]>
Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
a74a4b97 | 19 | #include <linux/version.h> |
e20d96d6 | 20 | #include <linux/fs.h> |
d98237b3 | 21 | #include <linux/blkdev.h> |
87cbda5c | 22 | #include <linux/scatterlist.h> |
22b0ebda | 23 | #include <linux/swap.h> |
0f7d52f4 | 24 | #include <linux/radix-tree.h> |
35b7e476 | 25 | #include <linux/writeback.h> |
5f39d397 | 26 | #include <linux/buffer_head.h> // for block_sync_page |
ce9adaa5 | 27 | #include <linux/workqueue.h> |
a74a4b97 CM |
28 | #include <linux/kthread.h> |
29 | #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,20) | |
30 | # include <linux/freezer.h> | |
31 | #else | |
32 | # include <linux/sched.h> | |
33 | #endif | |
a5eb62e3 | 34 | #include "crc32c.h" |
eb60ceac CM |
35 | #include "ctree.h" |
36 | #include "disk-io.h" | |
e089f05c | 37 | #include "transaction.h" |
0f7d52f4 | 38 | #include "btrfs_inode.h" |
0b86a832 | 39 | #include "volumes.h" |
db94535d | 40 | #include "print-tree.h" |
8b712842 | 41 | #include "async-thread.h" |
925baedd | 42 | #include "locking.h" |
017e5369 | 43 | #include "ref-cache.h" |
e02119d5 | 44 | #include "tree-log.h" |
eb60ceac | 45 | |
5f39d397 CM |
46 | #if 0 |
47 | static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) | |
7eccb903 | 48 | { |
5f39d397 CM |
49 | if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) { |
50 | printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n", | |
51 | (unsigned long long)extent_buffer_blocknr(buf), | |
52 | (unsigned long long)btrfs_header_blocknr(buf)); | |
39279cc3 | 53 | return 1; |
d98237b3 | 54 | } |
9a8dd150 | 55 | return 0; |
eb60ceac | 56 | } |
5f39d397 | 57 | #endif |
eb60ceac | 58 | |
d1310b2e | 59 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 60 | static void end_workqueue_fn(struct btrfs_work *work); |
ce9adaa5 CM |
61 | |
62 | struct end_io_wq { | |
63 | struct bio *bio; | |
64 | bio_end_io_t *end_io; | |
65 | void *private; | |
66 | struct btrfs_fs_info *info; | |
67 | int error; | |
22c59948 | 68 | int metadata; |
ce9adaa5 | 69 | struct list_head list; |
8b712842 | 70 | struct btrfs_work work; |
ce9adaa5 | 71 | }; |
0da5468f | 72 | |
44b8bd7e CM |
73 | struct async_submit_bio { |
74 | struct inode *inode; | |
75 | struct bio *bio; | |
76 | struct list_head list; | |
77 | extent_submit_bio_hook_t *submit_bio_hook; | |
78 | int rw; | |
79 | int mirror_num; | |
8b712842 | 80 | struct btrfs_work work; |
44b8bd7e CM |
81 | }; |
82 | ||
5f39d397 | 83 | struct extent_map *btree_get_extent(struct inode *inode, struct page *page, |
d1310b2e | 84 | size_t page_offset, u64 start, u64 len, |
5f39d397 | 85 | int create) |
7eccb903 | 86 | { |
5f39d397 CM |
87 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
88 | struct extent_map *em; | |
89 | int ret; | |
90 | ||
d1310b2e CM |
91 | spin_lock(&em_tree->lock); |
92 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
93 | if (em) { |
94 | em->bdev = | |
95 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
96 | spin_unlock(&em_tree->lock); | |
5f39d397 | 97 | goto out; |
a061fc8d CM |
98 | } |
99 | spin_unlock(&em_tree->lock); | |
7b13b7b1 | 100 | |
5f39d397 CM |
101 | em = alloc_extent_map(GFP_NOFS); |
102 | if (!em) { | |
103 | em = ERR_PTR(-ENOMEM); | |
104 | goto out; | |
105 | } | |
106 | em->start = 0; | |
0afbaf8c | 107 | em->len = (u64)-1; |
5f39d397 | 108 | em->block_start = 0; |
a061fc8d | 109 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
110 | |
111 | spin_lock(&em_tree->lock); | |
5f39d397 CM |
112 | ret = add_extent_mapping(em_tree, em); |
113 | if (ret == -EEXIST) { | |
0afbaf8c CM |
114 | u64 failed_start = em->start; |
115 | u64 failed_len = em->len; | |
116 | ||
117 | printk("failed to insert %Lu %Lu -> %Lu into tree\n", | |
118 | em->start, em->len, em->block_start); | |
5f39d397 | 119 | free_extent_map(em); |
7b13b7b1 | 120 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c CM |
121 | if (em) { |
122 | printk("after failing, found %Lu %Lu %Lu\n", | |
123 | em->start, em->len, em->block_start); | |
7b13b7b1 | 124 | ret = 0; |
0afbaf8c CM |
125 | } else { |
126 | em = lookup_extent_mapping(em_tree, failed_start, | |
127 | failed_len); | |
128 | if (em) { | |
129 | printk("double failure lookup gives us " | |
130 | "%Lu %Lu -> %Lu\n", em->start, | |
131 | em->len, em->block_start); | |
132 | free_extent_map(em); | |
133 | } | |
7b13b7b1 | 134 | ret = -EIO; |
0afbaf8c | 135 | } |
5f39d397 | 136 | } else if (ret) { |
7b13b7b1 CM |
137 | free_extent_map(em); |
138 | em = NULL; | |
5f39d397 | 139 | } |
7b13b7b1 CM |
140 | spin_unlock(&em_tree->lock); |
141 | ||
142 | if (ret) | |
143 | em = ERR_PTR(ret); | |
5f39d397 CM |
144 | out: |
145 | return em; | |
7eccb903 CM |
146 | } |
147 | ||
19c00ddc CM |
148 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
149 | { | |
a5eb62e3 | 150 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
151 | } |
152 | ||
153 | void btrfs_csum_final(u32 crc, char *result) | |
154 | { | |
155 | *(__le32 *)result = ~cpu_to_le32(crc); | |
156 | } | |
157 | ||
158 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, | |
159 | int verify) | |
160 | { | |
161 | char result[BTRFS_CRC32_SIZE]; | |
162 | unsigned long len; | |
163 | unsigned long cur_len; | |
164 | unsigned long offset = BTRFS_CSUM_SIZE; | |
165 | char *map_token = NULL; | |
166 | char *kaddr; | |
167 | unsigned long map_start; | |
168 | unsigned long map_len; | |
169 | int err; | |
170 | u32 crc = ~(u32)0; | |
171 | ||
172 | len = buf->len - offset; | |
173 | while(len > 0) { | |
174 | err = map_private_extent_buffer(buf, offset, 32, | |
175 | &map_token, &kaddr, | |
176 | &map_start, &map_len, KM_USER0); | |
177 | if (err) { | |
178 | printk("failed to map extent buffer! %lu\n", | |
179 | offset); | |
180 | return 1; | |
181 | } | |
182 | cur_len = min(len, map_len - (offset - map_start)); | |
183 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
184 | crc, cur_len); | |
185 | len -= cur_len; | |
186 | offset += cur_len; | |
187 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
188 | } | |
189 | btrfs_csum_final(crc, result); | |
190 | ||
191 | if (verify) { | |
e4204ded | 192 | /* FIXME, this is not good */ |
63b10fc4 | 193 | if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) { |
e4204ded CM |
194 | u32 val; |
195 | u32 found = 0; | |
196 | memcpy(&found, result, BTRFS_CRC32_SIZE); | |
197 | ||
198 | read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE); | |
199 | printk("btrfs: %s checksum verify failed on %llu " | |
2dd3e67b | 200 | "wanted %X found %X level %d\n", |
19c00ddc | 201 | root->fs_info->sb->s_id, |
2dd3e67b | 202 | buf->start, val, found, btrfs_header_level(buf)); |
19c00ddc CM |
203 | return 1; |
204 | } | |
205 | } else { | |
206 | write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE); | |
207 | } | |
208 | return 0; | |
209 | } | |
210 | ||
1259ab75 CM |
211 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
212 | struct extent_buffer *eb, u64 parent_transid) | |
213 | { | |
214 | int ret; | |
215 | ||
216 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
217 | return 0; | |
218 | ||
219 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
220 | if (extent_buffer_uptodate(io_tree, eb) && | |
221 | btrfs_header_generation(eb) == parent_transid) { | |
222 | ret = 0; | |
223 | goto out; | |
224 | } | |
225 | printk("parent transid verify failed on %llu wanted %llu found %llu\n", | |
226 | (unsigned long long)eb->start, | |
227 | (unsigned long long)parent_transid, | |
228 | (unsigned long long)btrfs_header_generation(eb)); | |
229 | ret = 1; | |
1259ab75 | 230 | clear_extent_buffer_uptodate(io_tree, eb); |
33958dc6 | 231 | out: |
1259ab75 CM |
232 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
233 | GFP_NOFS); | |
234 | return ret; | |
235 | ||
236 | } | |
237 | ||
f188591e CM |
238 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
239 | struct extent_buffer *eb, | |
ca7a79ad | 240 | u64 start, u64 parent_transid) |
f188591e CM |
241 | { |
242 | struct extent_io_tree *io_tree; | |
243 | int ret; | |
244 | int num_copies = 0; | |
245 | int mirror_num = 0; | |
246 | ||
247 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
248 | while (1) { | |
249 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
250 | btree_get_extent, mirror_num); | |
1259ab75 CM |
251 | if (!ret && |
252 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 253 | return ret; |
a1b32a59 | 254 | printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror_num); |
f188591e CM |
255 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
256 | eb->start, eb->len); | |
4235298e | 257 | if (num_copies == 1) |
f188591e | 258 | return ret; |
4235298e | 259 | |
f188591e | 260 | mirror_num++; |
4235298e | 261 | if (mirror_num > num_copies) |
f188591e | 262 | return ret; |
f188591e | 263 | } |
f188591e CM |
264 | return -EIO; |
265 | } | |
19c00ddc CM |
266 | |
267 | int csum_dirty_buffer(struct btrfs_root *root, struct page *page) | |
268 | { | |
d1310b2e | 269 | struct extent_io_tree *tree; |
35ebb934 | 270 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
271 | u64 found_start; |
272 | int found_level; | |
273 | unsigned long len; | |
274 | struct extent_buffer *eb; | |
f188591e CM |
275 | int ret; |
276 | ||
d1310b2e | 277 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
278 | |
279 | if (page->private == EXTENT_PAGE_PRIVATE) | |
280 | goto out; | |
