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 | */ | |
ec6b910f | 18 | #include <linux/sched.h> |
edbd8d4e | 19 | #include <linux/pagemap.h> |
74493f7a | 20 | #include "hash.h" |
a5eb62e3 | 21 | #include "crc32c.h" |
fec577fb CM |
22 | #include "ctree.h" |
23 | #include "disk-io.h" | |
24 | #include "print-tree.h" | |
e089f05c | 25 | #include "transaction.h" |
0b86a832 | 26 | #include "volumes.h" |
fec577fb | 27 | |
0b86a832 | 28 | #define BLOCK_GROUP_DATA EXTENT_WRITEBACK |
96b5179d | 29 | #define BLOCK_GROUP_METADATA EXTENT_UPTODATE |
0b86a832 CM |
30 | #define BLOCK_GROUP_SYSTEM EXTENT_NEW |
31 | ||
96b5179d CM |
32 | #define BLOCK_GROUP_DIRTY EXTENT_DIRTY |
33 | ||
e089f05c CM |
34 | static int finish_current_insert(struct btrfs_trans_handle *trans, struct |
35 | btrfs_root *extent_root); | |
e20d96d6 CM |
36 | static int del_pending_extents(struct btrfs_trans_handle *trans, struct |
37 | btrfs_root *extent_root); | |
d548ee51 | 38 | |
fec577fb | 39 | |
e37c9e69 CM |
40 | static int cache_block_group(struct btrfs_root *root, |
41 | struct btrfs_block_group_cache *block_group) | |
42 | { | |
43 | struct btrfs_path *path; | |
44 | int ret; | |
45 | struct btrfs_key key; | |
5f39d397 | 46 | struct extent_buffer *leaf; |
d1310b2e | 47 | struct extent_io_tree *free_space_cache; |
e37c9e69 | 48 | int slot; |
e37c9e69 CM |
49 | u64 last = 0; |
50 | u64 hole_size; | |
7d7d6068 | 51 | u64 first_free; |
e37c9e69 CM |
52 | int found = 0; |
53 | ||
00f5c795 CM |
54 | if (!block_group) |
55 | return 0; | |
56 | ||
e37c9e69 | 57 | root = root->fs_info->extent_root; |
f510cfec | 58 | free_space_cache = &root->fs_info->free_space_cache; |
e37c9e69 CM |
59 | |
60 | if (block_group->cached) | |
61 | return 0; | |
f510cfec | 62 | |
e37c9e69 CM |
63 | path = btrfs_alloc_path(); |
64 | if (!path) | |
65 | return -ENOMEM; | |
7d7d6068 | 66 | |
2cc58cf2 | 67 | path->reada = 2; |
7d7d6068 | 68 | first_free = block_group->key.objectid; |
e37c9e69 | 69 | key.objectid = block_group->key.objectid; |
e37c9e69 CM |
70 | key.offset = 0; |
71 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); | |
72 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
73 | if (ret < 0) | |
74 | return ret; | |
0b86a832 | 75 | ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY); |
d548ee51 Y |
76 | if (ret < 0) |
77 | return ret; | |
78 | if (ret == 0) { | |
79 | leaf = path->nodes[0]; | |
80 | btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); | |
81 | if (key.objectid + key.offset > first_free) | |
82 | first_free = key.objectid + key.offset; | |
83 | } | |
e37c9e69 | 84 | while(1) { |
5f39d397 | 85 | leaf = path->nodes[0]; |
e37c9e69 | 86 | slot = path->slots[0]; |
5f39d397 | 87 | if (slot >= btrfs_header_nritems(leaf)) { |
e37c9e69 | 88 | ret = btrfs_next_leaf(root, path); |
54aa1f4d CM |
89 | if (ret < 0) |
90 | goto err; | |
de428b63 | 91 | if (ret == 0) { |
e37c9e69 | 92 | continue; |
de428b63 | 93 | } else { |
e37c9e69 CM |
94 | break; |
95 | } | |
96 | } | |
5f39d397 | 97 | btrfs_item_key_to_cpu(leaf, &key, slot); |
7d7d6068 | 98 | if (key.objectid < block_group->key.objectid) { |
7d7d6068 Y |
99 | goto next; |
100 | } | |
e37c9e69 CM |
101 | if (key.objectid >= block_group->key.objectid + |
102 | block_group->key.offset) { | |
e37c9e69 CM |
103 | break; |
104 | } | |
7d7d6068 | 105 | |
e37c9e69 CM |
106 | if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) { |
107 | if (!found) { | |
7d7d6068 | 108 | last = first_free; |
e37c9e69 | 109 | found = 1; |
e37c9e69 | 110 | } |
f510cfec CM |
111 | if (key.objectid > last) { |
112 | hole_size = key.objectid - last; | |
113 | set_extent_dirty(free_space_cache, last, | |
114 | last + hole_size - 1, | |
115 | GFP_NOFS); | |
7d7d6068 Y |
116 | } |
117 | last = key.objectid + key.offset; | |
e37c9e69 | 118 | } |
7d7d6068 | 119 | next: |
e37c9e69 CM |
120 | path->slots[0]++; |
121 | } | |
122 | ||
7d7d6068 Y |
123 | if (!found) |
124 | last = first_free; | |
125 | if (block_group->key.objectid + | |
126 | block_group->key.offset > last) { | |
127 | hole_size = block_group->key.objectid + | |
128 | block_group->key.offset - last; | |
f510cfec CM |
129 | set_extent_dirty(free_space_cache, last, |
130 | last + hole_size - 1, GFP_NOFS); | |
7d7d6068 | 131 | } |
e37c9e69 | 132 | block_group->cached = 1; |
54aa1f4d | 133 | err: |
e37c9e69 CM |
134 | btrfs_free_path(path); |
135 | return 0; | |
136 | } | |
137 | ||
5276aeda CM |
138 | struct btrfs_block_group_cache *btrfs_lookup_block_group(struct |
139 | btrfs_fs_info *info, | |
db94535d | 140 | u64 bytenr) |
be744175 | 141 | { |
d1310b2e | 142 | struct extent_io_tree *block_group_cache; |
96b5179d CM |
143 | struct btrfs_block_group_cache *block_group = NULL; |
144 | u64 ptr; | |
145 | u64 start; | |
146 | u64 end; | |
be744175 CM |
147 | int ret; |
148 | ||
96b5179d CM |
149 | block_group_cache = &info->block_group_cache; |
150 | ret = find_first_extent_bit(block_group_cache, | |
db94535d | 151 | bytenr, &start, &end, |
0b86a832 CM |
152 | BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA | |
153 | BLOCK_GROUP_SYSTEM); | |
be744175 | 154 | if (ret) { |
96b5179d | 155 | return NULL; |
be744175 | 156 | } |
96b5179d CM |
157 | ret = get_state_private(block_group_cache, start, &ptr); |
158 | if (ret) | |
159 | return NULL; | |
160 | ||
ae2f5411 | 161 | block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr; |
5cf66426 | 162 | if (block_group->key.objectid <= bytenr && bytenr < |
96b5179d CM |
163 | block_group->key.objectid + block_group->key.offset) |
164 | return block_group; | |
be744175 CM |
165 | return NULL; |
166 | } | |
0b86a832 CM |
167 | |
168 | static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) | |
169 | { | |
593060d7 | 170 | return (cache->flags & bits) == bits; |
0b86a832 CM |
171 | } |
172 | ||
173 | static int noinline find_search_start(struct btrfs_root *root, | |
98ed5174 | 174 | struct btrfs_block_group_cache **cache_ret, |
0b86a832 | 175 | u64 *start_ret, int num, int data) |
e37c9e69 | 176 | { |
e37c9e69 CM |
177 | int ret; |
178 | struct btrfs_block_group_cache *cache = *cache_ret; | |
d7fc640e | 179 | struct extent_io_tree *free_space_cache; |
7d1660d4 | 180 | struct extent_state *state; |
e19caa5f | 181 | u64 last; |
f510cfec | 182 | u64 start = 0; |
257d0ce3 | 183 | u64 cache_miss = 0; |
c31f8830 | 184 | u64 total_fs_bytes; |
0b86a832 | 185 | u64 search_start = *start_ret; |
f84a8b36 | 186 | int wrapped = 0; |
e37c9e69 | 187 | |
0b86a832 | 188 | if (!cache) |
1a2b2ac7 | 189 | goto out; |
8f18cf13 | 190 | |
c31f8830 | 191 | total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); |
d7fc640e CM |
192 | free_space_cache = &root->fs_info->free_space_cache; |
193 | ||
e37c9e69 | 194 | again: |
54aa1f4d CM |
195 | ret = cache_block_group(root, cache); |
196 | if (ret) | |
197 | goto out; | |
f84a8b36 | 198 | |
e19caa5f | 199 | last = max(search_start, cache->key.objectid); |
8f18cf13 | 200 | if (!block_group_bits(cache, data) || cache->ro) { |
0b86a832 CM |
201 | goto new_group; |
202 | } | |
e19caa5f | 203 | |
7d1660d4 CM |
204 | spin_lock_irq(&free_space_cache->lock); |
205 | state = find_first_extent_bit_state(free_space_cache, last, EXTENT_DIRTY); | |
e37c9e69 | 206 | while(1) { |
7d1660d4 | 207 | if (!state) { |
257d0ce3 CM |
208 | if (!cache_miss) |
209 | cache_miss = last; | |
7d1660d4 | 210 | spin_unlock_irq(&free_space_cache->lock); |
e19caa5f CM |
211 | goto new_group; |
212 | } | |
f510cfec | 213 | |
7d1660d4 CM |
214 | start = max(last, state->start); |
215 | last = state->end + 1; | |
257d0ce3 CM |
216 | if (last - start < num) { |
217 | if (last == cache->key.objectid + cache->key.offset) | |
218 | cache_miss = start; | |
7d1660d4 CM |
219 | do { |
220 | state = extent_state_next(state); | |
221 | } while(state && !(state->state & EXTENT_DIRTY)); | |
f510cfec | 222 | continue; |
257d0ce3 | 223 | } |
7d1660d4 | 224 | spin_unlock_irq(&free_space_cache->lock); |
8f18cf13 CM |
225 | if (cache->ro) |
226 | goto new_group; | |
0b86a832 | 227 | if (start + num > cache->key.objectid + cache->key.offset) |
e37c9e69 | 228 | goto new_group; |
c31f8830 CM |
229 | if (start + num > total_fs_bytes) |
230 | goto new_group; | |
8790d502 | 231 | if (!block_group_bits(cache, data)) { |
611f0e00 | 232 | printk("block group bits don't match %Lu %d\n", cache->flags, data); |
8790d502 | 233 | } |
0b86a832 CM |
234 | *start_ret = start; |
235 | return 0; | |
8790d502 CM |
236 | } |
237 | out: | |
1a2b2ac7 CM |
238 | cache = btrfs_lookup_block_group(root->fs_info, search_start); |
239 | if (!cache) { | |
0b86a832 | 240 | printk("Unable to find block group for %Lu\n", search_start); |
1a2b2ac7 | 241 | WARN_ON(1); |
1a2b2ac7 | 242 | } |
0b86a832 | 243 | return -ENOSPC; |
e37c9e69 CM |
244 | |
245 | new_group: | |
e19caa5f | 246 | last = cache->key.objectid + cache->key.offset; |
f84a8b36 | 247 | wrapped: |
e19caa5f | 248 | cache = btrfs_lookup_block_group(root->fs_info, last); |
c31f8830 | 249 | if (!cache || cache->key.objectid >= total_fs_bytes) { |
0e4de584 | 250 | no_cache: |
f84a8b36 CM |
251 | if (!wrapped) { |
252 | wrapped = 1; | |
253 | last = search_start; | |
f84a8b36 CM |
254 | goto wrapped; |
255 | } | |
1a2b2ac7 | 256 | goto out; |
e37c9e69 | 257 | } |
257d0ce3 CM |
258 | if (cache_miss && !cache->cached) { |
259 | cache_block_group(root, cache); | |
260 | last = cache_miss; | |
257d0ce3 CM |
261 | cache = btrfs_lookup_block_group(root->fs_info, last); |
262 | } | |
1a2b2ac7 | 263 | cache = btrfs_find_block_group(root, cache, last, data, 0); |
0e4de584 CM |
264 | if (!cache) |
265 | goto no_cache; | |
e37c9e69 | 266 | *cache_ret = cache; |
257d0ce3 | 267 | cache_miss = 0; |
e37c9e69 CM |
268 | goto again; |
269 | } | |
270 | ||
84f54cfa CM |
271 | static u64 div_factor(u64 num, int factor) |
272 | { | |
257d0ce3 CM |
273 | if (factor == 10) |
274 | return num; | |
84f54cfa CM |
275 | num *= factor; |
276 | do_div(num, 10); | |
277 | return num; | |
278 | } | |
279 | ||
6324fbf3 CM |
280 | static int block_group_state_bits(u64 flags) |
281 | { | |
282 | int bits = 0; | |
283 | if (flags & BTRFS_BLOCK_GROUP_DATA) | |
284 | bits |= BLOCK_GROUP_DATA; | |
285 | if (flags & BTRFS_BLOCK_GROUP_METADATA) | |
286 | bits |= BLOCK_GROUP_METADATA; | |
287 | if (flags & BTRFS_BLOCK_GROUP_SYSTEM) | |
288 | bits |= BLOCK_GROUP_SYSTEM; | |
289 | return bits; | |
290 | } | |
291 | ||
31f3c99b CM |
292 | struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, |
293 | struct btrfs_block_group_cache | |
be744175 | 294 | *hint, u64 search_start, |
de428b63 | 295 | int data, int owner) |
cd1bc465 | 296 | { |
96b5179d | 297 | struct btrfs_block_group_cache *cache; |
d1310b2e | 298 | struct extent_io_tree *block_group_cache; |
31f3c99b | 299 | struct btrfs_block_group_cache *found_group = NULL; |
cd1bc465 CM |
300 | struct btrfs_fs_info *info = root->fs_info; |
301 | u64 used; | |
31f3c99b CM |
302 | u64 last = 0; |
303 | u64 hint_last; | |
96b5179d CM |
304 | u64 start; |
305 | u64 end; | |
306 | u64 free_check; | |
307 | u64 ptr; | |
c31f8830 | 308 | u64 total_fs_bytes; |
96b5179d | 309 | int bit; |
cd1bc465 | 310 | int ret; |
31f3c99b | 311 | int full_search = 0; |
bce4eae9 | 312 | int factor = 10; |
de428b63 | 313 | |
96b5179d | 314 | block_group_cache = &info->block_group_cache; |
c31f8830 | 315 | total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); |
96b5179d | 316 | |
de428b63 | 317 | if (!owner) |
bce4eae9 | 318 | factor = 10; |
be744175 | 319 | |
6324fbf3 | 320 | bit = block_group_state_bits(data); |
be744175 | 321 | |
c31f8830 | 322 | if (search_start && search_start < total_fs_bytes) { |
be744175 | 323 | struct btrfs_block_group_cache *shint; |
5276aeda | 324 | shint = btrfs_lookup_block_group(info, search_start); |
8f18cf13 | 325 | if (shint && block_group_bits(shint, data) && !shint->ro) { |
be744175 | 326 | used = btrfs_block_group_used(&shint->item); |
324ae4df Y |
327 | if (used + shint->pinned < |
328 | div_factor(shint->key.offset, factor)) { | |
be744175 CM |
329 | return shint; |
330 | } | |
331 | } | |
332 | } | |
8f18cf13 | 333 | if (hint && !hint->ro && block_group_bits(hint, data) && |
0b86a832 | 334 | hint->key.objectid < total_fs_bytes) { |
31f3c99b | 335 | used = btrfs_block_group_used(&hint->item); |
324ae4df Y |
336 | if (used + hint->pinned < |
337 | div_factor(hint->key.offset, factor)) { | |
31f3c99b CM |
338 | return hint; |
339 | } | |
e19caa5f | 340 | last = hint->key.objectid + hint->key.offset; |
31f3c99b CM |
341 | hint_last = last; |
342 | } else { | |
e37c9e69 CM |
343 | if (hint) |
344 | hint_last = max(hint->key.objectid, search_start); | |
345 | else | |
346 | hint_last = search_start; | |
347 | ||
c31f8830 CM |
348 | if (hint_last >= total_fs_bytes) |
349 | hint_last = search_start; | |
e37c9e69 | 350 | last = hint_last; |
31f3c99b | 351 | } |
31f3c99b | 352 | again: |
cd1bc465 | 353 | while(1) { |
96b5179d CM |
354 | ret = find_first_extent_bit(block_group_cache, last, |
355 | &start, &end, bit); | |
356 | if (ret) | |
cd1bc465 | 357 | break; |
96b5179d CM |
358 | |
359 | ret = get_state_private(block_group_cache, start, &ptr); | |
360 | if (ret) | |
361 | break; | |
362 | ||
ae2f5411 | 363 | cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; |
96b5179d CM |
364 | last = cache->key.objectid + cache->key.offset; |
365 | used = btrfs_block_group_used(&cache->item); | |
366 | ||
c31f8830 CM |
367 | if (cache->key.objectid > total_fs_bytes) |
368 | break; | |
369 | ||
8f18cf13 | 370 | if (!cache->ro && block_group_bits(cache, data)) { |
8790d502 CM |
371 | if (full_search) |
372 | free_check = cache->key.offset; | |
373 | else | |
