]>
Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
a74a4b97 | 19 | #include <linux/version.h> |
e20d96d6 | 20 | #include <linux/fs.h> |
d98237b3 | 21 | #include <linux/blkdev.h> |
87cbda5c | 22 | #include <linux/scatterlist.h> |
22b0ebda | 23 | #include <linux/swap.h> |
0f7d52f4 | 24 | #include <linux/radix-tree.h> |
35b7e476 | 25 | #include <linux/writeback.h> |
5f39d397 | 26 | #include <linux/buffer_head.h> // for block_sync_page |
ce9adaa5 | 27 | #include <linux/workqueue.h> |
a74a4b97 | 28 | #include <linux/kthread.h> |
a74a4b97 | 29 | # include <linux/freezer.h> |
a5eb62e3 | 30 | #include "crc32c.h" |
eb60ceac CM |
31 | #include "ctree.h" |
32 | #include "disk-io.h" | |
e089f05c | 33 | #include "transaction.h" |
0f7d52f4 | 34 | #include "btrfs_inode.h" |
0b86a832 | 35 | #include "volumes.h" |
db94535d | 36 | #include "print-tree.h" |
8b712842 | 37 | #include "async-thread.h" |
925baedd | 38 | #include "locking.h" |
017e5369 | 39 | #include "ref-cache.h" |
e02119d5 | 40 | #include "tree-log.h" |
eb60ceac | 41 | |
5f39d397 CM |
42 | #if 0 |
43 | static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) | |
7eccb903 | 44 | { |
5f39d397 CM |
45 | if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) { |
46 | printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n", | |
47 | (unsigned long long)extent_buffer_blocknr(buf), | |
48 | (unsigned long long)btrfs_header_blocknr(buf)); | |
39279cc3 | 49 | return 1; |
d98237b3 | 50 | } |
9a8dd150 | 51 | return 0; |
eb60ceac | 52 | } |
5f39d397 | 53 | #endif |
eb60ceac | 54 | |
d1310b2e | 55 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 56 | static void end_workqueue_fn(struct btrfs_work *work); |
ce9adaa5 CM |
57 | |
58 | struct end_io_wq { | |
59 | struct bio *bio; | |
60 | bio_end_io_t *end_io; | |
61 | void *private; | |
62 | struct btrfs_fs_info *info; | |
63 | int error; | |
22c59948 | 64 | int metadata; |
ce9adaa5 | 65 | struct list_head list; |
8b712842 | 66 | struct btrfs_work work; |
ce9adaa5 | 67 | }; |
0da5468f | 68 | |
44b8bd7e CM |
69 | struct async_submit_bio { |
70 | struct inode *inode; | |
71 | struct bio *bio; | |
72 | struct list_head list; | |
73 | extent_submit_bio_hook_t *submit_bio_hook; | |
74 | int rw; | |
75 | int mirror_num; | |
8b712842 | 76 | struct btrfs_work work; |
44b8bd7e CM |
77 | }; |
78 | ||
5f39d397 | 79 | struct extent_map *btree_get_extent(struct inode *inode, struct page *page, |
d1310b2e | 80 | size_t page_offset, u64 start, u64 len, |
5f39d397 | 81 | int create) |
7eccb903 | 82 | { |
5f39d397 CM |
83 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
84 | struct extent_map *em; | |
85 | int ret; | |
86 | ||
d1310b2e CM |
87 | spin_lock(&em_tree->lock); |
88 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
89 | if (em) { |
90 | em->bdev = | |
91 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
92 | spin_unlock(&em_tree->lock); | |
5f39d397 | 93 | goto out; |
a061fc8d CM |
94 | } |
95 | spin_unlock(&em_tree->lock); | |
7b13b7b1 | 96 | |
5f39d397 CM |
97 | em = alloc_extent_map(GFP_NOFS); |
98 | if (!em) { | |
99 | em = ERR_PTR(-ENOMEM); | |
100 | goto out; | |
101 | } | |
102 | em->start = 0; | |
0afbaf8c | 103 | em->len = (u64)-1; |
5f39d397 | 104 | em->block_start = 0; |
a061fc8d | 105 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
106 | |
107 | spin_lock(&em_tree->lock); | |
5f39d397 CM |
108 | ret = add_extent_mapping(em_tree, em); |
109 | if (ret == -EEXIST) { | |
0afbaf8c CM |
110 | u64 failed_start = em->start; |
111 | u64 failed_len = em->len; | |
112 | ||
113 | printk("failed to insert %Lu %Lu -> %Lu into tree\n", | |
114 | em->start, em->len, em->block_start); | |
5f39d397 | 115 | free_extent_map(em); |
7b13b7b1 | 116 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c CM |
117 | if (em) { |
118 | printk("after failing, found %Lu %Lu %Lu\n", | |
119 | em->start, em->len, em->block_start); | |
7b13b7b1 | 120 | ret = 0; |
0afbaf8c CM |
121 | } else { |
122 | em = lookup_extent_mapping(em_tree, failed_start, | |
123 | failed_len); | |
124 | if (em) { | |
125 | printk("double failure lookup gives us " | |
126 | "%Lu %Lu -> %Lu\n", em->start, | |
127 | em->len, em->block_start); | |
128 | free_extent_map(em); | |
129 | } | |
7b13b7b1 | 130 | ret = -EIO; |
0afbaf8c | 131 | } |
5f39d397 | 132 | } else if (ret) { |
7b13b7b1 CM |
133 | free_extent_map(em); |
134 | em = NULL; | |
5f39d397 | 135 | } |
7b13b7b1 CM |
136 | spin_unlock(&em_tree->lock); |
137 | ||
138 | if (ret) | |
139 | em = ERR_PTR(ret); | |
5f39d397 CM |
140 | out: |
141 | return em; | |
7eccb903 CM |
142 | } |
143 | ||
19c00ddc CM |
144 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
145 | { | |
a5eb62e3 | 146 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
147 | } |
148 | ||
149 | void btrfs_csum_final(u32 crc, char *result) | |
150 | { | |
151 | *(__le32 *)result = ~cpu_to_le32(crc); | |
152 | } | |
153 | ||
154 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, | |
155 | int verify) | |
156 | { | |
157 | char result[BTRFS_CRC32_SIZE]; | |
158 | unsigned long len; | |
159 | unsigned long cur_len; | |
160 | unsigned long offset = BTRFS_CSUM_SIZE; | |
161 | char *map_token = NULL; | |
162 | char *kaddr; | |
163 | unsigned long map_start; | |
164 | unsigned long map_len; | |
165 | int err; | |
166 | u32 crc = ~(u32)0; | |
167 | ||
168 | len = buf->len - offset; | |
169 | while(len > 0) { | |
170 | err = map_private_extent_buffer(buf, offset, 32, | |
171 | &map_token, &kaddr, | |
172 | &map_start, &map_len, KM_USER0); | |
173 | if (err) { | |
174 | printk("failed to map extent buffer! %lu\n", | |
175 | offset); | |
176 | return 1; | |
177 | } | |
178 | cur_len = min(len, map_len - (offset - map_start)); | |
179 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
180 | crc, cur_len); | |
181 | len -= cur_len; | |
182 | offset += cur_len; | |
183 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
184 | } | |
185 | btrfs_csum_final(crc, result); | |
186 | ||
187 | if (verify) { | |
e4204ded | 188 | /* FIXME, this is not good */ |
63b10fc4 | 189 | if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) { |
e4204ded CM |
190 | u32 val; |
191 | u32 found = 0; | |
192 | memcpy(&found, result, BTRFS_CRC32_SIZE); | |
193 | ||
194 | read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE); | |
195 | printk("btrfs: %s checksum verify failed on %llu " | |
2dd3e67b | 196 | "wanted %X found %X level %d\n", |
19c00ddc | 197 | root->fs_info->sb->s_id, |
2dd3e67b | 198 | buf->start, val, found, btrfs_header_level(buf)); |
19c00ddc CM |
199 | return 1; |
200 | } | |
201 | } else { | |
202 | write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE); | |
203 | } | |
204 | return 0; | |
205 | } | |
206 | ||
1259ab75 CM |
207 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
208 | struct extent_buffer *eb, u64 parent_transid) | |
209 | { | |
210 | int ret; | |
211 | ||
212 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
213 | return 0; | |
214 | ||
215 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
216 | if (extent_buffer_uptodate(io_tree, eb) && | |
217 | btrfs_header_generation(eb) == parent_transid) { | |
218 | ret = 0; | |
219 | goto out; | |
220 | } | |
221 | printk("parent transid verify failed on %llu wanted %llu found %llu\n", | |
222 | (unsigned long long)eb->start, | |
223 | (unsigned long long)parent_transid, | |
224 | (unsigned long long)btrfs_header_generation(eb)); | |
225 | ret = 1; | |
1259ab75 | 226 | clear_extent_buffer_uptodate(io_tree, eb); |
33958dc6 | 227 | out: |
1259ab75 CM |
228 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
229 | GFP_NOFS); | |
230 | return ret; | |
231 | ||
232 | } | |
233 | ||
f188591e CM |
234 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
235 | struct extent_buffer *eb, | |
ca7a79ad | 236 | u64 start, u64 parent_transid) |
f188591e CM |
237 | { |
238 | struct extent_io_tree *io_tree; | |
239 | int ret; | |
240 | int num_copies = 0; | |
241 | int mirror_num = 0; | |
242 | ||
243 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
244 | while (1) { | |
245 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
246 | btree_get_extent, mirror_num); | |
1259ab75 CM |
247 | if (!ret && |
248 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 249 | return ret; |
a1b32a59 | 250 | printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror_num); |
f188591e CM |
251 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
252 | eb->start, eb->len); | |
4235298e | 253 | if (num_copies == 1) |
f188591e | 254 | return ret; |
4235298e | 255 | |
f188591e | 256 | mirror_num++; |
4235298e | 257 | if (mirror_num > num_copies) |
f188591e | 258 | return ret; |
f188591e | 259 | } |
f188591e CM |
260 | return -EIO; |
261 | } | |
19c00ddc CM |
262 | |
263 | int csum_dirty_buffer(struct btrfs_root *root, struct page *page) | |
264 | { | |
d1310b2e | 265 | struct extent_io_tree *tree; |
35ebb934 | 266 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
267 | u64 found_start; |
268 | int found_level; | |
269 | unsigned long len; | |
270 | struct extent_buffer *eb; | |
f188591e CM |
271 | int ret; |
272 | ||
d1310b2e | 273 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
274 | |
275 | if (page->private == EXTENT_PAGE_PRIVATE) | |
276 | goto out; | |
277 | if (!page->private) | |
278 | goto out; | |
279 | len = page->private >> 2; | |
280 | if (len == 0) { | |
281 | WARN_ON(1); | |
282 | } | |
283 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
ca7a79ad CM |
284 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
