_bitmaps = btrfs_free_space_bitmaps(leaf, header);
670 	generation = btrfs_free_space_generation(leaf, header);
671 	btrfs_release_path(path);
672 
673 	if (BTRFS_I(inode)->generation != generation) {
674 		btrfs_err(root->fs_info,
675 			"free space inode generation (%llu) "
676 			"did not match free space cache generation (%llu)",
677 			(unsigned long long)BTRFS_I(inode)->generation,
678 			(unsigned long long)generation);
679 		return 0;
680 	}
681 
682 	if (!num_entries)
683 		return 0;
684 
685 	ret = io_ctl_init(&io_ctl, inode, root);
686 	if (ret)
687 		return ret;
688 
689 	ret = readahead_cache(inode);
690 	if (ret)
691 		goto out;
692 
693 	ret = io_ctl_prepare_pages(&io_ctl, inode, 1);
694 	if (ret)
695 		goto out;
696 
697 	ret = io_ctl_check_crc(&io_ctl, 0);
698 	if (ret)
699 		goto free_cache;
700 
701 	ret = io_ctl_check_generation(&io_ctl, generation);
702 	if (ret)
703 		goto free_cache;
704 
705 	while (num_entries) {
706 		e = kmem_cache_zalloc(btrfs_free_space_cachep,
707 				      GFP_NOFS);
708 		if (!e)
709 			goto free_cache;
710 
711 		ret = io_ctl_read_entry(&io_ctl, e, &type);
712 		if (ret) {
713 			kmem_cache_free(btrfs_free_space_cachep, e);
714 			goto free_cache;
715 		}
716 
717 		if (!e->bytes) {
718 			kmem_cache_free(btrfs_free_space_cachep, e);
719 			goto free_cache;
720 		}
721 
722 		if (type == BTRFS_FREE_SPACE_EXTENT) {
723 			spin_lock(&ctl->tree_lock);
724 			ret = link_free_space(ctl, e);
725 			spin_unlock(&ctl->tree_lock);
726 			if (ret) {
727 				btrfs_err(root->fs_info,
728 					"Duplicate entries in free space cache, dumping");
729 				kmem_cache_free(btrfs_free_space_cachep, e);
730 				goto free_cache;
731 			}
732 		} else {
733 			BUG_ON(!num_bitmaps);
734 			num_bitmaps--;
735 			e->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
736 			if (!e->bitmap) {
737 				kmem_cache_free(
738 					btrfs_free_space_cachep, e);
739 				goto free_cache;
740 			}
741 			spin_lock(&ctl->tree_lock);
742 			ret = link_free_space(ctl, e);
743 			ctl->total_bitmaps++;
744 			ctl->op->recalc_thresholds(ctl);
745 			spin_unlock(&ctl->tree_lock);
746 			if (ret) {
747 				btrfs_err(root->fs_info,
748 					"Duplicate entries in free space cache, dumping");
749 				kmem_cache_free(btrfs_free_space_cachep, e);
750 				goto free_cache;
751 			}
752 			list_add_tail(&e->list, &bitmaps);
753 		}
754 
755 		num_entries--;
756 	}
757 
758 	io_ctl_unmap_page(&io_ctl);
759 
760 	/*
761 	 * We add the bitmaps at the end of the entries in order that
762 	 * the bitmap entries are added to the cache.
763 	 */
764 	list_for_each_entry_safe(e, n, &bitmaps, list) {
765 		list_del_init(&e->list);
766 		ret = io_ctl_read_bitmap(&io_ctl, e);
767 		if (ret)
768 			goto free_cache;
769 	}
770 
771 	io_ctl_drop_pages(&io_ctl);
772 	merge_space_tree(ctl);
773 	ret = 1;
774 out:
775 	io_ctl_free(&io_ctl);
776 	return ret;
777 free_cache:
778 	io_ctl_drop_pages(&io_ctl);
779 	__btrfs_remove_free_space_cache(ctl);
780 	goto out;
781 }
782 
783 int load_free_space_cache(struct btrfs_fs_info *fs_info,
784 			  struct btrfs_block_group_cache *block_group)
785 {
786 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
787 	struct btrfs_root *root = fs_info->tree_root;
788 	struct inode *inode;
789 	struct btrfs_path *path;
790 	int ret = 0;
791 	bool matched;
792 	u64 used = btrfs_block_group_used(&block_group->item);
793 
794 	/*
795 	 * If this block group has been marked to be cleared for one reason or
796 	 * another then we can't trust the on disk cache, so just return.
797 	 */
798 	spin_lock(&block_group->lock);
799 	if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
800 		spin_unlock(&block_group->lock);
801 		return 0;
802 	}
803 	spin_unlock(&block_group->lock);
804 
805 	path = btrfs_alloc_path();
806 	if (!path)
807 		return 0;
808 	path->search_commit_root = 1;
809 	path->skip_locking = 1;
810 
811 	inode = lookup_free_space_inode(root, block_group, path);
812 	if (IS_ERR(inode)) {
813 		btrfs_free_path(path);
814 		return 0;
815 	}
816 
817 	/* We may have converted the inode and made the cache invalid. */
818 	spin_lock(&block_group->lock);
819 	if (block_group->disk_cache_state != BTRFS_DC_WRITTEN) {
820 		spin_unlock(&block_group->lock);
821 		btrfs_free_path(path);
822 		goto out;
823 	}
824 	spin_unlock(&block_group->lock);
825 
826 	ret = __load_free_space_cache(fs_info->tree_root, inode, ctl,
827 				      path, block_group->key.objectid);
828 	btrfs_free_path(path);
829 	if (ret <= 0)
830 		goto out;
831 
832 	spin_lock(&ctl->tree_lock);
833 	matched = (ctl->free_space == (block_group->key.offset - used -
834 				       block_group->bytes_super));
835 	spin_unlock(&ctl->tree_lock);
836 
837 	if (!matched) {
838 		__btrfs_remove_free_space_cache(ctl);
839 		btrfs_err(fs_info, "block group %llu has wrong amount of free space",
840 			block_group->key.objectid);
841 		ret = -1;
842 	}
843 out:
844 	if (ret < 0) {
845 		/* This cache is bogus, make sure it gets cleared */
846 		spin_lock(&block_group->lock);
847 		block_group->disk_cache_state = BTRFS_DC_CLEAR;
848 		spin_unlock(&block_group->lock);
849 		ret = 0;
850 
851 		btrfs_err(fs_info, "failed to load free space cache for block group %llu",
852 			block_group->key.objectid);
853 	}
854 
855 	iput(inode);
856 	return ret;
857 }
858 
859 /**
860  * __btrfs_write_out_cache - write out cached info to an inode
861  * @root - the root the inode belongs to
862  * @ctl - the free space cache we are going to write out
863  * @block_group - the block_group for this cache if it belongs to a block_group
864  * @trans - the trans handle
865  * @path - the path to use
866  * @offset - the offset for the key we'll insert
867  *
868  * This function writes out a free space cache struct to disk for quick recovery
869  * on mount.  This will return 0 if it was successfull in writing the cache out,
870  * and -1 if it was not.
871  */
872 static int __btrfs_write_out_cache(struct btrfs_root *root, struct inode *inode,
873 				   struct btrfs_free_space_ctl *ctl,
874 				   struct btrfs_block_group_cache *block_group,
875 				   struct btrfs_trans_handle *trans,
876 				   struct btrfs_path *path, u64 offset)
877 {
878 	struct btrfs_free_space_header *header;
879 	struct extent_buffer *leaf;
880 	struct rb_node *node;
881 	struct list_head *pos, *n;
882 	struct extent_state *cached_state = NULL;
883 	struct btrfs_free_cluster *cluster = NULL;
884 	struct extent_io_tree *unpin = NULL;
885 	struct io_ctl io_ctl;
886 	struct list_head bitmap_list;
887 	struct btrfs_key key;
888 	u64 start, extent_start, extent_end, len;
889 	int entries = 0;
890 	int bitmaps = 0;
891 	int ret;
892 	int err = -1;
893 
894 	INIT_LIST_HEAD(&bitmap_list);
895 
896 	if (!i_size_read(inode))
897 		return -1;
898 
899 	ret = io_ctl_init(&io_ctl, inode, root);
900 	if (ret)
901 		return -1;
902 
903 	/* Get the cluster for this block_group if it exists */
904 	if (block_group && !list_empty(&block_group->cluster_list))
905 		cluster = list_entry(block_group->cluster_list.next,
906 				     struct btrfs_free_cluster,
907 				     block_group_list);
908 
909 	/* Lock all pages first so we can lock the extent safely. */
910 	io_ctl_prepare_pages(&io_ctl, inode, 0);
911 
912 	lock_extent_bits(&BTRFS_I(inode)->io_tree, 0, i_size_read(inode) - 1,
913 			 0, &cached_state);
914 
915 	node = rb_first(&ctl->free_space_offset);
916 	if (!node && cluster) {
917 		node = rb_first(&cluster->root);
918 		cluster = NULL;
919 	}
920 
921 	/* Make sure we can fit our crcs into the first page */
922 	if (io_ctl.check_crcs &&
923 	    (io_ctl.num_pages * sizeof(u32)) >= PAGE_CACHE_SIZE) {
924 		WARN_ON(1);
925 		goto out_nospc;
926 	}
927 
928 	io_ctl_set_generation(&io_ctl, trans->transid);
929 
930 	/* Write out the extent entries */
931 	while (node) {
932 		struct btrfs_free_space *e;
933 
934 		e = rb_entry(node, struct btrfs_free_space, offset_index);
935 		entries++;
936 
937 		ret = io_ctl_add_entry(&io_ctl, e->offset, e->bytes,
938 				       e->bitmap);
939 		if (ret)
940 			goto out_nospc;
941 
942 		if (e->bitmap) {
943 			list_add_tail(&e->list, &bitmap_list);
944 			bitmaps++;
945 		}
946 		node = rb_next(node);
947 		if (!node && cluster) {
948 			node = rb_first(&cluster->root);
949 			cluster = NULL;
950 		}
951 	}
952 
953 	/*
954 	 * We want to add any pinned extents to our free space cache
955 	 * so we don't leak the space
956 	 */
957 
958 	/*
959 	 * We shouldn't have switched the pinned extents yet so this is the
960 	 * right one
961 	 */
962 	unpin = root->fs_info->pinned_extents;
963 
964 	if (block_group)
965 		start = block_group->key.objectid;
966 
967 	while (block_group && (start < block_group->key.objectid +
968 			       block_group->key.offset)) {
969 		ret = find_first_extent_bit(unpin, start,
970 					    &extent_start, &extent_end,
971 					    EXTENT_DIRTY, NULL);
972 		if (ret) {
973 			ret = 0;
974 			break;
975 		}
976 
977 		/* This pinned extent is out of our range */
978 		if (extent_start >= block_group->key.objectid +
979 		    block_group->key.offset)
980 			break;
981 
982 		extent_start = max(extent_start, start);
983 		extent_end = min(block_group->key.objectid +
984 				 block_group->key.offset, extent_end + 1);
985 		len = extent_end - extent_start;
986 
987 		entries++;
988 		ret = io_ctl_add_entry(&io_ctl, extent_start, len, NULL);
989 		if (ret)
990 			goto out_nospc;
991 
992 		start = extent_end;
993 	}
994 
995 	/* Write out the bitmaps */
996 	list_for_each_safe(pos, n, &bitmap_list) {
997 		struct btrfs_free_space *entry =
998 			list_entry(pos, struct btrfs_free_space, list);
999 
1000 		ret = io_ctl_add_bitmap(&io_ctl, entry->bitmap);
1001 		if (ret)
1002 			goto out_nospc;
1003 		list_del_init(&entry->list);
1004 	}
1005 
1006 	/* Zero out the rest of the pages just to make sure */
1007 	io_ctl_zero_remaining_pages(&io_ctl);
1008 
1009 	ret = btrfs_dirty_pages(root, inode, io_ctl.pages, io_ctl.num_pages,
1010 				0, i_size_read(inode), &cached_state);
1011 	io_ctl_drop_pages(&io_ctl);
1012 	unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
1013 			     i_size_read(inode) - 1, &cached_state, GFP_NOFS);
1014 
1015 	if (ret)
1016 		goto out;
1017 
1018 
1019 	btrfs_wait_ordered_range(inode, 0, (u64)-1);
1020 
1021 	key.objectid = BTRFS_FREE_SPACE_OBJECTID;
1022 	key.offset = offset;
1023 	key.type = 0;
1024 
1025 	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
1026 	if (ret < 0) {
1027 		clear_extent_bit(&BTRFS_I(inode)->io_tree, 0, inode->i_size - 1,
1028 				 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, NULL,
1029 				 GFP_NOFS);
1030 		goto out;
1031 	}
1032 	leaf = path->nodes[0];
1033 	if (ret > 0) {