374 | free_check = div_factor(cache->key.offset, | |
375 | factor); | |
6324fbf3 | 376 | |
8790d502 CM |
377 | if (used + cache->pinned < free_check) { |
378 | found_group = cache; | |
379 | goto found; | |
380 | } | |
6324fbf3 | 381 | } |
de428b63 | 382 | cond_resched(); |
cd1bc465 | 383 | } |
31f3c99b | 384 | if (!full_search) { |
be744175 | 385 | last = search_start; |
31f3c99b CM |
386 | full_search = 1; |
387 | goto again; | |
388 | } | |
be744175 | 389 | found: |
31f3c99b | 390 | return found_group; |
cd1bc465 CM |
391 | } |
392 | ||
7bb86316 | 393 | static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation, |
74493f7a CM |
394 | u64 owner, u64 owner_offset) |
395 | { | |
396 | u32 high_crc = ~(u32)0; | |
397 | u32 low_crc = ~(u32)0; | |
398 | __le64 lenum; | |
74493f7a | 399 | lenum = cpu_to_le64(root_objectid); |
a5eb62e3 | 400 | high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum)); |
7bb86316 | 401 | lenum = cpu_to_le64(ref_generation); |
a5eb62e3 | 402 | low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); |
21a4989d CM |
403 | if (owner >= BTRFS_FIRST_FREE_OBJECTID) { |
404 | lenum = cpu_to_le64(owner); | |
a5eb62e3 | 405 | low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); |
21a4989d | 406 | lenum = cpu_to_le64(owner_offset); |
a5eb62e3 | 407 | low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); |
21a4989d | 408 | } |
74493f7a CM |
409 | return ((u64)high_crc << 32) | (u64)low_crc; |
410 | } | |
411 | ||
7bb86316 CM |
412 | static int match_extent_ref(struct extent_buffer *leaf, |
413 | struct btrfs_extent_ref *disk_ref, | |
414 | struct btrfs_extent_ref *cpu_ref) | |
415 | { | |
416 | int ret; | |
417 | int len; | |
418 | ||
419 | if (cpu_ref->objectid) | |
420 | len = sizeof(*cpu_ref); | |
421 | else | |
422 | len = 2 * sizeof(u64); | |
423 | ret = memcmp_extent_buffer(leaf, cpu_ref, (unsigned long)disk_ref, | |
424 | len); | |
425 | return ret == 0; | |
426 | } | |
427 | ||
98ed5174 CM |
428 | static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, |
429 | struct btrfs_root *root, | |
430 | struct btrfs_path *path, u64 bytenr, | |
431 | u64 root_objectid, | |
432 | u64 ref_generation, u64 owner, | |
433 | u64 owner_offset, int del) | |
74493f7a CM |
434 | { |
435 | u64 hash; | |
436 | struct btrfs_key key; | |
7bb86316 | 437 | struct btrfs_key found_key; |
74493f7a | 438 | struct btrfs_extent_ref ref; |
7bb86316 CM |
439 | struct extent_buffer *leaf; |
440 | struct btrfs_extent_ref *disk_ref; | |
441 | int ret; | |
442 | int ret2; | |
443 | ||
444 | btrfs_set_stack_ref_root(&ref, root_objectid); | |
445 | btrfs_set_stack_ref_generation(&ref, ref_generation); | |
446 | btrfs_set_stack_ref_objectid(&ref, owner); | |
447 | btrfs_set_stack_ref_offset(&ref, owner_offset); | |
448 | ||
449 | hash = hash_extent_ref(root_objectid, ref_generation, owner, | |
450 | owner_offset); | |
451 | key.offset = hash; | |
452 | key.objectid = bytenr; | |
453 | key.type = BTRFS_EXTENT_REF_KEY; | |
454 | ||
455 | while (1) { | |
456 | ret = btrfs_search_slot(trans, root, &key, path, | |
457 | del ? -1 : 0, del); | |
458 | if (ret < 0) | |
459 | goto out; | |
460 | leaf = path->nodes[0]; | |
461 | if (ret != 0) { | |
462 | u32 nritems = btrfs_header_nritems(leaf); | |
463 | if (path->slots[0] >= nritems) { | |
464 | ret2 = btrfs_next_leaf(root, path); | |
465 | if (ret2) | |
466 | goto out; | |
467 | leaf = path->nodes[0]; | |
468 | } | |
469 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
470 | if (found_key.objectid != bytenr || | |
471 | found_key.type != BTRFS_EXTENT_REF_KEY) | |
472 | goto out; | |
473 | key.offset = found_key.offset; | |
474 | if (del) { | |
475 | btrfs_release_path(root, path); | |
476 | continue; | |
477 | } | |
478 | } | |
479 | disk_ref = btrfs_item_ptr(path->nodes[0], | |
480 | path->slots[0], | |
481 | struct btrfs_extent_ref); | |
482 | if (match_extent_ref(path->nodes[0], disk_ref, &ref)) { | |
483 | ret = 0; | |
484 | goto out; | |
485 | } | |
486 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
487 | key.offset = found_key.offset + 1; | |
488 | btrfs_release_path(root, path); | |
489 | } | |
490 | out: | |
491 | return ret; | |
492 | } | |
493 | ||
d8d5f3e1 CM |
494 | /* |
495 | * Back reference rules. Back refs have three main goals: | |
496 | * | |
497 | * 1) differentiate between all holders of references to an extent so that | |
498 | * when a reference is dropped we can make sure it was a valid reference | |
499 | * before freeing the extent. | |
500 | * | |
501 | * 2) Provide enough information to quickly find the holders of an extent | |
502 | * if we notice a given block is corrupted or bad. | |
503 | * | |
504 | * 3) Make it easy to migrate blocks for FS shrinking or storage pool | |
505 | * maintenance. This is actually the same as #2, but with a slightly | |
506 | * different use case. | |
507 | * | |
508 | * File extents can be referenced by: | |
509 | * | |
510 | * - multiple snapshots, subvolumes, or different generations in one subvol | |
511 | * - different files inside a single subvolume (in theory, not implemented yet) | |
512 | * - different offsets inside a file (bookend extents in file.c) | |
513 | * | |
514 | * The extent ref structure has fields for: | |
515 | * | |
516 | * - Objectid of the subvolume root | |
517 | * - Generation number of the tree holding the reference | |
518 | * - objectid of the file holding the reference | |
519 | * - offset in the file corresponding to the key holding the reference | |
520 | * | |
521 | * When a file extent is allocated the fields are filled in: | |
522 | * (root_key.objectid, trans->transid, inode objectid, offset in file) | |
523 | * | |
524 | * When a leaf is cow'd new references are added for every file extent found | |
525 | * in the leaf. It looks the same as the create case, but trans->transid | |
526 | * will be different when the block is cow'd. | |
527 | * | |
528 | * (root_key.objectid, trans->transid, inode objectid, offset in file) | |
529 | * | |
530 | * When a file extent is removed either during snapshot deletion or file | |
531 | * truncation, the corresponding back reference is found | |
532 | * by searching for: | |
533 | * | |
534 | * (btrfs_header_owner(leaf), btrfs_header_generation(leaf), | |
535 | * inode objectid, offset in file) | |
536 | * | |
537 | * Btree extents can be referenced by: | |
538 | * | |
539 | * - Different subvolumes | |
540 | * - Different generations of the same subvolume | |
541 | * | |
542 | * Storing sufficient information for a full reverse mapping of a btree | |
543 | * block would require storing the lowest key of the block in the backref, | |
544 | * and it would require updating that lowest key either before write out or | |
545 | * every time it changed. Instead, the objectid of the lowest key is stored | |
546 | * along with the level of the tree block. This provides a hint | |
547 | * about where in the btree the block can be found. Searches through the | |
548 | * btree only need to look for a pointer to that block, so they stop one | |
549 | * level higher than the level recorded in the backref. | |
550 | * | |
551 | * Some btrees do not do reference counting on their extents. These | |
552 | * include the extent tree and the tree of tree roots. Backrefs for these | |
553 | * trees always have a generation of zero. | |
554 | * | |
555 | * When a tree block is created, back references are inserted: | |
556 | * | |
f6dbff55 | 557 | * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid) |
d8d5f3e1 CM |
558 | * |
559 | * When a tree block is cow'd in a reference counted root, | |
560 | * new back references are added for all the blocks it points to. | |
561 | * These are of the form (trans->transid will have increased since creation): | |
562 | * | |
f6dbff55 | 563 | * (root->root_key.objectid, trans->transid, level, lowest_key_objectid) |
d8d5f3e1 CM |
564 | * |
565 | * Because the lowest_key_objectid and the level are just hints | |
566 | * they are not used when backrefs are deleted. When a backref is deleted: | |
567 | * | |
568 | * if backref was for a tree root: | |
569 | * root_objectid = root->root_key.objectid | |
570 | * else | |
571 | * root_objectid = btrfs_header_owner(parent) | |
572 | * | |
573 | * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0) | |
574 | * | |
575 | * Back Reference Key hashing: | |
576 | * | |
577 | * Back references have four fields, each 64 bits long. Unfortunately, | |
578 | * This is hashed into a single 64 bit number and placed into the key offset. | |
579 | * The key objectid corresponds to the first byte in the extent, and the | |
580 | * key type is set to BTRFS_EXTENT_REF_KEY | |
581 | */ | |
7bb86316 CM |
582 | int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, |
583 | struct btrfs_root *root, | |
584 | struct btrfs_path *path, u64 bytenr, | |
585 | u64 root_objectid, u64 ref_generation, | |
586 | u64 owner, u64 owner_offset) | |
587 | { | |
588 | u64 hash; | |
589 | struct btrfs_key key; | |
590 | struct btrfs_extent_ref ref; | |
591 | struct btrfs_extent_ref *disk_ref; | |
74493f7a CM |
592 | int ret; |
593 | ||
594 | btrfs_set_stack_ref_root(&ref, root_objectid); | |
7bb86316 | 595 | btrfs_set_stack_ref_generation(&ref, ref_generation); |
74493f7a CM |
596 | btrfs_set_stack_ref_objectid(&ref, owner); |
597 | btrfs_set_stack_ref_offset(&ref, owner_offset); | |
598 | ||
7bb86316 CM |
599 | hash = hash_extent_ref(root_objectid, ref_generation, owner, |
600 | owner_offset); | |
74493f7a CM |
601 | key.offset = hash; |
602 | key.objectid = bytenr; | |
603 | key.type = BTRFS_EXTENT_REF_KEY; | |
604 | ||
605 | ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(ref)); | |
606 | while (ret == -EEXIST) { | |
7bb86316 CM |
607 | disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], |
608 | struct btrfs_extent_ref); | |
609 | if (match_extent_ref(path->nodes[0], disk_ref, &ref)) | |
610 | goto out; | |
611 | key.offset++; | |
612 | btrfs_release_path(root, path); | |
613 | ret = btrfs_insert_empty_item(trans, root, path, &key, | |
614 | sizeof(ref)); | |
74493f7a | 615 | } |
7bb86316 CM |
616 | if (ret) |
617 | goto out; | |
618 | disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], | |
619 | struct btrfs_extent_ref); | |
620 | write_extent_buffer(path->nodes[0], &ref, (unsigned long)disk_ref, | |
621 | sizeof(ref)); | |
622 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
623 | out: | |
624 | btrfs_release_path(root, path); | |
625 | return ret; | |
74493f7a CM |
626 | } |
627 | ||
b18c6685 CM |
628 | int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, |
629 | struct btrfs_root *root, | |
74493f7a | 630 | u64 bytenr, u64 num_bytes, |
7bb86316 | 631 | u64 root_objectid, u64 ref_generation, |
74493f7a | 632 | u64 owner, u64 owner_offset) |
02217ed2 | 633 | { |
5caf2a00 | 634 | struct btrfs_path *path; |
02217ed2 | 635 | int ret; |
e2fa7227 | 636 | struct btrfs_key key; |
5f39d397 | 637 | struct extent_buffer *l; |
234b63a0 | 638 | struct btrfs_extent_item *item; |
cf27e1ee | 639 | u32 refs; |
037e6390 | 640 | |
db94535d | 641 | WARN_ON(num_bytes < root->sectorsize); |
5caf2a00 | 642 | path = btrfs_alloc_path(); |
54aa1f4d CM |
643 | if (!path) |
644 | return -ENOMEM; | |
26b8003f | 645 | |
3c12ac72 | 646 | path->reada = 1; |
db94535d | 647 | key.objectid = bytenr; |
62e2749e | 648 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
db94535d | 649 | key.offset = num_bytes; |
5caf2a00 | 650 | ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, |
9f5fae2f | 651 | 0, 1); |
54aa1f4d CM |
652 | if (ret < 0) |
653 | return ret; | |
a429e513 | 654 | if (ret != 0) { |
a28ec197 | 655 | BUG(); |
a429e513 | 656 | } |
02217ed2 | 657 | BUG_ON(ret != 0); |
5f39d397 | 658 | l = path->nodes[0]; |
5caf2a00 | 659 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); |
5f39d397 CM |
660 | refs = btrfs_extent_refs(l, item); |
661 | btrfs_set_extent_refs(l, item, refs + 1); | |
5caf2a00 | 662 | btrfs_mark_buffer_dirty(path->nodes[0]); |
a28ec197 | 663 | |
5caf2a00 | 664 | btrfs_release_path(root->fs_info->extent_root, path); |
7bb86316 | 665 | |
3c12ac72 | 666 | path->reada = 1; |
7bb86316 CM |
667 | ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root, |
668 | path, bytenr, root_objectid, | |
669 | ref_generation, owner, owner_offset); | |
670 | BUG_ON(ret); | |
9f5fae2f | 671 | finish_current_insert(trans, root->fs_info->extent_root); |
e20d96d6 | 672 | del_pending_extents(trans, root->fs_info->extent_root); |
74493f7a CM |
673 | |
674 | btrfs_free_path(path); | |
02217ed2 CM |
675 | return 0; |
676 | } | |
677 | ||
e9d0b13b CM |
678 | int btrfs_extent_post_op(struct btrfs_trans_handle *trans, |
679 | struct btrfs_root *root) | |
680 | { | |
681 | finish_current_insert(trans, root->fs_info->extent_root); | |
682 | del_pending_extents(trans, root->fs_info->extent_root); | |
683 | return 0; | |
684 | } | |
685 | ||
b18c6685 | 686 | static int lookup_extent_ref(struct btrfs_trans_handle *trans, |
db94535d CM |
687 | struct btrfs_root *root, u64 bytenr, |
688 | u64 num_bytes, u32 *refs) | |
a28ec197 | 689 | { |
5caf2a00 | 690 | struct btrfs_path *path; |
a28ec197 | 691 | int ret; |
e2fa7227 | 692 | struct btrfs_key key; |
5f39d397 | 693 | struct extent_buffer *l; |
234b63a0 | 694 | struct btrfs_extent_item *item; |
5caf2a00 | 695 | |
db94535d | 696 | WARN_ON(num_bytes < root->sectorsize); |
5caf2a00 | 697 | path = btrfs_alloc_path(); |
3c12ac72 | 698 | path->reada = 1; |
db94535d CM |
699 | key.objectid = bytenr; |
700 | key.offset = num_bytes; | |
62e2749e | 701 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
