Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. 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 | ||
ec6b910f | 19 | #include <linux/sched.h> |
7bb86316 | 20 | #include <linux/crc32c.h> |
edbd8d4e | 21 | #include <linux/pagemap.h> |
74493f7a | 22 | #include "hash.h" |
fec577fb CM |
23 | #include "ctree.h" |
24 | #include "disk-io.h" | |
25 | #include "print-tree.h" | |
e089f05c | 26 | #include "transaction.h" |
fec577fb | 27 | |
96b5179d CM |
28 | #define BLOCK_GROUP_DATA EXTENT_WRITEBACK |
29 | #define BLOCK_GROUP_METADATA EXTENT_UPTODATE | |
30 | #define BLOCK_GROUP_DIRTY EXTENT_DIRTY | |
31 | ||
e089f05c CM |
32 | static int finish_current_insert(struct btrfs_trans_handle *trans, struct |
33 | btrfs_root *extent_root); | |
e20d96d6 CM |
34 | static int del_pending_extents(struct btrfs_trans_handle *trans, struct |
35 | btrfs_root *extent_root); | |
d548ee51 Y |
36 | static int find_previous_extent(struct btrfs_root *root, |
37 | struct btrfs_path *path) | |
38 | { | |
39 | struct btrfs_key found_key; | |
40 | struct extent_buffer *leaf; | |
41 | int ret; | |
42 | ||
43 | while(1) { | |
44 | if (path->slots[0] == 0) { | |
45 | ret = btrfs_prev_leaf(root, path); | |
46 | if (ret != 0) | |
47 | return ret; | |
48 | } else { | |
49 | path->slots[0]--; | |
50 | } | |
51 | leaf = path->nodes[0]; | |
52 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
53 | if (found_key.type == BTRFS_EXTENT_ITEM_KEY) | |
54 | return 0; | |
55 | } | |
56 | return 1; | |
57 | } | |
fec577fb | 58 | |
e37c9e69 CM |
59 | static int cache_block_group(struct btrfs_root *root, |
60 | struct btrfs_block_group_cache *block_group) | |
61 | { | |
62 | struct btrfs_path *path; | |
63 | int ret; | |
64 | struct btrfs_key key; | |
5f39d397 | 65 | struct extent_buffer *leaf; |
f510cfec | 66 | struct extent_map_tree *free_space_cache; |
e37c9e69 | 67 | int slot; |
e37c9e69 CM |
68 | u64 last = 0; |
69 | u64 hole_size; | |
7d7d6068 | 70 | u64 first_free; |
e37c9e69 CM |
71 | int found = 0; |
72 | ||
00f5c795 CM |
73 | if (!block_group) |
74 | return 0; | |
75 | ||
e37c9e69 | 76 | root = root->fs_info->extent_root; |
f510cfec | 77 | free_space_cache = &root->fs_info->free_space_cache; |
e37c9e69 CM |
78 | |
79 | if (block_group->cached) | |
80 | return 0; | |
f510cfec | 81 | |
e37c9e69 CM |
82 | path = btrfs_alloc_path(); |
83 | if (!path) | |
84 | return -ENOMEM; | |
7d7d6068 | 85 | |
2cc58cf2 | 86 | path->reada = 2; |
7d7d6068 | 87 | first_free = block_group->key.objectid; |
e37c9e69 | 88 | key.objectid = block_group->key.objectid; |
e37c9e69 CM |
89 | key.offset = 0; |
90 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); | |
91 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
92 | if (ret < 0) | |
93 | return ret; | |
d548ee51 Y |
94 | ret = find_previous_extent(root, path); |
95 | if (ret < 0) | |
96 | return ret; | |
97 | if (ret == 0) { | |
98 | leaf = path->nodes[0]; | |
99 | btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); | |
100 | if (key.objectid + key.offset > first_free) | |
101 | first_free = key.objectid + key.offset; | |
102 | } | |
e37c9e69 | 103 | while(1) { |
5f39d397 | 104 | leaf = path->nodes[0]; |
e37c9e69 | 105 | slot = path->slots[0]; |
5f39d397 | 106 | if (slot >= btrfs_header_nritems(leaf)) { |
e37c9e69 | 107 | ret = btrfs_next_leaf(root, path); |
54aa1f4d CM |
108 | if (ret < 0) |
109 | goto err; | |
de428b63 | 110 | if (ret == 0) { |
e37c9e69 | 111 | continue; |
de428b63 | 112 | } else { |
e37c9e69 CM |
113 | break; |
114 | } | |
115 | } | |
5f39d397 | 116 | btrfs_item_key_to_cpu(leaf, &key, slot); |
7d7d6068 | 117 | if (key.objectid < block_group->key.objectid) { |
7d7d6068 Y |
118 | goto next; |
119 | } | |
e37c9e69 CM |
120 | if (key.objectid >= block_group->key.objectid + |
121 | block_group->key.offset) { | |
e37c9e69 CM |
122 | break; |
123 | } | |
7d7d6068 | 124 | |
e37c9e69 CM |
125 | if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) { |
126 | if (!found) { | |
7d7d6068 | 127 | last = first_free; |
e37c9e69 | 128 | found = 1; |
e37c9e69 | 129 | } |
f510cfec CM |
130 | if (key.objectid > last) { |
131 | hole_size = key.objectid - last; | |
132 | set_extent_dirty(free_space_cache, last, | |
133 | last + hole_size - 1, | |
134 | GFP_NOFS); | |
7d7d6068 Y |
135 | } |
136 | last = key.objectid + key.offset; | |
e37c9e69 | 137 | } |
7d7d6068 | 138 | next: |
e37c9e69 CM |
139 | path->slots[0]++; |
140 | } | |
141 | ||
7d7d6068 Y |
142 | if (!found) |
143 | last = first_free; | |
144 | if (block_group->key.objectid + | |
145 | block_group->key.offset > last) { | |
146 | hole_size = block_group->key.objectid + | |
147 | block_group->key.offset - last; | |
f510cfec CM |
148 | set_extent_dirty(free_space_cache, last, |
149 | last + hole_size - 1, GFP_NOFS); | |
7d7d6068 | 150 | } |
e37c9e69 | 151 | block_group->cached = 1; |
54aa1f4d | 152 | err: |
e37c9e69 CM |
153 | btrfs_free_path(path); |
154 | return 0; | |
155 | } | |
156 | ||
5276aeda CM |
157 | struct btrfs_block_group_cache *btrfs_lookup_block_group(struct |
158 | btrfs_fs_info *info, | |
db94535d | 159 | u64 bytenr) |
be744175 | 160 | { |
96b5179d CM |
161 | struct extent_map_tree *block_group_cache; |
162 | struct btrfs_block_group_cache *block_group = NULL; | |
163 | u64 ptr; | |
164 | u64 start; | |
165 | u64 end; | |
be744175 CM |
166 | int ret; |
167 | ||
96b5179d CM |
168 | block_group_cache = &info->block_group_cache; |
169 | ret = find_first_extent_bit(block_group_cache, | |
db94535d | 170 | bytenr, &start, &end, |
96b5179d | 171 | BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA); |
be744175 | 172 | if (ret) { |
96b5179d | 173 | return NULL; |
be744175 | 174 | } |
96b5179d CM |
175 | ret = get_state_private(block_group_cache, start, &ptr); |
176 | if (ret) | |
177 | return NULL; | |
178 | ||
ae2f5411 | 179 | block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr; |
5cf66426 | 180 | if (block_group->key.objectid <= bytenr && bytenr < |
96b5179d CM |
181 | block_group->key.objectid + block_group->key.offset) |
182 | return block_group; | |
be744175 CM |
183 | return NULL; |
184 | } | |
98ed5174 CM |
185 | static u64 noinline find_search_start(struct btrfs_root *root, |
186 | struct btrfs_block_group_cache **cache_ret, | |
d548ee51 | 187 | u64 search_start, int num, int data) |
e37c9e69 | 188 | { |
e37c9e69 CM |
189 | int ret; |
190 | struct btrfs_block_group_cache *cache = *cache_ret; | |
e19caa5f | 191 | u64 last; |
f510cfec CM |
192 | u64 start = 0; |
193 | u64 end = 0; | |
257d0ce3 | 194 | u64 cache_miss = 0; |
c31f8830 | 195 | u64 total_fs_bytes; |
f84a8b36 | 196 | int wrapped = 0; |
e37c9e69 | 197 | |
00f5c795 | 198 | if (!cache) { |
1a2b2ac7 | 199 | goto out; |
00f5c795 | 200 | } |
c31f8830 | 201 | total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); |
e37c9e69 | 202 | again: |
54aa1f4d CM |
203 | ret = cache_block_group(root, cache); |
204 | if (ret) | |
205 | goto out; | |
f84a8b36 | 206 | |
e19caa5f CM |
207 | last = max(search_start, cache->key.objectid); |
208 | ||
e37c9e69 | 209 | while(1) { |
f510cfec CM |
210 | ret = find_first_extent_bit(&root->fs_info->free_space_cache, |
211 | last, &start, &end, EXTENT_DIRTY); | |
e19caa5f | 212 | if (ret) { |
257d0ce3 CM |
213 | if (!cache_miss) |
214 | cache_miss = last; | |
e19caa5f CM |
215 | goto new_group; |
216 | } | |
f510cfec CM |
217 | |
218 | start = max(last, start); | |
219 | last = end + 1; | |
257d0ce3 CM |
220 | if (last - start < num) { |
221 | if (last == cache->key.objectid + cache->key.offset) | |
222 | cache_miss = start; | |
f510cfec | 223 | continue; |
257d0ce3 CM |
224 | } |
225 | if (data != BTRFS_BLOCK_GROUP_MIXED && | |
5cf66426 | 226 | start + num > cache->key.objectid + cache->key.offset) |
e37c9e69 | 227 | goto new_group; |
c31f8830 CM |
228 | if (start + num > total_fs_bytes) |
229 | goto new_group; | |
f510cfec | 230 | return start; |
e37c9e69 CM |
231 | } |
232 | out: | |
1a2b2ac7 CM |
233 | cache = btrfs_lookup_block_group(root->fs_info, search_start); |
234 | if (!cache) { | |
235 | printk("Unable to find block group for %Lu\n", | |
236 | search_start); | |
237 | WARN_ON(1); | |
238 | return search_start; | |
239 | } | |
1a5bc167 | 240 | return search_start; |
e37c9e69 CM |
241 | |
242 | new_group: | |
e19caa5f | 243 | last = cache->key.objectid + cache->key.offset; |
f84a8b36 | 244 | wrapped: |
e19caa5f | 245 | cache = btrfs_lookup_block_group(root->fs_info, last); |
c31f8830 | 246 | if (!cache || cache->key.objectid >= total_fs_bytes) { |
0e4de584 | 247 | no_cache: |
f84a8b36 CM |
248 | if (!wrapped) { |
249 | wrapped = 1; | |
250 | last = search_start; | |
251 | data = BTRFS_BLOCK_GROUP_MIXED; | |
252 | goto wrapped; | |
253 | } | |
1a2b2ac7 | 254 | goto out; |
e37c9e69 | 255 | } |
257d0ce3 CM |
256 | if (cache_miss && !cache->cached) { |
257 | cache_block_group(root, cache); | |
258 | last = cache_miss; | |
257d0ce3 CM |
259 | cache = btrfs_lookup_block_group(root->fs_info, last); |
260 | } | |
1a2b2ac7 | 261 | cache = btrfs_find_block_group(root, cache, last, data, 0); |
0e4de584 CM |
262 | if (!cache) |
263 | goto no_cache; | |
e37c9e69 | 264 | *cache_ret = cache; |
257d0ce3 | 265 | cache_miss = 0; |
e37c9e69 CM |
266 | goto again; |
267 | } | |
268 | ||
84f54cfa CM |
269 | static u64 div_factor(u64 num, int factor) |
270 | { | |
257d0ce3 CM |
271 | if (factor == 10) |
272 | return num; | |
84f54cfa CM |
273 | num *= factor; |
274 | do_div(num, 10); | |
275 | return num; | |
276 | } | |
277 | ||
31f3c99b CM |
278 | struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, |
279 | struct btrfs_block_group_cache | |
be744175 | 280 | *hint, u64 search_start, |
de428b63 | 281 | int data, int owner) |
cd1bc465 | 282 | { |
96b5179d CM |
283 | struct btrfs_block_group_cache *cache; |
284 | struct extent_map_tree *block_group_cache; | |
31f3c99b | 285 | struct btrfs_block_group_cache *found_group = NULL; |
cd1bc465 CM |
286 | struct btrfs_fs_info *info = root->fs_info; |
287 | u64 used; | |
31f3c99b CM |
288 | u64 last = 0; |
289 | u64 hint_last; | |
96b5179d CM |
290 | u64 start; |
291 | u64 end; | |
292 | u64 free_check; | |
293 | u64 ptr; | |
c31f8830 | 294 | u64 total_fs_bytes; |
96b5179d | 295 | int bit; |
cd1bc465 | 296 | int ret; |
31f3c99b | 297 | int full_search = 0; |
de428b63 | 298 | int factor = 8; |
1e2677e0 | 299 | int data_swap = 0; |
de428b63 | 300 | |
96b5179d | 301 | block_group_cache = &info->block_group_cache; |
c31f8830 | 302 | total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); |
96b5179d | 303 | |
de428b63 | 304 | if (!owner) |
f84a8b36 | 305 | factor = 8; |
be744175 | 306 | |
257d0ce3 | 307 | if (data == BTRFS_BLOCK_GROUP_MIXED) { |
f84a8b36 | 308 | bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA; |
257d0ce3 CM |
309 | factor = 10; |
310 | } else if (data) | |
96b5179d CM |
311 | bit = BLOCK_GROUP_DATA; |
312 | else | |
313 | bit = BLOCK_GROUP_METADATA; | |
be744175 | 314 | |
c31f8830 | 315 | if (search_start && search_start < total_fs_bytes) { |
be744175 | 316 | struct btrfs_block_group_cache *shint; |
5276aeda | 317 | shint = btrfs_lookup_block_group(info, search_start); |
f84a8b36 CM |
318 | if (shint && (shint->data == data || |
319 | shint->data == BTRFS_BLOCK_GROUP_MIXED)) { | |
be744175 | 320 | used = btrfs_block_group_used(&shint->item); |
324ae4df Y |
321 | if (used + shint->pinned < |
322 | div_factor(shint->key.offset, factor)) { | |
be744175 CM |
323 | return shint; |
324 | } | |
325 | } | |
326 | } | |
c31f8830 CM |
327 | if (hint && hint->key.objectid < total_fs_bytes && |
328 | (hint->data == data || hint->data == BTRFS_BLOCK_GROUP_MIXED)) { | |
31f3c99b | 329 | used = btrfs_block_group_used(&hint->item); |
324ae4df Y |
330 | if (used + hint->pinned < |
331 | div_factor(hint->key.offset, factor)) { | |
31f3c99b CM |
332 | return hint; |
333 | } | |
e19caa5f | 334 | last = hint->key.objectid + hint->key.offset; |
31f3c99b CM |
335 | hint_last = last; |
336 | } else { | |
e37c9e69 CM |
337 | if (hint) |
338 | hint_last = max(hint->key.objectid, search_start); | |
339 | else | |
340 | hint_last = search_start; | |
341 | ||
c31f8830 CM |
342 | if (hint_last >= total_fs_bytes) |
343 | hint_last = search_start; | |
e37c9e69 | 344 | last = hint_last; |
31f3c99b | 345 | } |
31f3c99b | 346 | again: |
cd1bc465 | 347 | while(1) { |
96b5179d CM |
348 | ret = find_first_extent_bit(block_group_cache, last, |
349 | &start, &end, bit); | |
350 | if (ret) | |
cd1bc465 | 351 | break; |
96b5179d CM |
352 | |
353 | ret = get_state_private(block_group_cache, start, &ptr); | |
354 | if (ret) | |
355 | break; | |
356 | ||
ae2f5411 | 357 | cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; |
96b5179d CM |
358 | last = cache->key.objectid + cache->key.offset; |
359 | used = btrfs_block_group_used(&cache->item); | |
360 | ||
c31f8830 CM |
361 | if (cache->key.objectid > total_fs_bytes) |
362 | break; | |
363 | ||
96b5179d CM |
364 | if (full_search) |