285 | btrfs_header_generation(eb)); | |
f188591e | 286 | BUG_ON(ret); |
19c00ddc CM |
287 | found_start = btrfs_header_bytenr(eb); |
288 | if (found_start != start) { | |
289 | printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", | |
290 | start, found_start, len); | |
55c69072 CM |
291 | WARN_ON(1); |
292 | goto err; | |
293 | } | |
294 | if (eb->first_page != page) { | |
295 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
296 | page->index); | |
297 | WARN_ON(1); | |
298 | goto err; | |
299 | } | |
300 | if (!PageUptodate(page)) { | |
301 | printk("csum not up to date page %lu\n", page->index); | |
302 | WARN_ON(1); | |
303 | goto err; | |
19c00ddc CM |
304 | } |
305 | found_level = btrfs_header_level(eb); | |
4bef0848 | 306 | |
19c00ddc | 307 | csum_tree_block(root, eb, 0); |
55c69072 | 308 | err: |
19c00ddc CM |
309 | free_extent_buffer(eb); |
310 | out: | |
311 | return 0; | |
312 | } | |
313 | ||
ce9adaa5 CM |
314 | int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
315 | struct extent_state *state) | |
316 | { | |
317 | struct extent_io_tree *tree; | |
318 | u64 found_start; | |
319 | int found_level; | |
320 | unsigned long len; | |
321 | struct extent_buffer *eb; | |
322 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 323 | int ret = 0; |
ce9adaa5 CM |
324 | |
325 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
326 | if (page->private == EXTENT_PAGE_PRIVATE) | |
327 | goto out; | |
328 | if (!page->private) | |
329 | goto out; | |
330 | len = page->private >> 2; | |
331 | if (len == 0) { | |
332 | WARN_ON(1); | |
333 | } | |
334 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
f188591e | 335 | |
ce9adaa5 | 336 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 337 | if (found_start != start) { |
a1b32a59 CM |
338 | printk("bad tree block start %llu %llu\n", |
339 | (unsigned long long)found_start, | |
340 | (unsigned long long)eb->start); | |
f188591e | 341 | ret = -EIO; |
ce9adaa5 CM |
342 | goto err; |
343 | } | |
344 | if (eb->first_page != page) { | |
345 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
346 | page->index); | |
347 | WARN_ON(1); | |
f188591e | 348 | ret = -EIO; |
ce9adaa5 CM |
349 | goto err; |
350 | } | |
1259ab75 CM |
351 | if (memcmp_extent_buffer(eb, root->fs_info->fsid, |
352 | (unsigned long)btrfs_header_fsid(eb), | |
353 | BTRFS_FSID_SIZE)) { | |
354 | printk("bad fsid on block %Lu\n", eb->start); | |
355 | ret = -EIO; | |
356 | goto err; | |
357 | } | |
ce9adaa5 CM |
358 | found_level = btrfs_header_level(eb); |
359 | ||
360 | ret = csum_tree_block(root, eb, 1); | |
f188591e CM |
361 | if (ret) |
362 | ret = -EIO; | |
ce9adaa5 CM |
363 | |
364 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
365 | end = eb->start + end - 1; | |
ce9adaa5 CM |
366 | err: |
367 | free_extent_buffer(eb); | |
368 | out: | |
f188591e | 369 | return ret; |
ce9adaa5 CM |
370 | } |
371 | ||
ce9adaa5 | 372 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
373 | { |
374 | struct end_io_wq *end_io_wq = bio->bi_private; | |
375 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 376 | |
ce9adaa5 | 377 | fs_info = end_io_wq->info; |
ce9adaa5 | 378 | end_io_wq->error = err; |
8b712842 CM |
379 | end_io_wq->work.func = end_workqueue_fn; |
380 | end_io_wq->work.flags = 0; | |
e6dcd2dc CM |
381 | if (bio->bi_rw & (1 << BIO_RW)) |
382 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
383 | &end_io_wq->work); | |
384 | else | |
385 | btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); | |
ce9adaa5 CM |
386 | } |
387 | ||
22c59948 CM |
388 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
389 | int metadata) | |
0b86a832 | 390 | { |
ce9adaa5 | 391 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
392 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
393 | if (!end_io_wq) | |
394 | return -ENOMEM; | |
395 | ||
396 | end_io_wq->private = bio->bi_private; | |
397 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 398 | end_io_wq->info = info; |
ce9adaa5 CM |
399 | end_io_wq->error = 0; |
400 | end_io_wq->bio = bio; | |
22c59948 | 401 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
402 | |
403 | bio->bi_private = end_io_wq; | |
404 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
405 | return 0; |
406 | } | |
407 | ||
b64a2851 | 408 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 409 | { |
4854ddd0 CM |
410 | unsigned long limit = min_t(unsigned long, |
411 | info->workers.max_workers, | |
412 | info->fs_devices->open_devices); | |
413 | return 256 * limit; | |
414 | } | |
0986fe9e | 415 | |
4854ddd0 CM |
416 | int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) |
417 | { | |
b64a2851 CM |
418 | return atomic_read(&info->nr_async_bios) > |
419 | btrfs_async_submit_limit(info); | |
0986fe9e CM |
420 | } |
421 | ||
8b712842 CM |
422 | static void run_one_async_submit(struct btrfs_work *work) |
423 | { | |
424 | struct btrfs_fs_info *fs_info; | |
425 | struct async_submit_bio *async; | |
4854ddd0 | 426 | int limit; |
8b712842 CM |
427 | |
428 | async = container_of(work, struct async_submit_bio, work); | |
429 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 430 | |
b64a2851 | 431 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
432 | limit = limit * 2 / 3; |
433 | ||
8b712842 | 434 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 435 | |
b64a2851 CM |
436 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
437 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
438 | wake_up(&fs_info->async_submit_wait); |
439 | ||
8b712842 CM |
440 | async->submit_bio_hook(async->inode, async->rw, async->bio, |
441 | async->mirror_num); | |
442 | kfree(async); | |
443 | } | |
444 | ||
44b8bd7e CM |
445 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
446 | int rw, struct bio *bio, int mirror_num, | |
447 | extent_submit_bio_hook_t *submit_bio_hook) | |
448 | { | |
449 | struct async_submit_bio *async; | |
b64a2851 | 450 | int limit = btrfs_async_submit_limit(fs_info); |
44b8bd7e CM |
451 | |
452 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
453 | if (!async) | |
454 | return -ENOMEM; | |
455 | ||
456 | async->inode = inode; | |
457 | async->rw = rw; | |
458 | async->bio = bio; | |
459 | async->mirror_num = mirror_num; | |
460 | async->submit_bio_hook = submit_bio_hook; | |
8b712842 CM |
461 | async->work.func = run_one_async_submit; |
462 | async->work.flags = 0; | |
8c8bee1d CM |
463 | |
464 | while(atomic_read(&fs_info->async_submit_draining) && | |
465 | atomic_read(&fs_info->nr_async_submits)) { | |
466 | wait_event(fs_info->async_submit_wait, | |
467 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
468 | } | |
469 | ||
cb03c743 | 470 | atomic_inc(&fs_info->nr_async_submits); |
8b712842 | 471 | btrfs_queue_worker(&fs_info->workers, &async->work); |
4854ddd0 | 472 | |
9473f16c CM |
473 | if (atomic_read(&fs_info->nr_async_submits) > limit) { |
474 | wait_event_timeout(fs_info->async_submit_wait, | |
4854ddd0 CM |
475 | (atomic_read(&fs_info->nr_async_submits) < limit), |
476 | HZ/10); | |
9473f16c CM |
477 | |
478 | wait_event_timeout(fs_info->async_submit_wait, | |
479 | (atomic_read(&fs_info->nr_async_bios) < limit), | |
480 | HZ/10); | |
481 | } | |
44b8bd7e CM |
482 | return 0; |
483 | } | |
484 | ||
ce3ed71a CM |
485 | static int btree_csum_one_bio(struct bio *bio) |
486 | { | |
487 | struct bio_vec *bvec = bio->bi_io_vec; | |
488 | int bio_index = 0; | |
489 | struct btrfs_root *root; | |
490 | ||
491 | WARN_ON(bio->bi_vcnt <= 0); | |
492 | while(bio_index < bio->bi_vcnt) { | |
493 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; | |
494 | csum_dirty_buffer(root, bvec->bv_page); | |
495 | bio_index++; | |
496 | bvec++; | |
497 | } | |
498 | return 0; | |
499 | } | |
500 | ||
44b8bd7e | 501 | static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
f188591e | 502 | int mirror_num) |
22c59948 CM |
503 | { |
504 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
22c59948 CM |
505 | int ret; |
506 | ||
8b712842 CM |
507 | /* |
508 | * when we're called for a write, we're already in the async | |
5443be45 | 509 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 510 | */ |
22c59948 | 511 | if (rw & (1 << BIO_RW)) { |
ce3ed71a | 512 | btree_csum_one_bio(bio); |
8b712842 | 513 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
5443be45 | 514 | mirror_num, 1); |
22c59948 CM |
515 | } |
516 | ||
8b712842 CM |
517 | /* |
518 | * called for a read, do the setup so that checksum validation | |
519 | * can happen in the async kernel threads | |
520 | */ | |
22c59948 CM |
521 | ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); |
522 | BUG_ON(ret); | |
ce9adaa5 | 523 | |
8b712842 | 524 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
525 | } |
526 | ||
44b8bd7e CM |
527 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
528 | int mirror_num) | |
529 | { | |
8b712842 CM |
530 | /* |
531 | * kthread helpers are used to submit writes so that checksumming | |
532 | * can happen in parallel across all CPUs | |
533 | */ | |
44b8bd7e CM |
534 | if (!(rw & (1 << BIO_RW))) { |
535 | return __btree_submit_bio_hook(inode, rw, bio, mirror_num); | |
536 | } | |
537 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
538 | inode, rw, bio, mirror_num, | |
539 | __btree_submit_bio_hook); | |
540 | } | |
541 | ||
0da5468f CM |
542 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
543 | { | |
d1310b2e CM |
544 | struct extent_io_tree *tree; |
545 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5443be45 CM |
546 | |
547 | if (current->flags & PF_MEMALLOC) { | |
