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 | ||
f188591e CM |
208 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
209 | struct extent_buffer *eb, | |
ca7a79ad | 210 | u64 start, u64 parent_transid) |
f188591e CM |
211 | { |
212 | struct extent_io_tree *io_tree; | |
213 | int ret; | |
214 | int num_copies = 0; | |
215 | int mirror_num = 0; | |
216 | ||
217 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
218 | while (1) { | |
219 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
220 | btree_get_extent, mirror_num); | |
4235298e | 221 | if (!ret) |
f188591e | 222 | return ret; |
4235298e | 223 | |
f188591e CM |
224 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
225 | eb->start, eb->len); | |
4235298e | 226 | if (num_copies == 1) |
f188591e | 227 | return ret; |
4235298e | 228 | |
f188591e | 229 | mirror_num++; |
4235298e | 230 | if (mirror_num > num_copies) |
f188591e | 231 | return ret; |
f188591e | 232 | } |
f188591e CM |
233 | return -EIO; |
234 | } | |
19c00ddc CM |
235 | |
236 | int csum_dirty_buffer(struct btrfs_root *root, struct page *page) | |
237 | { | |
d1310b2e | 238 | struct extent_io_tree *tree; |
35ebb934 | 239 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
240 | u64 found_start; |
241 | int found_level; | |
242 | unsigned long len; | |
243 | struct extent_buffer *eb; | |
f188591e CM |
244 | int ret; |
245 | ||
d1310b2e | 246 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
247 | |
248 | if (page->private == EXTENT_PAGE_PRIVATE) | |
249 | goto out; | |
250 | if (!page->private) | |
251 | goto out; | |
252 | len = page->private >> 2; | |
253 | if (len == 0) { | |
254 | WARN_ON(1); | |
255 | } | |
256 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
ca7a79ad CM |
257 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
258 | btrfs_header_generation(eb)); | |
f188591e | 259 | BUG_ON(ret); |
e18e4809 | 260 | btrfs_clear_buffer_defrag(eb); |
19c00ddc CM |
261 | found_start = btrfs_header_bytenr(eb); |
262 | if (found_start != start) { | |
263 | printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", | |
264 | start, found_start, len); | |
55c69072 CM |
265 | WARN_ON(1); |
266 | goto err; | |
267 | } | |
268 | if (eb->first_page != page) { | |
269 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
270 | page->index); | |
271 | WARN_ON(1); | |
272 | goto err; | |
273 | } | |
274 | if (!PageUptodate(page)) { | |
275 | printk("csum not up to date page %lu\n", page->index); | |
276 | WARN_ON(1); | |
277 | goto err; | |
19c00ddc CM |
278 | } |
279 | found_level = btrfs_header_level(eb); | |
63b10fc4 CM |
280 | spin_lock(&root->fs_info->hash_lock); |
281 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); | |
282 | spin_unlock(&root->fs_info->hash_lock); | |
19c00ddc | 283 | csum_tree_block(root, eb, 0); |
55c69072 | 284 | err: |
19c00ddc CM |
285 | free_extent_buffer(eb); |
286 | out: | |
287 | return 0; | |
288 | } | |
289 | ||
0da5468f | 290 | static int btree_writepage_io_hook(struct page *page, u64 start, u64 end) |
d98237b3 | 291 | { |
19c00ddc | 292 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
19c00ddc CM |
293 | |
294 | csum_dirty_buffer(root, page); | |
0da5468f CM |
295 | return 0; |
296 | } | |
297 | ||
ce9adaa5 CM |
298 | int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
299 | struct extent_state *state) | |
300 | { | |
301 | struct extent_io_tree *tree; | |
302 | u64 found_start; | |
303 | int found_level; | |
304 | unsigned long len; | |
305 | struct extent_buffer *eb; | |
306 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 307 | int ret = 0; |
ce9adaa5 CM |
308 | |
309 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
310 | if (page->private == EXTENT_PAGE_PRIVATE) | |
311 | goto out; | |
312 | if (!page->private) | |
313 | goto out; | |
314 | len = page->private >> 2; | |
315 | if (len == 0) { | |
316 | WARN_ON(1); | |
317 | } | |
318 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
f188591e | 319 | |
ce9adaa5 CM |
320 | btrfs_clear_buffer_defrag(eb); |
321 | found_start = btrfs_header_bytenr(eb); | |
322 | if (found_start != start) { | |
f188591e | 323 | ret = -EIO; |
ce9adaa5 CM |
324 | goto err; |
325 | } | |
326 | if (eb->first_page != page) { | |
327 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
328 | page->index); | |
329 | WARN_ON(1); | |
f188591e | 330 | ret = -EIO; |
ce9adaa5 CM |
331 | goto err; |
332 | } | |
333 | found_level = btrfs_header_level(eb); | |
334 | ||
335 | ret = csum_tree_block(root, eb, 1); | |
f188591e CM |
336 | if (ret) |
337 | ret = -EIO; | |
ce9adaa5 CM |
338 | |
339 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
340 | end = eb->start + end - 1; | |
341 | release_extent_buffer_tail_pages(eb); | |
342 | err: | |
343 | free_extent_buffer(eb); | |
344 | out: | |
f188591e | 345 | return ret; |
ce9adaa5 CM |
346 | } |
347 | ||
348 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) | |
349 | static void end_workqueue_bio(struct bio *bio, int err) | |
350 | #else | |
351 | static int end_workqueue_bio(struct bio *bio, | |
352 | unsigned int bytes_done, int err) | |
353 | #endif | |
354 | { | |
355 | struct end_io_wq *end_io_wq = bio->bi_private; | |
356 | struct btrfs_fs_info *fs_info; | |
357 | unsigned long flags; | |
358 | ||
359 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
360 | if (bio->bi_size) | |
361 | return 1; | |
362 | #endif | |
363 | ||
364 | fs_info = end_io_wq->info; | |
365 | spin_lock_irqsave(&fs_info->end_io_work_lock, flags); | |
366 | end_io_wq->error = err; | |
367 | list_add_tail(&end_io_wq->list, &fs_info->end_io_work_list); | |
368 | spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags); | |
369 | queue_work(end_io_workqueue, &fs_info->end_io_work); | |
370 | ||
371 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
372 | return 0; | |
373 | #endif | |
374 | } | |
375 | ||
22c59948 CM |
376 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
377 | int metadata) | |
0b86a832 | 378 | { |
ce9adaa5 | 379 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
380 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
381 | if (!end_io_wq) | |
382 | return -ENOMEM; | |
383 | ||
384 | end_io_wq->private = bio->bi_private; | |
385 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 386 | end_io_wq->info = info; |
ce9adaa5 CM |
387 | end_io_wq->error = 0; |
388 | end_io_wq->bio = bio; | |
22c59948 | 389 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
390 | |
391 | bio->bi_private = end_io_wq; | |
392 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
393 | return 0; |
394 | } | |
395 | ||
44b8bd7e CM |
396 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
397 | int rw, struct bio *bio, int mirror_num, | |
398 | extent_submit_bio_hook_t *submit_bio_hook) | |
399 | { | |
400 | struct async_submit_bio *async; | |
401 | ||
7b859fe7 CM |
402 | /* |
403 | * inline writerback should stay inline, only hop to the async | |
404 | * queue if we're pdflush | |
405 | */ | |
406 | if (!current_is_pdflush()) | |
407 | return submit_bio_hook(inode, rw, bio, mirror_num); | |
408 | ||
44b8bd7e CM |
409 | async = kmalloc(sizeof(*async), GFP_NOFS); |
410 | if (!async) | |
411 | return -ENOMEM; | |
412 | ||
413 | async->inode = inode; | |
414 | async->rw = rw; | |
415 | async->bio = bio; | |
416 | async->mirror_num = mirror_num; | |