365 | free_check = cache->key.offset; | |
366 | else | |
367 | free_check = div_factor(cache->key.offset, factor); | |
324ae4df | 368 | if (used + cache->pinned < free_check) { |
96b5179d CM |
369 | found_group = cache; |
370 | goto found; | |
cd1bc465 | 371 | } |
de428b63 | 372 | cond_resched(); |
cd1bc465 | 373 | } |
31f3c99b | 374 | if (!full_search) { |
be744175 | 375 | last = search_start; |
31f3c99b CM |
376 | full_search = 1; |
377 | goto again; | |
378 | } | |
1e2677e0 | 379 | if (!data_swap) { |
1e2677e0 | 380 | data_swap = 1; |
96b5179d | 381 | bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA; |
1e2677e0 CM |
382 | last = search_start; |
383 | goto again; | |
384 | } | |
be744175 | 385 | found: |
31f3c99b | 386 | return found_group; |
cd1bc465 CM |
387 | } |
388 | ||
7bb86316 | 389 | static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation, |
74493f7a CM |
390 | u64 owner, u64 owner_offset) |
391 | { | |
392 | u32 high_crc = ~(u32)0; | |
393 | u32 low_crc = ~(u32)0; | |
394 | __le64 lenum; | |
395 | ||
396 | lenum = cpu_to_le64(root_objectid); | |
397 | high_crc = crc32c(high_crc, &lenum, sizeof(lenum)); | |
7bb86316 CM |
398 | lenum = cpu_to_le64(ref_generation); |
399 | low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); | |
74493f7a | 400 | |
7bb86316 | 401 | #if 0 |
74493f7a CM |
402 | lenum = cpu_to_le64(owner); |
403 | low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); | |
74493f7a CM |
404 | lenum = cpu_to_le64(owner_offset); |
405 | low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); | |
7bb86316 | 406 | #endif |
74493f7a CM |
407 | return ((u64)high_crc << 32) | (u64)low_crc; |
408 | } | |
409 | ||
7bb86316 CM |
410 | static int match_extent_ref(struct extent_buffer *leaf, |
411 | struct btrfs_extent_ref *disk_ref, | |
412 | struct btrfs_extent_ref *cpu_ref) | |
413 | { | |
414 | int ret; | |
415 | int len; | |
416 | ||
417 | if (cpu_ref->objectid) | |
418 | len = sizeof(*cpu_ref); | |
419 | else | |
420 | len = 2 * sizeof(u64); | |
421 | ret = memcmp_extent_buffer(leaf, cpu_ref, (unsigned long)disk_ref, | |
422 | len); | |
423 | return ret == 0; | |
424 | } | |
425 | ||
98ed5174 CM |
426 | static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, |
427 | struct btrfs_root *root, | |
428 | struct btrfs_path *path, u64 bytenr, | |
429 | u64 root_objectid, | |
430 | u64 ref_generation, u64 owner, | |
431 | u64 owner_offset, int del) | |
74493f7a CM |
432 | { |
433 | u64 hash; | |
434 | struct btrfs_key key; | |
7bb86316 | 435 | struct btrfs_key found_key; |
74493f7a | 436 | struct btrfs_extent_ref ref; |
7bb86316 CM |
437 | struct extent_buffer *leaf; |
438 | struct btrfs_extent_ref *disk_ref; | |
439 | int ret; | |
440 | int ret2; | |
441 | ||
442 | btrfs_set_stack_ref_root(&ref, root_objectid); | |
443 | btrfs_set_stack_ref_generation(&ref, ref_generation); | |
444 | btrfs_set_stack_ref_objectid(&ref, owner); | |
445 | btrfs_set_stack_ref_offset(&ref, owner_offset); | |
446 | ||
447 | hash = hash_extent_ref(root_objectid, ref_generation, owner, | |
448 | owner_offset); | |
449 | key.offset = hash; | |
450 | key.objectid = bytenr; | |
451 | key.type = BTRFS_EXTENT_REF_KEY; | |
452 | ||
453 | while (1) { | |
454 | ret = btrfs_search_slot(trans, root, &key, path, | |
455 | del ? -1 : 0, del); | |
456 | if (ret < 0) | |
457 | goto out; | |
458 | leaf = path->nodes[0]; | |
459 | if (ret != 0) { | |
460 | u32 nritems = btrfs_header_nritems(leaf); | |
461 | if (path->slots[0] >= nritems) { | |
462 | ret2 = btrfs_next_leaf(root, path); | |
463 | if (ret2) | |
464 | goto out; | |
465 | leaf = path->nodes[0]; | |
466 | } | |
467 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
468 | if (found_key.objectid != bytenr || | |
469 | found_key.type != BTRFS_EXTENT_REF_KEY) | |
470 | goto out; | |
471 | key.offset = found_key.offset; | |
472 | if (del) { | |
473 | btrfs_release_path(root, path); | |
474 | continue; | |
475 | } | |
476 | } | |
477 | disk_ref = btrfs_item_ptr(path->nodes[0], | |
478 | path->slots[0], | |
479 | struct btrfs_extent_ref); | |
480 | if (match_extent_ref(path->nodes[0], disk_ref, &ref)) { | |
481 | ret = 0; | |
482 | goto out; | |
483 | } | |
484 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
485 | key.offset = found_key.offset + 1; | |
486 | btrfs_release_path(root, path); | |
487 | } | |
488 | out: | |
489 | return ret; | |
490 | } | |
491 | ||
d8d5f3e1 CM |
492 | /* |
493 | * Back reference rules. Back refs have three main goals: | |
494 | * | |
495 | * 1) differentiate between all holders of references to an extent so that | |
496 | * when a reference is dropped we can make sure it was a valid reference | |
497 | * before freeing the extent. | |
498 | * | |
499 | * 2) Provide enough information to quickly find the holders of an extent | |
500 | * if we notice a given block is corrupted or bad. | |
501 | * | |
502 | * 3) Make it easy to migrate blocks for FS shrinking or storage pool | |
503 | * maintenance. This is actually the same as #2, but with a slightly | |
504 | * different use case. | |
505 | * | |
506 | * File extents can be referenced by: | |
507 | * | |
508 | * - multiple snapshots, subvolumes, or different generations in one subvol | |
509 | * - different files inside a single subvolume (in theory, not implemented yet) | |
510 | * - different offsets inside a file (bookend extents in file.c) | |
511 | * | |
512 | * The extent ref structure has fields for: | |
513 | * | |
514 | * - Objectid of the subvolume root | |
515 | * - Generation number of the tree holding the reference | |
516 | * - objectid of the file holding the reference | |
517 | * - offset in the file corresponding to the key holding the reference | |
518 | * | |
519 | * When a file extent is allocated the fields are filled in: | |
520 | * (root_key.objectid, trans->transid, inode objectid, offset in file) | |
521 | * | |
522 | * When a leaf is cow'd new references are added for every file extent found | |
523 | * in the leaf. It looks the same as the create case, but trans->transid | |
524 | * will be different when the block is cow'd. | |
525 | * | |
526 | * (root_key.objectid, trans->transid, inode objectid, offset in file) | |
527 | * | |
528 | * When a file extent is removed either during snapshot deletion or file | |
529 | * truncation, the corresponding back reference is found | |
530 | * by searching for: | |
531 | * | |
532 | * (btrfs_header_owner(leaf), btrfs_header_generation(leaf), | |
533 | * inode objectid, offset in file) | |
534 | * | |
535 | * Btree extents can be referenced by: | |
536 | * | |
537 | * - Different subvolumes | |
538 | * - Different generations of the same subvolume | |
539 | * | |
540 | * Storing sufficient information for a full reverse mapping of a btree | |
541 | * block would require storing the lowest key of the block in the backref, | |
542 | * and it would require updating that lowest key either before write out or | |
543 | * every time it changed. Instead, the objectid of the lowest key is stored | |
544 | * along with the level of the tree block. This provides a hint | |
545 | * about where in the btree the block can be found. Searches through the | |
546 | * btree only need to look for a pointer to that block, so they stop one | |
547 | * level higher than the level recorded in the backref. | |
548 | * | |
549 | * Some btrees do not do reference counting on their extents. These | |
550 | * include the extent tree and the tree of tree roots. Backrefs for these | |
551 | * trees always have a generation of zero. | |
552 | * | |
553 | * When a tree block is created, back references are inserted: | |
554 | * | |
f6dbff55 | 555 | * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid) |
d8d5f3e1 CM |
556 | * |
557 | * When a tree block is cow'd in a reference counted root, | |
558 | * new back references are added for all the blocks it points to. | |
559 | * These are of the form (trans->transid will have increased since creation): | |
560 | * | |
f6dbff55 | 561 | * (root->root_key.objectid, trans->transid, level, lowest_key_objectid) |
d8d5f3e1 CM |
562 | * |
563 | * Because the lowest_key_objectid and the level are just hints | |
564 | * they are not used when backrefs are deleted. When a backref is deleted: | |
565 | * | |
566 | * if backref was for a tree root: | |
567 | * root_objectid = root->root_key.objectid | |
568 | * else | |
569 | * root_objectid = btrfs_header_owner(parent) | |
570 | * | |
571 | * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0) | |
572 | * | |
573 | * Back Reference Key hashing: | |
574 | * | |
575 | * Back references have four fields, each 64 bits long. Unfortunately, | |
576 | * This is hashed into a single 64 bit number and placed into the key offset. | |
577 | * The key objectid corresponds to the first byte in the extent, and the | |
578 | * key type is set to BTRFS_EXTENT_REF_KEY | |
579 | */ | |
7bb86316 CM |
580 | int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, |
581 | struct btrfs_root *root, | |
582 | struct btrfs_path *path, u64 bytenr, | |
583 | u64 root_objectid, u64 ref_generation, | |
584 | u64 owner, u64 owner_offset) | |
585 | { | |
586 | u64 hash; | |
587 | struct btrfs_key key; | |
588 | struct btrfs_extent_ref ref; | |
589 | struct btrfs_extent_ref *disk_ref; | |
74493f7a CM |
590 | int ret; |
591 | ||
592 | btrfs_set_stack_ref_root(&ref, root_objectid); | |
7bb86316 | 593 | btrfs_set_stack_ref_generation(&ref, ref_generation); |
74493f7a CM |
594 | btrfs_set_stack_ref_objectid(&ref, owner); |
595 | btrfs_set_stack_ref_offset(&ref, owner_offset); | |
596 | ||
7bb86316 CM |
597 | hash = hash_extent_ref(root_objectid, ref_generation, owner, |
598 | owner_offset); | |
74493f7a CM |
599 | key.offset = hash; |
600 | key.objectid = bytenr; | |
601 | key.type = BTRFS_EXTENT_REF_KEY; | |
602 | ||
603 | ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(ref)); | |
604 | while (ret == -EEXIST) { | |
7bb86316 CM |
605 | disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], |
606 | struct btrfs_extent_ref); | |
607 | if (match_extent_ref(path->nodes[0], disk_ref, &ref)) | |
608 | goto out; | |
609 | key.offset++; | |
610 | btrfs_release_path(root, path); | |
611 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
612 | sizeof(ref)); | |
74493f7a | 613 | } |
7bb86316 CM |
614 | if (ret) |
615 | goto out; | |
616 | disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
617 | struct btrfs_extent_ref); | |
618 | write_extent_buffer(path->nodes[0], &ref, (unsigned long)disk_ref, | |
619 | sizeof(ref)); | |
620 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
621 | out: | |
622 | btrfs_release_path(root, path); | |
623 | return ret; | |
74493f7a CM |
624 | } |
625 | ||
b18c6685 CM |
626 | int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, |
627 | struct btrfs_root *root, | |
74493f7a | 628 | u64 bytenr, u64 num_bytes, |
7bb86316 | 629 | u64 root_objectid, u64 ref_generation, |
74493f7a | 630 | u64 owner, u64 owner_offset) |
02217ed2 | 631 | { |
5caf2a00 | 632 | struct btrfs_path *path; |
02217ed2 | 633 | int ret; |
e2fa7227 | 634 | struct btrfs_key key; |
5f39d397 | 635 | struct extent_buffer *l; |
234b63a0 | 636 | struct btrfs_extent_item *item; |
cf27e1ee | 637 | u32 refs; |
037e6390 | 638 | |
db94535d | 639 | WARN_ON(num_bytes < root->sectorsize); |
5caf2a00 | 640 | path = btrfs_alloc_path(); |
54aa1f4d CM |
641 | if (!path) |
642 | return -ENOMEM; | |
26b8003f | 643 | |
b0331a4c | 644 | path->reada = 0; |
db94535d | 645 | key.objectid = bytenr; |
62e2749e | 646 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
db94535d | 647 | key.offset = num_bytes; |
5caf2a00 | 648 | ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, |
9f5fae2f | 649 | 0, 1); |
54aa1f4d CM |
650 | if (ret < 0) |
651 | return ret; | |
a429e513 | 652 | if (ret != 0) { |
a28ec197 | 653 | BUG(); |
a429e513 | 654 | } |
02217ed2 | 655 | BUG_ON(ret != 0); |
5f39d397 | 656 | l = path->nodes[0]; |
5caf2a00 | 657 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); |
5f39d397 CM |
658 | refs = btrfs_extent_refs(l, item); |
659 | btrfs_set_extent_refs(l, item, refs + 1); | |
5caf2a00 | 660 | btrfs_mark_buffer_dirty(path->nodes[0]); |
a28ec197 | 661 | |
5caf2a00 | 662 | btrfs_release_path(root->fs_info->extent_root, path); |
7bb86316 | 663 | |
b0331a4c | 664 | path->reada = 0; |
7bb86316 CM |
665 | ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root, |
666 | path, bytenr, root_objectid, | |
667 | ref_generation, owner, owner_offset); | |
668 | BUG_ON(ret); | |
9f5fae2f | 669 | finish_current_insert(trans, root->fs_info->extent_root); |
e20d96d6 | 670 | del_pending_extents(trans, root->fs_info->extent_root); |
74493f7a CM |
671 | |
672 | btrfs_free_path(path); | |
02217ed2 CM |
673 | return 0; |
674 | } | |
675 | ||
e9d0b13b CM |
676 | int btrfs_extent_post_op(struct btrfs_trans_handle *trans, |
677 | struct btrfs_root *root) | |