1034 		struct btrfs_key found_key;
1035 		BUG_ON(!path->slots[0]);
1036 		path->slots[0]--;
1037 		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1038 		if (found_key.objectid != BTRFS_FREE_SPACE_OBJECTID ||
1039 		    found_key.offset != offset) {
1040 			clear_extent_bit(&BTRFS_I(inode)->io_tree, 0,
1041 					 inode->i_size - 1,
1042 					 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0,
1043 					 NULL, GFP_NOFS);
1044 			btrfs_release_path(path);
1045 			goto out;
1046 		}
1047 	}
1048 
1049 	BTRFS_I(inode)->generation = trans->transid;
1050 	header = btrfs_item_ptr(leaf, path->slots[0],
1051 				struct btrfs_free_space_header);
1052 	btrfs_set_free_space_entries(leaf, header, entries);
1053 	btrfs_set_free_space_bitmaps(leaf, header, bitmaps);
1054 	btrfs_set_free_space_generation(leaf, header, trans->transid);
1055 	btrfs_mark_buffer_dirty(leaf);
1056 	btrfs_release_path(path);
1057 
1058 	err = 0;
1059 out:
1060 	io_ctl_free(&io_ctl);
1061 	if (err) {
1062 		invalidate_inode_pages2(inode->i_mapping);
1063 		BTRFS_I(inode)->generation = 0;
1064 	}
1065 	btrfs_update_inode(trans, root, inode);
1066 	return err;
1067 
1068 out_nospc:
1069 	list_for_each_safe(pos, n, &bitmap_list) {
1070 		struct btrfs_free_space *entry =
1071 			list_entry(pos, struct btrfs_free_space, list);
1072 		list_del_init(&entry->list);
1073 	}
1074 	io_ctl_drop_pages(&io_ctl);
1075 	unlock_extent_cached(&BTRFS_I(inode)->io_tree, 0,
1076 			     i_size_read(inode) - 1, &cached_state, GFP_NOFS);
1077 	goto out;
1078 }
1079 
1080 int btrfs_write_out_cache(struct btrfs_root *root,
1081 			  struct btrfs_trans_handle *trans,
1082 			  struct btrfs_block_group_cache *block_group,
1083 			  struct btrfs_path *path)
1084 {
1085 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1086 	struct inode *inode;
1087 	int ret = 0;
1088 
1089 	root = root->fs_info->tree_root;
1090 
1091 	spin_lock(&block_group->lock);
1092 	if (block_group->disk_cache_state < BTRFS_DC_SETUP) {
1093 		spin_unlock(&block_group->lock);
1094 		return 0;
1095 	}
1096 	spin_unlock(&block_group->lock);
1097 
1098 	inode = lookup_free_space_inode(root, block_group, path);
1099 	if (IS_ERR(inode))
1100 		return 0;
1101 
1102 	ret = __btrfs_write_out_cache(root, inode, ctl, block_group, trans,
1103 				      path, block_group->key.objectid);
1104 	if (ret) {
1105 		spin_lock(&block_group->lock);
1106 		block_group->disk_cache_state = BTRFS_DC_ERROR;
1107 		spin_unlock(&block_group->lock);
1108 		ret = 0;
1109 #ifdef DEBUG
1110 		btrfs_err(root->fs_info,
1111 			"failed to write free space cache for block group %llu",
1112 			block_group->key.objectid);
1113 #endif
1114 	}
1115 
1116 	iput(inode);
1117 	return ret;
1118 }
1119 
1120 static inline unsigned long offset_to_bit(u64 bitmap_start, u32 unit,
1121 					  u64 offset)
1122 {
1123 	BUG_ON(offset < bitmap_start);
1124 	offset -= bitmap_start;
1125 	return (unsigned long)(div_u64(offset, unit));
1126 }
1127 
1128 static inline unsigned long bytes_to_bits(u64 bytes, u32 unit)
1129 {
1130 	return (unsigned long)(div_u64(bytes, unit));
1131 }
1132 
1133 static inline u64 offset_to_bitmap(struct btrfs_free_space_ctl *ctl,
1134 				   u64 offset)
1135 {
1136 	u64 bitmap_start;
1137 	u64 bytes_per_bitmap;
1138 
1139 	bytes_per_bitmap = BITS_PER_BITMAP * ctl->unit;
1140 	bitmap_start = offset - ctl->start;
1141 	bitmap_start = div64_u64(bitmap_start, bytes_per_bitmap);
1142 	bitmap_start *= bytes_per_bitmap;
1143 	bitmap_start += ctl->start;
1144 
1145 	return bitmap_start;
1146 }
1147 
1148 static int tree_insert_offset(struct rb_root *root, u64 offset,
1149 			      struct rb_node *node, int bitmap)
1150 {
1151 	struct rb_node **p = &root->rb_node;
1152 	struct rb_node *parent = NULL;
1153 	struct btrfs_free_space *info;
1154 
1155 	while (*p) {
1156 		parent = *p;
1157 		info = rb_entry(parent, struct btrfs_free_space, offset_index);
1158 
1159 		if (offset < info->offset) {
1160 			p = &(*p)->rb_left;
1161 		} else if (offset > info->offset) {
1162 			p = &(*p)->rb_right;
1163 		} else {
1164 			/*
1165 			 * we could have a bitmap entry and an extent entry
1166 			 * share the same offset.  If this is the case, we want
1167 			 * the extent entry to always be found first if we do a
1168 			 * linear search through the tree, since we want to have
1169 			 * the quickest allocation time, and allocating from an
1170 			 * extent is faster than allocating from a bitmap.  So
1171 			 * if we're inserting a bitmap and we find an entry at
1172 			 * this offset, we want to go right, or after this entry
1173 			 * logically.  If we are inserting an extent and we've
1174 			 * found a bitmap, we want to go left, or before
1175 			 * logically.
1176 			 */
1177 			if (bitmap) {
1178 				if (info->bitmap) {
1179 					WARN_ON_ONCE(1);
1180 					return -EEXIST;
1181 				}
1182 				p = &(*p)->rb_right;
1183 			} else {
1184 				if (!info->bitmap) {
1185 					WARN_ON_ONCE(1);
1186 					return -EEXIST;
1187 				}
1188 				p = &(*p)->rb_left;
1189 			}
1190 		}
1191 	}
1192 
1193 	rb_link_node(node, parent, p);
1194 	rb_insert_color(node, root);
1195 
1196 	return 0;
1197 }
1198 
1199 /*
1200  * searches the tree for the given offset.
1201  *
1202  * fuzzy - If this is set, then we are trying to make an allocation, and we just
1203  * want a section that has at least bytes size and comes at or after the given
1204  * offset.
1205  */
1206 static struct btrfs_free_space *
1207 tree_search_offset(struct btrfs_free_space_ctl *ctl,
1208 		   u64 offset, int bitmap_only, int fuzzy)
1209 {
1210 	struct rb_node *n = ctl->free_space_offset.rb_node;
1211 	struct btrfs_free_space *entry, *prev = NULL;
1212 
1213 	/* find entry that is closest to the 'offset' */
1214 	while (1) {
1215 		if (!n) {
1216 			entry = NULL;
1217 			break;
1218 		}
1219 
1220 		entry = rb_entry(n, struct btrfs_free_space, offset_index);
1221 		prev = entry;
1222 
1223 		if (offset < entry->offset)
1224 			n = n->rb_left;
1225 		else if (offset > entry->offset)
1226 			n = n->rb_right;
1227 		else
1228 			break;
1229 	}
1230 
1231 	if (bitmap_only) {
1232 		if (!entry)
1233 			return NULL;
1234 		if (entry->bitmap)
1235 			return entry;
1236 
1237 		/*
1238 		 * bitmap entry and extent entry may share same offset,
1239 		 * in that case, bitmap entry comes after extent entry.
1240 		 */
1241 		n = rb_next(n);
1242 		if (!n)
1243 			return NULL;
1244 		entry = rb_entry(n, struct btrfs_free_space, offset_index);
1245 		if (entry->offset != offset)
1246 			return NULL;
1247 
1248 		WARN_ON(!entry->bitmap);
1249 		return entry;
1250 	} else if (entry) {
1251 		if (entry->bitmap) {
1252 			/*
1253 			 * if previous extent entry covers the offset,
1254 			 * we should return it instead of the bitmap entry
1255 			 */
1256 			n = rb_prev(&entry->offset_index);
1257 			if (n) {
1258 				prev = rb_entry(n, struct btrfs_free_space,
1259 						offset_index);
1260 				if (!prev->bitmap &&
1261 				    prev->offset + prev->bytes > offset)
1262 					entry = prev;
1263 			}
1264 		}
1265 		return entry;
1266 	}
1267 
1268 	if (!prev)
1269 		return NULL;
1270 
1271 	/* find last entry before the 'offset' */
1272 	entry = prev;
1273 	if (entry->offset > offset) {
1274 		n = rb_prev(&entry->offset_index);
1275 		if (n) {
1276 			entry = rb_entry(n, struct btrfs_free_space,
1277 					offset_index);
1278 			BUG_ON(entry->offset > offset);
1279 		} else {
1280 			if (fuzzy)
1281 				return entry;
1282 			else
1283 				return NULL;
1284 		}
1285 	}
1286 
1287 	if (entry->bitmap) {
1288 		n = rb_prev(&entry->offset_index);
1289 		if (n) {
1290 			prev = rb_entry(n, struct btrfs_free_space,
1291 					offset_index);
1292 			if (!prev->bitmap &&
1293 			    prev->offset + prev->bytes > offset)
1294 				return prev;
1295 		}
1296 		if (entry->offset + BITS_PER_BITMAP * ctl->unit > offset)
1297 			return entry;
1298 	} else if (entry->offset + entry->bytes > offset)
1299 		return entry;
1300 
1301 	if (!fuzzy)
1302 		return NULL;
1303 
1304 	while (1) {
1305 		if (entry->bitmap) {
1306 			if (entry->offset + BITS_PER_BITMAP *
1307 			    ctl->unit > offset)
1308 				break;
1309 		} else {
1310 			if (entry->offset + entry->bytes > offset)
1311 				break;
1312 		}
1313 
1314 		n = rb_next(&entry->offset_index);
1315 		if (!n)
1316 			return NULL;
1317 		entry = rb_entry(n, struct btrfs_free_space, offset_index);
1318 	}
1319 	return entry;
1320 }
1321 
1322 static inline void
1323 __unlink_free_space(struct btrfs_free_space_ctl *ctl,
1324 		    struct btrfs_free_space *info)
1325 {
1326 	rb_erase(&info->offset_index, &ctl->free_space_offset);
1327 	ctl->free_extents--;
1328 }
1329 
1330 static void unlink_free_space(struct btrfs_free_space_ctl *ctl,
1331 			      struct btrfs_free_space *info)
1332 {
1333 	__unlink_free_space(ctl, info);
1334 	ctl->free_space -= info->bytes;
1335 }
1336 
1337 static int link_free_space(struct btrfs_free_space_ctl *ctl,
1338 			   struct btrfs_free_space *info)
1339 {
1340 	int ret = 0;
1341 
1342 	BUG_ON(!info->bitmap && !info->bytes);
1343 	ret = tree_insert_offset(&ctl->free_space_offset, info->offset,
1344 				 &info->offset_index, (info->bitmap != NULL));
1345 	if (ret)
1346 		return ret;
1347 
1348 	ctl->free_space += info->bytes;
1349 	ctl->free_extents++;
1350 	return ret;
1351 }
1352 
1353 static void recalculate_thresholds(struct btrfs_free_space_ctl *ctl)
1354 {
1355 	struct btrfs_block_group_cache *block_group = ctl->private;
1356 	u64 max_bytes;
1357 	u64 bitmap_bytes;
1358 	u64 extent_bytes;
1359 	u64 size = block_group->key.offset;
1360 	u64 bytes_per_bg = BITS_PER_BITMAP * ctl->unit;
1361 	int max_bitmaps = div64_u64(size + bytes_per_bg - 1, bytes_per_bg);
1362 
1363 	max_bitmaps = max(max_bitmaps, 1);
1364 
1365 	BUG_ON(ctl->total_bitmaps > max_bitmaps);
1366 
1367 	/*
1368 	 * The goal is to keep the total amount of memory used per 1gb of space
1369 	 * at or below 32k, so we need to adjust how much memory we allow to be
1370 	 * used by extent based free space tracking
1371 	 */
1372 	if (size < 1024 * 1024 * 1024)
1373 		max_bytes = MAX_CACHE_BYTES_PER_GIG;
1374 	else
1375 		max_bytes = MAX_CACHE_BYTES_PER_GIG *
1376 			div64_u64(size, 1024 * 1024 * 1024);
1377 
1378 	/*
1379 	 * we want to account for 1 more bitmap than what we have so we can make
1380 	 * sure we don't go over our overall goal of MAX_CACHE_BYTES_PER_GIG as
1381 	 * we add more bitmaps.
1382 	 */
1383 	bitmap_bytes = (ctl->total_bitmaps + 1) * PAGE_CACHE_SIZE;
1384 
1385 	if (bitmap_bytes >= max_bytes) {
1386 		ctl->extents_thresh = 0;
1387 		return;
1388 	}
1389 
1390 	/*
1391 	 * we want the extent entry threshold to always be at most 1/2 the maxw
1392 	 * bytes we can have, or whatever is less than that.
