xref: /linux/fs/btrfs/free-space-tree.c (revision 9f388a653c8a481cbdbdedca081a1f9f3ba204a2)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) 2015 Facebook.  All rights reserved.
4  */
5 
6 #include <linux/kernel.h>
7 #include <linux/sched/mm.h>
8 #include "messages.h"
9 #include "ctree.h"
10 #include "disk-io.h"
11 #include "locking.h"
12 #include "free-space-tree.h"
13 #include "transaction.h"
14 #include "block-group.h"
15 #include "fs.h"
16 #include "accessors.h"
17 #include "extent-tree.h"
18 #include "root-tree.h"
19 
20 static int __add_block_group_free_space(struct btrfs_trans_handle *trans,
21 					struct btrfs_block_group *block_group,
22 					struct btrfs_path *path);
23 
btrfs_free_space_root(struct btrfs_block_group * block_group)24 static struct btrfs_root *btrfs_free_space_root(
25 				struct btrfs_block_group *block_group)
26 {
27 	struct btrfs_key key = {
28 		.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID,
29 		.type = BTRFS_ROOT_ITEM_KEY,
30 		.offset = 0,
31 	};
32 
33 	if (btrfs_fs_incompat(block_group->fs_info, EXTENT_TREE_V2))
34 		key.offset = block_group->global_root_id;
35 	return btrfs_global_root(block_group->fs_info, &key);
36 }
37 
btrfs_set_free_space_tree_thresholds(struct btrfs_block_group * cache)38 void btrfs_set_free_space_tree_thresholds(struct btrfs_block_group *cache)
39 {
40 	u32 bitmap_range;
41 	size_t bitmap_size;
42 	u64 num_bitmaps, total_bitmap_size;
43 
44 	if (WARN_ON(cache->length == 0))
45 		btrfs_warn(cache->fs_info, "block group %llu length is zero",
46 			   cache->start);
47 
48 	/*
49 	 * We convert to bitmaps when the disk space required for using extents
50 	 * exceeds that required for using bitmaps.
51 	 */
52 	bitmap_range = cache->fs_info->sectorsize * BTRFS_FREE_SPACE_BITMAP_BITS;
53 	num_bitmaps = div_u64(cache->length + bitmap_range - 1, bitmap_range);
54 	bitmap_size = sizeof(struct btrfs_item) + BTRFS_FREE_SPACE_BITMAP_SIZE;
55 	total_bitmap_size = num_bitmaps * bitmap_size;
56 	cache->bitmap_high_thresh = div_u64(total_bitmap_size,
57 					    sizeof(struct btrfs_item));
58 
59 	/*
60 	 * We allow for a small buffer between the high threshold and low
61 	 * threshold to avoid thrashing back and forth between the two formats.
62 	 */
63 	if (cache->bitmap_high_thresh > 100)
64 		cache->bitmap_low_thresh = cache->bitmap_high_thresh - 100;
65 	else
66 		cache->bitmap_low_thresh = 0;
67 }
68 
add_new_free_space_info(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path)69 static int add_new_free_space_info(struct btrfs_trans_handle *trans,
70 				   struct btrfs_block_group *block_group,
71 				   struct btrfs_path *path)
72 {
73 	struct btrfs_root *root = btrfs_free_space_root(block_group);
74 	struct btrfs_free_space_info *info;
75 	struct btrfs_key key;
76 	struct extent_buffer *leaf;
77 	int ret;
78 
79 	key.objectid = block_group->start;
80 	key.type = BTRFS_FREE_SPACE_INFO_KEY;
81 	key.offset = block_group->length;
82 
83 	ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(*info));
84 	if (ret)
85 		return ret;
86 
87 	leaf = path->nodes[0];
88 	info = btrfs_item_ptr(leaf, path->slots[0],
89 			      struct btrfs_free_space_info);
90 	btrfs_set_free_space_extent_count(leaf, info, 0);
91 	btrfs_set_free_space_flags(leaf, info, 0);
92 	btrfs_release_path(path);
93 	return 0;
94 }
95 
96 EXPORT_FOR_TESTS
btrfs_search_free_space_info(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path,int cow)97 struct btrfs_free_space_info *btrfs_search_free_space_info(
98 		struct btrfs_trans_handle *trans,
99 		struct btrfs_block_group *block_group,
100 		struct btrfs_path *path, int cow)
101 {
102 	struct btrfs_fs_info *fs_info = block_group->fs_info;
103 	struct btrfs_root *root = btrfs_free_space_root(block_group);
104 	struct btrfs_key key;
105 	int ret;
106 
107 	key.objectid = block_group->start;
108 	key.type = BTRFS_FREE_SPACE_INFO_KEY;
109 	key.offset = block_group->length;
110 
111 	ret = btrfs_search_slot(trans, root, &key, path, 0, cow);
112 	if (ret < 0)
113 		return ERR_PTR(ret);
114 	if (ret != 0) {
115 		btrfs_warn(fs_info, "missing free space info for %llu",
116 			   block_group->start);
117 		DEBUG_WARN();
118 		return ERR_PTR(-ENOENT);
119 	}
120 
121 	return btrfs_item_ptr(path->nodes[0], path->slots[0],
122 			      struct btrfs_free_space_info);
123 }
124 
125 /*
126  * btrfs_search_slot() but we're looking for the greatest key less than the
127  * passed key.
128  */
btrfs_search_prev_slot(struct btrfs_trans_handle * trans,struct btrfs_root * root,struct btrfs_key * key,struct btrfs_path * p,int ins_len,int cow)129 static int btrfs_search_prev_slot(struct btrfs_trans_handle *trans,
130 				  struct btrfs_root *root,
131 				  struct btrfs_key *key, struct btrfs_path *p,
132 				  int ins_len, int cow)
133 {
134 	int ret;
135 
136 	ret = btrfs_search_slot(trans, root, key, p, ins_len, cow);
137 	if (ret < 0)
138 		return ret;
139 
140 	if (unlikely(ret == 0)) {
141 		DEBUG_WARN();
142 		return -EIO;
143 	}
144 
145 	if (unlikely(p->slots[0] == 0)) {
146 		DEBUG_WARN("no previous slot found");
147 		return -EIO;
148 	}
149 	p->slots[0]--;
150 
151 	return 0;
152 }
153 
free_space_bitmap_size(const struct btrfs_fs_info * fs_info,u64 size)154 static inline u32 free_space_bitmap_size(const struct btrfs_fs_info *fs_info,
155 					 u64 size)
156 {
157 	return DIV_ROUND_UP(size >> fs_info->sectorsize_bits, BITS_PER_BYTE);
158 }
159 
alloc_bitmap(u32 bitmap_size)160 static unsigned long *alloc_bitmap(u32 bitmap_size)
161 {
162 	unsigned long *ret;
163 	unsigned int nofs_flag;
164 	u32 bitmap_rounded_size = round_up(bitmap_size, sizeof(unsigned long));
165 
166 	/*
167 	 * GFP_NOFS doesn't work with kvmalloc(), but we really can't recurse
168 	 * into the filesystem as the free space bitmap can be modified in the
169 	 * critical section of a transaction commit.
170 	 *
171 	 * TODO: push the memalloc_nofs_{save,restore}() to the caller where we
172 	 * know that recursion is unsafe.
