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 | ||
e20d96d6 | 19 | #include <linux/fs.h> |
d98237b3 | 20 | #include <linux/blkdev.h> |
87cbda5c | 21 | #include <linux/scatterlist.h> |
22b0ebda | 22 | #include <linux/swap.h> |
0f7d52f4 | 23 | #include <linux/radix-tree.h> |
35b7e476 | 24 | #include <linux/writeback.h> |
5f39d397 | 25 | #include <linux/buffer_head.h> // for block_sync_page |
ce9adaa5 | 26 | #include <linux/workqueue.h> |
a5eb62e3 | 27 | #include "crc32c.h" |
eb60ceac CM |
28 | #include "ctree.h" |
29 | #include "disk-io.h" | |
e089f05c | 30 | #include "transaction.h" |
0f7d52f4 | 31 | #include "btrfs_inode.h" |
0b86a832 | 32 | #include "volumes.h" |
db94535d | 33 | #include "print-tree.h" |
eb60ceac | 34 | |
5f39d397 CM |
35 | #if 0 |
36 | static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) | |
7eccb903 | 37 | { |
5f39d397 CM |
38 | if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) { |
39 | printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n", | |
40 | (unsigned long long)extent_buffer_blocknr(buf), | |
41 | (unsigned long long)btrfs_header_blocknr(buf)); | |
39279cc3 | 42 | return 1; |
d98237b3 | 43 | } |
9a8dd150 | 44 | return 0; |
eb60ceac | 45 | } |
5f39d397 | 46 | #endif |
eb60ceac | 47 | |
d1310b2e | 48 | static struct extent_io_ops btree_extent_io_ops; |
ce9adaa5 | 49 | static struct workqueue_struct *end_io_workqueue; |
44b8bd7e | 50 | static struct workqueue_struct *async_submit_workqueue; |
ce9adaa5 CM |
51 | |
52 | struct end_io_wq { | |
53 | struct bio *bio; | |
54 | bio_end_io_t *end_io; | |
55 | void *private; | |
56 | struct btrfs_fs_info *info; | |
57 | int error; | |
22c59948 | 58 | int metadata; |
ce9adaa5 CM |
59 | struct list_head list; |
60 | }; | |
0da5468f | 61 | |
44b8bd7e CM |
62 | struct async_submit_bio { |
63 | struct inode *inode; | |
64 | struct bio *bio; | |
65 | struct list_head list; | |
66 | extent_submit_bio_hook_t *submit_bio_hook; | |
67 | int rw; | |
68 | int mirror_num; | |
69 | }; | |
70 | ||
5f39d397 | 71 | struct extent_map *btree_get_extent(struct inode *inode, struct page *page, |
d1310b2e | 72 | size_t page_offset, u64 start, u64 len, |
5f39d397 | 73 | int create) |
7eccb903 | 74 | { |
5f39d397 CM |
75 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
76 | struct extent_map *em; | |
77 | int ret; | |
78 | ||
d1310b2e CM |
79 | spin_lock(&em_tree->lock); |
80 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
81 | if (em) { |
82 | em->bdev = | |
83 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
84 | spin_unlock(&em_tree->lock); | |
5f39d397 | 85 | goto out; |
a061fc8d CM |
86 | } |
87 | spin_unlock(&em_tree->lock); | |
7b13b7b1 | 88 | |
5f39d397 CM |
89 | em = alloc_extent_map(GFP_NOFS); |
90 | if (!em) { | |
91 | em = ERR_PTR(-ENOMEM); | |
92 | goto out; | |
93 | } | |
94 | em->start = 0; | |
0afbaf8c | 95 | em->len = (u64)-1; |
5f39d397 | 96 | em->block_start = 0; |
a061fc8d | 97 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
98 | |
99 | spin_lock(&em_tree->lock); | |
5f39d397 CM |
100 | ret = add_extent_mapping(em_tree, em); |
101 | if (ret == -EEXIST) { | |
0afbaf8c CM |
102 | u64 failed_start = em->start; |
103 | u64 failed_len = em->len; | |
104 | ||
105 | printk("failed to insert %Lu %Lu -> %Lu into tree\n", | |
106 | em->start, em->len, em->block_start); | |
5f39d397 | 107 | free_extent_map(em); |
7b13b7b1 | 108 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c CM |
109 | if (em) { |
110 | printk("after failing, found %Lu %Lu %Lu\n", | |
111 | em->start, em->len, em->block_start); | |
7b13b7b1 | 112 | ret = 0; |
0afbaf8c CM |
113 | } else { |
114 | em = lookup_extent_mapping(em_tree, failed_start, | |
115 | failed_len); | |
116 | if (em) { | |
117 | printk("double failure lookup gives us " | |
118 | "%Lu %Lu -> %Lu\n", em->start, | |
119 | em->len, em->block_start); | |
120 | free_extent_map(em); | |
121 | } | |
7b13b7b1 | 122 | ret = -EIO; |
0afbaf8c | 123 | } |
5f39d397 | 124 | } else if (ret) { |
7b13b7b1 CM |
125 | free_extent_map(em); |
126 | em = NULL; | |
5f39d397 | 127 | } |
7b13b7b1 CM |
128 | spin_unlock(&em_tree->lock); |
129 | ||
130 | if (ret) | |
131 | em = ERR_PTR(ret); | |
5f39d397 CM |
132 | out: |
133 | return em; | |
7eccb903 CM |
134 | } |
135 | ||
19c00ddc CM |
136 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
137 | { | |
a5eb62e3 | 138 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
139 | } |
140 | ||
141 | void btrfs_csum_final(u32 crc, char *result) | |
142 | { | |
143 | *(__le32 *)result = ~cpu_to_le32(crc); | |
144 | } | |
145 | ||
146 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, | |
147 | int verify) | |
148 | { | |
149 | char result[BTRFS_CRC32_SIZE]; | |
150 | unsigned long len; | |
151 | unsigned long cur_len; | |
152 | unsigned long offset = BTRFS_CSUM_SIZE; | |
153 | char *map_token = NULL; | |
154 | char *kaddr; | |
155 | unsigned long map_start; | |
156 | unsigned long map_len; | |
157 | int err; | |
158 | u32 crc = ~(u32)0; | |
159 | ||
160 | len = buf->len - offset; | |
161 | while(len > 0) { | |
162 | err = map_private_extent_buffer(buf, offset, 32, | |
163 | &map_token, &kaddr, | |
164 | &map_start, &map_len, KM_USER0); | |
165 | if (err) { | |
166 | printk("failed to map extent buffer! %lu\n", | |
167 | offset); | |
168 | return 1; | |
169 | } | |
170 | cur_len = min(len, map_len - (offset - map_start)); | |
171 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
172 | crc, cur_len); | |
173 | len -= cur_len; | |
174 | offset += cur_len; | |
175 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
176 | } | |
177 | btrfs_csum_final(crc, result); | |
178 | ||
179 | if (verify) { | |
e4204ded CM |
180 | int from_this_trans = 0; |
181 | ||
182 | if (root->fs_info->running_transaction && | |
183 | btrfs_header_generation(buf) == | |
184 | root->fs_info->running_transaction->transid) | |
185 | from_this_trans = 1; | |
186 | ||
187 | /* FIXME, this is not good */ | |
63b10fc4 | 188 | if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) { |
e4204ded CM |
189 | u32 val; |
190 | u32 found = 0; | |
191 | memcpy(&found, result, BTRFS_CRC32_SIZE); | |
192 | ||
193 | read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE); | |
194 | printk("btrfs: %s checksum verify failed on %llu " | |
63b10fc4 CM |
195 | "wanted %X found %X from_this_trans %d " |
196 | "level %d\n", | |
19c00ddc | 197 | root->fs_info->sb->s_id, |
63b10fc4 CM |
198 | buf->start, val, found, from_this_trans, |
199 | btrfs_header_level(buf)); | |
19c00ddc CM |
200 | return 1; |
201 | } | |
202 | } else { | |
203 | write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE); | |
204 | } | |
205 | return 0; | |
206 | } | |
207 | ||
1259ab75 CM |
208 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
209 | struct extent_buffer *eb, u64 parent_transid) | |
210 | { | |
211 | int ret; | |
212 | ||
213 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
214 | return 0; | |
215 | ||
216 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
217 | if (extent_buffer_uptodate(io_tree, eb) && | |
218 | btrfs_header_generation(eb) == parent_transid) { | |
219 | ret = 0; | |
220 | goto out; | |
221 | } | |
222 | printk("parent transid verify failed on %llu wanted %llu found %llu\n", | |
223 | (unsigned long long)eb->start, | |
224 | (unsigned long long)parent_transid, | |
225 | (unsigned long long)btrfs_header_generation(eb)); | |
226 | ret = 1; | |
227 | out: | |
228 | clear_extent_buffer_uptodate(io_tree, eb); | |
229 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, | |
230 | GFP_NOFS); | |
231 | return ret; | |
232 | ||
233 | } | |
234 | ||
f188591e CM |
235 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
236 | struct extent_buffer *eb, | |
ca7a79ad | 237 | u64 start, u64 parent_transid) |
f188591e CM |
238 | { |
239 | struct extent_io_tree *io_tree; | |
240 | int ret; | |
241 | int num_copies = 0; | |
242 | int mirror_num = 0; | |
243 | ||
244 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
245 | while (1) { | |
246 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
247 | btree_get_extent, mirror_num); | |
1259ab75 CM |
248 | if (!ret && |
249 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 250 | return ret; |
4235298e | 251 | |
f188591e CM |
252 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
253 | eb->start, eb->len); | |
4235298e | 254 | if (num_copies == 1) |