281 | if (!page->private) | |
282 | goto out; | |
283 | len = page->private >> 2; | |
284 | if (len == 0) { | |
285 | WARN_ON(1); | |
286 | } | |
287 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
ca7a79ad CM |
288 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
289 | btrfs_header_generation(eb)); | |
f188591e | 290 | BUG_ON(ret); |
19c00ddc CM |
291 | found_start = btrfs_header_bytenr(eb); |
292 | if (found_start != start) { | |
293 | printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", | |
294 | start, found_start, len); | |
55c69072 CM |
295 | WARN_ON(1); |
296 | goto err; | |
297 | } | |
298 | if (eb->first_page != page) { | |
299 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
300 | page->index); | |
301 | WARN_ON(1); | |
302 | goto err; | |
303 | } | |
304 | if (!PageUptodate(page)) { | |
305 | printk("csum not up to date page %lu\n", page->index); | |
306 | WARN_ON(1); | |
307 | goto err; | |
19c00ddc CM |
308 | } |
309 | found_level = btrfs_header_level(eb); | |
4bef0848 | 310 | |
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 | 348 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 349 | if (found_start != start) { |
a1b32a59 CM |
350 | printk("bad tree block start %llu %llu\n", |
351 | (unsigned long long)found_start, | |
352 | (unsigned long long)eb->start); | |
f188591e | 353 | ret = -EIO; |
ce9adaa5 CM |
354 | goto err; |
355 | } | |
356 | if (eb->first_page != page) { | |
357 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
358 | page->index); | |
359 | WARN_ON(1); | |
f188591e | 360 | ret = -EIO; |
ce9adaa5 CM |
361 | goto err; |
362 | } | |
1259ab75 CM |
363 | if (memcmp_extent_buffer(eb, root->fs_info->fsid, |
364 | (unsigned long)btrfs_header_fsid(eb), | |
365 | BTRFS_FSID_SIZE)) { | |
366 | printk("bad fsid on block %Lu\n", eb->start); | |
367 | ret = -EIO; | |
368 | goto err; | |
369 | } | |
ce9adaa5 CM |
370 | found_level = btrfs_header_level(eb); |
371 | ||
372 | ret = csum_tree_block(root, eb, 1); | |
f188591e CM |
373 | if (ret) |
374 | ret = -EIO; | |
ce9adaa5 CM |
375 | |
376 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
377 | end = eb->start + end - 1; | |
ce9adaa5 CM |
378 | err: |
379 | free_extent_buffer(eb); | |
380 | out: | |
f188591e | 381 | return ret; |
ce9adaa5 CM |
382 | } |
383 | ||
384 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) | |
385 | static void end_workqueue_bio(struct bio *bio, int err) | |
386 | #else | |
387 | static int end_workqueue_bio(struct bio *bio, | |
388 | unsigned int bytes_done, int err) | |
389 | #endif | |
390 | { | |
391 | struct end_io_wq *end_io_wq = bio->bi_private; | |
392 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 CM |
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; | |
ce9adaa5 | 400 | end_io_wq->error = err; |
8b712842 CM |
401 | end_io_wq->work.func = end_workqueue_fn; |
402 | end_io_wq->work.flags = 0; | |
e6dcd2dc CM |
403 | if (bio->bi_rw & (1 << BIO_RW)) |
404 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
405 | &end_io_wq->work); | |
406 | else | |
407 | btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); | |
ce9adaa5 CM |
408 | |
409 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
410 | return 0; | |
411 | #endif | |
412 | } | |
413 | ||
22c59948 CM |
414 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
415 | int metadata) | |
0b86a832 | 416 | { |
ce9adaa5 | 417 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
418 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
419 | if (!end_io_wq) | |
420 | return -ENOMEM; | |
421 | ||
422 | end_io_wq->private = bio->bi_private; | |
423 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 424 | end_io_wq->info = info; |
ce9adaa5 CM |
425 | end_io_wq->error = 0; |
426 | end_io_wq->bio = bio; | |
22c59948 | 427 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
428 | |
429 | bio->bi_private = end_io_wq; | |
430 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
431 | return 0; |
432 | } | |
433 | ||
b64a2851 | 434 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 435 | { |
4854ddd0 CM |
436 | unsigned long limit = min_t(unsigned long, |
437 | info->workers.max_workers, | |
438 | info->fs_devices->open_devices); | |
439 | return 256 * limit; | |
440 | } | |
0986fe9e | 441 | |
4854ddd0 CM |
442 | int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) |
443 | { | |
b64a2851 CM |
444 | return atomic_read(&info->nr_async_bios) > |
445 | btrfs_async_submit_limit(info); | |
0986fe9e CM |
446 | } |
447 | ||
8b712842 CM |
448 | static void run_one_async_submit(struct btrfs_work *work) |
449 | { | |
450 | struct btrfs_fs_info *fs_info; | |
451 | struct async_submit_bio *async; | |
4854ddd0 | 452 | int limit; |
8b712842 CM |
453 | |
454 | async = container_of(work, struct async_submit_bio, work); | |
455 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 456 | |
b64a2851 | 457 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
458 | limit = limit * 2 / 3; |
459 | ||
8b712842 | 460 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 461 | |
b64a2851 CM |
462 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
463 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
464 | wake_up(&fs_info->async_submit_wait); |
465 | ||
8b712842 CM |
466 | async->submit_bio_hook(async->inode, async->rw, async->bio, |
467 | async->mirror_num); | |
468 | kfree(async); | |
469 | } | |
470 | ||
44b8bd7e CM |
471 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
472 | int rw, struct bio *bio, int mirror_num, | |
473 | extent_submit_bio_hook_t *submit_bio_hook) | |
474 | { | |
475 | struct async_submit_bio *async; | |
b64a2851 | 476 | int limit = btrfs_async_submit_limit(fs_info); |
44b8bd7e CM |
477 | |
478 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
479 | if (!async) | |
480 | return -ENOMEM; | |
481 | ||
482 | async->inode = inode; | |
483 | async->rw = rw; | |
484 | async->bio = bio; | |
485 | async->mirror_num = mirror_num; | |
486 | async->submit_bio_hook = submit_bio_hook; | |
8b712842 CM |
487 | async->work.func = run_one_async_submit; |
488 | async->work.flags = 0; | |
cb03c743 | 489 | atomic_inc(&fs_info->nr_async_submits); |
8b712842 | 490 | btrfs_queue_worker(&fs_info->workers, &async->work); |
4854ddd0 | 491 | |
9473f16c CM |
492 | if (atomic_read(&fs_info->nr_async_submits) > limit) { |
493 | wait_event_timeout(fs_info->async_submit_wait, | |
4854ddd0 CM |
494 | (atomic_read(&fs_info->nr_async_submits) < limit), |
495 | HZ/10); | |
9473f16c CM |
496 | |
497 | wait_event_timeout(fs_info->async_submit_wait, | |
498 | (atomic_read(&fs_info->nr_async_bios) < limit), | |
499 | HZ/10); | |
500 | } | |
44b8bd7e CM |
501 | return 0; |
502 | } | |
503 | ||
504 | static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, | |
f188591e | 505 | int mirror_num) |
22c59948 CM |
506 | { |
507 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
508 | u64 offset; | |
509 | int ret; | |
510 | ||
511 | offset = bio->bi_sector << 9; | |
512 | ||
8b712842 CM |
513 | /* |
514 | * when we're called for a write, we're already in the async | |
5443be45 | 515 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 516 | */ |
22c59948 | 517 | if (rw & (1 << BIO_RW)) { |
8b712842 | 518 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, |
5443be45 | 519 | mirror_num, 1); |
22c59948 CM |
520 | } |
521 | ||
8b712842 CM |
522 | /* |
523 | * called for a read, do the setup so that checksum validation | |
524 | * can happen in the async kernel threads | |
525 | */ | |
22c59948 CM |
526 | ret = btrfs_bio_wq_end_io(root->fs_info, bio, 1); |
527 | BUG_ON(ret); | |
ce9adaa5 | 528 | |
8b712842 | 529 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
530 | } |
531 | ||
44b8bd7e CM |
532 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
533 | int mirror_num) | |
534 | { | |
8b712842 CM |
535 | /* |
536 | * kthread helpers are used to submit writes so that checksumming | |
537 | * can happen in parallel across all CPUs | |
538 | */ | |
44b8bd7e CM |
539 | if (!(rw & (1 << BIO_RW))) { |
540 | return __btree_submit_bio_hook(inode, rw, bio, mirror_num); | |
541 | } | |
542 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
543 | inode, rw, bio, mirror_num, | |
544 | __btree_submit_bio_hook); | |
545 | } | |
546 | ||
0da5468f CM |
547 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
548 | { | |
d1310b2e CM |
549 | struct extent_io_tree *tree; |
550 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5443be45 CM |
551 | |
552 | if (current->flags & PF_MEMALLOC) { | |
553 | redirty_page_for_writepage(wbc, page); | |
554 | unlock_page(page); | |
555 | return 0; | |
556 | } | |
5f39d397 CM |
557 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
558 | } | |
0da5468f CM |
559 | |
560 | static int btree_writepages(struct address_space *mapping, | |
561 | struct writeback_control *wbc) | |
562 | { | |
d1310b2e CM |
563 | struct extent_io_tree *tree; |
564 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 565 | if (wbc->sync_mode == WB_SYNC_NONE) { |
793955bc CM |
566 | u64 num_dirty; |
567 | u64 start = 0; | |
4854ddd0 | 568 | unsigned long thresh = 8 * 1024 * 1024; |
448d640b CM |
569 | |
570 | if (wbc->for_kupdate) | |
571 | return 0; | |
572 | ||
1832a6d5 CM |
573 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
574 | thresh, EXTENT_DIRTY); | |
793955bc CM |
575 | if (num_dirty < thresh) { |
576 | return 0; | |
577 | } | |
578 | } | |
