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 | { | |
593060d7 | 175 | return (cache->flags & bits) == 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; | |
6324fbf3 | 376 | } |
de428b63 | 377 | cond_resched(); |
cd1bc465 | 378 | } |
31f3c99b | 379 | if (!full_search) { |
be744175 | 380 | last = search_start; |
31f3c99b CM |
381 | full_search = 1; |
382 | goto again; | |
383 | } | |
be744175 | 384 | found: |
31f3c99b | 385 | return found_group; |
cd1bc465 CM |
386 | } |
387 | ||
7bb86316 | 388 | static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation, |
74493f7a CM |
389 | u64 owner, u64 owner_offset) |
390 | { | |
391 | u32 high_crc = ~(u32)0; | |
392 | u32 low_crc = ~(u32)0; | |
393 | __le64 lenum; | |
394 | ||
395 | lenum = cpu_to_le64(root_objectid); | |
396 | high_crc = crc32c(high_crc, &lenum, sizeof(lenum)); | |
7bb86316 CM |
397 | lenum = cpu_to_le64(ref_generation); |
398 | low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); | |
21a4989d CM |
399 | if (owner >= BTRFS_FIRST_FREE_OBJECTID) { |
400 | lenum = cpu_to_le64(owner); | |
401 | low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); | |
402 | lenum = cpu_to_le64(owner_offset); | |
403 | low_crc = crc32c(low_crc, &lenum, sizeof(lenum)); | |
404 | } | |
74493f7a CM |
405 | return ((u64)high_crc << 32) | (u64)low_crc; |
406 | } | |
407 | ||
7bb86316 CM |
408 | static int match_extent_ref(struct extent_buffer *leaf, |
409 | struct btrfs_extent_ref *disk_ref, | |
410 | struct btrfs_extent_ref *cpu_ref) | |
411 | { | |
412 | int ret; | |
413 | int len; | |
414 | ||
415 | if (cpu_ref->objectid) | |
416 | len = sizeof(*cpu_ref); | |
417 | else | |
418 | len = 2 * sizeof(u64); | |
419 | ret = memcmp_extent_buffer(leaf, cpu_ref, (unsigned long)disk_ref, | |
420 | len); | |
421 | return ret == 0; | |
422 | } | |
423 | ||
98ed5174 CM |
424 | static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, |
425 | struct btrfs_root *root, | |
426 | struct btrfs_path *path, u64 bytenr, | |
427 | u64 root_objectid, | |
428 | u64 ref_generation, u64 owner, | |
429 | u64 owner_offset, int del) | |
74493f7a CM |
430 | { |
431 | u64 hash; | |
432 | struct btrfs_key key; | |
7bb86316 | 433 | struct btrfs_key found_key; |
74493f7a | 434 | struct btrfs_extent_ref ref; |
7bb86316 CM |
435 | struct extent_buffer *leaf; |
436 | struct btrfs_extent_ref *disk_ref; | |
437 | int ret; | |
438 | int ret2; | |
439 | ||
440 | btrfs_set_stack_ref_root(&ref, root_objectid); | |
441 | btrfs_set_stack_ref_generation(&ref, ref_generation); | |
442 | btrfs_set_stack_ref_objectid(&ref, owner); | |
443 | btrfs_set_stack_ref_offset(&ref, owner_offset); | |
444 | ||
445 | hash = hash_extent_ref(root_objectid, ref_generation, owner, | |
446 | owner_offset); | |
447 | key.offset = hash; | |
448 | key.objectid = bytenr; | |
449 | key.type = BTRFS_EXTENT_REF_KEY; | |
450 | ||
451 | while (1) { | |
452 | ret = btrfs_search_slot(trans, root, &key, path, | |
453 | del ? -1 : 0, del); | |
454 | if (ret < 0) | |
455 | goto out; | |
456 | leaf = path->nodes[0]; | |
457 | if (ret != 0) { | |
458 | u32 nritems = btrfs_header_nritems(leaf); | |
459 | if (path->slots[0] >= nritems) { | |
460 | ret2 = btrfs_next_leaf(root, path); | |
461 | if (ret2) | |
462 | goto out; | |
463 | leaf = path->nodes[0]; | |
464 | } | |
465 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
466 | if (found_key.objectid != bytenr || | |
467 | found_key.type != BTRFS_EXTENT_REF_KEY) | |
468 | goto out; | |
469 | key.offset = found_key.offset; | |
470 | if (del) { | |
471 | btrfs_release_path(root, path); | |
472 | continue; | |
473 | } | |
474 | } | |
475 | disk_ref = btrfs_item_ptr(path->nodes[0], | |
476 | path->slots[0], | |
477 | struct btrfs_extent_ref); | |
478 | if (match_extent_ref(path->nodes[0], disk_ref, &ref)) { | |
479 | ret = 0; | |
480 | goto out; | |
481 | } | |
482 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
483 | key.offset = found_key.offset + 1; | |
484 | btrfs_release_path(root, path); | |
485 | } | |
486 | out: | |
487 | return ret; | |
488 | } | |
489 | ||
d8d5f3e1 CM |
490 | /* |
491 | * Back reference rules. Back refs have three main goals: | |
492 | * | |
493 | * 1) differentiate between all holders of references to an extent so that | |
494 | * when a reference is dropped we can make sure it was a valid reference | |
495 | * before freeing the extent. | |
496 | * | |
497 | * 2) Provide enough information to quickly find the holders of an extent | |
498 | * if we notice a given block is corrupted or bad. | |
499 | * | |
500 | * 3) Make it easy to migrate blocks for FS shrinking or storage pool | |
501 | * maintenance. This is actually the same as #2, but with a slightly | |
502 | * different use case. | |
503 | * | |
504 | * File extents can be referenced by: | |
505 | * | |
506 | * - multiple snapshots, subvolumes, or different generations in one subvol | |
507 | * - different files inside a single subvolume (in theory, not implemented yet) | |
508 | * - different offsets inside a file (bookend extents in file.c) | |
509 | * | |
510 | * The extent ref structure has fields for: | |
511 | * | |
512 | * - Objectid of the subvolume root | |
513 | * - Generation number of the tree holding the reference | |
514 | * - objectid of the file holding the reference | |
515 | * - offset in the file corresponding to the key holding the reference | |
516 | * | |
517 | * When a file extent is allocated the fields are filled in: | |
518 | * (root_key.objectid, trans->transid, inode objectid, offset in file) | |
519 | * | |
520 | * When a leaf is cow'd new references are added for every file extent found | |
521 | * in the leaf. It looks the same as the create case, but trans->transid | |
522 | * will be different when the block is cow'd. | |
523 | * | |
524 | * (root_key.objectid, trans->transid, inode objectid, offset in file) | |
525 | * | |
526 | * When a file extent is removed either during snapshot deletion or file | |
527 | * truncation, the corresponding back reference is found | |
528 | * by searching for: | |
529 | * | |
530 | * (btrfs_header_owner(leaf), btrfs_header_generation(leaf), | |
531 | * inode objectid, offset in file) | |
532 | * | |
533 | * Btree extents can be referenced by: | |
534 | * | |
535 | * - Different subvolumes | |
536 | * - Different generations of the same subvolume | |
537 | * | |
538 | * Storing sufficient information for a full reverse mapping of a btree | |
539 | * block would require storing the lowest key of the block in the backref, | |
540 | * and it would require updating that lowest key either before write out or | |
541 | * every time it changed. Instead, the objectid of the lowest key is stored | |
542 | * along with the level of the tree block. This provides a hint | |
543 | * about where in the btree the block can be found. Searches through the | |
544 | * btree only need to look for a pointer to that block, so they stop one | |
545 | * level higher than the level recorded in the backref. | |
546 | * | |
547 | * Some btrees do not do reference counting on their extents. These | |
548 | * include the extent tree and the tree of tree roots. Backrefs for these | |
549 | * trees always have a generation of zero. | |
550 | * | |
551 | * When a tree block is created, back references are inserted: | |
552 | * | |
f6dbff55 | 553 | * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid) |
d8d5f3e1 CM |
554 | * |
555 | * When a tree block is cow'd in a reference counted root, | |
556 | * new back references are added for all the blocks it points to. | |
557 | * These are of the form (trans->transid will have increased since creation): | |
558 | * | |
f6dbff55 | 559 | * (root->root_key.objectid, trans->transid, level, lowest_key_objectid) |
d8d5f3e1 CM |
560 | * |
561 | * Because the lowest_key_objectid and the level are just hints | |
562 | * they are not used when backrefs are deleted. When a backref is deleted: | |
563 | * | |
564 | * if backref was for a tree root: | |
565 | * root_objectid = root->root_key.objectid | |
566 | * else | |
567 | * root_objectid = btrfs_header_owner(parent) | |
568 | * | |
569 | * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0) | |
570 | * | |
571 | * Back Reference Key hashing: | |
572 | * | |
573 | * Back references have four fields, each 64 bits long. Unfortunately, | |
574 | * This is hashed into a single 64 bit number and placed into the key offset. | |
575 | * The key objectid corresponds to the first byte in the extent, and the | |
576 | * key type is set to BTRFS_EXTENT_REF_KEY | |
577 | */ | |
7bb86316 CM |
578 | int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, |
579 | struct btrfs_root *root, | |
580 | struct btrfs_path *path, u64 bytenr, | |
581 | u64 root_objectid, u64 ref_generation, | |
582 | u64 owner, u64 owner_offset) | |
583 | { | |
584 | u64 hash; | |
585 | struct btrfs_key key; | |
586 | struct btrfs_extent_ref ref; | |
587 | struct btrfs_extent_ref *disk_ref; | |
74493f7a CM |
588 | int ret; |
589 | ||
590 | btrfs_set_stack_ref_root(&ref, root_objectid); | |
7bb86316 | 591 | btrfs_set_stack_ref_generation(&ref, ref_generation); |
74493f7a CM |
592 | btrfs_set_stack_ref_objectid(&ref, owner); |
593 | btrfs_set_stack_ref_offset(&ref, owner_offset); | |
594 | ||
7bb86316 CM |
595 | hash = hash_extent_ref(root_objectid, ref_generation, owner, |
596 | owner_offset); | |
74493f7a CM |
597 | key.offset = hash; |
598 | key.objectid = bytenr; | |
599 | key.type = BTRFS_EXTENT_REF_KEY; | |
600 | ||
601 | ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(ref)); | |
602 | while (ret == -EEXIST) { | |
7bb86316 CM |
603 | disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], |
604 | struct btrfs_extent_ref); | |
605 | if (match_extent_ref(path->nodes[0], disk_ref, &ref)) | |
606 | goto out; | |
607 | key.offset++; | |
608 | btrfs_release_path(root, path); | |
609 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
610 | sizeof(ref)); | |
74493f7a | 611 | } |
7bb86316 CM |
612 | if (ret) |
613 | goto out; | |
614 | disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
615 | struct btrfs_extent_ref); | |
616 | write_extent_buffer(path->nodes[0], &ref, (unsigned long)disk_ref, | |
617 | sizeof(ref)); | |
618 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
619 | out: | |
620 | btrfs_release_path(root, path); | |
621 | return ret; | |
74493f7a CM |
622 | } |
623 | ||
b18c6685 CM |
624 | int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, |
625 | struct btrfs_root *root, | |
74493f7a | 626 | u64 bytenr, u64 num_bytes, |
7bb86316 | 627 | u64 root_objectid, u64 ref_generation, |
74493f7a | 628 | u64 owner, u64 owner_offset) |
02217ed2 | 629 | { |
5caf2a00 | 630 | struct btrfs_path *path; |
02217ed2 | 631 | int ret; |
e2fa7227 | 632 | struct btrfs_key key; |
5f39d397 | 633 | struct extent_buffer *l; |
234b63a0 | 634 | struct btrfs_extent_item *item; |
cf27e1ee | 635 | u32 refs; |
037e6390 | 636 | |
db94535d | 637 | WARN_ON(num_bytes < root->sectorsize); |
5caf2a00 | 638 | path = btrfs_alloc_path(); |
54aa1f4d CM |
639 | if (!path) |
640 | return -ENOMEM; | |
26b8003f | 641 | |
b0331a4c | 642 | path->reada = 0; |
db94535d | 643 | key.objectid = bytenr; |
62e2749e | 644 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
db94535d | 645 | key.offset = num_bytes; |
5caf2a00 | 646 | ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, |
9f5fae2f | 647 | 0, 1); |
54aa1f4d CM |
648 | if (ret < 0) |
649 | return ret; | |
a429e513 | 650 | if (ret != 0) { |
a28ec197 | 651 | BUG(); |
a429e513 | 652 | } |
02217ed2 | 653 | BUG_ON(ret != 0); |
5f39d397 | 654 | l = path->nodes[0]; |
5caf2a00 | 655 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); |
5f39d397 CM |
656 | refs = btrfs_extent_refs(l, item); |
657 | btrfs_set_extent_refs(l, item, refs + 1); | |
5caf2a00 | 658 | btrfs_mark_buffer_dirty(path->nodes[0]); |
a28ec197 | 659 | |