5caf2a00 | 702 | ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, |
9f5fae2f | 703 | 0, 0); |
54aa1f4d CM |
704 | if (ret < 0) |
705 | goto out; | |
5f39d397 CM |
706 | if (ret != 0) { |
707 | btrfs_print_leaf(root, path->nodes[0]); | |
db94535d | 708 | printk("failed to find block number %Lu\n", bytenr); |
a28ec197 | 709 | BUG(); |
5f39d397 CM |
710 | } |
711 | l = path->nodes[0]; | |
5caf2a00 | 712 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); |
5f39d397 | 713 | *refs = btrfs_extent_refs(l, item); |
54aa1f4d | 714 | out: |
5caf2a00 | 715 | btrfs_free_path(path); |
a28ec197 CM |
716 | return 0; |
717 | } | |
718 | ||
be20aa9d CM |
719 | u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, |
720 | struct btrfs_path *count_path, | |
721 | u64 first_extent) | |
722 | { | |
723 | struct btrfs_root *extent_root = root->fs_info->extent_root; | |
724 | struct btrfs_path *path; | |
725 | u64 bytenr; | |
726 | u64 found_objectid; | |
56b453c9 | 727 | u64 root_objectid = root->root_key.objectid; |
be20aa9d CM |
728 | u32 total_count = 0; |
729 | u32 cur_count; | |
be20aa9d CM |
730 | u32 nritems; |
731 | int ret; | |
732 | struct btrfs_key key; | |
733 | struct btrfs_key found_key; | |
734 | struct extent_buffer *l; | |
735 | struct btrfs_extent_item *item; | |
736 | struct btrfs_extent_ref *ref_item; | |
737 | int level = -1; | |
738 | ||
739 | path = btrfs_alloc_path(); | |
740 | again: | |
741 | if (level == -1) | |
742 | bytenr = first_extent; | |
743 | else | |
744 | bytenr = count_path->nodes[level]->start; | |
745 | ||
746 | cur_count = 0; | |
747 | key.objectid = bytenr; | |
748 | key.offset = 0; | |
749 | ||
750 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); | |
751 | ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); | |
752 | if (ret < 0) | |
753 | goto out; | |
754 | BUG_ON(ret == 0); | |
755 | ||
756 | l = path->nodes[0]; | |
757 | btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); | |
758 | ||
759 | if (found_key.objectid != bytenr || | |
760 | found_key.type != BTRFS_EXTENT_ITEM_KEY) { | |
761 | goto out; | |
762 | } | |
763 | ||
764 | item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); | |
be20aa9d | 765 | while (1) { |
bd09835d | 766 | l = path->nodes[0]; |
be20aa9d CM |
767 | nritems = btrfs_header_nritems(l); |
768 | if (path->slots[0] >= nritems) { | |
769 | ret = btrfs_next_leaf(extent_root, path); | |
770 | if (ret == 0) | |
771 | continue; | |
772 | break; | |
773 | } | |
774 | btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); | |
775 | if (found_key.objectid != bytenr) | |
776 | break; | |
bd09835d | 777 | |
be20aa9d CM |
778 | if (found_key.type != BTRFS_EXTENT_REF_KEY) { |
779 | path->slots[0]++; | |
780 | continue; | |
781 | } | |
782 | ||
783 | cur_count++; | |
784 | ref_item = btrfs_item_ptr(l, path->slots[0], | |
785 | struct btrfs_extent_ref); | |
786 | found_objectid = btrfs_ref_root(l, ref_item); | |
787 | ||
56b453c9 CM |
788 | if (found_objectid != root_objectid) { |
789 | total_count = 2; | |
4313b399 | 790 | goto out; |
56b453c9 CM |
791 | } |
792 | total_count = 1; | |
be20aa9d CM |
793 | path->slots[0]++; |
794 | } | |
795 | if (cur_count == 0) { | |
796 | total_count = 0; | |
797 | goto out; | |
798 | } | |
be20aa9d CM |
799 | if (level >= 0 && root->node == count_path->nodes[level]) |
800 | goto out; | |
801 | level++; | |
802 | btrfs_release_path(root, path); | |
803 | goto again; | |
804 | ||
805 | out: | |
806 | btrfs_free_path(path); | |
807 | return total_count; | |
be20aa9d | 808 | } |
c5739bba | 809 | int btrfs_inc_root_ref(struct btrfs_trans_handle *trans, |
7bb86316 | 810 | struct btrfs_root *root, u64 owner_objectid) |
c5739bba | 811 | { |
7bb86316 CM |
812 | u64 generation; |
813 | u64 key_objectid; | |
814 | u64 level; | |
815 | u32 nritems; | |
816 | struct btrfs_disk_key disk_key; | |
817 | ||
818 | level = btrfs_header_level(root->node); | |
819 | generation = trans->transid; | |
820 | nritems = btrfs_header_nritems(root->node); | |
821 | if (nritems > 0) { | |
822 | if (level == 0) | |
823 | btrfs_item_key(root->node, &disk_key, 0); | |
824 | else | |
825 | btrfs_node_key(root->node, &disk_key, 0); | |
826 | key_objectid = btrfs_disk_key_objectid(&disk_key); | |
827 | } else { | |
828 | key_objectid = 0; | |
829 | } | |
db94535d | 830 | return btrfs_inc_extent_ref(trans, root, root->node->start, |
7bb86316 | 831 | root->node->len, owner_objectid, |
f6dbff55 | 832 | generation, level, key_objectid); |
c5739bba CM |
833 | } |
834 | ||
e089f05c | 835 | int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 836 | struct extent_buffer *buf) |
02217ed2 | 837 | { |
db94535d | 838 | u64 bytenr; |
5f39d397 CM |
839 | u32 nritems; |
840 | struct btrfs_key key; | |
6407bf6d | 841 | struct btrfs_file_extent_item *fi; |
02217ed2 | 842 | int i; |
db94535d | 843 | int level; |
6407bf6d | 844 | int ret; |
54aa1f4d | 845 | int faili; |
a28ec197 | 846 | |
3768f368 | 847 | if (!root->ref_cows) |
a28ec197 | 848 | return 0; |
5f39d397 | 849 | |
db94535d | 850 | level = btrfs_header_level(buf); |
5f39d397 CM |
851 | nritems = btrfs_header_nritems(buf); |
852 | for (i = 0; i < nritems; i++) { | |
db94535d CM |
853 | if (level == 0) { |
854 | u64 disk_bytenr; | |
5f39d397 CM |
855 | btrfs_item_key_to_cpu(buf, &key, i); |
856 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
6407bf6d | 857 | continue; |
5f39d397 | 858 | fi = btrfs_item_ptr(buf, i, |
6407bf6d | 859 | struct btrfs_file_extent_item); |
5f39d397 | 860 | if (btrfs_file_extent_type(buf, fi) == |
236454df CM |
861 | BTRFS_FILE_EXTENT_INLINE) |
862 | continue; | |
db94535d CM |
863 | disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); |
864 | if (disk_bytenr == 0) | |
3a686375 | 865 | continue; |
db94535d | 866 | ret = btrfs_inc_extent_ref(trans, root, disk_bytenr, |
7bb86316 CM |
867 | btrfs_file_extent_disk_num_bytes(buf, fi), |
868 | root->root_key.objectid, trans->transid, | |
869 | key.objectid, key.offset); | |
54aa1f4d CM |
870 | if (ret) { |
871 | faili = i; | |
872 | goto fail; | |
873 | } | |
6407bf6d | 874 | } else { |
db94535d | 875 | bytenr = btrfs_node_blockptr(buf, i); |
6caab489 | 876 | btrfs_node_key_to_cpu(buf, &key, i); |
db94535d | 877 | ret = btrfs_inc_extent_ref(trans, root, bytenr, |
7bb86316 CM |
878 | btrfs_level_size(root, level - 1), |
879 | root->root_key.objectid, | |
f6dbff55 CM |
880 | trans->transid, |
881 | level - 1, key.objectid); | |
54aa1f4d CM |
882 | if (ret) { |
883 | faili = i; | |
884 | goto fail; | |
885 | } | |
6407bf6d | 886 | } |
02217ed2 CM |
887 | } |
888 | return 0; | |
54aa1f4d | 889 | fail: |
ccd467d6 | 890 | WARN_ON(1); |
7bb86316 | 891 | #if 0 |
54aa1f4d | 892 | for (i =0; i < faili; i++) { |
db94535d CM |
893 | if (level == 0) { |
894 | u64 disk_bytenr; | |
5f39d397 CM |
895 | btrfs_item_key_to_cpu(buf, &key, i); |
896 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
54aa1f4d | 897 | continue; |
5f39d397 | 898 | fi = btrfs_item_ptr(buf, i, |
54aa1f4d | 899 | struct btrfs_file_extent_item); |
5f39d397 | 900 | if (btrfs_file_extent_type(buf, fi) == |
54aa1f4d CM |
901 | BTRFS_FILE_EXTENT_INLINE) |
902 | continue; | |
db94535d CM |
903 | disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); |
904 | if (disk_bytenr == 0) | |
54aa1f4d | 905 | continue; |
db94535d CM |
906 | err = btrfs_free_extent(trans, root, disk_bytenr, |
907 | btrfs_file_extent_disk_num_bytes(buf, | |
5f39d397 | 908 | fi), 0); |
54aa1f4d CM |
909 | BUG_ON(err); |
910 | } else { | |
db94535d CM |
911 | bytenr = btrfs_node_blockptr(buf, i); |
912 | err = btrfs_free_extent(trans, root, bytenr, | |
913 | btrfs_level_size(root, level - 1), 0); | |
54aa1f4d CM |
914 | BUG_ON(err); |
915 | } | |
916 | } | |
7bb86316 | 917 | #endif |
54aa1f4d | 918 | return ret; |
02217ed2 CM |
919 | } |
920 | ||
9078a3e1 CM |
921 | static int write_one_cache_group(struct btrfs_trans_handle *trans, |
922 | struct btrfs_root *root, | |
923 | struct btrfs_path *path, | |
924 | struct btrfs_block_group_cache *cache) | |
925 | { | |
926 | int ret; | |
927 | int pending_ret; | |
928 | struct btrfs_root *extent_root = root->fs_info->extent_root; | |
5f39d397 CM |
929 | unsigned long bi; |
930 | struct extent_buffer *leaf; | |
9078a3e1 | 931 | |
9078a3e1 | 932 | ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1); |
54aa1f4d CM |
933 | if (ret < 0) |
934 | goto fail; | |
9078a3e1 | 935 | BUG_ON(ret); |
5f39d397 CM |
936 | |
937 | leaf = path->nodes[0]; | |
938 | bi = btrfs_item_ptr_offset(leaf, path->slots[0]); | |
939 | write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item)); | |
940 | btrfs_mark_buffer_dirty(leaf); | |
9078a3e1 | 941 | btrfs_release_path(extent_root, path); |
54aa1f4d | 942 | fail: |
9078a3e1 CM |
943 | finish_current_insert(trans, extent_root); |
944 | pending_ret = del_pending_extents(trans, extent_root); | |
945 | if (ret) | |
946 | return ret; | |
947 | if (pending_ret) | |
948 | return pending_ret; | |
949 | return 0; | |
950 | ||
951 | } | |
952 | ||
96b5179d CM |
953 | int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, |
954 | struct btrfs_root *root) | |
9078a3e1 | 955 | { |
d1310b2e | 956 | struct extent_io_tree *block_group_cache; |
96b5179d | 957 | struct btrfs_block_group_cache *cache; |
9078a3e1 CM |
958 | int ret; |
959 | int err = 0; | |
960 | int werr = 0; | |
9078a3e1 | 961 | struct btrfs_path *path; |
96b5179d CM |
962 | u64 last = 0; |
963 | u64 start; | |
964 | u64 end; | |
965 | u64 ptr; | |
9078a3e1 | 966 | |
96b5179d | 967 | block_group_cache = &root->fs_info->block_group_cache; |
9078a3e1 CM |
968 | path = btrfs_alloc_path(); |
969 | if (!path) | |
970 | return -ENOMEM; | |
971 | ||
972 | while(1) { | |
96b5179d CM |
973 | ret = find_first_extent_bit(block_group_cache, last, |
974 | &start, &end, BLOCK_GROUP_DIRTY); | |
975 | if (ret) | |
9078a3e1 | 976 | break; |
54aa1f4d | 977 | |
96b5179d CM |
978 | last = end + 1; |
979 | ret = get_state_private(block_group_cache, start, &ptr); | |
980 | if (ret) | |
981 | break; | |
ae2f5411 | 982 | cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; |
96b5179d CM |
983 | err = write_one_cache_group(trans, root, |
984 | path, cache); | |
985 | /* | |
986 | * if we fail to write the cache group, we want | |
987 | * to keep it marked dirty in hopes that a later | |
988 | * write will work | |
989 | */ | |
990 | if (err) { | |
991 | werr = err; | |
992 | continue; | |
9078a3e1 | 993 | } |
96b5179d CM |
994 | clear_extent_bits(block_group_cache, start, end, |
995 | BLOCK_GROUP_DIRTY, GFP_NOFS); | |
9078a3e1 CM |
996 | } |
997 | btrfs_free_path(path); | |
998 | return werr; | |
999 | } | |
1000 | ||
6324fbf3 CM |
1001 | static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info, |
1002 | u64 flags) | |
1003 | { | |
1004 | struct list_head *head = &info->space_info; | |
1005 | struct list_head *cur; | |
1006 | struct btrfs_space_info *found; | |
1007 | list_for_each(cur, head) { | |
1008 | found = list_entry(cur, struct btrfs_space_info, list); | |
1009 | if (found->flags == flags) | |
1010 | return found; | |
1011 | } | |
1012 | return NULL; | |
1013 | ||
1014 | } | |
1015 | ||
593060d7 CM |
1016 | static int update_space_info(struct btrfs_fs_info *info, u64 flags, |
1017 | u64 total_bytes, u64 bytes_used, | |
1018 | struct btrfs_space_info **space_info) | |
1019 | { | |
1020 | struct btrfs_space_info *found; | |
1021 | ||
1022 | found = __find_space_info(info, flags); | |
1023 | if (found) { | |
1024 | found->total_bytes += total_bytes; | |
1025 | found->bytes_used += bytes_used; | |
8f18cf13 | 1026 | found->full = 0; |
593060d7 CM |
1027 | WARN_ON(found->total_bytes < found->bytes_used); |
1028 | *space_info = found; | |
1029 | return 0; | |
1030 | } | |
1031 | found = kmalloc(sizeof(*found), GFP_NOFS); | |
1032 | if (!found) | |
1033 | return -ENOMEM; | |
1034 | ||
1035 | list_add(&found->list, &info->space_info); | |
1036 | found->flags = flags; | |
1037 | found->total_bytes = total_bytes; | |
1038 | found->bytes_used = bytes_used; | |
1039 | found->bytes_pinned = 0; | |
1040 | found->full = 0; | |
1041 | *space_info = found; | |
1042 | return 0; | |
1043 | } | |
1044 | ||
8790d502 CM |
1045 | static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags) |
1046 | { | |
1047 | u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 | | |
611f0e00 | 1048 | BTRFS_BLOCK_GROUP_RAID1 | |
321aecc6 | 1049 | BTRFS_BLOCK_GROUP_RAID10 | |
611f0e00 | 1050 | BTRFS_BLOCK_GROUP_DUP); |
8790d502 CM |
1051 | if (extra_flags) { |
1052 | if (flags & BTRFS_BLOCK_GROUP_DATA) | |
1053 | fs_info->avail_data_alloc_bits |= extra_flags; | |
1054 | if (flags & BTRFS_BLOCK_GROUP_METADATA) | |
1055 | fs_info->avail_metadata_alloc_bits |= extra_flags; | |
1056 | if (flags & BTRFS_BLOCK_GROUP_SYSTEM) | |
1057 | fs_info->avail_system_alloc_bits |= extra_flags; | |
1058 | } | |
1059 | } | |
593060d7 | 1060 | |
6324fbf3 CM |
1061 | static int do_chunk_alloc(struct btrfs_trans_handle *trans, |
1062 | struct btrfs_root *extent_root, u64 alloc_bytes, | |
1063 | u64 flags) | |
1064 | { | |
1065 | struct btrfs_space_info *space_info; | |
1066 | u64 thresh; | |
1067 | u64 start; | |
1068 | u64 num_bytes; | |
1069 | int ret; | |
1070 | ||
1071 | space_info = __find_space_info(extent_root->fs_info, flags); | |
593060d7 CM |
1072 | if (!space_info) { |
1073 | ret = update_space_info(extent_root->fs_info, flags, | |
1074 | 0, 0, &space_info); | |
1075 | BUG_ON(ret); | |
1076 | } | |
6324fbf3 CM |
1077 | BUG_ON(!space_info); |
1078 | ||
1079 | if (space_info->full) | |
1080 | return 0; | |
1081 | ||
8790d502 | 1082 | thresh = div_factor(space_info->total_bytes, 6); |