173 	 */
174 	nofs_flag = memalloc_nofs_save();
175 	ret = kvzalloc(bitmap_rounded_size, GFP_KERNEL);
176 	memalloc_nofs_restore(nofs_flag);
177 	return ret;
178 }
179 
le_bitmap_set(unsigned long * map,unsigned int start,int len)180 static void le_bitmap_set(unsigned long *map, unsigned int start, int len)
181 {
182 	u8 *p = ((u8 *)map) + BIT_BYTE(start);
183 	const unsigned int size = start + len;
184 	int bits_to_set = BITS_PER_BYTE - (start % BITS_PER_BYTE);
185 	u8 mask_to_set = BITMAP_FIRST_BYTE_MASK(start);
186 
187 	while (len - bits_to_set >= 0) {
188 		*p |= mask_to_set;
189 		len -= bits_to_set;
190 		bits_to_set = BITS_PER_BYTE;
191 		mask_to_set = ~0;
192 		p++;
193 	}
194 	if (len) {
195 		mask_to_set &= BITMAP_LAST_BYTE_MASK(size);
196 		*p |= mask_to_set;
197 	}
198 }
199 
200 EXPORT_FOR_TESTS
btrfs_convert_free_space_to_bitmaps(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path)201 int btrfs_convert_free_space_to_bitmaps(struct btrfs_trans_handle *trans,
202 					struct btrfs_block_group *block_group,
203 					struct btrfs_path *path)
204 {
205 	struct btrfs_fs_info *fs_info = trans->fs_info;
206 	struct btrfs_root *root = btrfs_free_space_root(block_group);
207 	struct btrfs_free_space_info *info;
208 	struct btrfs_key key, found_key;
209 	struct extent_buffer *leaf;
210 	unsigned long *bitmap;
211 	char *bitmap_cursor;
212 	u64 start, end;
213 	u64 bitmap_range, i;
214 	u32 bitmap_size, flags, expected_extent_count;
215 	u32 extent_count = 0;
216 	int done = 0, nr;
217 	int ret;
218 
219 	bitmap_size = free_space_bitmap_size(fs_info, block_group->length);
220 	bitmap = alloc_bitmap(bitmap_size);
221 	if (unlikely(!bitmap)) {
222 		ret = -ENOMEM;
223 		btrfs_abort_transaction(trans, ret);
224 		goto out;
225 	}
226 
227 	start = block_group->start;
228 	end = block_group->start + block_group->length;
229 
230 	key.objectid = end - 1;
231 	key.type = (u8)-1;
232 	key.offset = (u64)-1;
233 
234 	while (!done) {
235 		ret = btrfs_search_prev_slot(trans, root, &key, path, -1, 1);
236 		if (unlikely(ret)) {
237 			btrfs_abort_transaction(trans, ret);
238 			goto out;
239 		}
240 
241 		leaf = path->nodes[0];
242 		nr = 0;
243 		path->slots[0]++;
244 		while (path->slots[0] > 0) {
245 			btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0] - 1);
246 
247 			if (found_key.type == BTRFS_FREE_SPACE_INFO_KEY) {
248 				ASSERT(found_key.objectid == block_group->start);
249 				ASSERT(found_key.offset == block_group->length);
250 				done = 1;
251 				break;
252 			} else if (found_key.type == BTRFS_FREE_SPACE_EXTENT_KEY) {
253 				u64 first, last;
254 
255 				ASSERT(found_key.objectid >= start);
256 				ASSERT(found_key.objectid < end);
257 				ASSERT(found_key.objectid + found_key.offset <= end);
258 
259 				first = div_u64(found_key.objectid - start,
260 						fs_info->sectorsize);
261 				last = div_u64(found_key.objectid + found_key.offset - start,
262 					       fs_info->sectorsize);
263 				le_bitmap_set(bitmap, first, last - first);
264 
265 				extent_count++;
266 				nr++;
267 				path->slots[0]--;
268 			} else {
269 				ASSERT(0);
270 			}
271 		}
272 
273 		ret = btrfs_del_items(trans, root, path, path->slots[0], nr);
274 		if (unlikely(ret)) {
275 			btrfs_abort_transaction(trans, ret);
276 			goto out;
277 		}
278 		btrfs_release_path(path);
279 	}
280 
281 	info = btrfs_search_free_space_info(trans, block_group, path, 1);
282 	if (IS_ERR(info)) {
283 		ret = PTR_ERR(info);
284 		btrfs_abort_transaction(trans, ret);
285 		goto out;
286 	}
287 	leaf = path->nodes[0];
288 	flags = btrfs_free_space_flags(leaf, info);
289 	flags |= BTRFS_FREE_SPACE_USING_BITMAPS;
290 	block_group->using_free_space_bitmaps = true;
291 	block_group->using_free_space_bitmaps_cached = true;
292 	btrfs_set_free_space_flags(leaf, info, flags);
293 	expected_extent_count = btrfs_free_space_extent_count(leaf, info);
294 	btrfs_release_path(path);
295 
296 	if (unlikely(extent_count != expected_extent_count)) {
297 		btrfs_err(fs_info,
298 			  "incorrect extent count for %llu; counted %u, expected %u",
299 			  block_group->start, extent_count,
300 			  expected_extent_count);
301 		ret = -EIO;
302 		btrfs_abort_transaction(trans, ret);
303 		goto out;
304 	}
305 
306 	bitmap_cursor = (char *)bitmap;
307 	bitmap_range = fs_info->sectorsize * BTRFS_FREE_SPACE_BITMAP_BITS;
308 	i = start;
309 	while (i < end) {
310 		unsigned long ptr;
311 		u64 extent_size;
312 		u32 data_size;
313 
314 		extent_size = min(end - i, bitmap_range);
315 		data_size = free_space_bitmap_size(fs_info, extent_size);
316 
317 		key.objectid = i;
318 		key.type = BTRFS_FREE_SPACE_BITMAP_KEY;
319 		key.offset = extent_size;
320 
321 		ret = btrfs_insert_empty_item(trans, root, path, &key,
322 					      data_size);
323 		if (unlikely(ret)) {
324 			btrfs_abort_transaction(trans, ret);
325 			goto out;
326 		}
327 
328 		leaf = path->nodes[0];
329 		ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
330 		write_extent_buffer(leaf, bitmap_cursor, ptr,
331 				    data_size);
332 		btrfs_release_path(path);
333 
334 		i += extent_size;
335 		bitmap_cursor += data_size;
336 	}
337 
338 	ret = 0;
339 out:
340 	kvfree(bitmap);
341 	return ret;
342 }
343 
344 EXPORT_FOR_TESTS
btrfs_convert_free_space_to_extents(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path)345 int btrfs_convert_free_space_to_extents(struct btrfs_trans_handle *trans,
346 					struct btrfs_block_group *block_group,
347 					struct btrfs_path *path)
348 {
349 	struct btrfs_fs_info *fs_info = trans->fs_info;
350 	struct btrfs_root *root = btrfs_free_space_root(block_group);
351 	struct btrfs_free_space_info *info;
352 	struct btrfs_key key, found_key;
353 	struct extent_buffer *leaf;
354 	unsigned long *bitmap;
355 	u64 start, end;
356 	u32 bitmap_size, flags, expected_extent_count;
357 	unsigned long nrbits, start_bit, end_bit;
358 	u32 extent_count = 0;
359 	int done = 0, nr;
360 	int ret;
361 
362 	bitmap_size = free_space_bitmap_size(fs_info, block_group->length);
363 	bitmap = alloc_bitmap(bitmap_size);
364 	if (unlikely(!bitmap)) {
365 		ret = -ENOMEM;
366 		btrfs_abort_transaction(trans, ret);
367 		goto out;
368 	}
369 
370 	start = block_group->start;
371 	end = block_group->start + block_group->length;
372 
373 	key.objectid = end - 1;
374 	key.type = (u8)-1;
375 	key.offset = (u64)-1;
376 
377 	while (!done) {
378 		ret = btrfs_search_prev_slot(trans, root, &key, path, -1, 1);
379 		if (unlikely(ret)) {
380 			btrfs_abort_transaction(trans, ret);
381 			goto out;
382 		}
383 
384 		leaf = path->nodes[0];
385 		nr = 0;
386 		path->slots[0]++;
387 		while (path->slots[0] > 0) {
388 			btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0] - 1);
389 
390 			if (found_key.type == BTRFS_FREE_SPACE_INFO_KEY) {
391 				ASSERT(found_key.objectid == block_group->start);
392 				ASSERT(found_key.offset == block_group->length);
393 				done = 1;
394 				break;
395 			} else if (found_key.type == BTRFS_FREE_SPACE_BITMAP_KEY) {
396 				unsigned long ptr;
397 				char *bitmap_cursor;
398 				u32 bitmap_pos, data_size;
399 
400 				ASSERT(found_key.objectid >= start);
401 				ASSERT(found_key.objectid < end);
402 				ASSERT(found_key.objectid + found_key.offset <= end);
403 
404 				bitmap_pos = div_u64(found_key.objectid - start,
405 						     fs_info->sectorsize *
406 						     BITS_PER_BYTE);
407 				bitmap_cursor = ((char *)bitmap) + bitmap_pos;
408 				data_size = free_space_bitmap_size(fs_info,
409 								found_key.offset);
410 