5caf2a00 | 660 | btrfs_release_path(root->fs_info->extent_root, path); |
7bb86316 | 661 | |
b0331a4c | 662 | path->reada = 0; |
7bb86316 CM |
663 | ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root, |
664 | path, bytenr, root_objectid, | |
665 | ref_generation, owner, owner_offset); | |
666 | BUG_ON(ret); | |
9f5fae2f | 667 | finish_current_insert(trans, root->fs_info->extent_root); |
e20d96d6 | 668 | del_pending_extents(trans, root->fs_info->extent_root); |
74493f7a CM |
669 | |
670 | btrfs_free_path(path); | |
02217ed2 CM |
671 | return 0; |
672 | } | |
673 | ||
e9d0b13b CM |
674 | int btrfs_extent_post_op(struct btrfs_trans_handle *trans, |
675 | struct btrfs_root *root) | |
676 | { | |
677 | finish_current_insert(trans, root->fs_info->extent_root); | |
678 | del_pending_extents(trans, root->fs_info->extent_root); | |
679 | return 0; | |
680 | } | |
681 | ||
b18c6685 | 682 | static int lookup_extent_ref(struct btrfs_trans_handle *trans, |
db94535d CM |
683 | struct btrfs_root *root, u64 bytenr, |
684 | u64 num_bytes, u32 *refs) | |
a28ec197 | 685 | { |
5caf2a00 | 686 | struct btrfs_path *path; |
a28ec197 | 687 | int ret; |
e2fa7227 | 688 | struct btrfs_key key; |
5f39d397 | 689 | struct extent_buffer *l; |
234b63a0 | 690 | struct btrfs_extent_item *item; |
5caf2a00 | 691 | |
db94535d | 692 | WARN_ON(num_bytes < root->sectorsize); |
5caf2a00 | 693 | path = btrfs_alloc_path(); |
b0331a4c | 694 | path->reada = 0; |
db94535d CM |
695 | key.objectid = bytenr; |
696 | key.offset = num_bytes; | |
62e2749e | 697 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
5caf2a00 | 698 | ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, |
9f5fae2f | 699 | 0, 0); |
54aa1f4d CM |
700 | if (ret < 0) |
701 | goto out; | |
5f39d397 CM |
702 | if (ret != 0) { |
703 | btrfs_print_leaf(root, path->nodes[0]); | |
db94535d | 704 | printk("failed to find block number %Lu\n", bytenr); |
a28ec197 | 705 | BUG(); |
5f39d397 CM |
706 | } |
707 | l = path->nodes[0]; | |
5caf2a00 | 708 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); |
5f39d397 | 709 | *refs = btrfs_extent_refs(l, item); |
54aa1f4d | 710 | out: |
5caf2a00 | 711 | btrfs_free_path(path); |
a28ec197 CM |
712 | return 0; |
713 | } | |
714 | ||
be20aa9d CM |
715 | u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, |
716 | struct btrfs_path *count_path, | |
717 | u64 first_extent) | |
718 | { | |
719 | struct btrfs_root *extent_root = root->fs_info->extent_root; | |
720 | struct btrfs_path *path; | |
721 | u64 bytenr; | |
722 | u64 found_objectid; | |
56b453c9 | 723 | u64 root_objectid = root->root_key.objectid; |
be20aa9d CM |
724 | u32 total_count = 0; |
725 | u32 cur_count; | |
be20aa9d CM |
726 | u32 nritems; |
727 | int ret; | |
728 | struct btrfs_key key; | |
729 | struct btrfs_key found_key; | |
730 | struct extent_buffer *l; | |
731 | struct btrfs_extent_item *item; | |
732 | struct btrfs_extent_ref *ref_item; | |
733 | int level = -1; | |
734 | ||
735 | path = btrfs_alloc_path(); | |
736 | again: | |
737 | if (level == -1) | |
738 | bytenr = first_extent; | |
739 | else | |
740 | bytenr = count_path->nodes[level]->start; | |
741 | ||
742 | cur_count = 0; | |
743 | key.objectid = bytenr; | |
744 | key.offset = 0; | |
745 | ||
746 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); | |
747 | ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); | |
748 | if (ret < 0) | |
749 | goto out; | |
750 | BUG_ON(ret == 0); | |
751 | ||
752 | l = path->nodes[0]; | |
753 | btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); | |
754 | ||
755 | if (found_key.objectid != bytenr || | |
756 | found_key.type != BTRFS_EXTENT_ITEM_KEY) { | |
757 | goto out; | |
758 | } | |
759 | ||
760 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); | |
be20aa9d | 761 | while (1) { |
bd09835d | 762 | l = path->nodes[0]; |
be20aa9d CM |
763 | nritems = btrfs_header_nritems(l); |
764 | if (path->slots[0] >= nritems) { | |
765 | ret = btrfs_next_leaf(extent_root, path); | |
766 | if (ret == 0) | |
767 | continue; | |
768 | break; | |
769 | } | |
770 | btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); | |
771 | if (found_key.objectid != bytenr) | |
772 | break; | |
bd09835d | 773 | |
be20aa9d CM |
774 | if (found_key.type != BTRFS_EXTENT_REF_KEY) { |
775 | path->slots[0]++; | |
776 | continue; | |
777 | } | |
778 | ||
779 | cur_count++; | |
780 | ref_item = btrfs_item_ptr(l, path->slots[0], | |
781 | struct btrfs_extent_ref); | |
782 | found_objectid = btrfs_ref_root(l, ref_item); | |
783 | ||
56b453c9 CM |
784 | if (found_objectid != root_objectid) { |
785 | total_count = 2; | |
4313b399 | 786 | goto out; |
56b453c9 CM |
787 | } |
788 | total_count = 1; | |
be20aa9d CM |
789 | path->slots[0]++; |
790 | } | |
791 | if (cur_count == 0) { | |
792 | total_count = 0; | |
793 | goto out; | |
794 | } | |
be20aa9d CM |
795 | if (level >= 0 && root->node == count_path->nodes[level]) |
796 | goto out; | |
797 | level++; | |
798 | btrfs_release_path(root, path); | |
799 | goto again; | |
800 | ||
801 | out: | |
802 | btrfs_free_path(path); | |
803 | return total_count; | |
be20aa9d | 804 | } |
c5739bba | 805 | int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, |
7bb86316 | 806 | struct btrfs_root *root, u64 owner_objectid) |
c5739bba | 807 | { |
7bb86316 CM |
808 | u64 generation; |
809 | u64 key_objectid; | |
810 | u64 level; | |
811 | u32 nritems; | |
812 | struct btrfs_disk_key disk_key; | |
813 | ||
814 | level = btrfs_header_level(root->node); | |
815 | generation = trans->transid; | |
816 | nritems = btrfs_header_nritems(root->node); | |
817 | if (nritems > 0) { | |
818 | if (level == 0) | |
819 | btrfs_item_key(root->node, &disk_key, 0); | |
820 | else | |
821 | btrfs_node_key(root->node, &disk_key, 0); | |
822 | key_objectid = btrfs_disk_key_objectid(&disk_key); | |
823 | } else { | |
824 | key_objectid = 0; | |
825 | } | |
db94535d | 826 | return btrfs_inc_extent_ref(trans, root, root->node->start, |
7bb86316 | 827 | root->node->len, owner_objectid, |
f6dbff55 | 828 | generation, level, key_objectid); |
c5739bba CM |
829 | } |
830 | ||
e089f05c | 831 | int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 832 | struct extent_buffer *buf) |
02217ed2 | 833 | { |
db94535d | 834 | u64 bytenr; |
5f39d397 CM |
835 | u32 nritems; |
836 | struct btrfs_key key; | |
6407bf6d | 837 | struct btrfs_file_extent_item *fi; |
02217ed2 | 838 | int i; |
db94535d | 839 | int level; |
6407bf6d | 840 | int ret; |
54aa1f4d | 841 | int faili; |
a28ec197 | 842 | |
3768f368 | 843 | if (!root->ref_cows) |
a28ec197 | 844 | return 0; |
5f39d397 | 845 | |
db94535d | 846 | level = btrfs_header_level(buf); |
5f39d397 CM |
847 | nritems = btrfs_header_nritems(buf); |
848 | for (i = 0; i < nritems; i++) { | |
db94535d CM |
849 | if (level == 0) { |
850 | u64 disk_bytenr; | |
5f39d397 CM |
851 | btrfs_item_key_to_cpu(buf, &key, i); |
852 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
6407bf6d | 853 | continue; |
5f39d397 | 854 | fi = btrfs_item_ptr(buf, i, |
6407bf6d | 855 | struct btrfs_file_extent_item); |
5f39d397 | 856 | if (btrfs_file_extent_type(buf, fi) == |
236454df CM |
857 | BTRFS_FILE_EXTENT_INLINE) |
858 | continue; | |
db94535d CM |
859 | disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); |
860 | if (disk_bytenr == 0) | |
3a686375 | 861 | continue; |
db94535d | 862 | ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, |
7bb86316 CM |
863 | btrfs_file_extent_disk_num_bytes(buf, fi), |
864 | root->root_key.objectid, trans->transid, | |
865 | key.objectid, key.offset); | |
54aa1f4d CM |
866 | if (ret) { |
867 | faili = i; | |
868 | goto fail; | |
869 | } | |
6407bf6d | 870 | } else { |
db94535d | 871 | bytenr = btrfs_node_blockptr(buf, i); |
6caab489 | 872 | btrfs_node_key_to_cpu(buf, &key, i); |
db94535d | 873 | ret = btrfs_inc_extent_ref(trans, root, bytenr, |
7bb86316 CM |
874 | btrfs_level_size(root, level - 1), |
875 | root->root_key.objectid, | |
f6dbff55 CM |
876 | trans->transid, |
877 | level - 1, key.objectid); | |
54aa1f4d CM |
878 | if (ret) { |
879 | faili = i; | |
880 | goto fail; | |
881 | } | |
6407bf6d | 882 | } |
02217ed2 CM |
883 | } |
884 | return 0; | |
54aa1f4d | 885 | fail: |
ccd467d6 | 886 | WARN_ON(1); |
7bb86316 | 887 | #if 0 |
54aa1f4d | 888 | for (i =0; i < faili; i++) { |
db94535d CM |
889 | if (level == 0) { |
890 | u64 disk_bytenr; | |
5f39d397 CM |
891 | btrfs_item_key_to_cpu(buf, &key, i); |
892 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
54aa1f4d | 893 | continue; |
5f39d397 | 894 | fi = btrfs_item_ptr(buf, i, |
54aa1f4d | 895 | struct btrfs_file_extent_item); |
5f39d397 | 896 | if (btrfs_file_extent_type(buf, fi) == |
54aa1f4d CM |
897 | BTRFS_FILE_EXTENT_INLINE) |
898 | continue; | |
db94535d CM |
899 | disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); |
900 | if (disk_bytenr == 0) | |
54aa1f4d | 901 | continue; |
db94535d CM |
902 | err = btrfs_free_extent(trans, root, disk_bytenr, |
903 | btrfs_file_extent_disk_num_bytes(buf, | |
5f39d397 | 904 | fi), 0); |
54aa1f4d CM |
905 | BUG_ON(err); |
906 | } else { | |
db94535d CM |
907 | bytenr = btrfs_node_blockptr(buf, i); |
908 | err = btrfs_free_extent(trans, root, bytenr, | |
909 | btrfs_level_size(root, level - 1), 0); | |
54aa1f4d CM |
910 | BUG_ON(err); |
911 | } | |
912 | } | |
7bb86316 | 913 | #endif |
54aa1f4d | 914 | return ret; |
02217ed2 CM |
915 | } |
916 | ||
9078a3e1 CM |
917 | static int write_one_cache_group(struct btrfs_trans_handle *trans, |
918 | struct btrfs_root *root, | |
919 | struct btrfs_path *path, | |
920 | struct btrfs_block_group_cache *cache) | |
921 | { | |
922 | int ret; | |
923 | int pending_ret; | |
924 | struct btrfs_root *extent_root = root->fs_info->extent_root; | |
5f39d397 CM |
925 | unsigned long bi; |
926 | struct extent_buffer *leaf; | |
9078a3e1 | 927 | |
9078a3e1 | 928 | ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1); |
54aa1f4d CM |
929 | if (ret < 0) |
930 | goto fail; | |
9078a3e1 | 931 | BUG_ON(ret); |
5f39d397 CM |
932 | |
933 | leaf = path->nodes[0]; | |
934 | bi = btrfs_item_ptr_offset(leaf, path->slots[0]); | |
935 | write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item)); | |
936 | btrfs_mark_buffer_dirty(leaf); | |
9078a3e1 | 937 | btrfs_release_path(extent_root, path); |
54aa1f4d | 938 | fail: |
9078a3e1 CM |
939 | finish_current_insert(trans, extent_root); |
940 | pending_ret = del_pending_extents(trans, extent_root); | |
941 | if (ret) | |
942 | return ret; | |
943 | if (pending_ret) | |
944 | return pending_ret; | |
945 | return 0; | |
946 | ||
947 | } | |
948 | ||
96b5179d CM |
949 | int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, |
950 | struct btrfs_root *root) | |
9078a3e1 | 951 | { |
d1310b2e | 952 | struct extent_io_tree *block_group_cache; |
96b5179d | 953 | struct btrfs_block_group_cache *cache; |
9078a3e1 CM |
954 | int ret; |
955 | int err = 0; | |
956 | int werr = 0; | |
9078a3e1 | 957 | struct btrfs_path *path; |
96b5179d CM |
958 | u64 last = 0; |
959 | u64 start; | |
960 | u64 end; | |
961 | u64 ptr; | |
9078a3e1 | 962 | |
96b5179d | 963 | block_group_cache = &root->fs_info->block_group_cache; |
9078a3e1 CM |
964 | path = btrfs_alloc_path(); |
965 | if (!path) | |
966 | return -ENOMEM; | |
967 | ||
968 | while(1) { | |
96b5179d CM |
969 | ret = find_first_extent_bit(block_group_cache, last, |
970 | &start, &end, BLOCK_GROUP_DIRTY); | |
971 | if (ret) | |
9078a3e1 | 972 | break; |
54aa1f4d | 973 | |
96b5179d CM |
974 | last = end + 1; |
975 | ret = get_state_private(block_group_cache, start, &ptr); | |
976 | if (ret) | |
977 | break; | |
978 | ||
ae2f5411 | 979 | cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; |
96b5179d CM |
980 | err = write_one_cache_group(trans, root, |
981 | path, cache); | |
982 | /* | |
983 | * if we fail to write the cache group, we want | |
984 | * to keep it marked dirty in hopes that a later | |
985 | * write will work | |
986 | */ | |