1393 	 */
1394 	extent_bytes = max_bytes - bitmap_bytes;
1395 	extent_bytes = min_t(u64, extent_bytes, div64_u64(max_bytes, 2));
1396 
1397 	ctl->extents_thresh =
1398 		div64_u64(extent_bytes, (sizeof(struct btrfs_free_space)));
1399 }
1400 
1401 static inline void __bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
1402 				       struct btrfs_free_space *info,
1403 				       u64 offset, u64 bytes)
1404 {
1405 	unsigned long start, count;
1406 
1407 	start = offset_to_bit(info->offset, ctl->unit, offset);
1408 	count = bytes_to_bits(bytes, ctl->unit);
1409 	BUG_ON(start + count > BITS_PER_BITMAP);
1410 
1411 	bitmap_clear(info->bitmap, start, count);
1412 
1413 	info->bytes -= bytes;
1414 }
1415 
1416 static void bitmap_clear_bits(struct btrfs_free_space_ctl *ctl,
1417 			      struct btrfs_free_space *info, u64 offset,
1418 			      u64 bytes)
1419 {
1420 	__bitmap_clear_bits(ctl, info, offset, bytes);
1421 	ctl->free_space -= bytes;
1422 }
1423 
1424 static void bitmap_set_bits(struct btrfs_free_space_ctl *ctl,
1425 			    struct btrfs_free_space *info, u64 offset,
1426 			    u64 bytes)
1427 {
1428 	unsigned long start, count;
1429 
1430 	start = offset_to_bit(info->offset, ctl->unit, offset);
1431 	count = bytes_to_bits(bytes, ctl->unit);
1432 	BUG_ON(start + count > BITS_PER_BITMAP);
1433 
1434 	bitmap_set(info->bitmap, start, count);
1435 
1436 	info->bytes += bytes;
1437 	ctl->free_space += bytes;
1438 }
1439 
1440 static int search_bitmap(struct btrfs_free_space_ctl *ctl,
1441 			 struct btrfs_free_space *bitmap_info, u64 *offset,
1442 			 u64 *bytes)
1443 {
1444 	unsigned long found_bits = 0;
1445 	unsigned long bits, i;
1446 	unsigned long next_zero;
1447 
1448 	i = offset_to_bit(bitmap_info->offset, ctl->unit,
1449 			  max_t(u64, *offset, bitmap_info->offset));
1450 	bits = bytes_to_bits(*bytes, ctl->unit);
1451 
1452 	for_each_set_bit_from(i, bitmap_info->bitmap, BITS_PER_BITMAP) {
1453 		next_zero = find_next_zero_bit(bitmap_info->bitmap,
1454 					       BITS_PER_BITMAP, i);
1455 		if ((next_zero - i) >= bits) {
1456 			found_bits = next_zero - i;
1457 			break;
1458 		}
1459 		i = next_zero;
1460 	}
1461 
1462 	if (found_bits) {
1463 		*offset = (u64)(i * ctl->unit) + bitmap_info->offset;
1464 		*bytes = (u64)(found_bits) * ctl->unit;
1465 		return 0;
1466 	}
1467 
1468 	return -1;
1469 }
1470 
1471 static struct btrfs_free_space *
1472 find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes,
1473 		unsigned long align)
1474 {
1475 	struct btrfs_free_space *entry;
1476 	struct rb_node *node;
1477 	u64 ctl_off;
1478 	u64 tmp;
1479 	u64 align_off;
1480 	int ret;
1481 
1482 	if (!ctl->free_space_offset.rb_node)
1483 		return NULL;
1484 
1485 	entry = tree_search_offset(ctl, offset_to_bitmap(ctl, *offset), 0, 1);
1486 	if (!entry)
1487 		return NULL;
1488 
1489 	for (node = &entry->offset_index; node; node = rb_next(node)) {
1490 		entry = rb_entry(node, struct btrfs_free_space, offset_index);
1491 		if (entry->bytes < *bytes)
1492 			continue;
1493 
1494 		/* make sure the space returned is big enough
1495 		 * to match our requested alignment
1496 		 */
1497 		if (*bytes >= align) {
1498 			ctl_off = entry->offset - ctl->start;
1499 			tmp = ctl_off + align - 1;;
1500 			do_div(tmp, align);
1501 			tmp = tmp * align + ctl->start;
1502 			align_off = tmp - entry->offset;
1503 		} else {
1504 			align_off = 0;
1505 			tmp = entry->offset;
1506 		}
1507 
1508 		if (entry->bytes < *bytes + align_off)
1509 			continue;
1510 
1511 		if (entry->bitmap) {
1512 			ret = search_bitmap(ctl, entry, &tmp, bytes);
1513 			if (!ret) {
1514 				*offset = tmp;
1515 				return entry;
1516 			}
1517 			continue;
1518 		}
1519 
1520 		*offset = tmp;
1521 		*bytes = entry->bytes - align_off;
1522 		return entry;
1523 	}
1524 
1525 	return NULL;
1526 }
1527 
1528 static void add_new_bitmap(struct btrfs_free_space_ctl *ctl,
1529 			   struct btrfs_free_space *info, u64 offset)
1530 {
1531 	info->offset = offset_to_bitmap(ctl, offset);
1532 	info->bytes = 0;
1533 	INIT_LIST_HEAD(&info->list);
1534 	link_free_space(ctl, info);
1535 	ctl->total_bitmaps++;
1536 
1537 	ctl->op->recalc_thresholds(ctl);
1538 }
1539 
1540 static void free_bitmap(struct btrfs_free_space_ctl *ctl,
1541 			struct btrfs_free_space *bitmap_info)
1542 {
1543 	unlink_free_space(ctl, bitmap_info);
1544 	kfree(bitmap_info->bitmap);
1545 	kmem_cache_free(btrfs_free_space_cachep, bitmap_info);
1546 	ctl->total_bitmaps--;
1547 	ctl->op->recalc_thresholds(ctl);
1548 }
1549 
1550 static noinline int remove_from_bitmap(struct btrfs_free_space_ctl *ctl,
1551 			      struct btrfs_free_space *bitmap_info,
1552 			      u64 *offset, u64 *bytes)
1553 {
1554 	u64 end;
1555 	u64 search_start, search_bytes;
1556 	int ret;
1557 
1558 again:
1559 	end = bitmap_info->offset + (u64)(BITS_PER_BITMAP * ctl->unit) - 1;
1560 
1561 	/*
1562 	 * We need to search for bits in this bitmap.  We could only cover some
1563 	 * of the extent in this bitmap thanks to how we add space, so we need
1564 	 * to search for as much as it as we can and clear that amount, and then
1565 	 * go searching for the next bit.
1566 	 */
1567 	search_start = *offset;
1568 	search_bytes = ctl->unit;
1569 	search_bytes = min(search_bytes, end - search_start + 1);
1570 	ret = search_bitmap(ctl, bitmap_info, &search_start, &search_bytes);
1571 	if (ret < 0 || search_start != *offset)
1572 		return -EINVAL;
1573 
1574 	/* We may have found more bits than what we need */
1575 	search_bytes = min(search_bytes, *bytes);
1576 
1577 	/* Cannot clear past the end of the bitmap */
1578 	search_bytes = min(search_bytes, end - search_start + 1);
1579 
1580 	bitmap_clear_bits(ctl, bitmap_info, search_start, search_bytes);
1581 	*offset += search_bytes;
1582 	*bytes -= search_bytes;
1583 
1584 	if (*bytes) {
1585 		struct rb_node *next = rb_next(&bitmap_info->offset_index);
1586 		if (!bitmap_info->bytes)
1587 			free_bitmap(ctl, bitmap_info);
1588 
1589 		/*
1590 		 * no entry after this bitmap, but we still have bytes to
1591 		 * remove, so something has gone wrong.
1592 		 */
1593 		if (!next)
1594 			return -EINVAL;
1595 
1596 		bitmap_info = rb_entry(next, struct btrfs_free_space,
1597 				       offset_index);
1598 
1599 		/*
1600 		 * if the next entry isn't a bitmap we need to return to let the
1601 		 * extent stuff do its work.
1602 		 */
1603 		if (!bitmap_info->bitmap)
1604 			return -EAGAIN;
1605 
1606 		/*
1607 		 * Ok the next item is a bitmap, but it may not actually hold
1608 		 * the information for the rest of this free space stuff, so
1609 		 * look for it, and if we don't find it return so we can try
1610 		 * everything over again.
1611 		 */
1612 		search_start = *offset;
1613 		search_bytes = ctl->unit;
1614 		ret = search_bitmap(ctl, bitmap_info, &search_start,
1615 				    &search_bytes);
1616 		if (ret < 0 || search_start != *offset)
1617 			return -EAGAIN;
1618 
1619 		goto again;
1620 	} else if (!bitmap_info->bytes)
1621 		free_bitmap(ctl, bitmap_info);
1622 
1623 	return 0;
1624 }
1625 
1626 static u64 add_bytes_to_bitmap(struct btrfs_free_space_ctl *ctl,
1627 			       struct btrfs_free_space *info, u64 offset,
1628 			       u64 bytes)
1629 {
1630 	u64 bytes_to_set = 0;
1631 	u64 end;
1632 
1633 	end = info->offset + (u64)(BITS_PER_BITMAP * ctl->unit);
1634 
1635 	bytes_to_set = min(end - offset, bytes);
1636 
1637 	bitmap_set_bits(ctl, info, offset, bytes_to_set);
1638 
1639 	return bytes_to_set;
1640 
1641 }
1642 
1643 static bool use_bitmap(struct btrfs_free_space_ctl *ctl,
1644 		      struct btrfs_free_space *info)
1645 {
1646 	struct btrfs_block_group_cache *block_group = ctl->private;
1647 
1648 	/*
1649 	 * If we are below the extents threshold then we can add this as an
1650 	 * extent, and don't have to deal with the bitmap
1651 	 */
1652 	if (ctl->free_extents < ctl->extents_thresh) {
1653 		/*
1654 		 * If this block group has some small extents we don't want to
1655 		 * use up all of our free slots in the cache with them, we want
1656 		 * to reserve them to larger extents, however if we have plent
1657 		 * of cache left then go ahead an dadd them, no sense in adding
1658 		 * the overhead of a bitmap if we don't have to.
1659 		 */
1660 		if (info->bytes <= block_group->sectorsize * 4) {
1661 			if (ctl->free_extents * 2 <= ctl->extents_thresh)
1662 				return false;
1663 		} else {
1664 			return false;
1665 		}
1666 	}
1667 
1668 	/*
1669 	 * The original block groups from mkfs can be really small, like 8
1670 	 * megabytes, so don't bother with a bitmap for those entries.  However
1671 	 * some block groups can be smaller than what a bitmap would cover but
1672 	 * are still large enough that they could overflow the 32k memory limit,
1673 	 * so allow those block groups to still be allowed to have a bitmap
1674 	 * entry.
1675 	 */
1676 	if (((BITS_PER_BITMAP * ctl->unit) >> 1) > block_group->key.offset)
1677 		return false;
1678 
1679 	return true;
1680 }
1681 
1682 static struct btrfs_free_space_op free_space_op = {
1683 	.recalc_thresholds	= recalculate_thresholds,
1684 	.use_bitmap		= use_bitmap,
1685 };
1686 
1687 static int insert_into_bitmap(struct btrfs_free_space_ctl *ctl,
1688 			      struct btrfs_free_space *info)
1689 {
1690 	struct btrfs_free_space *bitmap_info;
1691 	struct btrfs_block_group_cache *block_group = NULL;
1692 	int added = 0;
1693 	u64 bytes, offset, bytes_added;
1694 	int ret;
1695 
1696 	bytes = info->bytes;
1697 	offset = info->offset;
1698 
1699 	if (!ctl->op->use_bitmap(ctl, info))
1700 		return 0;
1701 
1702 	if (ctl->op == &free_space_op)
1703 		block_group = ctl->private;
1704 again:
1705 	/*
1706 	 * Since we link bitmaps right into the cluster we need to see if we
1707 	 * have a cluster here, and if so and it has our bitmap we need to add
1708 	 * the free space to that bitmap.