0da5468f CM |
579 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
580 | } | |
581 | ||
5f39d397 CM |
582 | int btree_readpage(struct file *file, struct page *page) |
583 | { | |
d1310b2e CM |
584 | struct extent_io_tree *tree; |
585 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
586 | return extent_read_full_page(tree, page, btree_get_extent); |
587 | } | |
22b0ebda | 588 | |
70dec807 | 589 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 590 | { |
d1310b2e CM |
591 | struct extent_io_tree *tree; |
592 | struct extent_map_tree *map; | |
5f39d397 | 593 | int ret; |
d98237b3 | 594 | |
98509cfc CM |
595 | if (PageWriteback(page) || PageDirty(page)) |
596 | return 0; | |
597 | ||
d1310b2e CM |
598 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
599 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 600 | |
7b13b7b1 | 601 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
6af118ce CM |
602 | if (!ret) { |
603 | return 0; | |
604 | } | |
605 | ||
606 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
607 | if (ret == 1) { |
608 | ClearPagePrivate(page); | |
609 | set_page_private(page, 0); | |
610 | page_cache_release(page); | |
611 | } | |
6af118ce | 612 | |
d98237b3 CM |
613 | return ret; |
614 | } | |
615 | ||
5f39d397 | 616 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 617 | { |
d1310b2e CM |
618 | struct extent_io_tree *tree; |
619 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
620 | extent_invalidatepage(tree, page, offset); |
621 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 622 | if (PagePrivate(page)) { |
6af118ce CM |
623 | printk("warning page private not zero on page %Lu\n", |
624 | page_offset(page)); | |
9ad6b7bc CM |
625 | ClearPagePrivate(page); |
626 | set_page_private(page, 0); | |
627 | page_cache_release(page); | |
628 | } | |
d98237b3 CM |
629 | } |
630 | ||
5f39d397 | 631 | #if 0 |
d98237b3 | 632 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 633 | { |
87cbda5c | 634 | struct buffer_head *bh; |
0f7d52f4 | 635 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 636 | struct buffer_head *head; |
87cbda5c CM |
637 | if (!page_has_buffers(page)) { |
638 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
639 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
640 | } | |
641 | head = page_buffers(page); | |
642 | bh = head; | |
643 | do { | |
644 | if (buffer_dirty(bh)) | |
645 | csum_tree_block(root, bh, 0); | |
646 | bh = bh->b_this_page; | |
647 | } while (bh != head); | |
d98237b3 | 648 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 649 | } |
5f39d397 | 650 | #endif |
eb60ceac | 651 | |
d98237b3 CM |
652 | static struct address_space_operations btree_aops = { |
653 | .readpage = btree_readpage, | |
654 | .writepage = btree_writepage, | |
0da5468f | 655 | .writepages = btree_writepages, |
5f39d397 CM |
656 | .releasepage = btree_releasepage, |
657 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
658 | .sync_page = block_sync_page, |
659 | }; | |
660 | ||
ca7a79ad CM |
661 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
662 | u64 parent_transid) | |
090d1875 | 663 | { |
5f39d397 CM |
664 | struct extent_buffer *buf = NULL; |
665 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 666 | int ret = 0; |
090d1875 | 667 | |
db94535d | 668 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 669 | if (!buf) |
090d1875 | 670 | return 0; |
d1310b2e | 671 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 672 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 673 | free_extent_buffer(buf); |
de428b63 | 674 | return ret; |
090d1875 CM |
675 | } |
676 | ||
0999df54 CM |
677 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
678 | u64 bytenr, u32 blocksize) | |
679 | { | |
680 | struct inode *btree_inode = root->fs_info->btree_inode; | |
681 | struct extent_buffer *eb; | |
682 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
683 | bytenr, blocksize, GFP_NOFS); | |
684 | return eb; | |
685 | } | |
686 | ||
687 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
688 | u64 bytenr, u32 blocksize) | |
689 | { | |
690 | struct inode *btree_inode = root->fs_info->btree_inode; | |
691 | struct extent_buffer *eb; | |
692 | ||
693 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
694 | bytenr, blocksize, NULL, GFP_NOFS); | |
695 | return eb; | |
696 | } | |
697 | ||
698 | ||
e02119d5 CM |
699 | int btrfs_write_tree_block(struct extent_buffer *buf) |
700 | { | |
701 | return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start, | |
702 | buf->start + buf->len - 1, WB_SYNC_NONE); | |
703 | } | |
704 | ||
705 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
706 | { | |
707 | return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, | |
708 | buf->start, buf->start + buf->len -1); | |
709 | } | |
710 | ||
0999df54 | 711 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 712 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
713 | { |
714 | struct extent_buffer *buf = NULL; | |
715 | struct inode *btree_inode = root->fs_info->btree_inode; | |
716 | struct extent_io_tree *io_tree; | |
717 | int ret; | |
718 | ||
719 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
720 | ||
721 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
722 | if (!buf) | |
723 | return NULL; | |
0999df54 | 724 | |
ca7a79ad | 725 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
726 | |
727 | if (ret == 0) { | |
728 | buf->flags |= EXTENT_UPTODATE; | |
a1b32a59 CM |
729 | } else { |
730 | WARN_ON(1); | |
ce9adaa5 | 731 | } |
5f39d397 | 732 | return buf; |
ce9adaa5 | 733 | |
eb60ceac CM |
734 | } |
735 | ||
e089f05c | 736 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 737 | struct extent_buffer *buf) |
ed2ff2cb | 738 | { |
5f39d397 | 739 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 740 | if (btrfs_header_generation(buf) == |
925baedd CM |
741 | root->fs_info->running_transaction->transid) { |
742 | WARN_ON(!btrfs_tree_locked(buf)); | |
d1310b2e | 743 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 744 | buf); |
925baedd | 745 | } |
5f39d397 CM |
746 | return 0; |
747 | } | |
748 | ||
db94535d | 749 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 750 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 751 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 752 | u64 objectid) |
d97e63b6 | 753 | { |
cfaa7295 | 754 | root->node = NULL; |
0f7d52f4 | 755 | root->inode = NULL; |
a28ec197 | 756 | root->commit_root = NULL; |
31153d81 | 757 | root->ref_tree = NULL; |
db94535d CM |
758 | root->sectorsize = sectorsize; |
759 | root->nodesize = nodesize; | |
760 | root->leafsize = leafsize; | |
87ee04eb | 761 | root->stripesize = stripesize; |
123abc88 | 762 | root->ref_cows = 0; |
0b86a832 CM |
763 | root->track_dirty = 0; |
764 | ||
9f5fae2f | 765 | root->fs_info = fs_info; |
0f7d52f4 CM |
766 | root->objectid = objectid; |
767 | root->last_trans = 0; | |
1b05da2e CM |
768 | root->highest_inode = 0; |
769 | root->last_inode_alloc = 0; | |
58176a96 | 770 | root->name = NULL; |
4313b399 | 771 | root->in_sysfs = 0; |
0b86a832 CM |
772 | |
773 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 774 | INIT_LIST_HEAD(&root->orphan_list); |
bcc63abb | 775 | INIT_LIST_HEAD(&root->dead_list); |
925baedd | 776 | spin_lock_init(&root->node_lock); |
bcc63abb | 777 | spin_lock_init(&root->list_lock); |
a2135011 | 778 | mutex_init(&root->objectid_mutex); |
e02119d5 | 779 | mutex_init(&root->log_mutex); |
d0c803c4 CM |
780 | extent_io_tree_init(&root->dirty_log_pages, |
781 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
017e5369 CM |
782 | |
783 | btrfs_leaf_ref_tree_init(&root->ref_tree_struct); | |
784 | root->ref_tree = &root->ref_tree_struct; | |
785 | ||
3768f368 CM |
786 | memset(&root->root_key, 0, sizeof(root->root_key)); |
787 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 788 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 789 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 790 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 791 | init_completion(&root->kobj_unregister); |
6702ed49 CM |
792 | root->defrag_running = 0; |
793 | root->defrag_level = 0; | |
4d775673 | 794 | root->root_key.objectid = objectid; |
3768f368 CM |
795 | return 0; |
796 | } | |
797 | ||
db94535d | 798 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
799 | struct btrfs_fs_info *fs_info, |
800 | u64 objectid, | |
e20d96d6 | 801 | struct btrfs_root *root) |
3768f368 CM |
802 | { |
803 | int ret; | |
db94535d | 804 | u32 blocksize; |
3768f368 | 805 | |
db94535d | 806 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
807 | tree_root->sectorsize, tree_root->stripesize, |
808 | root, fs_info, objectid); | |
3768f368 CM |
809 | ret = btrfs_find_last_root(tree_root, objectid, |
810 | &root->root_item, &root->root_key); | |
811 | BUG_ON(ret); | |
812 | ||
db94535d CM |
813 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
814 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 815 | blocksize, 0); |
3768f368 | 816 | BUG_ON(!root->node); |
d97e63b6 CM |
817 | return 0; |
818 | } | |
819 | ||
e02119d5 CM |
820 | int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, |
821 | struct btrfs_fs_info *fs_info) | |
822 | { | |
823 | struct extent_buffer *eb; | |
d0c803c4 CM |
824 | struct btrfs_root *log_root_tree = fs_info->log_root_tree; |
825 | u64 start = 0; | |
826 | u64 end = 0; | |
e02119d5 CM |
827 | int ret; |