987 | if (err) { | |
988 | werr = err; | |
989 | continue; | |
9078a3e1 | 990 | } |
96b5179d CM |
991 | clear_extent_bits(block_group_cache, start, end, |
992 | BLOCK_GROUP_DIRTY, GFP_NOFS); | |
9078a3e1 CM |
993 | } |
994 | btrfs_free_path(path); | |
995 | return werr; | |
996 | } | |
997 | ||
6324fbf3 CM |
998 | static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info, |
999 | u64 flags) | |
1000 | { | |
1001 | struct list_head *head = &info->space_info; | |
1002 | struct list_head *cur; | |
1003 | struct btrfs_space_info *found; | |
1004 | list_for_each(cur, head) { | |
1005 | found = list_entry(cur, struct btrfs_space_info, list); | |
1006 | if (found->flags == flags) | |
1007 | return found; | |
1008 | } | |
1009 | return NULL; | |
1010 | ||
1011 | } | |
1012 | ||
593060d7 CM |
1013 | static int update_space_info(struct btrfs_fs_info *info, u64 flags, |
1014 | u64 total_bytes, u64 bytes_used, | |
1015 | struct btrfs_space_info **space_info) | |
1016 | { | |
1017 | struct btrfs_space_info *found; | |
1018 | ||
1019 | found = __find_space_info(info, flags); | |
1020 | if (found) { | |
1021 | found->total_bytes += total_bytes; | |
1022 | found->bytes_used += bytes_used; | |
1023 | WARN_ON(found->total_bytes < found->bytes_used); | |
1024 | *space_info = found; | |
1025 | return 0; | |
1026 | } | |
1027 | found = kmalloc(sizeof(*found), GFP_NOFS); | |
1028 | if (!found) | |
1029 | return -ENOMEM; | |
1030 | ||
1031 | list_add(&found->list, &info->space_info); | |
1032 | found->flags = flags; | |
1033 | found->total_bytes = total_bytes; | |
1034 | found->bytes_used = bytes_used; | |
1035 | found->bytes_pinned = 0; | |
1036 | found->full = 0; | |
1037 | *space_info = found; | |
1038 | return 0; | |
1039 | } | |
1040 | ||
1041 | ||
6324fbf3 CM |
1042 | static int do_chunk_alloc(struct btrfs_trans_handle *trans, |
1043 | struct btrfs_root *extent_root, u64 alloc_bytes, | |
1044 | u64 flags) | |
1045 | { | |
1046 | struct btrfs_space_info *space_info; | |
1047 | u64 thresh; | |
1048 | u64 start; | |
1049 | u64 num_bytes; | |
1050 | int ret; | |
1051 | ||
1052 | space_info = __find_space_info(extent_root->fs_info, flags); | |
593060d7 CM |
1053 | if (!space_info) { |
1054 | ret = update_space_info(extent_root->fs_info, flags, | |
1055 | 0, 0, &space_info); | |
1056 | BUG_ON(ret); | |
1057 | } | |
6324fbf3 CM |
1058 | BUG_ON(!space_info); |
1059 | ||
1060 | if (space_info->full) | |
1061 | return 0; | |
1062 | ||
1063 | thresh = div_factor(space_info->total_bytes, 7); | |
1064 | if ((space_info->bytes_used + space_info->bytes_pinned + alloc_bytes) < | |
1065 | thresh) | |
1066 | return 0; | |
1067 | ||
1068 | ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags); | |
1069 | if (ret == -ENOSPC) { | |
1070 | printk("space info full %Lu\n", flags); | |
1071 | space_info->full = 1; | |
1072 | return 0; | |
1073 | } | |
1074 | ||
1075 | BUG_ON(ret); | |
1076 | ||
1077 | ret = btrfs_make_block_group(trans, extent_root, 0, flags, | |
1078 | extent_root->fs_info->chunk_root->root_key.objectid, | |
1079 | start, num_bytes); | |
1080 | BUG_ON(ret); | |
593060d7 CM |
1081 | |
1082 | if (flags & BTRFS_BLOCK_GROUP_RAID0) { | |
1083 | if (flags & BTRFS_BLOCK_GROUP_DATA) { | |
1084 | extent_root->fs_info->extra_data_alloc_bits = | |
1085 | BTRFS_BLOCK_GROUP_RAID0; | |
1086 | } | |
1087 | if (flags & BTRFS_BLOCK_GROUP_METADATA) { | |
1088 | extent_root->fs_info->extra_alloc_bits = | |
1089 | BTRFS_BLOCK_GROUP_RAID0; | |
1090 | } | |
1091 | } | |
6324fbf3 CM |
1092 | return 0; |
1093 | } | |
1094 | ||
9078a3e1 CM |
1095 | static int update_block_group(struct btrfs_trans_handle *trans, |
1096 | struct btrfs_root *root, | |
db94535d | 1097 | u64 bytenr, u64 num_bytes, int alloc, |
0b86a832 | 1098 | int mark_free) |
9078a3e1 CM |
1099 | { |
1100 | struct btrfs_block_group_cache *cache; | |
1101 | struct btrfs_fs_info *info = root->fs_info; | |
db94535d | 1102 | u64 total = num_bytes; |
9078a3e1 | 1103 | u64 old_val; |
db94535d | 1104 | u64 byte_in_group; |
96b5179d CM |
1105 | u64 start; |
1106 | u64 end; | |
3e1ad54f | 1107 | |
9078a3e1 | 1108 | while(total) { |
db94535d | 1109 | cache = btrfs_lookup_block_group(info, bytenr); |
3e1ad54f | 1110 | if (!cache) { |
9078a3e1 | 1111 | return -1; |
cd1bc465 | 1112 | } |
db94535d CM |
1113 | byte_in_group = bytenr - cache->key.objectid; |
1114 | WARN_ON(byte_in_group > cache->key.offset); | |
96b5179d CM |
1115 | start = cache->key.objectid; |
1116 | end = start + cache->key.offset - 1; | |
1117 | set_extent_bits(&info->block_group_cache, start, end, | |
1118 | BLOCK_GROUP_DIRTY, GFP_NOFS); | |
9078a3e1 CM |
1119 | |
1120 | old_val = btrfs_block_group_used(&cache->item); | |
db94535d | 1121 | num_bytes = min(total, cache->key.offset - byte_in_group); |
cd1bc465 | 1122 | if (alloc) { |
db94535d | 1123 | old_val += num_bytes; |
6324fbf3 | 1124 | cache->space_info->bytes_used += num_bytes; |
cd1bc465 | 1125 | } else { |
db94535d | 1126 | old_val -= num_bytes; |
6324fbf3 | 1127 | cache->space_info->bytes_used -= num_bytes; |
f510cfec CM |
1128 | if (mark_free) { |
1129 | set_extent_dirty(&info->free_space_cache, | |
db94535d | 1130 | bytenr, bytenr + num_bytes - 1, |
f510cfec | 1131 | GFP_NOFS); |
e37c9e69 | 1132 | } |
cd1bc465 | 1133 | } |
9078a3e1 | 1134 | btrfs_set_block_group_used(&cache->item, old_val); |
db94535d CM |
1135 | total -= num_bytes; |
1136 | bytenr += num_bytes; | |
9078a3e1 CM |
1137 | } |
1138 | return 0; | |
1139 | } | |
6324fbf3 | 1140 | |
324ae4df Y |
1141 | static int update_pinned_extents(struct btrfs_root *root, |
1142 | u64 bytenr, u64 num, int pin) | |
1143 | { | |
1144 | u64 len; | |
1145 | struct btrfs_block_group_cache *cache; | |
1146 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1147 | ||
1148 | if (pin) { | |
1149 | set_extent_dirty(&fs_info->pinned_extents, | |
1150 | bytenr, bytenr + num - 1, GFP_NOFS); | |
1151 | } else { | |
1152 | clear_extent_dirty(&fs_info->pinned_extents, | |
1153 | bytenr, bytenr + num - 1, GFP_NOFS); | |
1154 | } | |
1155 | while (num > 0) { | |
1156 | cache = btrfs_lookup_block_group(fs_info, bytenr); | |
1157 | WARN_ON(!cache); | |
1158 | len = min(num, cache->key.offset - | |
1159 | (bytenr - cache->key.objectid)); | |
1160 | if (pin) { | |
1161 | cache->pinned += len; | |
6324fbf3 | 1162 | cache->space_info->bytes_pinned += len; |
324ae4df Y |
1163 | fs_info->total_pinned += len; |
1164 | } else { | |
1165 | cache->pinned -= len; | |
6324fbf3 | 1166 | cache->space_info->bytes_pinned -= len; |
324ae4df Y |
1167 | fs_info->total_pinned -= len; |
1168 | } | |
1169 | bytenr += len; | |
1170 | num -= len; | |
1171 | } | |
1172 | return 0; | |
1173 | } | |
9078a3e1 | 1174 | |
d1310b2e | 1175 | int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy) |
ccd467d6 | 1176 | { |
ccd467d6 | 1177 | u64 last = 0; |
1a5bc167 CM |
1178 | u64 start; |
1179 | u64 end; | |
d1310b2e | 1180 | struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents; |
ccd467d6 | 1181 | int ret; |
ccd467d6 CM |
1182 | |
1183 | while(1) { | |
1a5bc167 CM |
1184 | ret = find_first_extent_bit(pinned_extents, last, |
1185 | &start, &end, EXTENT_DIRTY); | |
1186 | if (ret) | |
ccd467d6 | 1187 | break; |
1a5bc167 CM |
1188 | set_extent_dirty(copy, start, end, GFP_NOFS); |
1189 | last = end + 1; | |
ccd467d6 CM |
1190 | } |
1191 | return 0; | |
1192 | } | |
1193 | ||
1194 | int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, | |
1195 | struct btrfs_root *root, | |
d1310b2e | 1196 | struct extent_io_tree *unpin) |
a28ec197 | 1197 | { |
1a5bc167 CM |
1198 | u64 start; |
1199 | u64 end; | |
a28ec197 | 1200 | int ret; |
d1310b2e | 1201 | struct extent_io_tree *free_space_cache; |
f510cfec | 1202 | free_space_cache = &root->fs_info->free_space_cache; |
a28ec197 CM |
1203 | |
1204 | while(1) { | |
1a5bc167 CM |
1205 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
1206 | EXTENT_DIRTY); | |
1207 | if (ret) | |
a28ec197 | 1208 | break; |
324ae4df | 1209 | update_pinned_extents(root, start, end + 1 - start, 0); |
1a5bc167 CM |
1210 | clear_extent_dirty(unpin, start, end, GFP_NOFS); |
1211 | set_extent_dirty(free_space_cache, start, end, GFP_NOFS); | |
a28ec197 CM |
1212 | } |
1213 | return 0; | |
1214 | } | |
1215 | ||
98ed5174 CM |
1216 | static int finish_current_insert(struct btrfs_trans_handle *trans, |
1217 | struct btrfs_root *extent_root) | |
037e6390 | 1218 | { |
7bb86316 CM |
1219 | u64 start; |
1220 | u64 end; | |
1221 | struct btrfs_fs_info *info = extent_root->fs_info; | |
d8d5f3e1 | 1222 | struct extent_buffer *eb; |
7bb86316 | 1223 | struct btrfs_path *path; |
e2fa7227 | 1224 | struct btrfs_key ins; |
d8d5f3e1 | 1225 | struct btrfs_disk_key first; |
234b63a0 | 1226 | struct btrfs_extent_item extent_item; |
037e6390 | 1227 | int ret; |
d8d5f3e1 | 1228 | int level; |
1a5bc167 | 1229 | int err = 0; |
037e6390 | 1230 | |
5f39d397 | 1231 | btrfs_set_stack_extent_refs(&extent_item, 1); |
62e2749e | 1232 | btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY); |
7bb86316 | 1233 | path = btrfs_alloc_path(); |
037e6390 | 1234 | |
26b8003f | 1235 | while(1) { |
1a5bc167 CM |
1236 | ret = find_first_extent_bit(&info->extent_ins, 0, &start, |
1237 | &end, EXTENT_LOCKED); | |
1238 | if (ret) | |
26b8003f CM |
1239 | break; |
1240 | ||
1a5bc167 CM |
1241 | ins.objectid = start; |
1242 | ins.offset = end + 1 - start; | |
1243 | err = btrfs_insert_item(trans, extent_root, &ins, | |
1244 | &extent_item, sizeof(extent_item)); | |
1245 | clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED, | |
1246 | GFP_NOFS); | |
d8d5f3e1 CM |
1247 | eb = read_tree_block(extent_root, ins.objectid, ins.offset); |
1248 | level = btrfs_header_level(eb); | |
1249 | if (level == 0) { | |
1250 | btrfs_item_key(eb, &first, 0); | |
1251 | } else { | |
1252 | btrfs_node_key(eb, &first, 0); | |
1253 | } | |
7bb86316 CM |
1254 | err = btrfs_insert_extent_backref(trans, extent_root, path, |
1255 | start, extent_root->root_key.objectid, | |
f6dbff55 CM |
1256 | 0, level, |
1257 | btrfs_disk_key_objectid(&first)); | |
7bb86316 | 1258 | BUG_ON(err); |
d8d5f3e1 | 1259 | free_extent_buffer(eb); |
037e6390 | 1260 | } |
7bb86316 | 1261 | btrfs_free_path(path); |
037e6390 CM |
1262 | return 0; |
1263 | } | |
1264 | ||
db94535d CM |
1265 | static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, |
1266 | int pending) | |
e20d96d6 | 1267 | { |
1a5bc167 | 1268 | int err = 0; |
5f39d397 | 1269 | struct extent_buffer *buf; |
8ef97622 | 1270 | |
f4b9aa8d | 1271 | if (!pending) { |
db94535d | 1272 | buf = btrfs_find_tree_block(root, bytenr, num_bytes); |
5f39d397 CM |
1273 | if (buf) { |
1274 | if (btrfs_buffer_uptodate(buf)) { | |
2c90e5d6 CM |
1275 | u64 transid = |
1276 | root->fs_info->running_transaction->transid; | |
dc17ff8f CM |
1277 | u64 header_transid = |
1278 | btrfs_header_generation(buf); | |
1279 | if (header_transid == transid) { | |
55c69072 | 1280 | clean_tree_block(NULL, root, buf); |
5f39d397 | 1281 | free_extent_buffer(buf); |
c549228f | 1282 | return 1; |
2c90e5d6 | 1283 | } |
f4b9aa8d | 1284 | } |
5f39d397 | 1285 | free_extent_buffer(buf); |
8ef97622 | 1286 | } |
324ae4df | 1287 | update_pinned_extents(root, bytenr, num_bytes, 1); |
f4b9aa8d | 1288 | } else { |
1a5bc167 | 1289 | set_extent_bits(&root->fs_info->pending_del, |
db94535d CM |
1290 | bytenr, bytenr + num_bytes - 1, |
1291 | EXTENT_LOCKED, GFP_NOFS); | |
f4b9aa8d | 1292 | } |
be744175 | 1293 | BUG_ON(err < 0); |
e20d96d6 CM |
1294 | return 0; |
1295 | } | |
1296 | ||
fec577fb | 1297 | /* |
a28ec197 | 1298 | * remove an extent from the root, returns 0 on success |
fec577fb | 1299 | */ |
e089f05c | 1300 | static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root |
7bb86316 CM |
1301 | *root, u64 bytenr, u64 num_bytes, |
1302 | u64 root_objectid, u64 ref_generation, | |
1303 | u64 owner_objectid, u64 owner_offset, int pin, | |
e37c9e69 | 1304 | int mark_free) |
a28ec197 | 1305 | { |