1709 	 */
1710 	if (block_group && !list_empty(&block_group->cluster_list)) {
1711 		struct btrfs_free_cluster *cluster;
1712 		struct rb_node *node;
1713 		struct btrfs_free_space *entry;
1714 
1715 		cluster = list_entry(block_group->cluster_list.next,
1716 				     struct btrfs_free_cluster,
1717 				     block_group_list);
1718 		spin_lock(&cluster->lock);
1719 		node = rb_first(&cluster->root);
1720 		if (!node) {
1721 			spin_unlock(&cluster->lock);
1722 			goto no_cluster_bitmap;
1723 		}
1724 
1725 		entry = rb_entry(node, struct btrfs_free_space, offset_index);
1726 		if (!entry->bitmap) {
1727 			spin_unlock(&cluster->lock);
1728 			goto no_cluster_bitmap;
1729 		}
1730 
1731 		if (entry->offset == offset_to_bitmap(ctl, offset)) {
1732 			bytes_added = add_bytes_to_bitmap(ctl, entry,
1733 							  offset, bytes);
1734 			bytes -= bytes_added;
1735 			offset += bytes_added;
1736 		}
1737 		spin_unlock(&cluster->lock);
1738 		if (!bytes) {
1739 			ret = 1;
1740 			goto out;
1741 		}
1742 	}
1743 
1744 no_cluster_bitmap:
1745 	bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
1746 					 1, 0);
1747 	if (!bitmap_info) {
1748 		BUG_ON(added);
1749 		goto new_bitmap;
1750 	}
1751 
1752 	bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes);
1753 	bytes -= bytes_added;
1754 	offset += bytes_added;
1755 	added = 0;
1756 
1757 	if (!bytes) {
1758 		ret = 1;
1759 		goto out;
1760 	} else
1761 		goto again;
1762 
1763 new_bitmap:
1764 	if (info && info->bitmap) {
1765 		add_new_bitmap(ctl, info, offset);
1766 		added = 1;
1767 		info = NULL;
1768 		goto again;
1769 	} else {
1770 		spin_unlock(&ctl->tree_lock);
1771 
1772 		/* no pre-allocated info, allocate a new one */
1773 		if (!info) {
1774 			info = kmem_cache_zalloc(btrfs_free_space_cachep,
1775 						 GFP_NOFS);
1776 			if (!info) {
1777 				spin_lock(&ctl->tree_lock);
1778 				ret = -ENOMEM;
1779 				goto out;
1780 			}
1781 		}
1782 
1783 		/* allocate the bitmap */
1784 		info->bitmap = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
1785 		spin_lock(&ctl->tree_lock);
1786 		if (!info->bitmap) {
1787 			ret = -ENOMEM;
1788 			goto out;
1789 		}
1790 		goto again;
1791 	}
1792 
1793 out:
1794 	if (info) {
1795 		if (info->bitmap)
1796 			kfree(info->bitmap);
1797 		kmem_cache_free(btrfs_free_space_cachep, info);
1798 	}
1799 
1800 	return ret;
1801 }
1802 
1803 static bool try_merge_free_space(struct btrfs_free_space_ctl *ctl,
1804 			  struct btrfs_free_space *info, bool update_stat)
1805 {
1806 	struct btrfs_free_space *left_info;
1807 	struct btrfs_free_space *right_info;
1808 	bool merged = false;
1809 	u64 offset = info->offset;
1810 	u64 bytes = info->bytes;
1811 
1812 	/*
1813 	 * first we want to see if there is free space adjacent to the range we
1814 	 * are adding, if there is remove that struct and add a new one to
1815 	 * cover the entire range
1816 	 */
1817 	right_info = tree_search_offset(ctl, offset + bytes, 0, 0);
1818 	if (right_info && rb_prev(&right_info->offset_index))
1819 		left_info = rb_entry(rb_prev(&right_info->offset_index),
1820 				     struct btrfs_free_space, offset_index);
1821 	else
1822 		left_info = tree_search_offset(ctl, offset - 1, 0, 0);
1823 
1824 	if (right_info && !right_info->bitmap) {
1825 		if (update_stat)
1826 			unlink_free_space(ctl, right_info);
1827 		else
1828 			__unlink_free_space(ctl, right_info);
1829 		info->bytes += right_info->bytes;
1830 		kmem_cache_free(btrfs_free_space_cachep, right_info);
1831 		merged = true;
1832 	}
1833 
1834 	if (left_info && !left_info->bitmap &&
1835 	    left_info->offset + left_info->bytes == offset) {
1836 		if (update_stat)
1837 			unlink_free_space(ctl, left_info);
1838 		else
1839 			__unlink_free_space(ctl, left_info);
1840 		info->offset = left_info->offset;
1841 		info->bytes += left_info->bytes;
1842 		kmem_cache_free(btrfs_free_space_cachep, left_info);
1843 		merged = true;
1844 	}
1845 
1846 	return merged;
1847 }
1848 
1849 int __btrfs_add_free_space(struct btrfs_free_space_ctl *ctl,
1850 			   u64 offset, u64 bytes)
1851 {
1852 	struct btrfs_free_space *info;
1853 	int ret = 0;
1854 
1855 	info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
1856 	if (!info)
1857 		return -ENOMEM;
1858 
1859 	info->offset = offset;
1860 	info->bytes = bytes;
1861 
1862 	spin_lock(&ctl->tree_lock);
1863 
1864 	if (try_merge_free_space(ctl, info, true))
1865 		goto link;
1866 
1867 	/*
1868 	 * There was no extent directly to the left or right of this new
1869 	 * extent then we know we're going to have to allocate a new extent, so
1870 	 * before we do that see if we need to drop this into a bitmap
1871 	 */
1872 	ret = insert_into_bitmap(ctl, info);
1873 	if (ret < 0) {
1874 		goto out;
1875 	} else if (ret) {
1876 		ret = 0;
1877 		goto out;
1878 	}
1879 link:
1880 	ret = link_free_space(ctl, info);
1881 	if (ret)
1882 		kmem_cache_free(btrfs_free_space_cachep, info);
1883 out:
1884 	spin_unlock(&ctl->tree_lock);
1885 
1886 	if (ret) {
1887 		printk(KERN_CRIT "btrfs: unable to add free space :%d\n", ret);
1888 		BUG_ON(ret == -EEXIST);
1889 	}
1890 
1891 	return ret;
1892 }
1893 
1894 int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
1895 			    u64 offset, u64 bytes)
1896 {
1897 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1898 	struct btrfs_free_space *info;
1899 	int ret;
1900 	bool re_search = false;
1901 
1902 	spin_lock(&ctl->tree_lock);
1903 
1904 again:
1905 	ret = 0;
1906 	if (!bytes)
1907 		goto out_lock;
1908 
1909 	info = tree_search_offset(ctl, offset, 0, 0);
1910 	if (!info) {
1911 		/*
1912 		 * oops didn't find an extent that matched the space we wanted
1913 		 * to remove, look for a bitmap instead
1914 		 */
1915 		info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
1916 					  1, 0);
1917 		if (!info) {
1918 			/*
1919 			 * If we found a partial bit of our free space in a
1920 			 * bitmap but then couldn't find the other part this may
1921 			 * be a problem, so WARN about it.
1922 			 */
1923 			WARN_ON(re_search);
1924 			goto out_lock;
1925 		}
1926 	}
1927 
1928 	re_search = false;
1929 	if (!info->bitmap) {
1930 		unlink_free_space(ctl, info);
1931 		if (offset == info->offset) {
1932 			u64 to_free = min(bytes, info->bytes);
1933 
1934 			info->bytes -= to_free;
1935 			info->offset += to_free;
1936 			if (info->bytes) {
1937 				ret = link_free_space(ctl, info);
1938 				WARN_ON(ret);
1939 			} else {
1940 				kmem_cache_free(btrfs_free_space_cachep, info);
1941 			}
1942 
1943 			offset += to_free;
1944 			bytes -= to_free;
1945 			goto again;
1946 		} else {
1947 			u64 old_end = info->bytes + info->offset;
1948 
1949 			info->bytes = offset - info->offset;
1950 			ret = link_free_space(ctl, info);
1951 			WARN_ON(ret);
1952 			if (ret)
1953 				goto out_lock;
1954 
1955 			/* Not enough bytes in this entry to satisfy us */
1956 			if (old_end < offset + bytes) {
1957 				bytes -= old_end - offset;
1958 				offset = old_end;
1959 				goto again;
1960 			} else if (old_end == offset + bytes) {
1961 				/* all done */
1962 				goto out_lock;
1963 			}
1964 			spin_unlock(&ctl->tree_lock);
1965 
1966 			ret = btrfs_add_free_space(block_group, offset + bytes,
1967 						   old_end - (offset + bytes));
1968 			WARN_ON(ret);
1969 			goto out;
1970 		}
1971 	}
1972 
1973 	ret = remove_from_bitmap(ctl, info, &offset, &bytes);
1974 	if (ret == -EAGAIN) {
1975 		re_search = true;
1976 		goto again;
1977 	}
1978 out_lock:
1979 	spin_unlock(&ctl->tree_lock);
1980 out:
1981 	return ret;
1982 }
1983 
1984 void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
1985 			   u64 bytes)
1986 {
1987 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1988 	struct btrfs_free_space *info;
1989 	struct rb_node *n;
1990 	int count = 0;
1991 
1992 	for (n = rb_first(&ctl->free_space_offset); n; n = rb_next(n)) {
1993 		info = rb_entry(n, struct btrfs_free_space, offset_index);
1994 		if (info->bytes >= bytes && !block_group->ro)
1995 			count++;
1996 		printk(KERN_CRIT "entry offset %llu, bytes %llu, bitmap %s\n",
1997 		       (unsigned long long)info->offset,
1998 		       (unsigned long long)info->bytes,
1999 		       (info->bitmap) ? "yes" : "no");
2000 	}
2001 	printk(KERN_INFO "block group has cluster?: %s\n",
2002 	       list_empty(&block_group->cluster_list) ? "no" : "yes");
2003 	printk(KERN_INFO "%d blocks of free space at or bigger than bytes is"
2004 	       "\n", count);
2005 }
2006 
2007 void btrfs_init_free_space_ctl(struct btrfs_block_group_cache *block_group)
2008 {
2009 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2010 
2011 	spin_lock_init(&ctl->tree_lock);
2012 	ctl->unit = block_group->sectorsize;
2013 	ctl->start = block_group->key.objectid;
2014 	ctl->private = block_group;
2015 	ctl->op = &free_space_op;
2016 
2017 	/*
2018 	 * we only want to have 32k of ram per block group for keeping
2019 	 * track of free space, and if we pass 1/2 of that we want to
2020 	 * start converting things over to using bitmaps
2021 	 */
2022 	ctl->extents_thresh = ((1024 * 32) / 2) /
2023 				sizeof(struct btrfs_free_space);
2024 }
2025 
2026 /*
2027  * for a given cluster, put all of its extents back into the free
2028  * space cache.  If the block group passed doesn't match the block group
2029  * pointed to by the cluster, someone else raced in and freed the
2030  * cluster already.  In that case, we just return without changing anything
2031  */
2032 static int
2033 __btrfs_return_cluster_to_free_space(
2034 			     struct btrfs_block_group_cache *block_group,
2035 			     struct btrfs_free_cluster *cluster)
2036 {
2037 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2038 	struct btrfs_free_space *entry;
2039 	struct rb_node *node;
2040 
2041 	spin_lock(&cluster->lock);
2042 	if (cluster->block_group != block_group)
2043 		goto out;
2044 
2045 	cluster->block_group = NULL;
2046 	cluster->window_start = 0;
2047 	list_del_init(&cluster->block_group_list);
2048 
2049 	node = rb_first(&cluster->root);
2050 	while (node) {
2051 		bool bitmap;
2052 
2053 		entry = rb_entry(node, struct btrfs_free_space, offset_index);
2054 		node = rb_next(&entry->offset_index);
2055 		rb_erase(&entry->offset_index, &cluster->root);
2056 
2057 		bitmap = (entry->bitmap != NULL);
2058 		if (!bitmap)
2059 			try_merge_free_space(ctl, entry, false);
2060 		tree_insert_offset(&ctl->free_space_offset,
2061 				   entry->offset, &entry->offset_index, bitmap);
2062 	}
2063 	cluster->root = RB_ROOT;
2064 
2065 out:
2066 	spin_unlock(&cluster->lock);
2067 	btrfs_put_block_group(block_group);
2068 	return 0;
2069 }
2070 
2071 static void __btrfs_remove_free_space_cache_locked(
2072 				struct btrfs_free_space_ctl *ctl)
2073 {
2074 	struct btrfs_free_space *info;
2075 	struct rb_node *node;
2076 
2077 	while ((node = rb_last(&ctl->free_space_offset)) != NULL) {
2078 		info = rb_entry(node, struct btrfs_free_space, offset_index);
2079 		if (!info->bitmap) {
2080 			unlink_free_space(ctl, info);
2081 			kmem_cache_free(btrfs_free_space_cachep, info);
2082 		} else {
2083 			free_bitmap(ctl, info);
2084 		}
2085 		if (need_resched()) {
2086 			spin_unlock(&ctl->tree_lock);
2087 			cond_resched();
2088 			spin_lock(&ctl->tree_lock);
2089 		}
2090 	}
2091 }
2092 
2093 void __btrfs_remove_free_space_cache(struct btrfs_free_space_ctl *ctl)
2094 {
2095 	spin_lock(&ctl->tree_lock);
2096 	__btrfs_remove_free_space_cache_locked(ctl);
2097 	spin_unlock(&ctl->tree_lock);
2098 }
2099 
2100 void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
2101 {
2102 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2103 	struct btrfs_free_cluster *cluster;
2104 	struct list_head *head;
2105 
2106 	spin_lock(&ctl->tree_lock);
2107 	while ((head = block_group->cluster_list.next) !=
2108 	       &block_group->cluster_list) {
2109 		cluster = list_entry(head, struct btrfs_free_cluster,
2110 				     block_group_list);
2111 
2112 		WARN_ON(cluster->block_group != block_group);
2113 		__btrfs_return_cluster_to_free_space(block_group, cluster);
2114 		if (need_resched()) {
2115 			spin_unlock(&ctl->tree_lock);
2116 			cond_resched();
2117 			spin_lock(&ctl->tree_lock);
2118 		}
2119 	}
2120 	__btrfs_remove_free_space_cache_locked(ctl);
2121 	spin_unlock(&ctl->tree_lock);
2122 
2123 }
2124 
2125 u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
2126 			       u64 offset, u64 bytes, u64 empty_size)
2127 {
2128 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2129 	struct btrfs_free_space *entry = NULL;
2130 	u64 bytes_search = bytes + empty_size;
2131 	u64 ret = 0;
2132 	u64 align_gap = 0;
2133 	u64 align_gap_len = 0;
2134 
2135 	spin_lock(&ctl->tree_lock);
2136 	entry = find_free_space(ctl, &offset, &bytes_search,
2137 				block_group->full_stripe_len);
2138 	if (!entry)
2139 		goto out;
2140 
2141 	ret = offset;
2142 	if (entry->bitmap) {
2143 		bitmap_clear_bits(ctl, entry, offset, bytes);
2144 		if (!entry->bytes)
2145 			free_bitmap(ctl, entry);
2146 	} else {
2147 
2148 		unlink_free_space(ctl, entry);
2149 		align_gap_len = offset - entry->offset;
2150 		align_gap = entry->offset;
2151 
2152 		entry->offset = offset + bytes;
2153 		WARN_ON(entry->bytes < bytes + align_gap_len);
2154 
2155 		entry->bytes -= bytes + align_gap_len;
2156 		if (!entry->bytes)
2157 			kmem_cache_free(btrfs_free_space_cachep, entry);
2158 		else
2159 			link_free_space(ctl, entry);
2160 	}
2161 
2162 out:
2163 	spin_unlock(&ctl->tree_lock);
2164 
2165 	if (align_gap_len)
2166 		__btrfs_add_free_space(ctl, align_gap, align_gap_len);
2167 	return ret;
2168 }
2169 
2170 /*
2171  * given a cluster, put all of its extents back into the free space
2172  * cache.  If a block group is passed, this function will only free
2173  * a cluster that belongs to the passed block group.