417 | async->submit_bio_hook = submit_bio_hook; | |
418 | ||
419 | spin_lock(&fs_info->async_submit_work_lock); | |
420 | list_add_tail(&async->list, &fs_info->async_submit_work_list); | |
421 | spin_unlock(&fs_info->async_submit_work_lock); | |
422 | ||
423 | queue_work(async_submit_workqueue, &fs_info->async_submit_work); | |
424 | return 0; | |
425 | } | |
426 | ||
427 | static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, | |
f188591e | 428 | int mirror_num) |
22c59948 CM |
429 | { |
430 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
431 | u64 offset; | |
432 | int ret; | |
433 | ||
434 | offset = bio->bi_sector << 9; | |
435 | ||
436 | if (rw & (1 << BIO_RW)) { | |
f188591e | 437 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num); |
22c59948 CM |
438 | } |
439 | ||
440 | ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); | |
441 | BUG_ON(ret); | |
ce9adaa5 | 442 | |
f188591e | 443 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num); |
0b86a832 CM |
444 | } |
445 | ||
44b8bd7e CM |
446 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
447 | int mirror_num) | |
448 | { | |
449 | if (!(rw & (1 << BIO_RW))) { | |
450 | return __btree_submit_bio_hook(inode, rw, bio, mirror_num); | |
451 | } | |
452 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
453 | inode, rw, bio, mirror_num, | |
454 | __btree_submit_bio_hook); | |
455 | } | |
456 | ||
0da5468f CM |
457 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
458 | { | |
d1310b2e CM |
459 | struct extent_io_tree *tree; |
460 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
461 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
462 | } | |
0da5468f CM |
463 | |
464 | static int btree_writepages(struct address_space *mapping, | |
465 | struct writeback_control *wbc) | |
466 | { | |
d1310b2e CM |
467 | struct extent_io_tree *tree; |
468 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 469 | if (wbc->sync_mode == WB_SYNC_NONE) { |
793955bc CM |
470 | u64 num_dirty; |
471 | u64 start = 0; | |
472 | unsigned long thresh = 96 * 1024 * 1024; | |
448d640b CM |
473 | |
474 | if (wbc->for_kupdate) | |
475 | return 0; | |
476 | ||
ca664626 CM |
477 | if (current_is_pdflush()) { |
478 | thresh = 96 * 1024 * 1024; | |
479 | } else { | |
480 | thresh = 8 * 1024 * 1024; | |
481 | } | |
1832a6d5 CM |
482 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
483 | thresh, EXTENT_DIRTY); | |
793955bc CM |
484 | if (num_dirty < thresh) { |
485 | return 0; | |
486 | } | |
487 | } | |
0da5468f CM |
488 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
489 | } | |
490 | ||
5f39d397 CM |
491 | int btree_readpage(struct file *file, struct page *page) |
492 | { | |
d1310b2e CM |
493 | struct extent_io_tree *tree; |
494 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
495 | return extent_read_full_page(tree, page, btree_get_extent); |
496 | } | |
22b0ebda | 497 | |
70dec807 | 498 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 499 | { |
d1310b2e CM |
500 | struct extent_io_tree *tree; |
501 | struct extent_map_tree *map; | |
5f39d397 | 502 | int ret; |
d98237b3 | 503 | |
3dd39914 CM |
504 | if (page_count(page) > 3) { |
505 | /* once for page->private, once for the caller, once | |
506 | * once for the page cache | |
507 | */ | |
508 | return 0; | |
509 | } | |
d1310b2e CM |
510 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
511 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
7b13b7b1 | 512 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
5f39d397 | 513 | if (ret == 1) { |
728131d8 | 514 | invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); |
5f39d397 CM |
515 | ClearPagePrivate(page); |
516 | set_page_private(page, 0); | |
517 | page_cache_release(page); | |
518 | } | |
d98237b3 CM |
519 | return ret; |
520 | } | |
521 | ||
5f39d397 | 522 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 523 | { |
d1310b2e CM |
524 | struct extent_io_tree *tree; |
525 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
526 | extent_invalidatepage(tree, page, offset); |
527 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 528 | if (PagePrivate(page)) { |
4ef64eae | 529 | invalidate_extent_lru(tree, page_offset(page), PAGE_CACHE_SIZE); |
9ad6b7bc CM |
530 | ClearPagePrivate(page); |
531 | set_page_private(page, 0); | |
532 | page_cache_release(page); | |
533 | } | |
d98237b3 CM |
534 | } |
535 | ||
5f39d397 | 536 | #if 0 |
d98237b3 | 537 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 538 | { |
87cbda5c | 539 | struct buffer_head *bh; |
0f7d52f4 | 540 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 541 | struct buffer_head *head; |
87cbda5c CM |
542 | if (!page_has_buffers(page)) { |
543 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
544 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
545 | } | |
546 | head = page_buffers(page); | |
547 | bh = head; | |
548 | do { | |
549 | if (buffer_dirty(bh)) | |
550 | csum_tree_block(root, bh, 0); | |
551 | bh = bh->b_this_page; | |
552 | } while (bh != head); | |
d98237b3 | 553 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 554 | } |
5f39d397 | 555 | #endif |
eb60ceac | 556 | |
d98237b3 CM |
557 | static struct address_space_operations btree_aops = { |
558 | .readpage = btree_readpage, | |
559 | .writepage = btree_writepage, | |
0da5468f | 560 | .writepages = btree_writepages, |
5f39d397 CM |
561 | .releasepage = btree_releasepage, |
562 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
563 | .sync_page = block_sync_page, |
564 | }; | |
565 | ||
ca7a79ad CM |
566 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
567 | u64 parent_transid) | |
090d1875 | 568 | { |
5f39d397 CM |
569 | struct extent_buffer *buf = NULL; |
570 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 571 | int ret = 0; |
090d1875 | 572 | |
db94535d | 573 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 574 | if (!buf) |
090d1875 | 575 | return 0; |
d1310b2e | 576 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 577 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 578 | free_extent_buffer(buf); |
de428b63 | 579 | return ret; |
090d1875 CM |
580 | } |
581 | ||
0b86a832 CM |
582 | static int close_all_devices(struct btrfs_fs_info *fs_info) |
583 | { | |
584 | struct list_head *list; | |
585 | struct list_head *next; | |
586 | struct btrfs_device *device; | |
587 | ||
8a4b83cc CM |
588 | list = &fs_info->fs_devices->devices; |
589 | list_for_each(next, list) { | |
0b86a832 | 590 | device = list_entry(next, struct btrfs_device, dev_list); |
a061fc8d | 591 | close_bdev_excl(device->bdev); |
8a4b83cc | 592 | device->bdev = NULL; |
0b86a832 CM |
593 | } |
594 | return 0; | |
595 | } | |
596 | ||
0999df54 CM |
597 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
598 | u64 bytenr, u32 blocksize) | |
599 | { | |
600 | struct inode *btree_inode = root->fs_info->btree_inode; | |
601 | struct extent_buffer *eb; | |
602 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
603 | bytenr, blocksize, GFP_NOFS); | |
604 | return eb; | |
605 | } | |
606 | ||
607 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
608 | u64 bytenr, u32 blocksize) | |
609 | { | |
610 | struct inode *btree_inode = root->fs_info->btree_inode; | |
611 | struct extent_buffer *eb; | |
612 | ||
613 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