f188591e | 255 | return ret; |
4235298e | 256 | |
f188591e | 257 | mirror_num++; |
4235298e | 258 | if (mirror_num > num_copies) |
f188591e | 259 | return ret; |
f188591e | 260 | } |
f188591e CM |
261 | return -EIO; |
262 | } | |
19c00ddc CM |
263 | |
264 | int csum_dirty_buffer(struct btrfs_root *root, struct page *page) | |
265 | { | |
d1310b2e | 266 | struct extent_io_tree *tree; |
35ebb934 | 267 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
268 | u64 found_start; |
269 | int found_level; | |
270 | unsigned long len; | |
271 | struct extent_buffer *eb; | |
f188591e CM |
272 | int ret; |
273 | ||
d1310b2e | 274 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
275 | |
276 | if (page->private == EXTENT_PAGE_PRIVATE) | |
277 | goto out; | |
278 | if (!page->private) | |
279 | goto out; | |
280 | len = page->private >> 2; | |
281 | if (len == 0) { | |
282 | WARN_ON(1); | |
283 | } | |
284 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
ca7a79ad CM |
285 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
286 | btrfs_header_generation(eb)); | |
f188591e | 287 | BUG_ON(ret); |
e18e4809 | 288 | btrfs_clear_buffer_defrag(eb); |
19c00ddc CM |
289 | found_start = btrfs_header_bytenr(eb); |
290 | if (found_start != start) { | |
291 | printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", | |
292 | start, found_start, len); | |
55c69072 CM |
293 | WARN_ON(1); |
294 | goto err; | |
295 | } | |
296 | if (eb->first_page != page) { | |
297 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
298 | page->index); | |
299 | WARN_ON(1); | |
300 | goto err; | |
301 | } | |
302 | if (!PageUptodate(page)) { | |
303 | printk("csum not up to date page %lu\n", page->index); | |
304 | WARN_ON(1); | |
305 | goto err; | |
19c00ddc CM |
306 | } |
307 | found_level = btrfs_header_level(eb); | |
63b10fc4 CM |
308 | spin_lock(&root->fs_info->hash_lock); |
309 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); | |
310 | spin_unlock(&root->fs_info->hash_lock); | |
19c00ddc | 311 | csum_tree_block(root, eb, 0); |
55c69072 | 312 | err: |
19c00ddc CM |
313 | free_extent_buffer(eb); |
314 | out: | |
315 | return 0; | |
316 | } | |
317 | ||
0da5468f | 318 | static int btree_writepage_io_hook(struct page *page, u64 start, u64 end) |
d98237b3 | 319 | { |
19c00ddc | 320 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
19c00ddc CM |
321 | |
322 | csum_dirty_buffer(root, page); | |
0da5468f CM |
323 | return 0; |
324 | } | |
325 | ||
ce9adaa5 CM |
326 | int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
327 | struct extent_state *state) | |
328 | { | |
329 | struct extent_io_tree *tree; | |
330 | u64 found_start; | |
331 | int found_level; | |
332 | unsigned long len; | |
333 | struct extent_buffer *eb; | |
334 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 335 | int ret = 0; |
ce9adaa5 CM |
336 | |
337 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
338 | if (page->private == EXTENT_PAGE_PRIVATE) | |
339 | goto out; | |
340 | if (!page->private) | |
341 | goto out; | |
342 | len = page->private >> 2; | |
343 | if (len == 0) { | |
344 | WARN_ON(1); | |
345 | } | |
346 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
f188591e | 347 | |
ce9adaa5 CM |
348 | btrfs_clear_buffer_defrag(eb); |
349 | found_start = btrfs_header_bytenr(eb); | |
350 | if (found_start != start) { | |
f188591e | 351 | ret = -EIO; |
ce9adaa5 CM |
352 | goto err; |
353 | } | |
354 | if (eb->first_page != page) { | |
355 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
356 | page->index); | |
357 | WARN_ON(1); | |
f188591e | 358 | ret = -EIO; |
ce9adaa5 CM |
359 | goto err; |
360 | } | |
1259ab75 CM |
361 | if (memcmp_extent_buffer(eb, root->fs_info->fsid, |
362 | (unsigned long)btrfs_header_fsid(eb), | |
363 | BTRFS_FSID_SIZE)) { | |
364 | printk("bad fsid on block %Lu\n", eb->start); | |
365 | ret = -EIO; | |
366 | goto err; | |
367 | } | |
ce9adaa5 CM |
368 | found_level = btrfs_header_level(eb); |
369 | ||
370 | ret = csum_tree_block(root, eb, 1); | |
f188591e CM |
371 | if (ret) |
372 | ret = -EIO; | |
ce9adaa5 CM |
373 | |
374 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
375 | end = eb->start + end - 1; | |
376 | release_extent_buffer_tail_pages(eb); | |
377 | err: | |
378 | free_extent_buffer(eb); | |
379 | out: | |
f188591e | 380 | return ret; |
ce9adaa5 CM |
381 | } |
382 | ||
383 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) | |
384 | static void end_workqueue_bio(struct bio *bio, int err) | |
385 | #else | |
386 | static int end_workqueue_bio(struct bio *bio, | |
387 | unsigned int bytes_done, int err) | |
388 | #endif | |
389 | { | |
390 | struct end_io_wq *end_io_wq = bio->bi_private; | |
391 | struct btrfs_fs_info *fs_info; | |
392 | unsigned long flags; | |
393 | ||
394 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
395 | if (bio->bi_size) | |
396 | return 1; | |
397 | #endif | |
398 | ||
399 | fs_info = end_io_wq->info; | |
400 | spin_lock_irqsave(&fs_info->end_io_work_lock, flags); | |
401 | end_io_wq->error = err; | |
402 | list_add_tail(&end_io_wq->list, &fs_info->end_io_work_list); | |
403 | spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags); | |
404 | queue_work(end_io_workqueue, &fs_info->end_io_work); | |
405 | ||
406 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
407 | return 0; | |
408 | #endif | |
409 | } | |
410 | ||
22c59948 CM |
411 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
412 | int metadata) | |
0b86a832 | 413 | { |
ce9adaa5 | 414 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
415 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
416 | if (!end_io_wq) | |
417 | return -ENOMEM; | |
418 | ||
419 | end_io_wq->private = bio->bi_private; | |
420 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 421 | end_io_wq->info = info; |
ce9adaa5 CM |
422 | end_io_wq->error = 0; |
423 | end_io_wq->bio = bio; | |
22c59948 | 424 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
425 | |
426 | bio->bi_private = end_io_wq; | |
427 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
428 | return 0; |
429 | } | |
430 | ||
44b8bd7e CM |
431 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
432 | int rw, struct bio *bio, int mirror_num, | |
433 | extent_submit_bio_hook_t *submit_bio_hook) | |
434 | { | |
435 | struct async_submit_bio *async; | |
436 | ||
437 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
438 | if (!async) | |
439 | return -ENOMEM; | |
440 | ||
441 | async->inode = inode; | |
442 | async->rw = rw; | |
443 | async->bio = bio; | |
444 | async->mirror_num = mirror_num; | |
445 | async->submit_bio_hook = submit_bio_hook; | |
446 | ||
447 | spin_lock(&fs_info->async_submit_work_lock); | |
448 | list_add_tail(&async->list, &fs_info->async_submit_work_list); | |
cb03c743 | 449 | atomic_inc(&fs_info->nr_async_submits); |
44b8bd7e CM |
450 | spin_unlock(&fs_info->async_submit_work_lock); |
451 | ||
452 | queue_work(async_submit_workqueue, &fs_info->async_submit_work); | |
453 | return 0; | |
454 | } | |
455 | ||
456 | static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, | |
f188591e | 457 | int mirror_num) |
22c59948 CM |
458 | { |
459 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
460 | u64 offset; | |
461 | int ret; | |
462 | ||
463 | offset = bio->bi_sector << 9; | |
464 | ||
465 | if (rw & (1 << BIO_RW)) { | |
f188591e | 466 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num); |
22c59948 CM |
467 | } |
468 | ||
469 | ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); | |
470 | BUG_ON(ret); | |
ce9adaa5 | 471 | |
f188591e | 472 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num); |
0b86a832 CM |
473 | } |
474 | ||
44b8bd7e CM |
475 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
476 | int mirror_num) | |
477 | { | |
478 | if (!(rw & (1 << BIO_RW))) { | |
479 | return __btree_submit_bio_hook(inode, rw, bio, mirror_num); | |
480 | } | |
481 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
482 | inode, rw, bio, mirror_num, | |
483 | __btree_submit_bio_hook); | |
484 | } | |
485 | ||
0da5468f CM |
486 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
487 | { | |
d1310b2e CM |
488 | struct extent_io_tree *tree; |
489 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
490 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
491 | } | |
0da5468f CM |
492 | |