411 				path->slots[0]--;
412 				ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
413 				read_extent_buffer(leaf, bitmap_cursor, ptr,
414 						   data_size);
415 
416 				nr++;
417 			} else {
418 				ASSERT(0);
419 			}
420 		}
421 
422 		ret = btrfs_del_items(trans, root, path, path->slots[0], nr);
423 		if (unlikely(ret)) {
424 			btrfs_abort_transaction(trans, ret);
425 			goto out;
426 		}
427 		btrfs_release_path(path);
428 	}
429 
430 	info = btrfs_search_free_space_info(trans, block_group, path, 1);
431 	if (IS_ERR(info)) {
432 		ret = PTR_ERR(info);
433 		btrfs_abort_transaction(trans, ret);
434 		goto out;
435 	}
436 	leaf = path->nodes[0];
437 	flags = btrfs_free_space_flags(leaf, info);
438 	flags &= ~BTRFS_FREE_SPACE_USING_BITMAPS;
439 	block_group->using_free_space_bitmaps = false;
440 	block_group->using_free_space_bitmaps_cached = true;
441 	btrfs_set_free_space_flags(leaf, info, flags);
442 	expected_extent_count = btrfs_free_space_extent_count(leaf, info);
443 	btrfs_release_path(path);
444 
445 	nrbits = block_group->length >> fs_info->sectorsize_bits;
446 	start_bit = find_next_bit_le(bitmap, nrbits, 0);
447 
448 	while (start_bit < nrbits) {
449 		end_bit = find_next_zero_bit_le(bitmap, nrbits, start_bit);
450 		ASSERT(start_bit < end_bit);
451 
452 		key.objectid = start + start_bit * fs_info->sectorsize;
453 		key.type = BTRFS_FREE_SPACE_EXTENT_KEY;
454 		key.offset = (end_bit - start_bit) * fs_info->sectorsize;
455 
456 		ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
457 		if (unlikely(ret)) {
458 			btrfs_abort_transaction(trans, ret);
459 			goto out;
460 		}
461 		btrfs_release_path(path);
462 
463 		extent_count++;
464 
465 		start_bit = find_next_bit_le(bitmap, nrbits, end_bit);
466 	}
467 
468 	if (unlikely(extent_count != expected_extent_count)) {
469 		btrfs_err(fs_info,
470 			  "incorrect extent count for %llu; counted %u, expected %u",
471 			  block_group->start, extent_count,
472 			  expected_extent_count);
473 		ret = -EIO;
474 		btrfs_abort_transaction(trans, ret);
475 		goto out;
476 	}
477 
478 	ret = 0;
479 out:
480 	kvfree(bitmap);
481 	return ret;
482 }
483 
update_free_space_extent_count(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path,int new_extents)484 static int update_free_space_extent_count(struct btrfs_trans_handle *trans,
485 					  struct btrfs_block_group *block_group,
486 					  struct btrfs_path *path,
487 					  int new_extents)
488 {
489 	struct btrfs_free_space_info *info;
490 	u32 flags;
491 	u32 extent_count;
492 	int ret = 0;
493 
494 	if (new_extents == 0)
495 		return 0;
496 
497 	info = btrfs_search_free_space_info(trans, block_group, path, 1);
498 	if (IS_ERR(info))
499 		return PTR_ERR(info);
500 
501 	flags = btrfs_free_space_flags(path->nodes[0], info);
502 	extent_count = btrfs_free_space_extent_count(path->nodes[0], info);
503 
504 	extent_count += new_extents;
505 	btrfs_set_free_space_extent_count(path->nodes[0], info, extent_count);
506 	btrfs_release_path(path);
507 
508 	if (!(flags & BTRFS_FREE_SPACE_USING_BITMAPS) &&
509 	    extent_count > block_group->bitmap_high_thresh) {
510 		ret = btrfs_convert_free_space_to_bitmaps(trans, block_group, path);
511 	} else if ((flags & BTRFS_FREE_SPACE_USING_BITMAPS) &&
512 		   extent_count < block_group->bitmap_low_thresh) {
513 		ret = btrfs_convert_free_space_to_extents(trans, block_group, path);
514 	}
515 
516 	return ret;
517 }
518 
519 EXPORT_FOR_TESTS
btrfs_free_space_test_bit(struct btrfs_block_group * block_group,struct btrfs_path * path,u64 offset)520 bool btrfs_free_space_test_bit(struct btrfs_block_group *block_group,
521 			       struct btrfs_path *path, u64 offset)
522 {
523 	struct extent_buffer *leaf;
524 	struct btrfs_key key;
525 	u64 found_start, found_end;
526 	unsigned long ptr, i;
527 
528 	leaf = path->nodes[0];
529 	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
530 	ASSERT(key.type == BTRFS_FREE_SPACE_BITMAP_KEY);
531 
532 	found_start = key.objectid;
533 	found_end = key.objectid + key.offset;
534 	ASSERT(offset >= found_start && offset < found_end);
535 
536 	ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
537 	i = div_u64(offset - found_start,
538 		    block_group->fs_info->sectorsize);
539 	return extent_buffer_test_bit(leaf, ptr, i);
540 }
541 
free_space_modify_bits(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path,u64 * start,u64 * size,bool set_bits)542 static void free_space_modify_bits(struct btrfs_trans_handle *trans,
543 				   struct btrfs_block_group *block_group,
544 				   struct btrfs_path *path, u64 *start, u64 *size,
545 				   bool set_bits)
546 {
547 	struct btrfs_fs_info *fs_info = block_group->fs_info;
548 	struct extent_buffer *leaf;
549 	struct btrfs_key key;
550 	u64 end = *start + *size;
551 	u64 found_start, found_end;
552 	unsigned long ptr, first, last;
553 
554 	leaf = path->nodes[0];
555 	btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
556 	ASSERT(key.type == BTRFS_FREE_SPACE_BITMAP_KEY);
557 
558 	found_start = key.objectid;
559 	found_end = key.objectid + key.offset;
560 	ASSERT(*start >= found_start && *start < found_end);
561 	ASSERT(end > found_start);
562 
563 	if (end > found_end)
564 		end = found_end;
565 
566 	ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
567 	first = (*start - found_start) >> fs_info->sectorsize_bits;
568 	last = (end - found_start) >> fs_info->sectorsize_bits;
569 	if (set_bits)
570 		extent_buffer_bitmap_set(leaf, ptr, first, last - first);
571 	else
572 		extent_buffer_bitmap_clear(leaf, ptr, first, last - first);
573 	btrfs_mark_buffer_dirty(trans, leaf);
574 
575 	*size -= end - *start;
576 	*start = end;
577 }
578 
579 /*
580  * We can't use btrfs_next_item() in modify_free_space_bitmap() because
581  * btrfs_next_leaf() doesn't get the path for writing. We can forgo the fancy
582  * tree walking in btrfs_next_leaf() anyways because we know exactly what we're
583  * looking for.
584  */
free_space_next_bitmap(struct btrfs_trans_handle * trans,struct btrfs_root * root,struct btrfs_path * p)585 static int free_space_next_bitmap(struct btrfs_trans_handle *trans,
586 				  struct btrfs_root *root, struct btrfs_path *p)
587 {
588 	struct btrfs_key key;
589 
590 	if (p->slots[0] + 1 < btrfs_header_nritems(p->nodes[0])) {
591 		p->slots[0]++;
592 		return 0;
593 	}
594 
595 	btrfs_item_key_to_cpu(p->nodes[0], &key, p->slots[0]);
596 	btrfs_release_path(p);
597 
598 	key.objectid += key.offset;
599 	key.type = (u8)-1;
600 	key.offset = (u64)-1;
601 
602 	return btrfs_search_prev_slot(trans, root, &key, p, 0, 1);
603 }
604 
605 /*
606  * If remove is 1, then we are removing free space, thus clearing bits in the
607  * bitmap. If remove is 0, then we are adding free space, thus setting bits in
608  * the bitmap.
609  */
modify_free_space_bitmap(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path,u64 start,u64 size,bool remove)610 static int modify_free_space_bitmap(struct btrfs_trans_handle *trans,
611 				    struct btrfs_block_group *block_group,
612 				    struct btrfs_path *path,
613 				    u64 start, u64 size, bool remove)
614 {
615 	struct btrfs_root *root = btrfs_free_space_root(block_group);
616 	struct btrfs_key key;
617 	u64 end = start + size;
618 	u64 cur_start, cur_size;
619 	bool prev_bit_set = false;
620 	bool next_bit_set = false;
621 	int new_extents;
622 	int ret;
623 
624 	/*
625 	 * Read the bit for the block immediately before the extent of space if
626 	 * that block is within the block group.