678 | { | |
679 | finish_current_insert(trans, root->fs_info->extent_root); | |
680 | del_pending_extents(trans, root->fs_info->extent_root); | |
681 | return 0; | |
682 | } | |
683 | ||
b18c6685 | 684 | static int lookup_extent_ref(struct btrfs_trans_handle *trans, |
db94535d CM |
685 | struct btrfs_root *root, u64 bytenr, |
686 | u64 num_bytes, u32 *refs) | |
a28ec197 | 687 | { |
5caf2a00 | 688 | struct btrfs_path *path; |
a28ec197 | 689 | int ret; |
e2fa7227 | 690 | struct btrfs_key key; |
5f39d397 | 691 | struct extent_buffer *l; |
234b63a0 | 692 | struct btrfs_extent_item *item; |
5caf2a00 | 693 | |
db94535d | 694 | WARN_ON(num_bytes < root->sectorsize); |
5caf2a00 | 695 | path = btrfs_alloc_path(); |
b0331a4c | 696 | path->reada = 0; |
db94535d CM |
697 | key.objectid = bytenr; |
698 | key.offset = num_bytes; | |
62e2749e | 699 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
5caf2a00 | 700 | ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, |
9f5fae2f | 701 | 0, 0); |
54aa1f4d CM |
702 | if (ret < 0) |
703 | goto out; | |
5f39d397 CM |
704 | if (ret != 0) { |
705 | btrfs_print_leaf(root, path->nodes[0]); | |
db94535d | 706 | printk("failed to find block number %Lu\n", bytenr); |
a28ec197 | 707 | BUG(); |
5f39d397 CM |
708 | } |
709 | l = path->nodes[0]; | |
5caf2a00 | 710 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); |
5f39d397 | 711 | *refs = btrfs_extent_refs(l, item); |
54aa1f4d | 712 | out: |
5caf2a00 | 713 | btrfs_free_path(path); |
a28ec197 CM |
714 | return 0; |
715 | } | |
716 | ||
be20aa9d CM |
717 | u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, |
718 | struct btrfs_path *count_path, | |
719 | u64 first_extent) | |
720 | { | |
721 | struct btrfs_root *extent_root = root->fs_info->extent_root; | |
722 | struct btrfs_path *path; | |
723 | u64 bytenr; | |
724 | u64 found_objectid; | |
56b453c9 | 725 | u64 root_objectid = root->root_key.objectid; |
be20aa9d CM |
726 | u32 total_count = 0; |
727 | u32 cur_count; | |
be20aa9d CM |
728 | u32 nritems; |
729 | int ret; | |
730 | struct btrfs_key key; | |
731 | struct btrfs_key found_key; | |
732 | struct extent_buffer *l; | |
733 | struct btrfs_extent_item *item; | |
734 | struct btrfs_extent_ref *ref_item; | |
735 | int level = -1; | |
736 | ||
737 | path = btrfs_alloc_path(); | |
738 | again: | |
739 | if (level == -1) | |
740 | bytenr = first_extent; | |
741 | else | |
742 | bytenr = count_path->nodes[level]->start; | |
743 | ||
744 | cur_count = 0; | |
745 | key.objectid = bytenr; | |
746 | key.offset = 0; | |
747 | ||
748 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); | |
749 | ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); | |
750 | if (ret < 0) | |
751 | goto out; | |
752 | BUG_ON(ret == 0); | |
753 | ||
754 | l = path->nodes[0]; | |
755 | btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); | |
756 | ||
757 | if (found_key.objectid != bytenr || | |
758 | found_key.type != BTRFS_EXTENT_ITEM_KEY) { | |
759 | goto out; | |
760 | } | |
761 | ||
762 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); | |
be20aa9d | 763 | while (1) { |
bd09835d | 764 | l = path->nodes[0]; |
be20aa9d CM |
765 | nritems = btrfs_header_nritems(l); |
766 | if (path->slots[0] >= nritems) { | |
767 | ret = btrfs_next_leaf(extent_root, path); | |
768 | if (ret == 0) | |
769 | continue; | |
770 | break; | |
771 | } | |
772 | btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); | |
773 | if (found_key.objectid != bytenr) | |
774 | break; | |
bd09835d | 775 | |
be20aa9d CM |
776 | if (found_key.type != BTRFS_EXTENT_REF_KEY) { |
777 | path->slots[0]++; | |
778 | continue; | |
779 | } | |
780 | ||
781 | cur_count++; | |
782 | ref_item = btrfs_item_ptr(l, path->slots[0], | |
783 | struct btrfs_extent_ref); | |
784 | found_objectid = btrfs_ref_root(l, ref_item); | |
785 | ||
56b453c9 CM |
786 | if (found_objectid != root_objectid) { |
787 | total_count = 2; | |
4313b399 | 788 | goto out; |
56b453c9 CM |
789 | } |
790 | total_count = 1; | |
be20aa9d CM |
791 | path->slots[0]++; |
792 | } | |
793 | if (cur_count == 0) { | |
794 | total_count = 0; | |
795 | goto out; | |
796 | } | |
be20aa9d CM |
797 | if (level >= 0 && root->node == count_path->nodes[level]) |
798 | goto out; | |
799 | level++; | |
800 | btrfs_release_path(root, path); | |
801 | goto again; | |
802 | ||
803 | out: | |
804 | btrfs_free_path(path); | |
805 | return total_count; | |
be20aa9d | 806 | } |
c5739bba | 807 | int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, |
7bb86316 | 808 | struct btrfs_root *root, u64 owner_objectid) |
c5739bba | 809 | { |
7bb86316 CM |
810 | u64 generation; |
811 | u64 key_objectid; | |
812 | u64 level; | |
813 | u32 nritems; | |
814 | struct btrfs_disk_key disk_key; | |
815 | ||
816 | level = btrfs_header_level(root->node); | |
817 | generation = trans->transid; | |
818 | nritems = btrfs_header_nritems(root->node); | |
819 | if (nritems > 0) { | |
820 | if (level == 0) | |
821 | btrfs_item_key(root->node, &disk_key, 0); | |
822 | else | |
823 | btrfs_node_key(root->node, &disk_key, 0); | |
824 | key_objectid = btrfs_disk_key_objectid(&disk_key); | |
825 | } else { | |
826 | key_objectid = 0; | |
827 | } | |
db94535d | 828 | return btrfs_inc_extent_ref(trans, root, root->node->start, |
7bb86316 | 829 | root->node->len, owner_objectid, |
f6dbff55 | 830 | generation, level, key_objectid); |
c5739bba CM |
831 | } |
832 | ||
e089f05c | 833 | int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 834 | struct extent_buffer *buf) |
02217ed2 | 835 | { |
db94535d | 836 | u64 bytenr; |
5f39d397 CM |
837 | u32 nritems; |
838 | struct btrfs_key key; | |
6407bf6d | 839 | struct btrfs_file_extent_item *fi; |
02217ed2 | 840 | int i; |
db94535d | 841 | int level; |
6407bf6d | 842 | int ret; |
54aa1f4d | 843 | int faili; |
a28ec197 | 844 | |
3768f368 | 845 | if (!root->ref_cows) |
a28ec197 | 846 | return 0; |
5f39d397 | 847 | |
db94535d | 848 | level = btrfs_header_level(buf); |
5f39d397 CM |
849 | nritems = btrfs_header_nritems(buf); |
850 | for (i = 0; i < nritems; i++) { | |
db94535d CM |
851 | if (level == 0) { |
852 | u64 disk_bytenr; | |
5f39d397 CM |
853 | btrfs_item_key_to_cpu(buf, &key, i); |
854 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
6407bf6d | 855 | continue; |
5f39d397 | 856 | fi = btrfs_item_ptr(buf, i, |
6407bf6d | 857 | struct btrfs_file_extent_item); |
5f39d397 | 858 | if (btrfs_file_extent_type(buf, fi) == |
236454df CM |
859 | BTRFS_FILE_EXTENT_INLINE) |
860 | continue; | |
db94535d CM |
861 | disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); |
862 | if (disk_bytenr == 0) | |
3a686375 | 863 | continue; |
db94535d | 864 | ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, |
7bb86316 CM |
865 | btrfs_file_extent_disk_num_bytes(buf, fi), |
866 | root->root_key.objectid, trans->transid, | |
867 | key.objectid, key.offset); | |
54aa1f4d CM |
868 | if (ret) { |
869 | faili = i; | |
870 | goto fail; | |
871 | } | |
6407bf6d | 872 | } else { |
db94535d | 873 | bytenr = btrfs_node_blockptr(buf, i); |
6caab489 | 874 | btrfs_node_key_to_cpu(buf, &key, i); |
db94535d | 875 | ret = btrfs_inc_extent_ref(trans, root, bytenr, |
7bb86316 CM |
876 | btrfs_level_size(root, level - 1), |
877 | root->root_key.objectid, | |
f6dbff55 CM |
878 | trans->transid, |
879 | level - 1, key.objectid); | |
54aa1f4d CM |
880 | if (ret) { |
881 | faili = i; | |
882 | goto fail; | |
883 | } | |
6407bf6d | 884 | } |
02217ed2 CM |
885 | } |
886 | return 0; | |
54aa1f4d | 887 | fail: |
ccd467d6 | 888 | WARN_ON(1); |
7bb86316 | 889 | #if 0 |
54aa1f4d | 890 | for (i =0; i < faili; i++) { |
db94535d CM |
891 | if (level == 0) { |
892 | u64 disk_bytenr; | |
5f39d397 CM |
893 | btrfs_item_key_to_cpu(buf, &key, i); |
894 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
54aa1f4d | 895 | continue; |
5f39d397 | 896 | fi = btrfs_item_ptr(buf, i, |
54aa1f4d | 897 | struct btrfs_file_extent_item); |
5f39d397 | 898 | if (btrfs_file_extent_type(buf, fi) == |
54aa1f4d CM |
899 | BTRFS_FILE_EXTENT_INLINE) |
900 | continue; | |
db94535d CM |
901 | disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); |
902 | if (disk_bytenr == 0) | |
54aa1f4d | 903 | continue; |
db94535d CM |
904 | err = btrfs_free_extent(trans, root, disk_bytenr, |
905 | btrfs_file_extent_disk_num_bytes(buf, | |
5f39d397 | 906 | fi), 0); |
54aa1f4d CM |
907 | BUG_ON(err); |
908 | } else { | |
db94535d CM |
909 | bytenr = btrfs_node_blockptr(buf, i); |
910 | err = btrfs_free_extent(trans, root, bytenr, | |
911 | btrfs_level_size(root, level - 1), 0); | |
54aa1f4d CM |
912 | BUG_ON(err); |
913 | } | |
914 | } | |
7bb86316 | 915 | #endif |
54aa1f4d | 916 | return ret; |
02217ed2 CM |
917 | } |
918 | ||
9078a3e1 CM |
919 | static int write_one_cache_group(struct btrfs_trans_handle *trans, |
920 | struct btrfs_root *root, | |
921 | struct btrfs_path *path, | |
922 | struct btrfs_block_group_cache *cache) | |
923 | { | |
924 | int ret; | |
925 | int pending_ret; | |
926 | struct btrfs_root *extent_root = root->fs_info->extent_root; | |
5f39d397 CM |
927 | unsigned long bi; |
928 | struct extent_buffer *leaf; | |
9078a3e1 | 929 | |
9078a3e1 | 930 | ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1); |
54aa1f4d CM |
931 | if (ret < 0) |
932 | goto fail; | |
9078a3e1 | 933 | BUG_ON(ret); |
5f39d397 CM |
934 | |
935 | leaf = path->nodes[0]; | |
936 | bi = btrfs_item_ptr_offset(leaf, path->slots[0]); | |
937 | write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item)); | |
938 | btrfs_mark_buffer_dirty(leaf); | |
9078a3e1 | 939 | btrfs_release_path(extent_root, path); |
54aa1f4d | 940 | fail: |
9078a3e1 CM |
941 | finish_current_insert(trans, extent_root); |
942 | pending_ret = del_pending_extents(trans, extent_root); | |
943 | if (ret) | |
944 | return ret; | |
945 | if (pending_ret) | |
946 | return pending_ret; | |
947 | return 0; | |
948 | ||
949 | } | |
950 | ||
96b5179d CM |
951 | int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, |
952 | struct btrfs_root *root) | |
9078a3e1 | 953 | { |
96b5179d CM |
954 | struct extent_map_tree *block_group_cache; |
955 | struct btrfs_block_group_cache *cache; | |
9078a3e1 CM |
956 | int ret; |
957 | int err = 0; | |
958 | int werr = 0; | |
9078a3e1 | 959 | struct btrfs_path *path; |
96b5179d CM |
960 | u64 last = 0; |
961 | u64 start; | |
962 | u64 end; | |
963 | u64 ptr; | |
9078a3e1 | 964 | |
96b5179d | 965 | block_group_cache = &root->fs_info->block_group_cache; |
9078a3e1 CM |
966 | path = btrfs_alloc_path(); |
967 | if (!path) | |
968 | return -ENOMEM; | |
969 | ||
970 | while(1) { | |
96b5179d CM |
971 | ret = find_first_extent_bit(block_group_cache, last, |
972 | &start, &end, BLOCK_GROUP_DIRTY); | |
973 | if (ret) | |
9078a3e1 | 974 | break; |
54aa1f4d | 975 | |
96b5179d CM |
976 | last = end + 1; |
977 | ret = get_state_private(block_group_cache, start, &ptr); | |
978 | if (ret) | |
979 | break; | |
980 | ||
ae2f5411 | 981 | cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; |
96b5179d CM |
982 | err = write_one_cache_group(trans, root, |
983 | path, cache); | |
984 | /* | |
985 | * if we fail to write the cache group, we want | |
986 | * to keep it marked dirty in hopes that a later | |
987 | * write will work | |
988 | */ | |
989 | if (err) { | |
990 | werr = err; | |
991 | continue; | |
9078a3e1 | 992 | } |
96b5179d CM |
993 | clear_extent_bits(block_group_cache, start, end, |
994 | BLOCK_GROUP_DIRTY, GFP_NOFS); | |
9078a3e1 CM |
995 | } |
996 | btrfs_free_path(path); | |
997 | return werr; | |
998 | } | |
999 | ||
1000 | static int update_block_group(struct btrfs_trans_handle *trans, | |
1001 | struct btrfs_root *root, | |
db94535d CM |
1002 | u64 bytenr, u64 num_bytes, int alloc, |
1003 | int mark_free, int data) | |
9078a3e1 CM |
1004 | { |
1005 | struct btrfs_block_group_cache *cache; | |
1006 | struct btrfs_fs_info *info = root->fs_info; | |
db94535d | 1007 | u64 total = num_bytes; |
9078a3e1 | 1008 | u64 old_val; |
db94535d | 1009 | u64 byte_in_group; |
96b5179d CM |
1010 | u64 start; |
1011 | u64 end; | |
3e1ad54f | 1012 | |
9078a3e1 | 1013 | while(total) { |
db94535d | 1014 | cache = btrfs_lookup_block_group(info, bytenr); |
3e1ad54f | 1015 | if (!cache) { |
9078a3e1 | 1016 | return -1; |
cd1bc465 | 1017 | } |
db94535d CM |
1018 | byte_in_group = bytenr - cache->key.objectid; |
1019 | WARN_ON(byte_in_group > cache->key.offset); | |
96b5179d CM |
1020 | start = cache->key.objectid; |
1021 | end = start + cache->key.offset - 1; | |
1022 | set_extent_bits(&info->block_group_cache, start, end, | |
1023 | BLOCK_GROUP_DIRTY, GFP_NOFS); | |
9078a3e1 CM |
1024 | |
1025 | old_val = btrfs_block_group_used(&cache->item); | |
db94535d | 1026 | num_bytes = min(total, cache->key.offset - byte_in_group); |
cd1bc465 | 1027 | if (alloc) { |
1e2677e0 | 1028 | if (cache->data != data && |
84f54cfa | 1029 | old_val < (cache->key.offset >> 1)) { |
96b5179d CM |