493 | static int btree_writepages(struct address_space *mapping, | |
494 | struct writeback_control *wbc) | |
495 | { | |
d1310b2e CM |
496 | struct extent_io_tree *tree; |
497 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 498 | if (wbc->sync_mode == WB_SYNC_NONE) { |
793955bc CM |
499 | u64 num_dirty; |
500 | u64 start = 0; | |
501 | unsigned long thresh = 96 * 1024 * 1024; | |
448d640b CM |
502 | |
503 | if (wbc->for_kupdate) | |
504 | return 0; | |
505 | ||
ca664626 CM |
506 | if (current_is_pdflush()) { |
507 | thresh = 96 * 1024 * 1024; | |
508 | } else { | |
509 | thresh = 8 * 1024 * 1024; | |
510 | } | |
1832a6d5 CM |
511 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
512 | thresh, EXTENT_DIRTY); | |
793955bc CM |
513 | if (num_dirty < thresh) { |
514 | return 0; | |
515 | } | |
516 | } | |
0da5468f CM |
517 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
518 | } | |
519 | ||
5f39d397 CM |
520 | int btree_readpage(struct file *file, struct page *page) |
521 | { | |
d1310b2e CM |
522 | struct extent_io_tree *tree; |
523 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
524 | return extent_read_full_page(tree, page, btree_get_extent); |
525 | } | |
22b0ebda | 526 | |
70dec807 | 527 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 528 | { |
d1310b2e CM |
529 | struct extent_io_tree *tree; |
530 | struct extent_map_tree *map; | |
5f39d397 | 531 | int ret; |
d98237b3 | 532 | |
3dd39914 CM |
533 | if (page_count(page) > 3) { |
534 | /* once for page->private, once for the caller, once | |
535 | * once for the page cache | |
536 | */ | |
537 | return 0; | |
538 | } | |
d1310b2e CM |
539 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
540 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
7b13b7b1 | 541 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
5f39d397 | 542 | if (ret == 1) { |
728131d8 | 543 | invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); |
5f39d397 CM |
544 | ClearPagePrivate(page); |
545 | set_page_private(page, 0); | |
546 | page_cache_release(page); | |
547 | } | |
d98237b3 CM |
548 | return ret; |
549 | } | |
550 | ||
5f39d397 | 551 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 552 | { |
d1310b2e CM |
553 | struct extent_io_tree *tree; |
554 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
555 | extent_invalidatepage(tree, page, offset); |
556 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 557 | if (PagePrivate(page)) { |
4ef64eae | 558 | invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); |
9ad6b7bc CM |
559 | ClearPagePrivate(page); |
560 | set_page_private(page, 0); | |
561 | page_cache_release(page); | |
562 | } | |
d98237b3 CM |
563 | } |
564 | ||
5f39d397 | 565 | #if 0 |
d98237b3 | 566 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 567 | { |
87cbda5c | 568 | struct buffer_head *bh; |
0f7d52f4 | 569 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 570 | struct buffer_head *head; |
87cbda5c CM |
571 | if (!page_has_buffers(page)) { |
572 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
573 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
574 | } | |
575 | head = page_buffers(page); | |
576 | bh = head; | |
577 | do { | |
578 | if (buffer_dirty(bh)) | |
579 | csum_tree_block(root, bh, 0); | |
580 | bh = bh->b_this_page; | |
581 | } while (bh != head); | |
d98237b3 | 582 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 583 | } |
5f39d397 | 584 | #endif |
eb60ceac | 585 | |
d98237b3 CM |
586 | static struct address_space_operations btree_aops = { |
587 | .readpage = btree_readpage, | |
588 | .writepage = btree_writepage, | |
0da5468f | 589 | .writepages = btree_writepages, |
5f39d397 CM |
590 | .releasepage = btree_releasepage, |
591 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
592 | .sync_page = block_sync_page, |
593 | }; | |
594 | ||
ca7a79ad CM |
595 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
596 | u64 parent_transid) | |
090d1875 | 597 | { |
5f39d397 CM |
598 | struct extent_buffer *buf = NULL; |
599 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 600 | int ret = 0; |
090d1875 | 601 | |
db94535d | 602 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 603 | if (!buf) |
090d1875 | 604 | return 0; |
d1310b2e | 605 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 606 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 607 | free_extent_buffer(buf); |
de428b63 | 608 | return ret; |
090d1875 CM |
609 | } |
610 | ||
0999df54 CM |
611 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
612 | u64 bytenr, u32 blocksize) | |
613 | { | |
614 | struct inode *btree_inode = root->fs_info->btree_inode; | |
615 | struct extent_buffer *eb; | |
616 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
617 | bytenr, blocksize, GFP_NOFS); | |
618 | return eb; | |
619 | } | |
620 | ||
621 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
622 | u64 bytenr, u32 blocksize) | |
623 | { | |
624 | struct inode *btree_inode = root->fs_info->btree_inode; | |
625 | struct extent_buffer *eb; | |
626 | ||
627 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
628 | bytenr, blocksize, NULL, GFP_NOFS); | |
629 | return eb; | |
630 | } | |
631 | ||
632 | ||
633 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, | |
ca7a79ad | 634 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
635 | { |
636 | struct extent_buffer *buf = NULL; | |
637 | struct inode *btree_inode = root->fs_info->btree_inode; | |
638 | struct extent_io_tree *io_tree; | |
639 | int ret; | |
640 | ||
641 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
642 | ||
643 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
644 | if (!buf) | |
645 | return NULL; | |
0999df54 | 646 | |
ca7a79ad | 647 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
648 | |
649 | if (ret == 0) { | |
650 | buf->flags |= EXTENT_UPTODATE; | |
651 | } | |
5f39d397 | 652 | return buf; |
ce9adaa5 | 653 | |
eb60ceac CM |
654 | } |
655 | ||
e089f05c | 656 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 657 | struct extent_buffer *buf) |
ed2ff2cb | 658 | { |
5f39d397 | 659 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 CM |
660 | if (btrfs_header_generation(buf) == |
661 | root->fs_info->running_transaction->transid) | |
d1310b2e | 662 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 663 | buf); |
5f39d397 CM |
664 | return 0; |
665 | } | |
666 | ||
667 | int wait_on_tree_block_writeback(struct btrfs_root *root, | |
668 | struct extent_buffer *buf) | |
669 | { | |
670 | struct inode *btree_inode = root->fs_info->btree_inode; | |
d1310b2e | 671 | wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
672 | buf); |
673 | return 0; | |
674 | } | |
675 | ||
db94535d | 676 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 677 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 678 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 679 | u64 objectid) |
d97e63b6 | 680 | { |
cfaa7295 | 681 | root->node = NULL; |
0f7d52f4 | 682 | root->inode = NULL; |
a28ec197 | 683 | root->commit_root = NULL; |
db94535d CM |
684 | root->sectorsize = sectorsize; |
685 | root->nodesize = nodesize; | |
686 | root->leafsize = leafsize; | |
87ee04eb | 687 | root->stripesize = stripesize; |
123abc88 | 688 | root->ref_cows = 0; |
0b86a832 CM |
689 | root->track_dirty = 0; |
690 | ||
9f5fae2f | 691 | root->fs_info = fs_info; |
0f7d52f4 CM |
692 | root->objectid = objectid; |
693 | root->last_trans = 0; | |
1b05da2e CM |
694 | root->highest_inode = 0; |
695 | root->last_inode_alloc = 0; | |
58176a96 | 696 | root->name = NULL; |
4313b399 | 697 | root->in_sysfs = 0; |
0b86a832 CM |
698 | |
699 | INIT_LIST_HEAD(&root->dirty_list); | |
3768f368 CM |
700 | memset(&root->root_key, 0, sizeof(root->root_key)); |
701 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 702 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 JB |
703 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
704 | init_completion(&root->kobj_unregister); | |
6702ed49 CM |
705 | root->defrag_running = 0; |
706 | root->defrag_level = 0; | |
4d775673 | 707 | root->root_key.objectid = objectid; |
3768f368 CM |
708 | return 0; |
709 | } | |
710 | ||
db94535d | 711 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
712 | struct btrfs_fs_info *fs_info, |
713 | u64 objectid, | |
e20d96d6 | 714 | struct btrfs_root *root) |
3768f368 CM |
715 | { |
716 | int ret; | |
db94535d | 717 | u32 blocksize; |
3768f368 | 718 | |