627 	 */
628 	if (start > block_group->start) {
629 		u64 prev_block = start - block_group->fs_info->sectorsize;
630 
631 		key.objectid = prev_block;
632 		key.type = (u8)-1;
633 		key.offset = (u64)-1;
634 
635 		ret = btrfs_search_prev_slot(trans, root, &key, path, 0, 1);
636 		if (ret)
637 			return ret;
638 
639 		prev_bit_set = btrfs_free_space_test_bit(block_group, path, prev_block);
640 
641 		/* The previous block may have been in the previous bitmap. */
642 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
643 		if (start >= key.objectid + key.offset) {
644 			ret = free_space_next_bitmap(trans, root, path);
645 			if (ret)
646 				return ret;
647 		}
648 	} else {
649 		key.objectid = start;
650 		key.type = (u8)-1;
651 		key.offset = (u64)-1;
652 
653 		ret = btrfs_search_prev_slot(trans, root, &key, path, 0, 1);
654 		if (ret)
655 			return ret;
656 	}
657 
658 	/*
659 	 * Iterate over all of the bitmaps overlapped by the extent of space,
660 	 * clearing/setting bits as required.
661 	 */
662 	cur_start = start;
663 	cur_size = size;
664 	while (1) {
665 		free_space_modify_bits(trans, block_group, path, &cur_start,
666 				       &cur_size, !remove);
667 		if (cur_size == 0)
668 			break;
669 		ret = free_space_next_bitmap(trans, root, path);
670 		if (ret)
671 			return ret;
672 	}
673 
674 	/*
675 	 * Read the bit for the block immediately after the extent of space if
676 	 * that block is within the block group.
677 	 */
678 	if (end < block_group->start + block_group->length) {
679 		/* The next block may be in the next bitmap. */
680 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
681 		if (end >= key.objectid + key.offset) {
682 			ret = free_space_next_bitmap(trans, root, path);
683 			if (ret)
684 				return ret;
685 		}
686 
687 		next_bit_set = btrfs_free_space_test_bit(block_group, path, end);
688 	}
689 
690 	if (remove) {
691 		new_extents = -1;
692 		if (prev_bit_set) {
693 			/* Leftover on the left. */
694 			new_extents++;
695 		}
696 		if (next_bit_set) {
697 			/* Leftover on the right. */
698 			new_extents++;
699 		}
700 	} else {
701 		new_extents = 1;
702 		if (prev_bit_set) {
703 			/* Merging with neighbor on the left. */
704 			new_extents--;
705 		}
706 		if (next_bit_set) {
707 			/* Merging with neighbor on the right. */
708 			new_extents--;
709 		}
710 	}
711 
712 	btrfs_release_path(path);
713 	return update_free_space_extent_count(trans, block_group, path, new_extents);
714 }
715 
remove_free_space_extent(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path,u64 start,u64 size)716 static int remove_free_space_extent(struct btrfs_trans_handle *trans,
717 				    struct btrfs_block_group *block_group,
718 				    struct btrfs_path *path,
719 				    u64 start, u64 size)
720 {
721 	struct btrfs_root *root = btrfs_free_space_root(block_group);
722 	struct btrfs_key key;
723 	u64 found_start, found_end;
724 	u64 end = start + size;
725 	int new_extents = -1;
726 	int ret;
727 
728 	key.objectid = start;
729 	key.type = (u8)-1;
730 	key.offset = (u64)-1;
731 
732 	ret = btrfs_search_prev_slot(trans, root, &key, path, -1, 1);
733 	if (ret)
734 		return ret;
735 
736 	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
737 
738 	ASSERT(key.type == BTRFS_FREE_SPACE_EXTENT_KEY);
739 
740 	found_start = key.objectid;
741 	found_end = key.objectid + key.offset;
742 	ASSERT(start >= found_start && end <= found_end);
743 
744 	/*
745 	 * Okay, now that we've found the free space extent which contains the
746 	 * free space that we are removing, there are four cases:
747 	 *
748 	 * 1. We're using the whole extent: delete the key we found and
749 	 * decrement the free space extent count.
750 	 * 2. We are using part of the extent starting at the beginning: delete
751 	 * the key we found and insert a new key representing the leftover at
752 	 * the end. There is no net change in the number of extents.
753 	 * 3. We are using part of the extent ending at the end: delete the key
754 	 * we found and insert a new key representing the leftover at the
755 	 * beginning. There is no net change in the number of extents.
756 	 * 4. We are using part of the extent in the middle: delete the key we
757 	 * found and insert two new keys representing the leftovers on each
758 	 * side. Where we used to have one extent, we now have two, so increment
759 	 * the extent count. We may need to convert the block group to bitmaps
760 	 * as a result.
761 	 */
762 
763 	/* Delete the existing key (cases 1-4). */
764 	ret = btrfs_del_item(trans, root, path);
765 	if (ret)
766 		return ret;
767 
768 	/* Add a key for leftovers at the beginning (cases 3 and 4). */
769 	if (start > found_start) {
770 		key.objectid = found_start;
771 		key.type = BTRFS_FREE_SPACE_EXTENT_KEY;
772 		key.offset = start - found_start;
773 
774 		btrfs_release_path(path);
775 		ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
776 		if (ret)
777 			return ret;
778 		new_extents++;
779 	}
780 
781 	/* Add a key for leftovers at the end (cases 2 and 4). */
782 	if (end < found_end) {
783 		key.objectid = end;
784 		key.type = BTRFS_FREE_SPACE_EXTENT_KEY;
785 		key.offset = found_end - end;
786 
787 		btrfs_release_path(path);
788 		ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
789 		if (ret)
790 			return ret;
791 		new_extents++;
792 	}
793 
794 	btrfs_release_path(path);
795 	return update_free_space_extent_count(trans, block_group, path, new_extents);
796 }
797 
using_bitmaps(struct btrfs_block_group * bg,struct btrfs_path * path)798 static int using_bitmaps(struct btrfs_block_group *bg, struct btrfs_path *path)
799 {
800 	struct btrfs_free_space_info *info;
801 	u32 flags;
802 
803 	if (bg->using_free_space_bitmaps_cached)
804 		return bg->using_free_space_bitmaps;
805 
806 	info = btrfs_search_free_space_info(NULL, bg, path, 0);
807 	if (IS_ERR(info))
808 		return PTR_ERR(info);
809 	flags = btrfs_free_space_flags(path->nodes[0], info);
810 	btrfs_release_path(path);
811 
812 	bg->using_free_space_bitmaps = (flags & BTRFS_FREE_SPACE_USING_BITMAPS);
813 	bg->using_free_space_bitmaps_cached = true;
814 
815 	return bg->using_free_space_bitmaps;
816 }
817 
818 EXPORT_FOR_TESTS
__btrfs_remove_from_free_space_tree(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path,u64 start,u64 size)819 int __btrfs_remove_from_free_space_tree(struct btrfs_trans_handle *trans,
820 					struct btrfs_block_group *block_group,
821 					struct btrfs_path *path, u64 start, u64 size)
822 {
823 	int ret;
824 
825 	ret = __add_block_group_free_space(trans, block_group, path);
826 	if (ret)
827 		return ret;
828 
829 	ret = using_bitmaps(block_group, path);
830 	if (ret < 0)
831 		return ret;
832 
833 	if (ret)
834 		return modify_free_space_bitmap(trans, block_group, path,
835 						start, size, true);
836 
837 	return remove_free_space_extent(trans, block_group, path, start, size);
838 }
839 
btrfs_remove_from_free_space_tree(struct btrfs_trans_handle * trans,u64 start,u64 size)840 int btrfs_remove_from_free_space_tree(struct btrfs_trans_handle *trans,
841 				      u64 start, u64 size)
842 {
843 	struct btrfs_block_group *block_group;
844 	struct btrfs_path *path;
845 	int ret;
846 
847 	if (!btrfs_fs_compat_ro(trans->fs_info, FREE_SPACE_TREE))
848 		return 0;
849 
850 	path = btrfs_alloc_path();
851 	if (unlikely(!path)) {
852 		ret = -ENOMEM;
853 		btrfs_abort_transaction(trans, ret);
854 		goto out;
855 	}
856 
857 	block_group = btrfs_lookup_block_group(trans->fs_info, start);
858 	if (unlikely(!block_group)) {
859 		DEBUG_WARN("no block group found for start=%llu", start);
860 		ret = -ENOENT;
861 		btrfs_abort_transaction(trans, ret);
862 		goto out;
863 	}
864 
865 	mutex_lock(&block_group->free_space_lock);
866 	ret = __btrfs_remove_from_free_space_tree(trans, block_group, path, start, size);
867 	mutex_unlock(&block_group->free_space_lock);
868 	if (ret)
869 		btrfs_abort_transaction(trans, ret);
870 
871 	btrfs_put_block_group(block_group);
872 out:
873 	btrfs_free_path(path);
874 	return ret;
875 }
876 
add_free_space_extent(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path,u64 start,u64 size)877 static int add_free_space_extent(struct btrfs_trans_handle *trans,
878 				 struct btrfs_block_group *block_group,
879 				 struct btrfs_path *path,
880 				 u64 start, u64 size)
881 {
882 	struct btrfs_root *root = btrfs_free_space_root(block_group);
883 	struct btrfs_key key, new_key;
884 	u64 found_start, found_end;
885 	u64 end = start + size;
886 	int new_extents = 1;
887 	int ret;
888 
889 	/*
890 	 * We are adding a new extent of free space, but we need to merge
891 	 * extents. There are four cases here:
892 	 *
893 	 * 1. The new extent does not have any immediate neighbors to merge
894 	 * with: add the new key and increment the free space extent count. We
895 	 * may need to convert the block group to bitmaps as a result.