2174  *
2175  * Otherwise, it'll get a reference on the block group pointed to by the
2176  * cluster and remove the cluster from it.
2177  */
2178 int btrfs_return_cluster_to_free_space(
2179 			       struct btrfs_block_group_cache *block_group,
2180 			       struct btrfs_free_cluster *cluster)
2181 {
2182 	struct btrfs_free_space_ctl *ctl;
2183 	int ret;
2184 
2185 	/* first, get a safe pointer to the block group */
2186 	spin_lock(&cluster->lock);
2187 	if (!block_group) {
2188 		block_group = cluster->block_group;
2189 		if (!block_group) {
2190 			spin_unlock(&cluster->lock);
2191 			return 0;
2192 		}
2193 	} else if (cluster->block_group != block_group) {
2194 		/* someone else has already freed it don't redo their work */
2195 		spin_unlock(&cluster->lock);
2196 		return 0;
2197 	}
2198 	atomic_inc(&block_group->count);
2199 	spin_unlock(&cluster->lock);
2200 
2201 	ctl = block_group->free_space_ctl;
2202 
2203 	/* now return any extents the cluster had on it */
2204 	spin_lock(&ctl->tree_lock);
2205 	ret = __btrfs_return_cluster_to_free_space(block_group, cluster);
2206 	spin_unlock(&ctl->tree_lock);
2207 
2208 	/* finally drop our ref */
2209 	btrfs_put_block_group(block_group);
2210 	return ret;
2211 }
2212 
2213 static u64 btrfs_alloc_from_bitmap(struct btrfs_block_group_cache *block_group,
2214 				   struct btrfs_free_cluster *cluster,
2215 				   struct btrfs_free_space *entry,
2216 				   u64 bytes, u64 min_start)
2217 {
2218 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2219 	int err;
2220 	u64 search_start = cluster->window_start;
2221 	u64 search_bytes = bytes;
2222 	u64 ret = 0;
2223 
2224 	search_start = min_start;
2225 	search_bytes = bytes;
2226 
2227 	err = search_bitmap(ctl, entry, &search_start, &search_bytes);
2228 	if (err)
2229 		return 0;
2230 
2231 	ret = search_start;
2232 	__bitmap_clear_bits(ctl, entry, ret, bytes);
2233 
2234 	return ret;
2235 }
2236 
2237 /*
2238  * given a cluster, try to allocate 'bytes' from it, returns 0
2239  * if it couldn't find anything suitably large, or a logical disk offset
2240  * if things worked out
2241  */
2242 u64 btrfs_alloc_from_cluster(struct btrfs_block_group_cache *block_group,
2243 			     struct btrfs_free_cluster *cluster, u64 bytes,
2244 			     u64 min_start)
2245 {
2246 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2247 	struct btrfs_free_space *entry = NULL;
2248 	struct rb_node *node;
2249 	u64 ret = 0;
2250 
2251 	spin_lock(&cluster->lock);
2252 	if (bytes > cluster->max_size)
2253 		goto out;
2254 
2255 	if (cluster->block_group != block_group)
2256 		goto out;
2257 
2258 	node = rb_first(&cluster->root);
2259 	if (!node)
2260 		goto out;
2261 
2262 	entry = rb_entry(node, struct btrfs_free_space, offset_index);
2263 	while(1) {
2264 		if (entry->bytes < bytes ||
2265 		    (!entry->bitmap && entry->offset < min_start)) {
2266 			node = rb_next(&entry->offset_index);
2267 			if (!node)
2268 				break;
2269 			entry = rb_entry(node, struct btrfs_free_space,
2270 					 offset_index);
2271 			continue;
2272 		}
2273 
2274 		if (entry->bitmap) {
2275 			ret = btrfs_alloc_from_bitmap(block_group,
2276 						      cluster, entry, bytes,
2277 						      cluster->window_start);
2278 			if (ret == 0) {
2279 				node = rb_next(&entry->offset_index);
2280 				if (!node)
2281 					break;
2282 				entry = rb_entry(node, struct btrfs_free_space,
2283 						 offset_index);
2284 				continue;
2285 			}
2286 			cluster->window_start += bytes;
2287 		} else {
2288 			ret = entry->offset;
2289 
2290 			entry->offset += bytes;
2291 			entry->bytes -= bytes;
2292 		}
2293 
2294 		if (entry->bytes == 0)
2295 			rb_erase(&entry->offset_index, &cluster->root);
2296 		break;
2297 	}
2298 out:
2299 	spin_unlock(&cluster->lock);
2300 
2301 	if (!ret)
2302 		return 0;
2303 
2304 	spin_lock(&ctl->tree_lock);
2305 
2306 	ctl->free_space -= bytes;
2307 	if (entry->bytes == 0) {
2308 		ctl->free_extents--;
2309 		if (entry->bitmap) {
2310 			kfree(entry->bitmap);
2311 			ctl->total_bitmaps--;
2312 			ctl->op->recalc_thresholds(ctl);
2313 		}
2314 		kmem_cache_free(btrfs_free_space_cachep, entry);
2315 	}
2316 
2317 	spin_unlock(&ctl->tree_lock);
2318 
2319 	return ret;
2320 }
2321 
2322 static int btrfs_bitmap_cluster(struct btrfs_block_group_cache *block_group,
2323 				struct btrfs_free_space *entry,
2324 				struct btrfs_free_cluster *cluster,
2325 				u64 offset, u64 bytes,
2326 				u64 cont1_bytes, u64 min_bytes)
2327 {
2328 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2329 	unsigned long next_zero;
2330 	unsigned long i;
2331 	unsigned long want_bits;
2332 	unsigned long min_bits;
2333 	unsigned long found_bits;
2334 	unsigned long start = 0;
2335 	unsigned long total_found = 0;
2336 	int ret;
2337 
2338 	i = offset_to_bit(entry->offset, ctl->unit,
2339 			  max_t(u64, offset, entry->offset));
2340 	want_bits = bytes_to_bits(bytes, ctl->unit);
2341 	min_bits = bytes_to_bits(min_bytes, ctl->unit);
2342 
2343 again:
2344 	found_bits = 0;
2345 	for_each_set_bit_from(i, entry->bitmap, BITS_PER_BITMAP) {
2346 		next_zero = find_next_zero_bit(entry->bitmap,
2347 					       BITS_PER_BITMAP, i);
2348 		if (next_zero - i >= min_bits) {
2349 			found_bits = next_zero - i;
2350 			break;
2351 		}
2352 		i = next_zero;
2353 	}
2354 
2355 	if (!found_bits)
2356 		return -ENOSPC;
2357 
2358 	if (!total_found) {
2359 		start = i;
2360 		cluster->max_size = 0;
2361 	}
2362 
2363 	total_found += found_bits;
2364 
2365 	if (cluster->max_size < found_bits * ctl->unit)
2366 		cluster->max_size = found_bits * ctl->unit;
2367 
2368 	if (total_found < want_bits || cluster->max_size < cont1_bytes) {
2369 		i = next_zero + 1;
2370 		goto again;
2371 	}
2372 
2373 	cluster->window_start = start * ctl->unit + entry->offset;
2374 	rb_erase(&entry->offset_index, &ctl->free_space_offset);
2375 	ret = tree_insert_offset(&cluster->root, entry->offset,
2376 				 &entry->offset_index, 1);
2377 	BUG_ON(ret); /* -EEXIST; Logic error */
2378 
2379 	trace_btrfs_setup_cluster(block_group, cluster,
2380 				  total_found * ctl->unit, 1);
2381 	return 0;
2382 }
2383 
2384 /*
2385  * This searches the block group for just extents to fill the cluster with.
2386  * Try to find a cluster with at least bytes total bytes, at least one
2387  * extent of cont1_bytes, and other clusters of at least min_bytes.
2388  */
2389 static noinline int
2390 setup_cluster_no_bitmap(struct btrfs_block_group_cache *block_group,
2391 			struct btrfs_free_cluster *cluster,
2392 			struct list_head *bitmaps, u64 offset, u64 bytes,
2393 			u64 cont1_bytes, u64 min_bytes)
2394 {
2395 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2396 	struct btrfs_free_space *first = NULL;
2397 	struct btrfs_free_space *entry = NULL;
2398 	struct btrfs_free_space *last;
2399 	struct rb_node *node;
2400 	u64 window_start;
2401 	u64 window_free;
2402 	u64 max_extent;
2403 	u64 total_size = 0;
2404 
2405 	entry = tree_search_offset(ctl, offset, 0, 1);
2406 	if (!entry)
2407 		return -ENOSPC;
2408 
2409 	/*
2410 	 * We don't want bitmaps, so just move along until we find a normal
2411 	 * extent entry.
2412 	 */
2413 	while (entry->bitmap || entry->bytes < min_bytes) {
2414 		if (entry->bitmap && list_empty(&entry->list))
2415 			list_add_tail(&entry->list, bitmaps);
2416 		node = rb_next(&entry->offset_index);
2417 		if (!node)
2418 			return -ENOSPC;
2419 		entry = rb_entry(node, struct btrfs_free_space, offset_index);
2420 	}
2421 
2422 	window_start = entry->offset;
2423 	window_free = entry->bytes;
2424 	max_extent = entry->bytes;
2425 	first = entry;
2426 	last = entry;
2427 
2428 	for (node = rb_next(&entry->offset_index); node;
2429 	     node = rb_next(&entry->offset_index)) {
2430 		entry = rb_entry(node, struct btrfs_free_space, offset_index);
2431 
2432 		if (entry->bitmap) {
2433 			if (list_empty(&entry->list))
2434 				list_add_tail(&entry->list, bitmaps);
2435 			continue;
2436 		}
2437 
2438 		if (entry->bytes < min_bytes)
2439 			continue;
2440 
2441 		last = entry;
2442 		window_free += entry->bytes;
2443 		if (entry->bytes > max_extent)
2444 			max_extent = entry->bytes;
2445 	}
2446 
2447 	if (window_free < bytes || max_extent < cont1_bytes)
2448 		return -ENOSPC;
2449 
2450 	cluster->window_start = first->offset;
2451 
2452 	node = &first->offset_index;
2453 
2454 	/*
2455 	 * now we've found our entries, pull them out of the free space
2456 	 * cache and put them into the cluster rbtree
2457 	 */
2458 	do {
2459 		int ret;
2460 
2461 		entry = rb_entry(node, struct btrfs_free_space, offset_index);
2462 		node = rb_next(&entry->offset_index);
2463 		if (entry->bitmap || entry->bytes < min_bytes)
2464 			continue;
2465 
2466 		rb_erase(&entry->offset_index, &ctl->free_space_offset);
2467 		ret = tree_insert_offset(&cluster->root, entry->offset,
2468 					 &entry->offset_index, 0);
2469 		total_size += entry->bytes;
2470 		BUG_ON(ret); /* -EEXIST; Logic error */
2471 	} while (node && entry != last);
2472 
2473 	cluster->max_size = max_extent;
2474 	trace_btrfs_setup_cluster(block_group, cluster, total_size, 0);
2475 	return 0;
2476 }
2477 
2478 /*
2479  * This specifically looks for bitmaps that may work in the cluster, we assume
2480  * that we have already failed to find extents that will work.