1030 | int bit_to_clear; |
1031 | int bit_to_set; | |
96b5179d | 1032 | cache->data = data; |
1e2677e0 | 1033 | if (data) { |
b97f9203 Y |
1034 | bit_to_clear = BLOCK_GROUP_METADATA; |
1035 | bit_to_set = BLOCK_GROUP_DATA; | |
f84a8b36 CM |
1036 | cache->item.flags &= |
1037 | ~BTRFS_BLOCK_GROUP_MIXED; | |
1e2677e0 CM |
1038 | cache->item.flags |= |
1039 | BTRFS_BLOCK_GROUP_DATA; | |
1040 | } else { | |
b97f9203 Y |
1041 | bit_to_clear = BLOCK_GROUP_DATA; |
1042 | bit_to_set = BLOCK_GROUP_METADATA; | |
f84a8b36 CM |
1043 | cache->item.flags &= |
1044 | ~BTRFS_BLOCK_GROUP_MIXED; | |
1e2677e0 CM |
1045 | cache->item.flags &= |
1046 | ~BTRFS_BLOCK_GROUP_DATA; | |
1047 | } | |
96b5179d CM |
1048 | clear_extent_bits(&info->block_group_cache, |
1049 | start, end, bit_to_clear, | |
1050 | GFP_NOFS); | |
1051 | set_extent_bits(&info->block_group_cache, | |
1052 | start, end, bit_to_set, | |
1053 | GFP_NOFS); | |
f84a8b36 CM |
1054 | } else if (cache->data != data && |
1055 | cache->data != BTRFS_BLOCK_GROUP_MIXED) { | |
1056 | cache->data = BTRFS_BLOCK_GROUP_MIXED; | |
1057 | set_extent_bits(&info->block_group_cache, | |
1058 | start, end, | |
1059 | BLOCK_GROUP_DATA | | |
1060 | BLOCK_GROUP_METADATA, | |
1061 | GFP_NOFS); | |
1e2677e0 | 1062 | } |
db94535d | 1063 | old_val += num_bytes; |
cd1bc465 | 1064 | } else { |
db94535d | 1065 | old_val -= num_bytes; |
f510cfec CM |
1066 | if (mark_free) { |
1067 | set_extent_dirty(&info->free_space_cache, | |
db94535d | 1068 | bytenr, bytenr + num_bytes - 1, |
f510cfec | 1069 | GFP_NOFS); |
e37c9e69 | 1070 | } |
cd1bc465 | 1071 | } |
9078a3e1 | 1072 | btrfs_set_block_group_used(&cache->item, old_val); |
db94535d CM |
1073 | total -= num_bytes; |
1074 | bytenr += num_bytes; | |
9078a3e1 CM |
1075 | } |
1076 | return 0; | |
1077 | } | |
324ae4df Y |
1078 | static int update_pinned_extents(struct btrfs_root *root, |
1079 | u64 bytenr, u64 num, int pin) | |
1080 | { | |
1081 | u64 len; | |
1082 | struct btrfs_block_group_cache *cache; | |
1083 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1084 | ||
1085 | if (pin) { | |
1086 | set_extent_dirty(&fs_info->pinned_extents, | |
1087 | bytenr, bytenr + num - 1, GFP_NOFS); | |
1088 | } else { | |
1089 | clear_extent_dirty(&fs_info->pinned_extents, | |
1090 | bytenr, bytenr + num - 1, GFP_NOFS); | |
1091 | } | |
1092 | while (num > 0) { | |
1093 | cache = btrfs_lookup_block_group(fs_info, bytenr); | |
1094 | WARN_ON(!cache); | |
1095 | len = min(num, cache->key.offset - | |
1096 | (bytenr - cache->key.objectid)); | |
1097 | if (pin) { | |
1098 | cache->pinned += len; | |
1099 | fs_info->total_pinned += len; | |
1100 | } else { | |
1101 | cache->pinned -= len; | |
1102 | fs_info->total_pinned -= len; | |
1103 | } | |
1104 | bytenr += len; | |
1105 | num -= len; | |
1106 | } | |
1107 | return 0; | |
1108 | } | |
9078a3e1 | 1109 | |
1a5bc167 | 1110 | int btrfs_copy_pinned(struct btrfs_root *root, struct extent_map_tree *copy) |
ccd467d6 | 1111 | { |
ccd467d6 | 1112 | u64 last = 0; |
1a5bc167 CM |
1113 | u64 start; |
1114 | u64 end; | |
1115 | struct extent_map_tree *pinned_extents = &root->fs_info->pinned_extents; | |
ccd467d6 | 1116 | int ret; |
ccd467d6 CM |
1117 | |
1118 | while(1) { | |
1a5bc167 CM |
1119 | ret = find_first_extent_bit(pinned_extents, last, |
1120 | &start, &end, EXTENT_DIRTY); | |
1121 | if (ret) | |
ccd467d6 | 1122 | break; |
1a5bc167 CM |
1123 | set_extent_dirty(copy, start, end, GFP_NOFS); |
1124 | last = end + 1; | |
ccd467d6 CM |
1125 | } |
1126 | return 0; | |
1127 | } | |
1128 | ||
1129 | int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, | |
1130 | struct btrfs_root *root, | |
1a5bc167 | 1131 | struct extent_map_tree *unpin) |
a28ec197 | 1132 | { |
1a5bc167 CM |
1133 | u64 start; |
1134 | u64 end; | |
a28ec197 | 1135 | int ret; |
f510cfec | 1136 | struct extent_map_tree *free_space_cache; |
f510cfec | 1137 | free_space_cache = &root->fs_info->free_space_cache; |
a28ec197 CM |
1138 | |
1139 | while(1) { | |
1a5bc167 CM |
1140 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
1141 | EXTENT_DIRTY); | |
1142 | if (ret) | |
a28ec197 | 1143 | break; |
324ae4df | 1144 | update_pinned_extents(root, start, end + 1 - start, 0); |
1a5bc167 CM |
1145 | clear_extent_dirty(unpin, start, end, GFP_NOFS); |
1146 | set_extent_dirty(free_space_cache, start, end, GFP_NOFS); | |
a28ec197 CM |
1147 | } |
1148 | return 0; | |
1149 | } | |
1150 | ||
98ed5174 CM |
1151 | static int finish_current_insert(struct btrfs_trans_handle *trans, |
1152 | struct btrfs_root *extent_root) | |
037e6390 | 1153 | { |
7bb86316 CM |
1154 | u64 start; |
1155 | u64 end; | |
1156 | struct btrfs_fs_info *info = extent_root->fs_info; | |
d8d5f3e1 | 1157 | struct extent_buffer *eb; |
7bb86316 | 1158 | struct btrfs_path *path; |
e2fa7227 | 1159 | struct btrfs_key ins; |
d8d5f3e1 | 1160 | struct btrfs_disk_key first; |
234b63a0 | 1161 | struct btrfs_extent_item extent_item; |
037e6390 | 1162 | int ret; |
d8d5f3e1 | 1163 | int level; |
1a5bc167 | 1164 | int err = 0; |
037e6390 | 1165 | |
5f39d397 | 1166 | btrfs_set_stack_extent_refs(&extent_item, 1); |
62e2749e | 1167 | btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY); |
7bb86316 | 1168 | path = btrfs_alloc_path(); |
037e6390 | 1169 | |
26b8003f | 1170 | while(1) { |
1a5bc167 CM |
1171 | ret = find_first_extent_bit(&info->extent_ins, 0, &start, |
1172 | &end, EXTENT_LOCKED); | |
1173 | if (ret) | |
26b8003f CM |
1174 | break; |
1175 | ||
1a5bc167 CM |
1176 | ins.objectid = start; |
1177 | ins.offset = end + 1 - start; | |
1178 | err = btrfs_insert_item(trans, extent_root, &ins, | |
1179 | &extent_item, sizeof(extent_item)); | |
1180 | clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED, | |
1181 | GFP_NOFS); | |
d8d5f3e1 CM |
1182 | eb = read_tree_block(extent_root, ins.objectid, ins.offset); |
1183 | level = btrfs_header_level(eb); | |
1184 | if (level == 0) { | |
1185 | btrfs_item_key(eb, &first, 0); | |
1186 | } else { | |
1187 | btrfs_node_key(eb, &first, 0); | |
1188 | } | |
7bb86316 CM |
1189 | err = btrfs_insert_extent_backref(trans, extent_root, path, |
1190 | start, extent_root->root_key.objectid, | |
f6dbff55 CM |
1191 | 0, level, |
1192 | btrfs_disk_key_objectid(&first)); | |
7bb86316 | 1193 | BUG_ON(err); |
d8d5f3e1 | 1194 | free_extent_buffer(eb); |
037e6390 | 1195 | } |
7bb86316 | 1196 | btrfs_free_path(path); |
037e6390 CM |
1197 | return 0; |
1198 | } | |
1199 | ||
db94535d CM |
1200 | static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, |
1201 | int pending) | |
e20d96d6 | 1202 | { |
1a5bc167 | 1203 | int err = 0; |
5f39d397 | 1204 | struct extent_buffer *buf; |
8ef97622 | 1205 | |
f4b9aa8d | 1206 | if (!pending) { |
db94535d | 1207 | buf = btrfs_find_tree_block(root, bytenr, num_bytes); |
5f39d397 CM |
1208 | if (buf) { |
1209 | if (btrfs_buffer_uptodate(buf)) { | |
2c90e5d6 CM |
1210 | u64 transid = |
1211 | root->fs_info->running_transaction->transid; | |
dc17ff8f CM |
1212 | u64 header_transid = |
1213 | btrfs_header_generation(buf); | |
1214 | if (header_transid == transid) { | |
55c69072 | 1215 | clean_tree_block(NULL, root, buf); |
5f39d397 | 1216 | free_extent_buffer(buf); |
c549228f | 1217 | return 1; |
2c90e5d6 | 1218 | } |
f4b9aa8d | 1219 | } |
5f39d397 | 1220 | free_extent_buffer(buf); |
8ef97622 | 1221 | } |
324ae4df | 1222 | update_pinned_extents(root, bytenr, num_bytes, 1); |
f4b9aa8d | 1223 | } else { |
1a5bc167 | 1224 | set_extent_bits(&root->fs_info->pending_del, |
db94535d CM |
1225 | bytenr, bytenr + num_bytes - 1, |
1226 | EXTENT_LOCKED, GFP_NOFS); | |
f4b9aa8d | 1227 | } |
be744175 | 1228 | BUG_ON(err < 0); |
e20d96d6 CM |
1229 | return 0; |
1230 | } | |
1231 | ||
fec577fb | 1232 | /* |
a28ec197 | 1233 | * remove an extent from the root, returns 0 on success |
fec577fb | 1234 | */ |
e089f05c | 1235 | static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root |
7bb86316 CM |
1236 | *root, u64 bytenr, u64 num_bytes, |
1237 | u64 root_objectid, u64 ref_generation, | |
1238 | u64 owner_objectid, u64 owner_offset, int pin, | |
e37c9e69 | 1239 | int mark_free) |
a28ec197 | 1240 | { |
5caf2a00 | 1241 | struct btrfs_path *path; |
e2fa7227 | 1242 | struct btrfs_key key; |
1261ec42 CM |
1243 | struct btrfs_fs_info *info = root->fs_info; |
1244 | struct btrfs_root *extent_root = info->extent_root; | |
5f39d397 | 1245 | struct extent_buffer *leaf; |
a28ec197 | 1246 | int ret; |
234b63a0 | 1247 | struct btrfs_extent_item *ei; |
cf27e1ee | 1248 | u32 refs; |
037e6390 | 1249 | |
db94535d | 1250 | key.objectid = bytenr; |
62e2749e | 1251 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
db94535d | 1252 | key.offset = num_bytes; |
5caf2a00 | 1253 | path = btrfs_alloc_path(); |
54aa1f4d CM |
1254 | if (!path) |
1255 | return -ENOMEM; | |
5f26f772 | 1256 | |
b0331a4c | 1257 | path->reada = 0; |
7bb86316 CM |
1258 | ret = lookup_extent_backref(trans, extent_root, path, |
1259 | bytenr, root_objectid, | |
1260 | ref_generation, | |
1261 | owner_objectid, owner_offset, 1); | |
1262 | if (ret == 0) { | |
1263 | ret = btrfs_del_item(trans, extent_root, path); | |
1264 | } else { | |
1265 | btrfs_print_leaf(extent_root, path->nodes[0]); | |
1266 | WARN_ON(1); | |
1267 | printk("Unable to find ref byte nr %Lu root %Lu " | |
1268 | " gen %Lu owner %Lu offset %Lu\n", bytenr, | |
1269 | root_objectid, ref_generation, owner_objectid, | |
1270 | owner_offset); | |
1271 | } | |
1272 | btrfs_release_path(extent_root, path); | |
54aa1f4d CM |
1273 | ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); |
1274 | if (ret < 0) | |
1275 | return ret; | |
1276 | BUG_ON(ret); | |
5f39d397 CM |
1277 | |
1278 | leaf = path->nodes[0]; | |
1279 | ei = btrfs_item_ptr(leaf, path->slots[0], | |
123abc88 | 1280 | struct btrfs_extent_item); |
5f39d397 CM |
1281 | refs = btrfs_extent_refs(leaf, ei); |
1282 | BUG_ON(refs == 0); | |
1283 | refs -= 1; | |
1284 | btrfs_set_extent_refs(leaf, ei, refs); | |
1285 | btrfs_mark_buffer_dirty(leaf); | |
1286 | ||
cf27e1ee | 1287 | if (refs == 0) { |
db94535d CM |
1288 | u64 super_used; |
1289 | u64 root_used; | |
78fae27e CM |
1290 | |
1291 | if (pin) { | |
db94535d | 1292 | ret = pin_down_bytes(root, bytenr, num_bytes, 0); |
c549228f Y |
1293 | if (ret > 0) |
1294 | mark_free = 1; | |
1295 | BUG_ON(ret < 0); | |
78fae27e CM |
1296 | } |
1297 | ||
58176a96 | 1298 | /* block accounting for super block */ |
db94535d CM |
1299 | super_used = btrfs_super_bytes_used(&info->super_copy); |
1300 | btrfs_set_super_bytes_used(&info->super_copy, | |
1301 | super_used - num_bytes); | |
58176a96 JB |
1302 | |
1303 | /* block accounting for root item */ | |
db94535d | 1304 | root_used = btrfs_root_used(&root->root_item); |
5f39d397 | 1305 | btrfs_set_root_used(&root->root_item, |
db94535d | 1306 | root_used - num_bytes); |
58176a96 | 1307 | |
5caf2a00 | 1308 | ret = btrfs_del_item(trans, extent_root, path); |
54aa1f4d CM |
1309 | if (ret) { |
1310 | return ret; | |
1311 | } | |
db94535d | 1312 | ret = update_block_group(trans, root, bytenr, num_bytes, 0, |
1e2677e0 | 1313 | mark_free, 0); |
9078a3e1 | 1314 | BUG_ON(ret); |
a28ec197 | 1315 | } |
5caf2a00 | 1316 | btrfs_free_path(path); |
e089f05c | 1317 | finish_current_insert(trans, extent_root); |
a28ec197 CM |
1318 | return ret; |
1319 | } | |
1320 | ||
a28ec197 CM |
1321 | /* |
1322 | * find all the blocks marked as pending in the radix tree and remove | |
1323 | * them from the extent map | |
1324 | */ | |
e089f05c CM |
1325 | static int del_pending_extents(struct btrfs_trans_handle *trans, struct |
1326 | btrfs_root *extent_root) | |
a28ec197 CM |
1327 | { |
1328 | int ret; | |
e20d96d6 | 1329 | int err = 0; |
1a5bc167 CM |
1330 | u64 start; |
1331 | u64 end; | |
1332 | struct extent_map_tree *pending_del; | |
1333 | struct extent_map_tree *pinned_extents; | |
8ef97622 | 1334 | |
1a5bc167 CM |
1335 | pending_del = &extent_root->fs_info->pending_del; |
1336 | pinned_extents = &extent_root->fs_info->pinned_extents; | |
a28ec197 CM |
1337 | |
1338 | while(1) { | |
1a5bc167 CM |
1339 | ret = find_first_extent_bit(pending_del, 0, &start, &end, |
1340 | EXTENT_LOCKED); | |
1341 | if (ret) | |
a28ec197 | 1342 | break; |
324ae4df | 1343 | update_pinned_extents(extent_root, start, end + 1 - start, 1); |
1a5bc167 CM |
1344 | clear_extent_bits(pending_del, start, end, EXTENT_LOCKED, |
1345 | GFP_NOFS); | |
1346 | ret = __free_extent(trans, extent_root, | |
7bb86316 CM |
1347 | start, end + 1 - start, |
1348 | extent_root->root_key.objectid, | |
1349 | 0, 0, 0, 0, 0); | |
1a5bc167 CM |
1350 | if (ret) |
1351 | err = ret; | |
fec577fb | 1352 | } |
e20d96d6 | 1353 | return err; |
fec577fb CM |
1354 | } |
1355 | ||
1356 | /* | |
1357 | * remove an extent from the root, returns 0 on success | |
1358 | */ | |
e089f05c | 1359 | int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root |
7bb86316 CM |
1360 | *root, u64 bytenr, u64 num_bytes, |
1361 | u64 root_objectid, u64 ref_generation, | |