6324fbf3 CM |
1083 | if ((space_info->bytes_used + space_info->bytes_pinned + alloc_bytes) < |
1084 | thresh) | |
1085 | return 0; | |
1086 | ||
1087 | ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags); | |
1088 | if (ret == -ENOSPC) { | |
1089 | printk("space info full %Lu\n", flags); | |
1090 | space_info->full = 1; | |
1091 | return 0; | |
1092 | } | |
1093 | ||
1094 | BUG_ON(ret); | |
1095 | ||
1096 | ret = btrfs_make_block_group(trans, extent_root, 0, flags, | |
e17cade2 | 1097 | BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes); |
6324fbf3 | 1098 | BUG_ON(ret); |
593060d7 | 1099 | |
6324fbf3 CM |
1100 | return 0; |
1101 | } | |
1102 | ||
9078a3e1 CM |
1103 | static int update_block_group(struct btrfs_trans_handle *trans, |
1104 | struct btrfs_root *root, | |
db94535d | 1105 | u64 bytenr, u64 num_bytes, int alloc, |
0b86a832 | 1106 | int mark_free) |
9078a3e1 CM |
1107 | { |
1108 | struct btrfs_block_group_cache *cache; | |
1109 | struct btrfs_fs_info *info = root->fs_info; | |
db94535d | 1110 | u64 total = num_bytes; |
9078a3e1 | 1111 | u64 old_val; |
db94535d | 1112 | u64 byte_in_group; |
96b5179d CM |
1113 | u64 start; |
1114 | u64 end; | |
3e1ad54f | 1115 | |
9078a3e1 | 1116 | while(total) { |
db94535d | 1117 | cache = btrfs_lookup_block_group(info, bytenr); |
3e1ad54f | 1118 | if (!cache) { |
9078a3e1 | 1119 | return -1; |
cd1bc465 | 1120 | } |
db94535d CM |
1121 | byte_in_group = bytenr - cache->key.objectid; |
1122 | WARN_ON(byte_in_group > cache->key.offset); | |
96b5179d CM |
1123 | start = cache->key.objectid; |
1124 | end = start + cache->key.offset - 1; | |
1125 | set_extent_bits(&info->block_group_cache, start, end, | |
1126 | BLOCK_GROUP_DIRTY, GFP_NOFS); | |
9078a3e1 CM |
1127 | |
1128 | old_val = btrfs_block_group_used(&cache->item); | |
db94535d | 1129 | num_bytes = min(total, cache->key.offset - byte_in_group); |
cd1bc465 | 1130 | if (alloc) { |
db94535d | 1131 | old_val += num_bytes; |
6324fbf3 | 1132 | cache->space_info->bytes_used += num_bytes; |
cd1bc465 | 1133 | } else { |
db94535d | 1134 | old_val -= num_bytes; |
6324fbf3 | 1135 | cache->space_info->bytes_used -= num_bytes; |
f510cfec CM |
1136 | if (mark_free) { |
1137 | set_extent_dirty(&info->free_space_cache, | |
db94535d | 1138 | bytenr, bytenr + num_bytes - 1, |
f510cfec | 1139 | GFP_NOFS); |
e37c9e69 | 1140 | } |
cd1bc465 | 1141 | } |
9078a3e1 | 1142 | btrfs_set_block_group_used(&cache->item, old_val); |
db94535d CM |
1143 | total -= num_bytes; |
1144 | bytenr += num_bytes; | |
9078a3e1 CM |
1145 | } |
1146 | return 0; | |
1147 | } | |
6324fbf3 | 1148 | |
324ae4df Y |
1149 | static int update_pinned_extents(struct btrfs_root *root, |
1150 | u64 bytenr, u64 num, int pin) | |
1151 | { | |
1152 | u64 len; | |
1153 | struct btrfs_block_group_cache *cache; | |
1154 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1155 | ||
1156 | if (pin) { | |
1157 | set_extent_dirty(&fs_info->pinned_extents, | |
1158 | bytenr, bytenr + num - 1, GFP_NOFS); | |
1159 | } else { | |
1160 | clear_extent_dirty(&fs_info->pinned_extents, | |
1161 | bytenr, bytenr + num - 1, GFP_NOFS); | |
1162 | } | |
1163 | while (num > 0) { | |
1164 | cache = btrfs_lookup_block_group(fs_info, bytenr); | |
1165 | WARN_ON(!cache); | |
1166 | len = min(num, cache->key.offset - | |
1167 | (bytenr - cache->key.objectid)); | |
1168 | if (pin) { | |
1169 | cache->pinned += len; | |
6324fbf3 | 1170 | cache->space_info->bytes_pinned += len; |
324ae4df Y |
1171 | fs_info->total_pinned += len; |
1172 | } else { | |
1173 | cache->pinned -= len; | |
6324fbf3 | 1174 | cache->space_info->bytes_pinned -= len; |
324ae4df Y |
1175 | fs_info->total_pinned -= len; |
1176 | } | |
1177 | bytenr += len; | |
1178 | num -= len; | |
1179 | } | |
1180 | return 0; | |
1181 | } | |
9078a3e1 | 1182 | |
d1310b2e | 1183 | int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy) |
ccd467d6 | 1184 | { |
ccd467d6 | 1185 | u64 last = 0; |
1a5bc167 CM |
1186 | u64 start; |
1187 | u64 end; | |
d1310b2e | 1188 | struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents; |
ccd467d6 | 1189 | int ret; |
ccd467d6 CM |
1190 | |
1191 | while(1) { | |
1a5bc167 CM |
1192 | ret = find_first_extent_bit(pinned_extents, last, |
1193 | &start, &end, EXTENT_DIRTY); | |
1194 | if (ret) | |
ccd467d6 | 1195 | break; |
1a5bc167 CM |
1196 | set_extent_dirty(copy, start, end, GFP_NOFS); |
1197 | last = end + 1; | |
ccd467d6 CM |
1198 | } |
1199 | return 0; | |
1200 | } | |
1201 | ||
1202 | int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, | |
1203 | struct btrfs_root *root, | |
d1310b2e | 1204 | struct extent_io_tree *unpin) |
a28ec197 | 1205 | { |
1a5bc167 CM |
1206 | u64 start; |
1207 | u64 end; | |
a28ec197 | 1208 | int ret; |
d1310b2e | 1209 | struct extent_io_tree *free_space_cache; |
f510cfec | 1210 | free_space_cache = &root->fs_info->free_space_cache; |
a28ec197 CM |
1211 | |
1212 | while(1) { | |
1a5bc167 CM |
1213 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
1214 | EXTENT_DIRTY); | |
1215 | if (ret) | |
a28ec197 | 1216 | break; |
324ae4df | 1217 | update_pinned_extents(root, start, end + 1 - start, 0); |
1a5bc167 CM |
1218 | clear_extent_dirty(unpin, start, end, GFP_NOFS); |
1219 | set_extent_dirty(free_space_cache, start, end, GFP_NOFS); | |
a28ec197 CM |
1220 | } |
1221 | return 0; | |
1222 | } | |
1223 | ||
98ed5174 CM |
1224 | static int finish_current_insert(struct btrfs_trans_handle *trans, |
1225 | struct btrfs_root *extent_root) | |
037e6390 | 1226 | { |
7bb86316 CM |
1227 | u64 start; |
1228 | u64 end; | |
1229 | struct btrfs_fs_info *info = extent_root->fs_info; | |
d8d5f3e1 | 1230 | struct extent_buffer *eb; |
7bb86316 | 1231 | struct btrfs_path *path; |
e2fa7227 | 1232 | struct btrfs_key ins; |
d8d5f3e1 | 1233 | struct btrfs_disk_key first; |
234b63a0 | 1234 | struct btrfs_extent_item extent_item; |
037e6390 | 1235 | int ret; |
d8d5f3e1 | 1236 | int level; |
1a5bc167 | 1237 | int err = 0; |
037e6390 | 1238 | |
5f39d397 | 1239 | btrfs_set_stack_extent_refs(&extent_item, 1); |
62e2749e | 1240 | btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY); |
7bb86316 | 1241 | path = btrfs_alloc_path(); |
037e6390 | 1242 | |
26b8003f | 1243 | while(1) { |
1a5bc167 CM |
1244 | ret = find_first_extent_bit(&info->extent_ins, 0, &start, |
1245 | &end, EXTENT_LOCKED); | |
1246 | if (ret) | |
26b8003f CM |
1247 | break; |
1248 | ||
1a5bc167 CM |
1249 | ins.objectid = start; |
1250 | ins.offset = end + 1 - start; | |
1251 | err = btrfs_insert_item(trans, extent_root, &ins, | |
1252 | &extent_item, sizeof(extent_item)); | |
1253 | clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED, | |
1254 | GFP_NOFS); | |
d8d5f3e1 CM |
1255 | eb = read_tree_block(extent_root, ins.objectid, ins.offset); |
1256 | level = btrfs_header_level(eb); | |
1257 | if (level == 0) { | |
1258 | btrfs_item_key(eb, &first, 0); | |
1259 | } else { | |
1260 | btrfs_node_key(eb, &first, 0); | |
1261 | } | |
7bb86316 CM |
1262 | err = btrfs_insert_extent_backref(trans, extent_root, path, |
1263 | start, extent_root->root_key.objectid, | |
f6dbff55 CM |
1264 | 0, level, |
1265 | btrfs_disk_key_objectid(&first)); | |
7bb86316 | 1266 | BUG_ON(err); |
d8d5f3e1 | 1267 | free_extent_buffer(eb); |
037e6390 | 1268 | } |
7bb86316 | 1269 | btrfs_free_path(path); |
037e6390 CM |
1270 | return 0; |
1271 | } | |
1272 | ||
db94535d CM |
1273 | static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, |
1274 | int pending) | |
e20d96d6 | 1275 | { |
1a5bc167 | 1276 | int err = 0; |
5f39d397 | 1277 | struct extent_buffer *buf; |
8ef97622 | 1278 | |
f4b9aa8d | 1279 | if (!pending) { |
db94535d | 1280 | buf = btrfs_find_tree_block(root, bytenr, num_bytes); |
5f39d397 CM |
1281 | if (buf) { |
1282 | if (btrfs_buffer_uptodate(buf)) { | |
2c90e5d6 CM |
1283 | u64 transid = |
1284 | root->fs_info->running_transaction->transid; | |
dc17ff8f CM |
1285 | u64 header_transid = |
1286 | btrfs_header_generation(buf); | |
6bc34676 CM |
1287 | if (header_transid == transid && |
1288 | !btrfs_header_flag(buf, | |
1289 | BTRFS_HEADER_FLAG_WRITTEN)) { | |
55c69072 | 1290 | clean_tree_block(NULL, root, buf); |
5f39d397 | 1291 | free_extent_buffer(buf); |
c549228f | 1292 | return 1; |
2c90e5d6 | 1293 | } |
f4b9aa8d | 1294 | } |
5f39d397 | 1295 | free_extent_buffer(buf); |
8ef97622 | 1296 | } |
324ae4df | 1297 | update_pinned_extents(root, bytenr, num_bytes, 1); |
f4b9aa8d | 1298 | } else { |
1a5bc167 | 1299 | set_extent_bits(&root->fs_info->pending_del, |
db94535d CM |
1300 | bytenr, bytenr + num_bytes - 1, |
1301 | EXTENT_LOCKED, GFP_NOFS); | |
f4b9aa8d | 1302 | } |
be744175 | 1303 | BUG_ON(err < 0); |
e20d96d6 CM |
1304 | return 0; |
1305 | } | |
1306 | ||
fec577fb | 1307 | /* |
a28ec197 | 1308 | * remove an extent from the root, returns 0 on success |
fec577fb | 1309 | */ |
e089f05c | 1310 | static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root |
7bb86316 CM |
1311 | *root, u64 bytenr, u64 num_bytes, |
1312 | u64 root_objectid, u64 ref_generation, | |
1313 | u64 owner_objectid, u64 owner_offset, int pin, | |
e37c9e69 | 1314 | int mark_free) |
a28ec197 | 1315 | { |
5caf2a00 | 1316 | struct btrfs_path *path; |
e2fa7227 | 1317 | struct btrfs_key key; |
1261ec42 CM |
1318 | struct btrfs_fs_info *info = root->fs_info; |
1319 | struct btrfs_root *extent_root = info->extent_root; | |
5f39d397 | 1320 | struct extent_buffer *leaf; |
a28ec197 | 1321 | int ret; |
952fccac CM |
1322 | int extent_slot = 0; |
1323 | int found_extent = 0; | |
1324 | int num_to_del = 1; | |
234b63a0 | 1325 | struct btrfs_extent_item *ei; |
cf27e1ee | 1326 | u32 refs; |
037e6390 | 1327 | |
db94535d | 1328 | key.objectid = bytenr; |
62e2749e | 1329 | btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); |
db94535d | 1330 | key.offset = num_bytes; |
5caf2a00 | 1331 | path = btrfs_alloc_path(); |
54aa1f4d CM |
1332 | if (!path) |
1333 | return -ENOMEM; | |
5f26f772 | 1334 | |
3c12ac72 | 1335 | path->reada = 1; |
7bb86316 CM |
1336 | ret = lookup_extent_backref(trans, extent_root, path, |
1337 | bytenr, root_objectid, | |
1338 | ref_generation, | |
1339 | owner_objectid, owner_offset, 1); | |
1340 | if (ret == 0) { | |
952fccac CM |
1341 | struct btrfs_key found_key; |
1342 | extent_slot = path->slots[0]; | |
1343 | while(extent_slot > 0) { | |
1344 | extent_slot--; | |
1345 | btrfs_item_key_to_cpu(path->nodes[0], &found_key, | |
1346 | extent_slot); | |
1347 | if (found_key.objectid != bytenr) | |
1348 | break; | |
1349 | if (found_key.type == BTRFS_EXTENT_ITEM_KEY && | |
1350 | found_key.offset == num_bytes) { | |
1351 | found_extent = 1; | |
1352 | break; | |
1353 | } | |
1354 | if (path->slots[0] - extent_slot > 5) | |
1355 | break; | |
1356 | } | |
1357 | if (!found_extent) | |
1358 | ret = btrfs_del_item(trans, extent_root, path); | |
7bb86316 CM |
1359 | } else { |
1360 | btrfs_print_leaf(extent_root, path->nodes[0]); | |
1361 | WARN_ON(1); | |
1362 | printk("Unable to find ref byte nr %Lu root %Lu " | |
1363 | " gen %Lu owner %Lu offset %Lu\n", bytenr, | |
1364 | root_objectid, ref_generation, owner_objectid, | |
1365 | owner_offset); | |
1366 | } | |
952fccac CM |
1367 | if (!found_extent) { |
1368 | btrfs_release_path(extent_root, path); | |
1369 | ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); | |
1370 | if (ret < 0) | |
1371 | return ret; | |
1372 | BUG_ON(ret); | |
1373 | extent_slot = path->slots[0]; | |
1374 | } | |
5f39d397 CM |
1375 | |
1376 | leaf = path->nodes[0]; | |
952fccac | 1377 | ei = btrfs_item_ptr(leaf, extent_slot, |
123abc88 | 1378 | struct btrfs_extent_item); |
5f39d397 CM |
1379 | refs = btrfs_extent_refs(leaf, ei); |
1380 | BUG_ON(refs == 0); | |
1381 | refs -= 1; | |
1382 | btrfs_set_extent_refs(leaf, ei, refs); | |
952fccac | 1383 | |
5f39d397 CM |
1384 | btrfs_mark_buffer_dirty(leaf); |
1385 | ||
952fccac CM |
1386 | if (refs == 0 && found_extent && path->slots[0] == extent_slot + 1) { |
1387 | /* if the back ref and the extent are next to each other | |
1388 | * they get deleted below in one shot | |
1389 | */ | |
1390 | path->slots[0] = extent_slot; | |
1391 | num_to_del = 2; | |
1392 | } else if (found_extent) { | |
1393 | /* otherwise delete the extent back ref */ | |
1394 | ret = btrfs_del_item(trans, extent_root, path); | |
1395 | BUG_ON(ret); | |
1396 | /* if refs are 0, we need to setup the path for deletion */ | |
1397 | if (refs == 0) { | |
1398 | btrfs_release_path(extent_root, path); | |
1399 | ret = btrfs_search_slot(trans, extent_root, &key, path, | |
1400 | -1, 1); | |
1401 | if (ret < 0) | |
1402 | return ret; | |
1403 | BUG_ON(ret); | |
1404 | } | |
1405 | } | |
1406 | ||
cf27e1ee | 1407 | if (refs == 0) { |
db94535d CM |
1408 | u64 super_used; |
1409 | u64 root_used; | |
78fae27e CM |
1410 | |
1411 | if (pin) { | |
db94535d | 1412 | ret = pin_down_bytes(root, bytenr, num_bytes, 0); |
c549228f Y |
1413 | if (ret > 0) |
1414 | mark_free = 1; | |
1415 | BUG_ON(ret < 0); | |
78fae27e CM |
1416 | } |
1417 | ||
58176a96 | 1418 | /* block accounting for super block */ |
db94535d CM |
1419 | super_used = btrfs_super_bytes_used(&info->super_copy); |
1420 | btrfs_set_super_bytes_used(&info->super_copy, | |
1421 | super_used - num_bytes); | |
58176a96 JB |
1422 | |
1423 | /* block accounting for root item */ | |
db94535d | 1424 | root_used = btrfs_root_used(&root->root_item); |
5f39d397 | 1425 | btrfs_set_root_used(&root->root_item, |
db94535d | 1426 | root_used - num_bytes); |
952fccac CM |
1427 | ret = btrfs_del_items(trans, extent_root, path, path->slots[0], |
1428 | num_to_del); | |
54aa1f4d CM |
1429 | if (ret) { |
1430 | return ret; | |
1431 | } | |
