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" |
0b86a832 | 27 | #include "volumes.h" |
fec577fb | 28 | |
0b86a832 | 29 | #define BLOCK_GROUP_DATA EXTENT_WRITEBACK |
96b5179d | 30 | #define BLOCK_GROUP_METADATA EXTENT_UPTODATE |
0b86a832 CM |
31 | #define BLOCK_GROUP_SYSTEM EXTENT_NEW |
32 | ||
96b5179d CM |
33 | #define BLOCK_GROUP_DIRTY EXTENT_DIRTY |
34 | ||
e089f05c CM |
35 | static int finish_current_insert(struct btrfs_trans_handle *trans, struct |
36 | btrfs_root *extent_root); | |
e20d96d6 CM |
37 | static int del_pending_extents(struct btrfs_trans_handle *trans, struct |
38 | btrfs_root *extent_root); | |
6324fbf3 CM |
39 | int btrfs_make_block_group(struct btrfs_trans_handle *trans, |
40 | struct btrfs_root *root, u64 bytes_used, | |
41 | u64 type, u64 chunk_tree, u64 chunk_objectid, | |
42 | u64 size); | |
d548ee51 | 43 | |
fec577fb | 44 | |
e37c9e69 CM |
45 | static int cache_block_group(struct btrfs_root *root, |
46 | struct btrfs_block_group_cache *block_group) | |
47 | { | |
48 | struct btrfs_path *path; | |
49 | int ret; | |
50 | struct btrfs_key key; | |
5f39d397 | 51 | struct extent_buffer *leaf; |
d1310b2e | 52 | struct extent_io_tree *free_space_cache; |
e37c9e69 | 53 | int slot; |
e37c9e69 CM |
54 | u64 last = 0; |
55 | u64 hole_size; | |
7d7d6068 | 56 | u64 first_free; |
e37c9e69 CM |
57 | int found = 0; |
58 | ||
00f5c795 CM |
59 | if (!block_group) |
60 | return 0; | |
61 | ||
e37c9e69 | 62 | root = root->fs_info->extent_root; |
f510cfec | 63 | free_space_cache = &root->fs_info->free_space_cache; |
e37c9e69 CM |
64 | |
65 | if (block_group->cached) | |
66 | return 0; | |
f510cfec | 67 | |
e37c9e69 CM |
68 | path = btrfs_alloc_path(); |
69 | if (!path) | |
70 | return -ENOMEM; | |
7d7d6068 | 71 | |
2cc58cf2 | 72 | path->reada = 2; |
7d7d6068 | 73 | first_free = block_group->key.objectid; |
e37c9e69 | 74 | key.objectid = block_group->key.objectid; |
e37c9e69 CM |
75 | key.offset = 0; |
76 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); | |
77 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
78 | if (ret < 0) | |
79 | return ret; | |
0b86a832 | 80 | ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY); |
d548ee51 Y |
81 | if (ret < 0) |
82 | return ret; | |
83 | if (ret == 0) { | |
84 | leaf = path->nodes[0]; | |
85 | btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); | |
86 | if (key.objectid + key.offset > first_free) | |
87 | first_free = key.objectid + key.offset; | |
88 | } | |
e37c9e69 | 89 | while(1) { |
5f39d397 | 90 | leaf = path->nodes[0]; |
e37c9e69 | 91 | slot = path->slots[0]; |
5f39d397 | 92 | if (slot >= btrfs_header_nritems(leaf)) { |
e37c9e69 | 93 | ret = btrfs_next_leaf(root, path); |
54aa1f4d CM |
94 | if (ret < 0) |
95 | goto err; | |
de428b63 | 96 | if (ret == 0) { |
e37c9e69 | 97 | continue; |
de428b63 | 98 | } else { |
e37c9e69 CM |
99 | break; |
100 | } | |
101 | } | |
5f39d397 | 102 | btrfs_item_key_to_cpu(leaf, &key, slot); |
7d7d6068 | 103 | if (key.objectid < block_group->key.objectid) { |
7d7d6068 Y |
104 | goto next; |
105 | } | |
e37c9e69 CM |
106 | if (key.objectid >= block_group->key.objectid + |
107 | block_group->key.offset) { | |
e37c9e69 CM |
108 | break; |
109 | } | |
7d7d6068 | 110 | |
e37c9e69 CM |
111 | if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) { |
112 | if (!found) { | |
7d7d6068 | 113 | last = first_free; |
e37c9e69 | 114 | found = 1; |
e37c9e69 | 115 | } |
f510cfec CM |
116 | if (key.objectid > last) { |
117 | hole_size = key.objectid - last; | |
118 | set_extent_dirty(free_space_cache, last, | |
119 | last + hole_size - 1, | |
120 | GFP_NOFS); | |
7d7d6068 Y |
121 | } |
122 | last = key.objectid + key.offset; | |
e37c9e69 | 123 | } |
7d7d6068 | 124 | next: |
e37c9e69 CM |
125 | path->slots[0]++; |
126 | } | |
127 | ||
7d7d6068 Y |
128 | if (!found) |
129 | last = first_free; | |
130 | if (block_group->key.objectid + | |
131 | block_group->key.offset > last) { | |
132 | hole_size = block_group->key.objectid + | |
133 | block_group->key.offset - last; | |
f510cfec CM |
134 | set_extent_dirty(free_space_cache, last, |
135 | last + hole_size - 1, GFP_NOFS); | |
7d7d6068 | 136 | } |
e37c9e69 | 137 | block_group->cached = 1; |
54aa1f4d | 138 | err: |
e37c9e69 CM |
139 | btrfs_free_path(path); |
140 | return 0; | |
141 | } | |
142 | ||
5276aeda CM |
143 | struct btrfs_block_group_cache *btrfs_lookup_block_group(struct |
144 | btrfs_fs_info *info, | |
db94535d | 145 | u64 bytenr) |
be744175 | 146 | { |
d1310b2e | 147 | struct extent_io_tree *block_group_cache; |
96b5179d CM |
148 | struct btrfs_block_group_cache *block_group = NULL; |
149 | u64 ptr; | |
150 | u64 start; | |
151 | u64 end; | |
be744175 CM |
152 | int ret; |
153 | ||
96b5179d CM |
154 | block_group_cache = &info->block_group_cache; |
155 | ret = find_first_extent_bit(block_group_cache, | |
db94535d | 156 | bytenr, &start, &end, |
0b86a832 CM |
157 | BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA | |
158 | BLOCK_GROUP_SYSTEM); | |
be744175 | 159 | if (ret) { |
96b5179d | 160 | return NULL; |
be744175 | 161 | } |
96b5179d CM |
162 | ret = get_state_private(block_group_cache, start, &ptr); |
163 | if (ret) | |
164 | return NULL; | |
165 | ||
ae2f5411 | 166 | block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr; |
5cf66426 | 167 | if (block_group->key.objectid <= bytenr && bytenr < |
96b5179d CM |
168 | block_group->key.objectid + block_group->key.offset) |
169 | return block_group; | |
be744175 CM |
170 | return NULL; |
171 | } | |
0b86a832 CM |
172 | |
173 | static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) | |
174 | { | |
6324fbf3 | 175 | return (cache->flags & bits); |
0b86a832 CM |
176 | } |
177 | ||
178 | static int noinline find_search_start(struct btrfs_root *root, | |
98ed5174 | 179 | struct btrfs_block_group_cache **cache_ret, |
0b86a832 | 180 | u64 *start_ret, int num, int data) |
e37c9e69 | 181 | { |
e37c9e69 CM |
182 | int ret; |
183 | struct btrfs_block_group_cache *cache = *cache_ret; | |
d7fc640e | 184 | struct extent_io_tree *free_space_cache; |
7d1660d4 | 185 | struct extent_state *state; |
e19caa5f | 186 | u64 last; |
f510cfec | 187 | u64 start = 0; |
257d0ce3 | 188 | u64 cache_miss = 0; |
c31f8830 | 189 | u64 total_fs_bytes; |
0b86a832 | 190 | u64 search_start = *start_ret; |
f84a8b36 | 191 | int wrapped = 0; |
e37c9e69 | 192 | |
0b86a832 | 193 | if (!cache) |
1a2b2ac7 | 194 | goto out; |
c31f8830 | 195 | total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); |
d7fc640e CM |
196 | free_space_cache = &root->fs_info->free_space_cache; |
197 | ||
e37c9e69 | 198 | again: |
54aa1f4d CM |
199 | ret = cache_block_group(root, cache); |
200 | if (ret) | |
201 | goto out; | |
f84a8b36 | 202 | |
e19caa5f | 203 | last = max(search_start, cache->key.objectid); |
0b86a832 CM |
204 | if (!block_group_bits(cache, data)) { |
205 | goto new_group; | |
206 | } | |
e19caa5f | 207 | |
7d1660d4 CM |
208 | spin_lock_irq(&free_space_cache->lock); |
209 | state = find_first_extent_bit_state(free_space_cache, last, EXTENT_DIRTY); | |
e37c9e69 | 210 | while(1) { |
7d1660d4 | 211 | if (!state) { |
257d0ce3 CM |
212 | if (!cache_miss) |
213 | cache_miss = last; | |
7d1660d4 | 214 | spin_unlock_irq(&free_space_cache->lock); |
e19caa5f CM |
215 | goto new_group; |
216 | } | |
f510cfec | 217 | |
7d1660d4 CM |
218 | start = max(last, state->start); |
219 | last = state->end + 1; | |
257d0ce3 CM |
220 | if (last - start < num) { |
221 | if (last == cache->key.objectid + cache->key.offset) | |
222 | cache_miss = start; | |
7d1660d4 CM |
223 | do { |
224 | state = extent_state_next(state); | |
225 | } while(state && !(state->state & EXTENT_DIRTY)); | |
f510cfec | 226 | continue; |
257d0ce3 | 227 | } |
7d1660d4 | 228 | spin_unlock_irq(&free_space_cache->lock); |
0b86a832 | 229 | if (start + num > cache->key.objectid + cache->key.offset) |
e37c9e69 | 230 | goto new_group; |
c31f8830 CM |
231 | if (start + num > total_fs_bytes) |
232 | goto new_group; | |
0b86a832 CM |
233 | *start_ret = start; |
234 | return 0; | |
7d1660d4 | 235 | } out: |
1a2b2ac7 CM |
236 | cache = btrfs_lookup_block_group(root->fs_info, search_start); |
237 | if (!cache) { | |
0b86a832 | 238 | printk("Unable to find block group for %Lu\n", search_start); |
1a2b2ac7 | 239 | WARN_ON(1); |
1a2b2ac7 | 240 | } |
0b86a832 | 241 | return -ENOSPC; |
e37c9e69 CM |
242 | |
243 | new_group: | |
e19caa5f | 244 | last = cache->key.objectid + cache->key.offset; |
f84a8b36 | 245 | wrapped: |
e19caa5f | 246 | cache = btrfs_lookup_block_group(root->fs_info, last); |
c31f8830 | 247 | if (!cache || cache->key.objectid >= total_fs_bytes) { |
0e4de584 | 248 | no_cache: |
f84a8b36 CM |
249 | if (!wrapped) { |
250 | wrapped = 1; | |
251 | last = search_start; | |
f84a8b36 CM |
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 | ||
6324fbf3 CM |
278 | static int block_group_state_bits(u64 flags) |
279 | { | |
280 | int bits = 0; | |
281 | if (flags & BTRFS_BLOCK_GROUP_DATA) | |
282 | bits |= BLOCK_GROUP_DATA; | |
283 | if (flags & BTRFS_BLOCK_GROUP_METADATA) | |
284 | bits |= BLOCK_GROUP_METADATA; | |
285 | if (flags & BTRFS_BLOCK_GROUP_SYSTEM) | |
286 | bits |= BLOCK_GROUP_SYSTEM; | |
287 | return bits; | |
288 | } | |
289 | ||
31f3c99b CM |
290 | struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, |
291 | struct btrfs_block_group_cache | |
be744175 | 292 | *hint, u64 search_start, |
de428b63 | 293 | int data, int owner) |
cd1bc465 | 294 | { |
96b5179d | 295 | struct btrfs_block_group_cache *cache; |
d1310b2e | 296 | struct extent_io_tree *block_group_cache; |
31f3c99b | 297 | struct btrfs_block_group_cache *found_group = NULL; |
cd1bc465 CM |
298 | struct btrfs_fs_info *info = root->fs_info; |
299 | u64 used; | |
31f3c99b CM |
300 | u64 last = 0; |
301 | u64 hint_last; | |
96b5179d CM |
302 | u64 start; |
303 | u64 end; | |
304 | u64 free_check; | |
305 | u64 ptr; | |
c31f8830 | 306 | u64 total_fs_bytes; |
96b5179d | 307 | int bit; |
cd1bc465 | 308 | int ret; |
31f3c99b | 309 | int full_search = 0; |
de428b63 CM |
310 | int factor = 8; |
311 | ||
96b5179d | 312 | block_group_cache = &info->block_group_cache; |
c31f8830 | 313 | total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); |
96b5179d | 314 | |
de428b63 | 315 | if (!owner) |
f84a8b36 | 316 | factor = 8; |
be744175 | 317 | |
6324fbf3 | 318 | bit = block_group_state_bits(data); |
be744175 | 319 | |
c31f8830 | 320 | if (search_start && search_start < total_fs_bytes) { |
be744175 | 321 | struct btrfs_block_group_cache *shint; |
5276aeda | 322 | shint = btrfs_lookup_block_group(info, search_start); |
0b86a832 | 323 | if (shint && block_group_bits(shint, data)) { |
be744175 | 324 | used = btrfs_block_group_used(&shint->item); |
324ae4df Y |
325 | if (used + shint->pinned < |
326 | div_factor(shint->key.offset, factor)) { | |
be744175 CM |
327 | return shint; |
328 | } | |
329 | } | |
330 | } | |
0b86a832 CM |
331 | if (hint && block_group_bits(hint, data) && |
332 | hint->key.objectid < total_fs_bytes) { | |
31f3c99b | 333 | used = btrfs_block_group_used(&hint->item); |
324ae4df Y |
334 | if (used + hint->pinned < |
335 | div_factor(hint->key.offset, factor)) { | |
31f3c99b CM |
336 | return hint; |
337 | } | |
e19caa5f | 338 | last = hint->key.objectid + hint->key.offset; |
31f3c99b CM |
339 | hint_last = last; |
340 | } else { | |
e37c9e69 CM |
341 | if (hint) |
342 | hint_last = max(hint->key.objectid, search_start); | |
343 | else | |
344 | hint_last = search_start; | |
345 | ||
c31f8830 CM |
346 | if (hint_last >= total_fs_bytes) |
347 | hint_last = search_start; | |
e37c9e69 | 348 | last = hint_last; |
31f3c99b | 349 | } |
31f3c99b | 350 | again: |
cd1bc465 | 351 | while(1) { |
96b5179d CM |
352 | ret = find_first_extent_bit(block_group_cache, last, |
353 | &start, &end, bit); | |
354 | if (ret) | |
cd1bc465 | 355 | break; |
96b5179d CM |
356 | |
357 | ret = get_state_private(block_group_cache, start, &ptr); | |
358 | if (ret) | |
359 | break; | |
360 | ||
ae2f5411 | 361 | cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; |
96b5179d CM |
362 | last = cache->key.objectid + cache->key.offset; |
363 | used = btrfs_block_group_used(&cache->item); | |
364 | ||
c31f8830 CM |
365 | if (cache->key.objectid > total_fs_bytes) |
366 | break; | |
367 | ||
96b5179d CM |
368 | if (full_search) |
369 | free_check = cache->key.offset; | |
370 | else | |
371 | free_check = div_factor(cache->key.offset, factor); | |
6324fbf3 | 372 | |
324ae4df | 373 | if (used + cache->pinned < free_check) { |
96b5179d CM |
374 | found_group = cache; |
375 | goto found; | |
cd1bc465 | 376 | } |
6324fbf3 CM |
377 | if (full_search) { |
378 | printk("failed on cache %Lu used %Lu total %Lu\n", | |
379 | cache->key.objectid, used, cache->key.offset); | |
380 | } | |
de428b63 | 381 | cond_resched(); |
cd1bc465 | 382 | } |
31f3c99b | 383 | if (!full_search) { |
be744175 | 384 | last = search_start; |
31f3c99b CM |
385 | full_search = 1; |
386 | goto again; | |
387 | } | |
be744175 | 388 | found: |
31f3c99b | 389 | return found_group; |