1362 | u64 owner_objectid, u64 owner_offset, int pin) | |
fec577fb | 1363 | { |
9f5fae2f | 1364 | struct btrfs_root *extent_root = root->fs_info->extent_root; |
fec577fb CM |
1365 | int pending_ret; |
1366 | int ret; | |
a28ec197 | 1367 | |
db94535d | 1368 | WARN_ON(num_bytes < root->sectorsize); |
7bb86316 CM |
1369 | if (!root->ref_cows) |
1370 | ref_generation = 0; | |
1371 | ||
fec577fb | 1372 | if (root == extent_root) { |
db94535d | 1373 | pin_down_bytes(root, bytenr, num_bytes, 1); |
fec577fb CM |
1374 | return 0; |
1375 | } | |
7bb86316 CM |
1376 | ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid, |
1377 | ref_generation, owner_objectid, owner_offset, | |
1378 | pin, pin == 0); | |
e20d96d6 | 1379 | pending_ret = del_pending_extents(trans, root->fs_info->extent_root); |
fec577fb CM |
1380 | return ret ? ret : pending_ret; |
1381 | } | |
1382 | ||
87ee04eb CM |
1383 | static u64 stripe_align(struct btrfs_root *root, u64 val) |
1384 | { | |
1385 | u64 mask = ((u64)root->stripesize - 1); | |
1386 | u64 ret = (val + mask) & ~mask; | |
1387 | return ret; | |
1388 | } | |
1389 | ||
fec577fb CM |
1390 | /* |
1391 | * walks the btree of allocated extents and find a hole of a given size. | |
1392 | * The key ins is changed to record the hole: | |
1393 | * ins->objectid == block start | |
62e2749e | 1394 | * ins->flags = BTRFS_EXTENT_ITEM_KEY |
fec577fb CM |
1395 | * ins->offset == number of blocks |
1396 | * Any available blocks before search_start are skipped. | |
1397 | */ | |
98ed5174 CM |
1398 | static int noinline find_free_extent(struct btrfs_trans_handle *trans, |
1399 | struct btrfs_root *orig_root, | |
1400 | u64 num_bytes, u64 empty_size, | |
1401 | u64 search_start, u64 search_end, | |
1402 | u64 hint_byte, struct btrfs_key *ins, | |
1403 | u64 exclude_start, u64 exclude_nr, | |
1404 | int data) | |
fec577fb | 1405 | { |
5caf2a00 | 1406 | struct btrfs_path *path; |
e2fa7227 | 1407 | struct btrfs_key key; |
fec577fb | 1408 | u64 hole_size = 0; |
87ee04eb CM |
1409 | u64 aligned; |
1410 | int ret; | |
fec577fb | 1411 | int slot = 0; |
db94535d | 1412 | u64 last_byte = 0; |
be744175 | 1413 | u64 orig_search_start = search_start; |
fec577fb | 1414 | int start_found; |
5f39d397 | 1415 | struct extent_buffer *l; |
9f5fae2f | 1416 | struct btrfs_root * root = orig_root->fs_info->extent_root; |
f2458e1d | 1417 | struct btrfs_fs_info *info = root->fs_info; |
db94535d | 1418 | u64 total_needed = num_bytes; |
e20d96d6 | 1419 | int level; |
be08c1b9 | 1420 | struct btrfs_block_group_cache *block_group; |
be744175 | 1421 | int full_scan = 0; |
fbdc762b | 1422 | int wrapped = 0; |
f84a8b36 | 1423 | u64 cached_start; |
fec577fb | 1424 | |
db94535d | 1425 | WARN_ON(num_bytes < root->sectorsize); |
b1a4d965 CM |
1426 | btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY); |
1427 | ||
5f39d397 CM |
1428 | level = btrfs_header_level(root->node); |
1429 | ||
015a739c | 1430 | if (num_bytes >= 32 * 1024 * 1024 && hint_byte) { |
257d0ce3 CM |
1431 | data = BTRFS_BLOCK_GROUP_MIXED; |
1432 | } | |
1433 | ||
c31f8830 CM |
1434 | search_end = min(search_end, |
1435 | btrfs_super_total_bytes(&info->super_copy)); | |
db94535d CM |
1436 | if (hint_byte) { |
1437 | block_group = btrfs_lookup_block_group(info, hint_byte); | |
1a2b2ac7 CM |
1438 | if (!block_group) |
1439 | hint_byte = search_start; | |
be744175 | 1440 | block_group = btrfs_find_block_group(root, block_group, |
db94535d | 1441 | hint_byte, data, 1); |
be744175 CM |
1442 | } else { |
1443 | block_group = btrfs_find_block_group(root, | |
1a2b2ac7 CM |
1444 | trans->block_group, |
1445 | search_start, data, 1); | |
be744175 CM |
1446 | } |
1447 | ||
6702ed49 | 1448 | total_needed += empty_size; |
e011599b | 1449 | path = btrfs_alloc_path(); |
be744175 | 1450 | check_failed: |
70b043f0 CM |
1451 | if (!block_group) { |
1452 | block_group = btrfs_lookup_block_group(info, search_start); | |
1453 | if (!block_group) | |
1454 | block_group = btrfs_lookup_block_group(info, | |
1455 | orig_search_start); | |
1456 | } | |
5e5745dc | 1457 | search_start = find_search_start(root, &block_group, search_start, |
d548ee51 | 1458 | total_needed, data); |
87ee04eb | 1459 | search_start = stripe_align(root, search_start); |
f84a8b36 | 1460 | cached_start = search_start; |
5caf2a00 | 1461 | btrfs_init_path(path); |
fec577fb CM |
1462 | ins->objectid = search_start; |
1463 | ins->offset = 0; | |
fec577fb | 1464 | start_found = 0; |
2cc58cf2 | 1465 | path->reada = 2; |
e37c9e69 | 1466 | |
5caf2a00 | 1467 | ret = btrfs_search_slot(trans, root, ins, path, 0, 0); |
0f70abe2 CM |
1468 | if (ret < 0) |
1469 | goto error; | |
d548ee51 Y |
1470 | ret = find_previous_extent(root, path); |
1471 | if (ret < 0) | |
1472 | goto error; | |
5f39d397 CM |
1473 | l = path->nodes[0]; |
1474 | btrfs_item_key_to_cpu(l, &key, path->slots[0]); | |
fec577fb | 1475 | while (1) { |
5f39d397 | 1476 | l = path->nodes[0]; |
5caf2a00 | 1477 | slot = path->slots[0]; |
5f39d397 | 1478 | if (slot >= btrfs_header_nritems(l)) { |
5caf2a00 | 1479 | ret = btrfs_next_leaf(root, path); |
fec577fb CM |
1480 | if (ret == 0) |
1481 | continue; | |
0f70abe2 CM |
1482 | if (ret < 0) |
1483 | goto error; | |
e19caa5f CM |
1484 | |
1485 | search_start = max(search_start, | |
1486 | block_group->key.objectid); | |
fec577fb | 1487 | if (!start_found) { |
87ee04eb CM |
1488 | aligned = stripe_align(root, search_start); |
1489 | ins->objectid = aligned; | |
1490 | if (aligned >= search_end) { | |
1491 | ret = -ENOSPC; | |
1492 | goto error; | |
1493 | } | |
1494 | ins->offset = search_end - aligned; | |
fec577fb CM |
1495 | start_found = 1; |
1496 | goto check_pending; | |
1497 | } | |
87ee04eb CM |
1498 | ins->objectid = stripe_align(root, |
1499 | last_byte > search_start ? | |
1500 | last_byte : search_start); | |
1501 | if (search_end <= ins->objectid) { | |
1502 | ret = -ENOSPC; | |
1503 | goto error; | |
1504 | } | |
3e1ad54f | 1505 | ins->offset = search_end - ins->objectid; |
e19caa5f | 1506 | BUG_ON(ins->objectid >= search_end); |
fec577fb CM |
1507 | goto check_pending; |
1508 | } | |
5f39d397 | 1509 | btrfs_item_key_to_cpu(l, &key, slot); |
96b5179d | 1510 | |
db94535d | 1511 | if (key.objectid >= search_start && key.objectid > last_byte && |
e37c9e69 | 1512 | start_found) { |
db94535d CM |
1513 | if (last_byte < search_start) |
1514 | last_byte = search_start; | |
87ee04eb CM |
1515 | aligned = stripe_align(root, last_byte); |
1516 | hole_size = key.objectid - aligned; | |
1517 | if (key.objectid > aligned && hole_size >= num_bytes) { | |
1518 | ins->objectid = aligned; | |
e37c9e69 CM |
1519 | ins->offset = hole_size; |
1520 | goto check_pending; | |
0579da42 | 1521 | } |
fec577fb | 1522 | } |
96b5179d | 1523 | if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY) { |
7bb86316 CM |
1524 | if (!start_found && btrfs_key_type(&key) == |
1525 | BTRFS_BLOCK_GROUP_ITEM_KEY) { | |
db94535d | 1526 | last_byte = key.objectid; |
96b5179d CM |
1527 | start_found = 1; |
1528 | } | |
e37c9e69 | 1529 | goto next; |
96b5179d CM |
1530 | } |
1531 | ||
e37c9e69 | 1532 | |
0579da42 | 1533 | start_found = 1; |
db94535d | 1534 | last_byte = key.objectid + key.offset; |
f510cfec | 1535 | |
257d0ce3 CM |
1536 | if (!full_scan && data != BTRFS_BLOCK_GROUP_MIXED && |
1537 | last_byte >= block_group->key.objectid + | |
be744175 CM |
1538 | block_group->key.offset) { |
1539 | btrfs_release_path(root, path); | |
1540 | search_start = block_group->key.objectid + | |
e19caa5f | 1541 | block_group->key.offset; |
be744175 CM |
1542 | goto new_group; |
1543 | } | |
9078a3e1 | 1544 | next: |
5caf2a00 | 1545 | path->slots[0]++; |
de428b63 | 1546 | cond_resched(); |
fec577fb | 1547 | } |
fec577fb CM |
1548 | check_pending: |
1549 | /* we have to make sure we didn't find an extent that has already | |
1550 | * been allocated by the map tree or the original allocation | |
1551 | */ | |
5caf2a00 | 1552 | btrfs_release_path(root, path); |
fec577fb | 1553 | BUG_ON(ins->objectid < search_start); |
e37c9e69 | 1554 | |
db94535d | 1555 | if (ins->objectid + num_bytes >= search_end) |
cf67582b | 1556 | goto enospc; |
257d0ce3 | 1557 | if (!full_scan && data != BTRFS_BLOCK_GROUP_MIXED && |
5cf66426 | 1558 | ins->objectid + num_bytes > block_group-> |
e19caa5f CM |
1559 | key.objectid + block_group->key.offset) { |
1560 | search_start = block_group->key.objectid + | |
1561 | block_group->key.offset; | |
1562 | goto new_group; | |
1563 | } | |
1a5bc167 | 1564 | if (test_range_bit(&info->extent_ins, ins->objectid, |
db94535d CM |
1565 | ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) { |
1566 | search_start = ins->objectid + num_bytes; | |
1a5bc167 CM |
1567 | goto new_group; |
1568 | } | |
1569 | if (test_range_bit(&info->pinned_extents, ins->objectid, | |
db94535d CM |
1570 | ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) { |
1571 | search_start = ins->objectid + num_bytes; | |
1a5bc167 | 1572 | goto new_group; |
fec577fb | 1573 | } |
db94535d | 1574 | if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start && |
f2654de4 CM |
1575 | ins->objectid < exclude_start + exclude_nr)) { |
1576 | search_start = exclude_start + exclude_nr; | |
1577 | goto new_group; | |
1578 | } | |
e37c9e69 | 1579 | if (!data) { |
5276aeda | 1580 | block_group = btrfs_lookup_block_group(info, ins->objectid); |
26b8003f CM |
1581 | if (block_group) |
1582 | trans->block_group = block_group; | |
f2458e1d | 1583 | } |
db94535d | 1584 | ins->offset = num_bytes; |
5caf2a00 | 1585 | btrfs_free_path(path); |
fec577fb | 1586 | return 0; |
be744175 CM |
1587 | |
1588 | new_group: | |
db94535d | 1589 | if (search_start + num_bytes >= search_end) { |
cf67582b | 1590 | enospc: |
be744175 | 1591 | search_start = orig_search_start; |
fbdc762b CM |
1592 | if (full_scan) { |
1593 | ret = -ENOSPC; | |
1594 | goto error; | |
1595 | } | |
6702ed49 CM |
1596 | if (wrapped) { |
1597 | if (!full_scan) | |
1598 | total_needed -= empty_size; | |
fbdc762b | 1599 | full_scan = 1; |
1a2b2ac7 | 1600 | data = BTRFS_BLOCK_GROUP_MIXED; |
6702ed49 | 1601 | } else |
fbdc762b | 1602 | wrapped = 1; |
be744175 | 1603 | } |
5276aeda | 1604 | block_group = btrfs_lookup_block_group(info, search_start); |
fbdc762b | 1605 | cond_resched(); |
1a2b2ac7 CM |
1606 | block_group = btrfs_find_block_group(root, block_group, |
1607 | search_start, data, 0); | |
be744175 CM |
1608 | goto check_failed; |
1609 | ||
0f70abe2 | 1610 | error: |
5caf2a00 CM |
1611 | btrfs_release_path(root, path); |
1612 | btrfs_free_path(path); | |
0f70abe2 | 1613 | return ret; |
fec577fb | 1614 | } |
fec577fb CM |
1615 | /* |
1616 | * finds a free extent and does all the dirty work required for allocation | |
1617 | * returns the key for the extent through ins, and a tree buffer for | |
1618 | * the first block of the extent through buf. | |
1619 | * | |
1620 | * returns 0 if everything worked, non-zero otherwise. | |
1621 | */ | |
4d775673 | 1622 | int btrfs_alloc_extent(struct btrfs_trans_handle *trans, |
7bb86316 CM |
1623 | struct btrfs_root *root, |
1624 | u64 num_bytes, u64 root_objectid, u64 ref_generation, | |
1625 | u64 owner, u64 owner_offset, | |
1626 | u64 empty_size, u64 hint_byte, | |
be08c1b9 | 1627 | u64 search_end, struct btrfs_key *ins, int data) |
fec577fb CM |
1628 | { |
1629 | int ret; | |
1630 | int pending_ret; | |
c31f8830 CM |
1631 | u64 super_used; |
1632 | u64 root_used; | |
fbdc762b | 1633 | u64 search_start = 0; |
edbd8d4e | 1634 | u64 new_hint; |
1261ec42 CM |
1635 | struct btrfs_fs_info *info = root->fs_info; |
1636 | struct btrfs_root *extent_root = info->extent_root; | |
234b63a0 | 1637 | struct btrfs_extent_item extent_item; |
7bb86316 | 1638 | struct btrfs_path *path; |
037e6390 | 1639 | |
5f39d397 | 1640 | btrfs_set_stack_extent_refs(&extent_item, 1); |
8f662a76 CM |
1641 | |
1642 | new_hint = max(hint_byte, root->fs_info->alloc_start); | |
edbd8d4e CM |
1643 | if (new_hint < btrfs_super_total_bytes(&info->super_copy)) |
1644 | hint_byte = new_hint; | |
8f662a76 | 1645 | |
db94535d CM |
1646 | WARN_ON(num_bytes < root->sectorsize); |
1647 | ret = find_free_extent(trans, root, num_bytes, empty_size, | |
1648 | search_start, search_end, hint_byte, ins, | |
26b8003f CM |
1649 | trans->alloc_exclude_start, |
1650 | trans->alloc_exclude_nr, data); | |
ccd467d6 | 1651 | BUG_ON(ret); |
f2654de4 CM |
1652 | if (ret) |
1653 | return ret; | |
fec577fb | 1654 | |
58176a96 | 1655 | /* block accounting for super block */ |
db94535d CM |
1656 | super_used = btrfs_super_bytes_used(&info->super_copy); |
1657 | btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes); | |
26b8003f | 1658 | |
58176a96 | 1659 | /* block accounting for root item */ |
db94535d CM |
1660 | root_used = btrfs_root_used(&root->root_item); |
1661 | btrfs_set_root_used(&root->root_item, root_used + num_bytes); | |
58176a96 | 1662 | |
f510cfec CM |
1663 | clear_extent_dirty(&root->fs_info->free_space_cache, |