5caf2a00 | 1306 | struct btrfs_path *path; |
e2fa7227 | 1307 | struct btrfs_key key; |
1261ec42 CM |
1308 | struct btrfs_fs_info *info = root->fs_info; |
1309 | struct btrfs_root *extent_root = info->extent_root; | |
5f39d397 | 1310 | struct extent_buffer *leaf; |
a28ec197 | 1311 | int ret; |
952fccac CM |
1312 | int extent_slot = 0; |
1313 | int found_extent = 0; | |
1314 | int num_to_del = 1; | |
234b63a0 | 1315 | struct btrfs_extent_item *ei; |
cf27e1ee | 1316 | u32 refs; |
037e6390 | 1317 | |
db94535d | 1318 | key.objectid = bytenr; |
62e2749e | 1319 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
db94535d | 1320 | key.offset = num_bytes; |
5caf2a00 | 1321 | path = btrfs_alloc_path(); |
54aa1f4d CM |
1322 | if (!path) |
1323 | return -ENOMEM; | |
5f26f772 | 1324 | |
b0331a4c | 1325 | path->reada = 0; |
7bb86316 CM |
1326 | ret = lookup_extent_backref(trans, extent_root, path, |
1327 | bytenr, root_objectid, | |
1328 | ref_generation, | |
1329 | owner_objectid, owner_offset, 1); | |
1330 | if (ret == 0) { | |
952fccac CM |
1331 | struct btrfs_key found_key; |
1332 | extent_slot = path->slots[0]; | |
1333 | while(extent_slot > 0) { | |
1334 | extent_slot--; | |
1335 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
1336 | extent_slot); | |
1337 | if (found_key.objectid != bytenr) | |
1338 | break; | |
1339 | if (found_key.type == BTRFS_EXTENT_ITEM_KEY && | |
1340 | found_key.offset == num_bytes) { | |
1341 | found_extent = 1; | |
1342 | break; | |
1343 | } | |
1344 | if (path->slots[0] - extent_slot > 5) | |
1345 | break; | |
1346 | } | |
1347 | if (!found_extent) | |
1348 | ret = btrfs_del_item(trans, extent_root, path); | |
7bb86316 CM |
1349 | } else { |
1350 | btrfs_print_leaf(extent_root, path->nodes[0]); | |
1351 | WARN_ON(1); | |
1352 | printk("Unable to find ref byte nr %Lu root %Lu " | |
1353 | " gen %Lu owner %Lu offset %Lu\n", bytenr, | |
1354 | root_objectid, ref_generation, owner_objectid, | |
1355 | owner_offset); | |
1356 | } | |
952fccac CM |
1357 | if (!found_extent) { |
1358 | btrfs_release_path(extent_root, path); | |
1359 | ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); | |
1360 | if (ret < 0) | |
1361 | return ret; | |
1362 | BUG_ON(ret); | |
1363 | extent_slot = path->slots[0]; | |
1364 | } | |
5f39d397 CM |
1365 | |
1366 | leaf = path->nodes[0]; | |
952fccac | 1367 | ei = btrfs_item_ptr(leaf, extent_slot, |
123abc88 | 1368 | struct btrfs_extent_item); |
5f39d397 CM |
1369 | refs = btrfs_extent_refs(leaf, ei); |
1370 | BUG_ON(refs == 0); | |
1371 | refs -= 1; | |
1372 | btrfs_set_extent_refs(leaf, ei, refs); | |
952fccac | 1373 | |
5f39d397 CM |
1374 | btrfs_mark_buffer_dirty(leaf); |
1375 | ||
952fccac CM |
1376 | if (refs == 0 && found_extent && path->slots[0] == extent_slot + 1) { |
1377 | /* if the back ref and the extent are next to each other | |
1378 | * they get deleted below in one shot | |
1379 | */ | |
1380 | path->slots[0] = extent_slot; | |
1381 | num_to_del = 2; | |
1382 | } else if (found_extent) { | |
1383 | /* otherwise delete the extent back ref */ | |
1384 | ret = btrfs_del_item(trans, extent_root, path); | |
1385 | BUG_ON(ret); | |
1386 | /* if refs are 0, we need to setup the path for deletion */ | |
1387 | if (refs == 0) { | |
1388 | btrfs_release_path(extent_root, path); | |
1389 | ret = btrfs_search_slot(trans, extent_root, &key, path, | |
1390 | -1, 1); | |
1391 | if (ret < 0) | |
1392 | return ret; | |
1393 | BUG_ON(ret); | |
1394 | } | |
1395 | } | |
1396 | ||
cf27e1ee | 1397 | if (refs == 0) { |
db94535d CM |
1398 | u64 super_used; |
1399 | u64 root_used; | |
78fae27e CM |
1400 | |
1401 | if (pin) { | |
db94535d | 1402 | ret = pin_down_bytes(root, bytenr, num_bytes, 0); |
c549228f Y |
1403 | if (ret > 0) |
1404 | mark_free = 1; | |
1405 | BUG_ON(ret < 0); | |
78fae27e CM |
1406 | } |
1407 | ||
58176a96 | 1408 | /* block accounting for super block */ |
db94535d CM |
1409 | super_used = btrfs_super_bytes_used(&info->super_copy); |
1410 | btrfs_set_super_bytes_used(&info->super_copy, | |
1411 | super_used - num_bytes); | |
58176a96 JB |
1412 | |
1413 | /* block accounting for root item */ | |
db94535d | 1414 | root_used = btrfs_root_used(&root->root_item); |
5f39d397 | 1415 | btrfs_set_root_used(&root->root_item, |
db94535d | 1416 | root_used - num_bytes); |
952fccac CM |
1417 | ret = btrfs_del_items(trans, extent_root, path, path->slots[0], |
1418 | num_to_del); | |
54aa1f4d CM |
1419 | if (ret) { |
1420 | return ret; | |
1421 | } | |
db94535d | 1422 | ret = update_block_group(trans, root, bytenr, num_bytes, 0, |
0b86a832 | 1423 | mark_free); |
9078a3e1 | 1424 | BUG_ON(ret); |
a28ec197 | 1425 | } |
5caf2a00 | 1426 | btrfs_free_path(path); |
e089f05c | 1427 | finish_current_insert(trans, extent_root); |
a28ec197 CM |
1428 | return ret; |
1429 | } | |
1430 | ||
a28ec197 CM |
1431 | /* |
1432 | * find all the blocks marked as pending in the radix tree and remove | |
1433 | * them from the extent map | |
1434 | */ | |
e089f05c CM |
1435 | static int del_pending_extents(struct btrfs_trans_handle *trans, struct |
1436 | btrfs_root *extent_root) | |
a28ec197 CM |
1437 | { |
1438 | int ret; | |
e20d96d6 | 1439 | int err = 0; |
1a5bc167 CM |
1440 | u64 start; |
1441 | u64 end; | |
d1310b2e CM |
1442 | struct extent_io_tree *pending_del; |
1443 | struct extent_io_tree *pinned_extents; | |
8ef97622 | 1444 | |
1a5bc167 CM |
1445 | pending_del = &extent_root->fs_info->pending_del; |
1446 | pinned_extents = &extent_root->fs_info->pinned_extents; | |
a28ec197 CM |
1447 | |
1448 | while(1) { | |
1a5bc167 CM |
1449 | ret = find_first_extent_bit(pending_del, 0, &start, &end, |
1450 | EXTENT_LOCKED); | |
1451 | if (ret) | |
a28ec197 | 1452 | break; |
324ae4df | 1453 | update_pinned_extents(extent_root, start, end + 1 - start, 1); |
1a5bc167 CM |
1454 | clear_extent_bits(pending_del, start, end, EXTENT_LOCKED, |
1455 | GFP_NOFS); | |
1456 | ret = __free_extent(trans, extent_root, | |
7bb86316 CM |
1457 | start, end + 1 - start, |
1458 | extent_root->root_key.objectid, | |
1459 | 0, 0, 0, 0, 0); | |
1a5bc167 CM |
1460 | if (ret) |
1461 | err = ret; | |
fec577fb | 1462 | } |
e20d96d6 | 1463 | return err; |
fec577fb CM |
1464 | } |
1465 | ||
1466 | /* | |
1467 | * remove an extent from the root, returns 0 on success | |
1468 | */ | |
e089f05c | 1469 | int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root |
7bb86316 CM |
1470 | *root, u64 bytenr, u64 num_bytes, |
1471 | u64 root_objectid, u64 ref_generation, | |
1472 | u64 owner_objectid, u64 owner_offset, int pin) | |
fec577fb | 1473 | { |
9f5fae2f | 1474 | struct btrfs_root *extent_root = root->fs_info->extent_root; |
fec577fb CM |
1475 | int pending_ret; |
1476 | int ret; | |
a28ec197 | 1477 | |
db94535d | 1478 | WARN_ON(num_bytes < root->sectorsize); |
7bb86316 CM |
1479 | if (!root->ref_cows) |
1480 | ref_generation = 0; | |
1481 | ||
fec577fb | 1482 | if (root == extent_root) { |
db94535d | 1483 | pin_down_bytes(root, bytenr, num_bytes, 1); |
fec577fb CM |
1484 | return 0; |
1485 | } | |
7bb86316 CM |
1486 | ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid, |
1487 | ref_generation, owner_objectid, owner_offset, | |
1488 | pin, pin == 0); | |
e20d96d6 | 1489 | pending_ret = del_pending_extents(trans, root->fs_info->extent_root); |
fec577fb CM |
1490 | return ret ? ret : pending_ret; |
1491 | } | |
1492 | ||
87ee04eb CM |
1493 | static u64 stripe_align(struct btrfs_root *root, u64 val) |
1494 | { | |
1495 | u64 mask = ((u64)root->stripesize - 1); | |
1496 | u64 ret = (val + mask) & ~mask; | |
1497 | return ret; | |
1498 | } | |
1499 | ||
fec577fb CM |
1500 | /* |
1501 | * walks the btree of allocated extents and find a hole of a given size. | |
1502 | * The key ins is changed to record the hole: | |
1503 | * ins->objectid == block start | |
62e2749e | 1504 | * ins->flags = BTRFS_EXTENT_ITEM_KEY |
fec577fb CM |
1505 | * ins->offset == number of blocks |
1506 | * Any available blocks before search_start are skipped. | |
1507 | */ | |
98ed5174 CM |
1508 | static int noinline find_free_extent(struct btrfs_trans_handle *trans, |
1509 | struct btrfs_root *orig_root, | |
1510 | u64 num_bytes, u64 empty_size, | |
1511 | u64 search_start, u64 search_end, | |
1512 | u64 hint_byte, struct btrfs_key *ins, | |
1513 | u64 exclude_start, u64 exclude_nr, | |
1514 | int data) | |
fec577fb | 1515 | { |
87ee04eb | 1516 | int ret; |
be744175 | 1517 | u64 orig_search_start = search_start; |
9f5fae2f | 1518 | struct btrfs_root * root = orig_root->fs_info->extent_root; |
f2458e1d | 1519 | struct btrfs_fs_info *info = root->fs_info; |
db94535d | 1520 | u64 total_needed = num_bytes; |
239b14b3 | 1521 | u64 *last_ptr = NULL; |
be08c1b9 | 1522 | struct btrfs_block_group_cache *block_group; |
be744175 | 1523 | int full_scan = 0; |
fbdc762b | 1524 | int wrapped = 0; |
239b14b3 | 1525 | int empty_cluster = 2 * 1024 * 1024; |
fec577fb | 1526 | |
db94535d | 1527 | WARN_ON(num_bytes < root->sectorsize); |
b1a4d965 CM |
1528 | btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY); |
1529 | ||
239b14b3 CM |
1530 | if (data & BTRFS_BLOCK_GROUP_METADATA) { |
1531 | last_ptr = &root->fs_info->last_alloc; | |
1532 | } | |
1533 | ||
1534 | if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) { | |
1535 | last_ptr = &root->fs_info->last_data_alloc; | |
1536 | } | |
1537 | ||
1538 | if (last_ptr) { | |
1539 | if (*last_ptr) | |
1540 | hint_byte = *last_ptr; | |
1541 | else { | |
1542 | empty_size += empty_cluster; | |
1543 | } | |
1544 | } | |
1545 | ||
7f93bf8d CM |
1546 | if (search_end == (u64)-1) |
1547 | search_end = btrfs_super_total_bytes(&info->super_copy); | |
0b86a832 | 1548 | |
db94535d CM |
1549 | if (hint_byte) { |
1550 | block_group = btrfs_lookup_block_group(info, hint_byte); | |
1a2b2ac7 CM |
1551 | if (!block_group) |
1552 | hint_byte = search_start; | |
be744175 | 1553 | block_group = btrfs_find_block_group(root, block_group, |
db94535d | 1554 | hint_byte, data, 1); |
239b14b3 CM |
1555 | if (last_ptr && *last_ptr == 0 && block_group) |
1556 | hint_byte = block_group->key.objectid; | |
be744175 CM |
1557 | } else { |
1558 | block_group = btrfs_find_block_group(root, | |
1a2b2ac7 CM |
1559 | trans->block_group, |
1560 | search_start, data, 1); | |
be744175 | 1561 | } |
239b14b3 | 1562 | search_start = max(search_start, hint_byte); |
be744175 | 1563 | |
6702ed49 | 1564 | total_needed += empty_size; |
0b86a832 | 1565 | |
be744175 | 1566 | check_failed: |
70b043f0 CM |
1567 | if (!block_group) { |
1568 | block_group = btrfs_lookup_block_group(info, search_start); | |
1569 | if (!block_group) | |
1570 | block_group = btrfs_lookup_block_group(info, | |
1571 | orig_search_start); | |
1572 | } | |
0b86a832 CM |
1573 | ret = find_search_start(root, &block_group, &search_start, |
1574 | total_needed, data); | |
239b14b3 CM |
1575 | if (ret == -ENOSPC && last_ptr && *last_ptr) { |
1576 | *last_ptr = 0; | |
1577 | block_group = btrfs_lookup_block_group(info, | |
1578 | orig_search_start); | |
1579 | search_start = orig_search_start; | |
1580 | ret = find_search_start(root, &block_group, &search_start, | |
1581 | total_needed, data); | |
1582 | } | |
1583 | if (ret == -ENOSPC) | |
1584 | goto enospc; | |
0b86a832 | 1585 | if (ret) |
d548ee51 | 1586 | goto error; |
e19caa5f | 1587 | |
239b14b3 CM |
1588 | if (last_ptr && *last_ptr && search_start != *last_ptr) { |
1589 | *last_ptr = 0; | |
1590 | if (!empty_size) { | |
1591 | empty_size += empty_cluster; | |
1592 | total_needed += empty_size; | |
1593 | } | |
1594 | block_group = btrfs_lookup_block_group(info, | |
1595 | orig_search_start); | |
1596 | search_start = orig_search_start; | |
1597 | ret = find_search_start(root, &block_group, | |
1598 | &search_start, total_needed, data); | |
1599 | if (ret == -ENOSPC) | |
1600 | goto enospc; | |
1601 | if (ret) | |
1602 | goto error; | |
1603 | } | |
1604 | ||
0b86a832 CM |
1605 | search_start = stripe_align(root, search_start); |