828 | ||
d0c803c4 | 829 | if (!log_root_tree) |
e02119d5 CM |
830 | return 0; |
831 | ||
d0c803c4 CM |
832 | while(1) { |
833 | ret = find_first_extent_bit(&log_root_tree->dirty_log_pages, | |
834 | 0, &start, &end, EXTENT_DIRTY); | |
835 | if (ret) | |
836 | break; | |
837 | ||
838 | clear_extent_dirty(&log_root_tree->dirty_log_pages, | |
839 | start, end, GFP_NOFS); | |
840 | } | |
e02119d5 CM |
841 | eb = fs_info->log_root_tree->node; |
842 | ||
843 | WARN_ON(btrfs_header_level(eb) != 0); | |
844 | WARN_ON(btrfs_header_nritems(eb) != 0); | |
845 | ||
d00aff00 CM |
846 | ret = btrfs_free_reserved_extent(fs_info->tree_root, |
847 | eb->start, eb->len); | |
e02119d5 CM |
848 | BUG_ON(ret); |
849 | ||
850 | free_extent_buffer(eb); | |
851 | kfree(fs_info->log_root_tree); | |
852 | fs_info->log_root_tree = NULL; | |
853 | return 0; | |
854 | } | |
855 | ||
856 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
857 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
858 | { |
859 | struct btrfs_root *root; | |
860 | struct btrfs_root *tree_root = fs_info->tree_root; | |
e02119d5 CM |
861 | |
862 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
863 | if (!root) | |
864 | return -ENOMEM; | |
865 | ||
866 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
867 | tree_root->sectorsize, tree_root->stripesize, | |
868 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
869 | ||
870 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
871 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
872 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
873 | root->ref_cows = 0; | |
874 | ||
875 | root->node = btrfs_alloc_free_block(trans, root, root->leafsize, | |
876 | BTRFS_TREE_LOG_OBJECTID, | |
877 | 0, 0, 0, 0, 0); | |
878 | ||
879 | btrfs_set_header_nritems(root->node, 0); | |
880 | btrfs_set_header_level(root->node, 0); | |
881 | btrfs_set_header_bytenr(root->node, root->node->start); | |
882 | btrfs_set_header_generation(root->node, trans->transid); | |
883 | btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID); | |
884 | ||
885 | write_extent_buffer(root->node, root->fs_info->fsid, | |
886 | (unsigned long)btrfs_header_fsid(root->node), | |
887 | BTRFS_FSID_SIZE); | |
888 | btrfs_mark_buffer_dirty(root->node); | |
889 | btrfs_tree_unlock(root->node); | |
890 | fs_info->log_root_tree = root; | |
891 | return 0; | |
892 | } | |
893 | ||
894 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
895 | struct btrfs_key *location) | |
896 | { | |
897 | struct btrfs_root *root; | |
898 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 899 | struct btrfs_path *path; |
5f39d397 | 900 | struct extent_buffer *l; |
1b05da2e | 901 | u64 highest_inode; |
db94535d | 902 | u32 blocksize; |
0f7d52f4 CM |
903 | int ret = 0; |
904 | ||
5eda7b5e | 905 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 906 | if (!root) |
0f7d52f4 | 907 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 908 | if (location->offset == (u64)-1) { |
db94535d | 909 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
910 | location->objectid, root); |
911 | if (ret) { | |
0f7d52f4 CM |
912 | kfree(root); |
913 | return ERR_PTR(ret); | |
914 | } | |
915 | goto insert; | |
916 | } | |
917 | ||
db94535d | 918 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
919 | tree_root->sectorsize, tree_root->stripesize, |
920 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
921 | |
922 | path = btrfs_alloc_path(); | |
923 | BUG_ON(!path); | |
924 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
925 | if (ret != 0) { | |
0f7d52f4 CM |
926 | if (ret > 0) |
927 | ret = -ENOENT; | |
928 | goto out; | |
929 | } | |
5f39d397 CM |
930 | l = path->nodes[0]; |
931 | read_extent_buffer(l, &root->root_item, | |
932 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 933 | sizeof(root->root_item)); |
44b36eb2 | 934 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
935 | ret = 0; |
936 | out: | |
937 | btrfs_release_path(root, path); | |
938 | btrfs_free_path(path); | |
939 | if (ret) { | |
940 | kfree(root); | |
941 | return ERR_PTR(ret); | |
942 | } | |
db94535d CM |
943 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
944 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
ca7a79ad | 945 | blocksize, 0); |
0f7d52f4 CM |
946 | BUG_ON(!root->node); |
947 | insert: | |
e02119d5 CM |
948 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
949 | root->ref_cows = 1; | |
950 | ret = btrfs_find_highest_inode(root, &highest_inode); | |
951 | if (ret == 0) { | |
952 | root->highest_inode = highest_inode; | |
953 | root->last_inode_alloc = highest_inode; | |
954 | } | |
5eda7b5e CM |
955 | } |
956 | return root; | |
957 | } | |
958 | ||
dc17ff8f CM |
959 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
960 | u64 root_objectid) | |
961 | { | |
962 | struct btrfs_root *root; | |
963 | ||
964 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
965 | return fs_info->tree_root; | |
966 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
967 | return fs_info->extent_root; | |
968 | ||
969 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
970 | (unsigned long)root_objectid); | |
971 | return root; | |
972 | } | |
973 | ||
edbd8d4e CM |
974 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
975 | struct btrfs_key *location) | |
5eda7b5e CM |
976 | { |
977 | struct btrfs_root *root; | |
978 | int ret; | |
979 | ||
edbd8d4e CM |
980 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
981 | return fs_info->tree_root; | |
982 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
983 | return fs_info->extent_root; | |
8f18cf13 CM |
984 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
985 | return fs_info->chunk_root; | |
986 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
987 | return fs_info->dev_root; | |
edbd8d4e | 988 | |
5eda7b5e CM |
989 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
990 | (unsigned long)location->objectid); | |
991 | if (root) | |
992 | return root; | |
993 | ||
e02119d5 | 994 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
995 | if (IS_ERR(root)) |
996 | return root; | |
2619ba1f CM |
997 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
998 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
999 | root); |
1000 | if (ret) { | |
5f39d397 | 1001 | free_extent_buffer(root->node); |
0f7d52f4 CM |
1002 | kfree(root); |
1003 | return ERR_PTR(ret); | |
1004 | } | |
edbd8d4e CM |
1005 | ret = btrfs_find_dead_roots(fs_info->tree_root, |
1006 | root->root_key.objectid, root); | |
1007 | BUG_ON(ret); | |
1008 | ||
1009 | return root; | |
1010 | } | |
1011 | ||
1012 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
1013 | struct btrfs_key *location, | |
1014 | const char *name, int namelen) | |
1015 | { | |
1016 | struct btrfs_root *root; | |
1017 | int ret; | |
1018 | ||
1019 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
1020 | if (!root) | |
1021 | return NULL; | |
58176a96 | 1022 | |
4313b399 CM |
1023 | if (root->in_sysfs) |
1024 | return root; | |
1025 | ||
58176a96 JB |
1026 | ret = btrfs_set_root_name(root, name, namelen); |
1027 | if (ret) { | |
5f39d397 | 1028 | free_extent_buffer(root->node); |
58176a96 JB |
1029 | kfree(root); |
1030 | return ERR_PTR(ret); | |
1031 | } | |
1032 | ||
1033 | ret = btrfs_sysfs_add_root(root); | |
1034 | if (ret) { | |
5f39d397 | 1035 | free_extent_buffer(root->node); |
58176a96 JB |
1036 | kfree(root->name); |
1037 | kfree(root); | |
1038 | return ERR_PTR(ret); | |
1039 | } | |
4313b399 | 1040 | root->in_sysfs = 1; |
0f7d52f4 CM |
1041 | return root; |
1042 | } | |
19c00ddc CM |
1043 | #if 0 |
1044 | static int add_hasher(struct btrfs_fs_info *info, char *type) { | |
1045 | struct btrfs_hasher *hasher; | |
1046 | ||
1047 | hasher = kmalloc(sizeof(*hasher), GFP_NOFS); | |
1048 | if (!hasher) | |
1049 | return -ENOMEM; | |
1050 | hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); | |
1051 | if (!hasher->hash_tfm) { | |
1052 | kfree(hasher); | |
1053 | return -EINVAL; | |
1054 | } | |
1055 | spin_lock(&info->hash_lock); | |
1056 | list_add(&hasher->list, &info->hashers); | |
1057 | spin_unlock(&info->hash_lock); | |
1058 | return 0; | |
1059 | } | |
1060 | #endif | |
04160088 CM |
1061 | |
1062 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
1063 | { | |
1064 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1065 | int ret = 0; | |
1066 | struct list_head *cur; | |
1067 | struct btrfs_device *device; | |
1068 | struct backing_dev_info *bdi; | |
1069 | ||
cb03c743 | 1070 | if ((bdi_bits & (1 << BDI_write_congested)) && |
777e6bd7 | 1071 | btrfs_congested_async(info, 0)) |
cb03c743 | 1072 | return 1; |
cb03c743 | 1073 | |
04160088 CM |
1074 | list_for_each(cur, &info->fs_devices->devices) { |
1075 | device = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1076 | if (!device->bdev) |
1077 | continue; | |
04160088 CM |
1078 | bdi = blk_get_backing_dev_info(device->bdev); |
1079 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1080 | ret = 1; | |
1081 | break; | |
1082 | } | |
1083 | } | |
1084 | return ret; | |
1085 | } | |
1086 | ||
38b66988 CM |
1087 | /* |
1088 | * this unplugs every device on the box, and it is only used when page | |
1089 | * is null | |
1090 | */ | |
1091 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
1092 | { | |
1093 | struct list_head *cur; | |
1094 | struct btrfs_device *device; | |
1095 | struct btrfs_fs_info *info; | |