548 | redirty_page_for_writepage(wbc, page); | |
549 | unlock_page(page); | |
550 | return 0; | |
551 | } | |
5f39d397 CM |
552 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
553 | } | |
0da5468f CM |
554 | |
555 | static int btree_writepages(struct address_space *mapping, | |
556 | struct writeback_control *wbc) | |
557 | { | |
d1310b2e CM |
558 | struct extent_io_tree *tree; |
559 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 560 | if (wbc->sync_mode == WB_SYNC_NONE) { |
793955bc CM |
561 | u64 num_dirty; |
562 | u64 start = 0; | |
24ab9cd8 | 563 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
564 | |
565 | if (wbc->for_kupdate) | |
566 | return 0; | |
567 | ||
1832a6d5 CM |
568 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
569 | thresh, EXTENT_DIRTY); | |
793955bc CM |
570 | if (num_dirty < thresh) { |
571 | return 0; | |
572 | } | |
573 | } | |
0da5468f CM |
574 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
575 | } | |
576 | ||
5f39d397 CM |
577 | int btree_readpage(struct file *file, struct page *page) |
578 | { | |
d1310b2e CM |
579 | struct extent_io_tree *tree; |
580 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
581 | return extent_read_full_page(tree, page, btree_get_extent); |
582 | } | |
22b0ebda | 583 | |
70dec807 | 584 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 585 | { |
d1310b2e CM |
586 | struct extent_io_tree *tree; |
587 | struct extent_map_tree *map; | |
5f39d397 | 588 | int ret; |
d98237b3 | 589 | |
98509cfc CM |
590 | if (PageWriteback(page) || PageDirty(page)) |
591 | return 0; | |
592 | ||
d1310b2e CM |
593 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
594 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 595 | |
7b13b7b1 | 596 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
6af118ce CM |
597 | if (!ret) { |
598 | return 0; | |
599 | } | |
600 | ||
601 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
602 | if (ret == 1) { |
603 | ClearPagePrivate(page); | |
604 | set_page_private(page, 0); | |
605 | page_cache_release(page); | |
606 | } | |
6af118ce | 607 | |
d98237b3 CM |
608 | return ret; |
609 | } | |
610 | ||
5f39d397 | 611 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 612 | { |
d1310b2e CM |
613 | struct extent_io_tree *tree; |
614 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
615 | extent_invalidatepage(tree, page, offset); |
616 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 617 | if (PagePrivate(page)) { |
6af118ce CM |
618 | printk("warning page private not zero on page %Lu\n", |
619 | page_offset(page)); | |
9ad6b7bc CM |
620 | ClearPagePrivate(page); |
621 | set_page_private(page, 0); | |
622 | page_cache_release(page); | |
623 | } | |
d98237b3 CM |
624 | } |
625 | ||
5f39d397 | 626 | #if 0 |
d98237b3 | 627 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 628 | { |
87cbda5c | 629 | struct buffer_head *bh; |
0f7d52f4 | 630 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 631 | struct buffer_head *head; |
87cbda5c CM |
632 | if (!page_has_buffers(page)) { |
633 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
634 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
635 | } | |
636 | head = page_buffers(page); | |
637 | bh = head; | |
638 | do { | |
639 | if (buffer_dirty(bh)) | |
640 | csum_tree_block(root, bh, 0); | |
641 | bh = bh->b_this_page; | |
642 | } while (bh != head); | |
d98237b3 | 643 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 644 | } |
5f39d397 | 645 | #endif |
eb60ceac | 646 | |
d98237b3 CM |
647 | static struct address_space_operations btree_aops = { |
648 | .readpage = btree_readpage, | |
649 | .writepage = btree_writepage, | |
0da5468f | 650 | .writepages = btree_writepages, |
5f39d397 CM |
651 | .releasepage = btree_releasepage, |
652 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
653 | .sync_page = block_sync_page, |
654 | }; | |
655 | ||
ca7a79ad CM |
656 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
657 | u64 parent_transid) | |
090d1875 | 658 | { |
5f39d397 CM |
659 | struct extent_buffer *buf = NULL; |
660 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 661 | int ret = 0; |
090d1875 | 662 | |
db94535d | 663 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 664 | if (!buf) |
090d1875 | 665 | return 0; |
d1310b2e | 666 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 667 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 668 | free_extent_buffer(buf); |
de428b63 | 669 | return ret; |
090d1875 CM |
670 | } |
671 | ||
0999df54 CM |
672 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
673 | u64 bytenr, u32 blocksize) | |
674 | { | |
675 | struct inode *btree_inode = root->fs_info->btree_inode; | |
676 | struct extent_buffer *eb; | |
677 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
678 | bytenr, blocksize, GFP_NOFS); | |
679 | return eb; | |
680 | } | |
681 | ||
682 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
683 | u64 bytenr, u32 blocksize) | |
684 | { | |
685 | struct inode *btree_inode = root->fs_info->btree_inode; | |
686 | struct extent_buffer *eb; | |
687 | ||
688 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
689 | bytenr, blocksize, NULL, GFP_NOFS); | |
690 | return eb; | |
691 | } | |
692 | ||
693 | ||
e02119d5 CM |
694 | int btrfs_write_tree_block(struct extent_buffer *buf) |
695 | { | |
696 | return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start, | |
24ab9cd8 | 697 | buf->start + buf->len - 1, WB_SYNC_ALL); |
e02119d5 CM |
698 | } |
699 | ||
700 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
701 | { | |
702 | return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, | |
703 | buf->start, buf->start + buf->len -1); | |
704 | } | |
705 | ||
0999df54 | 706 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 707 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
708 | { |
709 | struct extent_buffer *buf = NULL; | |
710 | struct inode *btree_inode = root->fs_info->btree_inode; | |
711 | struct extent_io_tree *io_tree; | |
712 | int ret; | |
713 | ||
714 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
715 | ||
716 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
717 | if (!buf) | |
718 | return NULL; | |
0999df54 | 719 | |
ca7a79ad | 720 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
721 | |
722 | if (ret == 0) { | |
723 | buf->flags |= EXTENT_UPTODATE; | |
a1b32a59 CM |
724 | } else { |
725 | WARN_ON(1); | |
ce9adaa5 | 726 | } |
5f39d397 | 727 | return buf; |
ce9adaa5 | 728 | |
eb60ceac CM |
729 | } |
730 | ||
e089f05c | 731 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 732 | struct extent_buffer *buf) |
ed2ff2cb | 733 | { |
5f39d397 | 734 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 735 | if (btrfs_header_generation(buf) == |
925baedd CM |
736 | root->fs_info->running_transaction->transid) { |
737 | WARN_ON(!btrfs_tree_locked(buf)); | |
d1310b2e | 738 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 739 | buf); |
925baedd | 740 | } |
5f39d397 CM |
741 | return 0; |
742 | } | |
743 | ||
db94535d | 744 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 745 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 746 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 747 | u64 objectid) |
d97e63b6 | 748 | { |
cfaa7295 | 749 | root->node = NULL; |
0f7d52f4 | 750 | root->inode = NULL; |
a28ec197 | 751 | root->commit_root = NULL; |
31153d81 | 752 | root->ref_tree = NULL; |
db94535d CM |
753 | root->sectorsize = sectorsize; |
754 | root->nodesize = nodesize; | |
755 | root->leafsize = leafsize; | |
87ee04eb | 756 | root->stripesize = stripesize; |
123abc88 | 757 | root->ref_cows = 0; |
0b86a832 CM |
758 | root->track_dirty = 0; |
759 | ||
9f5fae2f | 760 | root->fs_info = fs_info; |
0f7d52f4 CM |
761 | root->objectid = objectid; |
762 | root->last_trans = 0; | |
1b05da2e CM |
763 | root->highest_inode = 0; |
764 | root->last_inode_alloc = 0; | |
58176a96 | 765 | root->name = NULL; |
4313b399 | 766 | root->in_sysfs = 0; |
0b86a832 CM |
767 | |
768 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 769 | INIT_LIST_HEAD(&root->orphan_list); |
bcc63abb | 770 | INIT_LIST_HEAD(&root->dead_list); |
925baedd | 771 | spin_lock_init(&root->node_lock); |
bcc63abb | 772 | spin_lock_init(&root->list_lock); |
a2135011 | 773 | mutex_init(&root->objectid_mutex); |
e02119d5 | 774 | mutex_init(&root->log_mutex); |
d0c803c4 CM |
775 | extent_io_tree_init(&root->dirty_log_pages, |
776 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
017e5369 CM |
777 | |
778 | btrfs_leaf_ref_tree_init(&root->ref_tree_struct); | |
779 | root->ref_tree = &root->ref_tree_struct; | |
780 | ||
3768f368 CM |
781 | memset(&root->root_key, 0, sizeof(root->root_key)); |
782 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 783 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 784 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 785 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 786 | init_completion(&root->kobj_unregister); |
6702ed49 CM |
787 | root->defrag_running = 0; |
788 | root->defrag_level = 0; | |
4d775673 | 789 | root->root_key.objectid = objectid; |
3768f368 CM |
790 | return 0; |
791 | } | |
792 | ||
db94535d | 793 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
794 | struct btrfs_fs_info *fs_info, |
795 | u64 objectid, | |
e20d96d6 | 796 | struct btrfs_root *root) |
3768f368 CM |
797 | { |
798 | int ret; | |
db94535d | 799 | u32 blocksize; |
3768f368 | 800 | |
db94535d | 801 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
802 | tree_root->sectorsize, tree_root->stripesize, |