db94535d | 1432 | ret = update_block_group(trans, root, bytenr, num_bytes, 0, |
0b86a832 | 1433 | mark_free); |
9078a3e1 | 1434 | BUG_ON(ret); |
a28ec197 | 1435 | } |
5caf2a00 | 1436 | btrfs_free_path(path); |
e089f05c | 1437 | finish_current_insert(trans, extent_root); |
a28ec197 CM |
1438 | return ret; |
1439 | } | |
1440 | ||
a28ec197 CM |
1441 | /* |
1442 | * find all the blocks marked as pending in the radix tree and remove | |
1443 | * them from the extent map | |
1444 | */ | |
e089f05c CM |
1445 | static int del_pending_extents(struct btrfs_trans_handle *trans, struct |
1446 | btrfs_root *extent_root) | |
a28ec197 CM |
1447 | { |
1448 | int ret; | |
e20d96d6 | 1449 | int err = 0; |
1a5bc167 CM |
1450 | u64 start; |
1451 | u64 end; | |
d1310b2e CM |
1452 | struct extent_io_tree *pending_del; |
1453 | struct extent_io_tree *pinned_extents; | |
8ef97622 | 1454 | |
1a5bc167 CM |
1455 | pending_del = &extent_root->fs_info->pending_del; |
1456 | pinned_extents = &extent_root->fs_info->pinned_extents; | |
a28ec197 CM |
1457 | |
1458 | while(1) { | |
1a5bc167 CM |
1459 | ret = find_first_extent_bit(pending_del, 0, &start, &end, |
1460 | EXTENT_LOCKED); | |
1461 | if (ret) | |
a28ec197 | 1462 | break; |
324ae4df | 1463 | update_pinned_extents(extent_root, start, end + 1 - start, 1); |
1a5bc167 CM |
1464 | clear_extent_bits(pending_del, start, end, EXTENT_LOCKED, |
1465 | GFP_NOFS); | |
1466 | ret = __free_extent(trans, extent_root, | |
7bb86316 CM |
1467 | start, end + 1 - start, |
1468 | extent_root->root_key.objectid, | |
1469 | 0, 0, 0, 0, 0); | |
1a5bc167 CM |
1470 | if (ret) |
1471 | err = ret; | |
fec577fb | 1472 | } |
e20d96d6 | 1473 | return err; |
fec577fb CM |
1474 | } |
1475 | ||
1476 | /* | |
1477 | * remove an extent from the root, returns 0 on success | |
1478 | */ | |
e089f05c | 1479 | int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root |
7bb86316 CM |
1480 | *root, u64 bytenr, u64 num_bytes, |
1481 | u64 root_objectid, u64 ref_generation, | |
1482 | u64 owner_objectid, u64 owner_offset, int pin) | |
fec577fb | 1483 | { |
9f5fae2f | 1484 | struct btrfs_root *extent_root = root->fs_info->extent_root; |
fec577fb CM |
1485 | int pending_ret; |
1486 | int ret; | |
a28ec197 | 1487 | |
db94535d | 1488 | WARN_ON(num_bytes < root->sectorsize); |
7bb86316 CM |
1489 | if (!root->ref_cows) |
1490 | ref_generation = 0; | |
1491 | ||
fec577fb | 1492 | if (root == extent_root) { |
db94535d | 1493 | pin_down_bytes(root, bytenr, num_bytes, 1); |
fec577fb CM |
1494 | return 0; |
1495 | } | |
7bb86316 CM |
1496 | ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid, |
1497 | ref_generation, owner_objectid, owner_offset, | |
1498 | pin, pin == 0); | |
e20d96d6 | 1499 | pending_ret = del_pending_extents(trans, root->fs_info->extent_root); |
fec577fb CM |
1500 | return ret ? ret : pending_ret; |
1501 | } | |
1502 | ||
87ee04eb CM |
1503 | static u64 stripe_align(struct btrfs_root *root, u64 val) |
1504 | { | |
1505 | u64 mask = ((u64)root->stripesize - 1); | |
1506 | u64 ret = (val + mask) & ~mask; | |
1507 | return ret; | |
1508 | } | |
1509 | ||
fec577fb CM |
1510 | /* |
1511 | * walks the btree of allocated extents and find a hole of a given size. | |
1512 | * The key ins is changed to record the hole: | |
1513 | * ins->objectid == block start | |
62e2749e | 1514 | * ins->flags = BTRFS_EXTENT_ITEM_KEY |
fec577fb CM |
1515 | * ins->offset == number of blocks |
1516 | * Any available blocks before search_start are skipped. | |
1517 | */ | |
98ed5174 CM |
1518 | static int noinline find_free_extent(struct btrfs_trans_handle *trans, |
1519 | struct btrfs_root *orig_root, | |
1520 | u64 num_bytes, u64 empty_size, | |
1521 | u64 search_start, u64 search_end, | |
1522 | u64 hint_byte, struct btrfs_key *ins, | |
1523 | u64 exclude_start, u64 exclude_nr, | |
1524 | int data) | |
fec577fb | 1525 | { |
87ee04eb | 1526 | int ret; |
be744175 | 1527 | u64 orig_search_start = search_start; |
9f5fae2f | 1528 | struct btrfs_root * root = orig_root->fs_info->extent_root; |
f2458e1d | 1529 | struct btrfs_fs_info *info = root->fs_info; |
db94535d | 1530 | u64 total_needed = num_bytes; |
239b14b3 | 1531 | u64 *last_ptr = NULL; |
be08c1b9 | 1532 | struct btrfs_block_group_cache *block_group; |
be744175 | 1533 | int full_scan = 0; |
fbdc762b | 1534 | int wrapped = 0; |
239b14b3 | 1535 | int empty_cluster = 2 * 1024 * 1024; |
fec577fb | 1536 | |
db94535d | 1537 | WARN_ON(num_bytes < root->sectorsize); |
b1a4d965 CM |
1538 | btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY); |
1539 | ||
239b14b3 CM |
1540 | if (data & BTRFS_BLOCK_GROUP_METADATA) { |
1541 | last_ptr = &root->fs_info->last_alloc; | |
8790d502 | 1542 | empty_cluster = 256 * 1024; |
239b14b3 CM |
1543 | } |
1544 | ||
1545 | if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) { | |
1546 | last_ptr = &root->fs_info->last_data_alloc; | |
1547 | } | |
1548 | ||
1549 | if (last_ptr) { | |
1550 | if (*last_ptr) | |
1551 | hint_byte = *last_ptr; | |
1552 | else { | |
1553 | empty_size += empty_cluster; | |
1554 | } | |
1555 | } | |
1556 | ||
7f93bf8d CM |
1557 | if (search_end == (u64)-1) |
1558 | search_end = btrfs_super_total_bytes(&info->super_copy); | |
0b86a832 | 1559 | |
db94535d CM |
1560 | if (hint_byte) { |
1561 | block_group = btrfs_lookup_block_group(info, hint_byte); | |
1a2b2ac7 CM |
1562 | if (!block_group) |
1563 | hint_byte = search_start; | |
be744175 | 1564 | block_group = btrfs_find_block_group(root, block_group, |
db94535d | 1565 | hint_byte, data, 1); |
239b14b3 CM |
1566 | if (last_ptr && *last_ptr == 0 && block_group) |
1567 | hint_byte = block_group->key.objectid; | |
be744175 CM |
1568 | } else { |
1569 | block_group = btrfs_find_block_group(root, | |
1a2b2ac7 CM |
1570 | trans->block_group, |
1571 | search_start, data, 1); | |
be744175 | 1572 | } |
239b14b3 | 1573 | search_start = max(search_start, hint_byte); |
be744175 | 1574 | |
6702ed49 | 1575 | total_needed += empty_size; |
0b86a832 | 1576 | |
be744175 | 1577 | check_failed: |
70b043f0 CM |
1578 | if (!block_group) { |
1579 | block_group = btrfs_lookup_block_group(info, search_start); | |
1580 | if (!block_group) | |
1581 | block_group = btrfs_lookup_block_group(info, | |
1582 | orig_search_start); | |
1583 | } | |
0b86a832 CM |
1584 | ret = find_search_start(root, &block_group, &search_start, |
1585 | total_needed, data); | |
239b14b3 CM |
1586 | if (ret == -ENOSPC && last_ptr && *last_ptr) { |
1587 | *last_ptr = 0; | |
1588 | block_group = btrfs_lookup_block_group(info, | |
1589 | orig_search_start); | |
1590 | search_start = orig_search_start; | |
1591 | ret = find_search_start(root, &block_group, &search_start, | |
1592 | total_needed, data); | |
1593 | } | |
1594 | if (ret == -ENOSPC) | |
1595 | goto enospc; | |
0b86a832 | 1596 | if (ret) |
d548ee51 | 1597 | goto error; |
e19caa5f | 1598 | |
239b14b3 CM |
1599 | if (last_ptr && *last_ptr && search_start != *last_ptr) { |
1600 | *last_ptr = 0; | |
1601 | if (!empty_size) { | |
1602 | empty_size += empty_cluster; | |
1603 | total_needed += empty_size; | |
1604 | } | |
1605 | block_group = btrfs_lookup_block_group(info, | |
1606 | orig_search_start); | |
1607 | search_start = orig_search_start; | |
1608 | ret = find_search_start(root, &block_group, | |
1609 | &search_start, total_needed, data); | |
1610 | if (ret == -ENOSPC) | |
1611 | goto enospc; | |
1612 | if (ret) | |
1613 | goto error; | |
1614 | } | |
1615 | ||
0b86a832 CM |
1616 | search_start = stripe_align(root, search_start); |
1617 | ins->objectid = search_start; | |
1618 | ins->offset = num_bytes; | |
e37c9e69 | 1619 | |
db94535d | 1620 | if (ins->objectid + num_bytes >= search_end) |
cf67582b | 1621 | goto enospc; |
0b86a832 CM |
1622 | |
1623 | if (ins->objectid + num_bytes > | |
1624 | block_group->key.objectid + block_group->key.offset) { | |
e19caa5f CM |
1625 | search_start = block_group->key.objectid + |
1626 | block_group->key.offset; | |
1627 | goto new_group; | |
1628 | } | |
0b86a832 | 1629 | |
1a5bc167 | 1630 | if (test_range_bit(&info->extent_ins, ins->objectid, |
db94535d CM |
1631 | ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) { |
1632 | search_start = ins->objectid + num_bytes; | |
1a5bc167 CM |
1633 | goto new_group; |
1634 | } | |
0b86a832 | 1635 | |
1a5bc167 | 1636 | if (test_range_bit(&info->pinned_extents, ins->objectid, |
db94535d CM |
1637 | ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) { |
1638 | search_start = ins->objectid + num_bytes; | |
1a5bc167 | 1639 | goto new_group; |
fec577fb | 1640 | } |
0b86a832 | 1641 | |
db94535d | 1642 | if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start && |
f2654de4 CM |
1643 | ins->objectid < exclude_start + exclude_nr)) { |
1644 | search_start = exclude_start + exclude_nr; | |
1645 | goto new_group; | |
1646 | } | |
0b86a832 | 1647 | |
6324fbf3 | 1648 | if (!(data & BTRFS_BLOCK_GROUP_DATA)) { |
5276aeda | 1649 | block_group = btrfs_lookup_block_group(info, ins->objectid); |
26b8003f CM |
1650 | if (block_group) |
1651 | trans->block_group = block_group; | |
f2458e1d | 1652 | } |
db94535d | 1653 | ins->offset = num_bytes; |
239b14b3 CM |
1654 | if (last_ptr) { |
1655 | *last_ptr = ins->objectid + ins->offset; | |
1656 | if (*last_ptr == | |
1657 | btrfs_super_total_bytes(&root->fs_info->super_copy)) { | |
1658 | *last_ptr = 0; | |
1659 | } | |
1660 | } | |
fec577fb | 1661 | return 0; |
be744175 CM |
1662 | |
1663 | new_group: | |
db94535d | 1664 | if (search_start + num_bytes >= search_end) { |
cf67582b | 1665 | enospc: |
be744175 | 1666 | search_start = orig_search_start; |
fbdc762b CM |
1667 | if (full_scan) { |
1668 | ret = -ENOSPC; | |
1669 | goto error; | |
1670 | } | |
6702ed49 CM |
1671 | if (wrapped) { |
1672 | if (!full_scan) | |
1673 | total_needed -= empty_size; | |
fbdc762b | 1674 | full_scan = 1; |
6702ed49 | 1675 | } else |
fbdc762b | 1676 | wrapped = 1; |
be744175 | 1677 | } |
5276aeda | 1678 | block_group = btrfs_lookup_block_group(info, search_start); |
fbdc762b | 1679 | cond_resched(); |
1a2b2ac7 CM |
1680 | block_group = btrfs_find_block_group(root, block_group, |
1681 | search_start, data, 0); | |
be744175 CM |
1682 | goto check_failed; |
1683 | ||
0f70abe2 | 1684 | error: |
0f70abe2 | 1685 | return ret; |
fec577fb | 1686 | } |
fec577fb CM |
1687 | /* |
1688 | * finds a free extent and does all the dirty work required for allocation | |
1689 | * returns the key for the extent through ins, and a tree buffer for | |
1690 | * the first block of the extent through buf. | |
1691 | * | |
1692 | * returns 0 if everything worked, non-zero otherwise. | |
1693 | */ | |
4d775673 | 1694 | int btrfs_alloc_extent(struct btrfs_trans_handle *trans, |
7bb86316 | 1695 | struct btrfs_root *root, |
98d20f67 CM |
1696 | u64 num_bytes, u64 min_alloc_size, |
1697 | u64 root_objectid, u64 ref_generation, | |
7bb86316 CM |
1698 | u64 owner, u64 owner_offset, |
1699 | u64 empty_size, u64 hint_byte, | |
be08c1b9 | 1700 | u64 search_end, struct btrfs_key *ins, int data) |
fec577fb CM |
1701 | { |
1702 | int ret; | |
1703 | int pending_ret; | |
c31f8830 CM |
1704 | u64 super_used; |
1705 | u64 root_used; | |
fbdc762b | 1706 | u64 search_start = 0; |
8790d502 | 1707 | u64 alloc_profile; |
47e4bb98 | 1708 | u32 sizes[2]; |
1261ec42 CM |
1709 | struct btrfs_fs_info *info = root->fs_info; |
1710 | struct btrfs_root *extent_root = info->extent_root; | |
47e4bb98 CM |
1711 | struct btrfs_extent_item *extent_item; |
1712 | struct btrfs_extent_ref *ref; | |
7bb86316 | 1713 | struct btrfs_path *path; |
47e4bb98 | 1714 | struct btrfs_key keys[2]; |
8f662a76 | 1715 | |
6324fbf3 | 1716 | if (data) { |
8790d502 CM |
1717 | alloc_profile = info->avail_data_alloc_bits & |
1718 | info->data_alloc_profile; | |
1719 | data = BTRFS_BLOCK_GROUP_DATA | alloc_profile; | |
6324fbf3 | 1720 | } else if (root == root->fs_info->chunk_root) { |
8790d502 CM |
1721 | alloc_profile = info->avail_system_alloc_bits & |
1722 | info->system_alloc_profile; | |
1723 | data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile; | |
6324fbf3 | 1724 | } else { |
8790d502 CM |
1725 | alloc_profile = info->avail_metadata_alloc_bits & |
1726 | info->metadata_alloc_profile; | |
1727 | data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile; | |
6324fbf3 | 1728 | } |
98d20f67 | 1729 | again: |
3bf3d9e9 | 1730 | if (root->ref_cows) { |
593060d7 | 1731 | if (!(data & BTRFS_BLOCK_GROUP_METADATA)) { |
6324fbf3 | 1732 | ret = do_chunk_alloc(trans, root->fs_info->extent_root, |
239b14b3 | 1733 | 2 * 1024 * 1024, |
593060d7 | 1734 | BTRFS_BLOCK_GROUP_METADATA | |
8790d502 CM |
1735 | (info->metadata_alloc_profile & |
1736 | info->avail_metadata_alloc_bits)); | |
6324fbf3 CM |
1737 | BUG_ON(ret); |
1738 | } | |
1739 | ret = do_chunk_alloc(trans, root->fs_info->extent_root, | |
8790d502 | 1740 | num_bytes + 2 * 1024 * 1024, data); |
6324fbf3 CM |
1741 | BUG_ON(ret); |
1742 | } | |
0b86a832 | 1743 | |
db94535d CM |
1744 | WARN_ON(num_bytes < root->sectorsize); |
1745 | ret = find_free_extent(trans, root, num_bytes, empty_size, | |
1746 | search_start, search_end, hint_byte, ins, | |
26b8003f CM |
1747 | trans->alloc_exclude_start, |
1748 | trans->alloc_exclude_nr, data); | |
3b951516 | 1749 | |
98d20f67 CM |
1750 | if (ret == -ENOSPC && num_bytes > min_alloc_size) { |
1751 | num_bytes = num_bytes >> 1; | |
1752 | num_bytes = max(num_bytes, min_alloc_size); | |
1753 | goto again; | |
1754 | } | |
ccd467d6 | 1755 | BUG_ON(ret); |
f2654de4 CM |
1756 | if (ret) |
1757 | return ret; | |
fec577fb | 1758 | |
58176a96 | 1759 | /* block accounting for super block */ |
db94535d CM |
1760 | super_used = btrfs_super_bytes_used(&info->super_copy); |