db94535d | 719 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
720 | tree_root->sectorsize, tree_root->stripesize, |
721 | root, fs_info, objectid); | |
3768f368 CM |
722 | ret = btrfs_find_last_root(tree_root, objectid, |
723 | &root->root_item, &root->root_key); | |
724 | BUG_ON(ret); | |
725 | ||
db94535d CM |
726 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
727 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 728 | blocksize, 0); |
3768f368 | 729 | BUG_ON(!root->node); |
d97e63b6 CM |
730 | return 0; |
731 | } | |
732 | ||
5eda7b5e CM |
733 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, |
734 | struct btrfs_key *location) | |
0f7d52f4 CM |
735 | { |
736 | struct btrfs_root *root; | |
737 | struct btrfs_root *tree_root = fs_info->tree_root; | |
738 | struct btrfs_path *path; | |
5f39d397 | 739 | struct extent_buffer *l; |
1b05da2e | 740 | u64 highest_inode; |
db94535d | 741 | u32 blocksize; |
0f7d52f4 CM |
742 | int ret = 0; |
743 | ||
5eda7b5e | 744 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 745 | if (!root) |
0f7d52f4 | 746 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 747 | if (location->offset == (u64)-1) { |
db94535d | 748 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
749 | location->objectid, root); |
750 | if (ret) { | |
0f7d52f4 CM |
751 | kfree(root); |
752 | return ERR_PTR(ret); | |
753 | } | |
754 | goto insert; | |
755 | } | |
756 | ||
db94535d | 757 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
758 | tree_root->sectorsize, tree_root->stripesize, |
759 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
760 | |
761 | path = btrfs_alloc_path(); | |
762 | BUG_ON(!path); | |
763 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
764 | if (ret != 0) { | |
0f7d52f4 CM |
765 | if (ret > 0) |
766 | ret = -ENOENT; | |
767 | goto out; | |
768 | } | |
5f39d397 CM |
769 | l = path->nodes[0]; |
770 | read_extent_buffer(l, &root->root_item, | |
771 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 772 | sizeof(root->root_item)); |
44b36eb2 | 773 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
774 | ret = 0; |
775 | out: | |
776 | btrfs_release_path(root, path); | |
777 | btrfs_free_path(path); | |
778 | if (ret) { | |
779 | kfree(root); | |
780 | return ERR_PTR(ret); | |
781 | } | |
db94535d CM |
782 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
783 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 784 | blocksize, 0); |
0f7d52f4 CM |
785 | BUG_ON(!root->node); |
786 | insert: | |
0f7d52f4 | 787 | root->ref_cows = 1; |
5eda7b5e CM |
788 | ret = btrfs_find_highest_inode(root, &highest_inode); |
789 | if (ret == 0) { | |
790 | root->highest_inode = highest_inode; | |
791 | root->last_inode_alloc = highest_inode; | |
792 | } | |
793 | return root; | |
794 | } | |
795 | ||
dc17ff8f CM |
796 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
797 | u64 root_objectid) | |
798 | { | |
799 | struct btrfs_root *root; | |
800 | ||
801 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
802 | return fs_info->tree_root; | |
803 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
804 | return fs_info->extent_root; | |
805 | ||
806 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
807 | (unsigned long)root_objectid); | |
808 | return root; | |
809 | } | |
810 | ||
edbd8d4e CM |
811 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
812 | struct btrfs_key *location) | |
5eda7b5e CM |
813 | { |
814 | struct btrfs_root *root; | |
815 | int ret; | |
816 | ||
edbd8d4e CM |
817 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
818 | return fs_info->tree_root; | |
819 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
820 | return fs_info->extent_root; | |
8f18cf13 CM |
821 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
822 | return fs_info->chunk_root; | |
823 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
824 | return fs_info->dev_root; | |
edbd8d4e | 825 | |
5eda7b5e CM |
826 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
827 | (unsigned long)location->objectid); | |
828 | if (root) | |
829 | return root; | |
830 | ||
831 | root = btrfs_read_fs_root_no_radix(fs_info, location); | |
832 | if (IS_ERR(root)) | |
833 | return root; | |
2619ba1f CM |
834 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
835 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
836 | root); |
837 | if (ret) { | |
5f39d397 | 838 | free_extent_buffer(root->node); |
0f7d52f4 CM |
839 | kfree(root); |
840 | return ERR_PTR(ret); | |
841 | } | |
edbd8d4e CM |
842 | ret = btrfs_find_dead_roots(fs_info->tree_root, |
843 | root->root_key.objectid, root); | |
844 | BUG_ON(ret); | |
845 | ||
846 | return root; | |
847 | } | |
848 | ||
849 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
850 | struct btrfs_key *location, | |
851 | const char *name, int namelen) | |
852 | { | |
853 | struct btrfs_root *root; | |
854 | int ret; | |
855 | ||
856 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
857 | if (!root) | |
858 | return NULL; | |
58176a96 | 859 | |
4313b399 CM |
860 | if (root->in_sysfs) |
861 | return root; | |
862 | ||
58176a96 JB |
863 | ret = btrfs_set_root_name(root, name, namelen); |
864 | if (ret) { | |
5f39d397 | 865 | free_extent_buffer(root->node); |
58176a96 JB |
866 | kfree(root); |
867 | return ERR_PTR(ret); | |
868 | } | |
869 | ||
870 | ret = btrfs_sysfs_add_root(root); | |
871 | if (ret) { | |
5f39d397 | 872 | free_extent_buffer(root->node); |
58176a96 JB |
873 | kfree(root->name); |
874 | kfree(root); | |
875 | return ERR_PTR(ret); | |
876 | } | |
4313b399 | 877 | root->in_sysfs = 1; |
0f7d52f4 CM |
878 | return root; |
879 | } | |
19c00ddc CM |
880 | #if 0 |
881 | static int add_hasher(struct btrfs_fs_info *info, char *type) { | |
882 | struct btrfs_hasher *hasher; | |
883 | ||
884 | hasher = kmalloc(sizeof(*hasher), GFP_NOFS); | |
885 | if (!hasher) | |
886 | return -ENOMEM; | |
887 | hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); | |
888 | if (!hasher->hash_tfm) { | |
889 | kfree(hasher); | |
890 | return -EINVAL; | |
891 | } | |
892 | spin_lock(&info->hash_lock); | |
893 | list_add(&hasher->list, &info->hashers); | |
894 | spin_unlock(&info->hash_lock); | |
895 | return 0; | |
896 | } | |
897 | #endif | |
04160088 CM |
898 | |
899 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
900 | { | |
901 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
902 | int ret = 0; | |
cb03c743 | 903 | int limit = 256 * info->fs_devices->open_devices; |
04160088 CM |
904 | struct list_head *cur; |
905 | struct btrfs_device *device; | |
906 | struct backing_dev_info *bdi; | |
907 | ||
cb03c743 CM |
908 | if ((bdi_bits & (1 << BDI_write_congested)) && |
909 | atomic_read(&info->nr_async_submits) > limit) { | |
910 | return 1; | |
911 | } | |
912 | ||
04160088 CM |
913 | list_for_each(cur, &info->fs_devices->devices) { |
914 | device = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
915 | if (!device->bdev) |
916 | continue; | |
04160088 CM |
917 | bdi = blk_get_backing_dev_info(device->bdev); |
918 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
919 | ret = 1; | |
920 | break; | |
921 | } | |
922 | } | |
923 | return ret; | |
924 | } | |
925 | ||
38b66988 CM |
926 | /* |
927 | * this unplugs every device on the box, and it is only used when page | |
928 | * is null | |
929 | */ | |
930 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
931 | { | |
932 | struct list_head *cur; | |
933 | struct btrfs_device *device; | |
934 | struct btrfs_fs_info *info; | |
935 | ||
936 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
937 | list_for_each(cur, &info->fs_devices->devices) { | |
938 | device = list_entry(cur, struct btrfs_device, dev_list); | |
939 | bdi = blk_get_backing_dev_info(device->bdev); | |
940 | if (bdi->unplug_io_fn) { | |
941 | bdi->unplug_io_fn(bdi, page); | |
942 | } | |
943 | } | |
944 | } | |
945 | ||
04160088 CM |
946 | void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
947 | { | |
38b66988 | 948 | struct inode *inode; |
f2d8d74d CM |
949 | struct extent_map_tree *em_tree; |
950 | struct extent_map *em; | |
bcbfce8a | 951 | struct address_space *mapping; |
38b66988 CM |
952 | u64 offset; |
953 | ||
bcbfce8a | 954 | /* the generic O_DIRECT read code does this */ |
38b66988 CM |
955 | if (!page) { |
956 | __unplug_io_fn(bdi, page); | |
957 | return; | |
958 | } | |
959 | ||
bcbfce8a CM |