803 | root, fs_info, objectid); | |
3768f368 CM |
804 | ret = btrfs_find_last_root(tree_root, objectid, |
805 | &root->root_item, &root->root_key); | |
806 | BUG_ON(ret); | |
807 | ||
db94535d CM |
808 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
809 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 810 | blocksize, 0); |
3768f368 | 811 | BUG_ON(!root->node); |
d97e63b6 CM |
812 | return 0; |
813 | } | |
814 | ||
e02119d5 CM |
815 | int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, |
816 | struct btrfs_fs_info *fs_info) | |
817 | { | |
818 | struct extent_buffer *eb; | |
d0c803c4 CM |
819 | struct btrfs_root *log_root_tree = fs_info->log_root_tree; |
820 | u64 start = 0; | |
821 | u64 end = 0; | |
e02119d5 CM |
822 | int ret; |
823 | ||
d0c803c4 | 824 | if (!log_root_tree) |
e02119d5 CM |
825 | return 0; |
826 | ||
d0c803c4 CM |
827 | while(1) { |
828 | ret = find_first_extent_bit(&log_root_tree->dirty_log_pages, | |
829 | 0, &start, &end, EXTENT_DIRTY); | |
830 | if (ret) | |
831 | break; | |
832 | ||
833 | clear_extent_dirty(&log_root_tree->dirty_log_pages, | |
834 | start, end, GFP_NOFS); | |
835 | } | |
e02119d5 CM |
836 | eb = fs_info->log_root_tree->node; |
837 | ||
838 | WARN_ON(btrfs_header_level(eb) != 0); | |
839 | WARN_ON(btrfs_header_nritems(eb) != 0); | |
840 | ||
d00aff00 CM |
841 | ret = btrfs_free_reserved_extent(fs_info->tree_root, |
842 | eb->start, eb->len); | |
e02119d5 CM |
843 | BUG_ON(ret); |
844 | ||
845 | free_extent_buffer(eb); | |
846 | kfree(fs_info->log_root_tree); | |
847 | fs_info->log_root_tree = NULL; | |
848 | return 0; | |
849 | } | |
850 | ||
851 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
852 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
853 | { |
854 | struct btrfs_root *root; | |
855 | struct btrfs_root *tree_root = fs_info->tree_root; | |
e02119d5 CM |
856 | |
857 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
858 | if (!root) | |
859 | return -ENOMEM; | |
860 | ||
861 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
862 | tree_root->sectorsize, tree_root->stripesize, | |
863 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
864 | ||
865 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
866 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
867 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
868 | root->ref_cows = 0; | |
869 | ||
870 | root->node = btrfs_alloc_free_block(trans, root, root->leafsize, | |
31840ae1 ZY |
871 | 0, BTRFS_TREE_LOG_OBJECTID, |
872 | trans->transid, 0, 0, 0); | |
e02119d5 CM |
873 | |
874 | btrfs_set_header_nritems(root->node, 0); | |
875 | btrfs_set_header_level(root->node, 0); | |
876 | btrfs_set_header_bytenr(root->node, root->node->start); | |
877 | btrfs_set_header_generation(root->node, trans->transid); | |
878 | btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID); | |
879 | ||
880 | write_extent_buffer(root->node, root->fs_info->fsid, | |
881 | (unsigned long)btrfs_header_fsid(root->node), | |
882 | BTRFS_FSID_SIZE); | |
883 | btrfs_mark_buffer_dirty(root->node); | |
884 | btrfs_tree_unlock(root->node); | |
885 | fs_info->log_root_tree = root; | |
886 | return 0; | |
887 | } | |
888 | ||
889 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
890 | struct btrfs_key *location) | |
891 | { | |
892 | struct btrfs_root *root; | |
893 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 894 | struct btrfs_path *path; |
5f39d397 | 895 | struct extent_buffer *l; |
1b05da2e | 896 | u64 highest_inode; |
db94535d | 897 | u32 blocksize; |
0f7d52f4 CM |
898 | int ret = 0; |
899 | ||
5eda7b5e | 900 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 901 | if (!root) |
0f7d52f4 | 902 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 903 | if (location->offset == (u64)-1) { |
db94535d | 904 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
905 | location->objectid, root); |
906 | if (ret) { | |
0f7d52f4 CM |
907 | kfree(root); |
908 | return ERR_PTR(ret); | |
909 | } | |
910 | goto insert; | |
911 | } | |
912 | ||
db94535d | 913 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
914 | tree_root->sectorsize, tree_root->stripesize, |
915 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
916 | |
917 | path = btrfs_alloc_path(); | |
918 | BUG_ON(!path); | |
919 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
920 | if (ret != 0) { | |
0f7d52f4 CM |
921 | if (ret > 0) |
922 | ret = -ENOENT; | |
923 | goto out; | |
924 | } | |
5f39d397 CM |
925 | l = path->nodes[0]; |
926 | read_extent_buffer(l, &root->root_item, | |
927 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 928 | sizeof(root->root_item)); |
44b36eb2 | 929 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
930 | ret = 0; |
931 | out: | |
932 | btrfs_release_path(root, path); | |
933 | btrfs_free_path(path); | |
934 | if (ret) { | |
935 | kfree(root); | |
936 | return ERR_PTR(ret); | |
937 | } | |
db94535d CM |
938 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
939 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 940 | blocksize, 0); |
0f7d52f4 CM |
941 | BUG_ON(!root->node); |
942 | insert: | |
e02119d5 CM |
943 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
944 | root->ref_cows = 1; | |
945 | ret = btrfs_find_highest_inode(root, &highest_inode); | |
946 | if (ret == 0) { | |
947 | root->highest_inode = highest_inode; | |
948 | root->last_inode_alloc = highest_inode; | |
949 | } | |
5eda7b5e CM |
950 | } |
951 | return root; | |
952 | } | |
953 | ||
dc17ff8f CM |
954 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
955 | u64 root_objectid) | |
956 | { | |
957 | struct btrfs_root *root; | |
958 | ||
959 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
960 | return fs_info->tree_root; | |
961 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
962 | return fs_info->extent_root; | |
963 | ||
964 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
965 | (unsigned long)root_objectid); | |
966 | return root; | |
967 | } | |
968 | ||
edbd8d4e CM |
969 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
970 | struct btrfs_key *location) | |
5eda7b5e CM |
971 | { |
972 | struct btrfs_root *root; | |
973 | int ret; | |
974 | ||
edbd8d4e CM |
975 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
976 | return fs_info->tree_root; | |
977 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
978 | return fs_info->extent_root; | |
8f18cf13 CM |
979 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
980 | return fs_info->chunk_root; | |
981 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
982 | return fs_info->dev_root; | |
edbd8d4e | 983 | |
5eda7b5e CM |
984 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
985 | (unsigned long)location->objectid); | |
986 | if (root) | |
987 | return root; | |
988 | ||
e02119d5 | 989 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
990 | if (IS_ERR(root)) |
991 | return root; | |
2619ba1f CM |
992 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
993 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
994 | root); |
995 | if (ret) { | |
5f39d397 | 996 | free_extent_buffer(root->node); |
0f7d52f4 CM |
997 | kfree(root); |
998 | return ERR_PTR(ret); | |
999 | } | |
edbd8d4e CM |
1000 | ret = btrfs_find_dead_roots(fs_info->tree_root, |
1001 | root->root_key.objectid, root); | |
1002 | BUG_ON(ret); | |
1003 | ||
1004 | return root; | |
1005 | } | |
1006 | ||
1007 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
1008 | struct btrfs_key *location, | |
1009 | const char *name, int namelen) | |
1010 | { | |
1011 | struct btrfs_root *root; | |
1012 | int ret; | |
1013 | ||
1014 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
1015 | if (!root) | |
1016 | return NULL; | |
58176a96 | 1017 | |
4313b399 CM |
1018 | if (root->in_sysfs) |
1019 | return root; | |
1020 | ||
58176a96 JB |
1021 | ret = btrfs_set_root_name(root, name, namelen); |
1022 | if (ret) { | |
5f39d397 | 1023 | free_extent_buffer(root->node); |
58176a96 JB |
1024 | kfree(root); |
1025 | return ERR_PTR(ret); | |
1026 | } | |
1027 | ||
1028 | ret = btrfs_sysfs_add_root(root); | |
1029 | if (ret) { | |
5f39d397 | 1030 | free_extent_buffer(root->node); |
58176a96 JB |
1031 | kfree(root->name); |
1032 | kfree(root); | |
1033 | return ERR_PTR(ret); | |
1034 | } | |
4313b399 | 1035 | root->in_sysfs = 1; |
0f7d52f4 CM |
1036 | return root; |
1037 | } | |
19c00ddc CM |
1038 | #if 0 |
1039 | static int add_hasher(struct btrfs_fs_info *info, char *type) { | |
1040 | struct btrfs_hasher *hasher; | |
1041 | ||
1042 | hasher = kmalloc(sizeof(*hasher), GFP_NOFS); | |
1043 | if (!hasher) | |
1044 | return -ENOMEM; | |
1045 | hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); | |
1046 | if (!hasher->hash_tfm) { | |
1047 | kfree(hasher); | |
1048 | return -EINVAL; | |
1049 | } | |
1050 | spin_lock(&info->hash_lock); | |
1051 | list_add(&hasher->list, &info->hashers); | |
1052 | spin_unlock(&info->hash_lock); | |
1053 | return 0; | |
1054 | } | |
1055 | #endif | |
04160088 CM |
1056 | |
1057 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
1058 | { | |
1059 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1060 | int ret = 0; | |
1061 | struct list_head *cur; | |
1062 | struct btrfs_device *device; | |
1063 | struct backing_dev_info *bdi; | |
1064 | ||
cb03c743 | 1065 | if ((bdi_bits & (1 << BDI_write_congested)) && |
777e6bd7 | 1066 | btrfs_congested_async(info, 0)) |
cb03c743 | 1067 | return 1; |
cb03c743 | 1068 | |
04160088 CM |
1069 | list_for_each(cur, &info->fs_devices->devices) { |