1606 | ins->objectid = search_start; | |
1607 | ins->offset = num_bytes; | |
e37c9e69 | 1608 | |
db94535d | 1609 | if (ins->objectid + num_bytes >= search_end) |
cf67582b | 1610 | goto enospc; |
0b86a832 CM |
1611 | |
1612 | if (ins->objectid + num_bytes > | |
1613 | block_group->key.objectid + block_group->key.offset) { | |
e19caa5f CM |
1614 | search_start = block_group->key.objectid + |
1615 | block_group->key.offset; | |
1616 | goto new_group; | |
1617 | } | |
0b86a832 | 1618 | |
1a5bc167 | 1619 | if (test_range_bit(&info->extent_ins, ins->objectid, |
db94535d CM |
1620 | ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) { |
1621 | search_start = ins->objectid + num_bytes; | |
1a5bc167 CM |
1622 | goto new_group; |
1623 | } | |
0b86a832 | 1624 | |
1a5bc167 | 1625 | if (test_range_bit(&info->pinned_extents, ins->objectid, |
db94535d CM |
1626 | ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) { |
1627 | search_start = ins->objectid + num_bytes; | |
1a5bc167 | 1628 | goto new_group; |
fec577fb | 1629 | } |
0b86a832 | 1630 | |
db94535d | 1631 | if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start && |
f2654de4 CM |
1632 | ins->objectid < exclude_start + exclude_nr)) { |
1633 | search_start = exclude_start + exclude_nr; | |
1634 | goto new_group; | |
1635 | } | |
0b86a832 | 1636 | |
6324fbf3 | 1637 | if (!(data & BTRFS_BLOCK_GROUP_DATA)) { |
5276aeda | 1638 | block_group = btrfs_lookup_block_group(info, ins->objectid); |
26b8003f CM |
1639 | if (block_group) |
1640 | trans->block_group = block_group; | |
f2458e1d | 1641 | } |
db94535d | 1642 | ins->offset = num_bytes; |
239b14b3 CM |
1643 | if (last_ptr) { |
1644 | *last_ptr = ins->objectid + ins->offset; | |
1645 | if (*last_ptr == | |
1646 | btrfs_super_total_bytes(&root->fs_info->super_copy)) { | |
1647 | *last_ptr = 0; | |
1648 | } | |
1649 | } | |
fec577fb | 1650 | return 0; |
be744175 CM |
1651 | |
1652 | new_group: | |
db94535d | 1653 | if (search_start + num_bytes >= search_end) { |
cf67582b | 1654 | enospc: |
be744175 | 1655 | search_start = orig_search_start; |
fbdc762b CM |
1656 | if (full_scan) { |
1657 | ret = -ENOSPC; | |
1658 | goto error; | |
1659 | } | |
6702ed49 CM |
1660 | if (wrapped) { |
1661 | if (!full_scan) | |
1662 | total_needed -= empty_size; | |
fbdc762b | 1663 | full_scan = 1; |
6702ed49 | 1664 | } else |
fbdc762b | 1665 | wrapped = 1; |
be744175 | 1666 | } |
5276aeda | 1667 | block_group = btrfs_lookup_block_group(info, search_start); |
fbdc762b | 1668 | cond_resched(); |
1a2b2ac7 CM |
1669 | block_group = btrfs_find_block_group(root, block_group, |
1670 | search_start, data, 0); | |
be744175 CM |
1671 | goto check_failed; |
1672 | ||
0f70abe2 | 1673 | error: |
0f70abe2 | 1674 | return ret; |
fec577fb | 1675 | } |
fec577fb CM |
1676 | /* |
1677 | * finds a free extent and does all the dirty work required for allocation | |
1678 | * returns the key for the extent through ins, and a tree buffer for | |
1679 | * the first block of the extent through buf. | |
1680 | * | |
1681 | * returns 0 if everything worked, non-zero otherwise. | |
1682 | */ | |
4d775673 | 1683 | int btrfs_alloc_extent(struct btrfs_trans_handle *trans, |
7bb86316 CM |
1684 | struct btrfs_root *root, |
1685 | u64 num_bytes, u64 root_objectid, u64 ref_generation, | |
1686 | u64 owner, u64 owner_offset, | |
1687 | u64 empty_size, u64 hint_byte, | |
be08c1b9 | 1688 | u64 search_end, struct btrfs_key *ins, int data) |
fec577fb CM |
1689 | { |
1690 | int ret; | |
1691 | int pending_ret; | |
c31f8830 CM |
1692 | u64 super_used; |
1693 | u64 root_used; | |
fbdc762b | 1694 | u64 search_start = 0; |
edbd8d4e | 1695 | u64 new_hint; |
47e4bb98 | 1696 | u32 sizes[2]; |
1261ec42 CM |
1697 | struct btrfs_fs_info *info = root->fs_info; |
1698 | struct btrfs_root *extent_root = info->extent_root; | |
47e4bb98 CM |
1699 | struct btrfs_extent_item *extent_item; |
1700 | struct btrfs_extent_ref *ref; | |
7bb86316 | 1701 | struct btrfs_path *path; |
47e4bb98 | 1702 | struct btrfs_key keys[2]; |
593060d7 | 1703 | int extra_chunk_alloc_bits = 0; |
8f662a76 | 1704 | |
6324fbf3 | 1705 | if (data) { |
593060d7 | 1706 | data = BTRFS_BLOCK_GROUP_DATA | info->extra_data_alloc_bits; |
6324fbf3 CM |
1707 | } else if (root == root->fs_info->chunk_root) { |
1708 | data = BTRFS_BLOCK_GROUP_SYSTEM; | |
1709 | } else { | |
593060d7 | 1710 | data = BTRFS_BLOCK_GROUP_METADATA | info->extra_alloc_bits; |
6324fbf3 | 1711 | } |
593060d7 CM |
1712 | if (btrfs_super_num_devices(&info->super_copy) > 1 && |
1713 | !(data & BTRFS_BLOCK_GROUP_SYSTEM)) | |
1714 | extra_chunk_alloc_bits = BTRFS_BLOCK_GROUP_RAID0; | |
6324fbf3 CM |
1715 | |
1716 | if (root->ref_cows) { | |
593060d7 | 1717 | if (!(data & BTRFS_BLOCK_GROUP_METADATA)) { |
6324fbf3 | 1718 | ret = do_chunk_alloc(trans, root->fs_info->extent_root, |
239b14b3 | 1719 | 2 * 1024 * 1024, |
593060d7 CM |
1720 | BTRFS_BLOCK_GROUP_METADATA | |
1721 | info->extra_alloc_bits | | |
1722 | extra_chunk_alloc_bits); | |
6324fbf3 CM |
1723 | BUG_ON(ret); |
1724 | } | |
1725 | ret = do_chunk_alloc(trans, root->fs_info->extent_root, | |
593060d7 CM |
1726 | num_bytes + 2 * 1024 * 1024, data | |
1727 | extra_chunk_alloc_bits); | |
6324fbf3 CM |
1728 | BUG_ON(ret); |
1729 | } | |
0b86a832 | 1730 | |
8f662a76 | 1731 | new_hint = max(hint_byte, root->fs_info->alloc_start); |
edbd8d4e CM |
1732 | if (new_hint < btrfs_super_total_bytes(&info->super_copy)) |
1733 | hint_byte = new_hint; | |
8f662a76 | 1734 | |
db94535d CM |
1735 | WARN_ON(num_bytes < root->sectorsize); |
1736 | ret = find_free_extent(trans, root, num_bytes, empty_size, | |
1737 | search_start, search_end, hint_byte, ins, | |
26b8003f CM |
1738 | trans->alloc_exclude_start, |
1739 | trans->alloc_exclude_nr, data); | |
ccd467d6 | 1740 | BUG_ON(ret); |
f2654de4 CM |
1741 | if (ret) |
1742 | return ret; | |
fec577fb | 1743 | |
58176a96 | 1744 | /* block accounting for super block */ |
db94535d CM |
1745 | super_used = btrfs_super_bytes_used(&info->super_copy); |
1746 | btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes); | |
26b8003f | 1747 | |
58176a96 | 1748 | /* block accounting for root item */ |
db94535d CM |
1749 | root_used = btrfs_root_used(&root->root_item); |
1750 | btrfs_set_root_used(&root->root_item, root_used + num_bytes); | |
58176a96 | 1751 | |
f510cfec CM |
1752 | clear_extent_dirty(&root->fs_info->free_space_cache, |
1753 | ins->objectid, ins->objectid + ins->offset - 1, | |
1754 | GFP_NOFS); | |
1755 | ||
26b8003f | 1756 | if (root == extent_root) { |
1a5bc167 CM |
1757 | set_extent_bits(&root->fs_info->extent_ins, ins->objectid, |
1758 | ins->objectid + ins->offset - 1, | |
1759 | EXTENT_LOCKED, GFP_NOFS); | |
26b8003f CM |
1760 | goto update_block; |
1761 | } | |
1762 | ||
1763 | WARN_ON(trans->alloc_exclude_nr); | |
1764 | trans->alloc_exclude_start = ins->objectid; | |
1765 | trans->alloc_exclude_nr = ins->offset; | |
037e6390 | 1766 | |
47e4bb98 CM |
1767 | memcpy(&keys[0], ins, sizeof(*ins)); |
1768 | keys[1].offset = hash_extent_ref(root_objectid, ref_generation, | |
1769 | owner, owner_offset); | |
1770 | keys[1].objectid = ins->objectid; | |
1771 | keys[1].type = BTRFS_EXTENT_REF_KEY; | |
1772 | sizes[0] = sizeof(*extent_item); | |
1773 | sizes[1] = sizeof(*ref); | |
7bb86316 CM |
1774 | |
1775 | path = btrfs_alloc_path(); | |
1776 | BUG_ON(!path); | |
47e4bb98 CM |
1777 | |
1778 | ret = btrfs_insert_empty_items(trans, extent_root, path, keys, | |
1779 | sizes, 2); | |
26b8003f | 1780 | |
ccd467d6 | 1781 | BUG_ON(ret); |
47e4bb98 CM |
1782 | extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0], |
1783 | struct btrfs_extent_item); | |
1784 | btrfs_set_extent_refs(path->nodes[0], extent_item, 1); | |
1785 | ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1, | |
1786 | struct btrfs_extent_ref); | |
1787 | ||
1788 | btrfs_set_ref_root(path->nodes[0], ref, root_objectid); | |
1789 | btrfs_set_ref_generation(path->nodes[0], ref, ref_generation); | |
1790 | btrfs_set_ref_objectid(path->nodes[0], ref, owner); | |
1791 | btrfs_set_ref_offset(path->nodes[0], ref, owner_offset); | |
1792 | ||
1793 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
1794 | ||
1795 | trans->alloc_exclude_start = 0; | |
1796 | trans->alloc_exclude_nr = 0; | |
7bb86316 | 1797 | btrfs_free_path(path); |
e089f05c | 1798 | finish_current_insert(trans, extent_root); |
e20d96d6 | 1799 | pending_ret = del_pending_extents(trans, extent_root); |
f510cfec | 1800 | |
e37c9e69 | 1801 | if (ret) { |
037e6390 | 1802 | return ret; |
e37c9e69 CM |
1803 | } |
1804 | if (pending_ret) { | |
037e6390 | 1805 | return pending_ret; |
e37c9e69 | 1806 | } |
26b8003f CM |
1807 | |
1808 | update_block: | |
0b86a832 | 1809 | ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0); |
f5947066 CM |
1810 | if (ret) { |
1811 | printk("update block group failed for %Lu %Lu\n", | |
1812 | ins->objectid, ins->offset); | |
1813 | BUG(); | |
1814 | } | |
037e6390 | 1815 | return 0; |
fec577fb CM |
1816 | } |
1817 | ||
1818 | /* | |
1819 | * helper function to allocate a block for a given tree | |
1820 | * returns the tree buffer or NULL. | |
1821 | */ | |
5f39d397 | 1822 | struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, |
db94535d | 1823 | struct btrfs_root *root, |
7bb86316 CM |
1824 | u32 blocksize, |
1825 | u64 root_objectid, u64 hint, | |
1826 | u64 empty_size) | |
1827 | { | |
1828 | u64 ref_generation; | |
1829 | ||
1830 | if (root->ref_cows) | |
1831 | ref_generation = trans->transid; | |
1832 | else | |
1833 | ref_generation = 0; | |
1834 | ||
1835 | ||
1836 | return __btrfs_alloc_free_block(trans, root, blocksize, root_objectid, | |
1837 | ref_generation, 0, 0, hint, empty_size); | |
1838 | } | |
1839 | ||
1840 | /* | |
1841 | * helper function to allocate a block for a given tree | |
1842 | * returns the tree buffer or NULL. | |
1843 | */ | |
1844 | struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, | |
1845 | struct btrfs_root *root, | |
1846 | u32 blocksize, | |
1847 | u64 root_objectid, | |
1848 | u64 ref_generation, | |
1849 | u64 first_objectid, | |
1850 | int level, | |
1851 | u64 hint, | |
5f39d397 | 1852 | u64 empty_size) |
fec577fb | 1853 | { |
e2fa7227 | 1854 | struct btrfs_key ins; |
fec577fb | 1855 | int ret; |
5f39d397 | 1856 | struct extent_buffer *buf; |
fec577fb | 1857 | |
7bb86316 CM |
1858 | ret = btrfs_alloc_extent(trans, root, blocksize, |
1859 | root_objectid, ref_generation, | |
f6dbff55 | 1860 | level, first_objectid, empty_size, hint, |
db94535d | 1861 | (u64)-1, &ins, 0); |
fec577fb | 1862 | if (ret) { |
54aa1f4d CM |
1863 | BUG_ON(ret > 0); |
1864 | return ERR_PTR(ret); | |
fec577fb | 1865 | } |
db94535d | 1866 | buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize); |
54aa1f4d | 1867 | if (!buf) { |
7bb86316 CM |
1868 | btrfs_free_extent(trans, root, ins.objectid, blocksize, |
1869 | root->root_key.objectid, ref_generation, | |
1870 | 0, 0, 0); | |
54aa1f4d CM |
1871 | return ERR_PTR(-ENOMEM); |
1872 | } | |
55c69072 CM |
1873 | btrfs_set_header_generation(buf, trans->transid); |
1874 | clean_tree_block(trans, root, buf); | |
1875 | wait_on_tree_block_writeback(root, buf); | |
5f39d397 | 1876 | btrfs_set_buffer_uptodate(buf); |
55c69072 CM |
1877 | |
1878 | if (PageDirty(buf->first_page)) { | |
1879 | printk("page %lu dirty\n", buf->first_page->index); | |
1880 | WARN_ON(1); | |
1881 | } | |
1882 | ||
5f39d397 CM |
1883 | set_extent_dirty(&trans->transaction->dirty_pages, buf->start, |
1884 | buf->start + buf->len - 1, GFP_NOFS); | |
d1310b2e | 1885 | set_extent_bits(&BTRFS_I(root->fs_info->btree_inode)->io_tree, |
19c00ddc CM |
1886 | buf->start, buf->start + buf->len - 1, |
1887 | EXTENT_CSUM, GFP_NOFS); | |
1888 | buf->flags |= EXTENT_CSUM; | |
9afbb0b7 CM |
1889 | if (!btrfs_test_opt(root, SSD)) |
1890 | btrfs_set_buffer_defrag(buf); | |
d3c2fdcf | 1891 | trans->blocks_used++; |
fec577fb CM |
1892 | return buf; |
1893 | } | |
a28ec197 | 1894 | |