960 | /* |
961 | * page->mapping may change at any time. Get a consistent copy | |
962 | * and use that for everything below | |
963 | */ | |
964 | smp_mb(); | |
965 | mapping = page->mapping; | |
966 | if (!mapping) | |
967 | return; | |
968 | ||
969 | inode = mapping->host; | |
38b66988 | 970 | offset = page_offset(page); |
04160088 | 971 | |
f2d8d74d CM |
972 | em_tree = &BTRFS_I(inode)->extent_tree; |
973 | spin_lock(&em_tree->lock); | |
974 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
975 | spin_unlock(&em_tree->lock); | |
976 | if (!em) | |
977 | return; | |
978 | ||
979 | offset = offset - em->start; | |
980 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
981 | em->block_start + offset, page); | |
982 | free_extent_map(em); | |
04160088 CM |
983 | } |
984 | ||
985 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
986 | { | |
b248a415 | 987 | #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23) |
04160088 | 988 | bdi_init(bdi); |
b248a415 | 989 | #endif |
4575c9cc | 990 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
991 | bdi->state = 0; |
992 | bdi->capabilities = default_backing_dev_info.capabilities; | |
993 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
994 | bdi->unplug_io_data = info; | |
995 | bdi->congested_fn = btrfs_congested_fn; | |
996 | bdi->congested_data = info; | |
997 | return 0; | |
998 | } | |
999 | ||
ce9adaa5 CM |
1000 | static int bio_ready_for_csum(struct bio *bio) |
1001 | { | |
1002 | u64 length = 0; | |
1003 | u64 buf_len = 0; | |
1004 | u64 start = 0; | |
1005 | struct page *page; | |
1006 | struct extent_io_tree *io_tree = NULL; | |
1007 | struct btrfs_fs_info *info = NULL; | |
1008 | struct bio_vec *bvec; | |
1009 | int i; | |
1010 | int ret; | |
1011 | ||
1012 | bio_for_each_segment(bvec, bio, i) { | |
1013 | page = bvec->bv_page; | |
1014 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1015 | length += bvec->bv_len; | |
1016 | continue; | |
1017 | } | |
1018 | if (!page->private) { | |
1019 | length += bvec->bv_len; | |
1020 | continue; | |
1021 | } | |
1022 | length = bvec->bv_len; | |
1023 | buf_len = page->private >> 2; | |
1024 | start = page_offset(page) + bvec->bv_offset; | |
1025 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1026 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1027 | } | |
1028 | /* are we fully contained in this bio? */ | |
1029 | if (buf_len <= length) | |
1030 | return 1; | |
1031 | ||
1032 | ret = extent_range_uptodate(io_tree, start + length, | |
1033 | start + buf_len - 1); | |
1034 | if (ret == 1) | |
1035 | return ret; | |
1036 | return ret; | |
1037 | } | |
1038 | ||
1039 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) | |
44b8bd7e | 1040 | static void btrfs_end_io_csum(void *p) |
ce9adaa5 | 1041 | #else |
44b8bd7e | 1042 | static void btrfs_end_io_csum(struct work_struct *work) |
ce9adaa5 CM |
1043 | #endif |
1044 | { | |
1045 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) | |
1046 | struct btrfs_fs_info *fs_info = p; | |
1047 | #else | |
1048 | struct btrfs_fs_info *fs_info = container_of(work, | |
1049 | struct btrfs_fs_info, | |
1050 | end_io_work); | |
1051 | #endif | |
1052 | unsigned long flags; | |
1053 | struct end_io_wq *end_io_wq; | |
1054 | struct bio *bio; | |
1055 | struct list_head *next; | |
1056 | int error; | |
1057 | int was_empty; | |
1058 | ||
1059 | while(1) { | |
1060 | spin_lock_irqsave(&fs_info->end_io_work_lock, flags); | |
1061 | if (list_empty(&fs_info->end_io_work_list)) { | |
1062 | spin_unlock_irqrestore(&fs_info->end_io_work_lock, | |
1063 | flags); | |
1064 | return; | |
1065 | } | |
1066 | next = fs_info->end_io_work_list.next; | |
1067 | list_del(next); | |
1068 | spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags); | |
1069 | ||
1070 | end_io_wq = list_entry(next, struct end_io_wq, list); | |
1071 | ||
1072 | bio = end_io_wq->bio; | |
22c59948 | 1073 | if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { |
ce9adaa5 CM |
1074 | spin_lock_irqsave(&fs_info->end_io_work_lock, flags); |
1075 | was_empty = list_empty(&fs_info->end_io_work_list); | |
1076 | list_add_tail(&end_io_wq->list, | |
1077 | &fs_info->end_io_work_list); | |
1078 | spin_unlock_irqrestore(&fs_info->end_io_work_lock, | |
1079 | flags); | |
1080 | if (was_empty) | |
1081 | return; | |
1082 | continue; | |
1083 | } | |
1084 | error = end_io_wq->error; | |
1085 | bio->bi_private = end_io_wq->private; | |
1086 | bio->bi_end_io = end_io_wq->end_io; | |
1087 | kfree(end_io_wq); | |
73f61b2a | 1088 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) |
b248a415 | 1089 | bio_endio(bio, bio->bi_size, error); |
73f61b2a | 1090 | #else |
ce9adaa5 | 1091 | bio_endio(bio, error); |
73f61b2a | 1092 | #endif |
ce9adaa5 CM |
1093 | } |
1094 | } | |
1095 | ||
44b8bd7e CM |
1096 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
1097 | static void btrfs_async_submit_work(void *p) | |
1098 | #else | |
1099 | static void btrfs_async_submit_work(struct work_struct *work) | |
1100 | #endif | |
1101 | { | |
1102 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) | |
1103 | struct btrfs_fs_info *fs_info = p; | |
1104 | #else | |
1105 | struct btrfs_fs_info *fs_info = container_of(work, | |
1106 | struct btrfs_fs_info, | |
1107 | async_submit_work); | |
1108 | #endif | |
1109 | struct async_submit_bio *async; | |
1110 | struct list_head *next; | |
1111 | ||
1112 | while(1) { | |
1113 | spin_lock(&fs_info->async_submit_work_lock); | |
1114 | if (list_empty(&fs_info->async_submit_work_list)) { | |
1115 | spin_unlock(&fs_info->async_submit_work_lock); | |
1116 | return; | |
1117 | } | |
1118 | next = fs_info->async_submit_work_list.next; | |
1119 | list_del(next); | |
cb03c743 | 1120 | atomic_dec(&fs_info->nr_async_submits); |
44b8bd7e CM |
1121 | spin_unlock(&fs_info->async_submit_work_lock); |
1122 | ||
1123 | async = list_entry(next, struct async_submit_bio, list); | |
1124 | async->submit_bio_hook(async->inode, async->rw, async->bio, | |
1125 | async->mirror_num); | |
1126 | kfree(async); | |
1127 | } | |
1128 | } | |
1129 | ||
8a4b83cc | 1130 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1131 | struct btrfs_fs_devices *fs_devices, |
1132 | char *options) | |
2e635a27 | 1133 | { |
db94535d CM |
1134 | u32 sectorsize; |
1135 | u32 nodesize; | |
1136 | u32 leafsize; | |
1137 | u32 blocksize; | |
87ee04eb | 1138 | u32 stripesize; |
a061fc8d | 1139 | struct buffer_head *bh; |
e20d96d6 CM |
1140 | struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root), |
1141 | GFP_NOFS); | |
1142 | struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root), | |
1143 | GFP_NOFS); | |
8790d502 | 1144 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1145 | GFP_NOFS); |
0b86a832 CM |
1146 | struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root), |
1147 | GFP_NOFS); | |
1148 | struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root), | |
1149 | GFP_NOFS); | |
eb60ceac | 1150 | int ret; |
e58ca020 | 1151 | int err = -EINVAL; |
2c90e5d6 | 1152 | struct btrfs_super_block *disk_super; |
8790d502 | 1153 | |
39279cc3 CM |
1154 | if (!extent_root || !tree_root || !fs_info) { |
1155 | err = -ENOMEM; | |
1156 | goto fail; | |
1157 | } | |
ce9adaa5 CM |
1158 | end_io_workqueue = create_workqueue("btrfs-end-io"); |
1159 | BUG_ON(!end_io_workqueue); | |
44b8bd7e | 1160 | async_submit_workqueue = create_workqueue("btrfs-async-submit"); |
ce9adaa5 | 1161 | |
0f7d52f4 | 1162 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1163 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1164 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1165 | INIT_LIST_HEAD(&fs_info->hashers); |
ce9adaa5 | 1166 | INIT_LIST_HEAD(&fs_info->end_io_work_list); |
44b8bd7e | 1167 | INIT_LIST_HEAD(&fs_info->async_submit_work_list); |
19c00ddc | 1168 | spin_lock_init(&fs_info->hash_lock); |
ce9adaa5 | 1169 | spin_lock_init(&fs_info->end_io_work_lock); |
44b8bd7e | 1170 | spin_lock_init(&fs_info->async_submit_work_lock); |
1832a6d5 | 1171 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1172 | spin_lock_init(&fs_info->new_trans_lock); |
19c00ddc | 1173 | |
58176a96 | 1174 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1175 | fs_info->tree_root = tree_root; |
1176 | fs_info->extent_root = extent_root; | |
0b86a832 CM |
1177 | fs_info->chunk_root = chunk_root; |