614 | bytenr, blocksize, NULL, GFP_NOFS); | |
615 | return eb; | |
616 | } | |
617 | ||
618 | ||
619 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, | |
ca7a79ad | 620 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
621 | { |
622 | struct extent_buffer *buf = NULL; | |
623 | struct inode *btree_inode = root->fs_info->btree_inode; | |
624 | struct extent_io_tree *io_tree; | |
625 | int ret; | |
626 | ||
627 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
628 | ||
629 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
630 | if (!buf) | |
631 | return NULL; | |
0999df54 | 632 | |
ca7a79ad | 633 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
634 | |
635 | if (ret == 0) { | |
636 | buf->flags |= EXTENT_UPTODATE; | |
637 | } | |
5f39d397 | 638 | return buf; |
ce9adaa5 | 639 | |
eb60ceac CM |
640 | } |
641 | ||
e089f05c | 642 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 643 | struct extent_buffer *buf) |
ed2ff2cb | 644 | { |
5f39d397 | 645 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 CM |
646 | if (btrfs_header_generation(buf) == |
647 | root->fs_info->running_transaction->transid) | |
d1310b2e | 648 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 649 | buf); |
5f39d397 CM |
650 | return 0; |
651 | } | |
652 | ||
653 | int wait_on_tree_block_writeback(struct btrfs_root *root, | |
654 | struct extent_buffer *buf) | |
655 | { | |
656 | struct inode *btree_inode = root->fs_info->btree_inode; | |
d1310b2e | 657 | wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
658 | buf); |
659 | return 0; | |
660 | } | |
661 | ||
db94535d | 662 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 663 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 664 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 665 | u64 objectid) |
d97e63b6 | 666 | { |
cfaa7295 | 667 | root->node = NULL; |
0f7d52f4 | 668 | root->inode = NULL; |
a28ec197 | 669 | root->commit_root = NULL; |
db94535d CM |
670 | root->sectorsize = sectorsize; |
671 | root->nodesize = nodesize; | |
672 | root->leafsize = leafsize; | |
87ee04eb | 673 | root->stripesize = stripesize; |
123abc88 | 674 | root->ref_cows = 0; |
0b86a832 CM |
675 | root->track_dirty = 0; |
676 | ||
9f5fae2f | 677 | root->fs_info = fs_info; |
0f7d52f4 CM |
678 | root->objectid = objectid; |
679 | root->last_trans = 0; | |
1b05da2e CM |
680 | root->highest_inode = 0; |
681 | root->last_inode_alloc = 0; | |
58176a96 | 682 | root->name = NULL; |
4313b399 | 683 | root->in_sysfs = 0; |
0b86a832 CM |
684 | |
685 | INIT_LIST_HEAD(&root->dirty_list); | |
3768f368 CM |
686 | memset(&root->root_key, 0, sizeof(root->root_key)); |
687 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 688 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 JB |
689 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
690 | init_completion(&root->kobj_unregister); | |
6702ed49 CM |
691 | root->defrag_running = 0; |
692 | root->defrag_level = 0; | |
4d775673 | 693 | root->root_key.objectid = objectid; |
3768f368 CM |
694 | return 0; |
695 | } | |
696 | ||
db94535d | 697 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
698 | struct btrfs_fs_info *fs_info, |
699 | u64 objectid, | |
e20d96d6 | 700 | struct btrfs_root *root) |
3768f368 CM |
701 | { |
702 | int ret; | |
db94535d | 703 | u32 blocksize; |
3768f368 | 704 | |
db94535d | 705 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
706 | tree_root->sectorsize, tree_root->stripesize, |
707 | root, fs_info, objectid); | |
3768f368 CM |
708 | ret = btrfs_find_last_root(tree_root, objectid, |
709 | &root->root_item, &root->root_key); | |
710 | BUG_ON(ret); | |
711 | ||
db94535d CM |
712 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
713 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 714 | blocksize, 0); |
3768f368 | 715 | BUG_ON(!root->node); |
d97e63b6 CM |
716 | return 0; |
717 | } | |
718 | ||
5eda7b5e CM |
719 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info, |
720 | struct btrfs_key *location) | |
0f7d52f4 CM |
721 | { |
722 | struct btrfs_root *root; | |
723 | struct btrfs_root *tree_root = fs_info->tree_root; | |
724 | struct btrfs_path *path; | |
5f39d397 | 725 | struct extent_buffer *l; |
1b05da2e | 726 | u64 highest_inode; |
db94535d | 727 | u32 blocksize; |
0f7d52f4 CM |
728 | int ret = 0; |
729 | ||
5eda7b5e | 730 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 731 | if (!root) |
0f7d52f4 | 732 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 733 | if (location->offset == (u64)-1) { |
db94535d | 734 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
735 | location->objectid, root); |
736 | if (ret) { | |
0f7d52f4 CM |
737 | kfree(root); |
738 | return ERR_PTR(ret); | |
739 | } | |
740 | goto insert; | |
741 | } | |
742 | ||
db94535d | 743 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
744 | tree_root->sectorsize, tree_root->stripesize, |
745 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
746 | |
747 | path = btrfs_alloc_path(); | |
748 | BUG_ON(!path); | |
749 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
750 | if (ret != 0) { | |
0f7d52f4 CM |
751 | if (ret > 0) |
752 | ret = -ENOENT; | |
753 | goto out; | |
754 | } | |
5f39d397 CM |
755 | l = path->nodes[0]; |
756 | read_extent_buffer(l, &root->root_item, | |
757 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 758 | sizeof(root->root_item)); |
44b36eb2 | 759 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
760 | ret = 0; |
761 | out: | |
762 | btrfs_release_path(root, path); | |
763 | btrfs_free_path(path); | |
764 | if (ret) { | |
765 | kfree(root); | |
766 | return ERR_PTR(ret); | |
767 | } | |
db94535d CM |
768 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
769 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 770 | blocksize, 0); |
0f7d52f4 CM |
771 | BUG_ON(!root->node); |
772 | insert: | |
0f7d52f4 | 773 | root->ref_cows = 1; |
5eda7b5e CM |
774 | ret = btrfs_find_highest_inode(root, &highest_inode); |
775 | if (ret == 0) { | |
776 | root->highest_inode = highest_inode; | |
777 | root->last_inode_alloc = highest_inode; | |
778 | } | |
779 | return root; | |
780 | } | |
781 | ||
dc17ff8f CM |
782 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
783 | u64 root_objectid) | |
784 | { | |
785 | struct btrfs_root *root; | |
786 | ||
787 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
788 | return fs_info->tree_root; | |
789 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
790 | return fs_info->extent_root; | |
791 | ||
792 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
793 | (unsigned long)root_objectid); | |
794 | return root; | |
795 | } | |
796 | ||
edbd8d4e CM |
797 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
798 | struct btrfs_key *location) | |
5eda7b5e CM |
799 | { |
800 | struct btrfs_root *root; | |
801 | int ret; | |
802 | ||
edbd8d4e CM |
803 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
804 | return fs_info->tree_root; | |
805 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
806 | return fs_info->extent_root; | |
8f18cf13 CM |
807 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
808 | return fs_info->chunk_root; | |
809 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
810 | return fs_info->dev_root; | |
edbd8d4e | 811 | |
5eda7b5e CM |
812 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