2481  */
2482 static noinline int
2483 setup_cluster_bitmap(struct btrfs_block_group_cache *block_group,
2484 		     struct btrfs_free_cluster *cluster,
2485 		     struct list_head *bitmaps, u64 offset, u64 bytes,
2486 		     u64 cont1_bytes, u64 min_bytes)
2487 {
2488 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2489 	struct btrfs_free_space *entry;
2490 	int ret = -ENOSPC;
2491 	u64 bitmap_offset = offset_to_bitmap(ctl, offset);
2492 
2493 	if (ctl->total_bitmaps == 0)
2494 		return -ENOSPC;
2495 
2496 	/*
2497 	 * The bitmap that covers offset won't be in the list unless offset
2498 	 * is just its start offset.
2499 	 */
2500 	entry = list_first_entry(bitmaps, struct btrfs_free_space, list);
2501 	if (entry->offset != bitmap_offset) {
2502 		entry = tree_search_offset(ctl, bitmap_offset, 1, 0);
2503 		if (entry && list_empty(&entry->list))
2504 			list_add(&entry->list, bitmaps);
2505 	}
2506 
2507 	list_for_each_entry(entry, bitmaps, list) {
2508 		if (entry->bytes < bytes)
2509 			continue;
2510 		ret = btrfs_bitmap_cluster(block_group, entry, cluster, offset,
2511 					   bytes, cont1_bytes, min_bytes);
2512 		if (!ret)
2513 			return 0;
2514 	}
2515 
2516 	/*
2517 	 * The bitmaps list has all the bitmaps that record free space
2518 	 * starting after offset, so no more search is required.
2519 	 */
2520 	return -ENOSPC;
2521 }
2522 
2523 /*
2524  * here we try to find a cluster of blocks in a block group.  The goal
2525  * is to find at least bytes+empty_size.
2526  * We might not find them all in one contiguous area.
2527  *
2528  * returns zero and sets up cluster if things worked out, otherwise
2529  * it returns -enospc
2530  */
2531 int btrfs_find_space_cluster(struct btrfs_trans_handle *trans,
2532 			     struct btrfs_root *root,
2533 			     struct btrfs_block_group_cache *block_group,
2534 			     struct btrfs_free_cluster *cluster,
2535 			     u64 offset, u64 bytes, u64 empty_size)
2536 {
2537 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2538 	struct btrfs_free_space *entry, *tmp;
2539 	LIST_HEAD(bitmaps);
2540 	u64 min_bytes;
2541 	u64 cont1_bytes;
2542 	int ret;
2543 
2544 	/*
2545 	 * Choose the minimum extent size we'll require for this
2546 	 * cluster.  For SSD_SPREAD, don't allow any fragmentation.
2547 	 * For metadata, allow allocates with smaller extents.  For
2548 	 * data, keep it dense.
2549 	 */
2550 	if (btrfs_test_opt(root, SSD_SPREAD)) {
2551 		cont1_bytes = min_bytes = bytes + empty_size;
2552 	} else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) {
2553 		cont1_bytes = bytes;
2554 		min_bytes = block_group->sectorsize;
2555 	} else {
2556 		cont1_bytes = max(bytes, (bytes + empty_size) >> 2);
2557 		min_bytes = block_group->sectorsize;
2558 	}
2559 
2560 	spin_lock(&ctl->tree_lock);
2561 
2562 	/*
2563 	 * If we know we don't have enough space to make a cluster don't even
2564 	 * bother doing all the work to try and find one.
2565 	 */
2566 	if (ctl->free_space < bytes) {
2567 		spin_unlock(&ctl->tree_lock);
2568 		return -ENOSPC;
2569 	}
2570 
2571 	spin_lock(&cluster->lock);
2572 
2573 	/* someone already found a cluster, hooray */
2574 	if (cluster->block_group) {
2575 		ret = 0;
2576 		goto out;
2577 	}
2578 
2579 	trace_btrfs_find_cluster(block_group, offset, bytes, empty_size,
2580 				 min_bytes);
2581 
2582 	INIT_LIST_HEAD(&bitmaps);
2583 	ret = setup_cluster_no_bitmap(block_group, cluster, &bitmaps, offset,
2584 				      bytes + empty_size,
2585 				      cont1_bytes, min_bytes);
2586 	if (ret)
2587 		ret = setup_cluster_bitmap(block_group, cluster, &bitmaps,
2588 					   offset, bytes + empty_size,
2589 					   cont1_bytes, min_bytes);
2590 
2591 	/* Clear our temporary list */
2592 	list_for_each_entry_safe(entry, tmp, &bitmaps, list)
2593 		list_del_init(&entry->list);
2594 
2595 	if (!ret) {
2596 		atomic_inc(&block_group->count);
2597 		list_add_tail(&cluster->block_group_list,
2598 			      &block_group->cluster_list);
2599 		cluster->block_group = block_group;
2600 	} else {
2601 		trace_btrfs_failed_cluster_setup(block_group);
2602 	}
2603 out:
2604 	spin_unlock(&cluster->lock);
2605 	spin_unlock(&ctl->tree_lock);
2606 
2607 	return ret;
2608 }
2609 
2610 /*
2611  * simple code to zero out a cluster
2612  */
2613 void btrfs_init_free_cluster(struct btrfs_free_cluster *cluster)
2614 {
2615 	spin_lock_init(&cluster->lock);
2616 	spin_lock_init(&cluster->refill_lock);
2617 	cluster->root = RB_ROOT;
2618 	cluster->max_size = 0;
2619 	INIT_LIST_HEAD(&cluster->block_group_list);
2620 	cluster->block_group = NULL;
2621 }
2622 
2623 static int do_trimming(struct btrfs_block_group_cache *block_group,
2624 		       u64 *total_trimmed, u64 start, u64 bytes,
2625 		       u64 reserved_start, u64 reserved_bytes)
2626 {
2627 	struct btrfs_space_info *space_info = block_group->space_info;
2628 	struct btrfs_fs_info *fs_info = block_group->fs_info;
2629 	int ret;
2630 	int update = 0;
2631 	u64 trimmed = 0;
2632 
2633 	spin_lock(&space_info->lock);
2634 	spin_lock(&block_group->lock);
2635 	if (!block_group->ro) {
2636 		block_group->reserved += reserved_bytes;
2637 		space_info->bytes_reserved += reserved_bytes;
2638 		update = 1;
2639 	}
2640 	spin_unlock(&block_group->lock);
2641 	spin_unlock(&space_info->lock);
2642 
2643 	ret = btrfs_error_discard_extent(fs_info->extent_root,
2644 					 start, bytes, &trimmed);
2645 	if (!ret)
2646 		*total_trimmed += trimmed;
2647 
2648 	btrfs_add_free_space(block_group, reserved_start, reserved_bytes);
2649 
2650 	if (update) {
2651 		spin_lock(&space_info->lock);
2652 		spin_lock(&block_group->lock);
2653 		if (block_group->ro)
2654 			space_info->bytes_readonly += reserved_bytes;
2655 		block_group->reserved -= reserved_bytes;
2656 		space_info->bytes_reserved -= reserved_bytes;
2657 		spin_unlock(&space_info->lock);
2658 		spin_unlock(&block_group->lock);
2659 	}
2660 
2661 	return ret;
2662 }
2663 
2664 static int trim_no_bitmap(struct btrfs_block_group_cache *block_group,
2665 			  u64 *total_trimmed, u64 start, u64 end, u64 minlen)
2666 {
2667 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2668 	struct btrfs_free_space *entry;
2669 	struct rb_node *node;
2670 	int ret = 0;
2671 	u64 extent_start;
2672 	u64 extent_bytes;
2673 	u64 bytes;
2674 
2675 	while (start < end) {
2676 		spin_lock(&ctl->tree_lock);
2677 
2678 		if (ctl->free_space < minlen) {
2679 			spin_unlock(&ctl->tree_lock);
2680 			break;
2681 		}
2682 
2683 		entry = tree_search_offset(ctl, start, 0, 1);
2684 		if (!entry) {
2685 			spin_unlock(&ctl->tree_lock);
2686 			break;
2687 		}
2688 
2689 		/* skip bitmaps */
2690 		while (entry->bitmap) {
2691 			node = rb_next(&entry->offset_index);
2692 			if (!node) {
2693 				spin_unlock(&ctl->tree_lock);
2694 				goto out;
2695 			}
2696 			entry = rb_entry(node, struct btrfs_free_space,
2697 					 offset_index);
2698 		}
2699 
2700 		if (entry->offset >= end) {
2701 			spin_unlock(&ctl->tree_lock);
2702 			break;
2703 		}
2704 
2705 		extent_start = entry->offset;
2706 		extent_bytes = entry->bytes;
2707 		start = max(start, extent_start);
2708 		bytes = min(extent_start + extent_bytes, end) - start;
2709 		if (bytes < minlen) {
2710 			spin_unlock(&ctl->tree_lock);
2711 			goto next;
2712 		}
2713 
2714 		unlink_free_space(ctl, entry);
2715 		kmem_cache_free(btrfs_free_space_cachep, entry);
2716 
2717 		spin_unlock(&ctl->tree_lock);
2718 
2719 		ret = do_trimming(block_group, total_trimmed, start, bytes,
2720 				  extent_start, extent_bytes);
2721 		if (ret)
2722 			break;
2723 next:
2724 		start += bytes;
2725 
2726 		if (fatal_signal_pending(current)) {
2727 			ret = -ERESTARTSYS;
2728 			break;
2729 		}
2730 
2731 		cond_resched();
2732 	}
2733 out:
2734 	return ret;
2735 }
2736 
2737 static int trim_bitmaps(struct btrfs_block_group_cache *block_group,
2738 			u64 *total_trimmed, u64 start, u64 end, u64 minlen)
2739 {
2740 	struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
2741 	struct btrfs_free_space *entry;
2742 	int ret = 0;
2743 	int ret2;
2744 	u64 bytes;
2745 	u64 offset = offset_to_bitmap(ctl, start);
2746 
2747 	while (offset < end) {
2748 		bool next_bitmap = false;
2749 
2750 		spin_lock(&ctl->tree_lock);
2751 
2752 		if (ctl->free_space < minlen) {
2753 			spin_unlock(&ctl->tree_lock);
2754 			break;
2755 		}
2756 
2757 		entry = tree_search_offset(ctl, offset, 1, 0);
2758 		if (!entry) {
2759 			spin_unlock(&ctl->tree_lock);
2760 			next_bitmap = true;
2761 			goto next;
2762 		}
2763 
2764 		bytes = minlen;
2765 		ret2 = search_bitmap(ctl, entry, &start, &bytes);
2766 		if (ret2 || start >= end) {
2767 			spin_unlock(&ctl->tree_lock);
2768 			next_bitmap = true;
2769 			goto next;
2770 		}
2771 
2772 		bytes = min(bytes, end - start);
2773 		if (bytes < minlen) {
2774 			spin_unlock(&ctl->tree_lock);
2775 			goto next;
2776 		}
2777 
2778 		bitmap_clear_bits(ctl, entry, start, bytes);
2779 		if (entry->bytes == 0)
2780 			free_bitmap(ctl, entry);
2781 
2782 		spin_unlock(&ctl->tree_lock);
2783 
2784 		ret = do_trimming(block_group, total_trimmed, start, bytes,
2785 				  start, bytes);
2786 		if (ret)
2787 			break;
2788 next:
2789 		if (next_bitmap) {
2790 			offset += BITS_PER_BITMAP * ctl->unit;
2791 		} else {
2792 			start += bytes;
2793 			if (start >= offset + BITS_PER_BITMAP * ctl->unit)
2794 				offset += BITS_PER_BITMAP * ctl->unit;
2795 		}
2796 
2797 		if (fatal_signal_pending(current)) {
2798 			ret = -ERESTARTSYS;
2799 			break;
2800 		}
2801 
2802 		cond_resched();
2803 	}
2804 
2805 	return ret;
2806 }
2807 
2808 int btrfs_trim_block_group(struct btrfs_block_group_cache *block_group,
2809 			   u64 *trimmed, u64 start, u64 end, u64 minlen)
2810 {
2811 	int ret;
2812 
2813 	*trimmed = 0;
2814 
2815 	ret = trim_no_bitmap(block_group, trimmed, start, end, minlen);
2816 	if (ret)
2817 		return ret;
2818 
2819 	ret = trim_bitmaps(block_group, trimmed, start, end, minlen);
2820 
2821 	return ret;
2822 }
2823 
2824 /*
2825  * Find the left-most item in the cache tree, and then return the
2826  * smallest inode number in the item.