1178 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1179 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1180 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1181 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1182 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1183 | atomic_set(&fs_info->nr_async_submits, 0); |
e20d96d6 | 1184 | fs_info->sb = sb; |
c59f8951 | 1185 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1186 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1187 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1188 | fs_info->btree_inode = new_inode(sb); |
1189 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1190 | fs_info->btree_inode->i_nlink = 1; |
0afbaf8c | 1191 | |
a061fc8d CM |
1192 | sb->s_blocksize = 4096; |
1193 | sb->s_blocksize_bits = blksize_bits(4096); | |
1194 | ||
0afbaf8c CM |
1195 | /* |
1196 | * we set the i_size on the btree inode to the max possible int. | |
1197 | * the real end of the address space is determined by all of | |
1198 | * the devices in the system | |
1199 | */ | |
1200 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1201 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1202 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1203 | ||
d1310b2e | 1204 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1205 | fs_info->btree_inode->i_mapping, |
1206 | GFP_NOFS); | |
d1310b2e CM |
1207 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1208 | GFP_NOFS); | |
1209 | ||
1210 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1211 | |
d1310b2e | 1212 | extent_io_tree_init(&fs_info->free_space_cache, |
f510cfec | 1213 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1214 | extent_io_tree_init(&fs_info->block_group_cache, |
96b5179d | 1215 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1216 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1217 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1218 | extent_io_tree_init(&fs_info->pending_del, |
1a5bc167 | 1219 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1220 | extent_io_tree_init(&fs_info->extent_ins, |
1a5bc167 | 1221 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1222 | fs_info->do_barriers = 1; |
e18e4809 | 1223 | |
6da6abae | 1224 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
b248a415 | 1225 | INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum, fs_info); |
44b8bd7e CM |
1226 | INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work, |
1227 | fs_info); | |
6da6abae CM |
1228 | INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info); |
1229 | #else | |
b248a415 | 1230 | INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum); |
44b8bd7e | 1231 | INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work); |
08607c1b | 1232 | INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner); |
6da6abae | 1233 | #endif |
0f7d52f4 CM |
1234 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1235 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1236 | sizeof(struct btrfs_key)); | |
22b0ebda | 1237 | insert_inode_hash(fs_info->btree_inode); |
d98237b3 | 1238 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
39279cc3 | 1239 | |
79154b1b | 1240 | mutex_init(&fs_info->trans_mutex); |
d561c025 | 1241 | mutex_init(&fs_info->fs_mutex); |
3768f368 | 1242 | |
19c00ddc CM |
1243 | #if 0 |
1244 | ret = add_hasher(fs_info, "crc32c"); | |
1245 | if (ret) { | |
1246 | printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); | |
1247 | err = -ENOMEM; | |
1248 | goto fail_iput; | |
1249 | } | |
1250 | #endif | |
0b86a832 | 1251 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1252 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1253 | |
d98237b3 | 1254 | |
a061fc8d CM |
1255 | bh = __bread(fs_devices->latest_bdev, |
1256 | BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
1257 | if (!bh) | |
39279cc3 | 1258 | goto fail_iput; |
39279cc3 | 1259 | |
a061fc8d CM |
1260 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
1261 | brelse(bh); | |
5f39d397 | 1262 | |
a061fc8d | 1263 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1264 | |
5f39d397 | 1265 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1266 | if (!btrfs_super_root(disk_super)) |
39279cc3 | 1267 | goto fail_sb_buffer; |
0f7d52f4 | 1268 | |
dfe25020 CM |
1269 | btrfs_parse_options(options, tree_root, NULL); |
1270 | ||
a0af469b | 1271 | if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) { |
8a4b83cc CM |
1272 | printk("Btrfs: wanted %llu devices, but found %llu\n", |
1273 | (unsigned long long)btrfs_super_num_devices(disk_super), | |
a0af469b | 1274 | (unsigned long long)fs_devices->open_devices); |
dfe25020 CM |
1275 | if (btrfs_test_opt(tree_root, DEGRADED)) |
1276 | printk("continuing in degraded mode\n"); | |
1277 | else { | |
1278 | goto fail_sb_buffer; | |
1279 | } | |
8a4b83cc | 1280 | } |
dfe25020 | 1281 | |
4575c9cc CM |
1282 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
1283 | ||
db94535d CM |
1284 | nodesize = btrfs_super_nodesize(disk_super); |
1285 | leafsize = btrfs_super_leafsize(disk_super); | |
1286 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1287 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1288 | tree_root->nodesize = nodesize; |
1289 | tree_root->leafsize = leafsize; | |
1290 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1291 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1292 | |
1293 | sb->s_blocksize = sectorsize; | |
1294 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1295 | |
39279cc3 CM |
1296 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1297 | sizeof(disk_super->magic))) { | |
1298 | printk("btrfs: valid FS not found on %s\n", sb->s_id); | |
1299 | goto fail_sb_buffer; | |
1300 | } | |
19c00ddc | 1301 | |
0b86a832 | 1302 | mutex_lock(&fs_info->fs_mutex); |
0d81ba5d | 1303 | |
0b86a832 | 1304 | ret = btrfs_read_sys_array(tree_root); |
84eed90f CM |
1305 | if (ret) { |
1306 | printk("btrfs: failed to read the system array on %s\n", | |
1307 | sb->s_id); | |
1308 | goto fail_sys_array; | |
1309 | } | |
0b86a832 CM |
1310 | |
1311 | blocksize = btrfs_level_size(tree_root, | |
1312 | btrfs_super_chunk_root_level(disk_super)); | |
1313 | ||
1314 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1315 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1316 | ||
1317 | chunk_root->node = read_tree_block(chunk_root, | |
1318 | btrfs_super_chunk_root(disk_super), | |
ca7a79ad | 1319 | blocksize, 0); |
0b86a832 CM |
1320 | BUG_ON(!chunk_root->node); |
1321 | ||
e17cade2 CM |
1322 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
1323 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), | |
1324 | BTRFS_UUID_SIZE); | |
1325 | ||
0b86a832 CM |
1326 | ret = btrfs_read_chunk_tree(chunk_root); |
1327 | BUG_ON(ret); | |
1328 | ||
dfe25020 CM |
1329 | btrfs_close_extra_devices(fs_devices); |
1330 | ||
db94535d CM |
1331 | blocksize = btrfs_level_size(tree_root, |
1332 | btrfs_super_root_level(disk_super)); | |
19c00ddc | 1333 | |
0b86a832 | 1334 | |
e20d96d6 | 1335 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1336 | btrfs_super_root(disk_super), |
ca7a79ad | 1337 | blocksize, 0); |
39279cc3 CM |
1338 | if (!tree_root->node) |
1339 | goto fail_sb_buffer; | |
3768f368 | 1340 | |
db94535d CM |
1341 | |
1342 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1343 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1344 | if (ret) |
39279cc3 | 1345 | goto fail_tree_root; |
0b86a832 CM |
1346 | extent_root->track_dirty = 1; |
1347 | ||
1348 | ret = find_and_setup_root(tree_root, fs_info, | |
1349 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1350 | dev_root->track_dirty = 1; | |
1351 | ||
1352 | if (ret) | |
1353 | goto fail_extent_root; | |
3768f368 | 1354 | |
9078a3e1 CM |
1355 | btrfs_read_block_groups(extent_root); |
1356 | ||
0f7d52f4 | 1357 | fs_info->generation = btrfs_super_generation(disk_super) + 1; |
d18a2c44 CM |
1358 | fs_info->data_alloc_profile = (u64)-1; |
1359 | fs_info->metadata_alloc_profile = (u64)-1; | |
1360 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
1361 | ||
5be6f7f1 | 1362 | mutex_unlock(&fs_info->fs_mutex); |
0f7d52f4 | 1363 | return tree_root; |
39279cc3 | 1364 | |
0b86a832 CM |
1365 | fail_extent_root: |
1366 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1367 | fail_tree_root: |
5f39d397 | 1368 | free_extent_buffer(tree_root->node); |
84eed90f CM |
1369 | fail_sys_array: |