1664 | ins->objectid, ins->objectid + ins->offset - 1, | |
1665 | GFP_NOFS); | |
1666 | ||
26b8003f | 1667 | if (root == extent_root) { |
1a5bc167 CM |
1668 | set_extent_bits(&root->fs_info->extent_ins, ins->objectid, |
1669 | ins->objectid + ins->offset - 1, | |
1670 | EXTENT_LOCKED, GFP_NOFS); | |
e19caa5f | 1671 | WARN_ON(data == 1); |
26b8003f CM |
1672 | goto update_block; |
1673 | } | |
1674 | ||
1675 | WARN_ON(trans->alloc_exclude_nr); | |
1676 | trans->alloc_exclude_start = ins->objectid; | |
1677 | trans->alloc_exclude_nr = ins->offset; | |
e089f05c CM |
1678 | ret = btrfs_insert_item(trans, extent_root, ins, &extent_item, |
1679 | sizeof(extent_item)); | |
037e6390 | 1680 | |
26b8003f CM |
1681 | trans->alloc_exclude_start = 0; |
1682 | trans->alloc_exclude_nr = 0; | |
7bb86316 CM |
1683 | BUG_ON(ret); |
1684 | ||
1685 | path = btrfs_alloc_path(); | |
1686 | BUG_ON(!path); | |
1687 | ret = btrfs_insert_extent_backref(trans, extent_root, path, | |
1688 | ins->objectid, root_objectid, | |
1689 | ref_generation, owner, owner_offset); | |
26b8003f | 1690 | |
ccd467d6 | 1691 | BUG_ON(ret); |
7bb86316 | 1692 | btrfs_free_path(path); |
e089f05c | 1693 | finish_current_insert(trans, extent_root); |
e20d96d6 | 1694 | pending_ret = del_pending_extents(trans, extent_root); |
f510cfec | 1695 | |
e37c9e69 | 1696 | if (ret) { |
037e6390 | 1697 | return ret; |
e37c9e69 CM |
1698 | } |
1699 | if (pending_ret) { | |
037e6390 | 1700 | return pending_ret; |
e37c9e69 | 1701 | } |
26b8003f CM |
1702 | |
1703 | update_block: | |
1e2677e0 CM |
1704 | ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0, |
1705 | data); | |
fabb5681 | 1706 | BUG_ON(ret); |
037e6390 | 1707 | return 0; |
fec577fb CM |
1708 | } |
1709 | ||
1710 | /* | |
1711 | * helper function to allocate a block for a given tree | |
1712 | * returns the tree buffer or NULL. | |
1713 | */ | |
5f39d397 | 1714 | struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, |
db94535d | 1715 | struct btrfs_root *root, |
7bb86316 CM |
1716 | u32 blocksize, |
1717 | u64 root_objectid, u64 hint, | |
1718 | u64 empty_size) | |
1719 | { | |
1720 | u64 ref_generation; | |
1721 | ||
1722 | if (root->ref_cows) | |
1723 | ref_generation = trans->transid; | |
1724 | else | |
1725 | ref_generation = 0; | |
1726 | ||
1727 | ||
1728 | return __btrfs_alloc_free_block(trans, root, blocksize, root_objectid, | |
1729 | ref_generation, 0, 0, hint, empty_size); | |
1730 | } | |
1731 | ||
1732 | /* | |
1733 | * helper function to allocate a block for a given tree | |
1734 | * returns the tree buffer or NULL. | |
1735 | */ | |
1736 | struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, | |
1737 | struct btrfs_root *root, | |
1738 | u32 blocksize, | |
1739 | u64 root_objectid, | |
1740 | u64 ref_generation, | |
1741 | u64 first_objectid, | |
1742 | int level, | |
1743 | u64 hint, | |
5f39d397 | 1744 | u64 empty_size) |
fec577fb | 1745 | { |
e2fa7227 | 1746 | struct btrfs_key ins; |
fec577fb | 1747 | int ret; |
5f39d397 | 1748 | struct extent_buffer *buf; |
fec577fb | 1749 | |
7bb86316 CM |
1750 | ret = btrfs_alloc_extent(trans, root, blocksize, |
1751 | root_objectid, ref_generation, | |
f6dbff55 | 1752 | level, first_objectid, empty_size, hint, |
db94535d | 1753 | (u64)-1, &ins, 0); |
fec577fb | 1754 | if (ret) { |
54aa1f4d CM |
1755 | BUG_ON(ret > 0); |
1756 | return ERR_PTR(ret); | |
fec577fb | 1757 | } |
db94535d | 1758 | buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize); |
54aa1f4d | 1759 | if (!buf) { |
7bb86316 CM |
1760 | btrfs_free_extent(trans, root, ins.objectid, blocksize, |
1761 | root->root_key.objectid, ref_generation, | |
1762 | 0, 0, 0); | |
54aa1f4d CM |
1763 | return ERR_PTR(-ENOMEM); |
1764 | } | |
55c69072 CM |
1765 | btrfs_set_header_generation(buf, trans->transid); |
1766 | clean_tree_block(trans, root, buf); | |
1767 | wait_on_tree_block_writeback(root, buf); | |
5f39d397 | 1768 | btrfs_set_buffer_uptodate(buf); |
55c69072 CM |
1769 | |
1770 | if (PageDirty(buf->first_page)) { | |
1771 | printk("page %lu dirty\n", buf->first_page->index); | |
1772 | WARN_ON(1); | |
1773 | } | |
1774 | ||
5f39d397 CM |
1775 | set_extent_dirty(&trans->transaction->dirty_pages, buf->start, |
1776 | buf->start + buf->len - 1, GFP_NOFS); | |
19c00ddc CM |
1777 | set_extent_bits(&BTRFS_I(root->fs_info->btree_inode)->extent_tree, |
1778 | buf->start, buf->start + buf->len - 1, | |
1779 | EXTENT_CSUM, GFP_NOFS); | |
1780 | buf->flags |= EXTENT_CSUM; | |
6b80053d | 1781 | btrfs_set_buffer_defrag(buf); |
d3c2fdcf | 1782 | trans->blocks_used++; |
fec577fb CM |
1783 | return buf; |
1784 | } | |
a28ec197 | 1785 | |
98ed5174 CM |
1786 | static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans, |
1787 | struct btrfs_root *root, | |
1788 | struct extent_buffer *leaf) | |
6407bf6d | 1789 | { |
7bb86316 CM |
1790 | u64 leaf_owner; |
1791 | u64 leaf_generation; | |
5f39d397 | 1792 | struct btrfs_key key; |
6407bf6d CM |
1793 | struct btrfs_file_extent_item *fi; |
1794 | int i; | |
1795 | int nritems; | |
1796 | int ret; | |
1797 | ||
5f39d397 CM |
1798 | BUG_ON(!btrfs_is_leaf(leaf)); |
1799 | nritems = btrfs_header_nritems(leaf); | |
7bb86316 CM |
1800 | leaf_owner = btrfs_header_owner(leaf); |
1801 | leaf_generation = btrfs_header_generation(leaf); | |
1802 | ||
6407bf6d | 1803 | for (i = 0; i < nritems; i++) { |
db94535d | 1804 | u64 disk_bytenr; |
5f39d397 CM |
1805 | |
1806 | btrfs_item_key_to_cpu(leaf, &key, i); | |
1807 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
6407bf6d CM |
1808 | continue; |
1809 | fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item); | |
5f39d397 CM |
1810 | if (btrfs_file_extent_type(leaf, fi) == |
1811 | BTRFS_FILE_EXTENT_INLINE) | |
236454df | 1812 | continue; |
6407bf6d CM |
1813 | /* |
1814 | * FIXME make sure to insert a trans record that | |
1815 | * repeats the snapshot del on crash | |
1816 | */ | |
db94535d CM |
1817 | disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); |
1818 | if (disk_bytenr == 0) | |
3a686375 | 1819 | continue; |
db94535d | 1820 | ret = btrfs_free_extent(trans, root, disk_bytenr, |
7bb86316 CM |
1821 | btrfs_file_extent_disk_num_bytes(leaf, fi), |
1822 | leaf_owner, leaf_generation, | |
1823 | key.objectid, key.offset, 0); | |
6407bf6d CM |
1824 | BUG_ON(ret); |
1825 | } | |
1826 | return 0; | |
1827 | } | |
1828 | ||
98ed5174 CM |
1829 | static void noinline reada_walk_down(struct btrfs_root *root, |
1830 | struct extent_buffer *node) | |
e011599b CM |
1831 | { |
1832 | int i; | |
1833 | u32 nritems; | |
db94535d | 1834 | u64 bytenr; |
e011599b CM |
1835 | int ret; |
1836 | u32 refs; | |
db94535d CM |
1837 | int level; |
1838 | u32 blocksize; | |
e011599b | 1839 | |
5f39d397 | 1840 | nritems = btrfs_header_nritems(node); |
db94535d | 1841 | level = btrfs_header_level(node); |
e011599b | 1842 | for (i = 0; i < nritems; i++) { |
db94535d CM |
1843 | bytenr = btrfs_node_blockptr(node, i); |
1844 | blocksize = btrfs_level_size(root, level - 1); | |
1845 | ret = lookup_extent_ref(NULL, root, bytenr, blocksize, &refs); | |
e011599b CM |
1846 | BUG_ON(ret); |
1847 | if (refs != 1) | |
1848 | continue; | |
409eb95d | 1849 | mutex_unlock(&root->fs_info->fs_mutex); |
db94535d | 1850 | ret = readahead_tree_block(root, bytenr, blocksize); |
409eb95d CM |
1851 | cond_resched(); |
1852 | mutex_lock(&root->fs_info->fs_mutex); | |
e011599b CM |
1853 | if (ret) |
1854 | break; | |
1855 | } | |
1856 | } | |
1857 | ||
9aca1d51 CM |
1858 | /* |
1859 | * helper function for drop_snapshot, this walks down the tree dropping ref | |
1860 | * counts as it goes. | |
1861 | */ | |
98ed5174 CM |
1862 | static int noinline walk_down_tree(struct btrfs_trans_handle *trans, |
1863 | struct btrfs_root *root, | |
1864 | struct btrfs_path *path, int *level) | |
20524f02 | 1865 | { |
7bb86316 CM |
1866 | u64 root_owner; |
1867 | u64 root_gen; | |
1868 | u64 bytenr; | |
5f39d397 CM |
1869 | struct extent_buffer *next; |
1870 | struct extent_buffer *cur; | |
7bb86316 | 1871 | struct extent_buffer *parent; |
db94535d | 1872 | u32 blocksize; |
20524f02 CM |
1873 | int ret; |
1874 | u32 refs; | |
1875 | ||
5caf2a00 CM |
1876 | WARN_ON(*level < 0); |
1877 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
5f39d397 | 1878 | ret = lookup_extent_ref(trans, root, |
db94535d CM |
1879 | path->nodes[*level]->start, |
1880 | path->nodes[*level]->len, &refs); | |
20524f02 CM |
1881 | BUG_ON(ret); |
1882 | if (refs > 1) | |
1883 | goto out; | |
e011599b | 1884 | |
9aca1d51 CM |
1885 | /* |
1886 | * walk down to the last node level and free all the leaves | |
1887 | */ | |
6407bf6d | 1888 | while(*level >= 0) { |
5caf2a00 CM |
1889 | WARN_ON(*level < 0); |
1890 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
20524f02 | 1891 | cur = path->nodes[*level]; |
e011599b CM |
1892 | |
1893 | if (*level > 0 && path->slots[*level] == 0) | |
5f39d397 | 1894 | reada_walk_down(root, cur); |
e011599b | 1895 | |
5f39d397 | 1896 | if (btrfs_header_level(cur) != *level) |
2c90e5d6 | 1897 | WARN_ON(1); |
e011599b | 1898 | |
7518a238 | 1899 | if (path->slots[*level] >= |
5f39d397 | 1900 | btrfs_header_nritems(cur)) |
20524f02 | 1901 | break; |
6407bf6d CM |
1902 | if (*level == 0) { |
1903 | ret = drop_leaf_ref(trans, root, cur); | |
1904 | BUG_ON(ret); | |
1905 | break; | |
1906 | } | |
db94535d CM |
1907 | bytenr = btrfs_node_blockptr(cur, path->slots[*level]); |
1908 | blocksize = btrfs_level_size(root, *level - 1); | |
1909 | ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs); | |
6407bf6d CM |
1910 | BUG_ON(ret); |
1911 | if (refs != 1) { | |
7bb86316 CM |
1912 | parent = path->nodes[*level]; |
1913 | root_owner = btrfs_header_owner(parent); | |
1914 | root_gen = btrfs_header_generation(parent); | |
20524f02 | 1915 | path->slots[*level]++; |
db94535d | 1916 | ret = btrfs_free_extent(trans, root, bytenr, |
7bb86316 CM |
1917 | blocksize, root_owner, |
1918 | root_gen, 0, 0, 1); | |
20524f02 CM |
1919 | BUG_ON(ret); |
1920 | continue; | |
1921 | } | |
db94535d | 1922 | next = btrfs_find_tree_block(root, bytenr, blocksize); |
5f39d397 CM |
1923 | if (!next || !btrfs_buffer_uptodate(next)) { |
1924 | free_extent_buffer(next); | |
e9d0b13b | 1925 | mutex_unlock(&root->fs_info->fs_mutex); |
db94535d | 1926 | next = read_tree_block(root, bytenr, blocksize); |
e9d0b13b CM |
1927 | mutex_lock(&root->fs_info->fs_mutex); |
1928 | ||
1929 | /* we dropped the lock, check one more time */ | |
db94535d CM |
1930 | ret = lookup_extent_ref(trans, root, bytenr, |
1931 | blocksize, &refs); | |
e9d0b13b CM |
1932 | BUG_ON(ret); |
1933 | if (refs != 1) { | |
7bb86316 CM |
1934 | parent = path->nodes[*level]; |
1935 | root_owner = btrfs_header_owner(parent); | |
1936 | root_gen = btrfs_header_generation(parent); | |
1937 | ||
e9d0b13b | 1938 | path->slots[*level]++; |
5f39d397 | 1939 | free_extent_buffer(next); |
7bb86316 CM |
1940 | ret = btrfs_free_extent(trans, root, bytenr, |
1941 | blocksize, | |
1942 | root_owner, | |
1943 | root_gen, 0, 0, 1); | |
e9d0b13b CM |
1944 | BUG_ON(ret); |
1945 | continue; | |
1946 | } | |
1947 | } | |
5caf2a00 | 1948 | WARN_ON(*level <= 0); |
83e15a28 | 1949 | if (path->nodes[*level-1]) |
5f39d397 | 1950 | free_extent_buffer(path->nodes[*level-1]); |
20524f02 | 1951 | path->nodes[*level-1] = next; |
5f39d397 | 1952 | *level = btrfs_header_level(next); |
20524f02 CM |
1953 | path->slots[*level] = 0; |
1954 | } | |
1955 | out: | |
5caf2a00 CM |
1956 | WARN_ON(*level < 0); |
1957 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
7bb86316 CM |
1958 | |
1959 | if (path->nodes[*level] == root->node) { | |
1960 | root_owner = root->root_key.objectid; | |
1961 | parent = path->nodes[*level]; | |
1962 | } else { | |
1963 | parent = path->nodes[*level + 1]; | |
1964 | root_owner = btrfs_header_owner(parent); | |
1965 | } | |
1966 | ||
1967 | root_gen = btrfs_header_generation(parent); | |
db94535d | 1968 | ret = btrfs_free_extent(trans, root, path->nodes[*level]->start, |
7bb86316 CM |
1969 | path->nodes[*level]->len, |
1970 | root_owner, root_gen, 0, 0, 1); | |
5f39d397 | 1971 | free_extent_buffer(path->nodes[*level]); |
20524f02 CM |
1972 | path->nodes[*level] = NULL; |
1973 | *level += 1; | |
1974 | BUG_ON(ret); | |
1975 | return 0; | |
1976 | } | |
1977 | ||
9aca1d51 CM |
1978 | /* |
1979 | * helper for dropping snapshots. This walks back up the tree in the path | |
1980 | * to find the first node higher up where we haven't yet gone through | |
1981 | * all the slots | |
1982 | */ | |
98ed5174 CM |
1983 | static int noinline walk_up_tree(struct btrfs_trans_handle *trans, |
1984 | struct btrfs_root *root, | |
1985 | struct btrfs_path *path, int *level) | |
20524f02 | 1986 | { |
7bb86316 CM |
1987 | u64 root_owner; |
1988 | u64 root_gen; | |
1989 | struct btrfs_root_item *root_item = &root->root_item; | |
20524f02 CM |
1990 | int i; |
1991 | int slot; | |
1992 | int ret; | |