2827  *
2828  * Note: the returned inode number may not be the smallest one in
2829  * the tree, if the left-most item is a bitmap.
2830  */
2831 u64 btrfs_find_ino_for_alloc(struct btrfs_root *fs_root)
2832 {
2833 	struct btrfs_free_space_ctl *ctl = fs_root->free_ino_ctl;
2834 	struct btrfs_free_space *entry = NULL;
2835 	u64 ino = 0;
2836 
2837 	spin_lock(&ctl->tree_lock);
2838 
2839 	if (RB_EMPTY_ROOT(&ctl->free_space_offset))
2840 		goto out;
2841 
2842 	entry = rb_entry(rb_first(&ctl->free_space_offset),
2843 			 struct btrfs_free_space, offset_index);
2844 
2845 	if (!entry->bitmap) {
2846 		ino = entry->offset;
2847 
2848 		unlink_free_space(ctl, entry);
2849 		entry->offset++;
2850 		entry->bytes--;
2851 		if (!entry->bytes)
2852 			kmem_cache_free(btrfs_free_space_cachep, entry);
2853 		else
2854 			link_free_space(ctl, entry);
2855 	} else {
2856 		u64 offset = 0;
2857 		u64 count = 1;
2858 		int ret;
2859 
2860 		ret = search_bitmap(ctl, entry, &offset, &count);
2861 		/* Logic error; Should be empty if it can't find anything */
2862 		BUG_ON(ret);
2863 
2864 		ino = offset;
2865 		bitmap_clear_bits(ctl, entry, offset, 1);
2866 		if (entry->bytes == 0)
2867 			free_bitmap(ctl, entry);
2868 	}
2869 out:
2870 	spin_unlock(&ctl->tree_lock);
2871 
2872 	return ino;
2873 }
2874 
2875 struct inode *lookup_free_ino_inode(struct btrfs_root *root,
2876 				    struct btrfs_path *path)
2877 {
2878 	struct inode *inode = NULL;
2879 
2880 	spin_lock(&root->cache_lock);
2881 	if (root->cache_inode)
2882 		inode = igrab(root->cache_inode);
2883 	spin_unlock(&root->cache_lock);
2884 	if (inode)
2885 		return inode;
2886 
2887 	inode = __lookup_free_space_inode(root, path, 0);
2888 	if (IS_ERR(inode))
2889 		return inode;
2890 
2891 	spin_lock(&root->cache_lock);
2892 	if (!btrfs_fs_closing(root->fs_info))
2893 		root->cache_inode = igrab(inode);
2894 	spin_unlock(&root->cache_lock);
2895 
2896 	return inode;
2897 }
2898 
2899 int create_free_ino_inode(struct btrfs_root *root,
2900 			  struct btrfs_trans_handle *trans,
2901 			  struct btrfs_path *path)
2902 {
2903 	return __create_free_space_inode(root, trans, path,
2904 					 BTRFS_FREE_INO_OBJECTID, 0);
2905 }
2906 
2907 int load_free_ino_cache(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
2908 {
2909 	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
2910 	struct btrfs_path *path;
2911 	struct inode *inode;
2912 	int ret = 0;
2913 	u64 root_gen = btrfs_root_generation(&root->root_item);
2914 
2915 	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
2916 		return 0;
2917 
2918 	/*
2919 	 * If we're unmounting then just return, since this does a search on the
2920 	 * normal root and not the commit root and we could deadlock.
2921 	 */
2922 	if (btrfs_fs_closing(fs_info))
2923 		return 0;
2924 
2925 	path = btrfs_alloc_path();
2926 	if (!path)
2927 		return 0;
2928 
2929 	inode = lookup_free_ino_inode(root, path);
2930 	if (IS_ERR(inode))
2931 		goto out;
2932 
2933 	if (root_gen != BTRFS_I(inode)->generation)
2934 		goto out_put;
2935 
2936 	ret = __load_free_space_cache(root, inode, ctl, path, 0);
2937 
2938 	if (ret < 0)
2939 		btrfs_err(fs_info,
2940 			"failed to load free ino cache for root %llu",
2941 			root->root_key.objectid);
2942 out_put:
2943 	iput(inode);
2944 out:
2945 	btrfs_free_path(path);
2946 	return ret;
2947 }
2948 
2949 int btrfs_write_out_ino_cache(struct btrfs_root *root,
2950 			      struct btrfs_trans_handle *trans,
2951 			      struct btrfs_path *path)
2952 {
2953 	struct btrfs_free_space_ctl *ctl = root->free_ino_ctl;
2954 	struct inode *inode;
2955 	int ret;
2956 
2957 	if (!btrfs_test_opt(root, INODE_MAP_CACHE))
2958 		return 0;
2959 
2960 	inode = lookup_free_ino_inode(root, path);
2961 	if (IS_ERR(inode))
2962 		return 0;
2963 
2964 	ret = __btrfs_write_out_cache(root, inode, ctl, NULL, trans, path, 0);
2965 	if (ret) {
2966 		btrfs_delalloc_release_metadata(inode, inode->i_size);
2967 #ifdef DEBUG
2968 		btrfs_err(root->fs_info,
2969 			"failed to write free ino cache for root %llu",
2970 			root->root_key.objectid);
2971 #endif
2972 	}
2973 
2974 	iput(inode);
2975 	return ret;
2976 }
2977 
2978 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
2979 static struct btrfs_block_group_cache *init_test_block_group(void)
2980 {
2981 	struct btrfs_block_group_cache *cache;
2982 
2983 	cache = kzalloc(sizeof(*cache), GFP_NOFS);
2984 	if (!cache)
2985 		return NULL;
2986 	cache->free_space_ctl = kzalloc(sizeof(*cache->free_space_ctl),
2987 					GFP_NOFS);
2988 	if (!cache->free_space_ctl) {
2989 		kfree(cache);
2990 		return NULL;
2991 	}
2992 
2993 	cache->key.objectid = 0;
2994 	cache->key.offset = 1024 * 1024 * 1024;
2995 	cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
2996 	cache->sectorsize = 4096;
2997 
2998 	spin_lock_init(&cache->lock);
2999 	INIT_LIST_HEAD(&cache->list);
3000 	INIT_LIST_HEAD(&cache->cluster_list);
3001 	INIT_LIST_HEAD(&cache->new_bg_list);
3002 
3003 	btrfs_init_free_space_ctl(cache);
3004 
3005 	return cache;
3006 }
3007 
3008 /*
3009  * Checks to see if the given range is in the free space cache.  This is really
3010  * just used to check the absence of space, so if there is free space in the
3011  * range at all we will return 1.
3012  */
3013 static int check_exists(struct btrfs_block_group_cache *cache, u64 offset,
3014 			u64 bytes)
3015 {
3016 	struct btrfs_free_space_ctl *ctl = cache->free_space_ctl;
3017 	struct btrfs_free_space *info;
3018 	int ret = 0;
3019 
3020 	spin_lock(&ctl->tree_lock);
3021 	info = tree_search_offset(ctl, offset, 0, 0);
3022 	if (!info) {
3023 		info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
3024 					  1, 0);
3025 		if (!info)
3026 			goto out;
3027 	}
3028 
3029 have_info:
3030 	if (info->bitmap) {
3031 		u64 bit_off, bit_bytes;
3032 		struct rb_node *n;
3033 		struct btrfs_free_space *tmp;
3034 
3035 		bit_off = offset;
3036 		bit_bytes = ctl->unit;
3037 		ret = search_bitmap(ctl, info, &bit_off, &bit_bytes);
3038 		if (!ret) {
3039 			if (bit_off == offset) {
3040 				ret = 1;
3041 				goto out;
3042 			} else if (bit_off > offset &&
3043 				   offset + bytes > bit_off) {
3044 				ret = 1;
3045 				goto out;
3046 			}
3047 		}
3048 
3049 		n = rb_prev(&info->offset_index);
3050 		while (n) {
3051 			tmp = rb_entry(n, struct btrfs_free_space,
3052 				       offset_index);
3053 			if (tmp->offset + tmp->bytes < offset)
3054 				break;
3055 			if (offset + bytes < tmp->offset) {
3056 				n = rb_prev(&info->offset_index);
3057 				continue;
3058 			}
3059 			info = tmp;
3060 			goto have_info;
3061 		}
3062 
3063 		n = rb_next(&info->offset_index);
3064 		while (n) {
3065 			tmp = rb_entry(n, struct btrfs_free_space,
3066 				       offset_index);
3067 			if (offset + bytes < tmp->offset)
3068 				break;
3069 			if (tmp->offset + tmp->bytes < offset) {
3070 				n = rb_next(&info->offset_index);
3071 				continue;
3072 			}
3073 			info = tmp;
3074 			goto have_info;
3075 		}
3076 
3077 		goto out;
3078 	}
3079 
3080 	if (info->offset == offset) {
3081 		ret = 1;
3082 		goto out;
3083 	}
3084 
3085 	if (offset > info->offset && offset < info->offset + info->bytes)
3086 		ret = 1;
3087 out:
3088 	spin_unlock(&ctl->tree_lock);
3089 	return ret;
3090 }
3091 
3092 /*
3093  * Use this if you need to make a bitmap or extent entry specifically, it
3094  * doesn't do any of the merging that add_free_space does, this acts a lot like
3095  * how the free space cache loading stuff works, so you can get really weird
3096  * configurations.
3097  */
3098 static int add_free_space_entry(struct btrfs_block_group_cache *cache,
3099 				u64 offset, u64 bytes, bool bitmap)
3100 {
3101 	struct btrfs_free_space_ctl *ctl = cache->free_space_ctl;
3102 	struct btrfs_free_space *info = NULL, *bitmap_info;
3103 	void *map = NULL;
3104 	u64 bytes_added;
3105 	int ret;
3106 
3107 again:
3108 	if (!info) {
3109 		info = kmem_cache_zalloc(btrfs_free_space_cachep, GFP_NOFS);
3110 		if (!info)
3111 			return -ENOMEM;
3112 	}
3113 
3114 	if (!bitmap) {
3115 		spin_lock(&ctl->tree_lock);
3116 		info->offset = offset;
3117 		info->bytes = bytes;
3118 		ret = link_free_space(ctl, info);
3119 		spin_unlock(&ctl->tree_lock);
3120 		if (ret)
3121 			kmem_cache_free(btrfs_free_space_cachep, info);
3122 		return ret;
3123 	}
3124 
3125 	if (!map) {
3126 		map = kzalloc(PAGE_CACHE_SIZE, GFP_NOFS);
3127 		if (!map) {
3128 			kmem_cache_free(btrfs_free_space_cachep, info);
3129 			return -ENOMEM;
3130 		}
3131 	}
3132 
3133 	spin_lock(&ctl->tree_lock);
3134 	bitmap_info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
3135 					 1, 0);
3136 	if (!bitmap_info) {
3137 		info->bitmap = map;
3138 		map = NULL;
3139 		add_new_bitmap(ctl, info, offset);
3140 		bitmap_info = info;
3141 	}
3142 
3143 	bytes_added = add_bytes_to_bitmap(ctl, bitmap_info, offset, bytes);
3144 	bytes -= bytes_added;
3145 	offset += bytes_added;
3146 	spin_unlock(&ctl->tree_lock);
3147 
3148 	if (bytes)
3149 		goto again;
3150 
3151 	if (map)
3152 		kfree(map);
3153 	return 0;
3154 }
3155 
3156 /*
3157  * This test just does basic sanity checking, making sure we can add an exten
3158  * entry and remove space from either end and the middle, and make sure we can
3159  * remove space that covers adjacent extent entries.