1096 | ||
1097 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
1098 | list_for_each(cur, &info->fs_devices->devices) { | |
1099 | device = list_entry(cur, struct btrfs_device, dev_list); | |
1100 | bdi = blk_get_backing_dev_info(device->bdev); | |
1101 | if (bdi->unplug_io_fn) { | |
1102 | bdi->unplug_io_fn(bdi, page); | |
1103 | } | |
1104 | } | |
1105 | } | |
1106 | ||
04160088 CM |
1107 | void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
1108 | { | |
38b66988 | 1109 | struct inode *inode; |
f2d8d74d CM |
1110 | struct extent_map_tree *em_tree; |
1111 | struct extent_map *em; | |
bcbfce8a | 1112 | struct address_space *mapping; |
38b66988 CM |
1113 | u64 offset; |
1114 | ||
bcbfce8a | 1115 | /* the generic O_DIRECT read code does this */ |
38b66988 CM |
1116 | if (!page) { |
1117 | __unplug_io_fn(bdi, page); | |
1118 | return; | |
1119 | } | |
1120 | ||
bcbfce8a CM |
1121 | /* |
1122 | * page->mapping may change at any time. Get a consistent copy | |
1123 | * and use that for everything below | |
1124 | */ | |
1125 | smp_mb(); | |
1126 | mapping = page->mapping; | |
1127 | if (!mapping) | |
1128 | return; | |
1129 | ||
1130 | inode = mapping->host; | |
38b66988 | 1131 | offset = page_offset(page); |
04160088 | 1132 | |
f2d8d74d CM |
1133 | em_tree = &BTRFS_I(inode)->extent_tree; |
1134 | spin_lock(&em_tree->lock); | |
1135 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
1136 | spin_unlock(&em_tree->lock); | |
89642229 CM |
1137 | if (!em) { |
1138 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1139 | return; |
89642229 | 1140 | } |
f2d8d74d | 1141 | |
89642229 CM |
1142 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1143 | free_extent_map(em); | |
1144 | __unplug_io_fn(bdi, page); | |
1145 | return; | |
1146 | } | |
f2d8d74d CM |
1147 | offset = offset - em->start; |
1148 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1149 | em->block_start + offset, page); | |
1150 | free_extent_map(em); | |
04160088 CM |
1151 | } |
1152 | ||
1153 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
1154 | { | |
51ebc0d3 | 1155 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
04160088 | 1156 | bdi_init(bdi); |
b248a415 | 1157 | #endif |
4575c9cc | 1158 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1159 | bdi->state = 0; |
1160 | bdi->capabilities = default_backing_dev_info.capabilities; | |
1161 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
1162 | bdi->unplug_io_data = info; | |
1163 | bdi->congested_fn = btrfs_congested_fn; | |
1164 | bdi->congested_data = info; | |
1165 | return 0; | |
1166 | } | |
1167 | ||
ce9adaa5 CM |
1168 | static int bio_ready_for_csum(struct bio *bio) |
1169 | { | |
1170 | u64 length = 0; | |
1171 | u64 buf_len = 0; | |
1172 | u64 start = 0; | |
1173 | struct page *page; | |
1174 | struct extent_io_tree *io_tree = NULL; | |
1175 | struct btrfs_fs_info *info = NULL; | |
1176 | struct bio_vec *bvec; | |
1177 | int i; | |
1178 | int ret; | |
1179 | ||
1180 | bio_for_each_segment(bvec, bio, i) { | |
1181 | page = bvec->bv_page; | |
1182 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1183 | length += bvec->bv_len; | |
1184 | continue; | |
1185 | } | |
1186 | if (!page->private) { | |
1187 | length += bvec->bv_len; | |
1188 | continue; | |
1189 | } | |
1190 | length = bvec->bv_len; | |
1191 | buf_len = page->private >> 2; | |
1192 | start = page_offset(page) + bvec->bv_offset; | |
1193 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1194 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1195 | } | |
1196 | /* are we fully contained in this bio? */ | |
1197 | if (buf_len <= length) | |
1198 | return 1; | |
1199 | ||
1200 | ret = extent_range_uptodate(io_tree, start + length, | |
1201 | start + buf_len - 1); | |
1202 | if (ret == 1) | |
1203 | return ret; | |
1204 | return ret; | |
1205 | } | |
1206 | ||
8b712842 CM |
1207 | /* |
1208 | * called by the kthread helper functions to finally call the bio end_io | |
1209 | * functions. This is where read checksum verification actually happens | |
1210 | */ | |
1211 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1212 | { |
ce9adaa5 | 1213 | struct bio *bio; |
8b712842 CM |
1214 | struct end_io_wq *end_io_wq; |
1215 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1216 | int error; |
ce9adaa5 | 1217 | |
8b712842 CM |
1218 | end_io_wq = container_of(work, struct end_io_wq, work); |
1219 | bio = end_io_wq->bio; | |
1220 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1221 | |
8b712842 CM |
1222 | /* metadata bios are special because the whole tree block must |
1223 | * be checksummed at once. This makes sure the entire block is in | |
1224 | * ram and up to date before trying to verify things. For | |
1225 | * blocksize <= pagesize, it is basically a noop | |
1226 | */ | |
1227 | if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { | |
1228 | btrfs_queue_worker(&fs_info->endio_workers, | |
1229 | &end_io_wq->work); | |
1230 | return; | |
1231 | } | |
1232 | error = end_io_wq->error; | |
1233 | bio->bi_private = end_io_wq->private; | |
1234 | bio->bi_end_io = end_io_wq->end_io; | |
1235 | kfree(end_io_wq); | |
1236 | #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23) | |
1237 | bio_endio(bio, bio->bi_size, error); | |
44b8bd7e | 1238 | #else |
8b712842 | 1239 | bio_endio(bio, error); |
44b8bd7e | 1240 | #endif |
44b8bd7e CM |
1241 | } |
1242 | ||
a74a4b97 CM |
1243 | static int cleaner_kthread(void *arg) |
1244 | { | |
1245 | struct btrfs_root *root = arg; | |
1246 | ||
1247 | do { | |
1248 | smp_mb(); | |
1249 | if (root->fs_info->closing) | |
1250 | break; | |
1251 | ||
1252 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1253 | mutex_lock(&root->fs_info->cleaner_mutex); | |
a74a4b97 | 1254 | btrfs_clean_old_snapshots(root); |
a74a4b97 CM |
1255 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1256 | ||
1257 | if (freezing(current)) { | |
1258 | refrigerator(); | |
1259 | } else { | |
1260 | smp_mb(); | |
1261 | if (root->fs_info->closing) | |
1262 | break; | |
1263 | set_current_state(TASK_INTERRUPTIBLE); | |
1264 | schedule(); | |
1265 | __set_current_state(TASK_RUNNING); | |
1266 | } | |
1267 | } while (!kthread_should_stop()); | |
1268 | return 0; | |
1269 | } | |
1270 | ||
1271 | static int transaction_kthread(void *arg) | |
1272 | { | |
1273 | struct btrfs_root *root = arg; | |
1274 | struct btrfs_trans_handle *trans; | |
1275 | struct btrfs_transaction *cur; | |
1276 | unsigned long now; | |
1277 | unsigned long delay; | |
1278 | int ret; | |
1279 | ||
1280 | do { | |
1281 | smp_mb(); | |
1282 | if (root->fs_info->closing) | |
1283 | break; | |
1284 | ||
1285 | delay = HZ * 30; | |
1286 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1287 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1288 | ||
ab78c84d CM |
1289 | if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) { |
1290 | printk("btrfs: total reference cache size %Lu\n", | |
1291 | root->fs_info->total_ref_cache_size); | |
1292 | } | |
31153d81 | 1293 | |
a74a4b97 CM |
1294 | mutex_lock(&root->fs_info->trans_mutex); |
1295 | cur = root->fs_info->running_transaction; | |
1296 | if (!cur) { | |
1297 | mutex_unlock(&root->fs_info->trans_mutex); | |
1298 | goto sleep; | |
1299 | } | |
31153d81 | 1300 | |
a74a4b97 CM |
1301 | now = get_seconds(); |
1302 | if (now < cur->start_time || now - cur->start_time < 30) { | |
1303 | mutex_unlock(&root->fs_info->trans_mutex); | |
1304 | delay = HZ * 5; | |
1305 | goto sleep; | |
1306 | } | |
1307 | mutex_unlock(&root->fs_info->trans_mutex); | |
a74a4b97 CM |
1308 | trans = btrfs_start_transaction(root, 1); |
1309 | ret = btrfs_commit_transaction(trans, root); | |
1310 | sleep: | |
1311 | wake_up_process(root->fs_info->cleaner_kthread); | |
1312 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1313 | ||
1314 | if (freezing(current)) { | |
1315 | refrigerator(); | |
1316 | } else { | |
1317 | if (root->fs_info->closing) | |
1318 | break; | |
1319 | set_current_state(TASK_INTERRUPTIBLE); | |
1320 | schedule_timeout(delay); | |
1321 | __set_current_state(TASK_RUNNING); | |
1322 | } | |
1323 | } while (!kthread_should_stop()); | |
1324 | return 0; | |
1325 | } | |
1326 | ||
8a4b83cc | 1327 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1328 | struct btrfs_fs_devices *fs_devices, |
1329 | char *options) | |
2e635a27 | 1330 | { |
db94535d CM |
1331 | u32 sectorsize; |
1332 | u32 nodesize; | |
1333 | u32 leafsize; | |
1334 | u32 blocksize; | |
87ee04eb | 1335 | u32 stripesize; |
a061fc8d | 1336 | struct buffer_head *bh; |
e02119d5 | 1337 | struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1338 | GFP_NOFS); |
e02119d5 | 1339 | struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1340 | GFP_NOFS); |
8790d502 | 1341 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1342 | GFP_NOFS); |
e02119d5 | 1343 | struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1344 | GFP_NOFS); |
e02119d5 | 1345 | struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1346 | GFP_NOFS); |
e02119d5 CM |
1347 | struct btrfs_root *log_tree_root; |
1348 | ||
eb60ceac | 1349 | int ret; |
e58ca020 | 1350 | int err = -EINVAL; |
4543df7e | 1351 | |
2c90e5d6 | 1352 | struct btrfs_super_block *disk_super; |
8790d502 | 1353 | |
39279cc3 CM |
1354 | if (!extent_root || !tree_root || !fs_info) { |
1355 | err = -ENOMEM; | |
1356 | goto fail; | |
1357 | } | |