9f3a7427 | 1993 | |
234b63a0 | 1994 | for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { |
20524f02 | 1995 | slot = path->slots[i]; |
5f39d397 CM |
1996 | if (slot < btrfs_header_nritems(path->nodes[i]) - 1) { |
1997 | struct extent_buffer *node; | |
1998 | struct btrfs_disk_key disk_key; | |
1999 | node = path->nodes[i]; | |
20524f02 CM |
2000 | path->slots[i]++; |
2001 | *level = i; | |
9f3a7427 | 2002 | WARN_ON(*level == 0); |
5f39d397 | 2003 | btrfs_node_key(node, &disk_key, path->slots[i]); |
9f3a7427 | 2004 | memcpy(&root_item->drop_progress, |
5f39d397 | 2005 | &disk_key, sizeof(disk_key)); |
9f3a7427 | 2006 | root_item->drop_level = i; |
20524f02 CM |
2007 | return 0; |
2008 | } else { | |
7bb86316 CM |
2009 | if (path->nodes[*level] == root->node) { |
2010 | root_owner = root->root_key.objectid; | |
2011 | root_gen = | |
2012 | btrfs_header_generation(path->nodes[*level]); | |
2013 | } else { | |
2014 | struct extent_buffer *node; | |
2015 | node = path->nodes[*level + 1]; | |
2016 | root_owner = btrfs_header_owner(node); | |
2017 | root_gen = btrfs_header_generation(node); | |
2018 | } | |
e089f05c | 2019 | ret = btrfs_free_extent(trans, root, |
db94535d | 2020 | path->nodes[*level]->start, |
7bb86316 CM |
2021 | path->nodes[*level]->len, |
2022 | root_owner, root_gen, 0, 0, 1); | |
6407bf6d | 2023 | BUG_ON(ret); |
5f39d397 | 2024 | free_extent_buffer(path->nodes[*level]); |
83e15a28 | 2025 | path->nodes[*level] = NULL; |
20524f02 | 2026 | *level = i + 1; |
20524f02 CM |
2027 | } |
2028 | } | |
2029 | return 1; | |
2030 | } | |
2031 | ||
9aca1d51 CM |
2032 | /* |
2033 | * drop the reference count on the tree rooted at 'snap'. This traverses | |
2034 | * the tree freeing any blocks that have a ref count of zero after being | |
2035 | * decremented. | |
2036 | */ | |
e089f05c | 2037 | int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root |
9f3a7427 | 2038 | *root) |
20524f02 | 2039 | { |
3768f368 | 2040 | int ret = 0; |
9aca1d51 | 2041 | int wret; |
20524f02 | 2042 | int level; |
5caf2a00 | 2043 | struct btrfs_path *path; |
20524f02 CM |
2044 | int i; |
2045 | int orig_level; | |
9f3a7427 | 2046 | struct btrfs_root_item *root_item = &root->root_item; |
20524f02 | 2047 | |
5caf2a00 CM |
2048 | path = btrfs_alloc_path(); |
2049 | BUG_ON(!path); | |
20524f02 | 2050 | |
5f39d397 | 2051 | level = btrfs_header_level(root->node); |
20524f02 | 2052 | orig_level = level; |
9f3a7427 CM |
2053 | if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) { |
2054 | path->nodes[level] = root->node; | |
f510cfec | 2055 | extent_buffer_get(root->node); |
9f3a7427 CM |
2056 | path->slots[level] = 0; |
2057 | } else { | |
2058 | struct btrfs_key key; | |
5f39d397 CM |
2059 | struct btrfs_disk_key found_key; |
2060 | struct extent_buffer *node; | |
6702ed49 | 2061 | |
9f3a7427 | 2062 | btrfs_disk_key_to_cpu(&key, &root_item->drop_progress); |
6702ed49 CM |
2063 | level = root_item->drop_level; |
2064 | path->lowest_level = level; | |
9f3a7427 | 2065 | wret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
6702ed49 | 2066 | if (wret < 0) { |
9f3a7427 CM |
2067 | ret = wret; |
2068 | goto out; | |
2069 | } | |
5f39d397 CM |
2070 | node = path->nodes[level]; |
2071 | btrfs_node_key(node, &found_key, path->slots[level]); | |
2072 | WARN_ON(memcmp(&found_key, &root_item->drop_progress, | |
2073 | sizeof(found_key))); | |
9f3a7427 | 2074 | } |
20524f02 | 2075 | while(1) { |
5caf2a00 | 2076 | wret = walk_down_tree(trans, root, path, &level); |
9aca1d51 | 2077 | if (wret > 0) |
20524f02 | 2078 | break; |
9aca1d51 CM |
2079 | if (wret < 0) |
2080 | ret = wret; | |
2081 | ||
5caf2a00 | 2082 | wret = walk_up_tree(trans, root, path, &level); |
9aca1d51 | 2083 | if (wret > 0) |
20524f02 | 2084 | break; |
9aca1d51 CM |
2085 | if (wret < 0) |
2086 | ret = wret; | |
409eb95d | 2087 | ret = -EAGAIN; |
409eb95d | 2088 | break; |
20524f02 | 2089 | } |
83e15a28 | 2090 | for (i = 0; i <= orig_level; i++) { |
5caf2a00 | 2091 | if (path->nodes[i]) { |
5f39d397 | 2092 | free_extent_buffer(path->nodes[i]); |
0f82731f | 2093 | path->nodes[i] = NULL; |
83e15a28 | 2094 | } |
20524f02 | 2095 | } |
9f3a7427 | 2096 | out: |
5caf2a00 | 2097 | btrfs_free_path(path); |
9aca1d51 | 2098 | return ret; |
20524f02 | 2099 | } |
9078a3e1 | 2100 | |
96b5179d | 2101 | int btrfs_free_block_groups(struct btrfs_fs_info *info) |
9078a3e1 | 2102 | { |
96b5179d CM |
2103 | u64 start; |
2104 | u64 end; | |
b97f9203 | 2105 | u64 ptr; |
9078a3e1 | 2106 | int ret; |
9078a3e1 | 2107 | while(1) { |
96b5179d CM |
2108 | ret = find_first_extent_bit(&info->block_group_cache, 0, |
2109 | &start, &end, (unsigned int)-1); | |
2110 | if (ret) | |
9078a3e1 | 2111 | break; |
b97f9203 Y |
2112 | ret = get_state_private(&info->block_group_cache, start, &ptr); |
2113 | if (!ret) | |
2114 | kfree((void *)(unsigned long)ptr); | |
96b5179d CM |
2115 | clear_extent_bits(&info->block_group_cache, start, |
2116 | end, (unsigned int)-1, GFP_NOFS); | |
9078a3e1 | 2117 | } |
e37c9e69 | 2118 | while(1) { |
f510cfec CM |
2119 | ret = find_first_extent_bit(&info->free_space_cache, 0, |
2120 | &start, &end, EXTENT_DIRTY); | |
2121 | if (ret) | |
e37c9e69 | 2122 | break; |
f510cfec CM |
2123 | clear_extent_dirty(&info->free_space_cache, start, |
2124 | end, GFP_NOFS); | |
e37c9e69 | 2125 | } |
be744175 CM |
2126 | return 0; |
2127 | } | |
2128 | ||
98ed5174 CM |
2129 | static int noinline relocate_inode_pages(struct inode *inode, u64 start, |
2130 | u64 len) | |
edbd8d4e CM |
2131 | { |
2132 | u64 page_start; | |
2133 | u64 page_end; | |
2134 | u64 delalloc_start; | |
2135 | u64 existing_delalloc; | |
2136 | unsigned long last_index; | |
edbd8d4e CM |
2137 | unsigned long i; |
2138 | struct page *page; | |
2139 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2140 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | |
4313b399 CM |
2141 | struct file_ra_state *ra; |
2142 | ||
2143 | ra = kzalloc(sizeof(*ra), GFP_NOFS); | |
edbd8d4e CM |
2144 | |
2145 | mutex_lock(&inode->i_mutex); | |
4313b399 | 2146 | i = start >> PAGE_CACHE_SHIFT; |
edbd8d4e CM |
2147 | last_index = (start + len - 1) >> PAGE_CACHE_SHIFT; |
2148 | ||
4313b399 CM |
2149 | file_ra_state_init(ra, inode->i_mapping); |
2150 | btrfs_force_ra(inode->i_mapping, ra, NULL, i, last_index); | |
2151 | kfree(ra); | |
edbd8d4e | 2152 | |
4313b399 | 2153 | for (; i <= last_index; i++) { |
edbd8d4e CM |
2154 | page = grab_cache_page(inode->i_mapping, i); |
2155 | if (!page) | |
2156 | goto out_unlock; | |
2157 | if (!PageUptodate(page)) { | |
2158 | btrfs_readpage(NULL, page); | |
2159 | lock_page(page); | |
2160 | if (!PageUptodate(page)) { | |
2161 | unlock_page(page); | |
2162 | page_cache_release(page); | |
2163 | goto out_unlock; | |
2164 | } | |
2165 | } | |
2166 | page_start = (u64)page->index << PAGE_CACHE_SHIFT; | |
2167 | page_end = page_start + PAGE_CACHE_SIZE - 1; | |
2168 | ||
2169 | lock_extent(em_tree, page_start, page_end, GFP_NOFS); | |
2170 | ||
2171 | delalloc_start = page_start; | |
2172 | existing_delalloc = | |
2173 | count_range_bits(&BTRFS_I(inode)->extent_tree, | |
2174 | &delalloc_start, page_end, | |
2175 | PAGE_CACHE_SIZE, EXTENT_DELALLOC); | |
2176 | ||
2177 | set_extent_delalloc(em_tree, page_start, | |
2178 | page_end, GFP_NOFS); | |
2179 | ||
2180 | spin_lock(&root->fs_info->delalloc_lock); | |
2181 | root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE - | |
2182 | existing_delalloc; | |
2183 | spin_unlock(&root->fs_info->delalloc_lock); | |
2184 | ||
2185 | unlock_extent(em_tree, page_start, page_end, GFP_NOFS); | |
2186 | set_page_dirty(page); | |
2187 | unlock_page(page); | |
2188 | page_cache_release(page); | |
2189 | } | |
2190 | ||
2191 | out_unlock: | |
2192 | mutex_unlock(&inode->i_mutex); | |
2193 | return 0; | |
2194 | } | |
2195 | ||
4313b399 CM |
2196 | /* |
2197 | * note, this releases the path | |
2198 | */ | |
98ed5174 | 2199 | static int noinline relocate_one_reference(struct btrfs_root *extent_root, |
edbd8d4e | 2200 | struct btrfs_path *path, |
4313b399 | 2201 | struct btrfs_key *extent_key) |
edbd8d4e CM |
2202 | { |
2203 | struct inode *inode; | |
2204 | struct btrfs_root *found_root; | |
4313b399 CM |
2205 | struct btrfs_key *root_location; |
2206 | struct btrfs_extent_ref *ref; | |
2207 | u64 ref_root; | |
2208 | u64 ref_gen; | |
2209 | u64 ref_objectid; | |
2210 | u64 ref_offset; | |
edbd8d4e CM |
2211 | int ret; |
2212 | ||
4313b399 CM |
2213 | ref = btrfs_item_ptr(path->nodes[0], path->slots[0], |
2214 | struct btrfs_extent_ref); | |
2215 | ref_root = btrfs_ref_root(path->nodes[0], ref); | |
2216 | ref_gen = btrfs_ref_generation(path->nodes[0], ref); | |
2217 | ref_objectid = btrfs_ref_objectid(path->nodes[0], ref); | |
2218 | ref_offset = btrfs_ref_offset(path->nodes[0], ref); | |
2219 | btrfs_release_path(extent_root, path); | |
2220 | ||
2221 | root_location = kmalloc(sizeof(*root_location), GFP_NOFS); | |
2222 | root_location->objectid = ref_root; | |
edbd8d4e | 2223 | if (ref_gen == 0) |
4313b399 | 2224 | root_location->offset = 0; |
edbd8d4e | 2225 | else |
4313b399 CM |
2226 | root_location->offset = (u64)-1; |
2227 | root_location->type = BTRFS_ROOT_ITEM_KEY; | |
edbd8d4e CM |
2228 | |
2229 | found_root = btrfs_read_fs_root_no_name(extent_root->fs_info, | |
4313b399 | 2230 | root_location); |
edbd8d4e | 2231 | BUG_ON(!found_root); |
4313b399 | 2232 | kfree(root_location); |
edbd8d4e CM |
2233 | |
2234 | if (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID) { | |
2235 | mutex_unlock(&extent_root->fs_info->fs_mutex); | |
2236 | inode = btrfs_iget_locked(extent_root->fs_info->sb, | |
2237 | ref_objectid, found_root); | |
2238 | if (inode->i_state & I_NEW) { | |
2239 | /* the inode and parent dir are two different roots */ | |
2240 | BTRFS_I(inode)->root = found_root; | |
2241 | BTRFS_I(inode)->location.objectid = ref_objectid; | |
2242 | BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY; | |
2243 | BTRFS_I(inode)->location.offset = 0; | |
2244 | btrfs_read_locked_inode(inode); | |
2245 | unlock_new_inode(inode); | |
2246 | ||
2247 | } | |
2248 | /* this can happen if the reference is not against | |
2249 | * the latest version of the tree root | |
2250 | */ | |
2251 | if (is_bad_inode(inode)) { | |
2252 | mutex_lock(&extent_root->fs_info->fs_mutex); | |
2253 | goto out; | |
2254 | } | |
2255 | relocate_inode_pages(inode, ref_offset, extent_key->offset); | |
2256 | /* FIXME, data=ordered will help get rid of this */ | |
2257 | filemap_fdatawrite(inode->i_mapping); | |
2258 | iput(inode); | |
2259 | mutex_lock(&extent_root->fs_info->fs_mutex); | |
2260 | } else { | |
2261 | struct btrfs_trans_handle *trans; | |
2262 | struct btrfs_key found_key; | |
2263 | struct extent_buffer *eb; | |
2264 | int level; | |
2265 | int i; | |
2266 | ||
2267 | trans = btrfs_start_transaction(found_root, 1); | |
2268 | eb = read_tree_block(found_root, extent_key->objectid, | |
2269 | extent_key->offset); | |
2270 | level = btrfs_header_level(eb); | |
2271 | ||
2272 | if (level == 0) | |
2273 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
2274 | else | |
2275 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
2276 | ||
2277 | free_extent_buffer(eb); | |
2278 | ||
2279 | path->lowest_level = level; | |
8f662a76 | 2280 | path->reada = 2; |
edbd8d4e CM |
2281 | ret = btrfs_search_slot(trans, found_root, &found_key, path, |
2282 | 0, 1); | |
2283 | path->lowest_level = 0; | |
2284 | for (i = level; i < BTRFS_MAX_LEVEL; i++) { | |
2285 | if (!path->nodes[i]) | |
2286 | break; | |
2287 | free_extent_buffer(path->nodes[i]); | |
2288 | path->nodes[i] = NULL; | |
2289 | } | |
2290 | btrfs_release_path(found_root, path); | |
2291 | btrfs_end_transaction(trans, found_root); | |
2292 | } | |
2293 | ||
2294 | out: | |
2295 | return 0; | |
2296 | } | |
2297 | ||
98ed5174 CM |
2298 | static int noinline relocate_one_extent(struct btrfs_root *extent_root, |
2299 | struct btrfs_path *path, | |
2300 | struct btrfs_key *extent_key) | |
edbd8d4e CM |
2301 | { |
2302 | struct btrfs_key key; | |
2303 | struct btrfs_key found_key; | |
edbd8d4e | 2304 | struct extent_buffer *leaf; |
edbd8d4e CM |
2305 | u32 nritems; |
2306 | u32 item_size; | |
2307 | int ret = 0; | |
2308 | ||
2309 | key.objectid = extent_key->objectid; | |
2310 | key.type = BTRFS_EXTENT_REF_KEY; | |
2311 | key.offset = 0; | |
2312 | ||
2313 | while(1) { | |
2314 | ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); | |
2315 | ||
edbd8d4e CM |
2316 | if (ret < 0) |
2317 | goto out; | |
2318 | ||
2319 | ret = 0; | |
2320 | leaf = path->nodes[0]; | |
2321 | nritems = btrfs_header_nritems(leaf); | |
2322 | if (path->slots[0] == nritems) | |
2323 | goto out; | |
2324 | ||
2325 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2326 | if (found_key.objectid != extent_key->objectid) | |
2327 | break; | |
2328 | ||
2329 | if (found_key.type != BTRFS_EXTENT_REF_KEY) | |
2330 | break; | |
