]>
Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
a74a4b97 | 19 | #include <linux/version.h> |
e20d96d6 | 20 | #include <linux/fs.h> |
d98237b3 | 21 | #include <linux/blkdev.h> |
87cbda5c | 22 | #include <linux/scatterlist.h> |
22b0ebda | 23 | #include <linux/swap.h> |
0f7d52f4 | 24 | #include <linux/radix-tree.h> |
35b7e476 | 25 | #include <linux/writeback.h> |
5f39d397 | 26 | #include <linux/buffer_head.h> // for block_sync_page |
ce9adaa5 | 27 | #include <linux/workqueue.h> |
a74a4b97 | 28 | #include <linux/kthread.h> |
a74a4b97 | 29 | # include <linux/freezer.h> |
a5eb62e3 | 30 | #include "crc32c.h" |
eb60ceac CM |
31 | #include "ctree.h" |
32 | #include "disk-io.h" | |
e089f05c | 33 | #include "transaction.h" |
0f7d52f4 | 34 | #include "btrfs_inode.h" |
0b86a832 | 35 | #include "volumes.h" |
db94535d | 36 | #include "print-tree.h" |
8b712842 | 37 | #include "async-thread.h" |
925baedd | 38 | #include "locking.h" |
017e5369 | 39 | #include "ref-cache.h" |
e02119d5 | 40 | #include "tree-log.h" |
eb60ceac | 41 | |
5f39d397 CM |
42 | #if 0 |
43 | static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf) | |
7eccb903 | 44 | { |
5f39d397 CM |
45 | if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) { |
46 | printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n", | |
47 | (unsigned long long)extent_buffer_blocknr(buf), | |
48 | (unsigned long long)btrfs_header_blocknr(buf)); | |
39279cc3 | 49 | return 1; |
d98237b3 | 50 | } |
9a8dd150 | 51 | return 0; |
eb60ceac | 52 | } |
5f39d397 | 53 | #endif |
eb60ceac | 54 | |
d1310b2e | 55 | static struct extent_io_ops btree_extent_io_ops; |
8b712842 | 56 | static void end_workqueue_fn(struct btrfs_work *work); |
ce9adaa5 | 57 | |
d352ac68 CM |
58 | /* |
59 | * end_io_wq structs are used to do processing in task context when an IO is | |
60 | * complete. This is used during reads to verify checksums, and it is used | |
61 | * by writes to insert metadata for new file extents after IO is complete. | |
62 | */ | |
ce9adaa5 CM |
63 | struct end_io_wq { |
64 | struct bio *bio; | |
65 | bio_end_io_t *end_io; | |
66 | void *private; | |
67 | struct btrfs_fs_info *info; | |
68 | int error; | |
22c59948 | 69 | int metadata; |
ce9adaa5 | 70 | struct list_head list; |
8b712842 | 71 | struct btrfs_work work; |
ce9adaa5 | 72 | }; |
0da5468f | 73 | |
d352ac68 CM |
74 | /* |
75 | * async submit bios are used to offload expensive checksumming | |
76 | * onto the worker threads. They checksum file and metadata bios | |
77 | * just before they are sent down the IO stack. | |
78 | */ | |
44b8bd7e CM |
79 | struct async_submit_bio { |
80 | struct inode *inode; | |
81 | struct bio *bio; | |
82 | struct list_head list; | |
4a69a410 CM |
83 | extent_submit_bio_hook_t *submit_bio_start; |
84 | extent_submit_bio_hook_t *submit_bio_done; | |
44b8bd7e CM |
85 | int rw; |
86 | int mirror_num; | |
c8b97818 | 87 | unsigned long bio_flags; |
8b712842 | 88 | struct btrfs_work work; |
44b8bd7e CM |
89 | }; |
90 | ||
d352ac68 CM |
91 | /* |
92 | * extents on the btree inode are pretty simple, there's one extent | |
93 | * that covers the entire device | |
94 | */ | |
5f39d397 | 95 | struct extent_map *btree_get_extent(struct inode *inode, struct page *page, |
d1310b2e | 96 | size_t page_offset, u64 start, u64 len, |
5f39d397 | 97 | int create) |
7eccb903 | 98 | { |
5f39d397 CM |
99 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; |
100 | struct extent_map *em; | |
101 | int ret; | |
102 | ||
d1310b2e CM |
103 | spin_lock(&em_tree->lock); |