98ed5174 CM |
1895 | static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans, |
1896 | struct btrfs_root *root, | |
1897 | struct extent_buffer *leaf) | |
6407bf6d | 1898 | { |
7bb86316 CM |
1899 | u64 leaf_owner; |
1900 | u64 leaf_generation; | |
5f39d397 | 1901 | struct btrfs_key key; |
6407bf6d CM |
1902 | struct btrfs_file_extent_item *fi; |
1903 | int i; | |
1904 | int nritems; | |
1905 | int ret; | |
1906 | ||
5f39d397 CM |
1907 | BUG_ON(!btrfs_is_leaf(leaf)); |
1908 | nritems = btrfs_header_nritems(leaf); | |
7bb86316 CM |
1909 | leaf_owner = btrfs_header_owner(leaf); |
1910 | leaf_generation = btrfs_header_generation(leaf); | |
1911 | ||
6407bf6d | 1912 | for (i = 0; i < nritems; i++) { |
db94535d | 1913 | u64 disk_bytenr; |
5f39d397 CM |
1914 | |
1915 | btrfs_item_key_to_cpu(leaf, &key, i); | |
1916 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
6407bf6d CM |
1917 | continue; |
1918 | fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item); | |
5f39d397 CM |
1919 | if (btrfs_file_extent_type(leaf, fi) == |
1920 | BTRFS_FILE_EXTENT_INLINE) | |
236454df | 1921 | continue; |
6407bf6d CM |
1922 | /* |
1923 | * FIXME make sure to insert a trans record that | |
1924 | * repeats the snapshot del on crash | |
1925 | */ | |
db94535d CM |
1926 | disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); |
1927 | if (disk_bytenr == 0) | |
3a686375 | 1928 | continue; |
db94535d | 1929 | ret = btrfs_free_extent(trans, root, disk_bytenr, |
7bb86316 CM |
1930 | btrfs_file_extent_disk_num_bytes(leaf, fi), |
1931 | leaf_owner, leaf_generation, | |
1932 | key.objectid, key.offset, 0); | |
6407bf6d CM |
1933 | BUG_ON(ret); |
1934 | } | |
1935 | return 0; | |
1936 | } | |
1937 | ||
98ed5174 | 1938 | static void noinline reada_walk_down(struct btrfs_root *root, |
bea495e5 CM |
1939 | struct extent_buffer *node, |
1940 | int slot) | |
e011599b | 1941 | { |
db94535d | 1942 | u64 bytenr; |
bea495e5 CM |
1943 | u64 last = 0; |
1944 | u32 nritems; | |
e011599b | 1945 | u32 refs; |
db94535d | 1946 | u32 blocksize; |
bea495e5 CM |
1947 | int ret; |
1948 | int i; | |
1949 | int level; | |
1950 | int skipped = 0; | |
e011599b | 1951 | |
5f39d397 | 1952 | nritems = btrfs_header_nritems(node); |
db94535d | 1953 | level = btrfs_header_level(node); |
bea495e5 CM |
1954 | if (level) |
1955 | return; | |
1956 | ||
1957 | for (i = slot; i < nritems && skipped < 32; i++) { | |
db94535d | 1958 | bytenr = btrfs_node_blockptr(node, i); |
bea495e5 CM |
1959 | if (last && ((bytenr > last && bytenr - last > 32 * 1024) || |
1960 | (last > bytenr && last - bytenr > 32 * 1024))) { | |
1961 | skipped++; | |
e011599b | 1962 | continue; |
bea495e5 CM |
1963 | } |
1964 | blocksize = btrfs_level_size(root, level - 1); | |
1965 | if (i != slot) { | |
1966 | ret = lookup_extent_ref(NULL, root, bytenr, | |
1967 | blocksize, &refs); | |
1968 | BUG_ON(ret); | |
1969 | if (refs != 1) { | |
1970 | skipped++; | |
1971 | continue; | |
1972 | } | |
1973 | } | |
409eb95d | 1974 | mutex_unlock(&root->fs_info->fs_mutex); |
db94535d | 1975 | ret = readahead_tree_block(root, bytenr, blocksize); |
bea495e5 | 1976 | last = bytenr + blocksize; |
409eb95d CM |
1977 | cond_resched(); |
1978 | mutex_lock(&root->fs_info->fs_mutex); | |
e011599b CM |
1979 | if (ret) |
1980 | break; | |
1981 | } | |
1982 | } | |
1983 | ||
9aca1d51 CM |
1984 | /* |
1985 | * helper function for drop_snapshot, this walks down the tree dropping ref | |
1986 | * counts as it goes. | |
1987 | */ | |
98ed5174 CM |
1988 | static int noinline walk_down_tree(struct btrfs_trans_handle *trans, |
1989 | struct btrfs_root *root, | |
1990 | struct btrfs_path *path, int *level) | |
20524f02 | 1991 | { |
7bb86316 CM |
1992 | u64 root_owner; |
1993 | u64 root_gen; | |
1994 | u64 bytenr; | |
5f39d397 CM |
1995 | struct extent_buffer *next; |
1996 | struct extent_buffer *cur; | |
7bb86316 | 1997 | struct extent_buffer *parent; |
db94535d | 1998 | u32 blocksize; |
20524f02 CM |
1999 | int ret; |
2000 | u32 refs; | |
2001 | ||
5caf2a00 CM |
2002 | WARN_ON(*level < 0); |
2003 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
5f39d397 | 2004 | ret = lookup_extent_ref(trans, root, |
db94535d CM |
2005 | path->nodes[*level]->start, |
2006 | path->nodes[*level]->len, &refs); | |
20524f02 CM |
2007 | BUG_ON(ret); |
2008 | if (refs > 1) | |
2009 | goto out; | |
e011599b | 2010 | |
9aca1d51 CM |
2011 | /* |
2012 | * walk down to the last node level and free all the leaves | |
2013 | */ | |
6407bf6d | 2014 | while(*level >= 0) { |
5caf2a00 CM |
2015 | WARN_ON(*level < 0); |
2016 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
20524f02 | 2017 | cur = path->nodes[*level]; |
e011599b | 2018 | |
5f39d397 | 2019 | if (btrfs_header_level(cur) != *level) |
2c90e5d6 | 2020 | WARN_ON(1); |
e011599b | 2021 | |
7518a238 | 2022 | if (path->slots[*level] >= |
5f39d397 | 2023 | btrfs_header_nritems(cur)) |
20524f02 | 2024 | break; |
6407bf6d CM |
2025 | if (*level == 0) { |
2026 | ret = drop_leaf_ref(trans, root, cur); | |
2027 | BUG_ON(ret); | |
2028 | break; | |
2029 | } | |
db94535d CM |
2030 | bytenr = btrfs_node_blockptr(cur, path->slots[*level]); |
2031 | blocksize = btrfs_level_size(root, *level - 1); | |
2032 | ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs); | |
6407bf6d CM |
2033 | BUG_ON(ret); |
2034 | if (refs != 1) { | |
7bb86316 CM |
2035 | parent = path->nodes[*level]; |
2036 | root_owner = btrfs_header_owner(parent); | |
2037 | root_gen = btrfs_header_generation(parent); | |
20524f02 | 2038 | path->slots[*level]++; |
db94535d | 2039 | ret = btrfs_free_extent(trans, root, bytenr, |
7bb86316 CM |
2040 | blocksize, root_owner, |
2041 | root_gen, 0, 0, 1); | |
20524f02 CM |
2042 | BUG_ON(ret); |
2043 | continue; | |
2044 | } | |
db94535d | 2045 | next = btrfs_find_tree_block(root, bytenr, blocksize); |
5f39d397 CM |
2046 | if (!next || !btrfs_buffer_uptodate(next)) { |
2047 | free_extent_buffer(next); | |
bea495e5 | 2048 | reada_walk_down(root, cur, path->slots[*level]); |
db94535d | 2049 | next = read_tree_block(root, bytenr, blocksize); |
e9d0b13b | 2050 | |
ecbe2402 CM |
2051 | /* we used to drop the lock above, keep the |
2052 | * code to double check so that we won't forget | |
2053 | * when we drop the lock again in the future | |
2054 | */ | |
db94535d CM |
2055 | ret = lookup_extent_ref(trans, root, bytenr, |
2056 | blocksize, &refs); | |
e9d0b13b CM |
2057 | BUG_ON(ret); |
2058 | if (refs != 1) { | |
7bb86316 CM |
2059 | parent = path->nodes[*level]; |
2060 | root_owner = btrfs_header_owner(parent); | |
2061 | root_gen = btrfs_header_generation(parent); | |
2062 | ||
e9d0b13b | 2063 | path->slots[*level]++; |
5f39d397 | 2064 | free_extent_buffer(next); |
7bb86316 CM |
2065 | ret = btrfs_free_extent(trans, root, bytenr, |
2066 | blocksize, | |
2067 | root_owner, | |
2068 | root_gen, 0, 0, 1); | |
e9d0b13b CM |
2069 | BUG_ON(ret); |
2070 | continue; | |
2071 | } | |
0999df54 CM |
2072 | } else if (next) { |
2073 | btrfs_verify_block_csum(root, next); | |
e9d0b13b | 2074 | } |
5caf2a00 | 2075 | WARN_ON(*level <= 0); |
83e15a28 | 2076 | if (path->nodes[*level-1]) |
5f39d397 | 2077 | free_extent_buffer(path->nodes[*level-1]); |
20524f02 | 2078 | path->nodes[*level-1] = next; |
5f39d397 | 2079 | *level = btrfs_header_level(next); |
20524f02 CM |
2080 | path->slots[*level] = 0; |
2081 | } | |
2082 | out: | |
5caf2a00 CM |
2083 | WARN_ON(*level < 0); |
2084 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
7bb86316 CM |
2085 | |
2086 | if (path->nodes[*level] == root->node) { | |
2087 | root_owner = root->root_key.objectid; | |
2088 | parent = path->nodes[*level]; | |
2089 | } else { | |
2090 | parent = path->nodes[*level + 1]; | |
2091 | root_owner = btrfs_header_owner(parent); | |
2092 | } | |
2093 | ||
2094 | root_gen = btrfs_header_generation(parent); | |
db94535d | 2095 | ret = btrfs_free_extent(trans, root, path->nodes[*level]->start, |
7bb86316 CM |
2096 | path->nodes[*level]->len, |
2097 | root_owner, root_gen, 0, 0, 1); | |
5f39d397 | 2098 | free_extent_buffer(path->nodes[*level]); |
20524f02 CM |
2099 | path->nodes[*level] = NULL; |
2100 | *level += 1; | |
2101 | BUG_ON(ret); | |
2102 | return 0; | |
2103 | } | |
2104 | ||
9aca1d51 CM |
2105 | /* |
2106 | * helper for dropping snapshots. This walks back up the tree in the path | |
2107 | * to find the first node higher up where we haven't yet gone through | |
2108 | * all the slots | |
2109 | */ | |
98ed5174 CM |
2110 | static int noinline walk_up_tree(struct btrfs_trans_handle *trans, |
2111 | struct btrfs_root *root, | |
2112 | struct btrfs_path *path, int *level) | |
20524f02 | 2113 | { |
7bb86316 CM |
2114 | u64 root_owner; |
2115 | u64 root_gen; | |
2116 | struct btrfs_root_item *root_item = &root->root_item; | |
20524f02 CM |
2117 | int i; |
2118 | int slot; | |
2119 | int ret; | |
9f3a7427 | 2120 | |
234b63a0 | 2121 | for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { |
20524f02 | 2122 | slot = path->slots[i]; |
5f39d397 CM |
2123 | if (slot < btrfs_header_nritems(path->nodes[i]) - 1) { |
2124 | struct extent_buffer *node; | |
2125 | struct btrfs_disk_key disk_key; | |
2126 | node = path->nodes[i]; | |
20524f02 CM |
2127 | path->slots[i]++; |
2128 | *level = i; | |
9f3a7427 | 2129 | WARN_ON(*level == 0); |
5f39d397 | 2130 | btrfs_node_key(node, &disk_key, path->slots[i]); |
9f3a7427 | 2131 | memcpy(&root_item->drop_progress, |
5f39d397 | 2132 | &disk_key, sizeof(disk_key)); |
9f3a7427 | 2133 | root_item->drop_level = i; |
20524f02 CM |
2134 | return 0; |
2135 | } else { | |
7bb86316 CM |
2136 | if (path->nodes[*level] == root->node) { |
2137 | root_owner = root->root_key.objectid; | |
2138 | root_gen = | |
2139 | btrfs_header_generation(path->nodes[*level]); | |
2140 | } else { | |
2141 | struct extent_buffer *node; | |
2142 | node = path->nodes[*level + 1]; | |
2143 | root_owner = btrfs_header_owner(node); | |
2144 | root_gen = btrfs_header_generation(node); | |
2145 | } | |
e089f05c | 2146 | ret = btrfs_free_extent(trans, root, |
db94535d | 2147 | path->nodes[*level]->start, |
7bb86316 CM |
2148 | path->nodes[*level]->len, |
2149 | root_owner, root_gen, 0, 0, 1); | |
6407bf6d | 2150 | BUG_ON(ret); |
5f39d397 | 2151 | free_extent_buffer(path->nodes[*level]); |
83e15a28 | 2152 | path->nodes[*level] = NULL; |
20524f02 | 2153 | *level = i + 1; |
20524f02 CM |
2154 | } |
2155 | } | |
2156 | return 1; | |
2157 | } | |
2158 | ||
9aca1d51 CM |
2159 | /* |
2160 | * drop the reference count on the tree rooted at 'snap'. This traverses | |
2161 | * the tree freeing any blocks that have a ref count of zero after being | |
2162 | * decremented. | |
2163 | */ | |
e089f05c | 2164 | int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root |
9f3a7427 | 2165 | *root) |
20524f02 | 2166 | { |
3768f368 | 2167 | int ret = 0; |
9aca1d51 | 2168 | int wret; |
20524f02 | 2169 | int level; |
5caf2a00 | 2170 | struct btrfs_path *path; |
20524f02 CM |
2171 | int i; |
2172 | int orig_level; | |
9f3a7427 | 2173 | struct btrfs_root_item *root_item = &root->root_item; |
20524f02 | 2174 | |
5caf2a00 CM |
2175 | path = btrfs_alloc_path(); |
2176 | BUG_ON(!path); | |
20524f02 | 2177 | |
5f39d397 | 2178 | level = btrfs_header_level(root->node); |
20524f02 | 2179 | orig_level = level; |
9f3a7427 CM |
2180 | if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) { |
2181 | path->nodes[level] = root->node; | |
f510cfec | 2182 | extent_buffer_get(root->node); |
9f3a7427 CM |
2183 | path->slots[level] = 0; |
2184 | } else { | |
2185 | struct btrfs_key key; | |
5f39d397 CM |
2186 | struct btrfs_disk_key found_key; |
2187 | struct extent_buffer *node; | |
6702ed49 | 2188 | |
9f3a7427 | 2189 | btrfs_disk_key_to_cpu(&key, &root_item->drop_progress); |
6702ed49 CM |
2190 | level = root_item->drop_level; |
2191 | path->lowest_level = level; | |
9f3a7427 | 2192 | wret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
6702ed49 | 2193 | if (wret < 0) { |