896 	 * 2. The new extent has an immediate neighbor before it: remove the
897 	 * previous key and insert a new key combining both of them. There is no
898 	 * net change in the number of extents.
899 	 * 3. The new extent has an immediate neighbor after it: remove the next
900 	 * key and insert a new key combining both of them. There is no net
901 	 * change in the number of extents.
902 	 * 4. The new extent has immediate neighbors on both sides: remove both
903 	 * of the keys and insert a new key combining all of them. Where we used
904 	 * to have two extents, we now have one, so decrement the extent count.
905 	 */
906 
907 	new_key.objectid = start;
908 	new_key.type = BTRFS_FREE_SPACE_EXTENT_KEY;
909 	new_key.offset = size;
910 
911 	/* Search for a neighbor on the left. */
912 	if (start == block_group->start)
913 		goto right;
914 	key.objectid = start - 1;
915 	key.type = (u8)-1;
916 	key.offset = (u64)-1;
917 
918 	ret = btrfs_search_prev_slot(trans, root, &key, path, -1, 1);
919 	if (ret)
920 		return ret;
921 
922 	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
923 
924 	if (key.type != BTRFS_FREE_SPACE_EXTENT_KEY) {
925 		ASSERT(key.type == BTRFS_FREE_SPACE_INFO_KEY);
926 		btrfs_release_path(path);
927 		goto right;
928 	}
929 
930 	found_start = key.objectid;
931 	found_end = key.objectid + key.offset;
932 	ASSERT(found_start >= block_group->start &&
933 	       found_end > block_group->start);
934 	ASSERT(found_start < start && found_end <= start);
935 
936 	/*
937 	 * Delete the neighbor on the left and absorb it into the new key (cases
938 	 * 2 and 4).
939 	 */
940 	if (found_end == start) {
941 		ret = btrfs_del_item(trans, root, path);
942 		if (ret)
943 			return ret;
944 		new_key.objectid = found_start;
945 		new_key.offset += key.offset;
946 		new_extents--;
947 	}
948 	btrfs_release_path(path);
949 
950 right:
951 	/* Search for a neighbor on the right. */
952 	if (end == block_group->start + block_group->length)
953 		goto insert;
954 	key.objectid = end;
955 	key.type = (u8)-1;
956 	key.offset = (u64)-1;
957 
958 	ret = btrfs_search_prev_slot(trans, root, &key, path, -1, 1);
959 	if (ret)
960 		return ret;
961 
962 	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
963 
964 	if (key.type != BTRFS_FREE_SPACE_EXTENT_KEY) {
965 		ASSERT(key.type == BTRFS_FREE_SPACE_INFO_KEY);
966 		btrfs_release_path(path);
967 		goto insert;
968 	}
969 
970 	found_start = key.objectid;
971 	found_end = key.objectid + key.offset;
972 	ASSERT(found_start >= block_group->start &&
973 	       found_end > block_group->start);
974 	ASSERT((found_start < start && found_end <= start) ||
975 	       (found_start >= end && found_end > end));
976 
977 	/*
978 	 * Delete the neighbor on the right and absorb it into the new key
979 	 * (cases 3 and 4).
980 	 */
981 	if (found_start == end) {
982 		ret = btrfs_del_item(trans, root, path);
983 		if (ret)
984 			return ret;
985 		new_key.offset += key.offset;
986 		new_extents--;
987 	}
988 	btrfs_release_path(path);
989 
990 insert:
991 	/* Insert the new key (cases 1-4). */
992 	ret = btrfs_insert_empty_item(trans, root, path, &new_key, 0);
993 	if (ret)
994 		return ret;
995 
996 	btrfs_release_path(path);
997 	return update_free_space_extent_count(trans, block_group, path, new_extents);
998 }
999 
1000 EXPORT_FOR_TESTS
__btrfs_add_to_free_space_tree(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path,u64 start,u64 size)1001 int __btrfs_add_to_free_space_tree(struct btrfs_trans_handle *trans,
1002 				   struct btrfs_block_group *block_group,
1003 				   struct btrfs_path *path, u64 start, u64 size)
1004 {
1005 	int ret;
1006 
1007 	ret = __add_block_group_free_space(trans, block_group, path);
1008 	if (ret)
1009 		return ret;
1010 
1011 	ret = using_bitmaps(block_group, path);
1012 	if (ret < 0)
1013 		return ret;
1014 
1015 	if (ret)
1016 		return modify_free_space_bitmap(trans, block_group, path,
1017 						start, size, false);
1018 
1019 	return add_free_space_extent(trans, block_group, path, start, size);
1020 }
1021 
btrfs_add_to_free_space_tree(struct btrfs_trans_handle * trans,u64 start,u64 size)1022 int btrfs_add_to_free_space_tree(struct btrfs_trans_handle *trans,
1023 				 u64 start, u64 size)
1024 {
1025 	struct btrfs_block_group *block_group;
1026 	struct btrfs_path *path;
1027 	int ret;
1028 
1029 	if (!btrfs_fs_compat_ro(trans->fs_info, FREE_SPACE_TREE))
1030 		return 0;
1031 
1032 	path = btrfs_alloc_path();
1033 	if (unlikely(!path)) {
1034 		ret = -ENOMEM;
1035 		btrfs_abort_transaction(trans, ret);
1036 		goto out;
1037 	}
1038 
1039 	block_group = btrfs_lookup_block_group(trans->fs_info, start);
1040 	if (unlikely(!block_group)) {
1041 		DEBUG_WARN("no block group found for start=%llu", start);
1042 		ret = -ENOENT;
1043 		btrfs_abort_transaction(trans, ret);
1044 		goto out;
1045 	}
1046 
1047 	mutex_lock(&block_group->free_space_lock);
1048 	ret = __btrfs_add_to_free_space_tree(trans, block_group, path, start, size);
1049 	mutex_unlock(&block_group->free_space_lock);
1050 	if (ret)
1051 		btrfs_abort_transaction(trans, ret);
1052 
1053 	btrfs_put_block_group(block_group);
1054 out:
1055 	btrfs_free_path(path);
1056 	return ret;
1057 }
1058 
1059 /*
1060  * Populate the free space tree by walking the extent tree. Operations on the
1061  * extent tree that happen as a result of writes to the free space tree will go
1062  * through the normal add/remove hooks.
1063  */
populate_free_space_tree(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group)1064 static int populate_free_space_tree(struct btrfs_trans_handle *trans,
1065 				    struct btrfs_block_group *block_group)
1066 {
1067 	struct btrfs_root *extent_root;
1068 	BTRFS_PATH_AUTO_FREE(path);
1069 	BTRFS_PATH_AUTO_FREE(path2);
1070 	struct btrfs_key key;
1071 	u64 start, end;
1072 	int ret;
1073 
1074 	path = btrfs_alloc_path();
1075 	if (!path)
1076 		return -ENOMEM;
1077 
1078 	path2 = btrfs_alloc_path();
1079 	if (!path2)
1080 		return -ENOMEM;
1081 
1082 	path->reada = READA_FORWARD;
1083 
1084 	ret = add_new_free_space_info(trans, block_group, path2);
1085 	if (ret)
1086 		return ret;
1087 
1088 	mutex_lock(&block_group->free_space_lock);
1089 
1090 	/*
1091 	 * Iterate through all of the extent and metadata items in this block
1092 	 * group, adding the free space between them and the free space at the
1093 	 * end. Note that EXTENT_ITEM and METADATA_ITEM are less than
1094 	 * BLOCK_GROUP_ITEM, so an extent may precede the block group that it's
1095 	 * contained in.