cd1bc465 CM |
390 | } |
391 | ||
7bb86316 | 392 | static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation, |
74493f7a CM |
393 | u64 owner, u64 owner_offset) |
394 | { | |
395 | u32 high_crc = ~(u32)0; | |
396 | u32 low_crc = ~(u32)0; | |
397 | __le64 lenum; | |
398 | ||
399 | lenum = cpu_to_le64(root_objectid); | |
400 | high_crc = crc32c(high_crc, &lenum, sizeof(lenum)); | |
7bb86316 CM |
401 | lenum = cpu_to_le64(ref_generation); |
402 | low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); | |
21a4989d CM |
403 | if (owner >= BTRFS_FIRST_FREE_OBJECTID) { |
404 | lenum = cpu_to_le64(owner); | |
405 | low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); | |
406 | lenum = cpu_to_le64(owner_offset); | |
407 | low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); | |
408 | } | |
74493f7a CM |
409 | return ((u64)high_crc << 32) | (u64)low_crc; |
410 | } | |
411 | ||
7bb86316 CM |
412 | static int match_extent_ref(struct extent_buffer *leaf, |
413 | struct btrfs_extent_ref *disk_ref, | |
414 | struct btrfs_extent_ref *cpu_ref) | |
415 | { | |
416 | int ret; | |
417 | int len; | |
418 | ||
419 | if (cpu_ref->objectid) | |
420 | len = sizeof(*cpu_ref); | |
421 | else | |
422 | len = 2 * sizeof(u64); | |
423 | ret = memcmp_extent_buffer(leaf, cpu_ref, (unsigned long)disk_ref, | |
424 | len); | |
425 | return ret == 0; | |
426 | } | |
427 | ||
98ed5174 CM |
428 | static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, |
429 | struct btrfs_root *root, | |
430 | struct btrfs_path *path, u64 bytenr, | |
431 | u64 root_objectid, | |
432 | u64 ref_generation, u64 owner, | |
433 | u64 owner_offset, int del) | |
74493f7a CM |
434 | { |
435 | u64 hash; | |
436 | struct btrfs_key key; | |
7bb86316 | 437 | struct btrfs_key found_key; |
74493f7a | 438 | struct btrfs_extent_ref ref; |
7bb86316 CM |
439 | struct extent_buffer *leaf; |
440 | struct btrfs_extent_ref *disk_ref; | |
441 | int ret; | |
442 | int ret2; | |
443 | ||
444 | btrfs_set_stack_ref_root(&ref, root_objectid); | |
445 | btrfs_set_stack_ref_generation(&ref, ref_generation); | |
446 | btrfs_set_stack_ref_objectid(&ref, owner); | |
447 | btrfs_set_stack_ref_offset(&ref, owner_offset); | |
448 | ||
449 | hash = hash_extent_ref(root_objectid, ref_generation, owner, | |
450 | owner_offset); | |
451 | key.offset = hash; | |
452 | key.objectid = bytenr; | |
453 | key.type = BTRFS_EXTENT_REF_KEY; | |
454 | ||
455 | while (1) { | |
456 | ret = btrfs_search_slot(trans, root, &key, path, | |
457 | del ? -1 : 0, del); | |
458 | if (ret < 0) | |
459 | goto out; | |
460 | leaf = path->nodes[0]; | |
461 | if (ret != 0) { | |
462 | u32 nritems = btrfs_header_nritems(leaf); | |
463 | if (path->slots[0] >= nritems) { | |
464 | ret2 = btrfs_next_leaf(root, path); | |
465 | if (ret2) | |
466 | goto out; | |
467 | leaf = path->nodes[0]; | |
468 | } | |
469 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
470 | if (found_key.objectid != bytenr || | |
471 | found_key.type != BTRFS_EXTENT_REF_KEY) | |
472 | goto out; | |
473 | key.offset = found_key.offset; | |
474 | if (del) { | |
475 | btrfs_release_path(root, path); | |
476 | continue; | |
477 | } | |
478 | } | |
479 | disk_ref = btrfs_item_ptr(path->nodes[0], | |
480 | path->slots[0], | |
481 | struct btrfs_extent_ref); | |
482 | if (match_extent_ref(path->nodes[0], disk_ref, &ref)) { | |
483 | ret = 0; | |
484 | goto out; | |
485 | } | |
486 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
487 | key.offset = found_key.offset + 1; | |
488 | btrfs_release_path(root, path); | |
489 | } | |
490 | out: | |
491 | return ret; | |
492 | } | |
493 | ||
d8d5f3e1 CM |
494 | /* |
495 | * Back reference rules. Back refs have three main goals: | |
496 | * | |
497 | * 1) differentiate between all holders of references to an extent so that | |
498 | * when a reference is dropped we can make sure it was a valid reference | |
499 | * before freeing the extent. | |
500 | * | |
501 | * 2) Provide enough information to quickly find the holders of an extent | |
502 | * if we notice a given block is corrupted or bad. | |
503 | * | |
504 | * 3) Make it easy to migrate blocks for FS shrinking or storage pool | |
505 | * maintenance. This is actually the same as #2, but with a slightly | |
506 | * different use case. | |
507 | * | |
508 | * File extents can be referenced by: | |
509 | * | |
510 | * - multiple snapshots, subvolumes, or different generations in one subvol | |
511 | * - different files inside a single subvolume (in theory, not implemented yet) | |
512 | * - different offsets inside a file (bookend extents in file.c) | |
513 | * | |
514 | * The extent ref structure has fields for: | |
515 | * | |
516 | * - Objectid of the subvolume root | |
517 | * - Generation number of the tree holding the reference | |
518 | * - objectid of the file holding the reference | |
519 | * - offset in the file corresponding to the key holding the reference | |
520 | * | |
521 | * When a file extent is allocated the fields are filled in: | |
522 | * (root_key.objectid, trans->transid, inode objectid, offset in file) | |
523 | * | |
524 | * When a leaf is cow'd new references are added for every file extent found | |
525 | * in the leaf. It looks the same as the create case, but trans->transid | |
526 | * will be different when the block is cow'd. | |
527 | * | |
528 | * (root_key.objectid, trans->transid, inode objectid, offset in file) | |
529 | * | |
530 | * When a file extent is removed either during snapshot deletion or file | |
531 | * truncation, the corresponding back reference is found | |
532 | * by searching for: | |
533 | * | |
534 | * (btrfs_header_owner(leaf), btrfs_header_generation(leaf), | |
535 | * inode objectid, offset in file) | |
536 | * | |
537 | * Btree extents can be referenced by: | |
538 | * | |
539 | * - Different subvolumes | |
540 | * - Different generations of the same subvolume | |
541 | * | |
542 | * Storing sufficient information for a full reverse mapping of a btree | |
543 | * block would require storing the lowest key of the block in the backref, | |
544 | * and it would require updating that lowest key either before write out or | |
545 | * every time it changed. Instead, the objectid of the lowest key is stored | |
546 | * along with the level of the tree block. This provides a hint | |
547 | * about where in the btree the block can be found. Searches through the | |
548 | * btree only need to look for a pointer to that block, so they stop one | |
549 | * level higher than the level recorded in the backref. | |
550 | * | |
551 | * Some btrees do not do reference counting on their extents. These | |
552 | * include the extent tree and the tree of tree roots. Backrefs for these | |
553 | * trees always have a generation of zero. | |
554 | * | |
555 | * When a tree block is created, back references are inserted: | |
556 | * | |
f6dbff55 | 557 | * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid) |
d8d5f3e1 CM |
558 | * |
559 | * When a tree block is cow'd in a reference counted root, | |
560 | * new back references are added for all the blocks it points to. | |
561 | * These are of the form (trans->transid will have increased since creation): | |
562 | * | |
f6dbff55 | 563 | * (root->root_key.objectid, trans->transid, level, lowest_key_objectid) |
d8d5f3e1 CM |
564 | * |
565 | * Because the lowest_key_objectid and the level are just hints | |
566 | * they are not used when backrefs are deleted. When a backref is deleted: | |
567 | * | |
568 | * if backref was for a tree root: | |
569 | * root_objectid = root->root_key.objectid | |
570 | * else | |
571 | * root_objectid = btrfs_header_owner(parent) | |
572 | * | |
573 | * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0) | |
574 | * | |
575 | * Back Reference Key hashing: | |
576 | * | |
577 | * Back references have four fields, each 64 bits long. Unfortunately, | |
578 | * This is hashed into a single 64 bit number and placed into the key offset. | |
579 | * The key objectid corresponds to the first byte in the extent, and the | |
580 | * key type is set to BTRFS_EXTENT_REF_KEY | |
581 | */ | |
7bb86316 CM |
582 | int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, |
583 | struct btrfs_root *root, | |
584 | struct btrfs_path *path, u64 bytenr, | |
585 | u64 root_objectid, u64 ref_generation, | |
586 | u64 owner, u64 owner_offset) | |
587 | { | |
588 | u64 hash; | |
589 | struct btrfs_key key; | |
590 | struct btrfs_extent_ref ref; | |
591 | struct btrfs_extent_ref *disk_ref; | |
74493f7a CM |
592 | int ret; |
593 | ||
594 | btrfs_set_stack_ref_root(&ref, root_objectid); | |
7bb86316 | 595 | btrfs_set_stack_ref_generation(&ref, ref_generation); |
74493f7a CM |
596 | btrfs_set_stack_ref_objectid(&ref, owner); |
597 | btrfs_set_stack_ref_offset(&ref, owner_offset); | |
598 | ||
7bb86316 CM |
599 | hash = hash_extent_ref(root_objectid, ref_generation, owner, |
600 | owner_offset); | |
74493f7a CM |
601 | key.offset = hash; |
602 | key.objectid = bytenr; | |
603 | key.type = BTRFS_EXTENT_REF_KEY; | |
604 | ||
605 | ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(ref)); | |
606 | while (ret == -EEXIST) { | |
7bb86316 CM |
607 | disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], |
608 | struct btrfs_extent_ref); | |
609 | if (match_extent_ref(path->nodes[0], disk_ref, &ref)) | |
610 | goto out; | |
611 | key.offset++; | |
612 | btrfs_release_path(root, path); | |
613 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
614 | sizeof(ref)); | |
74493f7a | 615 | } |
7bb86316 CM |
616 | if (ret) |
617 | goto out; | |
618 | disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
619 | struct btrfs_extent_ref); | |
620 | write_extent_buffer(path->nodes[0], &ref, (unsigned long)disk_ref, | |
621 | sizeof(ref)); | |
622 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
623 | out: | |
624 | btrfs_release_path(root, path); | |
625 | return ret; | |
74493f7a CM |
626 | } |
627 | ||
b18c6685 CM |
628 | int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, |
629 | struct btrfs_root *root, | |
74493f7a | 630 | u64 bytenr, u64 num_bytes, |
7bb86316 | 631 | u64 root_objectid, u64 ref_generation, |
74493f7a | 632 | u64 owner, u64 owner_offset) |
02217ed2 | 633 | { |
5caf2a00 | 634 | struct btrfs_path *path; |
02217ed2 | 635 | int ret; |
e2fa7227 | 636 | struct btrfs_key key; |
5f39d397 | 637 | struct extent_buffer *l; |
234b63a0 | 638 | struct btrfs_extent_item *item; |
cf27e1ee | 639 | u32 refs; |
037e6390 | 640 | |
db94535d | 641 | WARN_ON(num_bytes < root->sectorsize); |
5caf2a00 | 642 | path = btrfs_alloc_path(); |
54aa1f4d CM |
643 | if (!path) |
644 | return -ENOMEM; | |
26b8003f | 645 | |
b0331a4c | 646 | path->reada = 0; |
db94535d | 647 | key.objectid = bytenr; |
62e2749e | 648 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
db94535d | 649 | key.offset = num_bytes; |
5caf2a00 | 650 | ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, |
9f5fae2f | 651 | 0, 1); |
54aa1f4d CM |
652 | if (ret < 0) |
653 | return ret; | |
a429e513 | 654 | if (ret != 0) { |
a28ec197 | 655 | BUG(); |
a429e513 | 656 | } |
02217ed2 | 657 | BUG_ON(ret != 0); |
5f39d397 | 658 | l = path->nodes[0]; |
5caf2a00 | 659 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); |
5f39d397 CM |
660 | refs = btrfs_extent_refs(l, item); |
661 | btrfs_set_extent_refs(l, item, refs + 1); | |
5caf2a00 | 662 | btrfs_mark_buffer_dirty(path->nodes[0]); |
a28ec197 | 663 | |
5caf2a00 | 664 | btrfs_release_path(root->fs_info->extent_root, path); |
7bb86316 | 665 | |
b0331a4c | 666 | path->reada = 0; |
7bb86316 CM |
667 | ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root, |
668 | path, bytenr, root_objectid, | |
669 | ref_generation, owner, owner_offset); | |
670 | BUG_ON(ret); | |
9f5fae2f | 671 | finish_current_insert(trans, root->fs_info->extent_root); |
e20d96d6 | 672 | del_pending_extents(trans, root->fs_info->extent_root); |
74493f7a CM |
673 | |
674 | btrfs_free_path(path); | |
02217ed2 CM |
675 | return 0; |
676 | } | |
677 | ||
e9d0b13b CM |
678 | int btrfs_extent_post_op(struct btrfs_trans_handle *trans, |
679 | struct btrfs_root *root) | |
680 | { | |
681 | finish_current_insert(trans, root->fs_info->extent_root); | |
682 | del_pending_extents(trans, root->fs_info->extent_root); | |
683 | return 0; | |
684 | } | |
685 | ||
b18c6685 | 686 | static int lookup_extent_ref(struct btrfs_trans_handle *trans, |
db94535d CM |
687 | struct btrfs_root *root, u64 bytenr, |
688 | u64 num_bytes, u32 *refs) | |
a28ec197 | 689 | { |
5caf2a00 | 690 | struct btrfs_path *path; |
a28ec197 | 691 | int ret; |
e2fa7227 | 692 | struct btrfs_key key; |
5f39d397 | 693 | struct extent_buffer *l; |
234b63a0 | 694 | struct btrfs_extent_item *item; |
5caf2a00 | 695 | |
db94535d | 696 | WARN_ON(num_bytes < root->sectorsize); |
5caf2a00 | 697 | path = btrfs_alloc_path(); |
b0331a4c | 698 | path->reada = 0; |
db94535d CM |
699 | key.objectid = bytenr; |
700 | key.offset = num_bytes; | |
62e2749e | 701 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