1761 | btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes); | |
26b8003f | 1762 | |
58176a96 | 1763 | /* block accounting for root item */ |
db94535d CM |
1764 | root_used = btrfs_root_used(&root->root_item); |
1765 | btrfs_set_root_used(&root->root_item, root_used + num_bytes); | |
58176a96 | 1766 | |
f510cfec CM |
1767 | clear_extent_dirty(&root->fs_info->free_space_cache, |
1768 | ins->objectid, ins->objectid + ins->offset - 1, | |
1769 | GFP_NOFS); | |
1770 | ||
26b8003f | 1771 | if (root == extent_root) { |
1a5bc167 CM |
1772 | set_extent_bits(&root->fs_info->extent_ins, ins->objectid, |
1773 | ins->objectid + ins->offset - 1, | |
1774 | EXTENT_LOCKED, GFP_NOFS); | |
26b8003f CM |
1775 | goto update_block; |
1776 | } | |
1777 | ||
1778 | WARN_ON(trans->alloc_exclude_nr); | |
1779 | trans->alloc_exclude_start = ins->objectid; | |
1780 | trans->alloc_exclude_nr = ins->offset; | |
037e6390 | 1781 | |
47e4bb98 CM |
1782 | memcpy(&keys[0], ins, sizeof(*ins)); |
1783 | keys[1].offset = hash_extent_ref(root_objectid, ref_generation, | |
1784 | owner, owner_offset); | |
1785 | keys[1].objectid = ins->objectid; | |
1786 | keys[1].type = BTRFS_EXTENT_REF_KEY; | |
1787 | sizes[0] = sizeof(*extent_item); | |
1788 | sizes[1] = sizeof(*ref); | |
7bb86316 CM |
1789 | |
1790 | path = btrfs_alloc_path(); | |
1791 | BUG_ON(!path); | |
47e4bb98 CM |
1792 | |
1793 | ret = btrfs_insert_empty_items(trans, extent_root, path, keys, | |
1794 | sizes, 2); | |
26b8003f | 1795 | |
ccd467d6 | 1796 | BUG_ON(ret); |
47e4bb98 CM |
1797 | extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0], |
1798 | struct btrfs_extent_item); | |
1799 | btrfs_set_extent_refs(path->nodes[0], extent_item, 1); | |
1800 | ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1, | |
1801 | struct btrfs_extent_ref); | |
1802 | ||
1803 | btrfs_set_ref_root(path->nodes[0], ref, root_objectid); | |
1804 | btrfs_set_ref_generation(path->nodes[0], ref, ref_generation); | |
1805 | btrfs_set_ref_objectid(path->nodes[0], ref, owner); | |
1806 | btrfs_set_ref_offset(path->nodes[0], ref, owner_offset); | |
1807 | ||
1808 | btrfs_mark_buffer_dirty(path->nodes[0]); | |
1809 | ||
1810 | trans->alloc_exclude_start = 0; | |
1811 | trans->alloc_exclude_nr = 0; | |
7bb86316 | 1812 | btrfs_free_path(path); |
e089f05c | 1813 | finish_current_insert(trans, extent_root); |
e20d96d6 | 1814 | pending_ret = del_pending_extents(trans, extent_root); |
f510cfec | 1815 | |
e37c9e69 | 1816 | if (ret) { |
037e6390 | 1817 | return ret; |
e37c9e69 CM |
1818 | } |
1819 | if (pending_ret) { | |
037e6390 | 1820 | return pending_ret; |
e37c9e69 | 1821 | } |
26b8003f CM |
1822 | |
1823 | update_block: | |
0b86a832 | 1824 | ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0); |
f5947066 CM |
1825 | if (ret) { |
1826 | printk("update block group failed for %Lu %Lu\n", | |
1827 | ins->objectid, ins->offset); | |
1828 | BUG(); | |
1829 | } | |
037e6390 | 1830 | return 0; |
fec577fb CM |
1831 | } |
1832 | ||
1833 | /* | |
1834 | * helper function to allocate a block for a given tree | |
1835 | * returns the tree buffer or NULL. | |
1836 | */ | |
5f39d397 | 1837 | struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, |
db94535d | 1838 | struct btrfs_root *root, |
7bb86316 CM |
1839 | u32 blocksize, |
1840 | u64 root_objectid, u64 hint, | |
1841 | u64 empty_size) | |
1842 | { | |
1843 | u64 ref_generation; | |
1844 | ||
1845 | if (root->ref_cows) | |
1846 | ref_generation = trans->transid; | |
1847 | else | |
1848 | ref_generation = 0; | |
1849 | ||
1850 | ||
1851 | return __btrfs_alloc_free_block(trans, root, blocksize, root_objectid, | |
1852 | ref_generation, 0, 0, hint, empty_size); | |
1853 | } | |
1854 | ||
1855 | /* | |
1856 | * helper function to allocate a block for a given tree | |
1857 | * returns the tree buffer or NULL. | |
1858 | */ | |
1859 | struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans, | |
1860 | struct btrfs_root *root, | |
1861 | u32 blocksize, | |
1862 | u64 root_objectid, | |
1863 | u64 ref_generation, | |
1864 | u64 first_objectid, | |
1865 | int level, | |
1866 | u64 hint, | |
5f39d397 | 1867 | u64 empty_size) |
fec577fb | 1868 | { |
e2fa7227 | 1869 | struct btrfs_key ins; |
fec577fb | 1870 | int ret; |
5f39d397 | 1871 | struct extent_buffer *buf; |
fec577fb | 1872 | |
98d20f67 | 1873 | ret = btrfs_alloc_extent(trans, root, blocksize, blocksize, |
7bb86316 | 1874 | root_objectid, ref_generation, |
f6dbff55 | 1875 | level, first_objectid, empty_size, hint, |
db94535d | 1876 | (u64)-1, &ins, 0); |
fec577fb | 1877 | if (ret) { |
54aa1f4d CM |
1878 | BUG_ON(ret > 0); |
1879 | return ERR_PTR(ret); | |
fec577fb | 1880 | } |
db94535d | 1881 | buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize); |
54aa1f4d | 1882 | if (!buf) { |
7bb86316 CM |
1883 | btrfs_free_extent(trans, root, ins.objectid, blocksize, |
1884 | root->root_key.objectid, ref_generation, | |
1885 | 0, 0, 0); | |
54aa1f4d CM |
1886 | return ERR_PTR(-ENOMEM); |
1887 | } | |
55c69072 CM |
1888 | btrfs_set_header_generation(buf, trans->transid); |
1889 | clean_tree_block(trans, root, buf); | |
5f39d397 | 1890 | btrfs_set_buffer_uptodate(buf); |
55c69072 CM |
1891 | |
1892 | if (PageDirty(buf->first_page)) { | |
1893 | printk("page %lu dirty\n", buf->first_page->index); | |
1894 | WARN_ON(1); | |
1895 | } | |
1896 | ||
5f39d397 CM |
1897 | set_extent_dirty(&trans->transaction->dirty_pages, buf->start, |
1898 | buf->start + buf->len - 1, GFP_NOFS); | |
9afbb0b7 CM |
1899 | if (!btrfs_test_opt(root, SSD)) |
1900 | btrfs_set_buffer_defrag(buf); | |
d3c2fdcf | 1901 | trans->blocks_used++; |
fec577fb CM |
1902 | return buf; |
1903 | } | |
a28ec197 | 1904 | |
98ed5174 CM |
1905 | static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans, |
1906 | struct btrfs_root *root, | |
1907 | struct extent_buffer *leaf) | |
6407bf6d | 1908 | { |
7bb86316 CM |
1909 | u64 leaf_owner; |
1910 | u64 leaf_generation; | |
5f39d397 | 1911 | struct btrfs_key key; |
6407bf6d CM |
1912 | struct btrfs_file_extent_item *fi; |
1913 | int i; | |
1914 | int nritems; | |
1915 | int ret; | |
1916 | ||
5f39d397 CM |
1917 | BUG_ON(!btrfs_is_leaf(leaf)); |
1918 | nritems = btrfs_header_nritems(leaf); | |
7bb86316 CM |
1919 | leaf_owner = btrfs_header_owner(leaf); |
1920 | leaf_generation = btrfs_header_generation(leaf); | |
1921 | ||
6407bf6d | 1922 | for (i = 0; i < nritems; i++) { |
db94535d | 1923 | u64 disk_bytenr; |
5f39d397 CM |
1924 | |
1925 | btrfs_item_key_to_cpu(leaf, &key, i); | |
1926 | if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) | |
6407bf6d CM |
1927 | continue; |
1928 | fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item); | |
5f39d397 CM |
1929 | if (btrfs_file_extent_type(leaf, fi) == |
1930 | BTRFS_FILE_EXTENT_INLINE) | |
236454df | 1931 | continue; |
6407bf6d CM |
1932 | /* |
1933 | * FIXME make sure to insert a trans record that | |
1934 | * repeats the snapshot del on crash | |
1935 | */ | |
db94535d CM |
1936 | disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); |
1937 | if (disk_bytenr == 0) | |
3a686375 | 1938 | continue; |
db94535d | 1939 | ret = btrfs_free_extent(trans, root, disk_bytenr, |
7bb86316 CM |
1940 | btrfs_file_extent_disk_num_bytes(leaf, fi), |
1941 | leaf_owner, leaf_generation, | |
1942 | key.objectid, key.offset, 0); | |
6407bf6d CM |
1943 | BUG_ON(ret); |
1944 | } | |
1945 | return 0; | |
1946 | } | |
1947 | ||
98ed5174 | 1948 | static void noinline reada_walk_down(struct btrfs_root *root, |
bea495e5 CM |
1949 | struct extent_buffer *node, |
1950 | int slot) | |
e011599b | 1951 | { |
db94535d | 1952 | u64 bytenr; |
bea495e5 CM |
1953 | u64 last = 0; |
1954 | u32 nritems; | |
e011599b | 1955 | u32 refs; |
db94535d | 1956 | u32 blocksize; |
bea495e5 CM |
1957 | int ret; |
1958 | int i; | |
1959 | int level; | |
1960 | int skipped = 0; | |
e011599b | 1961 | |
5f39d397 | 1962 | nritems = btrfs_header_nritems(node); |
db94535d | 1963 | level = btrfs_header_level(node); |
bea495e5 CM |
1964 | if (level) |
1965 | return; | |
1966 | ||
1967 | for (i = slot; i < nritems && skipped < 32; i++) { | |
db94535d | 1968 | bytenr = btrfs_node_blockptr(node, i); |
bea495e5 CM |
1969 | if (last && ((bytenr > last && bytenr - last > 32 * 1024) || |
1970 | (last > bytenr && last - bytenr > 32 * 1024))) { | |
1971 | skipped++; | |
e011599b | 1972 | continue; |
bea495e5 CM |
1973 | } |
1974 | blocksize = btrfs_level_size(root, level - 1); | |
1975 | if (i != slot) { | |
1976 | ret = lookup_extent_ref(NULL, root, bytenr, | |
1977 | blocksize, &refs); | |
1978 | BUG_ON(ret); | |
1979 | if (refs != 1) { | |
1980 | skipped++; | |
1981 | continue; | |
1982 | } | |
1983 | } | |
409eb95d | 1984 | mutex_unlock(&root->fs_info->fs_mutex); |
db94535d | 1985 | ret = readahead_tree_block(root, bytenr, blocksize); |
bea495e5 | 1986 | last = bytenr + blocksize; |
409eb95d CM |
1987 | cond_resched(); |
1988 | mutex_lock(&root->fs_info->fs_mutex); | |
e011599b CM |
1989 | if (ret) |
1990 | break; | |
1991 | } | |
1992 | } | |
1993 | ||
9aca1d51 CM |
1994 | /* |
1995 | * helper function for drop_snapshot, this walks down the tree dropping ref | |
1996 | * counts as it goes. | |
1997 | */ | |
98ed5174 CM |
1998 | static int noinline walk_down_tree(struct btrfs_trans_handle *trans, |
1999 | struct btrfs_root *root, | |
2000 | struct btrfs_path *path, int *level) | |
20524f02 | 2001 | { |
7bb86316 CM |
2002 | u64 root_owner; |
2003 | u64 root_gen; | |
2004 | u64 bytenr; | |
5f39d397 CM |
2005 | struct extent_buffer *next; |
2006 | struct extent_buffer *cur; | |
7bb86316 | 2007 | struct extent_buffer *parent; |
db94535d | 2008 | u32 blocksize; |
20524f02 CM |
2009 | int ret; |
2010 | u32 refs; | |
2011 | ||
5caf2a00 CM |
2012 | WARN_ON(*level < 0); |
2013 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
5f39d397 | 2014 | ret = lookup_extent_ref(trans, root, |
db94535d CM |
2015 | path->nodes[*level]->start, |
2016 | path->nodes[*level]->len, &refs); | |
20524f02 CM |
2017 | BUG_ON(ret); |
2018 | if (refs > 1) | |
2019 | goto out; | |
e011599b | 2020 | |
9aca1d51 CM |
2021 | /* |
2022 | * walk down to the last node level and free all the leaves | |
2023 | */ | |
6407bf6d | 2024 | while(*level >= 0) { |
5caf2a00 CM |
2025 | WARN_ON(*level < 0); |
2026 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
20524f02 | 2027 | cur = path->nodes[*level]; |
e011599b | 2028 | |
5f39d397 | 2029 | if (btrfs_header_level(cur) != *level) |
2c90e5d6 | 2030 | WARN_ON(1); |
e011599b | 2031 | |
7518a238 | 2032 | if (path->slots[*level] >= |
5f39d397 | 2033 | btrfs_header_nritems(cur)) |
20524f02 | 2034 | break; |
6407bf6d CM |
2035 | if (*level == 0) { |
2036 | ret = drop_leaf_ref(trans, root, cur); | |
2037 | BUG_ON(ret); | |
2038 | break; | |
2039 | } | |
db94535d CM |
2040 | bytenr = btrfs_node_blockptr(cur, path->slots[*level]); |
2041 | blocksize = btrfs_level_size(root, *level - 1); | |
2042 | ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs); | |
6407bf6d CM |
2043 | BUG_ON(ret); |
2044 | if (refs != 1) { | |
7bb86316 CM |
2045 | parent = path->nodes[*level]; |
2046 | root_owner = btrfs_header_owner(parent); | |
2047 | root_gen = btrfs_header_generation(parent); | |
20524f02 | 2048 | path->slots[*level]++; |
db94535d | 2049 | ret = btrfs_free_extent(trans, root, bytenr, |
7bb86316 CM |
2050 | blocksize, root_owner, |
2051 | root_gen, 0, 0, 1); | |
20524f02 CM |
2052 | BUG_ON(ret); |
2053 | continue; | |
2054 | } | |
db94535d | 2055 | next = btrfs_find_tree_block(root, bytenr, blocksize); |
5f39d397 CM |
2056 | if (!next || !btrfs_buffer_uptodate(next)) { |
2057 | free_extent_buffer(next); | |
bea495e5 | 2058 | reada_walk_down(root, cur, path->slots[*level]); |
8790d502 CM |
2059 | |
2060 | mutex_unlock(&root->fs_info->fs_mutex); | |
db94535d | 2061 | next = read_tree_block(root, bytenr, blocksize); |
8790d502 | 2062 | mutex_lock(&root->fs_info->fs_mutex); |
e9d0b13b | 2063 | |
8790d502 | 2064 | /* we've dropped the lock, double check */ |
db94535d CM |
2065 | ret = lookup_extent_ref(trans, root, bytenr, |
2066 | blocksize, &refs); | |
e9d0b13b CM |
2067 | BUG_ON(ret); |
2068 | if (refs != 1) { | |
7bb86316 CM |
2069 | parent = path->nodes[*level]; |
2070 | root_owner = btrfs_header_owner(parent); | |
2071 | root_gen = btrfs_header_generation(parent); | |
2072 | ||
e9d0b13b | 2073 | path->slots[*level]++; |
5f39d397 | 2074 | free_extent_buffer(next); |
7bb86316 CM |
2075 | ret = btrfs_free_extent(trans, root, bytenr, |
2076 | blocksize, | |
2077 | root_owner, | |
2078 | root_gen, 0, 0, 1); | |
e9d0b13b CM |
2079 | BUG_ON(ret); |
2080 | continue; | |
2081 | } | |
0999df54 CM |
2082 | } else if (next) { |
2083 | btrfs_verify_block_csum(root, next); | |
e9d0b13b | 2084 | } |
5caf2a00 | 2085 | WARN_ON(*level <= 0); |
83e15a28 | 2086 | if (path->nodes[*level-1]) |
5f39d397 | 2087 | free_extent_buffer(path->nodes[*level-1]); |
20524f02 | 2088 | path->nodes[*level-1] = next; |
5f39d397 | 2089 | *level = btrfs_header_level(next); |
20524f02 CM |
2090 | path->slots[*level] = 0; |
2091 | } | |
2092 | out: | |
5caf2a00 CM |
2093 | WARN_ON(*level < 0); |
2094 | WARN_ON(*level >= BTRFS_MAX_LEVEL); | |
7bb86316 CM |
2095 | |
2096 | if (path->nodes[*level] == root->node) { | |
2097 | root_owner = root->root_key.objectid; | |
2098 | parent = path->nodes[*level]; | |
2099 | } else { | |
2100 | parent = path->nodes[*level + 1]; | |
2101 | root_owner = btrfs_header_owner(parent); | |
2102 | } | |
2103 | ||
2104 | root_gen = btrfs_header_generation(parent); | |
db94535d | 2105 | ret = btrfs_free_extent(trans, root, path->nodes[*level]->start, |