2331 | ||
2332 | key.offset = found_key.offset + 1; | |
2333 | item_size = btrfs_item_size_nr(leaf, path->slots[0]); | |
2334 | ||
4313b399 | 2335 | ret = relocate_one_reference(extent_root, path, extent_key); |
edbd8d4e CM |
2336 | if (ret) |
2337 | goto out; | |
2338 | } | |
2339 | ret = 0; | |
2340 | out: | |
2341 | btrfs_release_path(extent_root, path); | |
2342 | return ret; | |
2343 | } | |
2344 | ||
edbd8d4e CM |
2345 | int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size) |
2346 | { | |
2347 | struct btrfs_trans_handle *trans; | |
2348 | struct btrfs_root *tree_root = root->fs_info->tree_root; | |
2349 | struct btrfs_path *path; | |
2350 | u64 cur_byte; | |
2351 | u64 total_found; | |
edbd8d4e CM |
2352 | struct btrfs_fs_info *info = root->fs_info; |
2353 | struct extent_map_tree *block_group_cache; | |
2354 | struct btrfs_key key; | |
73e48b27 | 2355 | struct btrfs_key found_key; |
edbd8d4e CM |
2356 | struct extent_buffer *leaf; |
2357 | u32 nritems; | |
2358 | int ret; | |
725c8463 | 2359 | int progress = 0; |
edbd8d4e CM |
2360 | |
2361 | btrfs_set_super_total_bytes(&info->super_copy, new_size); | |
c31f8830 CM |
2362 | clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1, |
2363 | GFP_NOFS); | |
edbd8d4e CM |
2364 | block_group_cache = &info->block_group_cache; |
2365 | path = btrfs_alloc_path(); | |
2366 | root = root->fs_info->extent_root; | |
8f662a76 | 2367 | path->reada = 2; |
edbd8d4e CM |
2368 | |
2369 | again: | |
2370 | total_found = 0; | |
2371 | key.objectid = new_size; | |
edbd8d4e CM |
2372 | key.offset = 0; |
2373 | key.type = 0; | |
73e48b27 | 2374 | cur_byte = key.objectid; |
4313b399 | 2375 | |
73e48b27 Y |
2376 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2377 | if (ret < 0) | |
2378 | goto out; | |
2379 | ||
2380 | ret = find_previous_extent(root, path); | |
2381 | if (ret < 0) | |
2382 | goto out; | |
2383 | if (ret == 0) { | |
2384 | leaf = path->nodes[0]; | |
2385 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2386 | if (found_key.objectid + found_key.offset > new_size) { | |
2387 | cur_byte = found_key.objectid; | |
2388 | key.objectid = cur_byte; | |
2389 | } | |
2390 | } | |
2391 | btrfs_release_path(root, path); | |
2392 | ||
2393 | while(1) { | |
edbd8d4e CM |
2394 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2395 | if (ret < 0) | |
2396 | goto out; | |
73e48b27 | 2397 | |
edbd8d4e | 2398 | leaf = path->nodes[0]; |
73e48b27 Y |
2399 | nritems = btrfs_header_nritems(leaf); |
2400 | next: | |
2401 | if (path->slots[0] >= nritems) { | |
2402 | ret = btrfs_next_leaf(root, path); | |
2403 | if (ret < 0) | |
2404 | goto out; | |
2405 | if (ret == 1) { | |
2406 | ret = 0; | |
2407 | break; | |
edbd8d4e | 2408 | } |
73e48b27 Y |
2409 | leaf = path->nodes[0]; |
2410 | nritems = btrfs_header_nritems(leaf); | |
edbd8d4e | 2411 | } |
73e48b27 Y |
2412 | |
2413 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
725c8463 CM |
2414 | |
2415 | if (progress && need_resched()) { | |
2416 | memcpy(&key, &found_key, sizeof(key)); | |
2417 | mutex_unlock(&root->fs_info->fs_mutex); | |
2418 | cond_resched(); | |
2419 | mutex_lock(&root->fs_info->fs_mutex); | |
2420 | btrfs_release_path(root, path); | |
2421 | btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
2422 | progress = 0; | |
2423 | goto next; | |
2424 | } | |
2425 | progress = 1; | |
2426 | ||
73e48b27 Y |
2427 | if (btrfs_key_type(&found_key) != BTRFS_EXTENT_ITEM_KEY || |
2428 | found_key.objectid + found_key.offset <= cur_byte) { | |
edbd8d4e | 2429 | path->slots[0]++; |
edbd8d4e CM |
2430 | goto next; |
2431 | } | |
73e48b27 | 2432 | |
edbd8d4e CM |
2433 | total_found++; |
2434 | cur_byte = found_key.objectid + found_key.offset; | |
2435 | key.objectid = cur_byte; | |
2436 | btrfs_release_path(root, path); | |
2437 | ret = relocate_one_extent(root, path, &found_key); | |
2438 | } | |
2439 | ||
2440 | btrfs_release_path(root, path); | |
2441 | ||
2442 | if (total_found > 0) { | |
2443 | trans = btrfs_start_transaction(tree_root, 1); | |
2444 | btrfs_commit_transaction(trans, tree_root); | |
2445 | ||
2446 | mutex_unlock(&root->fs_info->fs_mutex); | |
2447 | btrfs_clean_old_snapshots(tree_root); | |
2448 | mutex_lock(&root->fs_info->fs_mutex); | |
2449 | ||
2450 | trans = btrfs_start_transaction(tree_root, 1); | |
2451 | btrfs_commit_transaction(trans, tree_root); | |
2452 | goto again; | |
2453 | } | |
2454 | ||
2455 | trans = btrfs_start_transaction(root, 1); | |
2456 | key.objectid = new_size; | |
2457 | key.offset = 0; | |
2458 | key.type = 0; | |
2459 | while(1) { | |
4313b399 CM |
2460 | u64 ptr; |
2461 | ||
edbd8d4e CM |
2462 | ret = btrfs_search_slot(trans, root, &key, path, -1, 1); |
2463 | if (ret < 0) | |
2464 | goto out; | |
73e48b27 | 2465 | |
edbd8d4e CM |
2466 | leaf = path->nodes[0]; |
2467 | nritems = btrfs_header_nritems(leaf); | |
73e48b27 Y |
2468 | bg_next: |
2469 | if (path->slots[0] >= nritems) { | |
2470 | ret = btrfs_next_leaf(root, path); | |
2471 | if (ret < 0) | |
2472 | break; | |
2473 | if (ret == 1) { | |
2474 | ret = 0; | |
2475 | break; | |
edbd8d4e | 2476 | } |
73e48b27 | 2477 | leaf = path->nodes[0]; |
1372f8e6 CM |
2478 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); |
2479 | ||
2480 | /* | |
2481 | * btrfs_next_leaf doesn't cow buffers, we have to | |
2482 | * do the search again | |
2483 | */ | |
2484 | memcpy(&key, &found_key, sizeof(key)); | |
2485 | btrfs_release_path(root, path); | |
725c8463 | 2486 | goto resched_check; |
73e48b27 Y |
2487 | } |
2488 | ||
2489 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2490 | if (btrfs_key_type(&found_key) != BTRFS_BLOCK_GROUP_ITEM_KEY) { | |
2491 | printk("shrinker found key %Lu %u %Lu\n", | |
2492 | found_key.objectid, found_key.type, | |
2493 | found_key.offset); | |
2494 | path->slots[0]++; | |
edbd8d4e CM |
2495 | goto bg_next; |
2496 | } | |
edbd8d4e CM |
2497 | ret = get_state_private(&info->block_group_cache, |
2498 | found_key.objectid, &ptr); | |
2499 | if (!ret) | |
2500 | kfree((void *)(unsigned long)ptr); | |
2501 | ||
2502 | clear_extent_bits(&info->block_group_cache, found_key.objectid, | |
2503 | found_key.objectid + found_key.offset - 1, | |
2504 | (unsigned int)-1, GFP_NOFS); | |
2505 | ||
2506 | key.objectid = found_key.objectid + 1; | |
2507 | btrfs_del_item(trans, root, path); | |
2508 | btrfs_release_path(root, path); | |
725c8463 CM |
2509 | resched_check: |
2510 | if (need_resched()) { | |
2511 | mutex_unlock(&root->fs_info->fs_mutex); | |
2512 | cond_resched(); | |
2513 | mutex_lock(&root->fs_info->fs_mutex); | |
2514 | } | |
edbd8d4e CM |
2515 | } |
2516 | clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1, | |
2517 | GFP_NOFS); | |
2518 | btrfs_commit_transaction(trans, root); | |
2519 | out: | |
2520 | btrfs_free_path(path); | |
2521 | return ret; | |
2522 | } | |
2523 | ||
2524 | int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, | |
2525 | struct btrfs_root *root, u64 new_size) | |
2526 | { | |
2527 | struct btrfs_path *path; | |
2528 | u64 nr = 0; | |
2529 | u64 cur_byte; | |
2530 | u64 old_size; | |
f9ef6604 | 2531 | unsigned long rem; |
edbd8d4e CM |
2532 | struct btrfs_block_group_cache *cache; |
2533 | struct btrfs_block_group_item *item; | |
2534 | struct btrfs_fs_info *info = root->fs_info; | |
2535 | struct extent_map_tree *block_group_cache; | |
2536 | struct btrfs_key key; | |
2537 | struct extent_buffer *leaf; | |
2538 | int ret; | |
2539 | int bit; | |
2540 | ||
2541 | old_size = btrfs_super_total_bytes(&info->super_copy); | |
2542 | block_group_cache = &info->block_group_cache; | |
2543 | ||
2544 | root = info->extent_root; | |
2545 | ||
2546 | cache = btrfs_lookup_block_group(root->fs_info, old_size - 1); | |
2547 | ||
2548 | cur_byte = cache->key.objectid + cache->key.offset; | |
2549 | if (cur_byte >= new_size) | |
2550 | goto set_size; | |
2551 | ||
2552 | key.offset = BTRFS_BLOCK_GROUP_SIZE; | |
2553 | btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); | |
2554 | ||
2555 | path = btrfs_alloc_path(); | |
2556 | if (!path) | |
2557 | return -ENOMEM; | |
2558 | ||
2559 | while(cur_byte < new_size) { | |
2560 | key.objectid = cur_byte; | |
2561 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
2562 | sizeof(struct btrfs_block_group_item)); | |
2563 | BUG_ON(ret); | |
2564 | leaf = path->nodes[0]; | |
2565 | item = btrfs_item_ptr(leaf, path->slots[0], | |
2566 | struct btrfs_block_group_item); | |
2567 | ||
2568 | btrfs_set_disk_block_group_used(leaf, item, 0); | |
f9ef6604 CM |
2569 | div_long_long_rem(nr, 3, &rem); |
2570 | if (rem) { | |
edbd8d4e CM |
2571 | btrfs_set_disk_block_group_flags(leaf, item, |
2572 | BTRFS_BLOCK_GROUP_DATA); | |
2573 | } else { | |
2574 | btrfs_set_disk_block_group_flags(leaf, item, 0); | |
2575 | } | |
2576 | nr++; | |
2577 | ||
2578 | cache = kmalloc(sizeof(*cache), GFP_NOFS); | |
2579 | BUG_ON(!cache); | |
2580 | ||
2581 | read_extent_buffer(leaf, &cache->item, (unsigned long)item, | |
2582 | sizeof(cache->item)); | |
2583 | ||
2584 | memcpy(&cache->key, &key, sizeof(key)); | |
2585 | cache->cached = 0; | |
2586 | cache->pinned = 0; | |
2587 | cur_byte = key.objectid + key.offset; | |
2588 | btrfs_release_path(root, path); | |
2589 | ||
2590 | if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) { | |
2591 | bit = BLOCK_GROUP_DATA; | |
2592 | cache->data = BTRFS_BLOCK_GROUP_DATA; | |
2593 | } else { | |
2594 | bit = BLOCK_GROUP_METADATA; | |
2595 | cache->data = 0; | |
2596 | } | |
2597 | ||
2598 | /* use EXTENT_LOCKED to prevent merging */ | |
2599 | set_extent_bits(block_group_cache, key.objectid, | |
2600 | key.objectid + key.offset - 1, | |
2601 | bit | EXTENT_LOCKED, GFP_NOFS); | |
2602 | set_state_private(block_group_cache, key.objectid, | |
2603 | (unsigned long)cache); | |
2604 | } | |
2605 | btrfs_free_path(path); | |
2606 | set_size: | |
2607 | btrfs_set_super_total_bytes(&info->super_copy, new_size); | |
2608 | return 0; | |
2609 | } | |
2610 | ||
9078a3e1 CM |
2611 | int btrfs_read_block_groups(struct btrfs_root *root) |
2612 | { | |
2613 | struct btrfs_path *path; | |
2614 | int ret; | |
2615 | int err = 0; | |
96b5179d | 2616 | int bit; |
9078a3e1 | 2617 | struct btrfs_block_group_cache *cache; |
be744175 | 2618 | struct btrfs_fs_info *info = root->fs_info; |
96b5179d | 2619 | struct extent_map_tree *block_group_cache; |
9078a3e1 CM |
2620 | struct btrfs_key key; |
2621 | struct btrfs_key found_key; | |
5f39d397 | 2622 | struct extent_buffer *leaf; |
96b5179d CM |
2623 | |
2624 | block_group_cache = &info->block_group_cache; | |
9078a3e1 | 2625 | |
be744175 | 2626 | root = info->extent_root; |
9078a3e1 | 2627 | key.objectid = 0; |
db94535d | 2628 | key.offset = BTRFS_BLOCK_GROUP_SIZE; |
9078a3e1 CM |
2629 | btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); |
2630 | ||
2631 | path = btrfs_alloc_path(); | |
2632 | if (!path) | |
2633 | return -ENOMEM; | |
2634 | ||
2635 | while(1) { | |
be744175 | 2636 | ret = btrfs_search_slot(NULL, info->extent_root, |
9078a3e1 CM |
2637 | &key, path, 0, 0); |
2638 | if (ret != 0) { | |
2639 | err = ret; | |
2640 | break; | |
2641 | } | |
5f39d397 CM |
2642 | leaf = path->nodes[0]; |
2643 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
9078a3e1 CM |
2644 | cache = kmalloc(sizeof(*cache), GFP_NOFS); |
2645 | if (!cache) { | |
2646 | err = -1; | |
2647 | break; | |
2648 | } | |
3e1ad54f | 2649 | |
5f39d397 CM |
2650 | read_extent_buffer(leaf, &cache->item, |
2651 | btrfs_item_ptr_offset(leaf, path->slots[0]), | |
2652 | sizeof(cache->item)); | |
9078a3e1 | 2653 | memcpy(&cache->key, &found_key, sizeof(found_key)); |
e37c9e69 | 2654 | cache->cached = 0; |
324ae4df | 2655 | cache->pinned = 0; |
9078a3e1 CM |
2656 | key.objectid = found_key.objectid + found_key.offset; |
2657 | btrfs_release_path(root, path); | |
5f39d397 | 2658 | |
f84a8b36 CM |
2659 | if (cache->item.flags & BTRFS_BLOCK_GROUP_MIXED) { |
2660 | bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA; | |
2661 | cache->data = BTRFS_BLOCK_GROUP_MIXED; | |
2662 | } else if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) { | |
96b5179d | 2663 | bit = BLOCK_GROUP_DATA; |
f84a8b36 | 2664 | cache->data = BTRFS_BLOCK_GROUP_DATA; |
96b5179d CM |
2665 | } else { |
2666 | bit = BLOCK_GROUP_METADATA; | |
2667 | cache->data = 0; | |
31f3c99b | 2668 | } |
96b5179d CM |
2669 | |
2670 | /* use EXTENT_LOCKED to prevent merging */ | |
2671 | set_extent_bits(block_group_cache, found_key.objectid, | |
2672 | found_key.objectid + found_key.offset - 1, | |
2673 | bit | EXTENT_LOCKED, GFP_NOFS); | |
2674 | set_state_private(block_group_cache, found_key.objectid, | |
ae2f5411 | 2675 | (unsigned long)cache); |
96b5179d | 2676 | |
9078a3e1 | 2677 | if (key.objectid >= |
db94535d | 2678 | btrfs_super_total_bytes(&info->super_copy)) |
9078a3e1 CM |
2679 | break; |
2680 | } | |
2681 | ||
2682 | btrfs_free_path(path); | |
2683 | return 0; | |
2684 | } |