5caf2a00 | 702 | ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, |
9f5fae2f | 703 | 0, 0); |
54aa1f4d CM |
704 | if (ret < 0) |
705 | goto out; | |
5f39d397 CM |
706 | if (ret != 0) { |
707 | btrfs_print_leaf(root, path->nodes[0]); | |
db94535d | 708 | printk("failed to find block number %Lu\n", bytenr); |
a28ec197 | 709 | BUG(); |
5f39d397 CM |
710 | } |
711 | l = path->nodes[0]; | |
5caf2a00 | 712 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); |
5f39d397 | 713 | *refs = btrfs_extent_refs(l, item); |
54aa1f4d | 714 | out: |
5caf2a00 | 715 | btrfs_free_path(path); |
a28ec197 CM |
716 | return 0; |
717 | } | |
718 | ||
be20aa9d CM |
719 | u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, |
720 | struct btrfs_path *count_path, | |
721 | u64 first_extent) | |
722 | { | |
723 | struct btrfs_root *extent_root = root->fs_info->extent_root; | |
724 | struct btrfs_path *path; | |
725 | u64 bytenr; | |
726 | u64 found_objectid; | |
56b453c9 | 727 | u64 root_objectid = root->root_key.objectid; |
be20aa9d CM |
728 | u32 total_count = 0; |
729 | u32 cur_count; | |
be20aa9d CM |
730 | u32 nritems; |
731 | int ret; | |
732 | struct btrfs_key key; | |
733 | struct btrfs_key found_key; | |
734 | struct extent_buffer *l; | |
735 | struct btrfs_extent_item *item; | |
736 | struct btrfs_extent_ref *ref_item; | |
737 | int level = -1; | |
738 | ||
739 | path = btrfs_alloc_path(); | |
740 | again: | |
741 | if (level == -1) | |
742 | bytenr = first_extent; | |
743 | else | |
744 | bytenr = count_path->nodes[level]->start; | |
745 | ||
746 | cur_count = 0; | |
747 | key.objectid = bytenr; | |
748 | key.offset = 0; | |
749 | ||
750 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); | |
751 | ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); | |
752 | if (ret < 0) | |
753 | goto out; | |
754 | BUG_ON(ret == 0); | |
755 | ||
756 | l = path->nodes[0]; | |
757 | btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); | |
758 | ||
759 | if (found_key.objectid != bytenr || | |
760 | found_key.type != BTRFS_EXTENT_ITEM_KEY) { | |
761 | goto out; | |
762 | } | |
763 | ||
764 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); | |
be20aa9d | 765 | while (1) { |
bd09835d | 766 | l = path->nodes[0]; |
be20aa9d CM |
767 | nritems = btrfs_header_nritems(l); |
768 | if (path->slots[0] >= nritems) { | |
769 | ret = btrfs_next_leaf(extent_root, path); | |
770 | if (ret == 0) | |
771 | continue; | |
772 | break; | |
773 | } | |
774 | btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); | |
775 | if (found_key.objectid != bytenr) | |
776 | break; | |
bd09835d | 777 | |
be20aa9d CM |
778 | if (found_key.type != BTRFS_EXTENT_REF_KEY) { |
779 | path->slots[0]++; | |
780 | continue; | |
781 | } | |
782 | ||
783 | cur_count++; | |
784 | ref_item = btrfs_item_ptr(l, path->slots[0], | |
785 | struct btrfs_extent_ref); | |
786 | found_objectid = btrfs_ref_root(l, ref_item); | |
787 | ||
56b453c9 CM |
788 | if (found_objectid != root_objectid) { |
789 | total_count = 2; | |
4313b399 | 790 | goto out; |
56b453c9 CM |
791 | } |
792 | total_count = 1; | |
be20aa9d CM |
793 | path->slots[0]++; |
794 | } | |
795 | if (cur_count == 0) { | |
796 | total_count = 0; | |
797 | goto out; | |
798 | } | |
be20aa9d CM |
799 | if (level >= 0 && root->node == count_path->nodes[level]) |
800 | goto out; | |
801 | level++; | |
802 | btrfs_release_path(root, path); | |
803 | goto again; | |
804 | ||
805 | out: | |
806 | btrfs_free_path(path); | |
807 | return total_count; | |
be20aa9d | 808 | } |
c5739bba | 809 | int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, |
7bb86316 | 810 | struct btrfs_root *root, u64 owner_objectid) |
c5739bba | 811 | { |
7bb86316 CM |
812 | u64 generation; |
813 | u64 key_objectid; | |
814 | u64 level; | |
815 | u32 nritems; | |
816 | struct btrfs_disk_key disk_key; | |
817 | ||
818 | level = btrfs_header_level(root->node); | |
819 | generation = trans->transid; | |
820 | nritems = btrfs_header_nritems(root->node); | |
821 | if (nritems > 0) { | |
822 | if (level == 0) | |
823 | btrfs_item_key(root->node, &disk_key, 0); | |
824 | else | |
825 | btrfs_node_key(root->node, &disk_key, 0); | |
826 | key_objectid = btrfs_disk_key_objectid(&disk_key); | |
827 | } else { | |
828 | key_objectid = 0; | |
829 | } | |
db94535d | 830 | return btrfs_inc_extent_ref(trans, root, root->node->start, |
7bb86316 | 831 | root->node->len, owner_objectid, |
f6dbff55 | 832 | generation, level, key_objectid); |
c5739bba CM |
833 | } |
834 | ||
e089f05c | 835 | int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 836 | struct extent_buffer *buf) |
02217ed2 | 837 | { |
db94535d | 838 | u64 bytenr; |
5f39d397 CM |
839 | u32 nritems; |
840 | struct btrfs_key key; | |
6407bf6d | 841 | struct btrfs_file_extent_item *fi; |
02217ed2 | 842 | int i; |
db94535d | 843 | int level; |
6407bf6d | 844 | int ret; |
54aa1f4d | 845 | int faili; |
a28ec197 | 846 | |
3768f368 | 847 | if (!root->ref_cows) |
a28ec197 | 848 | return 0; |
5f39d397 | 849 | |
db94535d | 850 | level = btrfs_header_level(buf); |
5f39d397 CM |
851 | nritems = btrfs_header_nritems(buf); |
852 | for (i = 0; i < nritems; i++) { | |
db94535d CM |
853 | if (level == 0) { |
854 | u64 disk_bytenr; | |
5f39d397 CM |
855 | btrfs_item_key_to_cpu(buf, &key, i); |
856 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
6407bf6d | 857 | continue; |
5f39d397 | 858 | fi = btrfs_item_ptr(buf, i, |
6407bf6d | 859 | struct btrfs_file_extent_item); |
5f39d397 | 860 | if (btrfs_file_extent_type(buf, fi) == |
236454df CM |
861 | BTRFS_FILE_EXTENT_INLINE) |
862 | continue; | |
db94535d CM |
863 | disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); |
864 | if (disk_bytenr == 0) | |
3a686375 | 865 | continue; |
db94535d | 866 | ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, |
7bb86316 CM |
867 | btrfs_file_extent_disk_num_bytes(buf, fi), |
868 | root->root_key.objectid, trans->transid, | |
869 | key.objectid, key.offset); | |
54aa1f4d CM |
870 | if (ret) { |
871 | faili = i; | |
872 | goto fail; | |
873 | } | |
6407bf6d | 874 | } else { |
db94535d | 875 | bytenr = btrfs_node_blockptr(buf, i); |
6caab489 | 876 | btrfs_node_key_to_cpu(buf, &key, i); |
db94535d | 877 | ret = btrfs_inc_extent_ref(trans, root, bytenr, |
7bb86316 CM |
878 | btrfs_level_size(root, level - 1), |
879 | root->root_key.objectid, | |
f6dbff55 CM |
880 | trans->transid, |
881 | level - 1, key.objectid); | |
54aa1f4d CM |
882 | if (ret) { |
883 | faili = i; | |
884 | goto fail; | |
885 | } | |
6407bf6d | 886 | } |
02217ed2 CM |
887 | } |
888 | return 0; | |
54aa1f4d | 889 | fail: |
ccd467d6 | 890 | WARN_ON(1); |
7bb86316 | 891 | #if 0 |
54aa1f4d | 892 | for (i =0; i < faili; i++) { |
db94535d CM |
893 | if (level == 0) { |
894 | u64 disk_bytenr; | |
5f39d397 CM |
895 | btrfs_item_key_to_cpu(buf, &key, i); |
896 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
54aa1f4d | 897 | continue; |
5f39d397 | 898 | fi = btrfs_item_ptr(buf, i, |
54aa1f4d | 899 | struct btrfs_file_extent_item); |
5f39d397 | 900 | if (btrfs_file_extent_type(buf, fi) == |
54aa1f4d CM |
901 | BTRFS_FILE_EXTENT_INLINE) |
902 | continue; | |
db94535d CM |
903 | disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); |
904 | if (disk_bytenr == 0) | |
54aa1f4d | 905 | continue; |
db94535d CM |
906 | err = btrfs_free_extent(trans, root, disk_bytenr, |
907 | btrfs_file_extent_disk_num_bytes(buf, | |
5f39d397 | 908 | fi), 0); |
54aa1f4d CM |
909 | BUG_ON(err); |
910 | } else { | |
db94535d CM |
911 | bytenr = btrfs_node_blockptr(buf, i); |
912 | err = btrfs_free_extent(trans, root, bytenr, | |
913 | btrfs_level_size(root, level - 1), 0); | |
54aa1f4d CM |
914 | BUG_ON(err); |
915 | } | |
916 | } | |
7bb86316 | 917 | #endif |
54aa1f4d | 918 | return ret; |
02217ed2 CM |
919 | } |
920 | ||
9078a3e1 CM |
921 | static int write_one_cache_group(struct btrfs_trans_handle *trans, |
922 | struct btrfs_root *root, | |
923 | struct btrfs_path *path, | |
924 | struct btrfs_block_group_cache *cache) | |
925 | { | |
926 | int ret; | |
927 | int pending_ret; | |
928 | struct btrfs_root *extent_root = root->fs_info->extent_root; | |
5f39d397 CM |
929 | unsigned long bi; |
930 | struct extent_buffer *leaf; | |
9078a3e1 | 931 | |
9078a3e1 | 932 | ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1); |
54aa1f4d CM |
933 | if (ret < 0) |
934 | goto fail; | |
9078a3e1 | 935 | BUG_ON(ret); |
5f39d397 CM |
936 | |
937 | leaf = path->nodes[0]; | |
938 | bi = btrfs_item_ptr_offset(leaf, path->slots[0]); | |
939 | write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item)); | |
940 | btrfs_mark_buffer_dirty(leaf); | |
9078a3e1 | 941 | btrfs_release_path(extent_root, path); |
54aa1f4d | 942 | fail: |
9078a3e1 CM |
943 | finish_current_insert(trans, extent_root); |
944 | pending_ret = del_pending_extents(trans, extent_root); | |
945 | if (ret) | |
946 | return ret; | |
947 | if (pending_ret) | |
948 | return pending_ret; | |
949 | return 0; | |
950 | ||
951 | } | |
952 | ||
96b5179d CM |
953 | int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, |
954 | struct btrfs_root *root) | |
9078a3e1 | 955 | { |
d1310b2e | 956 | struct extent_io_tree *block_group_cache; |
96b5179d | 957 | struct btrfs_block_group_cache *cache; |
9078a3e1 CM |
958 | int ret; |
959 | int err = 0; | |
960 | int werr = 0; | |
9078a3e1 | 961 | struct btrfs_path *path; |
96b5179d CM |
962 | u64 last = 0; |
963 | u64 start; | |
964 | u64 end; | |
965 | u64 ptr; | |
9078a3e1 | 966 | |
96b5179d | 967 | block_group_cache = &root->fs_info->block_group_cache; |
9078a3e1 CM |
968 | path = btrfs_alloc_path(); |
969 | if (!path) | |
970 | return -ENOMEM; | |
971 | ||
972 | while(1) { | |
96b5179d CM |
973 | ret = find_first_extent_bit(block_group_cache, last, |
974 | &start, &end, BLOCK_GROUP_DIRTY); | |
975 | if (ret) | |
9078a3e1 | 976 | break; |
54aa1f4d | 977 | |
96b5179d CM |
978 | last = end + 1; |
979 | ret = get_state_private(block_group_cache, start, &ptr); | |
980 | if (ret) | |
981 | break; | |
982 | ||
ae2f5411 | 983 | cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; |
96b5179d CM |
984 | err = write_one_cache_group(trans, root, |
985 | path, cache); | |
986 | /* | |
987 | * if we fail to write the cache group, we want | |
988 | * to keep it marked dirty in hopes that a later | |
989 | * write will work | |
990 | */ | |
991 | if (err) { | |
992 | werr = err; | |
993 | continue; | |
9078a3e1 | 994 | } |
96b5179d CM |
995 | clear_extent_bits(block_group_cache, start, end, |
996 | BLOCK_GROUP_DIRTY, GFP_NOFS); | |
9078a3e1 CM |
997 | } |
998 | btrfs_free_path(path); | |
999 | return werr; | |
1000 | } | |
1001 | ||
6324fbf3 CM |
1002 | static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info, |
1003 | u64 flags) | |
1004 | { | |
1005 | struct list_head *head = &info->space_info; | |
1006 | struct list_head *cur; | |
1007 | struct btrfs_space_info *found; | |
1008 | list_for_each(cur, head) { | |
1009 | found = list_entry(cur, struct btrfs_space_info, list); | |
1010 | if (found->flags == flags) | |
1011 | return found; | |
1012 | } | |
1013 | return NULL; | |
1014 | ||
1015 | } | |
1016 | ||
1017 | static int do_chunk_alloc(struct btrfs_trans_handle *trans, | |
1018 | struct btrfs_root *extent_root, u64 alloc_bytes, | |
1019 | u64 flags) | |
1020 | { | |
1021 | struct btrfs_space_info *space_info; | |
1022 | u64 thresh; | |
1023 | u64 start; | |
1024 | u64 num_bytes; | |
1025 | int ret; | |
1026 | ||
1027 | space_info = __find_space_info(extent_root->fs_info, flags); | |
1028 | BUG_ON(!space_info); | |
1029 | ||
1030 | if (space_info->full) | |
1031 | return 0; | |
1032 | ||
1033 | thresh = div_factor(space_info->total_bytes, 7); | |
1034 | if ((space_info->bytes_used + space_info->bytes_pinned + alloc_bytes) < | |
1035 | thresh) | |
1036 | return 0; | |
1037 | ||
1038 | ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags); | |
1039 | if (ret == -ENOSPC) { | |
1040 | printk("space info full %Lu\n", flags); | |
1041 | space_info->full = 1; | |
1042 | return 0; | |
1043 | } | |
1044 | ||
1045 | BUG_ON(ret); | |
1046 | ||
1047 | ret = btrfs_make_block_group(trans, extent_root, 0, flags, | |
1048 | extent_root->fs_info->chunk_root->root_key.objectid, | |
1049 | start, num_bytes); | |
1050 | BUG_ON(ret); | |
1051 | return 0; | |
1052 | } | |
1053 | ||
9078a3e1 CM |
1054 | static int update_block_group(struct btrfs_trans_handle *trans, |
1055 | struct btrfs_root *root, | |
db94535d | 1056 | u64 bytenr, u64 num_bytes, int alloc, |
0b86a832 | 1057 | int mark_free) |
9078a3e1 CM |
1058 | { |
1059 | struct btrfs_block_group_cache *cache; | |
1060 | struct btrfs_fs_info *info = root->fs_info; | |
db94535d | 1061 | u64 total = num_bytes; |
9078a3e1 | 1062 | u64 old_val; |
db94535d | 1063 | u64 byte_in_group; |
96b5179d CM |
1064 | u64 start; |
1065 | u64 end; | |
3e1ad54f | 1066 | |
9078a3e1 | 1067 | while(total) { |
db94535d | 1068 | cache = btrfs_lookup_block_group(info, bytenr); |
3e1ad54f | 1069 | if (!cache) { |
9078a3e1 | 1070 | return -1; |