813 | (unsigned long)location->objectid); | |
814 | if (root) | |
815 | return root; | |
816 | ||
817 | root = btrfs_read_fs_root_no_radix(fs_info, location); | |
818 | if (IS_ERR(root)) | |
819 | return root; | |
2619ba1f CM |
820 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
821 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
822 | root); |
823 | if (ret) { | |
5f39d397 | 824 | free_extent_buffer(root->node); |
0f7d52f4 CM |
825 | kfree(root); |
826 | return ERR_PTR(ret); | |
827 | } | |
edbd8d4e CM |
828 | ret = btrfs_find_dead_roots(fs_info->tree_root, |
829 | root->root_key.objectid, root); | |
830 | BUG_ON(ret); | |
831 | ||
832 | return root; | |
833 | } | |
834 | ||
835 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
836 | struct btrfs_key *location, | |
837 | const char *name, int namelen) | |
838 | { | |
839 | struct btrfs_root *root; | |
840 | int ret; | |
841 | ||
842 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
843 | if (!root) | |
844 | return NULL; | |
58176a96 | 845 | |
4313b399 CM |
846 | if (root->in_sysfs) |
847 | return root; | |
848 | ||
58176a96 JB |
849 | ret = btrfs_set_root_name(root, name, namelen); |
850 | if (ret) { | |
5f39d397 | 851 | free_extent_buffer(root->node); |
58176a96 JB |
852 | kfree(root); |
853 | return ERR_PTR(ret); | |
854 | } | |
855 | ||
856 | ret = btrfs_sysfs_add_root(root); | |
857 | if (ret) { | |
5f39d397 | 858 | free_extent_buffer(root->node); |
58176a96 JB |
859 | kfree(root->name); |
860 | kfree(root); | |
861 | return ERR_PTR(ret); | |
862 | } | |
4313b399 | 863 | root->in_sysfs = 1; |
0f7d52f4 CM |
864 | return root; |
865 | } | |
19c00ddc CM |
866 | #if 0 |
867 | static int add_hasher(struct btrfs_fs_info *info, char *type) { | |
868 | struct btrfs_hasher *hasher; | |
869 | ||
870 | hasher = kmalloc(sizeof(*hasher), GFP_NOFS); | |
871 | if (!hasher) | |
872 | return -ENOMEM; | |
873 | hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); | |
874 | if (!hasher->hash_tfm) { | |
875 | kfree(hasher); | |
876 | return -EINVAL; | |
877 | } | |
878 | spin_lock(&info->hash_lock); | |
879 | list_add(&hasher->list, &info->hashers); | |
880 | spin_unlock(&info->hash_lock); | |
881 | return 0; | |
882 | } | |
883 | #endif | |
04160088 CM |
884 | |
885 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
886 | { | |
887 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
888 | int ret = 0; | |
889 | struct list_head *cur; | |
890 | struct btrfs_device *device; | |
891 | struct backing_dev_info *bdi; | |
892 | ||
893 | list_for_each(cur, &info->fs_devices->devices) { | |
894 | device = list_entry(cur, struct btrfs_device, dev_list); | |
895 | bdi = blk_get_backing_dev_info(device->bdev); | |
896 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
897 | ret = 1; | |
898 | break; | |
899 | } | |
900 | } | |
901 | return ret; | |
902 | } | |
903 | ||
38b66988 CM |
904 | /* |
905 | * this unplugs every device on the box, and it is only used when page | |
906 | * is null | |
907 | */ | |
908 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
909 | { | |
910 | struct list_head *cur; | |
911 | struct btrfs_device *device; | |
912 | struct btrfs_fs_info *info; | |
913 | ||
914 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
915 | list_for_each(cur, &info->fs_devices->devices) { | |
916 | device = list_entry(cur, struct btrfs_device, dev_list); | |
917 | bdi = blk_get_backing_dev_info(device->bdev); | |
918 | if (bdi->unplug_io_fn) { | |
919 | bdi->unplug_io_fn(bdi, page); | |
920 | } | |
921 | } | |
922 | } | |
923 | ||
04160088 CM |
924 | void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
925 | { | |
38b66988 | 926 | struct inode *inode; |
f2d8d74d CM |
927 | struct extent_map_tree *em_tree; |
928 | struct extent_map *em; | |
bcbfce8a | 929 | struct address_space *mapping; |
38b66988 CM |
930 | u64 offset; |
931 | ||
bcbfce8a | 932 | /* the generic O_DIRECT read code does this */ |
38b66988 CM |
933 | if (!page) { |
934 | __unplug_io_fn(bdi, page); | |
935 | return; | |
936 | } | |
937 | ||
bcbfce8a CM |
938 | /* |
939 | * page->mapping may change at any time. Get a consistent copy | |
940 | * and use that for everything below | |
941 | */ | |
942 | smp_mb(); | |
943 | mapping = page->mapping; | |
944 | if (!mapping) | |
945 | return; | |
946 | ||
947 | inode = mapping->host; | |
38b66988 | 948 | offset = page_offset(page); |
04160088 | 949 | |
f2d8d74d CM |
950 | em_tree = &BTRFS_I(inode)->extent_tree; |
951 | spin_lock(&em_tree->lock); | |
952 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
953 | spin_unlock(&em_tree->lock); | |
954 | if (!em) | |
955 | return; | |
956 | ||
957 | offset = offset - em->start; | |
958 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
959 | em->block_start + offset, page); | |
960 | free_extent_map(em); | |
04160088 CM |
961 | } |
962 | ||
963 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
964 | { | |
b248a415 | 965 | #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23) |
04160088 | 966 | bdi_init(bdi); |
b248a415 | 967 | #endif |
4575c9cc | 968 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
969 | bdi->state = 0; |
970 | bdi->capabilities = default_backing_dev_info.capabilities; | |
971 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
972 | bdi->unplug_io_data = info; | |
973 | bdi->congested_fn = btrfs_congested_fn; | |
974 | bdi->congested_data = info; | |
975 | return 0; | |
976 | } | |
977 | ||
ce9adaa5 CM |
978 | static int bio_ready_for_csum(struct bio *bio) |
979 | { | |
980 | u64 length = 0; | |
981 | u64 buf_len = 0; | |
982 | u64 start = 0; | |
983 | struct page *page; | |
984 | struct extent_io_tree *io_tree = NULL; | |
985 | struct btrfs_fs_info *info = NULL; | |
986 | struct bio_vec *bvec; | |
987 | int i; | |
988 | int ret; | |
989 | ||
990 | bio_for_each_segment(bvec, bio, i) { | |
991 | page = bvec->bv_page; | |
992 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
993 | length += bvec->bv_len; | |
994 | continue; | |
995 | } | |
996 | if (!page->private) { | |
997 | length += bvec->bv_len; | |
998 | continue; | |
999 | } | |
1000 | length = bvec->bv_len; | |
1001 | buf_len = page->private >> 2; | |
1002 | start = page_offset(page) + bvec->bv_offset; | |
1003 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1004 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1005 | } | |
1006 | /* are we fully contained in this bio? */ | |
1007 | if (buf_len <= length) | |
1008 | return 1; | |
1009 | ||
1010 | ret = extent_range_uptodate(io_tree, start + length, | |
1011 | start + buf_len - 1); | |
1012 | if (ret == 1) | |
1013 | return ret; | |
1014 | return ret; | |
1015 | } | |
1016 | ||
1017 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) | |
44b8bd7e | 1018 | static void btrfs_end_io_csum(void *p) |
ce9adaa5 | 1019 | #else |
44b8bd7e | 1020 | static void btrfs_end_io_csum(struct work_struct *work) |
ce9adaa5 CM |
1021 | #endif |
1022 | { | |
1023 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) | |
1024 | struct btrfs_fs_info *fs_info = p; | |
1025 | #else | |
1026 | struct btrfs_fs_info *fs_info = container_of(work, | |
1027 | struct btrfs_fs_info, | |
1028 | end_io_work); | |
1029 | #endif | |
1030 | unsigned long flags; | |
1031 | struct end_io_wq *end_io_wq; | |
1032 | struct bio *bio; | |