1096 	 */
1097 	key.objectid = block_group->start;
1098 	key.type = BTRFS_EXTENT_ITEM_KEY;
1099 	key.offset = 0;
1100 
1101 	extent_root = btrfs_extent_root(trans->fs_info, key.objectid);
1102 	ret = btrfs_search_slot_for_read(extent_root, &key, path, 1, 0);
1103 	if (ret < 0)
1104 		goto out_locked;
1105 	/*
1106 	 * If ret is 1 (no key found), it means this is an empty block group,
1107 	 * without any extents allocated from it and there's no block group
1108 	 * item (key BTRFS_BLOCK_GROUP_ITEM_KEY) located in the extent tree
1109 	 * because we are using the block group tree feature (so block group
1110 	 * items are stored in the block group tree) or this is a new block
1111 	 * group created in the current transaction and its block group item
1112 	 * was not yet inserted in the extent tree (that happens in
1113 	 * btrfs_create_pending_block_groups() -> insert_block_group_item()).
1114 	 * It also means there are no extents allocated for block groups with a
1115 	 * start offset beyond this block group's end offset (this is the last,
1116 	 * highest, block group).
1117 	 */
1118 	start = block_group->start;
1119 	end = block_group->start + block_group->length;
1120 	while (ret == 0) {
1121 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1122 
1123 		if (key.type == BTRFS_EXTENT_ITEM_KEY ||
1124 		    key.type == BTRFS_METADATA_ITEM_KEY) {
1125 			if (key.objectid >= end)
1126 				break;
1127 
1128 			if (start < key.objectid) {
1129 				ret = __btrfs_add_to_free_space_tree(trans,
1130 								     block_group,
1131 								     path2, start,
1132 								     key.objectid -
1133 								     start);
1134 				if (ret)
1135 					goto out_locked;
1136 			}
1137 			start = key.objectid;
1138 			if (key.type == BTRFS_METADATA_ITEM_KEY)
1139 				start += trans->fs_info->nodesize;
1140 			else
1141 				start += key.offset;
1142 		} else if (key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
1143 			if (key.objectid != block_group->start)
1144 				break;
1145 		}
1146 
1147 		ret = btrfs_next_item(extent_root, path);
1148 		if (ret < 0)
1149 			goto out_locked;
1150 	}
1151 	if (start < end) {
1152 		ret = __btrfs_add_to_free_space_tree(trans, block_group, path2,
1153 						     start, end - start);
1154 		if (ret)
1155 			goto out_locked;
1156 	}
1157 
1158 	ret = 0;
1159 out_locked:
1160 	mutex_unlock(&block_group->free_space_lock);
1161 
1162 	return ret;
1163 }
1164 
btrfs_create_free_space_tree(struct btrfs_fs_info * fs_info)1165 int btrfs_create_free_space_tree(struct btrfs_fs_info *fs_info)
1166 {
1167 	struct btrfs_trans_handle *trans;
1168 	struct btrfs_root *tree_root = fs_info->tree_root;
1169 	struct btrfs_root *free_space_root;
1170 	struct btrfs_block_group *block_group;
1171 	struct rb_node *node;
1172 	int ret;
1173 
1174 	trans = btrfs_start_transaction(tree_root, 0);
1175 	if (IS_ERR(trans))
1176 		return PTR_ERR(trans);
1177 
1178 	set_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags);
1179 	set_bit(BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED, &fs_info->flags);
1180 	free_space_root = btrfs_create_tree(trans,
1181 					    BTRFS_FREE_SPACE_TREE_OBJECTID);
1182 	if (IS_ERR(free_space_root)) {
1183 		ret = PTR_ERR(free_space_root);
1184 		btrfs_abort_transaction(trans, ret);
1185 		btrfs_end_transaction(trans);
1186 		goto out_clear;
1187 	}
1188 	ret = btrfs_global_root_insert(free_space_root);
1189 	if (unlikely(ret)) {
1190 		btrfs_put_root(free_space_root);
1191 		btrfs_abort_transaction(trans, ret);
1192 		btrfs_end_transaction(trans);
1193 		goto out_clear;
1194 	}
1195 
1196 	node = rb_first_cached(&fs_info->block_group_cache_tree);
1197 	while (node) {
1198 		block_group = rb_entry(node, struct btrfs_block_group,
1199 				       cache_node);
1200 		ret = populate_free_space_tree(trans, block_group);
1201 		if (unlikely(ret)) {
1202 			btrfs_abort_transaction(trans, ret);
1203 			btrfs_end_transaction(trans);
1204 			goto out_clear;
1205 		}
1206 		node = rb_next(node);
1207 	}
1208 
1209 	btrfs_set_fs_compat_ro(fs_info, FREE_SPACE_TREE);
1210 	btrfs_set_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID);
1211 	clear_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags);
1212 	ret = btrfs_commit_transaction(trans);
1213 
1214 	/*
1215 	 * Now that we've committed the transaction any reading of our commit
1216 	 * root will be safe, so we can cache from the free space tree now.
1217 	 */
1218 	clear_bit(BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED, &fs_info->flags);
1219 	return ret;
1220 
1221 out_clear:
1222 	clear_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags);
1223 	clear_bit(BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED, &fs_info->flags);
1224 	return ret;
1225 }
1226 
clear_free_space_tree(struct btrfs_trans_handle * trans,struct btrfs_root * root)1227 static int clear_free_space_tree(struct btrfs_trans_handle *trans,
1228 				 struct btrfs_root *root)
1229 {
1230 	BTRFS_PATH_AUTO_FREE(path);
1231 	struct btrfs_key key;
1232 	struct rb_node *node;
1233 	int nr;
1234 	int ret;
1235 
1236 	path = btrfs_alloc_path();
1237 	if (!path)
1238 		return -ENOMEM;
1239 
1240 	key.objectid = 0;
1241 	key.type = 0;
1242 	key.offset = 0;
1243 
1244 	while (1) {
1245 		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1246 		if (ret < 0)
1247 			return ret;
1248 
1249 		nr = btrfs_header_nritems(path->nodes[0]);
1250 		if (!nr)
1251 			break;
1252 
1253 		path->slots[0] = 0;
1254 		ret = btrfs_del_items(trans, root, path, 0, nr);
1255 		if (ret)
1256 			return ret;
1257 
1258 		btrfs_release_path(path);
1259 	}
1260 
1261 	node = rb_first_cached(&trans->fs_info->block_group_cache_tree);
1262 	while (node) {
1263 		struct btrfs_block_group *bg;
1264 
1265 		bg = rb_entry(node, struct btrfs_block_group, cache_node);
1266 		clear_bit(BLOCK_GROUP_FLAG_FREE_SPACE_ADDED, &bg->runtime_flags);
1267 		node = rb_next(node);
1268 		cond_resched();
1269 	}
1270 
1271 	return 0;
1272 }
1273 
btrfs_delete_free_space_tree(struct btrfs_fs_info * fs_info)1274 int btrfs_delete_free_space_tree(struct btrfs_fs_info *fs_info)
1275 {
1276 	struct btrfs_trans_handle *trans;
1277 	struct btrfs_root *tree_root = fs_info->tree_root;
1278 	struct btrfs_key key = {
1279 		.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID,
1280 		.type = BTRFS_ROOT_ITEM_KEY,
1281 		.offset = 0,
1282 	};
1283 	struct btrfs_root *free_space_root = btrfs_global_root(fs_info, &key);
1284 	int ret;
1285 
1286 	trans = btrfs_start_transaction(tree_root, 0);
1287 	if (IS_ERR(trans))
1288 		return PTR_ERR(trans);
1289 
1290 	btrfs_clear_fs_compat_ro(fs_info, FREE_SPACE_TREE);
1291 	btrfs_clear_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID);
1292 
1293 	ret = clear_free_space_tree(trans, free_space_root);
1294 	if (unlikely(ret)) {
1295 		btrfs_abort_transaction(trans, ret);
1296 		btrfs_end_transaction(trans);
1297 		return ret;
1298 	}
1299 
1300 	ret = btrfs_del_root(trans, &free_space_root->root_key);
1301 	if (unlikely(ret)) {