cd1bc465 | 1071 | } |
db94535d CM |
1072 | byte_in_group = bytenr - cache->key.objectid; |
1073 | WARN_ON(byte_in_group > cache->key.offset); | |
96b5179d CM |
1074 | start = cache->key.objectid; |
1075 | end = start + cache->key.offset - 1; | |
1076 | set_extent_bits(&info->block_group_cache, start, end, | |
1077 | BLOCK_GROUP_DIRTY, GFP_NOFS); | |
9078a3e1 CM |
1078 | |
1079 | old_val = btrfs_block_group_used(&cache->item); | |
db94535d | 1080 | num_bytes = min(total, cache->key.offset - byte_in_group); |
cd1bc465 | 1081 | if (alloc) { |
db94535d | 1082 | old_val += num_bytes; |
6324fbf3 | 1083 | cache->space_info->bytes_used += num_bytes; |
cd1bc465 | 1084 | } else { |
db94535d | 1085 | old_val -= num_bytes; |
6324fbf3 | 1086 | cache->space_info->bytes_used -= num_bytes; |
f510cfec CM |
1087 | if (mark_free) { |
1088 | set_extent_dirty(&info->free_space_cache, | |
db94535d | 1089 | bytenr, bytenr + num_bytes - 1, |
f510cfec | 1090 | GFP_NOFS); |
e37c9e69 | 1091 | } |
cd1bc465 | 1092 | } |
9078a3e1 | 1093 | btrfs_set_block_group_used(&cache->item, old_val); |
db94535d CM |
1094 | total -= num_bytes; |
1095 | bytenr += num_bytes; | |
9078a3e1 CM |
1096 | } |
1097 | return 0; | |
1098 | } | |
6324fbf3 | 1099 | |
324ae4df Y |
1100 | static int update_pinned_extents(struct btrfs_root *root, |
1101 | u64 bytenr, u64 num, int pin) | |
1102 | { | |
1103 | u64 len; | |
1104 | struct btrfs_block_group_cache *cache; | |
1105 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1106 | ||
1107 | if (pin) { | |
1108 | set_extent_dirty(&fs_info->pinned_extents, | |
1109 | bytenr, bytenr + num - 1, GFP_NOFS); | |
1110 | } else { | |
1111 | clear_extent_dirty(&fs_info->pinned_extents, | |
1112 | bytenr, bytenr + num - 1, GFP_NOFS); | |
1113 | } | |
1114 | while (num > 0) { | |
1115 | cache = btrfs_lookup_block_group(fs_info, bytenr); | |
1116 | WARN_ON(!cache); | |
1117 | len = min(num, cache->key.offset - | |
1118 | (bytenr - cache->key.objectid)); | |
1119 | if (pin) { | |
1120 | cache->pinned += len; | |
6324fbf3 | 1121 | cache->space_info->bytes_pinned += len; |
324ae4df Y |
1122 | fs_info->total_pinned += len; |
1123 | } else { | |
1124 | cache->pinned -= len; | |
6324fbf3 | 1125 | cache->space_info->bytes_pinned -= len; |
324ae4df Y |
1126 | fs_info->total_pinned -= len; |
1127 | } | |
1128 | bytenr += len; | |
1129 | num -= len; | |
1130 | } | |
1131 | return 0; | |
1132 | } | |
9078a3e1 | 1133 | |
d1310b2e | 1134 | int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy) |
ccd467d6 | 1135 | { |
ccd467d6 | 1136 | u64 last = 0; |
1a5bc167 CM |
1137 | u64 start; |
1138 | u64 end; | |
d1310b2e | 1139 | struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents; |
ccd467d6 | 1140 | int ret; |
ccd467d6 CM |
1141 | |
1142 | while(1) { | |
1a5bc167 CM |
1143 | ret = find_first_extent_bit(pinned_extents, last, |
1144 | &start, &end, EXTENT_DIRTY); | |
1145 | if (ret) | |
ccd467d6 | 1146 | break; |
1a5bc167 CM |
1147 | set_extent_dirty(copy, start, end, GFP_NOFS); |
1148 | last = end + 1; | |
ccd467d6 CM |
1149 | } |
1150 | return 0; | |
1151 | } | |
1152 | ||
1153 | int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, | |
1154 | struct btrfs_root *root, | |
d1310b2e | 1155 | struct extent_io_tree *unpin) |
a28ec197 | 1156 | { |
1a5bc167 CM |
1157 | u64 start; |
1158 | u64 end; | |
a28ec197 | 1159 | int ret; |
d1310b2e | 1160 | struct extent_io_tree *free_space_cache; |
f510cfec | 1161 | free_space_cache = &root->fs_info->free_space_cache; |
a28ec197 CM |
1162 | |
1163 | while(1) { | |
1a5bc167 CM |
1164 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
1165 | EXTENT_DIRTY); | |
1166 | if (ret) | |
a28ec197 | 1167 | break; |
324ae4df | 1168 | update_pinned_extents(root, start, end + 1 - start, 0); |
1a5bc167 CM |
1169 | clear_extent_dirty(unpin, start, end, GFP_NOFS); |
1170 | set_extent_dirty(free_space_cache, start, end, GFP_NOFS); | |
a28ec197 CM |
1171 | } |
1172 | return 0; | |
1173 | } | |
1174 | ||
98ed5174 CM |
1175 | static int finish_current_insert(struct btrfs_trans_handle *trans, |
1176 | struct btrfs_root *extent_root) | |
037e6390 | 1177 | { |
7bb86316 CM |
1178 | u64 start; |
1179 | u64 end; | |
1180 | struct btrfs_fs_info *info = extent_root->fs_info; | |
d8d5f3e1 | 1181 | struct extent_buffer *eb; |
7bb86316 | 1182 | struct btrfs_path *path; |
e2fa7227 | 1183 | struct btrfs_key ins; |
d8d5f3e1 | 1184 | struct btrfs_disk_key first; |
234b63a0 | 1185 | struct btrfs_extent_item extent_item; |
037e6390 | 1186 | int ret; |
d8d5f3e1 | 1187 | int level; |
1a5bc167 | 1188 | int err = 0; |
037e6390 | 1189 | |
5f39d397 | 1190 | btrfs_set_stack_extent_refs(&extent_item, 1); |
62e2749e | 1191 | btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY); |
7bb86316 | 1192 | path = btrfs_alloc_path(); |
037e6390 | 1193 | |
26b8003f | 1194 | while(1) { |
1a5bc167 CM |
1195 | ret = find_first_extent_bit(&info->extent_ins, 0, &start, |
1196 | &end, EXTENT_LOCKED); | |
1197 | if (ret) | |
26b8003f CM |
1198 | break; |
1199 | ||
1a5bc167 CM |
1200 | ins.objectid = start; |
1201 | ins.offset = end + 1 - start; | |
1202 | err = btrfs_insert_item(trans, extent_root, &ins, | |
1203 | &extent_item, sizeof(extent_item)); | |
1204 | clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED, | |
1205 | GFP_NOFS); | |
d8d5f3e1 CM |
1206 | eb = read_tree_block(extent_root, ins.objectid, ins.offset); |
1207 | level = btrfs_header_level(eb); | |
1208 | if (level == 0) { | |
1209 | btrfs_item_key(eb, &first, 0); | |
1210 | } else { | |
1211 | btrfs_node_key(eb, &first, 0); | |
1212 | } | |
7bb86316 CM |
1213 | err = btrfs_insert_extent_backref(trans, extent_root, path, |
1214 | start, extent_root->root_key.objectid, | |
f6dbff55 CM |
1215 | 0, level, |
1216 | btrfs_disk_key_objectid(&first)); | |
7bb86316 | 1217 | BUG_ON(err); |
d8d5f3e1 | 1218 | free_extent_buffer(eb); |
037e6390 | 1219 | } |
7bb86316 | 1220 | btrfs_free_path(path); |
037e6390 CM |
1221 | return 0; |
1222 | } | |
1223 | ||
db94535d CM |
1224 | static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, |
1225 | int pending) | |
e20d96d6 | 1226 | { |
1a5bc167 | 1227 | int err = 0; |
5f39d397 | 1228 | struct extent_buffer *buf; |
8ef97622 | 1229 | |
f4b9aa8d | 1230 | if (!pending) { |
db94535d | 1231 | buf = btrfs_find_tree_block(root, bytenr, num_bytes); |
5f39d397 CM |
1232 | if (buf) { |
1233 | if (btrfs_buffer_uptodate(buf)) { | |
2c90e5d6 CM |
1234 | u64 transid = |
1235 | root->fs_info->running_transaction->transid; | |
dc17ff8f CM |
1236 | u64 header_transid = |
1237 | btrfs_header_generation(buf); | |
1238 | if (header_transid == transid) { | |
55c69072 | 1239 | clean_tree_block(NULL, root, buf); |
5f39d397 | 1240 | free_extent_buffer(buf); |
c549228f | 1241 | return 1; |
2c90e5d6 | 1242 | } |
f4b9aa8d | 1243 | } |
5f39d397 | 1244 | free_extent_buffer(buf); |
8ef97622 | 1245 | } |
324ae4df | 1246 | update_pinned_extents(root, bytenr, num_bytes, 1); |
f4b9aa8d | 1247 | } else { |
1a5bc167 | 1248 | set_extent_bits(&root->fs_info->pending_del, |
db94535d CM |
1249 | bytenr, bytenr + num_bytes - 1, |
1250 | EXTENT_LOCKED, GFP_NOFS); | |
f4b9aa8d | 1251 | } |
be744175 | 1252 | BUG_ON(err < 0); |
e20d96d6 CM |
1253 | return 0; |
1254 | } | |
1255 | ||
fec577fb | 1256 | /* |
a28ec197 | 1257 | * remove an extent from the root, returns 0 on success |
fec577fb | 1258 | */ |
e089f05c | 1259 | static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root |
7bb86316 CM |
1260 | *root, u64 bytenr, u64 num_bytes, |
1261 | u64 root_objectid, u64 ref_generation, | |
1262 | u64 owner_objectid, u64 owner_offset, int pin, | |
e37c9e69 | 1263 | int mark_free) |
a28ec197 | 1264 | { |
5caf2a00 | 1265 | struct btrfs_path *path; |
e2fa7227 | 1266 | struct btrfs_key key; |
1261ec42 CM |
1267 | struct btrfs_fs_info *info = root->fs_info; |
1268 | struct btrfs_root *extent_root = info->extent_root; | |
5f39d397 | 1269 | struct extent_buffer *leaf; |
a28ec197 | 1270 | int ret; |
952fccac CM |
1271 | int extent_slot = 0; |
1272 | int found_extent = 0; | |
1273 | int num_to_del = 1; | |
234b63a0 | 1274 | struct btrfs_extent_item *ei; |
cf27e1ee | 1275 | u32 refs; |
037e6390 | 1276 | |
db94535d | 1277 | key.objectid = bytenr; |
62e2749e | 1278 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
db94535d | 1279 | key.offset = num_bytes; |
5caf2a00 | 1280 | path = btrfs_alloc_path(); |
54aa1f4d CM |
1281 | if (!path) |
1282 | return -ENOMEM; | |
5f26f772 | 1283 | |
b0331a4c | 1284 | path->reada = 0; |
7bb86316 CM |
1285 | ret = lookup_extent_backref(trans, extent_root, path, |
1286 | bytenr, root_objectid, | |
1287 | ref_generation, | |
1288 | owner_objectid, owner_offset, 1); | |
1289 | if (ret == 0) { | |
952fccac CM |
1290 | struct btrfs_key found_key; |
1291 | extent_slot = path->slots[0]; | |
1292 | while(extent_slot > 0) { | |
1293 | extent_slot--; | |
1294 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
1295 | extent_slot); | |
1296 | if (found_key.objectid != bytenr) | |
1297 | break; | |
1298 | if (found_key.type == BTRFS_EXTENT_ITEM_KEY && | |
1299 | found_key.offset == num_bytes) { | |
1300 | found_extent = 1; | |
1301 | break; | |
1302 | } | |
1303 | if (path->slots[0] - extent_slot > 5) | |
1304 | break; | |
1305 | } | |
1306 | if (!found_extent) | |
1307 | ret = btrfs_del_item(trans, extent_root, path); | |
7bb86316 CM |
1308 | } else { |
1309 | btrfs_print_leaf(extent_root, path->nodes[0]); | |
1310 | WARN_ON(1); | |
1311 | printk("Unable to find ref byte nr %Lu root %Lu " | |
1312 | " gen %Lu owner %Lu offset %Lu\n", bytenr, | |
1313 | root_objectid, ref_generation, owner_objectid, | |
1314 | owner_offset); | |
1315 | } | |
952fccac CM |
1316 | if (!found_extent) { |
1317 | btrfs_release_path(extent_root, path); | |
1318 | ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); | |
1319 | if (ret < 0) | |
1320 | return ret; | |
1321 | BUG_ON(ret); | |
1322 | extent_slot = path->slots[0]; | |
1323 | } | |
5f39d397 CM |
1324 | |
1325 | leaf = path->nodes[0]; | |
952fccac | 1326 | ei = btrfs_item_ptr(leaf, extent_slot, |
123abc88 | 1327 | struct btrfs_extent_item); |
5f39d397 CM |
1328 | refs = btrfs_extent_refs(leaf, ei); |
1329 | BUG_ON(refs == 0); | |
1330 | refs -= 1; | |
1331 | btrfs_set_extent_refs(leaf, ei, refs); | |
952fccac | 1332 | |
5f39d397 CM |
1333 | btrfs_mark_buffer_dirty(leaf); |
1334 | ||
952fccac CM |
1335 | if (refs == 0 && found_extent && path->slots[0] == extent_slot + 1) { |
1336 | /* if the back ref and the extent are next to each other | |
1337 | * they get deleted below in one shot | |
1338 | */ | |
1339 | path->slots[0] = extent_slot; | |
1340 | num_to_del = 2; | |
1341 | } else if (found_extent) { | |
1342 | /* otherwise delete the extent back ref */ | |
1343 | ret = btrfs_del_item(trans, extent_root, path); | |
1344 | BUG_ON(ret); | |
1345 | /* if refs are 0, we need to setup the path for deletion */ | |
1346 | if (refs == 0) { | |
1347 | btrfs_release_path(extent_root, path); | |
1348 | ret = btrfs_search_slot(trans, extent_root, &key, path, | |
1349 | -1, 1); | |
1350 | if (ret < 0) | |
1351 | return ret; | |
1352 | BUG_ON(ret); | |
1353 | } | |
1354 | } | |
1355 | ||
cf27e1ee | 1356 | if (refs == 0) { |
db94535d CM |
1357 | u64 super_used; |
1358 | u64 root_used; | |
78fae27e CM |
1359 | |
1360 | if (pin) { | |
db94535d | 1361 | ret = pin_down_bytes(root, bytenr, num_bytes, 0); |
c549228f Y |
1362 | if (ret > 0) |
1363 | mark_free = 1; | |
1364 | BUG_ON(ret < 0); | |
78fae27e CM |
1365 | } |
1366 | ||
58176a96 | 1367 | /* block accounting for super block */ |
db94535d CM |
1368 | super_used = btrfs_super_bytes_used(&info->super_copy); |
1369 | btrfs_set_super_bytes_used(&info->super_copy, | |
1370 | super_used - num_bytes); | |
58176a96 JB |
1371 | |
1372 | /* block accounting for root item */ | |
db94535d | 1373 | root_used = btrfs_root_used(&root->root_item); |
5f39d397 | 1374 | btrfs_set_root_used(&root->root_item, |
db94535d | 1375 | root_used - num_bytes); |
952fccac CM |
1376 | ret = btrfs_del_items(trans, extent_root, path, path->slots[0], |
1377 | num_to_del); | |
54aa1f4d CM |
1378 | if (ret) { |
1379 | return ret; | |
1380 | } | |
db94535d | 1381 | ret = update_block_group(trans, root, bytenr, num_bytes, 0, |
0b86a832 | 1382 | mark_free); |
9078a3e1 | 1383 | BUG_ON(ret); |
a28ec197 | 1384 | } |
5caf2a00 | 1385 | btrfs_free_path(path); |
e089f05c | 1386 | finish_current_insert(trans, extent_root); |
a28ec197 CM |
1387 | return ret; |
1388 | } | |
1389 | ||
a28ec197 CM |
1390 | /* |
1391 | * find all the blocks marked as pending in the radix tree and remove | |
1392 | * them from the extent map | |
1393 | */ | |
e089f05c CM |
1394 | static int del_pending_extents(struct btrfs_trans_handle *trans, struct |
1395 | btrfs_root *extent_root) | |
a28ec197 CM |
1396 | { |
1397 | int ret; | |
e20d96d6 | 1398 | int err = 0; |
1a5bc167 CM |
1399 | u64 start; |
1400 | u64 end; | |
d1310b2e CM |
1401 | struct extent_io_tree *pending_del; |
1402 | struct extent_io_tree *pinned_extents; | |
8ef97622 | 1403 | |
1a5bc167 CM |