7bb86316 CM |
2106 | path->nodes[*level]->len, |
2107 | root_owner, root_gen, 0, 0, 1); | |
5f39d397 | 2108 | free_extent_buffer(path->nodes[*level]); |
20524f02 CM |
2109 | path->nodes[*level] = NULL; |
2110 | *level += 1; | |
2111 | BUG_ON(ret); | |
2112 | return 0; | |
2113 | } | |
2114 | ||
9aca1d51 CM |
2115 | /* |
2116 | * helper for dropping snapshots. This walks back up the tree in the path | |
2117 | * to find the first node higher up where we haven't yet gone through | |
2118 | * all the slots | |
2119 | */ | |
98ed5174 CM |
2120 | static int noinline walk_up_tree(struct btrfs_trans_handle *trans, |
2121 | struct btrfs_root *root, | |
2122 | struct btrfs_path *path, int *level) | |
20524f02 | 2123 | { |
7bb86316 CM |
2124 | u64 root_owner; |
2125 | u64 root_gen; | |
2126 | struct btrfs_root_item *root_item = &root->root_item; | |
20524f02 CM |
2127 | int i; |
2128 | int slot; | |
2129 | int ret; | |
9f3a7427 | 2130 | |
234b63a0 | 2131 | for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { |
20524f02 | 2132 | slot = path->slots[i]; |
5f39d397 CM |
2133 | if (slot < btrfs_header_nritems(path->nodes[i]) - 1) { |
2134 | struct extent_buffer *node; | |
2135 | struct btrfs_disk_key disk_key; | |
2136 | node = path->nodes[i]; | |
20524f02 CM |
2137 | path->slots[i]++; |
2138 | *level = i; | |
9f3a7427 | 2139 | WARN_ON(*level == 0); |
5f39d397 | 2140 | btrfs_node_key(node, &disk_key, path->slots[i]); |
9f3a7427 | 2141 | memcpy(&root_item->drop_progress, |
5f39d397 | 2142 | &disk_key, sizeof(disk_key)); |
9f3a7427 | 2143 | root_item->drop_level = i; |
20524f02 CM |
2144 | return 0; |
2145 | } else { | |
7bb86316 CM |
2146 | if (path->nodes[*level] == root->node) { |
2147 | root_owner = root->root_key.objectid; | |
2148 | root_gen = | |
2149 | btrfs_header_generation(path->nodes[*level]); | |
2150 | } else { | |
2151 | struct extent_buffer *node; | |
2152 | node = path->nodes[*level + 1]; | |
2153 | root_owner = btrfs_header_owner(node); | |
2154 | root_gen = btrfs_header_generation(node); | |
2155 | } | |
e089f05c | 2156 | ret = btrfs_free_extent(trans, root, |
db94535d | 2157 | path->nodes[*level]->start, |
7bb86316 CM |
2158 | path->nodes[*level]->len, |
2159 | root_owner, root_gen, 0, 0, 1); | |
6407bf6d | 2160 | BUG_ON(ret); |
5f39d397 | 2161 | free_extent_buffer(path->nodes[*level]); |
83e15a28 | 2162 | path->nodes[*level] = NULL; |
20524f02 | 2163 | *level = i + 1; |
20524f02 CM |
2164 | } |
2165 | } | |
2166 | return 1; | |
2167 | } | |
2168 | ||
9aca1d51 CM |
2169 | /* |
2170 | * drop the reference count on the tree rooted at 'snap'. This traverses | |
2171 | * the tree freeing any blocks that have a ref count of zero after being | |
2172 | * decremented. | |
2173 | */ | |
e089f05c | 2174 | int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root |
9f3a7427 | 2175 | *root) |
20524f02 | 2176 | { |
3768f368 | 2177 | int ret = 0; |
9aca1d51 | 2178 | int wret; |
20524f02 | 2179 | int level; |
5caf2a00 | 2180 | struct btrfs_path *path; |
20524f02 CM |
2181 | int i; |
2182 | int orig_level; | |
9f3a7427 | 2183 | struct btrfs_root_item *root_item = &root->root_item; |
20524f02 | 2184 | |
5caf2a00 CM |
2185 | path = btrfs_alloc_path(); |
2186 | BUG_ON(!path); | |
20524f02 | 2187 | |
5f39d397 | 2188 | level = btrfs_header_level(root->node); |
20524f02 | 2189 | orig_level = level; |
9f3a7427 CM |
2190 | if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) { |
2191 | path->nodes[level] = root->node; | |
f510cfec | 2192 | extent_buffer_get(root->node); |
9f3a7427 CM |
2193 | path->slots[level] = 0; |
2194 | } else { | |
2195 | struct btrfs_key key; | |
5f39d397 CM |
2196 | struct btrfs_disk_key found_key; |
2197 | struct extent_buffer *node; | |
6702ed49 | 2198 | |
9f3a7427 | 2199 | btrfs_disk_key_to_cpu(&key, &root_item->drop_progress); |
6702ed49 CM |
2200 | level = root_item->drop_level; |
2201 | path->lowest_level = level; | |
9f3a7427 | 2202 | wret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
6702ed49 | 2203 | if (wret < 0) { |
9f3a7427 CM |
2204 | ret = wret; |
2205 | goto out; | |
2206 | } | |
5f39d397 CM |
2207 | node = path->nodes[level]; |
2208 | btrfs_node_key(node, &found_key, path->slots[level]); | |
2209 | WARN_ON(memcmp(&found_key, &root_item->drop_progress, | |
2210 | sizeof(found_key))); | |
9f3a7427 | 2211 | } |
20524f02 | 2212 | while(1) { |
5caf2a00 | 2213 | wret = walk_down_tree(trans, root, path, &level); |
9aca1d51 | 2214 | if (wret > 0) |
20524f02 | 2215 | break; |
9aca1d51 CM |
2216 | if (wret < 0) |
2217 | ret = wret; | |
2218 | ||
5caf2a00 | 2219 | wret = walk_up_tree(trans, root, path, &level); |
9aca1d51 | 2220 | if (wret > 0) |
20524f02 | 2221 | break; |
9aca1d51 CM |
2222 | if (wret < 0) |
2223 | ret = wret; | |
409eb95d | 2224 | ret = -EAGAIN; |
409eb95d | 2225 | break; |
20524f02 | 2226 | } |
83e15a28 | 2227 | for (i = 0; i <= orig_level; i++) { |
5caf2a00 | 2228 | if (path->nodes[i]) { |
5f39d397 | 2229 | free_extent_buffer(path->nodes[i]); |
0f82731f | 2230 | path->nodes[i] = NULL; |
83e15a28 | 2231 | } |
20524f02 | 2232 | } |
9f3a7427 | 2233 | out: |
5caf2a00 | 2234 | btrfs_free_path(path); |
9aca1d51 | 2235 | return ret; |
20524f02 | 2236 | } |
9078a3e1 | 2237 | |
96b5179d | 2238 | int btrfs_free_block_groups(struct btrfs_fs_info *info) |
9078a3e1 | 2239 | { |
96b5179d CM |
2240 | u64 start; |
2241 | u64 end; | |
b97f9203 | 2242 | u64 ptr; |
9078a3e1 | 2243 | int ret; |
9078a3e1 | 2244 | while(1) { |
96b5179d CM |
2245 | ret = find_first_extent_bit(&info->block_group_cache, 0, |
2246 | &start, &end, (unsigned int)-1); | |
2247 | if (ret) | |
9078a3e1 | 2248 | break; |
b97f9203 Y |
2249 | ret = get_state_private(&info->block_group_cache, start, &ptr); |
2250 | if (!ret) | |
2251 | kfree((void *)(unsigned long)ptr); | |
96b5179d CM |
2252 | clear_extent_bits(&info->block_group_cache, start, |
2253 | end, (unsigned int)-1, GFP_NOFS); | |
9078a3e1 | 2254 | } |
e37c9e69 | 2255 | while(1) { |
f510cfec CM |
2256 | ret = find_first_extent_bit(&info->free_space_cache, 0, |
2257 | &start, &end, EXTENT_DIRTY); | |
2258 | if (ret) | |
e37c9e69 | 2259 | break; |
f510cfec CM |
2260 | clear_extent_dirty(&info->free_space_cache, start, |
2261 | end, GFP_NOFS); | |
e37c9e69 | 2262 | } |
be744175 CM |
2263 | return 0; |
2264 | } | |
2265 | ||
8e7bf94f CM |
2266 | static unsigned long calc_ra(unsigned long start, unsigned long last, |
2267 | unsigned long nr) | |
2268 | { | |
2269 | return min(last, start + nr - 1); | |
2270 | } | |
2271 | ||
98ed5174 CM |
2272 | static int noinline relocate_inode_pages(struct inode *inode, u64 start, |
2273 | u64 len) | |
edbd8d4e CM |
2274 | { |
2275 | u64 page_start; | |
2276 | u64 page_end; | |
2277 | u64 delalloc_start; | |
2278 | u64 existing_delalloc; | |
2279 | unsigned long last_index; | |
edbd8d4e CM |
2280 | unsigned long i; |
2281 | struct page *page; | |
d1310b2e | 2282 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; |
4313b399 | 2283 | struct file_ra_state *ra; |
8e7bf94f CM |
2284 | unsigned long total_read = 0; |
2285 | unsigned long ra_pages; | |
4313b399 CM |
2286 | |
2287 | ra = kzalloc(sizeof(*ra), GFP_NOFS); | |
edbd8d4e CM |
2288 | |
2289 | mutex_lock(&inode->i_mutex); | |
4313b399 | 2290 | i = start >> PAGE_CACHE_SHIFT; |
edbd8d4e CM |
2291 | last_index = (start + len - 1) >> PAGE_CACHE_SHIFT; |
2292 | ||
8e7bf94f CM |
2293 | ra_pages = BTRFS_I(inode)->root->fs_info->bdi.ra_pages; |
2294 | ||
4313b399 | 2295 | file_ra_state_init(ra, inode->i_mapping); |
4313b399 | 2296 | kfree(ra); |
edbd8d4e | 2297 | |
4313b399 | 2298 | for (; i <= last_index; i++) { |
8e7bf94f CM |
2299 | if (total_read % ra_pages == 0) { |
2300 | btrfs_force_ra(inode->i_mapping, ra, NULL, i, | |
2301 | calc_ra(i, last_index, ra_pages)); | |
2302 | } | |
2303 | total_read++; | |
edbd8d4e CM |
2304 | page = grab_cache_page(inode->i_mapping, i); |
2305 | if (!page) | |
2306 | goto out_unlock; | |
2307 | if (!PageUptodate(page)) { | |
2308 | btrfs_readpage(NULL, page); | |
2309 | lock_page(page); | |
2310 | if (!PageUptodate(page)) { | |
2311 | unlock_page(page); | |
2312 | page_cache_release(page); | |
2313 | goto out_unlock; | |
2314 | } | |
2315 | } | |
2316 | page_start = (u64)page->index << PAGE_CACHE_SHIFT; | |
2317 | page_end = page_start + PAGE_CACHE_SIZE - 1; | |
2318 | ||
d1310b2e | 2319 | lock_extent(io_tree, page_start, page_end, GFP_NOFS); |
edbd8d4e CM |
2320 | |
2321 | delalloc_start = page_start; | |
d1310b2e CM |
2322 | existing_delalloc = count_range_bits(io_tree, |
2323 | &delalloc_start, page_end, | |
2324 | PAGE_CACHE_SIZE, EXTENT_DELALLOC); | |
edbd8d4e | 2325 | |
d1310b2e | 2326 | set_extent_delalloc(io_tree, page_start, |
edbd8d4e CM |
2327 | page_end, GFP_NOFS); |
2328 | ||
d1310b2e | 2329 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); |
edbd8d4e CM |
2330 | set_page_dirty(page); |
2331 | unlock_page(page); | |
2332 | page_cache_release(page); | |
2333 | } | |
2334 | ||
2335 | out_unlock: | |
2336 | mutex_unlock(&inode->i_mutex); | |
2337 | return 0; | |
2338 | } | |
2339 | ||
4313b399 CM |
2340 | /* |
2341 | * note, this releases the path | |
2342 | */ | |
98ed5174 | 2343 | static int noinline relocate_one_reference(struct btrfs_root *extent_root, |
edbd8d4e | 2344 | struct btrfs_path *path, |
4313b399 | 2345 | struct btrfs_key *extent_key) |
edbd8d4e CM |
2346 | { |
2347 | struct inode *inode; | |
2348 | struct btrfs_root *found_root; | |
4313b399 CM |
2349 | struct btrfs_key *root_location; |
2350 | struct btrfs_extent_ref *ref; | |
2351 | u64 ref_root; | |
2352 | u64 ref_gen; | |
2353 | u64 ref_objectid; | |
2354 | u64 ref_offset; | |
edbd8d4e CM |
2355 | int ret; |
2356 | ||
4313b399 CM |
2357 | ref = btrfs_item_ptr(path->nodes[0], path->slots[0], |
2358 | struct btrfs_extent_ref); | |
2359 | ref_root = btrfs_ref_root(path->nodes[0], ref); | |
2360 | ref_gen = btrfs_ref_generation(path->nodes[0], ref); | |
2361 | ref_objectid = btrfs_ref_objectid(path->nodes[0], ref); | |
2362 | ref_offset = btrfs_ref_offset(path->nodes[0], ref); | |
2363 | btrfs_release_path(extent_root, path); | |
2364 | ||
2365 | root_location = kmalloc(sizeof(*root_location), GFP_NOFS); | |
2366 | root_location->objectid = ref_root; | |
edbd8d4e | 2367 | if (ref_gen == 0) |
4313b399 | 2368 | root_location->offset = 0; |
edbd8d4e | 2369 | else |
4313b399 CM |
2370 | root_location->offset = (u64)-1; |
2371 | root_location->type = BTRFS_ROOT_ITEM_KEY; | |
edbd8d4e CM |
2372 | |
2373 | found_root = btrfs_read_fs_root_no_name(extent_root->fs_info, | |
4313b399 | 2374 | root_location); |
edbd8d4e | 2375 | BUG_ON(!found_root); |
4313b399 | 2376 | kfree(root_location); |
edbd8d4e CM |
2377 | |
2378 | if (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID) { | |
2379 | mutex_unlock(&extent_root->fs_info->fs_mutex); | |
2380 | inode = btrfs_iget_locked(extent_root->fs_info->sb, | |
2381 | ref_objectid, found_root); | |
2382 | if (inode->i_state & I_NEW) { | |
2383 | /* the inode and parent dir are two different roots */ | |
2384 | BTRFS_I(inode)->root = found_root; | |
2385 | BTRFS_I(inode)->location.objectid = ref_objectid; | |
2386 | BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY; | |
2387 | BTRFS_I(inode)->location.offset = 0; | |
2388 | btrfs_read_locked_inode(inode); | |
2389 | unlock_new_inode(inode); | |
2390 | ||
2391 | } | |
2392 | /* this can happen if the reference is not against | |
2393 | * the latest version of the tree root | |
2394 | */ | |
2395 | if (is_bad_inode(inode)) { | |
2396 | mutex_lock(&extent_root->fs_info->fs_mutex); | |
2397 | goto out; | |
2398 | } | |
2399 | relocate_inode_pages(inode, ref_offset, extent_key->offset); | |
2400 | /* FIXME, data=ordered will help get rid of this */ | |
2401 | filemap_fdatawrite(inode->i_mapping); | |
2402 | iput(inode); | |
2403 | mutex_lock(&extent_root->fs_info->fs_mutex); | |
2404 | } else { | |
2405 | struct btrfs_trans_handle *trans; | |
2406 | struct btrfs_key found_key; | |
2407 | struct extent_buffer *eb; | |
2408 | int level; | |
2409 | int i; | |
2410 | ||
2411 | trans = btrfs_start_transaction(found_root, 1); | |
2412 | eb = read_tree_block(found_root, extent_key->objectid, | |
2413 | extent_key->offset); | |
2414 | level = btrfs_header_level(eb); | |
2415 | ||
2416 | if (level == 0) | |
2417 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
2418 | else | |
2419 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
2420 | ||
2421 | free_extent_buffer(eb); | |
2422 | ||
2423 | path->lowest_level = level; | |
8f662a76 | 2424 | path->reada = 2; |
edbd8d4e CM |
2425 | ret = btrfs_search_slot(trans, found_root, &found_key, path, |
2426 | 0, 1); | |
2427 | path->lowest_level = 0; | |
2428 | for (i = level; i < BTRFS_MAX_LEVEL; i++) { | |
2429 | if (!path->nodes[i]) | |
2430 | break; | |
2431 | free_extent_buffer(path->nodes[i]); | |
2432 | path->nodes[i] = NULL; | |
2433 | } | |
2434 | btrfs_release_path(found_root, path); | |
2435 | btrfs_end_transaction(trans, found_root); | |
2436 | } | |
2437 | ||
2438 | out: | |
2439 | return 0; | |
2440 | } | |
2441 | ||
98ed5174 CM |
2442 | static int noinline relocate_one_extent(struct btrfs_root *extent_root, |
2443 | struct btrfs_path *path, | |
2444 | struct btrfs_key *extent_key) | |
edbd8d4e CM |
2445 | { |
2446 | struct btrfs_key key; | |
2447 | struct btrfs_key found_key; | |
edbd8d4e | 2448 | struct extent_buffer *leaf; |
edbd8d4e CM |
2449 | u32 nritems; |
2450 | u32 item_size; | |
2451 | int ret = 0; | |
2452 | ||
2453 | key.objectid = extent_key->objectid; | |
2454 | key.type = BTRFS_EXTENT_REF_KEY; | |
2455 | key.offset = 0; | |
2456 | ||
2457 | while(1) { | |