0f7d52f4 | 1358 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1359 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1360 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1361 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1362 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
19c00ddc | 1363 | spin_lock_init(&fs_info->hash_lock); |
1832a6d5 | 1364 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1365 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1366 | spin_lock_init(&fs_info->ref_cache_lock); |
19c00ddc | 1367 | |
58176a96 | 1368 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1369 | fs_info->tree_root = tree_root; |
1370 | fs_info->extent_root = extent_root; | |
0b86a832 CM |
1371 | fs_info->chunk_root = chunk_root; |
1372 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1373 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1374 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1375 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1376 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1377 | atomic_set(&fs_info->nr_async_submits, 0); |
0986fe9e | 1378 | atomic_set(&fs_info->nr_async_bios, 0); |
a2135011 | 1379 | atomic_set(&fs_info->throttles, 0); |
ab78c84d | 1380 | atomic_set(&fs_info->throttle_gen, 0); |
e20d96d6 | 1381 | fs_info->sb = sb; |
c59f8951 | 1382 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1383 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1384 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1385 | fs_info->btree_inode = new_inode(sb); |
1386 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1387 | fs_info->btree_inode->i_nlink = 1; |
4543df7e | 1388 | fs_info->thread_pool_size = min(num_online_cpus() + 2, 8); |
0afbaf8c | 1389 | |
3eaa2885 CM |
1390 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1391 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1392 | ||
a061fc8d CM |
1393 | sb->s_blocksize = 4096; |
1394 | sb->s_blocksize_bits = blksize_bits(4096); | |
1395 | ||
0afbaf8c CM |
1396 | /* |
1397 | * we set the i_size on the btree inode to the max possible int. | |
1398 | * the real end of the address space is determined by all of | |
1399 | * the devices in the system | |
1400 | */ | |
1401 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1402 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1403 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1404 | ||
d1310b2e | 1405 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1406 | fs_info->btree_inode->i_mapping, |
1407 | GFP_NOFS); | |
d1310b2e CM |
1408 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1409 | GFP_NOFS); | |
1410 | ||
1411 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1412 | |
d1310b2e | 1413 | extent_io_tree_init(&fs_info->free_space_cache, |
f510cfec | 1414 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1415 | extent_io_tree_init(&fs_info->block_group_cache, |
96b5179d | 1416 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1417 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1418 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1419 | extent_io_tree_init(&fs_info->pending_del, |
1a5bc167 | 1420 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1421 | extent_io_tree_init(&fs_info->extent_ins, |
1a5bc167 | 1422 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1423 | fs_info->do_barriers = 1; |
e18e4809 | 1424 | |
0f7d52f4 CM |
1425 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1426 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1427 | sizeof(struct btrfs_key)); | |
22b0ebda | 1428 | insert_inode_hash(fs_info->btree_inode); |
39279cc3 | 1429 | |
79154b1b | 1430 | mutex_init(&fs_info->trans_mutex); |
e02119d5 | 1431 | mutex_init(&fs_info->tree_log_mutex); |
a2135011 | 1432 | mutex_init(&fs_info->drop_mutex); |
925baedd CM |
1433 | mutex_init(&fs_info->alloc_mutex); |
1434 | mutex_init(&fs_info->chunk_mutex); | |
a74a4b97 CM |
1435 | mutex_init(&fs_info->transaction_kthread_mutex); |
1436 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1437 | mutex_init(&fs_info->volume_mutex); |
e6dcd2dc | 1438 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1439 | init_waitqueue_head(&fs_info->transaction_wait); |
4854ddd0 | 1440 | init_waitqueue_head(&fs_info->async_submit_wait); |
e02119d5 CM |
1441 | init_waitqueue_head(&fs_info->tree_log_wait); |
1442 | atomic_set(&fs_info->tree_log_commit, 0); | |
1443 | atomic_set(&fs_info->tree_log_writers, 0); | |
1444 | fs_info->tree_log_transid = 0; | |
3768f368 | 1445 | |
19c00ddc CM |
1446 | #if 0 |
1447 | ret = add_hasher(fs_info, "crc32c"); | |
1448 | if (ret) { | |
1449 | printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); | |
1450 | err = -ENOMEM; | |
1451 | goto fail_iput; | |
1452 | } | |
1453 | #endif | |
0b86a832 | 1454 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1455 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1456 | |
d98237b3 | 1457 | |
a061fc8d CM |
1458 | bh = __bread(fs_devices->latest_bdev, |
1459 | BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
1460 | if (!bh) | |
39279cc3 | 1461 | goto fail_iput; |
39279cc3 | 1462 | |
a061fc8d CM |
1463 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
1464 | brelse(bh); | |
5f39d397 | 1465 | |
a061fc8d | 1466 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1467 | |
5f39d397 | 1468 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1469 | if (!btrfs_super_root(disk_super)) |
39279cc3 | 1470 | goto fail_sb_buffer; |
0f7d52f4 | 1471 | |
edf24abe CH |
1472 | err = btrfs_parse_options(tree_root, options); |
1473 | if (err) | |
1474 | goto fail_sb_buffer; | |
dfe25020 | 1475 | |
4543df7e CM |
1476 | /* |
1477 | * we need to start all the end_io workers up front because the | |
1478 | * queue work function gets called at interrupt time, and so it | |
1479 | * cannot dynamically grow. | |
1480 | */ | |
5443be45 CM |
1481 | btrfs_init_workers(&fs_info->workers, "worker", |
1482 | fs_info->thread_pool_size); | |
1483 | btrfs_init_workers(&fs_info->submit_workers, "submit", | |
b720d209 CM |
1484 | min_t(u64, fs_devices->num_devices, |
1485 | fs_info->thread_pool_size)); | |
61b49440 CM |
1486 | |
1487 | /* a higher idle thresh on the submit workers makes it much more | |
1488 | * likely that bios will be send down in a sane order to the | |
1489 | * devices | |
1490 | */ | |
1491 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 CM |
1492 | |
1493 | /* fs_info->workers is responsible for checksumming file data | |
1494 | * blocks and metadata. Using a larger idle thresh allows each | |
1495 | * worker thread to operate on things in roughly the order they | |
1496 | * were sent by the writeback daemons, improving overall locality | |
1497 | * of the IO going down the pipe. | |
1498 | */ | |
1499 | fs_info->workers.idle_thresh = 128; | |
61b49440 | 1500 | |
5443be45 CM |
1501 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); |
1502 | btrfs_init_workers(&fs_info->endio_workers, "endio", | |
1503 | fs_info->thread_pool_size); | |
1504 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", | |
e6dcd2dc | 1505 | fs_info->thread_pool_size); |
61b49440 CM |
1506 | |
1507 | /* | |
1508 | * endios are largely parallel and should have a very | |
1509 | * low idle thresh | |
1510 | */ | |
1511 | fs_info->endio_workers.idle_thresh = 4; | |
3a5f1d45 | 1512 | fs_info->endio_write_workers.idle_thresh = 64; |
61b49440 | 1513 | |
4543df7e | 1514 | btrfs_start_workers(&fs_info->workers, 1); |
1cc127b5 | 1515 | btrfs_start_workers(&fs_info->submit_workers, 1); |
247e743c | 1516 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
4543df7e | 1517 | btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
e6dcd2dc CM |
1518 | btrfs_start_workers(&fs_info->endio_write_workers, |
1519 | fs_info->thread_pool_size); | |
4543df7e | 1520 | |
edf24abe | 1521 | err = -EINVAL; |
a0af469b | 1522 | if (btrfs_super_num_devices(disk_super) > fs_devices->open_devices) { |
8a4b83cc CM |
1523 | printk("Btrfs: wanted %llu devices, but found %llu\n", |
1524 | (unsigned long long)btrfs_super_num_devices(disk_super), | |
a0af469b | 1525 | (unsigned long long)fs_devices->open_devices); |
dfe25020 CM |
1526 | if (btrfs_test_opt(tree_root, DEGRADED)) |
1527 | printk("continuing in degraded mode\n"); | |
1528 | else { | |
1529 | goto fail_sb_buffer; | |
1530 | } | |
8a4b83cc | 1531 | } |
dfe25020 | 1532 | |
4575c9cc CM |
1533 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
1534 | ||
db94535d CM |
1535 | nodesize = btrfs_super_nodesize(disk_super); |
1536 | leafsize = btrfs_super_leafsize(disk_super); | |
1537 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1538 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1539 | tree_root->nodesize = nodesize; |
1540 | tree_root->leafsize = leafsize; | |
1541 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1542 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1543 | |
1544 | sb->s_blocksize = sectorsize; | |
1545 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1546 | |
39279cc3 CM |
1547 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1548 | sizeof(disk_super->magic))) { | |
1549 | printk("btrfs: valid FS not found on %s\n", sb->s_id); | |
1550 | goto fail_sb_buffer; | |
1551 | } | |
19c00ddc | 1552 | |
925baedd | 1553 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1554 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1555 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f CM |