1404 | pending_del = &extent_root->fs_info->pending_del; |
1405 | pinned_extents = &extent_root->fs_info->pinned_extents; | |
a28ec197 CM |
1406 | |
1407 | while(1) { | |
1a5bc167 CM |
1408 | ret = find_first_extent_bit(pending_del, 0, &start, &end, |
1409 | EXTENT_LOCKED); | |
1410 | if (ret) | |
a28ec197 | 1411 | break; |
324ae4df | 1412 | update_pinned_extents(extent_root, start, end + 1 - start, 1); |
1a5bc167 CM |
1413 | clear_extent_bits(pending_del, start, end, EXTENT_LOCKED, |
1414 | GFP_NOFS); | |
1415 | ret = __free_extent(trans, extent_root, | |
7bb86316 CM |
1416 | start, end + 1 - start, |
1417 | extent_root->root_key.objectid, | |
1418 | 0, 0, 0, 0, 0); | |
1a5bc167 CM |
1419 | if (ret) |
1420 | err = ret; | |
fec577fb | 1421 | } |
e20d96d6 | 1422 | return err; |
fec577fb CM |
1423 | } |
1424 | ||
1425 | /* | |
1426 | * remove an extent from the root, returns 0 on success | |
1427 | */ | |
e089f05c | 1428 | int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root |
7bb86316 CM |
1429 | *root, u64 bytenr, u64 num_bytes, |
1430 | u64 root_objectid, u64 ref_generation, | |
1431 | u64 owner_objectid, u64 owner_offset, int pin) | |
fec577fb | 1432 | { |
9f5fae2f | 1433 | struct btrfs_root *extent_root = root->fs_info->extent_root; |
fec577fb CM |
1434 | int pending_ret; |
1435 | int ret; | |
a28ec197 | 1436 | |
db94535d | 1437 | WARN_ON(num_bytes < root->sectorsize); |
7bb86316 CM |
1438 | if (!root->ref_cows) |
1439 | ref_generation = 0; | |
1440 | ||
fec577fb | 1441 | if (root == extent_root) { |
db94535d | 1442 | pin_down_bytes(root, bytenr, num_bytes, 1); |
fec577fb CM |
1443 | return 0; |
1444 | } | |
7bb86316 CM |
1445 | ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid, |
1446 | ref_generation, owner_objectid, owner_offset, | |
1447 | pin, pin == 0); | |
e20d96d6 | 1448 | pending_ret = del_pending_extents(trans, root->fs_info->extent_root); |
fec577fb CM |
1449 | return ret ? ret : pending_ret; |
1450 | } | |
1451 | ||
87ee04eb CM |
1452 | static u64 stripe_align(struct btrfs_root *root, u64 val) |
1453 | { | |
1454 | u64 mask = ((u64)root->stripesize - 1); | |
1455 | u64 ret = (val + mask) & ~mask; | |
1456 | return ret; | |
1457 | } | |
1458 | ||
fec577fb CM |
1459 | /* |
1460 | * walks the btree of allocated extents and find a hole of a given size. | |
1461 | * The key ins is changed to record the hole: | |
1462 | * ins->objectid == block start | |
62e2749e | 1463 | * ins->flags = BTRFS_EXTENT_ITEM_KEY |
fec577fb CM |
1464 | * ins->offset == number of blocks |
1465 | * Any available blocks before search_start are skipped. | |
1466 | */ | |
98ed5174 CM |
1467 | static int noinline find_free_extent(struct btrfs_trans_handle *trans, |
1468 | struct btrfs_root *orig_root, | |
1469 | u64 num_bytes, u64 empty_size, | |
1470 | u64 search_start, u64 search_end, | |
1471 | u64 hint_byte, struct btrfs_key *ins, | |
1472 | u64 exclude_start, u64 exclude_nr, | |
1473 | int data) | |
fec577fb | 1474 | { |
87ee04eb | 1475 | int ret; |
be744175 | 1476 | u64 orig_search_start = search_start; |
9f5fae2f | 1477 | struct btrfs_root * root = orig_root->fs_info->extent_root; |
f2458e1d | 1478 | struct btrfs_fs_info *info = root->fs_info; |
db94535d | 1479 | u64 total_needed = num_bytes; |
be08c1b9 | 1480 | struct btrfs_block_group_cache *block_group; |
be744175 | 1481 | int full_scan = 0; |
fbdc762b | 1482 | int wrapped = 0; |
fec577fb | 1483 | |
db94535d | 1484 | WARN_ON(num_bytes < root->sectorsize); |
b1a4d965 CM |
1485 | btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY); |
1486 | ||
7f93bf8d CM |
1487 | if (search_end == (u64)-1) |
1488 | search_end = btrfs_super_total_bytes(&info->super_copy); | |
0b86a832 | 1489 | |
db94535d CM |
1490 | if (hint_byte) { |
1491 | block_group = btrfs_lookup_block_group(info, hint_byte); | |
1a2b2ac7 CM |
1492 | if (!block_group) |
1493 | hint_byte = search_start; | |
be744175 | 1494 | block_group = btrfs_find_block_group(root, block_group, |
db94535d | 1495 | hint_byte, data, 1); |
be744175 CM |
1496 | } else { |
1497 | block_group = btrfs_find_block_group(root, | |
1a2b2ac7 CM |
1498 | trans->block_group, |
1499 | search_start, data, 1); | |
be744175 CM |
1500 | } |
1501 | ||
6702ed49 | 1502 | total_needed += empty_size; |
0b86a832 | 1503 | |
be744175 | 1504 | check_failed: |
70b043f0 CM |
1505 | if (!block_group) { |
1506 | block_group = btrfs_lookup_block_group(info, search_start); | |
1507 | if (!block_group) | |
1508 | block_group = btrfs_lookup_block_group(info, | |
1509 | orig_search_start); | |
1510 | } | |
0b86a832 CM |
1511 | ret = find_search_start(root, &block_group, &search_start, |
1512 | total_needed, data); | |
1513 | if (ret) | |
d548ee51 | 1514 | goto error; |
e19caa5f | 1515 | |
0b86a832 CM |
1516 | search_start = stripe_align(root, search_start); |
1517 | ins->objectid = search_start; | |
1518 | ins->offset = num_bytes; | |
e37c9e69 | 1519 | |
db94535d | 1520 | if (ins->objectid + num_bytes >= search_end) |
cf67582b | 1521 | goto enospc; |
0b86a832 CM |
1522 | |
1523 | if (ins->objectid + num_bytes > | |
1524 | block_group->key.objectid + block_group->key.offset) { | |
e19caa5f CM |
1525 | search_start = block_group->key.objectid + |
1526 | block_group->key.offset; | |
1527 | goto new_group; | |
1528 | } | |
0b86a832 | 1529 | |
1a5bc167 | 1530 | if (test_range_bit(&info->extent_ins, ins->objectid, |
db94535d CM |
1531 | ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) { |
1532 | search_start = ins->objectid + num_bytes; | |
1a5bc167 CM |
1533 | goto new_group; |
1534 | } | |
0b86a832 | 1535 | |
1a5bc167 | 1536 | if (test_range_bit(&info->pinned_extents, ins->objectid, |
db94535d CM |
1537 | ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) { |
1538 | search_start = ins->objectid + num_bytes; | |
1a5bc167 | 1539 | goto new_group; |
fec577fb | 1540 | } |
0b86a832 | 1541 | |
db94535d | 1542 | if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start && |
f2654de4 CM |
1543 | ins->objectid < exclude_start + exclude_nr)) { |
1544 | search_start = exclude_start + exclude_nr; | |
1545 | goto new_group; | |
1546 | } | |
0b86a832 | 1547 | |
6324fbf3 | 1548 | if (!(data & BTRFS_BLOCK_GROUP_DATA)) { |
5276aeda | 1549 | block_group = btrfs_lookup_block_group(info, ins->objectid); |
26b8003f CM |
1550 | if (block_group) |
1551 | trans->block_group = block_group; | |
f2458e1d | 1552 | } |
db94535d | 1553 | ins->offset = num_bytes; |
fec577fb | 1554 | return 0; |
be744175 CM |
1555 | |
1556 | new_group: | |
db94535d | 1557 | if (search_start + num_bytes >= search_end) { |
cf67582b | 1558 | enospc: |
be744175 | 1559 | search_start = orig_search_start; |
fbdc762b CM |
1560 | if (full_scan) { |
1561 | ret = -ENOSPC; | |
1562 | goto error; | |
1563 | } | |
6702ed49 CM |
1564 | if (wrapped) { |
1565 | if (!full_scan) | |
1566 | total_needed -= empty_size; | |
fbdc762b | 1567 | full_scan = 1; |
6702ed49 | 1568 | } else |
fbdc762b | 1569 | wrapped = 1; |
be744175 | 1570 | } |
5276aeda | 1571 | block_group = btrfs_lookup_block_group(info, search_start); |
fbdc762b | 1572 | cond_resched(); |
1a2b2ac7 CM |
1573 | block_group = btrfs_find_block_group(root, block_group, |
1574 | search_start, data, 0); | |
be744175 CM |
1575 | goto check_failed; |
1576 | ||
0f70abe2 | 1577 | error: |
0f70abe2 | 1578 | return ret; |
fec577fb | 1579 | } |
fec577fb CM |
1580 | /* |
1581 | * finds a free extent and does all the dirty work required for allocation | |
1582 | * returns the key for the extent through ins, and a tree buffer for | |
1583 | * the first block of the extent through buf. | |
1584 | * | |
1585 | * returns 0 if everything worked, non-zero otherwise. | |
1586 | */ | |
4d775673 | 1587 | int btrfs_alloc_extent(struct btrfs_trans_handle *trans, |
7bb86316 CM |
1588 | struct btrfs_root *root, |
1589 | u64 num_bytes, u64 root_objectid, u64 ref_generation, | |
1590 | u64 owner, u64 owner_offset, | |
1591 | u64 empty_size, u64 hint_byte, | |
be08c1b9 | 1592 | u64 search_end, struct btrfs_key *ins, int data) |
fec577fb CM |
1593 | { |
1594 | int ret; | |
1595 | int pending_ret; | |
c31f8830 CM |
1596 | u64 super_used; |
1597 | u64 root_used; | |
fbdc762b | 1598 | u64 search_start = 0; |
edbd8d4e | 1599 | u64 new_hint; |
47e4bb98 | 1600 | u32 sizes[2]; |
1261ec42 CM |
1601 | struct btrfs_fs_info *info = root->fs_info; |
1602 | struct btrfs_root *extent_root = info->extent_root; | |
47e4bb98 CM |
1603 | struct btrfs_extent_item *extent_item; |
1604 | struct btrfs_extent_ref *ref; | |
7bb86316 | 1605 | struct btrfs_path *path; |
47e4bb98 | 1606 | struct btrfs_key keys[2]; |
8f662a76 | 1607 | |
6324fbf3 CM |
1608 | if (data) { |
1609 | data = BTRFS_BLOCK_GROUP_DATA; | |
1610 | } else if (root == root->fs_info->chunk_root) { | |
1611 | data = BTRFS_BLOCK_GROUP_SYSTEM; | |
1612 | } else { | |
1613 | data = BTRFS_BLOCK_GROUP_METADATA; | |
1614 | } | |
1615 | ||
1616 | if (root->ref_cows) { | |
1617 | if (data != BTRFS_BLOCK_GROUP_METADATA) { | |
1618 | ret = do_chunk_alloc(trans, root->fs_info->extent_root, | |
1619 | num_bytes, | |
1620 | BTRFS_BLOCK_GROUP_METADATA); | |
1621 | BUG_ON(ret); | |
1622 | } | |
1623 | ret = do_chunk_alloc(trans, root->fs_info->extent_root, | |
1624 | num_bytes, data); | |
1625 | BUG_ON(ret); | |
1626 | } | |
0b86a832 | 1627 | |
8f662a76 | 1628 | new_hint = max(hint_byte, root->fs_info->alloc_start); |
edbd8d4e CM |
1629 | if (new_hint < btrfs_super_total_bytes(&info->super_copy)) |
1630 | hint_byte = new_hint; | |
8f662a76 | 1631 | |
db94535d CM |
1632 | WARN_ON(num_bytes < root->sectorsize); |
1633 | ret = find_free_extent(trans, root, num_bytes, empty_size, | |
1634 | search_start, search_end, hint_byte, ins, | |
26b8003f CM |
1635 | trans->alloc_exclude_start, |
1636 | trans->alloc_exclude_nr, data); | |
ccd467d6 | 1637 | BUG_ON(ret); |
f2654de4 CM |
1638 | if (ret) |
1639 | return ret; | |
fec577fb | 1640 | |
58176a96 | 1641 | /* block accounting for super block */ |
db94535d CM |
1642 | super_used = btrfs_super_bytes_used(&info->super_copy); |
1643 | btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes); | |
26b8003f | 1644 | |
58176a96 | 1645 | /* block accounting for root item */ |
db94535d CM |
1646 | root_used = btrfs_root_used(&root->root_item); |
1647 | btrfs_set_root_used(&root->root_item, root_used + num_bytes); | |
58176a96 | 1648 | |
f510cfec CM |
1649 | clear_extent_dirty(&root->fs_info->free_space_cache, |
1650 | ins->objectid, ins->objectid + ins->offset - 1, | |
1651 | GFP_NOFS); | |
1652 | ||
26b8003f | 1653 | if (root == extent_root) { |
1a5bc167 CM |
1654 | set_extent_bits(&root->fs_info->extent_ins, ins->objectid, |
1655 | ins->objectid + ins->offset - 1, | |
1656 | EXTENT_LOCKED, GFP_NOFS); | |
26b8003f CM |
1657 | goto update_block; |
1658 | } | |
1659 | ||
1660 | WARN_ON(trans->alloc_exclude_nr); | |
1661 | trans->alloc_exclude_start = ins->objectid; | |
1662 | trans->alloc_exclude_nr = ins->offset; | |
037e6390 | 1663 | |
47e4bb98 CM |
1664 | memcpy(&keys[0], ins, sizeof(*ins)); |
1665 | keys[1].offset = hash_extent_ref(root_objectid, ref_generation, | |
1666 | owner, owner_offset); | |
1667 | keys[1].objectid = ins->objectid; | |
1668 | keys[1].type = BTRFS_EXTENT_REF_KEY; | |
1669 | sizes[0] = sizeof(*extent_item); | |
1670 | sizes[1] = sizeof(*ref); | |
7bb86316 CM |
1671 | |
1672 | path = btrfs_alloc_path(); | |
1673 | BUG_ON(!path); | |
47e4bb98 CM |
1674 | |
1675 | ret = btrfs_insert_empty_items(trans, extent_root, path, keys, | |
1676 | sizes, 2); | |
26b8003f | 1677 | |
ccd467d6 | 1678 | BUG_ON(ret); |
47e4bb98 CM |
1679 | extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0], |
1680 | struct btrfs_extent_item); | |
1681 | btrfs_set_extent_refs(path->nodes[0], extent_item, 1); | |
1682 | ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1, | |
1683 | struct btrfs_extent_ref); | |
1684 | ||
1685 | btrfs_set_ref_root(path->nodes[0], ref, root_objectid); | |
1686 | btrfs_set_ref_generation(path->nodes[0], ref, ref_generation); | |
1687 | btrfs_set_ref_objectid(path->nodes[0], ref, owner); | |
1688 | btrfs_set_ref_offset(path->nodes[0], ref, owner_offset); | |
1689 | ||
1690 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
1691 | ||
1692 | trans->alloc_exclude_start = 0; | |
1693 | trans->alloc_exclude_nr = 0; | |
7bb86316 | 1694 | btrfs_free_path(path); |
e089f05c | 1695 | finish_current_insert(trans, extent_root); |
e20d96d6 | 1696 | pending_ret = del_pending_extents(trans, extent_root); |
f510cfec | 1697 | |
e37c9e69 | 1698 | if (ret) { |
037e6390 | 1699 | return ret; |
e37c9e69 CM |
1700 | } |
1701 | if (pending_ret) { | |
037e6390 | 1702 | return pending_ret; |
e37c9e69 | 1703 | } |
26b8003f CM |
1704 | |
1705 | update_block: | |
0b86a832 | 1706 | ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0); |
f5947066 CM |
1707 | if (ret) { |
1708 | printk("update block group failed for %Lu %Lu\n", | |
1709 | ins->objectid, ins->offset); | |
1710 | BUG(); | |
1711 | } | |
037e6390 | 1712 | return 0; |
fec577fb CM |
1713 | } |
1714 | ||
1715 | /* | |
1716 | * helper function to allocate a block for a given tree | |