2458 | ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); | |
2459 | ||
edbd8d4e CM |
2460 | if (ret < 0) |
2461 | goto out; | |
2462 | ||
2463 | ret = 0; | |
2464 | leaf = path->nodes[0]; | |
2465 | nritems = btrfs_header_nritems(leaf); | |
2466 | if (path->slots[0] == nritems) | |
2467 | goto out; | |
2468 | ||
2469 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2470 | if (found_key.objectid != extent_key->objectid) | |
2471 | break; | |
2472 | ||
2473 | if (found_key.type != BTRFS_EXTENT_REF_KEY) | |
2474 | break; | |
2475 | ||
2476 | key.offset = found_key.offset + 1; | |
2477 | item_size = btrfs_item_size_nr(leaf, path->slots[0]); | |
2478 | ||
4313b399 | 2479 | ret = relocate_one_reference(extent_root, path, extent_key); |
edbd8d4e CM |
2480 | if (ret) |
2481 | goto out; | |
2482 | } | |
2483 | ret = 0; | |
2484 | out: | |
2485 | btrfs_release_path(extent_root, path); | |
2486 | return ret; | |
2487 | } | |
2488 | ||
8f18cf13 | 2489 | int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start) |
edbd8d4e CM |
2490 | { |
2491 | struct btrfs_trans_handle *trans; | |
2492 | struct btrfs_root *tree_root = root->fs_info->tree_root; | |
2493 | struct btrfs_path *path; | |
2494 | u64 cur_byte; | |
2495 | u64 total_found; | |
8f18cf13 CM |
2496 | u64 shrink_last_byte; |
2497 | struct btrfs_block_group_cache *shrink_block_group; | |
edbd8d4e | 2498 | struct btrfs_fs_info *info = root->fs_info; |
edbd8d4e | 2499 | struct btrfs_key key; |
73e48b27 | 2500 | struct btrfs_key found_key; |
edbd8d4e CM |
2501 | struct extent_buffer *leaf; |
2502 | u32 nritems; | |
2503 | int ret; | |
725c8463 | 2504 | int progress = 0; |
edbd8d4e | 2505 | |
8f18cf13 CM |
2506 | shrink_block_group = btrfs_lookup_block_group(root->fs_info, |
2507 | shrink_start); | |
2508 | BUG_ON(!shrink_block_group); | |
2509 | ||
2510 | shrink_last_byte = shrink_start + shrink_block_group->key.offset; | |
2511 | ||
2512 | shrink_block_group->space_info->total_bytes -= | |
2513 | shrink_block_group->key.offset; | |
2514 | printk("shrink_extent_tree %Lu -> %Lu type %Lu\n", shrink_start, shrink_last_byte, shrink_block_group->flags); | |
edbd8d4e CM |
2515 | path = btrfs_alloc_path(); |
2516 | root = root->fs_info->extent_root; | |
8f662a76 | 2517 | path->reada = 2; |
edbd8d4e CM |
2518 | |
2519 | again: | |
8f18cf13 CM |
2520 | trans = btrfs_start_transaction(root, 1); |
2521 | do_chunk_alloc(trans, root->fs_info->extent_root, | |
2522 | btrfs_block_group_used(&shrink_block_group->item) + | |
2523 | 2 * 1024 * 1024, shrink_block_group->flags); | |
2524 | btrfs_end_transaction(trans, root); | |
2525 | shrink_block_group->ro = 1; | |
2526 | ||
edbd8d4e | 2527 | total_found = 0; |
8f18cf13 | 2528 | key.objectid = shrink_start; |
edbd8d4e CM |
2529 | key.offset = 0; |
2530 | key.type = 0; | |
73e48b27 | 2531 | cur_byte = key.objectid; |
4313b399 | 2532 | |
73e48b27 Y |
2533 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2534 | if (ret < 0) | |
2535 | goto out; | |
2536 | ||
0b86a832 | 2537 | ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY); |
73e48b27 Y |
2538 | if (ret < 0) |
2539 | goto out; | |
8f18cf13 | 2540 | |
73e48b27 Y |
2541 | if (ret == 0) { |
2542 | leaf = path->nodes[0]; | |
2543 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
8f18cf13 CM |
2544 | if (found_key.objectid + found_key.offset > shrink_start && |
2545 | found_key.objectid < shrink_last_byte) { | |
73e48b27 Y |
2546 | cur_byte = found_key.objectid; |
2547 | key.objectid = cur_byte; | |
2548 | } | |
2549 | } | |
2550 | btrfs_release_path(root, path); | |
2551 | ||
2552 | while(1) { | |
edbd8d4e CM |
2553 | ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
2554 | if (ret < 0) | |
2555 | goto out; | |
73e48b27 | 2556 | |
edbd8d4e | 2557 | leaf = path->nodes[0]; |
73e48b27 Y |
2558 | nritems = btrfs_header_nritems(leaf); |
2559 | next: | |
2560 | if (path->slots[0] >= nritems) { | |
2561 | ret = btrfs_next_leaf(root, path); | |
2562 | if (ret < 0) | |
2563 | goto out; | |
2564 | if (ret == 1) { | |
2565 | ret = 0; | |
2566 | break; | |
edbd8d4e | 2567 | } |
73e48b27 Y |
2568 | leaf = path->nodes[0]; |
2569 | nritems = btrfs_header_nritems(leaf); | |
edbd8d4e | 2570 | } |
73e48b27 Y |
2571 | |
2572 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
725c8463 | 2573 | |
8f18cf13 CM |
2574 | if (found_key.objectid >= shrink_last_byte) |
2575 | break; | |
2576 | ||
725c8463 CM |
2577 | if (progress && need_resched()) { |
2578 | memcpy(&key, &found_key, sizeof(key)); | |
2579 | mutex_unlock(&root->fs_info->fs_mutex); | |
2580 | cond_resched(); | |
2581 | mutex_lock(&root->fs_info->fs_mutex); | |
2582 | btrfs_release_path(root, path); | |
2583 | btrfs_search_slot(NULL, root, &key, path, 0, 0); | |
2584 | progress = 0; | |
2585 | goto next; | |
2586 | } | |
2587 | progress = 1; | |
2588 | ||
73e48b27 Y |
2589 | if (btrfs_key_type(&found_key) != BTRFS_EXTENT_ITEM_KEY || |
2590 | found_key.objectid + found_key.offset <= cur_byte) { | |
edbd8d4e | 2591 | path->slots[0]++; |
edbd8d4e CM |
2592 | goto next; |
2593 | } | |
73e48b27 | 2594 | |
edbd8d4e CM |
2595 | total_found++; |
2596 | cur_byte = found_key.objectid + found_key.offset; | |
2597 | key.objectid = cur_byte; | |
2598 | btrfs_release_path(root, path); | |
2599 | ret = relocate_one_extent(root, path, &found_key); | |
2600 | } | |
2601 | ||
2602 | btrfs_release_path(root, path); | |
2603 | ||
2604 | if (total_found > 0) { | |
2605 | trans = btrfs_start_transaction(tree_root, 1); | |
2606 | btrfs_commit_transaction(trans, tree_root); | |
2607 | ||
2608 | mutex_unlock(&root->fs_info->fs_mutex); | |
2609 | btrfs_clean_old_snapshots(tree_root); | |
2610 | mutex_lock(&root->fs_info->fs_mutex); | |
2611 | ||
2612 | trans = btrfs_start_transaction(tree_root, 1); | |
2613 | btrfs_commit_transaction(trans, tree_root); | |
2614 | goto again; | |
2615 | } | |
2616 | ||
8f18cf13 CM |
2617 | /* |
2618 | * we've freed all the extents, now remove the block | |
2619 | * group item from the tree | |
2620 | */ | |
edbd8d4e | 2621 | trans = btrfs_start_transaction(root, 1); |
8f18cf13 | 2622 | memcpy(&key, &shrink_block_group->key, sizeof(key)); |
1372f8e6 | 2623 | |
8f18cf13 CM |
2624 | ret = btrfs_search_slot(trans, root, &key, path, -1, 1); |
2625 | if (ret > 0) | |
2626 | ret = -EIO; | |
2627 | if (ret < 0) | |
2628 | goto out; | |
73e48b27 | 2629 | |
8f18cf13 CM |
2630 | leaf = path->nodes[0]; |
2631 | nritems = btrfs_header_nritems(leaf); | |
2632 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
2633 | kfree(shrink_block_group); | |
edbd8d4e | 2634 | |
8f18cf13 CM |
2635 | clear_extent_bits(&info->block_group_cache, found_key.objectid, |
2636 | found_key.objectid + found_key.offset - 1, | |
2637 | (unsigned int)-1, GFP_NOFS); | |
edbd8d4e | 2638 | |
8f18cf13 CM |
2639 | btrfs_del_item(trans, root, path); |
2640 | clear_extent_dirty(&info->free_space_cache, | |
2641 | shrink_start, shrink_last_byte - 1, | |
edbd8d4e CM |
2642 | GFP_NOFS); |
2643 | btrfs_commit_transaction(trans, root); | |
2644 | out: | |
2645 | btrfs_free_path(path); | |
2646 | return ret; | |
2647 | } | |
2648 | ||
0b86a832 CM |
2649 | int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path, |
2650 | struct btrfs_key *key) | |
2651 | { | |
2652 | int ret; | |
2653 | struct btrfs_key found_key; | |
2654 | struct extent_buffer *leaf; | |
2655 | int slot; | |
edbd8d4e | 2656 | |
0b86a832 CM |
2657 | ret = btrfs_search_slot(NULL, root, key, path, 0, 0); |
2658 | if (ret < 0) | |
2659 | return ret; | |
2660 | while(1) { | |
2661 | slot = path->slots[0]; | |
edbd8d4e | 2662 | leaf = path->nodes[0]; |
0b86a832 CM |
2663 | if (slot >= btrfs_header_nritems(leaf)) { |
2664 | ret = btrfs_next_leaf(root, path); | |
2665 | if (ret == 0) | |
2666 | continue; | |
2667 | if (ret < 0) | |
2668 | goto error; | |
2669 | break; | |
edbd8d4e | 2670 | } |
0b86a832 | 2671 | btrfs_item_key_to_cpu(leaf, &found_key, slot); |
edbd8d4e | 2672 | |
0b86a832 CM |
2673 | if (found_key.objectid >= key->objectid && |
2674 | found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) | |
2675 | return 0; | |
2676 | path->slots[0]++; | |
edbd8d4e | 2677 | } |
0b86a832 CM |
2678 | ret = -ENOENT; |
2679 | error: | |
2680 | return ret; | |
edbd8d4e CM |
2681 | } |
2682 | ||
9078a3e1 CM |
2683 | int btrfs_read_block_groups(struct btrfs_root *root) |
2684 | { | |
2685 | struct btrfs_path *path; | |
2686 | int ret; | |
96b5179d | 2687 | int bit; |
9078a3e1 | 2688 | struct btrfs_block_group_cache *cache; |
be744175 | 2689 | struct btrfs_fs_info *info = root->fs_info; |
6324fbf3 | 2690 | struct btrfs_space_info *space_info; |
d1310b2e | 2691 | struct extent_io_tree *block_group_cache; |
9078a3e1 CM |
2692 | struct btrfs_key key; |
2693 | struct btrfs_key found_key; | |
5f39d397 | 2694 | struct extent_buffer *leaf; |
96b5179d CM |
2695 | |
2696 | block_group_cache = &info->block_group_cache; | |
be744175 | 2697 | root = info->extent_root; |
9078a3e1 | 2698 | key.objectid = 0; |
0b86a832 | 2699 | key.offset = 0; |
9078a3e1 | 2700 | btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); |
9078a3e1 CM |
2701 | path = btrfs_alloc_path(); |
2702 | if (!path) | |
2703 | return -ENOMEM; | |
2704 | ||
2705 | while(1) { | |
0b86a832 CM |
2706 | ret = find_first_block_group(root, path, &key); |
2707 | if (ret > 0) { | |
2708 | ret = 0; | |
2709 | goto error; | |
9078a3e1 | 2710 | } |
0b86a832 CM |
2711 | if (ret != 0) |
2712 | goto error; | |
2713 | ||
5f39d397 CM |
2714 | leaf = path->nodes[0]; |
2715 | btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); | |
8f18cf13 | 2716 | cache = kzalloc(sizeof(*cache), GFP_NOFS); |
9078a3e1 | 2717 | if (!cache) { |
0b86a832 | 2718 | ret = -ENOMEM; |
9078a3e1 CM |
2719 | break; |
2720 | } | |
3e1ad54f | 2721 | |
5f39d397 CM |
2722 | read_extent_buffer(leaf, &cache->item, |
2723 | btrfs_item_ptr_offset(leaf, path->slots[0]), | |
2724 | sizeof(cache->item)); | |
9078a3e1 | 2725 | memcpy(&cache->key, &found_key, sizeof(found_key)); |
0b86a832 | 2726 | |
9078a3e1 CM |
2727 | key.objectid = found_key.objectid + found_key.offset; |
2728 | btrfs_release_path(root, path); | |
0b86a832 CM |
2729 | cache->flags = btrfs_block_group_flags(&cache->item); |
2730 | bit = 0; | |
2731 | if (cache->flags & BTRFS_BLOCK_GROUP_DATA) { | |
96b5179d | 2732 | bit = BLOCK_GROUP_DATA; |
0b86a832 CM |
2733 | } else if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) { |
2734 | bit = BLOCK_GROUP_SYSTEM; | |
2735 | } else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { | |
96b5179d | 2736 | bit = BLOCK_GROUP_METADATA; |
31f3c99b | 2737 | } |
8790d502 | 2738 | set_avail_alloc_bits(info, cache->flags); |
96b5179d | 2739 | |
6324fbf3 CM |
2740 | ret = update_space_info(info, cache->flags, found_key.offset, |
2741 | btrfs_block_group_used(&cache->item), | |
2742 | &space_info); | |
2743 | BUG_ON(ret); | |
2744 | cache->space_info = space_info; | |
2745 | ||
96b5179d CM |
2746 | /* use EXTENT_LOCKED to prevent merging */ |
2747 | set_extent_bits(block_group_cache, found_key.objectid, | |
2748 | found_key.objectid + found_key.offset - 1, | |
2749 | bit | EXTENT_LOCKED, GFP_NOFS); | |
2750 | set_state_private(block_group_cache, found_key.objectid, | |
ae2f5411 | 2751 | (unsigned long)cache); |
96b5179d | 2752 | |
9078a3e1 | 2753 | if (key.objectid >= |
db94535d | 2754 | btrfs_super_total_bytes(&info->super_copy)) |
9078a3e1 CM |
2755 | break; |
2756 | } | |
0b86a832 CM |
2757 | ret = 0; |
2758 | error: | |
9078a3e1 | 2759 | btrfs_free_path(path); |
0b86a832 | 2760 | return ret; |
9078a3e1 | 2761 | } |
6324fbf3 CM |
2762 | |
2763 | int btrfs_make_block_group(struct btrfs_trans_handle *trans, | |
2764 | struct btrfs_root *root, u64 bytes_used, | |
e17cade2 | 2765 | u64 type, u64 chunk_objectid, u64 chunk_offset, |
6324fbf3 CM |
2766 | u64 size) |
2767 | { | |
2768 | int ret; | |
2769 | int bit = 0; | |
2770 | struct btrfs_root *extent_root; | |
2771 | struct btrfs_block_group_cache *cache; | |
2772 | struct extent_io_tree *block_group_cache; | |
2773 | ||
2774 | extent_root = root->fs_info->extent_root; | |
2775 | block_group_cache = &root->fs_info->block_group_cache; | |
2776 | ||
8f18cf13 | 2777 | cache = kzalloc(sizeof(*cache), GFP_NOFS); |
6324fbf3 | 2778 | BUG_ON(!cache); |
e17cade2 | 2779 | cache->key.objectid = chunk_offset; |
6324fbf3 | 2780 | cache->key.offset = size; |
e17cade2 | 2781 | |
6324fbf3 CM |
2782 | btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY); |
2783 | memset(&cache->item, 0, sizeof(cache->item)); | |
2784 | btrfs_set_block_group_used(&cache->item, bytes_used); | |
6324fbf3 CM |
2785 | btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid); |
2786 | cache->flags = type; | |
2787 | btrfs_set_block_group_flags(&cache->item, type); | |
2788 | ||
2789 | ret = update_space_info(root->fs_info, cache->flags, size, bytes_used, | |
2790 | &cache->space_info); | |
2791 | BUG_ON(ret); | |
2792 | ||
d18a2c44 | 2793 | bit = block_group_state_bits(type); |
e17cade2 CM |
2794 | set_extent_bits(block_group_cache, chunk_offset, |
2795 | chunk_offset + size - 1, | |
6324fbf3 | 2796 | bit | EXTENT_LOCKED, GFP_NOFS); |
6324fbf3 | 2797 | |
e17cade2 CM |
2798 | set_state_private(block_group_cache, chunk_offset, |
2799 | (unsigned long)cache); | |
6324fbf3 CM |
2800 | ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item, |
2801 | sizeof(cache->item)); | |
2802 | BUG_ON(ret); | |
2803 | ||
2804 | finish_current_insert(trans, extent_root); | |
2805 | ret = del_pending_extents(trans, extent_root); | |
2806 | BUG_ON(ret); | |
d18a2c44 | 2807 | set_avail_alloc_bits(extent_root->fs_info, type); |
6324fbf3 CM |
2808 | return 0; |
2809 | } |