9f3a7427 CM |
2194 | ret = wret; |
2195 | goto out; | |
2196 | } | |
5f39d397 CM |
2197 | node = path->nodes[level]; |
2198 | btrfs_node_key(node, &found_key, path->slots[level]); | |
2199 | WARN_ON(memcmp(&found_key, &root_item->drop_progress, | |
2200 | sizeof(found_key))); | |
9f3a7427 | 2201 | } |
20524f02 | 2202 | while(1) { |
5caf2a00 | 2203 | wret = walk_down_tree(trans, root, path, &level); |
9aca1d51 | 2204 | if (wret > 0) |
20524f02 | 2205 | break; |
9aca1d51 CM |
2206 | if (wret < 0) |
2207 | ret = wret; | |
2208 | ||
5caf2a00 | 2209 | wret = walk_up_tree(trans, root, path, &level); |
9aca1d51 | 2210 | if (wret > 0) |
20524f02 | 2211 | break; |
9aca1d51 CM |
2212 | if (wret < 0) |
2213 | ret = wret; | |
409eb95d | 2214 | ret = -EAGAIN; |
409eb95d | 2215 | break; |
20524f02 | 2216 | } |
83e15a28 | 2217 | for (i = 0; i <= orig_level; i++) { |
5caf2a00 | 2218 | if (path->nodes[i]) { |
5f39d397 | 2219 | free_extent_buffer(path->nodes[i]); |
0f82731f | 2220 | path->nodes[i] = NULL; |
83e15a28 | 2221 | } |
20524f02 | 2222 | } |
9f3a7427 | 2223 | out: |
5caf2a00 | 2224 | btrfs_free_path(path); |
9aca1d51 | 2225 | return ret; |
20524f02 | 2226 | } |
9078a3e1 | 2227 | |
96b5179d | 2228 | int btrfs_free_block_groups(struct btrfs_fs_info *info) |
9078a3e1 | 2229 | { |
96b5179d CM |
2230 | u64 start; |
2231 | u64 end; | |
b97f9203 | 2232 | u64 ptr; |
9078a3e1 | 2233 | int ret; |
9078a3e1 | 2234 | while(1) { |
96b5179d CM |
2235 | ret = find_first_extent_bit(&info->block_group_cache, 0, |
2236 | &start, &end, (unsigned int)-1); | |
2237 | if (ret) | |
9078a3e1 | 2238 | break; |
b97f9203 Y |
2239 | ret = get_state_private(&info->block_group_cache, start, &ptr); |
2240 | if (!ret) | |
2241 | kfree((void *)(unsigned long)ptr); | |
96b5179d CM |
2242 | clear_extent_bits(&info->block_group_cache, start, |
2243 | end, (unsigned int)-1, GFP_NOFS); | |
9078a3e1 | 2244 | } |
e37c9e69 | 2245 | while(1) { |
f510cfec CM |
2246 | ret = find_first_extent_bit(&info->free_space_cache, 0, |
2247 | &start, &end, EXTENT_DIRTY); | |
2248 | if (ret) | |
e37c9e69 | 2249 | break; |
f510cfec CM |
2250 | clear_extent_dirty(&info->free_space_cache, start, |
2251 | end, GFP_NOFS); | |
e37c9e69 | 2252 | } |
be744175 CM |
2253 | return 0; |
2254 | } | |
2255 | ||
98ed5174 CM |
2256 | static int noinline relocate_inode_pages(struct inode *inode, u64 start, |
2257 | u64 len) | |
edbd8d4e CM |
2258 | { |
2259 | u64 page_start; | |
2260 | u64 page_end; | |
2261 | u64 delalloc_start; | |
2262 | u64 existing_delalloc; | |
2263 | unsigned long last_index; | |
edbd8d4e CM |
2264 | unsigned long i; |
2265 | struct page *page; | |
d1310b2e | 2266 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
4313b399 CM |
2267 | struct file_ra_state *ra; |
2268 | ||
2269 | ra = kzalloc(sizeof(*ra), GFP_NOFS); | |
edbd8d4e CM |
2270 | |
2271 | mutex_lock(&inode->i_mutex); | |
4313b399 | 2272 | i = start >> PAGE_CACHE_SHIFT; |
edbd8d4e CM |
2273 | last_index = (start + len - 1) >> PAGE_CACHE_SHIFT; |
2274 | ||
4313b399 CM |
2275 | file_ra_state_init(ra, inode->i_mapping); |
2276 | btrfs_force_ra(inode->i_mapping, ra, NULL, i, last_index); | |
2277 | kfree(ra); | |
edbd8d4e | 2278 | |
4313b399 | 2279 | for (; i <= last_index; i++) { |
edbd8d4e CM |
2280 | page = grab_cache_page(inode->i_mapping, i); |
2281 | if (!page) | |
2282 | goto out_unlock; | |
2283 | if (!PageUptodate(page)) { | |
2284 | btrfs_readpage(NULL, page); | |
2285 | lock_page(page); | |
2286 | if (!PageUptodate(page)) { | |
2287 | unlock_page(page); | |
2288 | page_cache_release(page); | |
2289 | goto out_unlock; | |
2290 | } | |
2291 | } | |
2292 | page_start = (u64)page->index << PAGE_CACHE_SHIFT; | |
2293 | page_end = page_start + PAGE_CACHE_SIZE - 1; | |
2294 | ||
d1310b2e | 2295 | lock_extent(io_tree, page_start, page_end, GFP_NOFS); |
edbd8d4e CM |
2296 | |
2297 | delalloc_start = page_start; | |
d1310b2e CM |
2298 | existing_delalloc = count_range_bits(io_tree, |
2299 | &delalloc_start, page_end, | |
2300 | PAGE_CACHE_SIZE, EXTENT_DELALLOC); | |
edbd8d4e | 2301 | |
d1310b2e | 2302 | set_extent_delalloc(io_tree, page_start, |
edbd8d4e CM |
2303 | page_end, GFP_NOFS); |
2304 | ||
d1310b2e | 2305 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); |
edbd8d4e CM |
2306 | set_page_dirty(page); |
2307 | unlock_page(page); | |
2308 | page_cache_release(page); | |
2309 | } | |
2310 | ||
2311 | out_unlock: | |
2312 | mutex_unlock(&inode->i_mutex); | |
2313 | return 0; | |
2314 | } | |
2315 | ||
4313b399 CM |
2316 | /* |
2317 | * note, this releases the path | |
2318 | */ | |
98ed5174 | 2319 | static int noinline relocate_one_reference(struct btrfs_root *extent_root, |
edbd8d4e | 2320 | struct btrfs_path *path, |
4313b399 | 2321 | struct btrfs_key *extent_key) |
edbd8d4e CM |
2322 | { |
2323 | struct inode *inode; | |
2324 | struct btrfs_root *found_root; | |
4313b399 CM |
2325 | struct btrfs_key *root_location; |
2326 | struct btrfs_extent_ref *ref; | |
2327 | u64 ref_root; | |
2328 | u64 ref_gen; | |
2329 | u64 ref_objectid; | |
2330 | u64 ref_offset; | |
edbd8d4e CM |
2331 | int ret; |
2332 | ||
4313b399 CM |
2333 | ref = btrfs_item_ptr(path->nodes[0], path->slots[0], |
2334 | struct btrfs_extent_ref); | |
2335 | ref_root = btrfs_ref_root(path->nodes[0], ref); | |
2336 | ref_gen = btrfs_ref_generation(path->nodes[0], ref); | |
2337 | ref_objectid = btrfs_ref_objectid(path->nodes[0], ref); | |
2338 | ref_offset = btrfs_ref_offset(path->nodes[0], ref); | |
2339 | btrfs_release_path(extent_root, path); | |
2340 | ||
2341 | root_location = kmalloc(sizeof(*root_location), GFP_NOFS); | |
2342 | root_location->objectid = ref_root; | |
edbd8d4e | 2343 | if (ref_gen == 0) |
4313b399 | 2344 | root_location->offset = 0; |
edbd8d4e | 2345 | else |
4313b399 CM |
2346 | root_location->offset = (u64)-1; |
2347 | root_location->type = BTRFS_ROOT_ITEM_KEY; | |
edbd8d4e CM |
2348 | |
2349 | found_root = btrfs_read_fs_root_no_name(extent_root->fs_info, | |
4313b399 | 2350 | root_location); |
edbd8d4e | 2351 | BUG_ON(!found_root); |
4313b399 | 2352 | kfree(root_location); |
edbd8d4e CM |
2353 | |
2354 | if (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID) { | |
2355 | mutex_unlock(&extent_root->fs_info->fs_mutex); | |
2356 | inode = btrfs_iget_locked(extent_root->fs_info->sb, | |
2357 | ref_objectid, found_root); | |
2358 | if (inode->i_state & I_NEW) { | |
2359 | /* the inode and parent dir are two different roots */ | |
2360 | BTRFS_I(inode)->root = found_root; | |
2361 | BTRFS_I(inode)->location.objectid = ref_objectid; | |
2362 | BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY; | |
2363 | BTRFS_I(inode)->location.offset = 0; | |
2364 | btrfs_read_locked_inode(inode); | |
2365 | unlock_new_inode(inode); | |
2366 | ||
2367 | } | |
2368 | /* this can happen if the reference is not against | |
2369 | * the latest version of the tree root | |
2370 | */ | |
2371 | if (is_bad_inode(inode)) { | |
2372 | mutex_lock(&extent_root->fs_info->fs_mutex); | |
2373 | goto out; | |
2374 | } | |
2375 | relocate_inode_pages(inode, ref_offset, extent_key->offset); | |
2376 | /* FIXME, data=ordered will help get rid of this */ | |
2377 | filemap_fdatawrite(inode->i_mapping); | |
2378 | iput(inode); | |
2379 | mutex_lock(&extent_root->fs_info->fs_mutex); | |
2380 | } else { | |
2381 | struct btrfs_trans_handle *trans; | |
2382 | struct btrfs_key found_key; | |
2383 | struct extent_buffer *eb; | |
2384 | int level; | |
2385 | int i; | |
2386 | ||
2387 | trans = btrfs_start_transaction(found_root, 1); | |
2388 | eb = read_tree_block(found_root, extent_key->objectid, | |
2389 | extent_key->offset); | |
2390 | level = btrfs_header_level(eb); | |
2391 | ||
2392 | if (level == 0) | |
2393 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
2394 | else | |
2395 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
2396 | ||
2397 | free_extent_buffer(eb); | |
2398 | ||
2399 | path->lowest_level = level; | |
8f662a76 | 2400 | path->reada = 2; |
edbd8d4e CM |
2401 | ret = btrfs_search_slot(trans, found_root, &found_key, path, |
2402 | 0, 1); | |
2403 | path->lowest_level = 0; | |
2404 | for (i = level; i < BTRFS_MAX_LEVEL; i++) { | |
2405 | if (!path->nodes[i]) | |
2406 | break; | |
2407 | free_extent_buffer(path->nodes[i]); | |
2408 | path->nodes[i] = NULL; | |
2409 | } | |
2410 | btrfs_release_path(found_root, path); | |
2411 | btrfs_end_transaction(trans, found_root); | |
2412 | } | |
2413 | ||
2414 | out: | |
2415 | return 0; | |
2416 | } | |
2417 | ||
98ed5174 CM |
2418 | static int noinline relocate_one_extent(struct btrfs_root *extent_root, |
2419 | struct btrfs_path *path, | |
2420 | struct btrfs_key *extent_key) | |
edbd8d4e CM |
2421 | { |
2422 | struct btrfs_key key; | |
2423 | struct btrfs_key found_key; | |
edbd8d4e | 2424 | struct extent_buffer *leaf; |
edbd8d4e CM |
2425 | u32 nritems; |
2426 | u32 item_size; | |
2427 | int ret = 0; | |
2428 | ||
2429 | key.objectid = extent_key->objectid; | |
2430 | key.type = BTRFS_EXTENT_REF_KEY; | |
2431 | key.offset = 0; | |
2432 | ||
2433 | while(1) { | |
2434 | ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); | |
2435 | ||
edbd8d4e CM |
2436 | if (ret < 0) |
2437 | goto out; | |
2438 | ||
2439 | ret = 0; | |
2440 | leaf = path->nodes[0]; | |
2441 | nritems = btrfs_header_nritems(leaf); | |
2442 | if (path->slots[0] == nritems) | |
2443 | goto out; | |
2444 | ||
2445 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2446 | if (found_key.objectid != extent_key->objectid) | |
2447 | break; | |
2448 | ||
2449 | if (found_key.type != BTRFS_EXTENT_REF_KEY) | |
2450 | break; | |
2451 | ||
2452 | key.offset = found_key.offset + 1; | |
2453 | item_size = btrfs_item_size_nr(leaf, path->slots[0]); | |
2454 | ||
4313b399 | 2455 | ret = relocate_one_reference(extent_root, path, extent_key); |
edbd8d4e CM |
2456 | if (ret) |
2457 | goto out; | |
2458 | } | |
2459 | ret = 0; | |
2460 | out: | |
2461 | btrfs_release_path(extent_root, path); | |
2462 | return ret; | |
2463 | } | |
2464 | ||
edbd8d4e CM |
2465 | int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 new_size) |
2466 | { | |
2467 | struct btrfs_trans_handle *trans; | |
2468 | struct btrfs_root *tree_root = root->fs_info->tree_root; | |
2469 | struct btrfs_path *path; | |
2470 | u64 cur_byte; | |
2471 | u64 total_found; | |
edbd8d4e | 2472 | struct btrfs_fs_info *info = root->fs_info; |
d1310b2e | 2473 | struct extent_io_tree *block_group_cache; |
edbd8d4e | 2474 | struct btrfs_key key; |
73e48b27 | 2475 | struct btrfs_key found_key; |
edbd8d4e CM |
2476 | struct extent_buffer *leaf; |
2477 | u32 nritems; | |
2478 | int ret; | |
725c8463 | 2479 | int progress = 0; |
edbd8d4e CM |
2480 | |
2481 | btrfs_set_super_total_bytes(&info->super_copy, new_size); | |
c31f8830 CM |
2482 | clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1, |
2483 | GFP_NOFS); | |
edbd8d4e CM |
2484 | block_group_cache = &info->block_group_cache; |
2485 | path = btrfs_alloc_path(); | |
2486 | root = root->fs_info->extent_root; | |
8f662a76 | 2487 | path->reada = 2; |
edbd8d4e CM |
2488 | |
2489 | again: | |
2490 | total_found = 0; | |
2491 | key.objectid = new_size; | |
edbd8d4e CM |
2492 | key.offset = 0; |
2493 | key.type = 0; | |
73e48b27 | 2494 | cur_byte = key.objectid; |
4313b399 | 2495 | |
73e48b27 Y |
2496 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2497 | if (ret < 0) | |
2498 | goto out; | |
2499 | ||
0b86a832 | 2500 | ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY); |
73e48b27 Y |
2501 | if (ret < 0) |
2502 | goto out; | |
2503 | if (ret == 0) { | |
2504 | leaf = path->nodes[0]; | |
2505 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2506 | if (found_key.objectid + found_key.offset > new_size) { | |
2507 | cur_byte = found_key.objectid; | |
2508 | key.objectid = cur_byte; | |
2509 | } | |
2510 | } | |
2511 | btrfs_release_path(root, path); | |
2512 | ||
2513 | while(1) { | |