1717 | * returns the tree buffer or NULL. | |
1718 | */ | |
5f39d397 | 1719 | struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, |
db94535d | 1720 | struct btrfs_root *root, |
7bb86316 CM |
1721 | u32 blocksize, |
1722 | u64 root_objectid, u64 hint, | |
1723 | u64 empty_size) | |
1724 | { | |
1725 | u64 ref_generation; | |
1726 | ||
1727 | if (root->ref_cows) | |
1728 | ref_generation = trans->transid; | |
1729 | else | |
1730 | ref_generation = 0; | |
1731 | ||
1732 | ||
1733 | return __btrfs_alloc_free_block(trans, root, blocksize, root_objectid, | |
1734 | ref_generation, 0, 0, hint, empty_size); | |
1735 | } | |
1736 | ||
1737 | /* | |
1738 | * helper function to allocate a block for a given tree | |
1739 | * returns the tree buffer or NULL. | |
1740 | */ | |
1741 | struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, | |
1742 | struct btrfs_root *root, | |
1743 | u32 blocksize, | |
1744 | u64 root_objectid, | |
1745 | u64 ref_generation, | |
1746 | u64 first_objectid, | |
1747 | int level, | |
1748 | u64 hint, | |
5f39d397 | 1749 | u64 empty_size) |
fec577fb | 1750 | { |
e2fa7227 | 1751 | struct btrfs_key ins; |
fec577fb | 1752 | int ret; |
5f39d397 | 1753 | struct extent_buffer *buf; |
fec577fb | 1754 | |
7bb86316 CM |
1755 | ret = btrfs_alloc_extent(trans, root, blocksize, |
1756 | root_objectid, ref_generation, | |
f6dbff55 | 1757 | level, first_objectid, empty_size, hint, |
db94535d | 1758 | (u64)-1, &ins, 0); |
fec577fb | 1759 | if (ret) { |
54aa1f4d CM |
1760 | BUG_ON(ret > 0); |
1761 | return ERR_PTR(ret); | |
fec577fb | 1762 | } |
db94535d | 1763 | buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize); |
54aa1f4d | 1764 | if (!buf) { |
7bb86316 CM |
1765 | btrfs_free_extent(trans, root, ins.objectid, blocksize, |
1766 | root->root_key.objectid, ref_generation, | |
1767 | 0, 0, 0); | |
54aa1f4d CM |
1768 | return ERR_PTR(-ENOMEM); |
1769 | } | |
55c69072 CM |
1770 | btrfs_set_header_generation(buf, trans->transid); |
1771 | clean_tree_block(trans, root, buf); | |
1772 | wait_on_tree_block_writeback(root, buf); | |
5f39d397 | 1773 | btrfs_set_buffer_uptodate(buf); |
55c69072 CM |
1774 | |
1775 | if (PageDirty(buf->first_page)) { | |
1776 | printk("page %lu dirty\n", buf->first_page->index); | |
1777 | WARN_ON(1); | |
1778 | } | |
1779 | ||
5f39d397 CM |
1780 | set_extent_dirty(&trans->transaction->dirty_pages, buf->start, |
1781 | buf->start + buf->len - 1, GFP_NOFS); | |
d1310b2e | 1782 | set_extent_bits(&BTRFS_I(root->fs_info->btree_inode)->io_tree, |
19c00ddc CM |
1783 | buf->start, buf->start + buf->len - 1, |
1784 | EXTENT_CSUM, GFP_NOFS); | |
1785 | buf->flags |= EXTENT_CSUM; | |
9afbb0b7 CM |
1786 | if (!btrfs_test_opt(root, SSD)) |
1787 | btrfs_set_buffer_defrag(buf); | |
d3c2fdcf | 1788 | trans->blocks_used++; |
fec577fb CM |
1789 | return buf; |
1790 | } | |
a28ec197 | 1791 | |
98ed5174 CM |
1792 | static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans, |
1793 | struct btrfs_root *root, | |
1794 | struct extent_buffer *leaf) | |
6407bf6d | 1795 | { |
7bb86316 CM |
1796 | u64 leaf_owner; |
1797 | u64 leaf_generation; | |
5f39d397 | 1798 | struct btrfs_key key; |
6407bf6d CM |
1799 | struct btrfs_file_extent_item *fi; |
1800 | int i; | |
1801 | int nritems; | |
1802 | int ret; | |
1803 | ||
5f39d397 CM |
1804 | BUG_ON(!btrfs_is_leaf(leaf)); |
1805 | nritems = btrfs_header_nritems(leaf); | |
7bb86316 CM |
1806 | leaf_owner = btrfs_header_owner(leaf); |
1807 | leaf_generation = btrfs_header_generation(leaf); | |
1808 | ||
6407bf6d | 1809 | for (i = 0; i < nritems; i++) { |
db94535d | 1810 | u64 disk_bytenr; |
5f39d397 CM |
1811 | |
1812 | btrfs_item_key_to_cpu(leaf, &key, i); | |
1813 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
6407bf6d CM |
1814 | continue; |
1815 | fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item); | |
5f39d397 CM |
1816 | if (btrfs_file_extent_type(leaf, fi) == |
1817 | BTRFS_FILE_EXTENT_INLINE) | |
236454df | 1818 | continue; |
6407bf6d CM |
1819 | /* |
1820 | * FIXME make sure to insert a trans record that | |
1821 | * repeats the snapshot del on crash | |
1822 | */ | |
db94535d CM |
1823 | disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); |
1824 | if (disk_bytenr == 0) | |
3a686375 | 1825 | continue; |
db94535d | 1826 | ret = btrfs_free_extent(trans, root, disk_bytenr, |
7bb86316 CM |
1827 | btrfs_file_extent_disk_num_bytes(leaf, fi), |
1828 | leaf_owner, leaf_generation, | |
1829 | key.objectid, key.offset, 0); | |
6407bf6d CM |
1830 | BUG_ON(ret); |
1831 | } | |
1832 | return 0; | |
1833 | } | |
1834 | ||
98ed5174 | 1835 | static void noinline reada_walk_down(struct btrfs_root *root, |
bea495e5 CM |
1836 | struct extent_buffer *node, |
1837 | int slot) | |
e011599b | 1838 | { |
db94535d | 1839 | u64 bytenr; |
bea495e5 CM |
1840 | u64 last = 0; |
1841 | u32 nritems; | |
e011599b | 1842 | u32 refs; |
db94535d | 1843 | u32 blocksize; |
bea495e5 CM |
1844 | int ret; |
1845 | int i; | |
1846 | int level; | |
1847 | int skipped = 0; | |
e011599b | 1848 | |
5f39d397 | 1849 | nritems = btrfs_header_nritems(node); |
db94535d | 1850 | level = btrfs_header_level(node); |
bea495e5 CM |
1851 | if (level) |
1852 | return; | |
1853 | ||
1854 | for (i = slot; i < nritems && skipped < 32; i++) { | |
db94535d | 1855 | bytenr = btrfs_node_blockptr(node, i); |
bea495e5 CM |
1856 | if (last && ((bytenr > last && bytenr - last > 32 * 1024) || |
1857 | (last > bytenr && last - bytenr > 32 * 1024))) { | |
1858 | skipped++; | |
e011599b | 1859 | continue; |
bea495e5 CM |
1860 | } |
1861 | blocksize = btrfs_level_size(root, level - 1); | |
1862 | if (i != slot) { | |
1863 | ret = lookup_extent_ref(NULL, root, bytenr, | |
1864 | blocksize, &refs); | |
1865 | BUG_ON(ret); | |
1866 | if (refs != 1) { | |
1867 | skipped++; | |
1868 | continue; | |
1869 | } | |
1870 | } | |
409eb95d | 1871 | mutex_unlock(&root->fs_info->fs_mutex); |
db94535d | 1872 | ret = readahead_tree_block(root, bytenr, blocksize); |
bea495e5 | 1873 | last = bytenr + blocksize; |
409eb95d CM |
1874 | cond_resched(); |
1875 | mutex_lock(&root->fs_info->fs_mutex); | |
e011599b CM |
1876 | if (ret) |
1877 | break; | |
1878 | } | |
1879 | } | |
1880 | ||
9aca1d51 CM |
1881 | /* |
1882 | * helper function for drop_snapshot, this walks down the tree dropping ref | |
1883 | * counts as it goes. | |
1884 | */ | |
98ed5174 CM |
1885 | static int noinline walk_down_tree(struct btrfs_trans_handle *trans, |
1886 | struct btrfs_root *root, | |
1887 | struct btrfs_path *path, int *level) | |
20524f02 | 1888 | { |
7bb86316 CM |
1889 | u64 root_owner; |
1890 | u64 root_gen; | |
1891 | u64 bytenr; | |
5f39d397 CM |
1892 | struct extent_buffer *next; |
1893 | struct extent_buffer *cur; | |
7bb86316 | 1894 | struct extent_buffer *parent; |
db94535d | 1895 | u32 blocksize; |
20524f02 CM |
1896 | int ret; |
1897 | u32 refs; | |
1898 | ||
5caf2a00 CM |
1899 | WARN_ON(*level < 0); |
1900 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
5f39d397 | 1901 | ret = lookup_extent_ref(trans, root, |
db94535d CM |
1902 | path->nodes[*level]->start, |
1903 | path->nodes[*level]->len, &refs); | |
20524f02 CM |
1904 | BUG_ON(ret); |
1905 | if (refs > 1) | |
1906 | goto out; | |
e011599b | 1907 | |
9aca1d51 CM |
1908 | /* |
1909 | * walk down to the last node level and free all the leaves | |
1910 | */ | |
6407bf6d | 1911 | while(*level >= 0) { |
5caf2a00 CM |
1912 | WARN_ON(*level < 0); |
1913 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
20524f02 | 1914 | cur = path->nodes[*level]; |
e011599b | 1915 | |
5f39d397 | 1916 | if (btrfs_header_level(cur) != *level) |
2c90e5d6 | 1917 | WARN_ON(1); |
e011599b | 1918 | |
7518a238 | 1919 | if (path->slots[*level] >= |
5f39d397 | 1920 | btrfs_header_nritems(cur)) |
20524f02 | 1921 | break; |
6407bf6d CM |
1922 | if (*level == 0) { |
1923 | ret = drop_leaf_ref(trans, root, cur); | |
1924 | BUG_ON(ret); | |
1925 | break; | |
1926 | } | |
db94535d CM |
1927 | bytenr = btrfs_node_blockptr(cur, path->slots[*level]); |
1928 | blocksize = btrfs_level_size(root, *level - 1); | |
1929 | ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs); | |
6407bf6d CM |
1930 | BUG_ON(ret); |
1931 | if (refs != 1) { | |
7bb86316 CM |
1932 | parent = path->nodes[*level]; |
1933 | root_owner = btrfs_header_owner(parent); | |
1934 | root_gen = btrfs_header_generation(parent); | |
20524f02 | 1935 | path->slots[*level]++; |
db94535d | 1936 | ret = btrfs_free_extent(trans, root, bytenr, |
7bb86316 CM |
1937 | blocksize, root_owner, |
1938 | root_gen, 0, 0, 1); | |
20524f02 CM |
1939 | BUG_ON(ret); |
1940 | continue; | |
1941 | } | |
db94535d | 1942 | next = btrfs_find_tree_block(root, bytenr, blocksize); |
5f39d397 CM |
1943 | if (!next || !btrfs_buffer_uptodate(next)) { |
1944 | free_extent_buffer(next); | |
bea495e5 | 1945 | reada_walk_down(root, cur, path->slots[*level]); |
e9d0b13b | 1946 | mutex_unlock(&root->fs_info->fs_mutex); |
db94535d | 1947 | next = read_tree_block(root, bytenr, blocksize); |
e9d0b13b CM |
1948 | mutex_lock(&root->fs_info->fs_mutex); |
1949 | ||
1950 | /* we dropped the lock, check one more time */ | |
db94535d CM |
1951 | ret = lookup_extent_ref(trans, root, bytenr, |
1952 | blocksize, &refs); | |
e9d0b13b CM |
1953 | BUG_ON(ret); |
1954 | if (refs != 1) { | |
7bb86316 CM |
1955 | parent = path->nodes[*level]; |
1956 | root_owner = btrfs_header_owner(parent); | |
1957 | root_gen = btrfs_header_generation(parent); | |
1958 | ||
e9d0b13b | 1959 | path->slots[*level]++; |
5f39d397 | 1960 | free_extent_buffer(next); |
7bb86316 CM |
1961 | ret = btrfs_free_extent(trans, root, bytenr, |
1962 | blocksize, | |
1963 | root_owner, | |
1964 | root_gen, 0, 0, 1); | |
e9d0b13b CM |
1965 | BUG_ON(ret); |
1966 | continue; | |
1967 | } | |
1968 | } | |
5caf2a00 | 1969 | WARN_ON(*level <= 0); |
83e15a28 | 1970 | if (path->nodes[*level-1]) |
5f39d397 | 1971 | free_extent_buffer(path->nodes[*level-1]); |
20524f02 | 1972 | path->nodes[*level-1] = next; |
5f39d397 | 1973 | *level = btrfs_header_level(next); |
20524f02 CM |
1974 | path->slots[*level] = 0; |
1975 | } | |
1976 | out: | |
5caf2a00 CM |
1977 | WARN_ON(*level < 0); |
1978 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
7bb86316 CM |
1979 | |
1980 | if (path->nodes[*level] == root->node) { | |
1981 | root_owner = root->root_key.objectid; | |
1982 | parent = path->nodes[*level]; | |
1983 | } else { | |
1984 | parent = path->nodes[*level + 1]; | |
1985 | root_owner = btrfs_header_owner(parent); | |
1986 | } | |
1987 | ||
1988 | root_gen = btrfs_header_generation(parent); | |
db94535d | 1989 | ret = btrfs_free_extent(trans, root, path->nodes[*level]->start, |
7bb86316 CM |
1990 | path->nodes[*level]->len, |
1991 | root_owner, root_gen, 0, 0, 1); | |
5f39d397 | 1992 | free_extent_buffer(path->nodes[*level]); |
20524f02 CM |
1993 | path->nodes[*level] = NULL; |
1994 | *level += 1; | |
1995 | BUG_ON(ret); | |
1996 | return 0; | |
1997 | } | |
1998 | ||
9aca1d51 CM |
1999 | /* |
2000 | * helper for dropping snapshots. This walks back up the tree in the path | |
2001 | * to find the first node higher up where we haven't yet gone through | |
2002 | * all the slots | |
2003 | */ | |
98ed5174 CM |
2004 | static int noinline walk_up_tree(struct btrfs_trans_handle *trans, |
2005 | struct btrfs_root *root, | |
2006 | struct btrfs_path *path, int *level) | |
20524f02 | 2007 | { |
7bb86316 CM |
2008 | u64 root_owner; |
2009 | u64 root_gen; | |
2010 | struct btrfs_root_item *root_item = &root->root_item; | |
20524f02 CM |
2011 | int i; |
2012 | int slot; | |
2013 | int ret; | |
9f3a7427 | 2014 | |
234b63a0 | 2015 | for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { |
20524f02 | 2016 | slot = path->slots[i]; |
5f39d397 CM |
2017 | if (slot < btrfs_header_nritems(path->nodes[i]) - 1) { |
2018 | struct extent_buffer *node; | |
2019 | struct btrfs_disk_key disk_key; | |
2020 | node = path->nodes[i]; | |
20524f02 CM |
2021 | path->slots[i]++; |
2022 | *level = i; | |
9f3a7427 | 2023 | WARN_ON(*level == 0); |
5f39d397 | 2024 | btrfs_node_key(node, &disk_key, path->slots[i]); |
9f3a7427 | 2025 | memcpy(&root_item->drop_progress, |
5f39d397 | 2026 | &disk_key, sizeof(disk_key)); |
9f3a7427 | 2027 | root_item->drop_level = i; |
20524f02 CM |
2028 | return 0; |
2029 | } else { | |
7bb86316 CM |
2030 | if (path->nodes[*level] == root->node) { |
2031 | root_owner = root->root_key.objectid; | |
2032 | root_gen = | |
2033 | btrfs_header_generation(path->nodes[*level]); | |
2034 | } else { | |
2035 | struct extent_buffer *node; | |
2036 | node = path->nodes[*level + 1]; | |
2037 | root_owner = btrfs_header_owner(node); | |
2038 | root_gen = btrfs_header_generation(node); | |
2039 | } | |
e089f05c | 2040 | ret = btrfs_free_extent(trans, root, |
db94535d | 2041 | path->nodes[*level]->start, |
7bb86316 CM |
2042 | path->nodes[*level]->len, |
2043 | root_owner, root_gen, 0, 0, 1); | |
6407bf6d | 2044 | BUG_ON(ret); |
5f39d397 | 2045 | free_extent_buffer(path->nodes[*level]); |