1070 | device = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1071 | if (!device->bdev) |
1072 | continue; | |
04160088 CM |
1073 | bdi = blk_get_backing_dev_info(device->bdev); |
1074 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1075 | ret = 1; | |
1076 | break; | |
1077 | } | |
1078 | } | |
1079 | return ret; | |
1080 | } | |
1081 | ||
38b66988 CM |
1082 | /* |
1083 | * this unplugs every device on the box, and it is only used when page | |
1084 | * is null | |
1085 | */ | |
1086 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
1087 | { | |
1088 | struct list_head *cur; | |
1089 | struct btrfs_device *device; | |
1090 | struct btrfs_fs_info *info; | |
1091 | ||
1092 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
1093 | list_for_each(cur, &info->fs_devices->devices) { | |
1094 | device = list_entry(cur, struct btrfs_device, dev_list); | |
1095 | bdi = blk_get_backing_dev_info(device->bdev); | |
1096 | if (bdi->unplug_io_fn) { | |
1097 | bdi->unplug_io_fn(bdi, page); | |
1098 | } | |
1099 | } | |
1100 | } | |
1101 | ||
04160088 CM |
1102 | void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
1103 | { | |
38b66988 | 1104 | struct inode *inode; |
f2d8d74d CM |
1105 | struct extent_map_tree *em_tree; |
1106 | struct extent_map *em; | |
bcbfce8a | 1107 | struct address_space *mapping; |
38b66988 CM |
1108 | u64 offset; |
1109 | ||
bcbfce8a | 1110 | /* the generic O_DIRECT read code does this */ |
38b66988 CM |
1111 | if (!page) { |
1112 | __unplug_io_fn(bdi, page); | |
1113 | return; | |
1114 | } | |
1115 | ||
bcbfce8a CM |
1116 | /* |
1117 | * page->mapping may change at any time. Get a consistent copy | |
1118 | * and use that for everything below | |
1119 | */ | |
1120 | smp_mb(); | |
1121 | mapping = page->mapping; | |
1122 | if (!mapping) | |
1123 | return; | |
1124 | ||
1125 | inode = mapping->host; | |
38b66988 | 1126 | offset = page_offset(page); |
04160088 | 1127 | |
f2d8d74d CM |
1128 | em_tree = &BTRFS_I(inode)->extent_tree; |
1129 | spin_lock(&em_tree->lock); | |
1130 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
1131 | spin_unlock(&em_tree->lock); | |
89642229 CM |
1132 | if (!em) { |
1133 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1134 | return; |
89642229 | 1135 | } |
f2d8d74d | 1136 | |
89642229 CM |
1137 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1138 | free_extent_map(em); | |
1139 | __unplug_io_fn(bdi, page); | |
1140 | return; | |
1141 | } | |
f2d8d74d CM |
1142 | offset = offset - em->start; |
1143 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1144 | em->block_start + offset, page); | |
1145 | free_extent_map(em); | |
04160088 CM |
1146 | } |
1147 | ||
1148 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
1149 | { | |
1150 | bdi_init(bdi); | |
4575c9cc | 1151 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1152 | bdi->state = 0; |
1153 | bdi->capabilities = default_backing_dev_info.capabilities; | |
1154 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
1155 | bdi->unplug_io_data = info; | |
1156 | bdi->congested_fn = btrfs_congested_fn; | |
1157 | bdi->congested_data = info; | |
1158 | return 0; | |
1159 | } | |
1160 | ||
ce9adaa5 CM |
1161 | static int bio_ready_for_csum(struct bio *bio) |
1162 | { | |
1163 | u64 length = 0; | |
1164 | u64 buf_len = 0; | |
1165 | u64 start = 0; | |
1166 | struct page *page; | |
1167 | struct extent_io_tree *io_tree = NULL; | |
1168 | struct btrfs_fs_info *info = NULL; | |
1169 | struct bio_vec *bvec; | |
1170 | int i; | |
1171 | int ret; | |
1172 | ||
1173 | bio_for_each_segment(bvec, bio, i) { | |
1174 | page = bvec->bv_page; | |
1175 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1176 | length += bvec->bv_len; | |
1177 | continue; | |
1178 | } | |
1179 | if (!page->private) { | |
1180 | length += bvec->bv_len; | |
1181 | continue; | |
1182 | } | |
1183 | length = bvec->bv_len; | |
1184 | buf_len = page->private >> 2; | |
1185 | start = page_offset(page) + bvec->bv_offset; | |
1186 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1187 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1188 | } | |
1189 | /* are we fully contained in this bio? */ | |
1190 | if (buf_len <= length) | |
1191 | return 1; | |
1192 | ||
1193 | ret = extent_range_uptodate(io_tree, start + length, | |
1194 | start + buf_len - 1); | |
1195 | if (ret == 1) | |
1196 | return ret; | |
1197 | return ret; | |
1198 | } | |
1199 | ||
8b712842 CM |
1200 | /* |
1201 | * called by the kthread helper functions to finally call the bio end_io | |
1202 | * functions. This is where read checksum verification actually happens | |
1203 | */ | |
1204 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1205 | { |
ce9adaa5 | 1206 | struct bio *bio; |
8b712842 CM |
1207 | struct end_io_wq *end_io_wq; |
1208 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1209 | int error; |
ce9adaa5 | 1210 | |
8b712842 CM |
1211 | end_io_wq = container_of(work, struct end_io_wq, work); |
1212 | bio = end_io_wq->bio; | |
1213 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1214 | |
8b712842 CM |
1215 | /* metadata bios are special because the whole tree block must |
1216 | * be checksummed at once. This makes sure the entire block is in | |
1217 | * ram and up to date before trying to verify things. For | |
1218 | * blocksize <= pagesize, it is basically a noop | |
1219 | */ | |
1220 | if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { | |
1221 | btrfs_queue_worker(&fs_info->endio_workers, | |
1222 | &end_io_wq->work); | |
1223 | return; | |
1224 | } | |
1225 | error = end_io_wq->error; | |
1226 | bio->bi_private = end_io_wq->private; | |
1227 | bio->bi_end_io = end_io_wq->end_io; | |
1228 | kfree(end_io_wq); | |
8b712842 | 1229 | bio_endio(bio, error); |
44b8bd7e CM |
1230 | } |
1231 | ||
a74a4b97 CM |
1232 | static int cleaner_kthread(void *arg) |
1233 | { | |
1234 | struct btrfs_root *root = arg; | |
1235 | ||
1236 | do { | |
1237 | smp_mb(); | |
1238 | if (root->fs_info->closing) | |
1239 | break; | |
1240 | ||
1241 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1242 | mutex_lock(&root->fs_info->cleaner_mutex); | |
a74a4b97 | 1243 | btrfs_clean_old_snapshots(root); |
a74a4b97 CM |
1244 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1245 | ||
1246 | if (freezing(current)) { | |
1247 | refrigerator(); | |
1248 | } else { | |
1249 | smp_mb(); | |
1250 | if (root->fs_info->closing) | |
1251 | break; | |
1252 | set_current_state(TASK_INTERRUPTIBLE); | |
1253 | schedule(); | |
1254 | __set_current_state(TASK_RUNNING); | |
1255 | } | |
1256 | } while (!kthread_should_stop()); | |
1257 | return 0; | |
1258 | } | |
1259 | ||
1260 | static int transaction_kthread(void *arg) | |
1261 | { | |
1262 | struct btrfs_root *root = arg; | |
1263 | struct btrfs_trans_handle *trans; | |
1264 | struct btrfs_transaction *cur; | |
1265 | unsigned long now; | |
1266 | unsigned long delay; | |
1267 | int ret; | |
1268 | ||
1269 | do { | |
1270 | smp_mb(); | |
1271 | if (root->fs_info->closing) | |
1272 | break; | |
1273 | ||
1274 | delay = HZ * 30; | |
1275 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1276 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1277 | ||
ab78c84d CM |
1278 | if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) { |
1279 | printk("btrfs: total reference cache size %Lu\n", | |
1280 | root->fs_info->total_ref_cache_size); | |
1281 | } | |
31153d81 | 1282 | |
a74a4b97 CM |
1283 | mutex_lock(&root->fs_info->trans_mutex); |
1284 | cur = root->fs_info->running_transaction; | |
1285 | if (!cur) { | |
1286 | mutex_unlock(&root->fs_info->trans_mutex); | |
1287 | goto sleep; | |
1288 | } | |
31153d81 | 1289 | |
a74a4b97 CM |
1290 | now = get_seconds(); |
1291 | if (now < cur->start_time || now - cur->start_time < 30) { | |
1292 | mutex_unlock(&root->fs_info->trans_mutex); | |
1293 | delay = HZ * 5; | |
1294 | goto sleep; | |
1295 | } | |
1296 | mutex_unlock(&root->fs_info->trans_mutex); | |
a74a4b97 CM |
1297 | trans = btrfs_start_transaction(root, 1); |
1298 | ret = btrfs_commit_transaction(trans, root); | |
1299 | sleep: | |
1300 | wake_up_process(root->fs_info->cleaner_kthread); | |
1301 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1302 | ||
1303 | if (freezing(current)) { | |
1304 | refrigerator(); | |
1305 | } else { | |
1306 | if (root->fs_info->closing) | |
1307 | break; | |
1308 | set_current_state(TASK_INTERRUPTIBLE); | |
1309 | schedule_timeout(delay); | |
1310 | __set_current_state(TASK_RUNNING); | |
1311 | } | |
1312 | } while (!kthread_should_stop()); | |
1313 | return 0; | |
1314 | } | |
1315 | ||
8a4b83cc | 1316 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1317 | struct btrfs_fs_devices *fs_devices, |
1318 | char *options) | |
2e635a27 | 1319 | { |
db94535d CM |
1320 | u32 sectorsize; |
1321 | u32 nodesize; | |
1322 | u32 leafsize; | |
1323 | u32 blocksize; | |
87ee04eb | 1324 | u32 stripesize; |
a061fc8d | 1325 | struct buffer_head *bh; |
e02119d5 | 1326 | struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1327 | GFP_NOFS); |
e02119d5 | 1328 | struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1329 | GFP_NOFS); |
8790d502 | 1330 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1331 | GFP_NOFS); |
e02119d5 | 1332 | struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1333 | GFP_NOFS); |
e02119d5 | 1334 | struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1335 | GFP_NOFS); |
e02119d5 CM |
1336 | struct btrfs_root *log_tree_root; |
1337 | ||
eb60ceac | 1338 | int ret; |
e58ca020 | 1339 | int err = -EINVAL; |
4543df7e | 1340 | |