3160  */
3161 static int test_extents(struct btrfs_block_group_cache *cache)
3162 {
3163 	int ret = 0;
3164 
3165 	printk(KERN_ERR "Running extent only tests\n");
3166 
3167 	/* First just make sure we can remove an entire entry */
3168 	ret = btrfs_add_free_space(cache, 0, 4 * 1024 * 1024);
3169 	if (ret) {
3170 		printk(KERN_ERR "Error adding initial extents %d\n", ret);
3171 		return ret;
3172 	}
3173 
3174 	ret = btrfs_remove_free_space(cache, 0, 4 * 1024 * 1024);
3175 	if (ret) {
3176 		printk(KERN_ERR "Error removing extent %d\n", ret);
3177 		return ret;
3178 	}
3179 
3180 	if (check_exists(cache, 0, 4 * 1024 * 1024)) {
3181 		printk(KERN_ERR "Full remove left some lingering space\n");
3182 		return -1;
3183 	}
3184 
3185 	/* Ok edge and middle cases now */
3186 	ret = btrfs_add_free_space(cache, 0, 4 * 1024 * 1024);
3187 	if (ret) {
3188 		printk(KERN_ERR "Error adding half extent %d\n", ret);
3189 		return ret;
3190 	}
3191 
3192 	ret = btrfs_remove_free_space(cache, 3 * 1024 * 1024, 1 * 1024 * 1024);
3193 	if (ret) {
3194 		printk(KERN_ERR "Error removing tail end %d\n", ret);
3195 		return ret;
3196 	}
3197 
3198 	ret = btrfs_remove_free_space(cache, 0, 1 * 1024 * 1024);
3199 	if (ret) {
3200 		printk(KERN_ERR "Error removing front end %d\n", ret);
3201 		return ret;
3202 	}
3203 
3204 	ret = btrfs_remove_free_space(cache, 2 * 1024 * 1024, 4096);
3205 	if (ret) {
3206 		printk(KERN_ERR "Error removing middle peice %d\n", ret);
3207 		return ret;
3208 	}
3209 
3210 	if (check_exists(cache, 0, 1 * 1024 * 1024)) {
3211 		printk(KERN_ERR "Still have space at the front\n");
3212 		return -1;
3213 	}
3214 
3215 	if (check_exists(cache, 2 * 1024 * 1024, 4096)) {
3216 		printk(KERN_ERR "Still have space in the middle\n");
3217 		return -1;
3218 	}
3219 
3220 	if (check_exists(cache, 3 * 1024 * 1024, 1 * 1024 * 1024)) {
3221 		printk(KERN_ERR "Still have space at the end\n");
3222 		return -1;
3223 	}
3224 
3225 	/* Cleanup */
3226 	__btrfs_remove_free_space_cache(cache->free_space_ctl);
3227 
3228 	return 0;
3229 }
3230 
3231 static int test_bitmaps(struct btrfs_block_group_cache *cache)
3232 {
3233 	u64 next_bitmap_offset;
3234 	int ret;
3235 
3236 	printk(KERN_ERR "Running bitmap only tests\n");
3237 
3238 	ret = add_free_space_entry(cache, 0, 4 * 1024 * 1024, 1);
3239 	if (ret) {
3240 		printk(KERN_ERR "Couldn't create a bitmap entry %d\n", ret);
3241 		return ret;
3242 	}
3243 
3244 	ret = btrfs_remove_free_space(cache, 0, 4 * 1024 * 1024);
3245 	if (ret) {
3246 		printk(KERN_ERR "Error removing bitmap full range %d\n", ret);
3247 		return ret;
3248 	}
3249 
3250 	if (check_exists(cache, 0, 4 * 1024 * 1024)) {
3251 		printk(KERN_ERR "Left some space in bitmap\n");
3252 		return -1;
3253 	}
3254 
3255 	ret = add_free_space_entry(cache, 0, 4 * 1024 * 1024, 1);
3256 	if (ret) {
3257 		printk(KERN_ERR "Couldn't add to our bitmap entry %d\n", ret);
3258 		return ret;
3259 	}
3260 
3261 	ret = btrfs_remove_free_space(cache, 1 * 1024 * 1024, 2 * 1024 * 1024);
3262 	if (ret) {
3263 		printk(KERN_ERR "Couldn't remove middle chunk %d\n", ret);
3264 		return ret;
3265 	}
3266 
3267 	/*
3268 	 * The first bitmap we have starts at offset 0 so the next one is just
3269 	 * at the end of the first bitmap.
3270 	 */
3271 	next_bitmap_offset = (u64)(BITS_PER_BITMAP * 4096);
3272 
3273 	/* Test a bit straddling two bitmaps */
3274 	ret = add_free_space_entry(cache, next_bitmap_offset -
3275 				   (2 * 1024 * 1024), 4 * 1024 * 1024, 1);
3276 	if (ret) {
3277 		printk(KERN_ERR "Couldn't add space that straddles two bitmaps"
3278 		       " %d\n", ret);
3279 		return ret;
3280 	}
3281 
3282 	ret = btrfs_remove_free_space(cache, next_bitmap_offset -
3283 				      (1 * 1024 * 1024), 2 * 1024 * 1024);
3284 	if (ret) {
3285 		printk(KERN_ERR "Couldn't remove overlapping space %d\n", ret);
3286 		return ret;
3287 	}
3288 
3289 	if (check_exists(cache, next_bitmap_offset - (1 * 1024 * 1024),
3290 			 2 * 1024 * 1024)) {
3291 		printk(KERN_ERR "Left some space when removing overlapping\n");
3292 		return -1;
3293 	}
3294 
3295 	__btrfs_remove_free_space_cache(cache->free_space_ctl);
3296 
3297 	return 0;
3298 }
3299 
3300 /* This is the high grade jackassery */
3301 static int test_bitmaps_and_extents(struct btrfs_block_group_cache *cache)
3302 {
3303 	u64 bitmap_offset = (u64)(BITS_PER_BITMAP * 4096);
3304 	int ret;
3305 
3306 	printk(KERN_ERR "Running bitmap and extent tests\n");
3307 
3308 	/*
3309 	 * First let's do something simple, an extent at the same offset as the
3310 	 * bitmap, but the free space completely in the extent and then
3311 	 * completely in the bitmap.
3312 	 */
3313 	ret = add_free_space_entry(cache, 4 * 1024 * 1024, 1 * 1024 * 1024, 1);
3314 	if (ret) {
3315 		printk(KERN_ERR "Couldn't create bitmap entry %d\n", ret);
3316 		return ret;
3317 	}
3318 
3319 	ret = add_free_space_entry(cache, 0, 1 * 1024 * 1024, 0);
3320 	if (ret) {
3321 		printk(KERN_ERR "Couldn't add extent entry %d\n", ret);
3322 		return ret;
3323 	}
3324 
3325 	ret = btrfs_remove_free_space(cache, 0, 1 * 1024 * 1024);
3326 	if (ret) {
3327 		printk(KERN_ERR "Couldn't remove extent entry %d\n", ret);
3328 		return ret;
3329 	}
3330 
3331 	if (check_exists(cache, 0, 1 * 1024 * 1024)) {
3332 		printk(KERN_ERR "Left remnants after our remove\n");
3333 		return -1;
3334 	}
3335 
3336 	/* Now to add back the extent entry and remove from the bitmap */
3337 	ret = add_free_space_entry(cache, 0, 1 * 1024 * 1024, 0);
3338 	if (ret) {
3339 		printk(KERN_ERR "Couldn't re-add extent entry %d\n", ret);
3340 		return ret;
3341 	}
3342 
3343 	ret = btrfs_remove_free_space(cache, 4 * 1024 * 1024, 1 * 1024 * 1024);
3344 	if (ret) {
3345 		printk(KERN_ERR "Couldn't remove from bitmap %d\n", ret);
3346 		return ret;
3347 	}
3348 
3349 	if (check_exists(cache, 4 * 1024 * 1024, 1 * 1024 * 1024)) {
3350 		printk(KERN_ERR "Left remnants in the bitmap\n");
3351 		return -1;
3352 	}
3353 
3354 	/*
3355 	 * Ok so a little more evil, extent entry and bitmap at the same offset,
3356 	 * removing an overlapping chunk.
3357 	 */
3358 	ret = add_free_space_entry(cache, 1 * 1024 * 1024, 4 * 1024 * 1024, 1);
3359 	if (ret) {
3360 		printk(KERN_ERR "Couldn't add to a bitmap %d\n", ret);
3361 		return ret;
3362 	}
3363 
3364 	ret = btrfs_remove_free_space(cache, 512 * 1024, 3 * 1024 * 1024);
3365 	if (ret) {
3366 		printk(KERN_ERR "Couldn't remove overlapping space %d\n", ret);
3367 		return ret;
3368 	}
3369 
3370 	if (check_exists(cache, 512 * 1024, 3 * 1024 * 1024)) {
3371 		printk(KERN_ERR "Left over peices after removing "
3372 		       "overlapping\n");
3373 		return -1;
3374 	}
3375 
3376 	__btrfs_remove_free_space_cache(cache->free_space_ctl);
3377 
3378 	/* Now with the extent entry offset into the bitmap */
3379 	ret = add_free_space_entry(cache, 4 * 1024 * 1024, 4 * 1024 * 1024, 1);
3380 	if (ret) {
3381 		printk(KERN_ERR "Couldn't add space to the bitmap %d\n", ret);
3382 		return ret;
3383 	}
3384 
3385 	ret = add_free_space_entry(cache, 2 * 1024 * 1024, 2 * 1024 * 1024, 0);
3386 	if (ret) {
3387 		printk(KERN_ERR "Couldn't add extent to the cache %d\n", ret);
3388 		return ret;
3389 	}
3390 
3391 	ret = btrfs_remove_free_space(cache, 3 * 1024 * 1024, 4 * 1024 * 1024);
3392 	if (ret) {
3393 		printk(KERN_ERR "Problem removing overlapping space %d\n", ret);
3394 		return ret;
3395 	}
3396 
3397 	if (check_exists(cache, 3 * 1024 * 1024, 4 * 1024 * 1024)) {
3398 		printk(KERN_ERR "Left something behind when removing space");
3399 		return -1;
3400 	}
3401 
3402 	/*
3403 	 * This has blown up in the past, the extent entry starts before the
3404 	 * bitmap entry, but we're trying to remove an offset that falls
3405 	 * completely within the bitmap range and is in both the extent entry
3406 	 * and the bitmap entry, looks like this
3407 	 *
3408 	 *   [ extent ]
3409 	 *      [ bitmap ]
3410 	 *        [ del ]
3411 	 */
3412 	__btrfs_remove_free_space_cache(cache->free_space_ctl);
3413 	ret = add_free_space_entry(cache, bitmap_offset + 4 * 1024 * 1024,
3414 				   4 * 1024 * 1024, 1);
3415 	if (ret) {
3416 		printk(KERN_ERR "Couldn't add bitmap %d\n", ret);
3417 		return ret;
3418 	}
3419 
3420 	ret = add_free_space_entry(cache, bitmap_offset - 1 * 1024 * 1024,
3421 				   5 * 1024 * 1024, 0);
3422 	if (ret) {
3423 		printk(KERN_ERR "Couldn't add extent entry %d\n", ret);
3424 		return ret;
3425 	}
3426 
3427 	ret = btrfs_remove_free_space(cache, bitmap_offset + 1 * 1024 * 1024,
3428 				      5 * 1024 * 1024);
3429 	if (ret) {
3430 		printk(KERN_ERR "Failed to free our space %d\n", ret);
3431 		return ret;
3432 	}
3433 
3434 	if (check_exists(cache, bitmap_offset + 1 * 1024 * 1024,
3435 			 5 * 1024 * 1024)) {
3436 		printk(KERN_ERR "Left stuff over\n");
3437 		return -1;
3438 	}
3439 
3440 	__btrfs_remove_free_space_cache(cache->free_space_ctl);
3441 
3442 	/*
3443 	 * This blew up before, we have part of the free space in a bitmap and
3444 	 * then the entirety of the rest of the space in an extent.  This used
3445 	 * to return -EAGAIN back from btrfs_remove_extent, make sure this
3446 	 * doesn't happen.
3447 	 */
3448 	ret = add_free_space_entry(cache, 1 * 1024 * 1024, 2 * 1024 * 1024, 1);
3449 	if (ret) {
3450 		printk(KERN_ERR "Couldn't add bitmap entry %d\n", ret);
3451 		return ret;
3452 	}
3453 
3454 	ret = add_free_space_entry(cache, 3 * 1024 * 1024, 1 * 1024 * 1024, 0);
3455 	if (ret) {
3456 		printk(KERN_ERR "Couldn't add extent entry %d\n", ret);
3457 		return ret;
3458 	}
3459 
3460 	ret = btrfs_remove_free_space(cache, 1 * 1024 * 1024, 3 * 1024 * 1024);
3461 	if (ret) {
3462 		printk(KERN_ERR "Error removing bitmap and extent "
3463 		       "overlapping %d\n", ret);
3464 		return ret;
3465 	}
3466 
3467 	__btrfs_remove_free_space_cache(cache->free_space_ctl);
3468 	return 0;
3469 }
3470 
3471 void btrfs_test_free_space_cache(void)
3472 {
3473 	struct btrfs_block_group_cache *cache;
3474 
3475 	printk(KERN_ERR "Running btrfs free space cache tests\n");
3476 
3477 	cache = init_test_block_group();
3478 	if (!cache) {
3479 		printk(KERN_ERR "Couldn't run the tests\n");
3480 		return;
3481 	}
3482 
3483 	if (test_extents(cache))
3484 		goto out;
3485 	if (test_bitmaps(cache))
3486 		goto out;
3487 	if (test_bitmaps_and_extents(cache))
3488 		goto out;
3489 out:
3490 	__btrfs_remove_free_space_cache(cache->free_space_ctl);
3491 	kfree(cache->free_space_ctl);
3492 	kfree(cache);
3493 	printk(KERN_ERR "Free space cache tests finished\n");
3494 }
3495 #endif /* CONFIG_BTRFS_FS_RUN_SANITY_TESTS */
3496