83e15a28 | 2046 | path->nodes[*level] = NULL; |
20524f02 | 2047 | *level = i + 1; |
20524f02 CM |
2048 | } |
2049 | } | |
2050 | return 1; | |
2051 | } | |
2052 | ||
9aca1d51 CM |
2053 | /* |
2054 | * drop the reference count on the tree rooted at 'snap'. This traverses | |
2055 | * the tree freeing any blocks that have a ref count of zero after being | |
2056 | * decremented. | |
2057 | */ | |
e089f05c | 2058 | int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root |
9f3a7427 | 2059 | *root) |
20524f02 | 2060 | { |
3768f368 | 2061 | int ret = 0; |
9aca1d51 | 2062 | int wret; |
20524f02 | 2063 | int level; |
5caf2a00 | 2064 | struct btrfs_path *path; |
20524f02 CM |
2065 | int i; |
2066 | int orig_level; | |
9f3a7427 | 2067 | struct btrfs_root_item *root_item = &root->root_item; |
20524f02 | 2068 | |
5caf2a00 CM |
2069 | path = btrfs_alloc_path(); |
2070 | BUG_ON(!path); | |
20524f02 | 2071 | |
5f39d397 | 2072 | level = btrfs_header_level(root->node); |
20524f02 | 2073 | orig_level = level; |
9f3a7427 CM |
2074 | if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) { |
2075 | path->nodes[level] = root->node; | |
f510cfec | 2076 | extent_buffer_get(root->node); |
9f3a7427 CM |
2077 | path->slots[level] = 0; |
2078 | } else { | |
2079 | struct btrfs_key key; | |
5f39d397 CM |
2080 | struct btrfs_disk_key found_key; |
2081 | struct extent_buffer *node; | |
6702ed49 | 2082 | |
9f3a7427 | 2083 | btrfs_disk_key_to_cpu(&key, &root_item->drop_progress); |
6702ed49 CM |
2084 | level = root_item->drop_level; |
2085 | path->lowest_level = level; | |
9f3a7427 | 2086 | wret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
6702ed49 | 2087 | if (wret < 0) { |
9f3a7427 CM |
2088 | ret = wret; |
2089 | goto out; | |
2090 | } | |
5f39d397 CM |
2091 | node = path->nodes[level]; |
2092 | btrfs_node_key(node, &found_key, path->slots[level]); | |
2093 | WARN_ON(memcmp(&found_key, &root_item->drop_progress, | |
2094 | sizeof(found_key))); | |
9f3a7427 | 2095 | } |
20524f02 | 2096 | while(1) { |
5caf2a00 | 2097 | wret = walk_down_tree(trans, root, path, &level); |
9aca1d51 | 2098 | if (wret > 0) |
20524f02 | 2099 | break; |
9aca1d51 CM |
2100 | if (wret < 0) |
2101 | ret = wret; | |
2102 | ||
5caf2a00 | 2103 | wret = walk_up_tree(trans, root, path, &level); |
9aca1d51 | 2104 | if (wret > 0) |
20524f02 | 2105 | break; |
9aca1d51 CM |
2106 | if (wret < 0) |
2107 | ret = wret; | |
409eb95d | 2108 | ret = -EAGAIN; |
409eb95d | 2109 | break; |
20524f02 | 2110 | } |
83e15a28 | 2111 | for (i = 0; i <= orig_level; i++) { |
5caf2a00 | 2112 | if (path->nodes[i]) { |
5f39d397 | 2113 | free_extent_buffer(path->nodes[i]); |
0f82731f | 2114 | path->nodes[i] = NULL; |
83e15a28 | 2115 | } |
20524f02 | 2116 | } |
9f3a7427 | 2117 | out: |
5caf2a00 | 2118 | btrfs_free_path(path); |
9aca1d51 | 2119 | return ret; |
20524f02 | 2120 | } |
9078a3e1 | 2121 | |
96b5179d | 2122 | int btrfs_free_block_groups(struct btrfs_fs_info *info) |
9078a3e1 | 2123 | { |
96b5179d CM |
2124 | u64 start; |
2125 | u64 end; | |
b97f9203 | 2126 | u64 ptr; |
9078a3e1 | 2127 | int ret; |
9078a3e1 | 2128 | while(1) { |
96b5179d CM |
2129 | ret = find_first_extent_bit(&info->block_group_cache, 0, |
2130 | &start, &end, (unsigned int)-1); | |
2131 | if (ret) | |
9078a3e1 | 2132 | break; |
b97f9203 Y |
2133 | ret = get_state_private(&info->block_group_cache, start, &ptr); |
2134 | if (!ret) | |
2135 | kfree((void *)(unsigned long)ptr); | |
96b5179d CM |
2136 | clear_extent_bits(&info->block_group_cache, start, |
2137 | end, (unsigned int)-1, GFP_NOFS); | |
9078a3e1 | 2138 | } |
e37c9e69 | 2139 | while(1) { |
f510cfec CM |
2140 | ret = find_first_extent_bit(&info->free_space_cache, 0, |
2141 | &start, &end, EXTENT_DIRTY); | |
2142 | if (ret) | |
e37c9e69 | 2143 | break; |
f510cfec CM |
2144 | clear_extent_dirty(&info->free_space_cache, start, |
2145 | end, GFP_NOFS); | |
e37c9e69 | 2146 | } |
be744175 CM |
2147 | return 0; |
2148 | } | |
2149 | ||
98ed5174 CM |
2150 | static int noinline relocate_inode_pages(struct inode *inode, u64 start, |
2151 | u64 len) | |
edbd8d4e CM |
2152 | { |
2153 | u64 page_start; | |
2154 | u64 page_end; | |
2155 | u64 delalloc_start; | |
2156 | u64 existing_delalloc; | |
2157 | unsigned long last_index; | |
edbd8d4e CM |
2158 | unsigned long i; |
2159 | struct page *page; | |
d1310b2e | 2160 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
4313b399 CM |
2161 | struct file_ra_state *ra; |
2162 | ||
2163 | ra = kzalloc(sizeof(*ra), GFP_NOFS); | |
edbd8d4e CM |
2164 | |
2165 | mutex_lock(&inode->i_mutex); | |
4313b399 | 2166 | i = start >> PAGE_CACHE_SHIFT; |
edbd8d4e CM |
2167 | last_index = (start + len - 1) >> PAGE_CACHE_SHIFT; |
2168 | ||
4313b399 CM |
2169 | file_ra_state_init(ra, inode->i_mapping); |
2170 | btrfs_force_ra(inode->i_mapping, ra, NULL, i, last_index); | |
2171 | kfree(ra); | |
edbd8d4e | 2172 | |
4313b399 | 2173 | for (; i <= last_index; i++) { |
edbd8d4e CM |
2174 | page = grab_cache_page(inode->i_mapping, i); |
2175 | if (!page) | |
2176 | goto out_unlock; | |
2177 | if (!PageUptodate(page)) { | |
2178 | btrfs_readpage(NULL, page); | |
2179 | lock_page(page); | |
2180 | if (!PageUptodate(page)) { | |
2181 | unlock_page(page); | |
2182 | page_cache_release(page); | |
2183 | goto out_unlock; | |
2184 | } | |
2185 | } | |
2186 | page_start = (u64)page->index << PAGE_CACHE_SHIFT; | |
2187 | page_end = page_start + PAGE_CACHE_SIZE - 1; | |
2188 | ||
d1310b2e | 2189 | lock_extent(io_tree, page_start, page_end, GFP_NOFS); |
edbd8d4e CM |
2190 | |
2191 | delalloc_start = page_start; | |
d1310b2e CM |
2192 | existing_delalloc = count_range_bits(io_tree, |
2193 | &delalloc_start, page_end, | |
2194 | PAGE_CACHE_SIZE, EXTENT_DELALLOC); | |
edbd8d4e | 2195 | |
d1310b2e | 2196 | set_extent_delalloc(io_tree, page_start, |
edbd8d4e CM |
2197 | page_end, GFP_NOFS); |
2198 | ||
d1310b2e | 2199 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); |
edbd8d4e CM |
2200 | set_page_dirty(page); |
2201 | unlock_page(page); | |
2202 | page_cache_release(page); | |
2203 | } | |
2204 | ||
2205 | out_unlock: | |
2206 | mutex_unlock(&inode->i_mutex); | |
2207 | return 0; | |
2208 | } | |
2209 | ||
4313b399 CM |
2210 | /* |
2211 | * note, this releases the path | |
2212 | */ | |
98ed5174 | 2213 | static int noinline relocate_one_reference(struct btrfs_root *extent_root, |
edbd8d4e | 2214 | struct btrfs_path *path, |
4313b399 | 2215 | struct btrfs_key *extent_key) |
edbd8d4e CM |
2216 | { |
2217 | struct inode *inode; | |
2218 | struct btrfs_root *found_root; | |
4313b399 CM |
2219 | struct btrfs_key *root_location; |
2220 | struct btrfs_extent_ref *ref; | |
2221 | u64 ref_root; | |
2222 | u64 ref_gen; | |
2223 | u64 ref_objectid; | |
2224 | u64 ref_offset; | |
edbd8d4e CM |
2225 | int ret; |
2226 | ||
4313b399 CM |
2227 | ref = btrfs_item_ptr(path->nodes[0], path->slots[0], |
2228 | struct btrfs_extent_ref); | |
2229 | ref_root = btrfs_ref_root(path->nodes[0], ref); | |
2230 | ref_gen = btrfs_ref_generation(path->nodes[0], ref); | |
2231 | ref_objectid = btrfs_ref_objectid(path->nodes[0], ref); | |
2232 | ref_offset = btrfs_ref_offset(path->nodes[0], ref); | |
2233 | btrfs_release_path(extent_root, path); | |
2234 | ||
2235 | root_location = kmalloc(sizeof(*root_location), GFP_NOFS); | |
2236 | root_location->objectid = ref_root; | |
edbd8d4e | 2237 | if (ref_gen == 0) |
4313b399 | 2238 | root_location->offset = 0; |
edbd8d4e | 2239 | else |
4313b399 CM |
2240 | root_location->offset = (u64)-1; |
2241 | root_location->type = BTRFS_ROOT_ITEM_KEY; | |
edbd8d4e CM |
2242 | |
2243 | found_root = btrfs_read_fs_root_no_name(extent_root->fs_info, | |
4313b399 | 2244 | root_location); |
edbd8d4e | 2245 | BUG_ON(!found_root); |
4313b399 | 2246 | kfree(root_location); |
edbd8d4e CM |
2247 | |
2248 | if (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID) { | |
2249 | mutex_unlock(&extent_root->fs_info->fs_mutex); | |
2250 | inode = btrfs_iget_locked(extent_root->fs_info->sb, | |
2251 | ref_objectid, found_root); | |
2252 | if (inode->i_state & I_NEW) { | |
2253 | /* the inode and parent dir are two different roots */ | |
2254 | BTRFS_I(inode)->root = found_root; | |
2255 | BTRFS_I(inode)->location.objectid = ref_objectid; | |
2256 | BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY; | |
2257 | BTRFS_I(inode)->location.offset = 0; | |
2258 | btrfs_read_locked_inode(inode); | |
2259 | unlock_new_inode(inode); | |
2260 | ||
2261 | } | |
2262 | /* this can happen if the reference is not against | |
2263 | * the latest version of the tree root | |
2264 | */ | |
2265 | if (is_bad_inode(inode)) { | |
2266 | mutex_lock(&extent_root->fs_info->fs_mutex); | |
2267 | goto out; | |
2268 | } | |
2269 | relocate_inode_pages(inode, ref_offset, extent_key->offset); | |
2270 | /* FIXME, data=ordered will help get rid of this */ | |
2271 | filemap_fdatawrite(inode->i_mapping); | |
2272 | iput(inode); | |
2273 | mutex_lock(&extent_root->fs_info->fs_mutex); | |
2274 | } else { | |
2275 | struct btrfs_trans_handle *trans; | |
2276 | struct btrfs_key found_key; | |
2277 | struct extent_buffer *eb; | |
2278 | int level; | |
2279 | int i; | |
2280 | ||
2281 | trans = btrfs_start_transaction(found_root, 1); | |
2282 | eb = read_tree_block(found_root, extent_key->objectid, | |
2283 | extent_key->offset); | |
2284 | level = btrfs_header_level(eb); | |
2285 | ||
2286 | if (level == 0) | |
2287 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
2288 | else | |
2289 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
2290 | ||
2291 | free_extent_buffer(eb); | |
2292 | ||
2293 | path->lowest_level = level; | |
8f662a76 | 2294 | path->reada = 2; |
edbd8d4e CM |
2295 | ret = btrfs_search_slot(trans, found_root, &found_key, path, |
2296 | 0, 1); | |
2297 | path->lowest_level = 0; | |
2298 | for (i = level; i < BTRFS_MAX_LEVEL; i++) { | |
2299 | if (!path->nodes[i]) | |
2300 | break; | |
2301 | free_extent_buffer(path->nodes[i]); | |
2302 | path->nodes[i] = NULL; | |
2303 | } | |
2304 | btrfs_release_path(found_root, path); | |
2305 | btrfs_end_transaction(trans, found_root); | |
2306 | } | |
2307 | ||
2308 | out: | |
2309 | return 0; | |
2310 | } | |
2311 | ||
98ed5174 CM |
2312 | static int noinline relocate_one_extent(struct btrfs_root *extent_root, |
2313 | struct btrfs_path *path, | |
2314 | struct btrfs_key *extent_key) | |
edbd8d4e CM |
2315 | { |
2316 | struct btrfs_key key; | |
2317 | struct btrfs_key found_key; | |
edbd8d4e | 2318 | struct extent_buffer *leaf; |
edbd8d4e CM |
2319 | u32 nritems; |
2320 | u32 item_size; | |
2321 | int ret = 0; | |
2322 | ||
2323 | key.objectid = extent_key->objectid; | |
2324 | key.type = BTRFS_EXTENT_REF_KEY; | |
2325 | key.offset = 0; | |
2326 | ||
2327 | while(1) { | |
2328 | ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); | |
2329 | ||
edbd8d4e CM |
2330 | if (ret < 0) |
2331 | goto out; | |
2332 | ||
2333 | ret = 0; | |
2334 | leaf = path->nodes[0]; | |
2335 | nritems = btrfs_header_nritems(leaf); | |
2336 | if (path->slots[0] == nritems) | |
2337 | goto out; | |
2338 | ||
2339 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2340 | if (found_key.objectid != extent_key->objectid) | |
2341 | break; | |
2342 | ||
2343 | if (found_key.type != BTRFS_EXTENT_REF_KEY) | |
2344 | break; | |
2345 | ||
2346 | key.offset = found_key.offset + 1; | |
2347 | item_size = btrfs_item_size_nr(leaf, path->slots[0]); | |
2348 | ||
4313b399 | 2349 | ret = relocate_one_reference(extent_root, path, extent_key); |
edbd8d4e CM |
2350 | if (ret) |
2351 | goto out; | |
2352 | } | |
2353 | ret = 0; | |
2354 | out: | |
2355 | btrfs_release_path(extent_root, path); | |
2356 | return ret; | |
2357 | } | |
2358 | ||
edbd8d4e CM |
2359 | int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size) |
2360 | { | |
2361 | struct btrfs_trans_handle *trans; | |
2362 | struct btrfs_root *tree_root = root->fs_info->tree_root; | |
2363 | struct btrfs_path *path; | |
2364 | u64 cur_byte; | |
2365 | u64 total_found; | |
edbd8d4e | 2366 | struct btrfs_fs_info *info = root->fs_info; |
d1310b2e | 2367 | struct extent_io_tree *block_group_cache; |
edbd8d4e | 2368 | struct btrfs_key key; |
73e48b27 | 2369 | struct btrfs_key found_key; |
edbd8d4e CM |
2370 | struct extent_buffer *leaf; |
2371 | u32 nritems; | |
2372 | int ret; | |
725c8463 | 2373 | int progress = 0; |
edbd8d4e CM |
2374 | |
2375 | btrfs_set_super_total_bytes(&info->super_copy, new_size); | |
c31f8830 CM |
2376 | clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1, |
2377 | GFP_NOFS); | |
edbd8d4e CM |
2378 | block_group_cache = &info->block_group_cache; |
2379 | path = btrfs_alloc_path(); | |
2380 | root = root->fs_info->extent_root; | |
8f662a76 | 2381 | path->reada = 2; |
edbd8d4e CM |
2382 | |
2383 | again: | |
2384 | total_found = 0; | |
2385 | key.objectid = new_size; | |
edbd8d4e CM |
2386 | key.offset = 0; |
2387 | key.type = 0; | |
73e48b27 | 2388 | cur_byte = key.objectid; |
4313b399 | 2389 | |
73e48b27 Y |
2390 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2391 | if (ret < 0) | |