1370 | mutex_unlock(&fs_info->fs_mutex); | |
39279cc3 | 1371 | fail_sb_buffer: |
2d2ae547 | 1372 | extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); |
39279cc3 CM |
1373 | fail_iput: |
1374 | iput(fs_info->btree_inode); | |
1375 | fail: | |
dfe25020 | 1376 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f CM |
1377 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
1378 | ||
39279cc3 CM |
1379 | kfree(extent_root); |
1380 | kfree(tree_root); | |
b248a415 | 1381 | #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23) |
2d2ae547 | 1382 | bdi_destroy(&fs_info->bdi); |
b248a415 | 1383 | #endif |
39279cc3 CM |
1384 | kfree(fs_info); |
1385 | return ERR_PTR(err); | |
eb60ceac CM |
1386 | } |
1387 | ||
f2984462 CM |
1388 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1389 | { | |
1390 | char b[BDEVNAME_SIZE]; | |
1391 | ||
1392 | if (uptodate) { | |
1393 | set_buffer_uptodate(bh); | |
1394 | } else { | |
1395 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1396 | printk(KERN_WARNING "lost page write due to " | |
1397 | "I/O error on %s\n", | |
1398 | bdevname(bh->b_bdev, b)); | |
1399 | } | |
1259ab75 CM |
1400 | /* note, we dont' set_buffer_write_io_error because we have |
1401 | * our own ways of dealing with the IO errors | |
1402 | */ | |
f2984462 CM |
1403 | clear_buffer_uptodate(bh); |
1404 | } | |
1405 | unlock_buffer(bh); | |
1406 | put_bh(bh); | |
1407 | } | |
1408 | ||
1409 | int write_all_supers(struct btrfs_root *root) | |
1410 | { | |
1411 | struct list_head *cur; | |
1412 | struct list_head *head = &root->fs_info->fs_devices->devices; | |
1413 | struct btrfs_device *dev; | |
a061fc8d | 1414 | struct btrfs_super_block *sb; |
f2984462 CM |
1415 | struct btrfs_dev_item *dev_item; |
1416 | struct buffer_head *bh; | |
1417 | int ret; | |
1418 | int do_barriers; | |
a236aed1 CM |
1419 | int max_errors; |
1420 | int total_errors = 0; | |
a061fc8d CM |
1421 | u32 crc; |
1422 | u64 flags; | |
f2984462 | 1423 | |
a236aed1 | 1424 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
1425 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
1426 | ||
a061fc8d CM |
1427 | sb = &root->fs_info->super_for_commit; |
1428 | dev_item = &sb->dev_item; | |
f2984462 CM |
1429 | list_for_each(cur, head) { |
1430 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1431 | if (!dev->bdev) { |
1432 | total_errors++; | |
1433 | continue; | |
1434 | } | |
1435 | if (!dev->in_fs_metadata) | |
1436 | continue; | |
1437 | ||
a061fc8d CM |
1438 | btrfs_set_stack_device_type(dev_item, dev->type); |
1439 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
1440 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
1441 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
1442 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
1443 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
1444 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
1445 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
1446 | flags = btrfs_super_flags(sb); | |
1447 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
1448 | ||
1449 | ||
1450 | crc = ~(u32)0; | |
1451 | crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc, | |
1452 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
1453 | btrfs_csum_final(crc, sb->csum); | |
1454 | ||
1455 | bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096, | |
f2984462 CM |
1456 | BTRFS_SUPER_INFO_SIZE); |
1457 | ||
a061fc8d | 1458 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
f2984462 CM |
1459 | dev->pending_io = bh; |
1460 | ||
1461 | get_bh(bh); | |
1462 | set_buffer_uptodate(bh); | |
1463 | lock_buffer(bh); | |
1464 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
1465 | ||
1466 | if (do_barriers && dev->barriers) { | |
1467 | ret = submit_bh(WRITE_BARRIER, bh); | |
1468 | if (ret == -EOPNOTSUPP) { | |
1469 | printk("btrfs: disabling barriers on dev %s\n", | |
1470 | dev->name); | |
1471 | set_buffer_uptodate(bh); | |
1472 | dev->barriers = 0; | |
1473 | get_bh(bh); | |
1474 | lock_buffer(bh); | |
1475 | ret = submit_bh(WRITE, bh); | |
1476 | } | |
1477 | } else { | |
1478 | ret = submit_bh(WRITE, bh); | |
1479 | } | |
a236aed1 CM |
1480 | if (ret) |
1481 | total_errors++; | |
f2984462 | 1482 | } |
a236aed1 CM |
1483 | if (total_errors > max_errors) { |
1484 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1485 | BUG(); | |
1486 | } | |
1487 | total_errors = 0; | |
f2984462 CM |
1488 | |
1489 | list_for_each(cur, head) { | |
1490 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1491 | if (!dev->bdev) |
1492 | continue; | |
1493 | if (!dev->in_fs_metadata) | |
1494 | continue; | |
1495 | ||
f2984462 CM |
1496 | BUG_ON(!dev->pending_io); |
1497 | bh = dev->pending_io; | |
1498 | wait_on_buffer(bh); | |
1499 | if (!buffer_uptodate(dev->pending_io)) { | |
1500 | if (do_barriers && dev->barriers) { | |
1501 | printk("btrfs: disabling barriers on dev %s\n", | |
1502 | dev->name); | |
1503 | set_buffer_uptodate(bh); | |
1504 | get_bh(bh); | |
1505 | lock_buffer(bh); | |
1506 | dev->barriers = 0; | |
1507 | ret = submit_bh(WRITE, bh); | |
1508 | BUG_ON(ret); | |
1509 | wait_on_buffer(bh); | |
1259ab75 CM |
1510 | if (!buffer_uptodate(bh)) |
1511 | total_errors++; | |
f2984462 | 1512 | } else { |
a236aed1 | 1513 | total_errors++; |
f2984462 CM |
1514 | } |
1515 | ||
1516 | } | |
1517 | dev->pending_io = NULL; | |
1518 | brelse(bh); | |
1519 | } | |
a236aed1 CM |
1520 | if (total_errors > max_errors) { |
1521 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1522 | BUG(); | |
1523 | } | |
f2984462 CM |
1524 | return 0; |
1525 | } | |
1526 | ||
e089f05c | 1527 | int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root |
79154b1b | 1528 | *root) |
eb60ceac | 1529 | { |
e66f709b | 1530 | int ret; |
5f39d397 | 1531 | |
f2984462 | 1532 | ret = write_all_supers(root); |
5f39d397 | 1533 | return ret; |
cfaa7295 CM |
1534 | } |
1535 | ||
5eda7b5e | 1536 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
1537 | { |
1538 | radix_tree_delete(&fs_info->fs_roots_radix, | |
1539 | (unsigned long)root->root_key.objectid); | |
b99aa6cb CM |
1540 | if (root->in_sysfs) |
1541 | btrfs_sysfs_del_root(root); | |
2619ba1f CM |
1542 | if (root->inode) |
1543 | iput(root->inode); | |
1544 | if (root->node) | |
5f39d397 | 1545 | free_extent_buffer(root->node); |
2619ba1f | 1546 | if (root->commit_root) |
5f39d397 | 1547 | free_extent_buffer(root->commit_root); |
58176a96 JB |
1548 | if (root->name) |
1549 | kfree(root->name); | |
2619ba1f CM |
1550 | kfree(root); |
1551 | return 0; | |
1552 | } | |
1553 | ||
35b7e476 | 1554 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
1555 | { |
1556 | int ret; | |
1557 | struct btrfs_root *gang[8]; | |
1558 | int i; | |
1559 | ||
1560 | while(1) { | |
1561 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1562 | (void **)gang, 0, | |
1563 | ARRAY_SIZE(gang)); | |
1564 | if (!ret) | |
1565 | break; | |
2619ba1f | 1566 | for (i = 0; i < ret; i++) |
5eda7b5e | 1567 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
1568 | } |
1569 | return 0; | |
1570 | } | |
b4100d64 | 1571 | |
e20d96d6 | 1572 | int close_ctree(struct btrfs_root *root) |
cfaa7295 | 1573 | { |
3768f368 | 1574 | int ret; |
e089f05c | 1575 | struct btrfs_trans_handle *trans; |
0f7d52f4 | 1576 | struct btrfs_fs_info *fs_info = root->fs_info; |
e089f05c | 1577 | |
facda1e7 | 1578 | fs_info->closing = 1; |
08607c1b | 1579 | btrfs_transaction_flush_work(root); |
0f7d52f4 | 1580 | mutex_lock(&fs_info->fs_mutex); |
6702ed49 | 1581 | btrfs_defrag_dirty_roots(root->fs_info); |
79154b1b | 1582 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 1583 | ret = btrfs_commit_transaction(trans, root); |
79154b1b CM |
1584 | /* run commit again to drop the original snapshot */ |
1585 | trans = btrfs_start_transaction(root, 1); | |
1586 | btrfs_commit_transaction(trans, root); | |
1587 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 1588 | BUG_ON(ret); |
d6bfde87 | 1589 | |
79154b1b | 1590 | write_ctree_super(NULL, root); |
0f7d52f4 CM |
1591 | mutex_unlock(&fs_info->fs_mutex); |
1592 | ||
9ad6b7bc CM |
1593 | btrfs_transaction_flush_work(root); |
1594 | ||
b0c68f8b CM |
1595 | if (fs_info->delalloc_bytes) { |
1596 | printk("btrfs: at unmount delalloc count %Lu\n", | |
1597 | fs_info->delalloc_bytes); | |
1598 | } | |
0f7d52f4 | 1599 | if (fs_info->extent_root->node) |