2c90e5d6 | 1341 | struct btrfs_super_block *disk_super; |
8790d502 | 1342 | |
39279cc3 CM |
1343 | if (!extent_root || !tree_root || !fs_info) { |
1344 | err = -ENOMEM; | |
1345 | goto fail; | |
1346 | } | |
0f7d52f4 | 1347 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1348 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1349 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1350 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1351 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
19c00ddc | 1352 | spin_lock_init(&fs_info->hash_lock); |
1832a6d5 | 1353 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1354 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1355 | spin_lock_init(&fs_info->ref_cache_lock); |
19c00ddc | 1356 | |
58176a96 | 1357 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1358 | fs_info->tree_root = tree_root; |
1359 | fs_info->extent_root = extent_root; | |
0b86a832 CM |
1360 | fs_info->chunk_root = chunk_root; |
1361 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1362 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1363 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1364 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1365 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1366 | atomic_set(&fs_info->nr_async_submits, 0); |
8c8bee1d | 1367 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1368 | atomic_set(&fs_info->nr_async_bios, 0); |
a2135011 | 1369 | atomic_set(&fs_info->throttles, 0); |
ab78c84d | 1370 | atomic_set(&fs_info->throttle_gen, 0); |
e20d96d6 | 1371 | fs_info->sb = sb; |
c59f8951 | 1372 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1373 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1374 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1375 | fs_info->btree_inode = new_inode(sb); |
1376 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1377 | fs_info->btree_inode->i_nlink = 1; |
4543df7e | 1378 | fs_info->thread_pool_size = min(num_online_cpus() + 2, 8); |
0afbaf8c | 1379 | |
3eaa2885 CM |
1380 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1381 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1382 | ||
a061fc8d CM |
1383 | sb->s_blocksize = 4096; |
1384 | sb->s_blocksize_bits = blksize_bits(4096); | |
1385 | ||
0afbaf8c CM |
1386 | /* |
1387 | * we set the i_size on the btree inode to the max possible int. | |
1388 | * the real end of the address space is determined by all of | |
1389 | * the devices in the system | |
1390 | */ | |
1391 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1392 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1393 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1394 | ||
d1310b2e | 1395 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1396 | fs_info->btree_inode->i_mapping, |
1397 | GFP_NOFS); | |
d1310b2e CM |
1398 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1399 | GFP_NOFS); | |
1400 | ||
1401 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1402 | |
0f9dd46c JB |
1403 | spin_lock_init(&fs_info->block_group_cache_lock); |
1404 | fs_info->block_group_cache_tree.rb_node = NULL; | |
1405 | ||
d1310b2e | 1406 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1407 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1408 | extent_io_tree_init(&fs_info->pending_del, |
1a5bc167 | 1409 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1410 | extent_io_tree_init(&fs_info->extent_ins, |
1a5bc167 | 1411 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1412 | fs_info->do_barriers = 1; |
e18e4809 | 1413 | |
1a40e23b ZY |
1414 | extent_io_tree_init(&fs_info->reloc_mapping_tree, |
1415 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
1416 | INIT_LIST_HEAD(&fs_info->dead_reloc_roots); | |
1417 | btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree); | |
e4657689 ZY |
1418 | btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree); |
1419 | ||
0f7d52f4 CM |
1420 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1421 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1422 | sizeof(struct btrfs_key)); | |
22b0ebda | 1423 | insert_inode_hash(fs_info->btree_inode); |
39279cc3 | 1424 | |
79154b1b | 1425 | mutex_init(&fs_info->trans_mutex); |
e02119d5 | 1426 | mutex_init(&fs_info->tree_log_mutex); |
a2135011 | 1427 | mutex_init(&fs_info->drop_mutex); |
925baedd CM |
1428 | mutex_init(&fs_info->alloc_mutex); |
1429 | mutex_init(&fs_info->chunk_mutex); | |
a74a4b97 CM |
1430 | mutex_init(&fs_info->transaction_kthread_mutex); |
1431 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1432 | mutex_init(&fs_info->volume_mutex); |
1a40e23b | 1433 | mutex_init(&fs_info->tree_reloc_mutex); |
e6dcd2dc | 1434 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1435 | init_waitqueue_head(&fs_info->transaction_wait); |
4854ddd0 | 1436 | init_waitqueue_head(&fs_info->async_submit_wait); |
e02119d5 CM |
1437 | init_waitqueue_head(&fs_info->tree_log_wait); |
1438 | atomic_set(&fs_info->tree_log_commit, 0); | |
1439 | atomic_set(&fs_info->tree_log_writers, 0); | |
1440 | fs_info->tree_log_transid = 0; | |
3768f368 | 1441 | |
19c00ddc CM |
1442 | #if 0 |
1443 | ret = add_hasher(fs_info, "crc32c"); | |
1444 | if (ret) { | |
1445 | printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); | |
1446 | err = -ENOMEM; | |
1447 | goto fail_iput; | |
1448 | } | |
1449 | #endif | |
0b86a832 | 1450 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1451 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1452 | |
d98237b3 | 1453 | |
a061fc8d CM |
1454 | bh = __bread(fs_devices->latest_bdev, |
1455 | BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
1456 | if (!bh) | |
39279cc3 | 1457 | goto fail_iput; |
39279cc3 | 1458 | |
a061fc8d CM |
1459 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
1460 | brelse(bh); | |
5f39d397 | 1461 | |
a061fc8d | 1462 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1463 | |
5f39d397 | 1464 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1465 | if (!btrfs_super_root(disk_super)) |
39279cc3 | 1466 | goto fail_sb_buffer; |
0f7d52f4 | 1467 | |
edf24abe CH |
1468 | err = btrfs_parse_options(tree_root, options); |
1469 | if (err) | |
1470 | goto fail_sb_buffer; | |
dfe25020 | 1471 | |
4543df7e CM |
1472 | /* |
1473 | * we need to start all the end_io workers up front because the | |
1474 | * queue work function gets called at interrupt time, and so it | |
1475 | * cannot dynamically grow. | |
1476 | */ | |
5443be45 CM |
1477 | btrfs_init_workers(&fs_info->workers, "worker", |
1478 | fs_info->thread_pool_size); | |
1479 | btrfs_init_workers(&fs_info->submit_workers, "submit", | |
b720d209 CM |
1480 | min_t(u64, fs_devices->num_devices, |
1481 | fs_info->thread_pool_size)); | |
61b49440 CM |
1482 | |
1483 | /* a higher idle thresh on the submit workers makes it much more | |
1484 | * likely that bios will be send down in a sane order to the | |
1485 | * devices | |
1486 | */ | |
1487 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 CM |
1488 | |
1489 | /* fs_info->workers is responsible for checksumming file data | |
1490 | * blocks and metadata. Using a larger idle thresh allows each | |
1491 | * worker thread to operate on things in roughly the order they | |
1492 | * were sent by the writeback daemons, improving overall locality | |
1493 | * of the IO going down the pipe. | |
1494 | */ | |
1495 | fs_info->workers.idle_thresh = 128; | |
61b49440 | 1496 | |
5443be45 CM |
1497 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); |
1498 | btrfs_init_workers(&fs_info->endio_workers, "endio", | |
1499 | fs_info->thread_pool_size); | |
1500 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", | |
e6dcd2dc | 1501 | fs_info->thread_pool_size); |
61b49440 CM |
1502 | |
1503 | /* | |
1504 | * endios are largely parallel and should have a very | |
1505 | * low idle thresh | |
1506 | */ | |
1507 | fs_info->endio_workers.idle_thresh = 4; | |
3a5f1d45 | 1508 | fs_info->endio_write_workers.idle_thresh = 64; |
61b49440 | 1509 | |
4543df7e | 1510 | btrfs_start_workers(&fs_info->workers, 1); |
1cc127b5 | 1511 | btrfs_start_workers(&fs_info->submit_workers, 1); |
247e743c | 1512 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
4543df7e | 1513 | btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
e6dcd2dc CM |
1514 | btrfs_start_workers(&fs_info->endio_write_workers, |
1515 | fs_info->thread_pool_size); | |
4543df7e | 1516 | |
edf24abe | 1517 | err = -EINVAL; |
a0af469b | 1518 | if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) { |
8a4b83cc CM |
1519 | printk("Btrfs: wanted %llu devices, but found %llu\n", |
1520 | (unsigned long long)btrfs_super_num_devices(disk_super), | |
a0af469b | 1521 | (unsigned long long)fs_devices->open_devices); |
dfe25020 CM |
1522 | if (btrfs_test_opt(tree_root, DEGRADED)) |
1523 | printk("continuing in degraded mode\n"); | |
1524 | else { | |
1525 | goto fail_sb_buffer; | |
1526 | } | |
8a4b83cc | 1527 | } |
dfe25020 | 1528 | |
4575c9cc CM |
1529 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
1530 | ||
db94535d CM |
1531 | nodesize = btrfs_super_nodesize(disk_super); |
1532 | leafsize = btrfs_super_leafsize(disk_super); | |
1533 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1534 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1535 | tree_root->nodesize = nodesize; |
1536 | tree_root->leafsize = leafsize; | |
1537 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1538 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1539 | |
1540 | sb->s_blocksize = sectorsize; | |
1541 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1542 | |