2392 | goto out; | |
2393 | ||
0b86a832 | 2394 | ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY); |
73e48b27 Y |
2395 | if (ret < 0) |
2396 | goto out; | |
2397 | if (ret == 0) { | |
2398 | leaf = path->nodes[0]; | |
2399 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2400 | if (found_key.objectid + found_key.offset > new_size) { | |
2401 | cur_byte = found_key.objectid; | |
2402 | key.objectid = cur_byte; | |
2403 | } | |
2404 | } | |
2405 | btrfs_release_path(root, path); | |
2406 | ||
2407 | while(1) { | |
edbd8d4e CM |
2408 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2409 | if (ret < 0) | |
2410 | goto out; | |
73e48b27 | 2411 | |
edbd8d4e | 2412 | leaf = path->nodes[0]; |
73e48b27 Y |
2413 | nritems = btrfs_header_nritems(leaf); |
2414 | next: | |
2415 | if (path->slots[0] >= nritems) { | |
2416 | ret = btrfs_next_leaf(root, path); | |
2417 | if (ret < 0) | |
2418 | goto out; | |
2419 | if (ret == 1) { | |
2420 | ret = 0; | |
2421 | break; | |
edbd8d4e | 2422 | } |
73e48b27 Y |
2423 | leaf = path->nodes[0]; |
2424 | nritems = btrfs_header_nritems(leaf); | |
edbd8d4e | 2425 | } |
73e48b27 Y |
2426 | |
2427 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
725c8463 CM |
2428 | |
2429 | if (progress && need_resched()) { | |
2430 | memcpy(&key, &found_key, sizeof(key)); | |
2431 | mutex_unlock(&root->fs_info->fs_mutex); | |
2432 | cond_resched(); | |
2433 | mutex_lock(&root->fs_info->fs_mutex); | |
2434 | btrfs_release_path(root, path); | |
2435 | btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
2436 | progress = 0; | |
2437 | goto next; | |
2438 | } | |
2439 | progress = 1; | |
2440 | ||
73e48b27 Y |
2441 | if (btrfs_key_type(&found_key) != BTRFS_EXTENT_ITEM_KEY || |
2442 | found_key.objectid + found_key.offset <= cur_byte) { | |
edbd8d4e | 2443 | path->slots[0]++; |
edbd8d4e CM |
2444 | goto next; |
2445 | } | |
73e48b27 | 2446 | |
edbd8d4e CM |
2447 | total_found++; |
2448 | cur_byte = found_key.objectid + found_key.offset; | |
2449 | key.objectid = cur_byte; | |
2450 | btrfs_release_path(root, path); | |
2451 | ret = relocate_one_extent(root, path, &found_key); | |
2452 | } | |
2453 | ||
2454 | btrfs_release_path(root, path); | |
2455 | ||
2456 | if (total_found > 0) { | |
2457 | trans = btrfs_start_transaction(tree_root, 1); | |
2458 | btrfs_commit_transaction(trans, tree_root); | |
2459 | ||
2460 | mutex_unlock(&root->fs_info->fs_mutex); | |
2461 | btrfs_clean_old_snapshots(tree_root); | |
2462 | mutex_lock(&root->fs_info->fs_mutex); | |
2463 | ||
2464 | trans = btrfs_start_transaction(tree_root, 1); | |
2465 | btrfs_commit_transaction(trans, tree_root); | |
2466 | goto again; | |
2467 | } | |
2468 | ||
2469 | trans = btrfs_start_transaction(root, 1); | |
2470 | key.objectid = new_size; | |
2471 | key.offset = 0; | |
2472 | key.type = 0; | |
2473 | while(1) { | |
4313b399 CM |
2474 | u64 ptr; |
2475 | ||
edbd8d4e CM |
2476 | ret = btrfs_search_slot(trans, root, &key, path, -1, 1); |
2477 | if (ret < 0) | |
2478 | goto out; | |
73e48b27 | 2479 | |
edbd8d4e CM |
2480 | leaf = path->nodes[0]; |
2481 | nritems = btrfs_header_nritems(leaf); | |
73e48b27 Y |
2482 | bg_next: |
2483 | if (path->slots[0] >= nritems) { | |
2484 | ret = btrfs_next_leaf(root, path); | |
2485 | if (ret < 0) | |
2486 | break; | |
2487 | if (ret == 1) { | |
2488 | ret = 0; | |
2489 | break; | |
edbd8d4e | 2490 | } |
73e48b27 | 2491 | leaf = path->nodes[0]; |
1372f8e6 CM |
2492 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); |
2493 | ||
2494 | /* | |
2495 | * btrfs_next_leaf doesn't cow buffers, we have to | |
2496 | * do the search again | |
2497 | */ | |
2498 | memcpy(&key, &found_key, sizeof(key)); | |
2499 | btrfs_release_path(root, path); | |
725c8463 | 2500 | goto resched_check; |
73e48b27 Y |
2501 | } |
2502 | ||
2503 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2504 | if (btrfs_key_type(&found_key) != BTRFS_BLOCK_GROUP_ITEM_KEY) { | |
2505 | printk("shrinker found key %Lu %u %Lu\n", | |
2506 | found_key.objectid, found_key.type, | |
2507 | found_key.offset); | |
2508 | path->slots[0]++; | |
edbd8d4e CM |
2509 | goto bg_next; |
2510 | } | |
edbd8d4e CM |
2511 | ret = get_state_private(&info->block_group_cache, |
2512 | found_key.objectid, &ptr); | |
2513 | if (!ret) | |
2514 | kfree((void *)(unsigned long)ptr); | |
2515 | ||
2516 | clear_extent_bits(&info->block_group_cache, found_key.objectid, | |
2517 | found_key.objectid + found_key.offset - 1, | |
2518 | (unsigned int)-1, GFP_NOFS); | |
2519 | ||
2520 | key.objectid = found_key.objectid + 1; | |
2521 | btrfs_del_item(trans, root, path); | |
2522 | btrfs_release_path(root, path); | |
725c8463 CM |
2523 | resched_check: |
2524 | if (need_resched()) { | |
2525 | mutex_unlock(&root->fs_info->fs_mutex); | |
2526 | cond_resched(); | |
2527 | mutex_lock(&root->fs_info->fs_mutex); | |
2528 | } | |
edbd8d4e CM |
2529 | } |
2530 | clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1, | |
2531 | GFP_NOFS); | |
2532 | btrfs_commit_transaction(trans, root); | |
2533 | out: | |
2534 | btrfs_free_path(path); | |
2535 | return ret; | |
2536 | } | |
2537 | ||
2538 | int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, | |
2539 | struct btrfs_root *root, u64 new_size) | |
2540 | { | |
0b86a832 CM |
2541 | btrfs_set_super_total_bytes(&root->fs_info->super_copy, new_size); |
2542 | return 0; | |
2543 | } | |
edbd8d4e | 2544 | |
0b86a832 CM |
2545 | int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path, |
2546 | struct btrfs_key *key) | |
2547 | { | |
2548 | int ret; | |
2549 | struct btrfs_key found_key; | |
2550 | struct extent_buffer *leaf; | |
2551 | int slot; | |
edbd8d4e | 2552 | |
0b86a832 CM |
2553 | ret = btrfs_search_slot(NULL, root, key, path, 0, 0); |
2554 | if (ret < 0) | |
2555 | return ret; | |
2556 | while(1) { | |
2557 | slot = path->slots[0]; | |
edbd8d4e | 2558 | leaf = path->nodes[0]; |
0b86a832 CM |
2559 | if (slot >= btrfs_header_nritems(leaf)) { |
2560 | ret = btrfs_next_leaf(root, path); | |
2561 | if (ret == 0) | |
2562 | continue; | |
2563 | if (ret < 0) | |
2564 | goto error; | |
2565 | break; | |
edbd8d4e | 2566 | } |
0b86a832 | 2567 | btrfs_item_key_to_cpu(leaf, &found_key, slot); |
edbd8d4e | 2568 | |
0b86a832 CM |
2569 | if (found_key.objectid >= key->objectid && |
2570 | found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) | |
2571 | return 0; | |
2572 | path->slots[0]++; | |
edbd8d4e | 2573 | } |
0b86a832 CM |
2574 | ret = -ENOENT; |
2575 | error: | |
2576 | return ret; | |
edbd8d4e CM |
2577 | } |
2578 | ||
6324fbf3 CM |
2579 | static int update_space_info(struct btrfs_fs_info *info, u64 flags, |
2580 | u64 total_bytes, u64 bytes_used, | |
2581 | struct btrfs_space_info **space_info) | |
2582 | { | |
2583 | struct btrfs_space_info *found; | |
2584 | ||
2585 | found = __find_space_info(info, flags); | |
2586 | if (found) { | |
2587 | found->total_bytes += total_bytes; | |
2588 | found->bytes_used += bytes_used; | |
2589 | WARN_ON(found->total_bytes < found->bytes_used); | |
2590 | *space_info = found; | |
2591 | return 0; | |
2592 | } | |
2593 | found = kmalloc(sizeof(*found), GFP_NOFS); | |
2594 | if (!found) | |
2595 | return -ENOMEM; | |
2596 | ||
2597 | list_add(&found->list, &info->space_info); | |
2598 | found->flags = flags; | |
2599 | found->total_bytes = total_bytes; | |
2600 | found->bytes_used = bytes_used; | |
2601 | found->bytes_pinned = 0; | |
2602 | found->full = 0; | |
2603 | *space_info = found; | |
2604 | return 0; | |
2605 | } | |
2606 | ||
9078a3e1 CM |
2607 | int btrfs_read_block_groups(struct btrfs_root *root) |
2608 | { | |
2609 | struct btrfs_path *path; | |
2610 | int ret; | |
96b5179d | 2611 | int bit; |
9078a3e1 | 2612 | struct btrfs_block_group_cache *cache; |
be744175 | 2613 | struct btrfs_fs_info *info = root->fs_info; |
6324fbf3 | 2614 | struct btrfs_space_info *space_info; |
d1310b2e | 2615 | struct extent_io_tree *block_group_cache; |
9078a3e1 CM |
2616 | struct btrfs_key key; |
2617 | struct btrfs_key found_key; | |
5f39d397 | 2618 | struct extent_buffer *leaf; |
96b5179d CM |
2619 | |
2620 | block_group_cache = &info->block_group_cache; | |
be744175 | 2621 | root = info->extent_root; |
9078a3e1 | 2622 | key.objectid = 0; |
0b86a832 | 2623 | key.offset = 0; |
9078a3e1 | 2624 | btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); |
9078a3e1 CM |
2625 | path = btrfs_alloc_path(); |
2626 | if (!path) | |
2627 | return -ENOMEM; | |
2628 | ||
2629 | while(1) { | |
0b86a832 CM |
2630 | ret = find_first_block_group(root, path, &key); |
2631 | if (ret > 0) { | |
2632 | ret = 0; | |
2633 | goto error; | |
9078a3e1 | 2634 | } |
0b86a832 CM |
2635 | if (ret != 0) |
2636 | goto error; | |
2637 | ||
5f39d397 CM |
2638 | leaf = path->nodes[0]; |
2639 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
9078a3e1 CM |
2640 | cache = kmalloc(sizeof(*cache), GFP_NOFS); |
2641 | if (!cache) { | |
0b86a832 | 2642 | ret = -ENOMEM; |
9078a3e1 CM |
2643 | break; |
2644 | } | |
3e1ad54f | 2645 | |
5f39d397 CM |
2646 | read_extent_buffer(leaf, &cache->item, |
2647 | btrfs_item_ptr_offset(leaf, path->slots[0]), | |
2648 | sizeof(cache->item)); | |
9078a3e1 | 2649 | memcpy(&cache->key, &found_key, sizeof(found_key)); |
e37c9e69 | 2650 | cache->cached = 0; |
324ae4df | 2651 | cache->pinned = 0; |
0b86a832 | 2652 | |
9078a3e1 CM |
2653 | key.objectid = found_key.objectid + found_key.offset; |
2654 | btrfs_release_path(root, path); | |
0b86a832 CM |
2655 | cache->flags = btrfs_block_group_flags(&cache->item); |
2656 | bit = 0; | |
2657 | if (cache->flags & BTRFS_BLOCK_GROUP_DATA) { | |
96b5179d | 2658 | bit = BLOCK_GROUP_DATA; |
0b86a832 CM |
2659 | } else if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) { |
2660 | bit = BLOCK_GROUP_SYSTEM; | |
2661 | } else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { | |
96b5179d | 2662 | bit = BLOCK_GROUP_METADATA; |
31f3c99b | 2663 | } |
96b5179d | 2664 | |
6324fbf3 CM |
2665 | ret = update_space_info(info, cache->flags, found_key.offset, |
2666 | btrfs_block_group_used(&cache->item), | |
2667 | &space_info); | |
2668 | BUG_ON(ret); | |
2669 | cache->space_info = space_info; | |
2670 | ||
96b5179d CM |
2671 | /* use EXTENT_LOCKED to prevent merging */ |
2672 | set_extent_bits(block_group_cache, found_key.objectid, | |
2673 | found_key.objectid + found_key.offset - 1, | |
2674 | bit | EXTENT_LOCKED, GFP_NOFS); | |
2675 | set_state_private(block_group_cache, found_key.objectid, | |
ae2f5411 | 2676 | (unsigned long)cache); |
96b5179d | 2677 | |
9078a3e1 | 2678 | if (key.objectid >= |
db94535d | 2679 | btrfs_super_total_bytes(&info->super_copy)) |
9078a3e1 CM |
2680 | break; |
2681 | } | |
0b86a832 CM |
2682 | ret = 0; |
2683 | error: | |
9078a3e1 | 2684 | btrfs_free_path(path); |
0b86a832 | 2685 | return ret; |
9078a3e1 | 2686 | } |
6324fbf3 CM |
2687 | |
2688 | int btrfs_make_block_group(struct btrfs_trans_handle *trans, | |
2689 | struct btrfs_root *root, u64 bytes_used, | |
2690 | u64 type, u64 chunk_tree, u64 chunk_objectid, | |
2691 | u64 size) | |
2692 | { | |
2693 | int ret; | |
2694 | int bit = 0; | |
2695 | struct btrfs_root *extent_root; | |
2696 | struct btrfs_block_group_cache *cache; | |
2697 | struct extent_io_tree *block_group_cache; | |
2698 | ||
2699 | extent_root = root->fs_info->extent_root; | |
2700 | block_group_cache = &root->fs_info->block_group_cache; | |
2701 | ||
2702 | cache = kmalloc(sizeof(*cache), GFP_NOFS); | |
2703 | BUG_ON(!cache); | |
2704 | cache->key.objectid = chunk_objectid; | |
2705 | cache->key.offset = size; | |
2706 | cache->cached = 0; | |
2707 | cache->pinned = 0; | |
2708 | btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY); | |
2709 | memset(&cache->item, 0, sizeof(cache->item)); | |
2710 | btrfs_set_block_group_used(&cache->item, bytes_used); | |
2711 | btrfs_set_block_group_chunk_tree(&cache->item, chunk_tree); | |
2712 | btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid); | |
2713 | cache->flags = type; | |
2714 | btrfs_set_block_group_flags(&cache->item, type); | |
2715 | ||
2716 | ret = update_space_info(root->fs_info, cache->flags, size, bytes_used, | |
2717 | &cache->space_info); | |
2718 | BUG_ON(ret); | |
2719 | ||
2720 | if (type & BTRFS_BLOCK_GROUP_DATA) { | |
2721 | bit = BLOCK_GROUP_DATA; | |
2722 | } else if (type & BTRFS_BLOCK_GROUP_SYSTEM) { | |
2723 | bit = BLOCK_GROUP_SYSTEM; | |
2724 | } else if (type & BTRFS_BLOCK_GROUP_METADATA) { | |
2725 | bit = BLOCK_GROUP_METADATA; | |
2726 | } | |
2727 | set_extent_bits(block_group_cache, chunk_objectid, | |
2728 | chunk_objectid + size - 1, | |
2729 | bit | EXTENT_LOCKED, GFP_NOFS); | |
2730 | set_state_private(block_group_cache, chunk_objectid, | |
2731 | (unsigned long)cache); | |
2732 | ||
2733 | ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item, | |
2734 | sizeof(cache->item)); | |
2735 | BUG_ON(ret); | |
2736 | ||
2737 | finish_current_insert(trans, extent_root); | |
2738 | ret = del_pending_extents(trans, extent_root); | |
2739 | BUG_ON(ret); | |
2740 | return 0; | |
2741 | } |