1033 | struct list_head *next; | |
1034 | int error; | |
1035 | int was_empty; | |
1036 | ||
1037 | while(1) { | |
1038 | spin_lock_irqsave(&fs_info->end_io_work_lock, flags); | |
1039 | if (list_empty(&fs_info->end_io_work_list)) { | |
1040 | spin_unlock_irqrestore(&fs_info->end_io_work_lock, | |
1041 | flags); | |
1042 | return; | |
1043 | } | |
1044 | next = fs_info->end_io_work_list.next; | |
1045 | list_del(next); | |
1046 | spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags); | |
1047 | ||
1048 | end_io_wq = list_entry(next, struct end_io_wq, list); | |
1049 | ||
1050 | bio = end_io_wq->bio; | |
22c59948 | 1051 | if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { |
ce9adaa5 CM |
1052 | spin_lock_irqsave(&fs_info->end_io_work_lock, flags); |
1053 | was_empty = list_empty(&fs_info->end_io_work_list); | |
1054 | list_add_tail(&end_io_wq->list, | |
1055 | &fs_info->end_io_work_list); | |
1056 | spin_unlock_irqrestore(&fs_info->end_io_work_lock, | |
1057 | flags); | |
1058 | if (was_empty) | |
1059 | return; | |
1060 | continue; | |
1061 | } | |
1062 | error = end_io_wq->error; | |
1063 | bio->bi_private = end_io_wq->private; | |
1064 | bio->bi_end_io = end_io_wq->end_io; | |
1065 | kfree(end_io_wq); | |
73f61b2a | 1066 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) |
b248a415 | 1067 | bio_endio(bio, bio->bi_size, error); |
73f61b2a | 1068 | #else |
ce9adaa5 | 1069 | bio_endio(bio, error); |
73f61b2a | 1070 | #endif |
ce9adaa5 CM |
1071 | } |
1072 | } | |
1073 | ||
44b8bd7e CM |
1074 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
1075 | static void btrfs_async_submit_work(void *p) | |
1076 | #else | |
1077 | static void btrfs_async_submit_work(struct work_struct *work) | |
1078 | #endif | |
1079 | { | |
1080 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) | |
1081 | struct btrfs_fs_info *fs_info = p; | |
1082 | #else | |
1083 | struct btrfs_fs_info *fs_info = container_of(work, | |
1084 | struct btrfs_fs_info, | |
1085 | async_submit_work); | |
1086 | #endif | |
1087 | struct async_submit_bio *async; | |
1088 | struct list_head *next; | |
1089 | ||
1090 | while(1) { | |
1091 | spin_lock(&fs_info->async_submit_work_lock); | |
1092 | if (list_empty(&fs_info->async_submit_work_list)) { | |
1093 | spin_unlock(&fs_info->async_submit_work_lock); | |
1094 | return; | |
1095 | } | |
1096 | next = fs_info->async_submit_work_list.next; | |
1097 | list_del(next); | |
1098 | spin_unlock(&fs_info->async_submit_work_lock); | |
1099 | ||
1100 | async = list_entry(next, struct async_submit_bio, list); | |
1101 | async->submit_bio_hook(async->inode, async->rw, async->bio, | |
1102 | async->mirror_num); | |
1103 | kfree(async); | |
1104 | } | |
1105 | } | |
1106 | ||
8a4b83cc CM |
1107 | struct btrfs_root *open_ctree(struct super_block *sb, |
1108 | struct btrfs_fs_devices *fs_devices) | |
2e635a27 | 1109 | { |
db94535d CM |
1110 | u32 sectorsize; |
1111 | u32 nodesize; | |
1112 | u32 leafsize; | |
1113 | u32 blocksize; | |
87ee04eb | 1114 | u32 stripesize; |
a061fc8d | 1115 | struct buffer_head *bh; |
e20d96d6 CM |
1116 | struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root), |
1117 | GFP_NOFS); | |
1118 | struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root), | |
1119 | GFP_NOFS); | |
8790d502 | 1120 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1121 | GFP_NOFS); |
0b86a832 CM |
1122 | struct btrfs_root *chunk_root = kmalloc(sizeof(struct btrfs_root), |
1123 | GFP_NOFS); | |
1124 | struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root), | |
1125 | GFP_NOFS); | |
eb60ceac | 1126 | int ret; |
e58ca020 | 1127 | int err = -EINVAL; |
2c90e5d6 | 1128 | struct btrfs_super_block *disk_super; |
8790d502 | 1129 | |
39279cc3 CM |
1130 | if (!extent_root || !tree_root || !fs_info) { |
1131 | err = -ENOMEM; | |
1132 | goto fail; | |
1133 | } | |
ce9adaa5 CM |
1134 | end_io_workqueue = create_workqueue("btrfs-end-io"); |
1135 | BUG_ON(!end_io_workqueue); | |
44b8bd7e | 1136 | async_submit_workqueue = create_workqueue("btrfs-async-submit"); |
ce9adaa5 | 1137 | |
0f7d52f4 | 1138 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1139 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1140 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1141 | INIT_LIST_HEAD(&fs_info->hashers); |
ce9adaa5 | 1142 | INIT_LIST_HEAD(&fs_info->end_io_work_list); |
44b8bd7e | 1143 | INIT_LIST_HEAD(&fs_info->async_submit_work_list); |
19c00ddc | 1144 | spin_lock_init(&fs_info->hash_lock); |
ce9adaa5 | 1145 | spin_lock_init(&fs_info->end_io_work_lock); |
44b8bd7e | 1146 | spin_lock_init(&fs_info->async_submit_work_lock); |
1832a6d5 | 1147 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1148 | spin_lock_init(&fs_info->new_trans_lock); |
19c00ddc | 1149 | |
58176a96 | 1150 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1151 | fs_info->tree_root = tree_root; |
1152 | fs_info->extent_root = extent_root; | |
0b86a832 CM |
1153 | fs_info->chunk_root = chunk_root; |
1154 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1155 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1156 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1157 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1158 | btrfs_mapping_init(&fs_info->mapping_tree); |
e20d96d6 | 1159 | fs_info->sb = sb; |
c59f8951 | 1160 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1161 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1162 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1163 | fs_info->btree_inode = new_inode(sb); |
1164 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1165 | fs_info->btree_inode->i_nlink = 1; |
0afbaf8c | 1166 | |
a061fc8d CM |
1167 | sb->s_blocksize = 4096; |
1168 | sb->s_blocksize_bits = blksize_bits(4096); | |
1169 | ||
0afbaf8c CM |
1170 | /* |
1171 | * we set the i_size on the btree inode to the max possible int. | |
1172 | * the real end of the address space is determined by all of | |
1173 | * the devices in the system | |
1174 | */ | |
1175 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1176 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1177 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1178 | ||
d1310b2e | 1179 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1180 | fs_info->btree_inode->i_mapping, |
1181 | GFP_NOFS); | |
d1310b2e CM |
1182 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1183 | GFP_NOFS); | |
1184 | ||
1185 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1186 | |
d1310b2e | 1187 | extent_io_tree_init(&fs_info->free_space_cache, |
f510cfec | 1188 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1189 | extent_io_tree_init(&fs_info->block_group_cache, |
96b5179d | 1190 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1191 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1192 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1193 | extent_io_tree_init(&fs_info->pending_del, |
1a5bc167 | 1194 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1195 | extent_io_tree_init(&fs_info->extent_ins, |
1a5bc167 | 1196 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1197 | fs_info->do_barriers = 1; |
e18e4809 | 1198 | |
6da6abae | 1199 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) |
b248a415 | 1200 | INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum, fs_info); |