edbd8d4e CM |
2514 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2515 | if (ret < 0) | |
2516 | goto out; | |
73e48b27 | 2517 | |
edbd8d4e | 2518 | leaf = path->nodes[0]; |
73e48b27 Y |
2519 | nritems = btrfs_header_nritems(leaf); |
2520 | next: | |
2521 | if (path->slots[0] >= nritems) { | |
2522 | ret = btrfs_next_leaf(root, path); | |
2523 | if (ret < 0) | |
2524 | goto out; | |
2525 | if (ret == 1) { | |
2526 | ret = 0; | |
2527 | break; | |
edbd8d4e | 2528 | } |
73e48b27 Y |
2529 | leaf = path->nodes[0]; |
2530 | nritems = btrfs_header_nritems(leaf); | |
edbd8d4e | 2531 | } |
73e48b27 Y |
2532 | |
2533 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
725c8463 CM |
2534 | |
2535 | if (progress && need_resched()) { | |
2536 | memcpy(&key, &found_key, sizeof(key)); | |
2537 | mutex_unlock(&root->fs_info->fs_mutex); | |
2538 | cond_resched(); | |
2539 | mutex_lock(&root->fs_info->fs_mutex); | |
2540 | btrfs_release_path(root, path); | |
2541 | btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
2542 | progress = 0; | |
2543 | goto next; | |
2544 | } | |
2545 | progress = 1; | |
2546 | ||
73e48b27 Y |
2547 | if (btrfs_key_type(&found_key) != BTRFS_EXTENT_ITEM_KEY || |
2548 | found_key.objectid + found_key.offset <= cur_byte) { | |
edbd8d4e | 2549 | path->slots[0]++; |
edbd8d4e CM |
2550 | goto next; |
2551 | } | |
73e48b27 | 2552 | |
edbd8d4e CM |
2553 | total_found++; |
2554 | cur_byte = found_key.objectid + found_key.offset; | |
2555 | key.objectid = cur_byte; | |
2556 | btrfs_release_path(root, path); | |
2557 | ret = relocate_one_extent(root, path, &found_key); | |
2558 | } | |
2559 | ||
2560 | btrfs_release_path(root, path); | |
2561 | ||
2562 | if (total_found > 0) { | |
2563 | trans = btrfs_start_transaction(tree_root, 1); | |
2564 | btrfs_commit_transaction(trans, tree_root); | |
2565 | ||
2566 | mutex_unlock(&root->fs_info->fs_mutex); | |
2567 | btrfs_clean_old_snapshots(tree_root); | |
2568 | mutex_lock(&root->fs_info->fs_mutex); | |
2569 | ||
2570 | trans = btrfs_start_transaction(tree_root, 1); | |
2571 | btrfs_commit_transaction(trans, tree_root); | |
2572 | goto again; | |
2573 | } | |
2574 | ||
2575 | trans = btrfs_start_transaction(root, 1); | |
2576 | key.objectid = new_size; | |
2577 | key.offset = 0; | |
2578 | key.type = 0; | |
2579 | while(1) { | |
4313b399 CM |
2580 | u64 ptr; |
2581 | ||
edbd8d4e CM |
2582 | ret = btrfs_search_slot(trans, root, &key, path, -1, 1); |
2583 | if (ret < 0) | |
2584 | goto out; | |
73e48b27 | 2585 | |
edbd8d4e CM |
2586 | leaf = path->nodes[0]; |
2587 | nritems = btrfs_header_nritems(leaf); | |
73e48b27 Y |
2588 | bg_next: |
2589 | if (path->slots[0] >= nritems) { | |
2590 | ret = btrfs_next_leaf(root, path); | |
2591 | if (ret < 0) | |
2592 | break; | |
2593 | if (ret == 1) { | |
2594 | ret = 0; | |
2595 | break; | |
edbd8d4e | 2596 | } |
73e48b27 | 2597 | leaf = path->nodes[0]; |
1372f8e6 CM |
2598 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); |
2599 | ||
2600 | /* | |
2601 | * btrfs_next_leaf doesn't cow buffers, we have to | |
2602 | * do the search again | |
2603 | */ | |
2604 | memcpy(&key, &found_key, sizeof(key)); | |
2605 | btrfs_release_path(root, path); | |
725c8463 | 2606 | goto resched_check; |
73e48b27 Y |
2607 | } |
2608 | ||
2609 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2610 | if (btrfs_key_type(&found_key) != BTRFS_BLOCK_GROUP_ITEM_KEY) { | |
2611 | printk("shrinker found key %Lu %u %Lu\n", | |
2612 | found_key.objectid, found_key.type, | |
2613 | found_key.offset); | |
2614 | path->slots[0]++; | |
edbd8d4e CM |
2615 | goto bg_next; |
2616 | } | |
edbd8d4e CM |
2617 | ret = get_state_private(&info->block_group_cache, |
2618 | found_key.objectid, &ptr); | |
2619 | if (!ret) | |
2620 | kfree((void *)(unsigned long)ptr); | |
2621 | ||
2622 | clear_extent_bits(&info->block_group_cache, found_key.objectid, | |
2623 | found_key.objectid + found_key.offset - 1, | |
2624 | (unsigned int)-1, GFP_NOFS); | |
2625 | ||
2626 | key.objectid = found_key.objectid + 1; | |
2627 | btrfs_del_item(trans, root, path); | |
2628 | btrfs_release_path(root, path); | |
725c8463 CM |
2629 | resched_check: |
2630 | if (need_resched()) { | |
2631 | mutex_unlock(&root->fs_info->fs_mutex); | |
2632 | cond_resched(); | |
2633 | mutex_lock(&root->fs_info->fs_mutex); | |
2634 | } | |
edbd8d4e CM |
2635 | } |
2636 | clear_extent_dirty(&info->free_space_cache, new_size, (u64)-1, | |
2637 | GFP_NOFS); | |
2638 | btrfs_commit_transaction(trans, root); | |
2639 | out: | |
2640 | btrfs_free_path(path); | |
2641 | return ret; | |
2642 | } | |
2643 | ||
2644 | int btrfs_grow_extent_tree(struct btrfs_trans_handle *trans, | |
2645 | struct btrfs_root *root, u64 new_size) | |
2646 | { | |
0b86a832 CM |
2647 | btrfs_set_super_total_bytes(&root->fs_info->super_copy, new_size); |
2648 | return 0; | |
2649 | } | |
edbd8d4e | 2650 | |
0b86a832 CM |
2651 | int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path, |
2652 | struct btrfs_key *key) | |
2653 | { | |
2654 | int ret; | |
2655 | struct btrfs_key found_key; | |
2656 | struct extent_buffer *leaf; | |
2657 | int slot; | |
edbd8d4e | 2658 | |
0b86a832 CM |
2659 | ret = btrfs_search_slot(NULL, root, key, path, 0, 0); |
2660 | if (ret < 0) | |
2661 | return ret; | |
2662 | while(1) { | |
2663 | slot = path->slots[0]; | |
edbd8d4e | 2664 | leaf = path->nodes[0]; |
0b86a832 CM |
2665 | if (slot >= btrfs_header_nritems(leaf)) { |
2666 | ret = btrfs_next_leaf(root, path); | |
2667 | if (ret == 0) | |
2668 | continue; | |
2669 | if (ret < 0) | |
2670 | goto error; | |
2671 | break; | |
edbd8d4e | 2672 | } |
0b86a832 | 2673 | btrfs_item_key_to_cpu(leaf, &found_key, slot); |
edbd8d4e | 2674 | |
0b86a832 CM |
2675 | if (found_key.objectid >= key->objectid && |
2676 | found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) | |
2677 | return 0; | |
2678 | path->slots[0]++; | |
edbd8d4e | 2679 | } |
0b86a832 CM |
2680 | ret = -ENOENT; |
2681 | error: | |
2682 | return ret; | |
edbd8d4e CM |
2683 | } |
2684 | ||
9078a3e1 CM |
2685 | int btrfs_read_block_groups(struct btrfs_root *root) |
2686 | { | |
2687 | struct btrfs_path *path; | |
2688 | int ret; | |
96b5179d | 2689 | int bit; |
9078a3e1 | 2690 | struct btrfs_block_group_cache *cache; |
be744175 | 2691 | struct btrfs_fs_info *info = root->fs_info; |
6324fbf3 | 2692 | struct btrfs_space_info *space_info; |
d1310b2e | 2693 | struct extent_io_tree *block_group_cache; |
9078a3e1 CM |
2694 | struct btrfs_key key; |
2695 | struct btrfs_key found_key; | |
5f39d397 | 2696 | struct extent_buffer *leaf; |
96b5179d CM |
2697 | |
2698 | block_group_cache = &info->block_group_cache; | |
be744175 | 2699 | root = info->extent_root; |
9078a3e1 | 2700 | key.objectid = 0; |
0b86a832 | 2701 | key.offset = 0; |
9078a3e1 | 2702 | btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); |
9078a3e1 CM |
2703 | path = btrfs_alloc_path(); |
2704 | if (!path) | |
2705 | return -ENOMEM; | |
2706 | ||
2707 | while(1) { | |
0b86a832 CM |
2708 | ret = find_first_block_group(root, path, &key); |
2709 | if (ret > 0) { | |
2710 | ret = 0; | |
2711 | goto error; | |
9078a3e1 | 2712 | } |
0b86a832 CM |
2713 | if (ret != 0) |
2714 | goto error; | |
2715 | ||
5f39d397 CM |
2716 | leaf = path->nodes[0]; |
2717 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
9078a3e1 CM |
2718 | cache = kmalloc(sizeof(*cache), GFP_NOFS); |
2719 | if (!cache) { | |
0b86a832 | 2720 | ret = -ENOMEM; |
9078a3e1 CM |
2721 | break; |
2722 | } | |
3e1ad54f | 2723 | |
5f39d397 CM |
2724 | read_extent_buffer(leaf, &cache->item, |
2725 | btrfs_item_ptr_offset(leaf, path->slots[0]), | |
2726 | sizeof(cache->item)); | |
9078a3e1 | 2727 | memcpy(&cache->key, &found_key, sizeof(found_key)); |
e37c9e69 | 2728 | cache->cached = 0; |
324ae4df | 2729 | cache->pinned = 0; |
0b86a832 | 2730 | |
9078a3e1 CM |
2731 | key.objectid = found_key.objectid + found_key.offset; |
2732 | btrfs_release_path(root, path); | |
0b86a832 CM |
2733 | cache->flags = btrfs_block_group_flags(&cache->item); |
2734 | bit = 0; | |
2735 | if (cache->flags & BTRFS_BLOCK_GROUP_DATA) { | |
96b5179d | 2736 | bit = BLOCK_GROUP_DATA; |
0b86a832 CM |
2737 | } else if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) { |
2738 | bit = BLOCK_GROUP_SYSTEM; | |
2739 | } else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { | |
96b5179d | 2740 | bit = BLOCK_GROUP_METADATA; |
31f3c99b | 2741 | } |
593060d7 CM |
2742 | if (cache->flags & BTRFS_BLOCK_GROUP_RAID0) { |
2743 | if (cache->flags & BTRFS_BLOCK_GROUP_DATA) { | |
2744 | info->extra_data_alloc_bits = | |
2745 | BTRFS_BLOCK_GROUP_RAID0; | |
2746 | } | |
2747 | if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { | |
2748 | info->extra_alloc_bits = | |
2749 | BTRFS_BLOCK_GROUP_RAID0; | |
2750 | } | |
2751 | } | |
96b5179d | 2752 | |
6324fbf3 CM |
2753 | ret = update_space_info(info, cache->flags, found_key.offset, |
2754 | btrfs_block_group_used(&cache->item), | |
2755 | &space_info); | |
2756 | BUG_ON(ret); | |
2757 | cache->space_info = space_info; | |
2758 | ||
96b5179d CM |
2759 | /* use EXTENT_LOCKED to prevent merging */ |
2760 | set_extent_bits(block_group_cache, found_key.objectid, | |
2761 | found_key.objectid + found_key.offset - 1, | |
2762 | bit | EXTENT_LOCKED, GFP_NOFS); | |
2763 | set_state_private(block_group_cache, found_key.objectid, | |
ae2f5411 | 2764 | (unsigned long)cache); |
96b5179d | 2765 | |
9078a3e1 | 2766 | if (key.objectid >= |
db94535d | 2767 | btrfs_super_total_bytes(&info->super_copy)) |
9078a3e1 CM |
2768 | break; |
2769 | } | |
0b86a832 CM |
2770 | ret = 0; |
2771 | error: | |
9078a3e1 | 2772 | btrfs_free_path(path); |
0b86a832 | 2773 | return ret; |
9078a3e1 | 2774 | } |
6324fbf3 CM |
2775 | |
2776 | int btrfs_make_block_group(struct btrfs_trans_handle *trans, | |
2777 | struct btrfs_root *root, u64 bytes_used, | |
2778 | u64 type, u64 chunk_tree, u64 chunk_objectid, | |
2779 | u64 size) | |
2780 | { | |
2781 | int ret; | |
2782 | int bit = 0; | |
2783 | struct btrfs_root *extent_root; | |
2784 | struct btrfs_block_group_cache *cache; | |
2785 | struct extent_io_tree *block_group_cache; | |
2786 | ||
2787 | extent_root = root->fs_info->extent_root; | |
2788 | block_group_cache = &root->fs_info->block_group_cache; | |
2789 | ||
2790 | cache = kmalloc(sizeof(*cache), GFP_NOFS); | |
2791 | BUG_ON(!cache); | |
2792 | cache->key.objectid = chunk_objectid; | |
2793 | cache->key.offset = size; | |
2794 | cache->cached = 0; | |
2795 | cache->pinned = 0; | |
2796 | btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY); | |
2797 | memset(&cache->item, 0, sizeof(cache->item)); | |
2798 | btrfs_set_block_group_used(&cache->item, bytes_used); | |
2799 | btrfs_set_block_group_chunk_tree(&cache->item, chunk_tree); | |
2800 | btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid); | |
2801 | cache->flags = type; | |
2802 | btrfs_set_block_group_flags(&cache->item, type); | |
2803 | ||
2804 | ret = update_space_info(root->fs_info, cache->flags, size, bytes_used, | |
2805 | &cache->space_info); | |
2806 | BUG_ON(ret); | |
2807 | ||
2808 | if (type & BTRFS_BLOCK_GROUP_DATA) { | |
2809 | bit = BLOCK_GROUP_DATA; | |
2810 | } else if (type & BTRFS_BLOCK_GROUP_SYSTEM) { | |
2811 | bit = BLOCK_GROUP_SYSTEM; | |
2812 | } else if (type & BTRFS_BLOCK_GROUP_METADATA) { | |
2813 | bit = BLOCK_GROUP_METADATA; | |
2814 | } | |
2815 | set_extent_bits(block_group_cache, chunk_objectid, | |
2816 | chunk_objectid + size - 1, | |
2817 | bit | EXTENT_LOCKED, GFP_NOFS); | |
2818 | set_state_private(block_group_cache, chunk_objectid, | |
2819 | (unsigned long)cache); | |
2820 | ||
2821 | ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item, | |
2822 | sizeof(cache->item)); | |
2823 | BUG_ON(ret); | |
2824 | ||
2825 | finish_current_insert(trans, extent_root); | |
2826 | ret = del_pending_extents(trans, extent_root); | |
2827 | BUG_ON(ret); | |
2828 | return 0; | |
2829 | } |