1556 | if (ret) { |
1557 | printk("btrfs: failed to read the system array on %s\n", | |
1558 | sb->s_id); | |
1559 | goto fail_sys_array; | |
1560 | } | |
0b86a832 CM |
1561 | |
1562 | blocksize = btrfs_level_size(tree_root, | |
1563 | btrfs_super_chunk_root_level(disk_super)); | |
1564 | ||
1565 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1566 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1567 | ||
1568 | chunk_root->node = read_tree_block(chunk_root, | |
1569 | btrfs_super_chunk_root(disk_super), | |
ca7a79ad | 1570 | blocksize, 0); |
0b86a832 CM |
1571 | BUG_ON(!chunk_root->node); |
1572 | ||
e17cade2 CM |
1573 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
1574 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), | |
1575 | BTRFS_UUID_SIZE); | |
1576 | ||
925baedd | 1577 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1578 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1579 | mutex_unlock(&fs_info->chunk_mutex); |
0b86a832 CM |
1580 | BUG_ON(ret); |
1581 | ||
dfe25020 CM |
1582 | btrfs_close_extra_devices(fs_devices); |
1583 | ||
db94535d CM |
1584 | blocksize = btrfs_level_size(tree_root, |
1585 | btrfs_super_root_level(disk_super)); | |
19c00ddc | 1586 | |
0b86a832 | 1587 | |
e20d96d6 | 1588 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1589 | btrfs_super_root(disk_super), |
ca7a79ad | 1590 | blocksize, 0); |
39279cc3 CM |
1591 | if (!tree_root->node) |
1592 | goto fail_sb_buffer; | |
3768f368 | 1593 | |
db94535d CM |
1594 | |
1595 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1596 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1597 | if (ret) |
39279cc3 | 1598 | goto fail_tree_root; |
0b86a832 CM |
1599 | extent_root->track_dirty = 1; |
1600 | ||
1601 | ret = find_and_setup_root(tree_root, fs_info, | |
1602 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1603 | dev_root->track_dirty = 1; | |
1604 | ||
1605 | if (ret) | |
1606 | goto fail_extent_root; | |
3768f368 | 1607 | |
9078a3e1 CM |
1608 | btrfs_read_block_groups(extent_root); |
1609 | ||
0f7d52f4 | 1610 | fs_info->generation = btrfs_super_generation(disk_super) + 1; |
d18a2c44 CM |
1611 | fs_info->data_alloc_profile = (u64)-1; |
1612 | fs_info->metadata_alloc_profile = (u64)-1; | |
1613 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
a74a4b97 CM |
1614 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1615 | "btrfs-cleaner"); | |
1616 | if (!fs_info->cleaner_kthread) | |
1617 | goto fail_extent_root; | |
1618 | ||
1619 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1620 | tree_root, | |
1621 | "btrfs-transaction"); | |
1622 | if (!fs_info->transaction_kthread) | |
3f157a2f | 1623 | goto fail_cleaner; |
a74a4b97 | 1624 | |
e02119d5 CM |
1625 | if (btrfs_super_log_root(disk_super) != 0) { |
1626 | u32 blocksize; | |
1627 | u64 bytenr = btrfs_super_log_root(disk_super); | |
1628 | ||
1629 | blocksize = | |
1630 | btrfs_level_size(tree_root, | |
1631 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 1632 | |
e02119d5 CM |
1633 | log_tree_root = kzalloc(sizeof(struct btrfs_root), |
1634 | GFP_NOFS); | |
1635 | ||
1636 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1637 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1638 | ||
1639 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
1640 | blocksize, 0); | |
1641 | ret = btrfs_recover_log_trees(log_tree_root); | |
1642 | BUG_ON(ret); | |
1643 | } | |
1644 | fs_info->last_trans_committed = btrfs_super_generation(disk_super); | |
0f7d52f4 | 1645 | return tree_root; |
39279cc3 | 1646 | |
3f157a2f | 1647 | fail_cleaner: |
a74a4b97 | 1648 | kthread_stop(fs_info->cleaner_kthread); |
0b86a832 CM |
1649 | fail_extent_root: |
1650 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1651 | fail_tree_root: |
5f39d397 | 1652 | free_extent_buffer(tree_root->node); |
84eed90f | 1653 | fail_sys_array: |
39279cc3 | 1654 | fail_sb_buffer: |
247e743c | 1655 | btrfs_stop_workers(&fs_info->fixup_workers); |
8b712842 CM |
1656 | btrfs_stop_workers(&fs_info->workers); |
1657 | btrfs_stop_workers(&fs_info->endio_workers); | |
e6dcd2dc | 1658 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1659 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e CM |
1660 | fail_iput: |
1661 | iput(fs_info->btree_inode); | |
39279cc3 | 1662 | fail: |
dfe25020 | 1663 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f CM |
1664 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
1665 | ||
39279cc3 CM |
1666 | kfree(extent_root); |
1667 | kfree(tree_root); | |
51ebc0d3 | 1668 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
2d2ae547 | 1669 | bdi_destroy(&fs_info->bdi); |
b248a415 | 1670 | #endif |
39279cc3 CM |
1671 | kfree(fs_info); |
1672 | return ERR_PTR(err); | |
eb60ceac CM |
1673 | } |
1674 | ||
f2984462 CM |
1675 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1676 | { | |
1677 | char b[BDEVNAME_SIZE]; | |
1678 | ||
1679 | if (uptodate) { | |
1680 | set_buffer_uptodate(bh); | |
1681 | } else { | |
1682 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1683 | printk(KERN_WARNING "lost page write due to " | |
1684 | "I/O error on %s\n", | |
1685 | bdevname(bh->b_bdev, b)); | |
1686 | } | |
1259ab75 CM |
1687 | /* note, we dont' set_buffer_write_io_error because we have |
1688 | * our own ways of dealing with the IO errors | |
1689 | */ | |
f2984462 CM |
1690 | clear_buffer_uptodate(bh); |
1691 | } | |
1692 | unlock_buffer(bh); | |
1693 | put_bh(bh); | |
1694 | } | |
1695 | ||
1696 | int write_all_supers(struct btrfs_root *root) | |
1697 | { | |
1698 | struct list_head *cur; | |
1699 | struct list_head *head = &root->fs_info->fs_devices->devices; | |
1700 | struct btrfs_device *dev; | |
a061fc8d | 1701 | struct btrfs_super_block *sb; |
f2984462 CM |
1702 | struct btrfs_dev_item *dev_item; |
1703 | struct buffer_head *bh; | |
1704 | int ret; | |
1705 | int do_barriers; | |
a236aed1 CM |
1706 | int max_errors; |
1707 | int total_errors = 0; | |
a061fc8d CM |
1708 | u32 crc; |
1709 | u64 flags; | |
f2984462 | 1710 | |
a236aed1 | 1711 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
1712 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
1713 | ||
a061fc8d CM |
1714 | sb = &root->fs_info->super_for_commit; |
1715 | dev_item = &sb->dev_item; | |
f2984462 CM |
1716 | list_for_each(cur, head) { |
1717 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1718 | if (!dev->bdev) { |
1719 | total_errors++; | |
1720 | continue; | |
1721 | } | |
1722 | if (!dev->in_fs_metadata) | |
1723 | continue; | |
1724 | ||
a061fc8d CM |
1725 | btrfs_set_stack_device_type(dev_item, dev->type); |
1726 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
1727 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
1728 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
1729 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
1730 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
1731 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
1732 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
1733 | flags = btrfs_super_flags(sb); | |
1734 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
1735 | ||
1736 | ||
1737 | crc = ~(u32)0; | |
1738 | crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc, | |
1739 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
1740 | btrfs_csum_final(crc, sb->csum); | |
1741 | ||
1742 | bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096, | |
f2984462 CM |
1743 | BTRFS_SUPER_INFO_SIZE); |
1744 | ||
a061fc8d | 1745 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
f2984462 CM |
1746 | dev->pending_io = bh; |
1747 | ||
1748 | get_bh(bh); | |
1749 | set_buffer_uptodate(bh); | |
1750 | lock_buffer(bh); | |
1751 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
1752 | ||
1753 | if (do_barriers && dev->barriers) { | |
1754 | ret = submit_bh(WRITE_BARRIER, bh); | |
1755 | if (ret == -EOPNOTSUPP) { | |
1756 | printk("btrfs: disabling barriers on dev %s\n", | |
1757 | dev->name); | |
1758 | set_buffer_uptodate(bh); | |
1759 | dev->barriers = 0; | |
1760 | get_bh(bh); | |
1761 | lock_buffer(bh); | |
1762 | ret = submit_bh(WRITE, bh); | |
1763 | } | |
1764 | } else { | |
1765 | ret = submit_bh(WRITE, bh); | |
1766 | } | |
a236aed1 CM |
1767 | if (ret) |
1768 | total_errors++; | |
f2984462 | 1769 | } |
a236aed1 CM |
1770 | if (total_errors > max_errors) { |
1771 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1772 | BUG(); | |
1773 | } | |
1774 | total_errors = 0; | |
f2984462 CM |
1775 | |
1776 | list_for_each(cur, head) { | |
1777 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1778 | if (!dev->bdev) |
1779 | continue; | |
1780 | if (!dev->in_fs_metadata) | |
1781 | continue; | |
1782 | ||
f2984462 CM |
1783 | BUG_ON(!dev->pending_io); |
1784 | bh = dev->pending_io; | |
1785 | wait_on_buffer(bh); | |
1786 | if (!buffer_uptodate(dev->pending_io)) { | |
1787 | if (do_barriers && dev->barriers) { | |
1788 | printk("btrfs: disabling barriers on dev %s\n", | |
1789 | dev->name); | |
1790 | set_buffer_uptodate(bh); | |
1791 | get_bh(bh); | |
1792 | lock_buffer(bh); | |
1793 | dev->barriers = 0; | |
1794 | ret = submit_bh(WRITE, bh); | |
1795 | BUG_ON(ret); | |
1796 | wait_on_buffer(bh); | |
1259ab75 CM |
1797 | if (!buffer_uptodate(bh)) |
1798 | total_errors++; | |
f2984462 | 1799 | } else { |
a236aed1 | 1800 | total_errors++; |
f2984462 CM |
1801 | } |
1802 | ||
1803 | } | |
1804 | dev->pending_io = NULL; | |