1302 		btrfs_abort_transaction(trans, ret);
1303 		btrfs_end_transaction(trans);
1304 		return ret;
1305 	}
1306 
1307 	btrfs_global_root_delete(free_space_root);
1308 
1309 	spin_lock(&fs_info->trans_lock);
1310 	list_del(&free_space_root->dirty_list);
1311 	spin_unlock(&fs_info->trans_lock);
1312 
1313 	btrfs_tree_lock(free_space_root->node);
1314 	btrfs_clear_buffer_dirty(trans, free_space_root->node);
1315 	btrfs_tree_unlock(free_space_root->node);
1316 	ret = btrfs_free_tree_block(trans, btrfs_root_id(free_space_root),
1317 				    free_space_root->node, 0, 1);
1318 	btrfs_put_root(free_space_root);
1319 	if (unlikely(ret < 0)) {
1320 		btrfs_abort_transaction(trans, ret);
1321 		btrfs_end_transaction(trans);
1322 		return ret;
1323 	}
1324 
1325 	return btrfs_commit_transaction(trans);
1326 }
1327 
btrfs_rebuild_free_space_tree(struct btrfs_fs_info * fs_info)1328 int btrfs_rebuild_free_space_tree(struct btrfs_fs_info *fs_info)
1329 {
1330 	struct btrfs_trans_handle *trans;
1331 	struct btrfs_key key = {
1332 		.objectid = BTRFS_FREE_SPACE_TREE_OBJECTID,
1333 		.type = BTRFS_ROOT_ITEM_KEY,
1334 		.offset = 0,
1335 	};
1336 	struct btrfs_root *free_space_root = btrfs_global_root(fs_info, &key);
1337 	struct rb_node *node;
1338 	int ret;
1339 
1340 	trans = btrfs_start_transaction(free_space_root, 1);
1341 	if (IS_ERR(trans))
1342 		return PTR_ERR(trans);
1343 
1344 	set_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags);
1345 	set_bit(BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED, &fs_info->flags);
1346 
1347 	ret = clear_free_space_tree(trans, free_space_root);
1348 	if (unlikely(ret)) {
1349 		btrfs_abort_transaction(trans, ret);
1350 		btrfs_end_transaction(trans);
1351 		return ret;
1352 	}
1353 
1354 	node = rb_first_cached(&fs_info->block_group_cache_tree);
1355 	while (node) {
1356 		struct btrfs_block_group *block_group;
1357 
1358 		block_group = rb_entry(node, struct btrfs_block_group,
1359 				       cache_node);
1360 
1361 		if (test_bit(BLOCK_GROUP_FLAG_FREE_SPACE_ADDED,
1362 			     &block_group->runtime_flags))
1363 			goto next;
1364 
1365 		ret = populate_free_space_tree(trans, block_group);
1366 		if (unlikely(ret)) {
1367 			btrfs_abort_transaction(trans, ret);
1368 			btrfs_end_transaction(trans);
1369 			return ret;
1370 		}
1371 next:
1372 		if (btrfs_should_end_transaction(trans)) {
1373 			btrfs_end_transaction(trans);
1374 			trans = btrfs_start_transaction(free_space_root, 1);
1375 			if (IS_ERR(trans))
1376 				return PTR_ERR(trans);
1377 		}
1378 		node = rb_next(node);
1379 	}
1380 
1381 	btrfs_set_fs_compat_ro(fs_info, FREE_SPACE_TREE);
1382 	btrfs_set_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID);
1383 	clear_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags);
1384 
1385 	ret = btrfs_commit_transaction(trans);
1386 	clear_bit(BTRFS_FS_FREE_SPACE_TREE_UNTRUSTED, &fs_info->flags);
1387 	return ret;
1388 }
1389 
__add_block_group_free_space(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group,struct btrfs_path * path)1390 static int __add_block_group_free_space(struct btrfs_trans_handle *trans,
1391 					struct btrfs_block_group *block_group,
1392 					struct btrfs_path *path)
1393 {
1394 	bool own_path = false;
1395 	int ret;
1396 
1397 	if (!test_and_clear_bit(BLOCK_GROUP_FLAG_NEEDS_FREE_SPACE,
1398 				&block_group->runtime_flags))
1399 		return 0;
1400 
1401 	/*
1402 	 * While rebuilding the free space tree we may allocate new metadata
1403 	 * block groups while modifying the free space tree.
1404 	 *
1405 	 * Because during the rebuild (at btrfs_rebuild_free_space_tree()) we
1406 	 * can use multiple transactions, every time btrfs_end_transaction() is
1407 	 * called at btrfs_rebuild_free_space_tree() we finish the creation of
1408 	 * new block groups by calling btrfs_create_pending_block_groups(), and
1409 	 * that in turn calls us, through add_block_group_free_space(), to add
1410 	 * a free space info item and a free space extent item for the block
1411 	 * group.
1412 	 *
1413 	 * Then later btrfs_rebuild_free_space_tree() may find such new block
1414 	 * groups and processes them with populate_free_space_tree(), which can
1415 	 * fail with EEXIST since there are already items for the block group in
1416 	 * the free space tree. Notice that we say "may find" because a new
1417 	 * block group may be added to the block groups rbtree in a node before
1418 	 * or after the block group currently being processed by the rebuild
1419 	 * process. So signal the rebuild process to skip such new block groups
1420 	 * if it finds them.
1421 	 */
1422 	set_bit(BLOCK_GROUP_FLAG_FREE_SPACE_ADDED, &block_group->runtime_flags);
1423 
1424 	if (!path) {
1425 		path = btrfs_alloc_path();
1426 		if (unlikely(!path)) {
1427 			btrfs_abort_transaction(trans, -ENOMEM);
1428 			return -ENOMEM;
1429 		}
1430 		own_path = true;
1431 	}
1432 
1433 	ret = add_new_free_space_info(trans, block_group, path);
1434 	if (unlikely(ret)) {
1435 		btrfs_abort_transaction(trans, ret);
1436 		goto out;
1437 	}
1438 
1439 	ret = __btrfs_add_to_free_space_tree(trans, block_group, path,
1440 					     block_group->start, block_group->length);
1441 	if (ret)
1442 		btrfs_abort_transaction(trans, ret);
1443 
1444 out:
1445 	if (own_path)
1446 		btrfs_free_path(path);
1447 
1448 	return ret;
1449 }
1450 
btrfs_add_block_group_free_space(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group)1451 int btrfs_add_block_group_free_space(struct btrfs_trans_handle *trans,
1452 				     struct btrfs_block_group *block_group)
1453 {
1454 	int ret;
1455 
1456 	if (!btrfs_fs_compat_ro(trans->fs_info, FREE_SPACE_TREE))
1457 		return 0;
1458 
1459 	mutex_lock(&block_group->free_space_lock);
1460 	ret = __add_block_group_free_space(trans, block_group, NULL);
1461 	mutex_unlock(&block_group->free_space_lock);
1462 	return ret;
1463 }
1464 
btrfs_remove_block_group_free_space(struct btrfs_trans_handle * trans,struct btrfs_block_group * block_group)1465 int btrfs_remove_block_group_free_space(struct btrfs_trans_handle *trans,
1466 					struct btrfs_block_group *block_group)
1467 {
1468 	struct btrfs_root *root = btrfs_free_space_root(block_group);
1469 	struct btrfs_path *path;
1470 	struct btrfs_key key, found_key;
1471 	struct extent_buffer *leaf;
1472 	u64 start, end;
1473 	int done = 0, nr;
1474 	int ret;
1475 
1476 	if (!btrfs_fs_compat_ro(trans->fs_info, FREE_SPACE_TREE))
1477 		return 0;
1478 
1479 	if (test_bit(BLOCK_GROUP_FLAG_NEEDS_FREE_SPACE, &block_group->runtime_flags)) {
1480 		/* We never added this block group to the free space tree. */
1481 		return 0;
1482 	}
1483 
1484 	path = btrfs_alloc_path();
1485 	if (unlikely(!path)) {
1486 		ret = -ENOMEM;
1487 		btrfs_abort_transaction(trans, ret);
1488 		goto out;
1489 	}
1490 
1491 	start = block_group->start;
1492 	end = block_group->start + block_group->length;
1493 
1494 	key.objectid = end - 1;
1495 	key.type = (u8)-1;
1496 	key.offset = (u64)-1;
1497 
1498 	while (!done) {