39279cc3 CM |
1543 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1544 | sizeof(disk_super->magic))) { | |
1545 | printk("btrfs: valid FS not found on %s\n", sb->s_id); | |
1546 | goto fail_sb_buffer; | |
1547 | } | |
19c00ddc | 1548 | |
925baedd | 1549 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1550 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1551 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f CM |
1552 | if (ret) { |
1553 | printk("btrfs: failed to read the system array on %s\n", | |
1554 | sb->s_id); | |
1555 | goto fail_sys_array; | |
1556 | } | |
0b86a832 CM |
1557 | |
1558 | blocksize = btrfs_level_size(tree_root, | |
1559 | btrfs_super_chunk_root_level(disk_super)); | |
1560 | ||
1561 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1562 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1563 | ||
1564 | chunk_root->node = read_tree_block(chunk_root, | |
1565 | btrfs_super_chunk_root(disk_super), | |
ca7a79ad | 1566 | blocksize, 0); |
0b86a832 CM |
1567 | BUG_ON(!chunk_root->node); |
1568 | ||
e17cade2 CM |
1569 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
1570 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), | |
1571 | BTRFS_UUID_SIZE); | |
1572 | ||
925baedd | 1573 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1574 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1575 | mutex_unlock(&fs_info->chunk_mutex); |
0b86a832 CM |
1576 | BUG_ON(ret); |
1577 | ||
dfe25020 CM |
1578 | btrfs_close_extra_devices(fs_devices); |
1579 | ||
db94535d CM |
1580 | blocksize = btrfs_level_size(tree_root, |
1581 | btrfs_super_root_level(disk_super)); | |
19c00ddc | 1582 | |
0b86a832 | 1583 | |
e20d96d6 | 1584 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1585 | btrfs_super_root(disk_super), |
ca7a79ad | 1586 | blocksize, 0); |
39279cc3 CM |
1587 | if (!tree_root->node) |
1588 | goto fail_sb_buffer; | |
3768f368 | 1589 | |
db94535d CM |
1590 | |
1591 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1592 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1593 | if (ret) |
39279cc3 | 1594 | goto fail_tree_root; |
0b86a832 CM |
1595 | extent_root->track_dirty = 1; |
1596 | ||
1597 | ret = find_and_setup_root(tree_root, fs_info, | |
1598 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1599 | dev_root->track_dirty = 1; | |
1600 | ||
1601 | if (ret) | |
1602 | goto fail_extent_root; | |
3768f368 | 1603 | |
9078a3e1 CM |
1604 | btrfs_read_block_groups(extent_root); |
1605 | ||
0f7d52f4 | 1606 | fs_info->generation = btrfs_super_generation(disk_super) + 1; |
d18a2c44 CM |
1607 | fs_info->data_alloc_profile = (u64)-1; |
1608 | fs_info->metadata_alloc_profile = (u64)-1; | |
1609 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
a74a4b97 CM |
1610 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1611 | "btrfs-cleaner"); | |
1612 | if (!fs_info->cleaner_kthread) | |
1613 | goto fail_extent_root; | |
1614 | ||
1615 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1616 | tree_root, | |
1617 | "btrfs-transaction"); | |
1618 | if (!fs_info->transaction_kthread) | |
3f157a2f | 1619 | goto fail_cleaner; |
a74a4b97 | 1620 | |
e02119d5 CM |
1621 | if (btrfs_super_log_root(disk_super) != 0) { |
1622 | u32 blocksize; | |
1623 | u64 bytenr = btrfs_super_log_root(disk_super); | |
1624 | ||
1625 | blocksize = | |
1626 | btrfs_level_size(tree_root, | |
1627 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 1628 | |
e02119d5 CM |
1629 | log_tree_root = kzalloc(sizeof(struct btrfs_root), |
1630 | GFP_NOFS); | |
1631 | ||
1632 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1633 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1634 | ||
1635 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
1636 | blocksize, 0); | |
1637 | ret = btrfs_recover_log_trees(log_tree_root); | |
1638 | BUG_ON(ret); | |
1639 | } | |
1a40e23b ZY |
1640 | |
1641 | ret = btrfs_cleanup_reloc_trees(tree_root); | |
1642 | BUG_ON(ret); | |
1643 | ||
e02119d5 | 1644 | fs_info->last_trans_committed = btrfs_super_generation(disk_super); |
0f7d52f4 | 1645 | return tree_root; |
39279cc3 | 1646 | |
3f157a2f | 1647 | fail_cleaner: |
a74a4b97 | 1648 | kthread_stop(fs_info->cleaner_kthread); |
0b86a832 CM |
1649 | fail_extent_root: |
1650 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1651 | fail_tree_root: |
5f39d397 | 1652 | free_extent_buffer(tree_root->node); |
84eed90f | 1653 | fail_sys_array: |
39279cc3 | 1654 | fail_sb_buffer: |
247e743c | 1655 | btrfs_stop_workers(&fs_info->fixup_workers); |
8b712842 CM |
1656 | btrfs_stop_workers(&fs_info->workers); |
1657 | btrfs_stop_workers(&fs_info->endio_workers); | |
e6dcd2dc | 1658 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1659 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e CM |
1660 | fail_iput: |
1661 | iput(fs_info->btree_inode); | |
39279cc3 | 1662 | fail: |
dfe25020 | 1663 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f CM |
1664 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
1665 | ||
39279cc3 CM |
1666 | kfree(extent_root); |
1667 | kfree(tree_root); | |
2d2ae547 | 1668 | bdi_destroy(&fs_info->bdi); |
39279cc3 CM |
1669 | kfree(fs_info); |
1670 | return ERR_PTR(err); | |
eb60ceac CM |
1671 | } |
1672 | ||
f2984462 CM |
1673 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1674 | { | |
1675 | char b[BDEVNAME_SIZE]; | |
1676 | ||
1677 | if (uptodate) { | |
1678 | set_buffer_uptodate(bh); | |
1679 | } else { | |
1680 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1681 | printk(KERN_WARNING "lost page write due to " | |
1682 | "I/O error on %s\n", | |
1683 | bdevname(bh->b_bdev, b)); | |
1684 | } | |
1259ab75 CM |
1685 | /* note, we dont' set_buffer_write_io_error because we have |
1686 | * our own ways of dealing with the IO errors | |
1687 | */ | |
f2984462 CM |
1688 | clear_buffer_uptodate(bh); |
1689 | } | |
1690 | unlock_buffer(bh); | |
1691 | put_bh(bh); | |
1692 | } | |
1693 | ||
1694 | int write_all_supers(struct btrfs_root *root) | |
1695 | { | |
1696 | struct list_head *cur; | |
1697 | struct list_head *head = &root->fs_info->fs_devices->devices; | |
1698 | struct btrfs_device *dev; | |
a061fc8d | 1699 | struct btrfs_super_block *sb; |
f2984462 CM |
1700 | struct btrfs_dev_item *dev_item; |
1701 | struct buffer_head *bh; | |
1702 | int ret; | |
1703 | int do_barriers; | |
a236aed1 CM |
1704 | int max_errors; |
1705 | int total_errors = 0; | |
a061fc8d CM |
1706 | u32 crc; |
1707 | u64 flags; | |
f2984462 | 1708 | |
a236aed1 | 1709 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
1710 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
1711 | ||
a061fc8d CM |
1712 | sb = &root->fs_info->super_for_commit; |
1713 | dev_item = &sb->dev_item; | |
f2984462 CM |
1714 | list_for_each(cur, head) { |
1715 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1716 | if (!dev->bdev) { |
1717 | total_errors++; | |
1718 | continue; | |
1719 | } | |
1720 | if (!dev->in_fs_metadata) | |
1721 | continue; | |
1722 | ||
a061fc8d CM |
1723 | btrfs_set_stack_device_type(dev_item, dev->type); |
1724 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
1725 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
1726 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
1727 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
1728 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
1729 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
1730 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
1731 | flags = btrfs_super_flags(sb); | |
1732 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
1733 | ||
1734 | ||
1735 | crc = ~(u32)0; | |
1736 | crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc, | |
1737 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
1738 | btrfs_csum_final(crc, sb->csum); | |
1739 | ||
1740 | bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096, | |
f2984462 CM |
1741 | BTRFS_SUPER_INFO_SIZE); |
1742 | ||
a061fc8d | 1743 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
f2984462 CM |
1744 | dev->pending_io = bh; |
1745 | ||
1746 | get_bh(bh); | |
1747 | set_buffer_uptodate(bh); | |
1748 | lock_buffer(bh); | |
1749 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
1750 | ||
1751 | if (do_barriers && dev->barriers) { | |
1752 | ret = submit_bh(WRITE_BARRIER, bh); | |
1753 | if (ret == -EOPNOTSUPP) { | |
1754 | printk("btrfs: disabling barriers on dev %s\n", | |
1755 | dev->name); | |
1756 | set_buffer_uptodate(bh); | |
1757 | dev->barriers = 0; | |
1758 | get_bh(bh); | |
1759 | lock_buffer(bh); | |
1760 | ret = submit_bh(WRITE, bh); | |
1761 | } | |
1762 | } else { | |
1763 | ret = submit_bh(WRITE, bh); | |
1764 | } | |
a236aed1 CM |
1765 | if (ret) |
1766 | total_errors++; | |
f2984462 | 1767 | } |
a236aed1 CM |
1768 | if (total_errors > max_errors) { |
1769 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1770 | BUG(); | |
1771 | } | |
1772 | total_errors = 0; | |
f2984462 CM |
1773 | |
1774 | list_for_each(cur, head) { | |
1775 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1776 | if (!dev->bdev) |
1777 | continue; | |
1778 | if (!dev->in_fs_metadata) | |
1779 | continue; | |
1780 | ||
f2984462 CM |
1781 | BUG_ON(!dev->pending_io); |
1782 | bh = dev->pending_io; | |
1783 | wait_on_buffer(bh); | |
1784 | if (!buffer_uptodate(dev->pending_io)) { | |
1785 | if (do_barriers && dev->barriers) { | |
1786 | printk("btrfs: disabling barriers on dev %s\n", | |
1787 | dev->name); | |
1788 | set_buffer_uptodate(bh); | |
1789 | get_bh(bh); | |
1790 | lock_buffer(bh); | |
1791 | dev->barriers = 0; | |