1805 | brelse(bh); | |
1806 | } | |
a236aed1 CM |
1807 | if (total_errors > max_errors) { |
1808 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1809 | BUG(); | |
1810 | } | |
f2984462 CM |
1811 | return 0; |
1812 | } | |
1813 | ||
e089f05c | 1814 | int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root |
79154b1b | 1815 | *root) |
eb60ceac | 1816 | { |
e66f709b | 1817 | int ret; |
5f39d397 | 1818 | |
f2984462 | 1819 | ret = write_all_supers(root); |
5f39d397 | 1820 | return ret; |
cfaa7295 CM |
1821 | } |
1822 | ||
5eda7b5e | 1823 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
1824 | { |
1825 | radix_tree_delete(&fs_info->fs_roots_radix, | |
1826 | (unsigned long)root->root_key.objectid); | |
b99aa6cb CM |
1827 | if (root->in_sysfs) |
1828 | btrfs_sysfs_del_root(root); | |
2619ba1f CM |
1829 | if (root->inode) |
1830 | iput(root->inode); | |
1831 | if (root->node) | |
5f39d397 | 1832 | free_extent_buffer(root->node); |
2619ba1f | 1833 | if (root->commit_root) |
5f39d397 | 1834 | free_extent_buffer(root->commit_root); |
58176a96 JB |
1835 | if (root->name) |
1836 | kfree(root->name); | |
2619ba1f CM |
1837 | kfree(root); |
1838 | return 0; | |
1839 | } | |
1840 | ||
35b7e476 | 1841 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
1842 | { |
1843 | int ret; | |
1844 | struct btrfs_root *gang[8]; | |
1845 | int i; | |
1846 | ||
1847 | while(1) { | |
1848 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1849 | (void **)gang, 0, | |
1850 | ARRAY_SIZE(gang)); | |
1851 | if (!ret) | |
1852 | break; | |
2619ba1f | 1853 | for (i = 0; i < ret; i++) |
5eda7b5e | 1854 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
1855 | } |
1856 | return 0; | |
1857 | } | |
b4100d64 | 1858 | |
e20d96d6 | 1859 | int close_ctree(struct btrfs_root *root) |
cfaa7295 | 1860 | { |
3768f368 | 1861 | int ret; |
e089f05c | 1862 | struct btrfs_trans_handle *trans; |
0f7d52f4 | 1863 | struct btrfs_fs_info *fs_info = root->fs_info; |
e089f05c | 1864 | |
facda1e7 | 1865 | fs_info->closing = 1; |
a2135011 CM |
1866 | smp_mb(); |
1867 | ||
a74a4b97 CM |
1868 | kthread_stop(root->fs_info->transaction_kthread); |
1869 | kthread_stop(root->fs_info->cleaner_kthread); | |
1870 | ||
a74a4b97 | 1871 | btrfs_clean_old_snapshots(root); |
79154b1b | 1872 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 1873 | ret = btrfs_commit_transaction(trans, root); |
79154b1b CM |
1874 | /* run commit again to drop the original snapshot */ |
1875 | trans = btrfs_start_transaction(root, 1); | |
1876 | btrfs_commit_transaction(trans, root); | |
1877 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 1878 | BUG_ON(ret); |
d6bfde87 | 1879 | |
79154b1b | 1880 | write_ctree_super(NULL, root); |
0f7d52f4 | 1881 | |
b0c68f8b CM |
1882 | if (fs_info->delalloc_bytes) { |
1883 | printk("btrfs: at unmount delalloc count %Lu\n", | |
1884 | fs_info->delalloc_bytes); | |
1885 | } | |
31153d81 YZ |
1886 | if (fs_info->total_ref_cache_size) { |
1887 | printk("btrfs: at umount reference cache size %Lu\n", | |
1888 | fs_info->total_ref_cache_size); | |
1889 | } | |
bcc63abb | 1890 | |
0f7d52f4 | 1891 | if (fs_info->extent_root->node) |
5f39d397 | 1892 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 1893 | |
0f7d52f4 | 1894 | if (fs_info->tree_root->node) |
5f39d397 | 1895 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 1896 | |
0b86a832 CM |
1897 | if (root->fs_info->chunk_root->node); |
1898 | free_extent_buffer(root->fs_info->chunk_root->node); | |
1899 | ||
1900 | if (root->fs_info->dev_root->node); | |
1901 | free_extent_buffer(root->fs_info->dev_root->node); | |
1902 | ||
9078a3e1 | 1903 | btrfs_free_block_groups(root->fs_info); |
4ca8b41e | 1904 | fs_info->closing = 2; |
0f7d52f4 | 1905 | del_fs_roots(fs_info); |
d10c5f31 CM |
1906 | |
1907 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
1908 | ||
db94535d | 1909 | truncate_inode_pages(fs_info->btree_inode->i_mapping, 0); |
9ad6b7bc | 1910 | |
247e743c | 1911 | btrfs_stop_workers(&fs_info->fixup_workers); |
8b712842 CM |
1912 | btrfs_stop_workers(&fs_info->workers); |
1913 | btrfs_stop_workers(&fs_info->endio_workers); | |
e6dcd2dc | 1914 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1915 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 1916 | |
db94535d | 1917 | iput(fs_info->btree_inode); |
19c00ddc CM |
1918 | #if 0 |
1919 | while(!list_empty(&fs_info->hashers)) { | |
1920 | struct btrfs_hasher *hasher; | |
1921 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
1922 | hashers); | |
1923 | list_del(&hasher->hashers); | |
1924 | crypto_free_hash(&fs_info->hash_tfm); | |
1925 | kfree(hasher); | |
1926 | } | |
1927 | #endif | |
dfe25020 | 1928 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 1929 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 1930 | |
51ebc0d3 | 1931 | #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23) |
04160088 | 1932 | bdi_destroy(&fs_info->bdi); |
b248a415 | 1933 | #endif |
0b86a832 | 1934 | |
0f7d52f4 | 1935 | kfree(fs_info->extent_root); |
0f7d52f4 | 1936 | kfree(fs_info->tree_root); |
0b86a832 CM |
1937 | kfree(fs_info->chunk_root); |
1938 | kfree(fs_info->dev_root); | |
eb60ceac CM |
1939 | return 0; |
1940 | } | |
1941 | ||
1259ab75 | 1942 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 1943 | { |
1259ab75 | 1944 | int ret; |
810191ff | 1945 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
1946 | |
1947 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
1948 | if (!ret) | |
1949 | return ret; | |
1950 | ||
1951 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
1952 | parent_transid); | |
1953 | return !ret; | |
5f39d397 CM |
1954 | } |
1955 | ||
1956 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 1957 | { |
810191ff | 1958 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 1959 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
1960 | buf); |
1961 | } | |
6702ed49 | 1962 | |
5f39d397 CM |
1963 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
1964 | { | |
810191ff | 1965 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
1966 | u64 transid = btrfs_header_generation(buf); |
1967 | struct inode *btree_inode = root->fs_info->btree_inode; | |
6702ed49 | 1968 | |
925baedd | 1969 | WARN_ON(!btrfs_tree_locked(buf)); |
ccd467d6 CM |
1970 | if (transid != root->fs_info->generation) { |
1971 | printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", | |
db94535d | 1972 | (unsigned long long)buf->start, |
ccd467d6 CM |
1973 | transid, root->fs_info->generation); |
1974 | WARN_ON(1); | |
1975 | } | |
d1310b2e | 1976 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); |
eb60ceac CM |
1977 | } |
1978 | ||
d3c2fdcf | 1979 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 1980 | { |
188de649 CM |
1981 | /* |
1982 | * looks as though older kernels can get into trouble with | |
1983 | * this code, they end up stuck in balance_dirty_pages forever | |
1984 | */ | |
d6bfde87 CM |
1985 | struct extent_io_tree *tree; |
1986 | u64 num_dirty; | |
1987 | u64 start = 0; | |
b64a2851 | 1988 | unsigned long thresh = 96 * 1024 * 1024; |
d6bfde87 CM |
1989 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
1990 | ||
b64a2851 | 1991 | if (current_is_pdflush() || current->flags & PF_MEMALLOC) |
d6bfde87 CM |
1992 | return; |
1993 | ||
1994 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
1995 | thresh, EXTENT_DIRTY); | |
1996 | if (num_dirty > thresh) { | |
1997 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 1998 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 1999 | } |
188de649 | 2000 | return; |
35b7e476 | 2001 | } |
6b80053d | 2002 | |
ca7a79ad | 2003 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 2004 | { |
810191ff | 2005 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 2006 | int ret; |
ca7a79ad | 2007 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
2008 | if (ret == 0) { |
2009 | buf->flags |= EXTENT_UPTODATE; | |
2010 | } | |
2011 | return ret; | |
6b80053d | 2012 | } |
0da5468f | 2013 | |
4bef0848 CM |
2014 | int btree_lock_page_hook(struct page *page) |
2015 | { | |
2016 | struct inode *inode = page->mapping->host; | |
2017 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2018 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
2019 | struct extent_buffer *eb; | |
2020 | unsigned long len; | |
2021 | u64 bytenr = page_offset(page); | |
2022 | ||
2023 | if (page->private == EXTENT_PAGE_PRIVATE) | |
2024 | goto out; | |
2025 | ||
2026 | len = page->private >> 2; | |
2027 | eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS); | |
2028 | if (!eb) | |
2029 | goto out; | |
2030 | ||
2031 | btrfs_tree_lock(eb); | |
2032 | spin_lock(&root->fs_info->hash_lock); | |
2033 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); | |
2034 | spin_unlock(&root->fs_info->hash_lock); | |
2035 | btrfs_tree_unlock(eb); | |
2036 | free_extent_buffer(eb); | |
2037 | out: | |
2038 | lock_page(page); | |
2039 | return 0; | |
2040 | } | |
2041 | ||
d1310b2e | 2042 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 2043 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
0da5468f | 2044 | .writepage_io_hook = btree_writepage_io_hook, |
ce9adaa5 | 2045 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 2046 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
2047 | /* note we're sharing with inode.c for the merge bio hook */ |
2048 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 2049 | }; |