1499 		ret = btrfs_search_prev_slot(trans, root, &key, path, -1, 1);
1500 		if (unlikely(ret)) {
1501 			btrfs_abort_transaction(trans, ret);
1502 			goto out;
1503 		}
1504 
1505 		leaf = path->nodes[0];
1506 		nr = 0;
1507 		path->slots[0]++;
1508 		while (path->slots[0] > 0) {
1509 			btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0] - 1);
1510 
1511 			if (found_key.type == BTRFS_FREE_SPACE_INFO_KEY) {
1512 				ASSERT(found_key.objectid == block_group->start);
1513 				ASSERT(found_key.offset == block_group->length);
1514 				done = 1;
1515 				nr++;
1516 				path->slots[0]--;
1517 				break;
1518 			} else if (found_key.type == BTRFS_FREE_SPACE_EXTENT_KEY ||
1519 				   found_key.type == BTRFS_FREE_SPACE_BITMAP_KEY) {
1520 				ASSERT(found_key.objectid >= start);
1521 				ASSERT(found_key.objectid < end);
1522 				ASSERT(found_key.objectid + found_key.offset <= end);
1523 				nr++;
1524 				path->slots[0]--;
1525 			} else {
1526 				ASSERT(0);
1527 			}
1528 		}
1529 
1530 		ret = btrfs_del_items(trans, root, path, path->slots[0], nr);
1531 		if (unlikely(ret)) {
1532 			btrfs_abort_transaction(trans, ret);
1533 			goto out;
1534 		}
1535 		btrfs_release_path(path);
1536 	}
1537 
1538 	ret = 0;
1539 out:
1540 	btrfs_free_path(path);
1541 	return ret;
1542 }
1543 
load_free_space_bitmaps(struct btrfs_caching_control * caching_ctl,struct btrfs_path * path,u32 expected_extent_count)1544 static int load_free_space_bitmaps(struct btrfs_caching_control *caching_ctl,
1545 				   struct btrfs_path *path,
1546 				   u32 expected_extent_count)
1547 {
1548 	struct btrfs_block_group *block_group;
1549 	struct btrfs_fs_info *fs_info;
1550 	struct btrfs_root *root;
1551 	struct btrfs_key key;
1552 	bool prev_bit_set = false;
1553 	/* Initialize to silence GCC. */
1554 	u64 extent_start = 0;
1555 	u64 end, offset;
1556 	u64 total_found = 0;
1557 	u32 extent_count = 0;
1558 	int ret;
1559 
1560 	block_group = caching_ctl->block_group;
1561 	fs_info = block_group->fs_info;
1562 	root = btrfs_free_space_root(block_group);
1563 
1564 	end = block_group->start + block_group->length;
1565 
1566 	while (1) {
1567 		ret = btrfs_next_item(root, path);
1568 		if (ret < 0)
1569 			return ret;
1570 		if (ret)
1571 			break;
1572 
1573 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1574 
1575 		if (key.type == BTRFS_FREE_SPACE_INFO_KEY)
1576 			break;
1577 
1578 		ASSERT(key.type == BTRFS_FREE_SPACE_BITMAP_KEY);
1579 		ASSERT(key.objectid < end && key.objectid + key.offset <= end);
1580 
1581 		offset = key.objectid;
1582 		while (offset < key.objectid + key.offset) {
1583 			bool bit_set;
1584 
1585 			bit_set = btrfs_free_space_test_bit(block_group, path, offset);
1586 			if (!prev_bit_set && bit_set) {
1587 				extent_start = offset;
1588 			} else if (prev_bit_set && !bit_set) {
1589 				u64 space_added;
1590 
1591 				ret = btrfs_add_new_free_space(block_group,
1592 							       extent_start,
1593 							       offset,
1594 							       &space_added);
1595 				if (ret)
1596 					return ret;
1597 				total_found += space_added;
1598 				if (total_found > CACHING_CTL_WAKE_UP) {
1599 					total_found = 0;
1600 					wake_up(&caching_ctl->wait);
1601 				}
1602 				extent_count++;
1603 			}
1604 			prev_bit_set = bit_set;
1605 			offset += fs_info->sectorsize;
1606 		}
1607 	}
1608 	if (prev_bit_set) {
1609 		ret = btrfs_add_new_free_space(block_group, extent_start, end, NULL);
1610 		if (ret)
1611 			return ret;
1612 		extent_count++;
1613 	}
1614 
1615 	if (unlikely(extent_count != expected_extent_count)) {
1616 		btrfs_err(fs_info,
1617 			  "incorrect extent count for %llu; counted %u, expected %u",
1618 			  block_group->start, extent_count,
1619 			  expected_extent_count);
1620 		DEBUG_WARN();
1621 		return -EIO;
1622 	}
1623 
1624 	return 0;
1625 }
1626 
load_free_space_extents(struct btrfs_caching_control * caching_ctl,struct btrfs_path * path,u32 expected_extent_count)1627 static int load_free_space_extents(struct btrfs_caching_control *caching_ctl,
1628 				   struct btrfs_path *path,
1629 				   u32 expected_extent_count)
1630 {
1631 	struct btrfs_block_group *block_group;
1632 	struct btrfs_fs_info *fs_info;
1633 	struct btrfs_root *root;
1634 	struct btrfs_key key;
1635 	u64 end;
1636 	u64 total_found = 0;
1637 	u32 extent_count = 0;
1638 	int ret;
1639 
1640 	block_group = caching_ctl->block_group;
1641 	fs_info = block_group->fs_info;
1642 	root = btrfs_free_space_root(block_group);
1643 
1644 	end = block_group->start + block_group->length;
1645 
1646 	while (1) {
1647 		u64 space_added;
1648 
1649 		ret = btrfs_next_item(root, path);
1650 		if (ret < 0)
1651 			return ret;
1652 		if (ret)
1653 			break;
1654 
1655 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1656 
1657 		if (key.type == BTRFS_FREE_SPACE_INFO_KEY)
1658 			break;
1659 
1660 		ASSERT(key.type == BTRFS_FREE_SPACE_EXTENT_KEY);
1661 		ASSERT(key.objectid < end && key.objectid + key.offset <= end);
1662 
1663 		ret = btrfs_add_new_free_space(block_group, key.objectid,
1664 					       key.objectid + key.offset,
1665 					       &space_added);
1666 		if (ret)
1667 			return ret;
1668 		total_found += space_added;
1669 		if (total_found > CACHING_CTL_WAKE_UP) {
1670 			total_found = 0;
1671 			wake_up(&caching_ctl->wait);
1672 		}
1673 		extent_count++;
1674 	}
1675 
1676 	if (unlikely(extent_count != expected_extent_count)) {
1677 		btrfs_err(fs_info,
1678 			  "incorrect extent count for %llu; counted %u, expected %u",
1679 			  block_group->start, extent_count,
1680 			  expected_extent_count);
1681 		DEBUG_WARN();
1682 		return -EIO;
1683 	}
1684 
1685 	return 0;
1686 }
1687 
btrfs_load_free_space_tree(struct btrfs_caching_control * caching_ctl)1688 int btrfs_load_free_space_tree(struct btrfs_caching_control *caching_ctl)
1689 {
1690 	struct btrfs_block_group *block_group;
1691 	struct btrfs_free_space_info *info;
1692 	BTRFS_PATH_AUTO_FREE(path);
1693 	u32 extent_count, flags;
1694 
1695 	block_group = caching_ctl->block_group;
1696 
1697 	path = btrfs_alloc_path();
1698 	if (!path)
1699 		return -ENOMEM;
1700 
1701 	/*
1702 	 * Just like caching_thread() doesn't want to deadlock on the extent
1703 	 * tree, we don't want to deadlock on the free space tree.
1704 	 */
1705 	path->skip_locking = 1;
1706 	path->search_commit_root = 1;
1707 	path->reada = READA_FORWARD;
1708 
1709 	info = btrfs_search_free_space_info(NULL, block_group, path, 0);
1710 	if (IS_ERR(info))
1711 		return PTR_ERR(info);
1712 
1713 	extent_count = btrfs_free_space_extent_count(path->nodes[0], info);
1714 	flags = btrfs_free_space_flags(path->nodes[0], info);
1715 
1716 	/*
1717 	 * We left path pointing to the free space info item, so now
1718 	 * load_free_space_foo can just iterate through the free space tree from
1719 	 * there.
1720 	 */
1721 	if (flags & BTRFS_FREE_SPACE_USING_BITMAPS)
1722 		return load_free_space_bitmaps(caching_ctl, path, extent_count);
1723 	else
1724 		return load_free_space_extents(caching_ctl, path, extent_count);
1725 }
1726