1792 | ret = submit_bh(WRITE, bh); | |
1793 | BUG_ON(ret); | |
1794 | wait_on_buffer(bh); | |
1259ab75 CM |
1795 | if (!buffer_uptodate(bh)) |
1796 | total_errors++; | |
f2984462 | 1797 | } else { |
a236aed1 | 1798 | total_errors++; |
f2984462 CM |
1799 | } |
1800 | ||
1801 | } | |
1802 | dev->pending_io = NULL; | |
1803 | brelse(bh); | |
1804 | } | |
a236aed1 CM |
1805 | if (total_errors > max_errors) { |
1806 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1807 | BUG(); | |
1808 | } | |
f2984462 CM |
1809 | return 0; |
1810 | } | |
1811 | ||
e089f05c | 1812 | int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root |
79154b1b | 1813 | *root) |
eb60ceac | 1814 | { |
e66f709b | 1815 | int ret; |
5f39d397 | 1816 | |
f2984462 | 1817 | ret = write_all_supers(root); |
5f39d397 | 1818 | return ret; |
cfaa7295 CM |
1819 | } |
1820 | ||
5eda7b5e | 1821 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
1822 | { |
1823 | radix_tree_delete(&fs_info->fs_roots_radix, | |
1824 | (unsigned long)root->root_key.objectid); | |
b99aa6cb CM |
1825 | if (root->in_sysfs) |
1826 | btrfs_sysfs_del_root(root); | |
2619ba1f CM |
1827 | if (root->inode) |
1828 | iput(root->inode); | |
1829 | if (root->node) | |
5f39d397 | 1830 | free_extent_buffer(root->node); |
2619ba1f | 1831 | if (root->commit_root) |
5f39d397 | 1832 | free_extent_buffer(root->commit_root); |
58176a96 JB |
1833 | if (root->name) |
1834 | kfree(root->name); | |
2619ba1f CM |
1835 | kfree(root); |
1836 | return 0; | |
1837 | } | |
1838 | ||
35b7e476 | 1839 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
1840 | { |
1841 | int ret; | |
1842 | struct btrfs_root *gang[8]; | |
1843 | int i; | |
1844 | ||
1845 | while(1) { | |
1846 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1847 | (void **)gang, 0, | |
1848 | ARRAY_SIZE(gang)); | |
1849 | if (!ret) | |
1850 | break; | |
2619ba1f | 1851 | for (i = 0; i < ret; i++) |
5eda7b5e | 1852 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
1853 | } |
1854 | return 0; | |
1855 | } | |
b4100d64 | 1856 | |
e20d96d6 | 1857 | int close_ctree(struct btrfs_root *root) |
cfaa7295 | 1858 | { |
3768f368 | 1859 | int ret; |
e089f05c | 1860 | struct btrfs_trans_handle *trans; |
0f7d52f4 | 1861 | struct btrfs_fs_info *fs_info = root->fs_info; |
e089f05c | 1862 | |
facda1e7 | 1863 | fs_info->closing = 1; |
a2135011 CM |
1864 | smp_mb(); |
1865 | ||
a74a4b97 CM |
1866 | kthread_stop(root->fs_info->transaction_kthread); |
1867 | kthread_stop(root->fs_info->cleaner_kthread); | |
1868 | ||
a74a4b97 | 1869 | btrfs_clean_old_snapshots(root); |
79154b1b | 1870 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 1871 | ret = btrfs_commit_transaction(trans, root); |
79154b1b CM |
1872 | /* run commit again to drop the original snapshot */ |
1873 | trans = btrfs_start_transaction(root, 1); | |
1874 | btrfs_commit_transaction(trans, root); | |
1875 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 1876 | BUG_ON(ret); |
d6bfde87 | 1877 | |
79154b1b | 1878 | write_ctree_super(NULL, root); |
0f7d52f4 | 1879 | |
b0c68f8b CM |
1880 | if (fs_info->delalloc_bytes) { |
1881 | printk("btrfs: at unmount delalloc count %Lu\n", | |
1882 | fs_info->delalloc_bytes); | |
1883 | } | |
31153d81 YZ |
1884 | if (fs_info->total_ref_cache_size) { |
1885 | printk("btrfs: at umount reference cache size %Lu\n", | |
1886 | fs_info->total_ref_cache_size); | |
1887 | } | |
bcc63abb | 1888 | |
0f7d52f4 | 1889 | if (fs_info->extent_root->node) |
5f39d397 | 1890 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 1891 | |
0f7d52f4 | 1892 | if (fs_info->tree_root->node) |
5f39d397 | 1893 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 1894 | |
0b86a832 CM |
1895 | if (root->fs_info->chunk_root->node); |
1896 | free_extent_buffer(root->fs_info->chunk_root->node); | |
1897 | ||
1898 | if (root->fs_info->dev_root->node); | |
1899 | free_extent_buffer(root->fs_info->dev_root->node); | |
1900 | ||
9078a3e1 | 1901 | btrfs_free_block_groups(root->fs_info); |
4ca8b41e | 1902 | fs_info->closing = 2; |
0f7d52f4 | 1903 | del_fs_roots(fs_info); |
d10c5f31 CM |
1904 | |
1905 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1906 | ||
db94535d | 1907 | truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); |
9ad6b7bc | 1908 | |
247e743c | 1909 | btrfs_stop_workers(&fs_info->fixup_workers); |
8b712842 CM |
1910 | btrfs_stop_workers(&fs_info->workers); |
1911 | btrfs_stop_workers(&fs_info->endio_workers); | |
e6dcd2dc | 1912 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1913 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 1914 | |
db94535d | 1915 | iput(fs_info->btree_inode); |
19c00ddc CM |
1916 | #if 0 |
1917 | while(!list_empty(&fs_info->hashers)) { | |
1918 | struct btrfs_hasher *hasher; | |
1919 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
1920 | hashers); | |
1921 | list_del(&hasher->hashers); | |
1922 | crypto_free_hash(&fs_info->hash_tfm); | |
1923 | kfree(hasher); | |
1924 | } | |
1925 | #endif | |
dfe25020 | 1926 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 1927 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 1928 | |
04160088 | 1929 | bdi_destroy(&fs_info->bdi); |
0b86a832 | 1930 | |
0f7d52f4 | 1931 | kfree(fs_info->extent_root); |
0f7d52f4 | 1932 | kfree(fs_info->tree_root); |
0b86a832 CM |
1933 | kfree(fs_info->chunk_root); |
1934 | kfree(fs_info->dev_root); | |
eb60ceac CM |
1935 | return 0; |
1936 | } | |
1937 | ||
1259ab75 | 1938 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 1939 | { |
1259ab75 | 1940 | int ret; |
810191ff | 1941 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
1942 | |
1943 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
1944 | if (!ret) | |
1945 | return ret; | |
1946 | ||
1947 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
1948 | parent_transid); | |
1949 | return !ret; | |
5f39d397 CM |
1950 | } |
1951 | ||
1952 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 1953 | { |
810191ff | 1954 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 1955 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
1956 | buf); |
1957 | } | |
6702ed49 | 1958 | |
5f39d397 CM |
1959 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
1960 | { | |
810191ff | 1961 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
1962 | u64 transid = btrfs_header_generation(buf); |
1963 | struct inode *btree_inode = root->fs_info->btree_inode; | |
6702ed49 | 1964 | |
925baedd | 1965 | WARN_ON(!btrfs_tree_locked(buf)); |
ccd467d6 CM |
1966 | if (transid != root->fs_info->generation) { |
1967 | printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", | |
db94535d | 1968 | (unsigned long long)buf->start, |
ccd467d6 CM |
1969 | transid, root->fs_info->generation); |
1970 | WARN_ON(1); | |
1971 | } | |
d1310b2e | 1972 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); |
eb60ceac CM |
1973 | } |
1974 | ||
d3c2fdcf | 1975 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 1976 | { |
188de649 CM |
1977 | /* |
1978 | * looks as though older kernels can get into trouble with | |
1979 | * this code, they end up stuck in balance_dirty_pages forever | |
1980 | */ | |
d6bfde87 CM |
1981 | struct extent_io_tree *tree; |
1982 | u64 num_dirty; | |
1983 | u64 start = 0; | |
b64a2851 | 1984 | unsigned long thresh = 96 * 1024 * 1024; |
d6bfde87 CM |
1985 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
1986 | ||
b64a2851 | 1987 | if (current_is_pdflush() || current->flags & PF_MEMALLOC) |
d6bfde87 CM |
1988 | return; |
1989 | ||
1990 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
1991 | thresh, EXTENT_DIRTY); | |
1992 | if (num_dirty > thresh) { | |
1993 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 1994 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 1995 | } |
188de649 | 1996 | return; |
35b7e476 | 1997 | } |
6b80053d | 1998 | |
ca7a79ad | 1999 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 2000 | { |
810191ff | 2001 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 2002 | int ret; |
ca7a79ad | 2003 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
2004 | if (ret == 0) { |
2005 | buf->flags |= EXTENT_UPTODATE; | |
2006 | } | |
2007 | return ret; | |
6b80053d | 2008 | } |
0da5468f | 2009 | |
4bef0848 CM |
2010 | int btree_lock_page_hook(struct page *page) |
2011 | { | |
2012 | struct inode *inode = page->mapping->host; | |
2013 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2014 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
2015 | struct extent_buffer *eb; | |
2016 | unsigned long len; | |
2017 | u64 bytenr = page_offset(page); | |
2018 | ||
2019 | if (page->private == EXTENT_PAGE_PRIVATE) | |
2020 | goto out; | |
2021 | ||
2022 | len = page->private >> 2; | |
2023 | eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS); | |
2024 | if (!eb) | |
2025 | goto out; | |
2026 | ||
2027 | btrfs_tree_lock(eb); | |
2028 | spin_lock(&root->fs_info->hash_lock); | |
2029 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); | |
2030 | spin_unlock(&root->fs_info->hash_lock); | |
2031 | btrfs_tree_unlock(eb); | |
2032 | free_extent_buffer(eb); | |
2033 | out: | |
2034 | lock_page(page); | |
2035 | return 0; | |
2036 | } | |
2037 | ||
d1310b2e | 2038 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 2039 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 2040 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 2041 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
2042 | /* note we're sharing with inode.c for the merge bio hook */ |
2043 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 2044 | }; |