5f39d397 | 1600 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 1601 | |
0f7d52f4 | 1602 | if (fs_info->tree_root->node) |
5f39d397 | 1603 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 1604 | |
0b86a832 CM |
1605 | if (root->fs_info->chunk_root->node); |
1606 | free_extent_buffer(root->fs_info->chunk_root->node); | |
1607 | ||
1608 | if (root->fs_info->dev_root->node); | |
1609 | free_extent_buffer(root->fs_info->dev_root->node); | |
1610 | ||
9078a3e1 | 1611 | btrfs_free_block_groups(root->fs_info); |
0f7d52f4 | 1612 | del_fs_roots(fs_info); |
d10c5f31 CM |
1613 | |
1614 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1615 | ||
d1310b2e CM |
1616 | extent_io_tree_empty_lru(&fs_info->free_space_cache); |
1617 | extent_io_tree_empty_lru(&fs_info->block_group_cache); | |
1618 | extent_io_tree_empty_lru(&fs_info->pinned_extents); | |
1619 | extent_io_tree_empty_lru(&fs_info->pending_del); | |
1620 | extent_io_tree_empty_lru(&fs_info->extent_ins); | |
1621 | extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); | |
d10c5f31 | 1622 | |
9ad6b7bc | 1623 | flush_workqueue(async_submit_workqueue); |
d6bfde87 | 1624 | flush_workqueue(end_io_workqueue); |
9ad6b7bc | 1625 | |
db94535d | 1626 | truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); |
9ad6b7bc | 1627 | |
44b8bd7e CM |
1628 | flush_workqueue(async_submit_workqueue); |
1629 | destroy_workqueue(async_submit_workqueue); | |
1630 | ||
d6bfde87 CM |
1631 | flush_workqueue(end_io_workqueue); |
1632 | destroy_workqueue(end_io_workqueue); | |
1633 | ||
db94535d | 1634 | iput(fs_info->btree_inode); |
19c00ddc CM |
1635 | #if 0 |
1636 | while(!list_empty(&fs_info->hashers)) { | |
1637 | struct btrfs_hasher *hasher; | |
1638 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
1639 | hashers); | |
1640 | list_del(&hasher->hashers); | |
1641 | crypto_free_hash(&fs_info->hash_tfm); | |
1642 | kfree(hasher); | |
1643 | } | |
1644 | #endif | |
dfe25020 | 1645 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 1646 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 CM |
1647 | |
1648 | #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23) | |
04160088 | 1649 | bdi_destroy(&fs_info->bdi); |
b248a415 | 1650 | #endif |
0b86a832 | 1651 | |
0f7d52f4 | 1652 | kfree(fs_info->extent_root); |
0f7d52f4 | 1653 | kfree(fs_info->tree_root); |
0b86a832 CM |
1654 | kfree(fs_info->chunk_root); |
1655 | kfree(fs_info->dev_root); | |
eb60ceac CM |
1656 | return 0; |
1657 | } | |
1658 | ||
1259ab75 | 1659 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 1660 | { |
1259ab75 | 1661 | int ret; |
810191ff | 1662 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
1663 | |
1664 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
1665 | if (!ret) | |
1666 | return ret; | |
1667 | ||
1668 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
1669 | parent_transid); | |
1670 | return !ret; | |
5f39d397 CM |
1671 | } |
1672 | ||
1673 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 1674 | { |
810191ff | 1675 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 1676 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
1677 | buf); |
1678 | } | |
6702ed49 | 1679 | |
5f39d397 CM |
1680 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
1681 | { | |
810191ff | 1682 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
1683 | u64 transid = btrfs_header_generation(buf); |
1684 | struct inode *btree_inode = root->fs_info->btree_inode; | |
6702ed49 | 1685 | |
ccd467d6 CM |
1686 | if (transid != root->fs_info->generation) { |
1687 | printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", | |
db94535d | 1688 | (unsigned long long)buf->start, |
ccd467d6 CM |
1689 | transid, root->fs_info->generation); |
1690 | WARN_ON(1); | |
1691 | } | |
d1310b2e | 1692 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); |
eb60ceac CM |
1693 | } |
1694 | ||
e2008b61 CM |
1695 | void btrfs_throttle(struct btrfs_root *root) |
1696 | { | |
55c69072 CM |
1697 | struct backing_dev_info *bdi; |
1698 | ||
a061fc8d | 1699 | bdi = &root->fs_info->bdi; |
04005cc7 CM |
1700 | if (root->fs_info->throttles && bdi_write_congested(bdi)) { |
1701 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18) | |
55c69072 | 1702 | congestion_wait(WRITE, HZ/20); |
04005cc7 CM |
1703 | #else |
1704 | blk_congestion_wait(WRITE, HZ/20); | |
1705 | #endif | |
1706 | } | |
e2008b61 CM |
1707 | } |
1708 | ||
d3c2fdcf | 1709 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 1710 | { |
188de649 CM |
1711 | /* |
1712 | * looks as though older kernels can get into trouble with | |
1713 | * this code, they end up stuck in balance_dirty_pages forever | |
1714 | */ | |
d6bfde87 CM |
1715 | struct extent_io_tree *tree; |
1716 | u64 num_dirty; | |
1717 | u64 start = 0; | |
1718 | unsigned long thresh = 16 * 1024 * 1024; | |
1719 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
1720 | ||
1721 | if (current_is_pdflush()) | |
1722 | return; | |
1723 | ||
1724 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
1725 | thresh, EXTENT_DIRTY); | |
1726 | if (num_dirty > thresh) { | |
1727 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 1728 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 1729 | } |
188de649 | 1730 | return; |
35b7e476 | 1731 | } |
6b80053d CM |
1732 | |
1733 | void btrfs_set_buffer_defrag(struct extent_buffer *buf) | |
1734 | { | |
810191ff | 1735 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1736 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1737 | set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start, |
6b80053d CM |
1738 | buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS); |
1739 | } | |
1740 | ||
1741 | void btrfs_set_buffer_defrag_done(struct extent_buffer *buf) | |
1742 | { | |
810191ff | 1743 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1744 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1745 | set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start, |
6b80053d CM |
1746 | buf->start + buf->len - 1, EXTENT_DEFRAG_DONE, |
1747 | GFP_NOFS); | |
1748 | } | |
1749 | ||
1750 | int btrfs_buffer_defrag(struct extent_buffer *buf) | |
1751 | { | |
810191ff | 1752 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1753 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1754 | return test_range_bit(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1755 | buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0); |
1756 | } | |
1757 | ||
1758 | int btrfs_buffer_defrag_done(struct extent_buffer *buf) | |
1759 | { | |
810191ff | 1760 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1761 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1762 | return test_range_bit(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1763 | buf->start, buf->start + buf->len - 1, |
1764 | EXTENT_DEFRAG_DONE, 0); | |
1765 | } | |
1766 | ||
1767 | int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf) | |
1768 | { | |
810191ff | 1769 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1770 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1771 | return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1772 | buf->start, buf->start + buf->len - 1, |
1773 | EXTENT_DEFRAG_DONE, GFP_NOFS); | |
1774 | } | |
1775 | ||
1776 | int btrfs_clear_buffer_defrag(struct extent_buffer *buf) | |
1777 | { | |
810191ff | 1778 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1779 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1780 | return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1781 | buf->start, buf->start + buf->len - 1, |
1782 | EXTENT_DEFRAG, GFP_NOFS); | |
1783 | } | |
1784 | ||
ca7a79ad | 1785 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 1786 | { |
810191ff | 1787 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 1788 | int ret; |
ca7a79ad | 1789 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
1790 | if (ret == 0) { |
1791 | buf->flags |= EXTENT_UPTODATE; | |
1792 | } | |
1793 | return ret; | |
6b80053d | 1794 | } |
0da5468f | 1795 | |
d1310b2e | 1796 | static struct extent_io_ops btree_extent_io_ops = { |
0da5468f | 1797 | .writepage_io_hook = btree_writepage_io_hook, |
ce9adaa5 | 1798 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 1799 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
1800 | /* note we're sharing with inode.c for the merge bio hook */ |
1801 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 1802 | }; |