44b8bd7e CM |
1201 | INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work, |
1202 | fs_info); | |
6da6abae CM |
1203 | INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info); |
1204 | #else | |
b248a415 | 1205 | INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum); |
44b8bd7e | 1206 | INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work); |
08607c1b | 1207 | INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner); |
6da6abae | 1208 | #endif |
0f7d52f4 CM |
1209 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1210 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1211 | sizeof(struct btrfs_key)); | |
22b0ebda | 1212 | insert_inode_hash(fs_info->btree_inode); |
d98237b3 | 1213 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); |
39279cc3 | 1214 | |
79154b1b | 1215 | mutex_init(&fs_info->trans_mutex); |
d561c025 | 1216 | mutex_init(&fs_info->fs_mutex); |
3768f368 | 1217 | |
19c00ddc CM |
1218 | #if 0 |
1219 | ret = add_hasher(fs_info, "crc32c"); | |
1220 | if (ret) { | |
1221 | printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); | |
1222 | err = -ENOMEM; | |
1223 | goto fail_iput; | |
1224 | } | |
1225 | #endif | |
0b86a832 | 1226 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1227 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1228 | |
d98237b3 | 1229 | |
a061fc8d CM |
1230 | bh = __bread(fs_devices->latest_bdev, |
1231 | BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
1232 | if (!bh) | |
39279cc3 | 1233 | goto fail_iput; |
39279cc3 | 1234 | |
a061fc8d CM |
1235 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
1236 | brelse(bh); | |
5f39d397 | 1237 | |
a061fc8d | 1238 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1239 | |
5f39d397 | 1240 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1241 | if (!btrfs_super_root(disk_super)) |
39279cc3 | 1242 | goto fail_sb_buffer; |
0f7d52f4 | 1243 | |
8a4b83cc CM |
1244 | if (btrfs_super_num_devices(disk_super) != fs_devices->num_devices) { |
1245 | printk("Btrfs: wanted %llu devices, but found %llu\n", | |
1246 | (unsigned long long)btrfs_super_num_devices(disk_super), | |
1247 | (unsigned long long)fs_devices->num_devices); | |
1248 | goto fail_sb_buffer; | |
1249 | } | |
4575c9cc CM |
1250 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
1251 | ||
db94535d CM |
1252 | nodesize = btrfs_super_nodesize(disk_super); |
1253 | leafsize = btrfs_super_leafsize(disk_super); | |
1254 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1255 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1256 | tree_root->nodesize = nodesize; |
1257 | tree_root->leafsize = leafsize; | |
1258 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1259 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1260 | |
1261 | sb->s_blocksize = sectorsize; | |
1262 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1263 | |
39279cc3 CM |
1264 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1265 | sizeof(disk_super->magic))) { | |
1266 | printk("btrfs: valid FS not found on %s\n", sb->s_id); | |
1267 | goto fail_sb_buffer; | |
1268 | } | |
19c00ddc | 1269 | |
0b86a832 | 1270 | mutex_lock(&fs_info->fs_mutex); |
0d81ba5d | 1271 | |
0b86a832 | 1272 | ret = btrfs_read_sys_array(tree_root); |
84eed90f CM |
1273 | if (ret) { |
1274 | printk("btrfs: failed to read the system array on %s\n", | |
1275 | sb->s_id); | |
1276 | goto fail_sys_array; | |
1277 | } | |
0b86a832 CM |
1278 | |
1279 | blocksize = btrfs_level_size(tree_root, | |
1280 | btrfs_super_chunk_root_level(disk_super)); | |
1281 | ||
1282 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1283 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1284 | ||
1285 | chunk_root->node = read_tree_block(chunk_root, | |
1286 | btrfs_super_chunk_root(disk_super), | |
ca7a79ad | 1287 | blocksize, 0); |
0b86a832 CM |
1288 | BUG_ON(!chunk_root->node); |
1289 | ||
e17cade2 CM |
1290 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
1291 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), | |
1292 | BTRFS_UUID_SIZE); | |
1293 | ||
0b86a832 CM |
1294 | ret = btrfs_read_chunk_tree(chunk_root); |
1295 | BUG_ON(ret); | |
1296 | ||
db94535d CM |
1297 | blocksize = btrfs_level_size(tree_root, |
1298 | btrfs_super_root_level(disk_super)); | |
19c00ddc | 1299 | |
0b86a832 | 1300 | |
e20d96d6 | 1301 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1302 | btrfs_super_root(disk_super), |
ca7a79ad | 1303 | blocksize, 0); |
39279cc3 CM |
1304 | if (!tree_root->node) |
1305 | goto fail_sb_buffer; | |
3768f368 | 1306 | |
db94535d CM |
1307 | |
1308 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1309 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1310 | if (ret) |
39279cc3 | 1311 | goto fail_tree_root; |
0b86a832 CM |
1312 | extent_root->track_dirty = 1; |
1313 | ||
1314 | ret = find_and_setup_root(tree_root, fs_info, | |
1315 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1316 | dev_root->track_dirty = 1; | |
1317 | ||
1318 | if (ret) | |
1319 | goto fail_extent_root; | |
3768f368 | 1320 | |
9078a3e1 CM |
1321 | btrfs_read_block_groups(extent_root); |
1322 | ||
0f7d52f4 | 1323 | fs_info->generation = btrfs_super_generation(disk_super) + 1; |
d18a2c44 CM |
1324 | fs_info->data_alloc_profile = (u64)-1; |
1325 | fs_info->metadata_alloc_profile = (u64)-1; | |
1326 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
1327 | ||
5be6f7f1 | 1328 | mutex_unlock(&fs_info->fs_mutex); |
0f7d52f4 | 1329 | return tree_root; |
39279cc3 | 1330 | |
0b86a832 CM |
1331 | fail_extent_root: |
1332 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1333 | fail_tree_root: |
5f39d397 | 1334 | free_extent_buffer(tree_root->node); |
84eed90f CM |
1335 | fail_sys_array: |
1336 | mutex_unlock(&fs_info->fs_mutex); | |
39279cc3 | 1337 | fail_sb_buffer: |
2d2ae547 | 1338 | extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); |
39279cc3 CM |
1339 | fail_iput: |
1340 | iput(fs_info->btree_inode); | |
1341 | fail: | |
8a4b83cc | 1342 | close_all_devices(fs_info); |
84eed90f CM |
1343 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
1344 | ||
39279cc3 CM |
1345 | kfree(extent_root); |
1346 | kfree(tree_root); | |
b248a415 | 1347 | #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23) |
2d2ae547 | 1348 | bdi_destroy(&fs_info->bdi); |
b248a415 | 1349 | #endif |
39279cc3 CM |
1350 | kfree(fs_info); |
1351 | return ERR_PTR(err); | |
eb60ceac CM |
1352 | } |
1353 | ||
f2984462 CM |
1354 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1355 | { | |
1356 | char b[BDEVNAME_SIZE]; | |
1357 | ||
1358 | if (uptodate) { | |
1359 | set_buffer_uptodate(bh); | |
1360 | } else { | |
1361 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1362 | printk(KERN_WARNING "lost page write due to " | |
1363 | "I/O error on %s\n", | |
1364 | bdevname(bh->b_bdev, b)); | |
1365 | } | |
1366 | set_buffer_write_io_error(bh); | |
1367 | clear_buffer_uptodate(bh); | |
1368 | } | |
1369 | unlock_buffer(bh); | |
1370 | put_bh(bh); | |
1371 | } | |
1372 | ||
1373 | int write_all_supers(struct btrfs_root *root) | |
1374 | { | |
1375 | struct list_head *cur; | |
1376 | struct list_head *head = &root->fs_info->fs_devices->devices; | |
1377 | struct btrfs_device *dev; | |
a061fc8d | 1378 | struct btrfs_super_block *sb; |
f2984462 CM |
1379 | struct btrfs_dev_item *dev_item; |
1380 | struct buffer_head *bh; | |
1381 | int ret; | |
1382 | int do_barriers; | |
a236aed1 CM |
1383 | int max_errors; |
1384 | int total_errors = 0; | |
a061fc8d CM |
1385 | u32 crc; |
1386 | u64 flags; | |
f2984462 | 1387 | |
a236aed1 | 1388 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
1389 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
1390 | ||
a061fc8d CM |
1391 | sb = &root->fs_info->super_for_commit; |
1392 | dev_item = &sb->dev_item; | |
f2984462 CM |
1393 | list_for_each(cur, head) { |
1394 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
a061fc8d CM |
1395 | btrfs_set_stack_device_type(dev_item, dev->type); |
1396 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
1397 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
1398 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
1399 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
1400 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
1401 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
1402 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
1403 | flags = btrfs_super_flags(sb); | |
1404 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
1405 | ||
1406 | ||
1407 | crc = ~(u32)0; | |
1408 | crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc, | |
1409 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
1410 | btrfs_csum_final(crc, sb->csum); | |
1411 | ||
1412 | bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096, | |
f2984462 CM |
1413 | BTRFS_SUPER_INFO_SIZE); |
1414 | ||
a061fc8d | 1415 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
f2984462 CM |
1416 | dev->pending_io = bh; |
1417 | ||
1418 | get_bh(bh); | |
1419 | set_buffer_uptodate(bh); | |
1420 | lock_buffer(bh); | |
1421 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
1422 | ||
1423 | if (do_barriers && dev->barriers) { | |
1424 | ret = submit_bh(WRITE_BARRIER, bh); | |
1425 | if (ret == -EOPNOTSUPP) { | |
1426 | printk("btrfs: disabling barriers on dev %s\n", | |
1427 | dev->name); | |
1428 | set_buffer_uptodate(bh); | |
1429 | dev->barriers = 0; | |
1430 | get_bh(bh); | |
1431 | lock_buffer(bh); | |
1432 | ret = submit_bh(WRITE, bh); | |
1433 | } | |
1434 | } else { | |
1435 | ret = submit_bh(WRITE, bh); | |
1436 | } | |
a236aed1 CM |
1437 | if (ret) |
1438 | total_errors++; | |
f2984462 | 1439 | } |
a236aed1 CM |
1440 | if (total_errors > max_errors) { |
1441 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1442 | BUG(); | |
1443 | } | |
1444 | total_errors = 0; | |
f2984462 CM |
1445 | |
1446 | list_for_each(cur, head) { | |
1447 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
1448 | BUG_ON(!dev->pending_io); | |
1449 | bh = dev->pending_io; | |
1450 | wait_on_buffer(bh); | |
1451 | if (!buffer_uptodate(dev->pending_io)) { | |
1452 | if (do_barriers && dev->barriers) { | |
1453 | printk("btrfs: disabling barriers on dev %s\n", | |
1454 | dev->name); | |
1455 | set_buffer_uptodate(bh); | |
1456 | get_bh(bh); | |
1457 | lock_buffer(bh); | |
1458 | dev->barriers = 0; | |
1459 | ret = submit_bh(WRITE, bh); | |
1460 | BUG_ON(ret); | |
1461 | wait_on_buffer(bh); | |
1462 | BUG_ON(!buffer_uptodate(bh)); | |
1463 | } else { | |
a236aed1 | 1464 | total_errors++; |
f2984462 CM |
1465 | } |
1466 | ||
1467 | } | |
1468 | dev->pending_io = NULL; | |
1469 | brelse(bh); | |
1470 | } | |
a236aed1 CM |
1471 | if (total_errors > max_errors) { |
1472 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1473 | BUG(); | |
1474 | } | |
f2984462 CM |
1475 | return 0; |
1476 | } | |
1477 | ||
e089f05c | 1478 | int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root |
79154b1b | 1479 | *root) |
eb60ceac | 1480 | { |
e66f709b | 1481 | int ret; |
5f39d397 | 1482 | |
f2984462 | 1483 | ret = write_all_supers(root); |
5f39d397 | 1484 | return ret; |
cfaa7295 CM |
1485 | } |
1486 | ||
5eda7b5e | 1487 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
1488 | { |
1489 | radix_tree_delete(&fs_info->fs_roots_radix, | |
1490 | (unsigned long)root->root_key.objectid); | |
b99aa6cb CM |
1491 | if (root->in_sysfs) |
1492 | btrfs_sysfs_del_root(root); | |
2619ba1f CM |
1493 | if (root->inode) |
1494 | iput(root->inode); | |
1495 | if (root->node) | |
5f39d397 | 1496 | free_extent_buffer(root->node); |
2619ba1f | 1497 | if (root->commit_root) |
5f39d397 | 1498 | free_extent_buffer(root->commit_root); |
58176a96 JB |
1499 | if (root->name) |
1500 | kfree(root->name); | |
2619ba1f CM |
1501 | kfree(root); |
1502 | return 0; | |
1503 | } | |
1504 | ||
35b7e476 | 1505 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
1506 | { |
1507 | int ret; | |
1508 | struct btrfs_root *gang[8]; | |
1509 | int i; | |
1510 | ||
1511 | while(1) { | |
1512 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1513 | (void **)gang, 0, | |
1514 | ARRAY_SIZE(gang)); | |
1515 | if (!ret) | |
1516 | break; | |
2619ba1f | 1517 | for (i = 0; i < ret; i++) |
5eda7b5e | 1518 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
1519 | } |
1520 | return 0; | |
1521 | } | |
b4100d64 | 1522 | |
e20d96d6 | 1523 | int close_ctree(struct btrfs_root *root) |
cfaa7295 | 1524 | { |
3768f368 | 1525 | int ret; |
e089f05c | 1526 | struct btrfs_trans_handle *trans; |
0f7d52f4 | 1527 | struct btrfs_fs_info *fs_info = root->fs_info; |
e089f05c | 1528 | |
facda1e7 | 1529 | fs_info->closing = 1; |
08607c1b | 1530 | btrfs_transaction_flush_work(root); |
0f7d52f4 | 1531 | mutex_lock(&fs_info->fs_mutex); |
6702ed49 | 1532 | btrfs_defrag_dirty_roots(root->fs_info); |
79154b1b | 1533 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 1534 | ret = btrfs_commit_transaction(trans, root); |
79154b1b CM |
1535 | /* run commit again to drop the original snapshot */ |
1536 | trans = btrfs_start_transaction(root, 1); | |
1537 | btrfs_commit_transaction(trans, root); | |
1538 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 1539 | BUG_ON(ret); |
d6bfde87 | 1540 | |
79154b1b | 1541 | write_ctree_super(NULL, root); |
0f7d52f4 CM |
1542 | mutex_unlock(&fs_info->fs_mutex); |
1543 | ||
9ad6b7bc CM |
1544 | btrfs_transaction_flush_work(root); |
1545 | ||
b0c68f8b CM |
1546 | if (fs_info->delalloc_bytes) { |
1547 | printk("btrfs: at unmount delalloc count %Lu\n", | |
1548 | fs_info->delalloc_bytes); | |
1549 | } | |
0f7d52f4 | 1550 | if (fs_info->extent_root->node) |
5f39d397 | 1551 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 1552 | |
0f7d52f4 | 1553 | if (fs_info->tree_root->node) |
5f39d397 | 1554 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 1555 | |
0b86a832 CM |
1556 | if (root->fs_info->chunk_root->node); |
1557 | free_extent_buffer(root->fs_info->chunk_root->node); | |
1558 | ||
1559 | if (root->fs_info->dev_root->node); | |
1560 | free_extent_buffer(root->fs_info->dev_root->node); | |
1561 | ||
9078a3e1 | 1562 | btrfs_free_block_groups(root->fs_info); |
0f7d52f4 | 1563 | del_fs_roots(fs_info); |
d10c5f31 CM |
1564 | |
1565 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1566 | ||
d1310b2e CM |
1567 | extent_io_tree_empty_lru(&fs_info->free_space_cache); |
1568 | extent_io_tree_empty_lru(&fs_info->block_group_cache); | |
1569 | extent_io_tree_empty_lru(&fs_info->pinned_extents); | |
1570 | extent_io_tree_empty_lru(&fs_info->pending_del); | |
1571 | extent_io_tree_empty_lru(&fs_info->extent_ins); | |
1572 | extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree); | |
d10c5f31 | 1573 | |
9ad6b7bc | 1574 | flush_workqueue(async_submit_workqueue); |
d6bfde87 | 1575 | flush_workqueue(end_io_workqueue); |
9ad6b7bc | 1576 | |
db94535d | 1577 | truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); |
9ad6b7bc | 1578 | |
44b8bd7e CM |
1579 | flush_workqueue(async_submit_workqueue); |
1580 | destroy_workqueue(async_submit_workqueue); | |
1581 | ||
d6bfde87 CM |
1582 | flush_workqueue(end_io_workqueue); |
1583 | destroy_workqueue(end_io_workqueue); | |
1584 | ||
db94535d | 1585 | iput(fs_info->btree_inode); |
19c00ddc CM |
1586 | #if 0 |
1587 | while(!list_empty(&fs_info->hashers)) { | |
1588 | struct btrfs_hasher *hasher; | |
1589 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
1590 | hashers); | |
1591 | list_del(&hasher->hashers); | |
1592 | crypto_free_hash(&fs_info->hash_tfm); | |
1593 | kfree(hasher); | |
1594 | } | |
1595 | #endif | |
0b86a832 CM |
1596 | close_all_devices(fs_info); |
1597 | btrfs_mapping_tree_free(&fs_info->mapping_tree); | |
b248a415 CM |
1598 | |
1599 | #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23) | |
04160088 | 1600 | bdi_destroy(&fs_info->bdi); |
b248a415 | 1601 | #endif |
0b86a832 | 1602 | |
0f7d52f4 | 1603 | kfree(fs_info->extent_root); |
0f7d52f4 | 1604 | kfree(fs_info->tree_root); |
0b86a832 CM |
1605 | kfree(fs_info->chunk_root); |
1606 | kfree(fs_info->dev_root); | |
eb60ceac CM |
1607 | return 0; |
1608 | } | |
1609 | ||
5f39d397 CM |
1610 | int btrfs_buffer_uptodate(struct extent_buffer *buf) |
1611 | { | |
810191ff | 1612 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 1613 | return extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); |
5f39d397 CM |
1614 | } |
1615 | ||
1616 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 1617 | { |
810191ff | 1618 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 1619 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
1620 | buf); |
1621 | } | |
6702ed49 | 1622 | |
5f39d397 CM |
1623 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
1624 | { | |
810191ff | 1625 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
1626 | u64 transid = btrfs_header_generation(buf); |
1627 | struct inode *btree_inode = root->fs_info->btree_inode; | |
6702ed49 | 1628 | |
ccd467d6 CM |
1629 | if (transid != root->fs_info->generation) { |
1630 | printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", | |
db94535d | 1631 | (unsigned long long)buf->start, |
ccd467d6 CM |
1632 | transid, root->fs_info->generation); |
1633 | WARN_ON(1); | |
1634 | } | |
d1310b2e | 1635 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); |
eb60ceac CM |
1636 | } |
1637 | ||
e2008b61 CM |
1638 | void btrfs_throttle(struct btrfs_root *root) |
1639 | { | |
55c69072 CM |
1640 | struct backing_dev_info *bdi; |
1641 | ||
a061fc8d | 1642 | bdi = &root->fs_info->bdi; |
04005cc7 CM |
1643 | if (root->fs_info->throttles && bdi_write_congested(bdi)) { |
1644 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,18) | |
55c69072 | 1645 | congestion_wait(WRITE, HZ/20); |
04005cc7 CM |
1646 | #else |
1647 | blk_congestion_wait(WRITE, HZ/20); | |
1648 | #endif | |
1649 | } | |
e2008b61 CM |
1650 | } |
1651 | ||
d3c2fdcf | 1652 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 1653 | { |
188de649 CM |
1654 | /* |
1655 | * looks as though older kernels can get into trouble with | |
1656 | * this code, they end up stuck in balance_dirty_pages forever | |
1657 | */ | |
1658 | #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23) | |
d6bfde87 CM |
1659 | struct extent_io_tree *tree; |
1660 | u64 num_dirty; | |
1661 | u64 start = 0; | |
1662 | unsigned long thresh = 16 * 1024 * 1024; | |
1663 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
1664 | ||
1665 | if (current_is_pdflush()) | |
1666 | return; | |
1667 | ||
1668 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
1669 | thresh, EXTENT_DIRTY); | |
1670 | if (num_dirty > thresh) { | |
1671 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 1672 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 1673 | } |
188de649 CM |
1674 | #else |
1675 | return; | |
1676 | #endif | |
35b7e476 | 1677 | } |
6b80053d CM |
1678 | |
1679 | void btrfs_set_buffer_defrag(struct extent_buffer *buf) | |
1680 | { | |
810191ff | 1681 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1682 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1683 | set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start, |
6b80053d CM |
1684 | buf->start + buf->len - 1, EXTENT_DEFRAG, GFP_NOFS); |
1685 | } | |
1686 | ||
1687 | void btrfs_set_buffer_defrag_done(struct extent_buffer *buf) | |
1688 | { | |
810191ff | 1689 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1690 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1691 | set_extent_bits(&BTRFS_I(btree_inode)->io_tree, buf->start, |
6b80053d CM |
1692 | buf->start + buf->len - 1, EXTENT_DEFRAG_DONE, |
1693 | GFP_NOFS); | |
1694 | } | |
1695 | ||
1696 | int btrfs_buffer_defrag(struct extent_buffer *buf) | |
1697 | { | |
810191ff | 1698 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1699 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1700 | return test_range_bit(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1701 | buf->start, buf->start + buf->len - 1, EXTENT_DEFRAG, 0); |
1702 | } | |
1703 | ||
1704 | int btrfs_buffer_defrag_done(struct extent_buffer *buf) | |
1705 | { | |
810191ff | 1706 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1707 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1708 | return test_range_bit(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1709 | buf->start, buf->start + buf->len - 1, |
1710 | EXTENT_DEFRAG_DONE, 0); | |
1711 | } | |
1712 | ||
1713 | int btrfs_clear_buffer_defrag_done(struct extent_buffer *buf) | |
1714 | { | |
810191ff | 1715 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1716 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1717 | return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1718 | buf->start, buf->start + buf->len - 1, |
1719 | EXTENT_DEFRAG_DONE, GFP_NOFS); | |
1720 | } | |
1721 | ||
1722 | int btrfs_clear_buffer_defrag(struct extent_buffer *buf) | |
1723 | { | |
810191ff | 1724 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
6b80053d | 1725 | struct inode *btree_inode = root->fs_info->btree_inode; |
d1310b2e | 1726 | return clear_extent_bits(&BTRFS_I(btree_inode)->io_tree, |
6b80053d CM |
1727 | buf->start, buf->start + buf->len - 1, |
1728 | EXTENT_DEFRAG, GFP_NOFS); | |
1729 | } | |
1730 | ||
ca7a79ad | 1731 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 1732 | { |
810191ff | 1733 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 1734 | int ret; |
ca7a79ad | 1735 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
1736 | if (ret == 0) { |
1737 | buf->flags |= EXTENT_UPTODATE; | |
1738 | } | |
1739 | return ret; | |
6b80053d | 1740 | } |
0da5468f | 1741 | |
d1310b2e | 1742 | static struct extent_io_ops btree_extent_io_ops = { |
0da5468f | 1743 | .writepage_io_hook = btree_writepage_io_hook, |
ce9adaa5 | 1744 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 1745 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
1746 | /* note we're sharing with inode.c for the merge bio hook */ |
1747 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 1748 | }; |