104 | em = lookup_extent_mapping(em_tree, start, len); | |
a061fc8d CM |
105 | if (em) { |
106 | em->bdev = | |
107 | BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; | |
108 | spin_unlock(&em_tree->lock); | |
5f39d397 | 109 | goto out; |
a061fc8d CM |
110 | } |
111 | spin_unlock(&em_tree->lock); | |
7b13b7b1 | 112 | |
5f39d397 CM |
113 | em = alloc_extent_map(GFP_NOFS); |
114 | if (!em) { | |
115 | em = ERR_PTR(-ENOMEM); | |
116 | goto out; | |
117 | } | |
118 | em->start = 0; | |
0afbaf8c | 119 | em->len = (u64)-1; |
c8b97818 | 120 | em->block_len = (u64)-1; |
5f39d397 | 121 | em->block_start = 0; |
a061fc8d | 122 | em->bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; |
d1310b2e CM |
123 | |
124 | spin_lock(&em_tree->lock); | |
5f39d397 CM |
125 | ret = add_extent_mapping(em_tree, em); |
126 | if (ret == -EEXIST) { | |
0afbaf8c CM |
127 | u64 failed_start = em->start; |
128 | u64 failed_len = em->len; | |
129 | ||
130 | printk("failed to insert %Lu %Lu -> %Lu into tree\n", | |
131 | em->start, em->len, em->block_start); | |
5f39d397 | 132 | free_extent_map(em); |
7b13b7b1 | 133 | em = lookup_extent_mapping(em_tree, start, len); |
0afbaf8c CM |
134 | if (em) { |
135 | printk("after failing, found %Lu %Lu %Lu\n", | |
136 | em->start, em->len, em->block_start); | |
7b13b7b1 | 137 | ret = 0; |
0afbaf8c CM |
138 | } else { |
139 | em = lookup_extent_mapping(em_tree, failed_start, | |
140 | failed_len); | |
141 | if (em) { | |
142 | printk("double failure lookup gives us " | |
143 | "%Lu %Lu -> %Lu\n", em->start, | |
144 | em->len, em->block_start); | |
145 | free_extent_map(em); | |
146 | } | |
7b13b7b1 | 147 | ret = -EIO; |
0afbaf8c | 148 | } |
5f39d397 | 149 | } else if (ret) { |
7b13b7b1 CM |
150 | free_extent_map(em); |
151 | em = NULL; | |
5f39d397 | 152 | } |
7b13b7b1 CM |
153 | spin_unlock(&em_tree->lock); |
154 | ||
155 | if (ret) | |
156 | em = ERR_PTR(ret); | |
5f39d397 CM |
157 | out: |
158 | return em; | |
7eccb903 CM |
159 | } |
160 | ||
19c00ddc CM |
161 | u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len) |
162 | { | |
a5eb62e3 | 163 | return btrfs_crc32c(seed, data, len); |
19c00ddc CM |
164 | } |
165 | ||
166 | void btrfs_csum_final(u32 crc, char *result) | |
167 | { | |
168 | *(__le32 *)result = ~cpu_to_le32(crc); | |
169 | } | |
170 | ||
d352ac68 CM |
171 | /* |
172 | * compute the csum for a btree block, and either verify it or write it | |
173 | * into the csum field of the block. | |
174 | */ | |
19c00ddc CM |
175 | static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf, |
176 | int verify) | |
177 | { | |
178 | char result[BTRFS_CRC32_SIZE]; | |
179 | unsigned long len; | |
180 | unsigned long cur_len; | |
181 | unsigned long offset = BTRFS_CSUM_SIZE; | |
182 | char *map_token = NULL; | |
183 | char *kaddr; | |
184 | unsigned long map_start; | |
185 | unsigned long map_len; | |
186 | int err; | |
187 | u32 crc = ~(u32)0; | |
188 | ||
189 | len = buf->len - offset; | |
190 | while(len > 0) { | |
191 | err = map_private_extent_buffer(buf, offset, 32, | |
192 | &map_token, &kaddr, | |
193 | &map_start, &map_len, KM_USER0); | |
194 | if (err) { | |
195 | printk("failed to map extent buffer! %lu\n", | |
196 | offset); | |
197 | return 1; | |
198 | } | |
199 | cur_len = min(len, map_len - (offset - map_start)); | |
200 | crc = btrfs_csum_data(root, kaddr + offset - map_start, | |
201 | crc, cur_len); | |
202 | len -= cur_len; | |
203 | offset += cur_len; | |
204 | unmap_extent_buffer(buf, map_token, KM_USER0); | |
205 | } | |
206 | btrfs_csum_final(crc, result); | |
207 | ||
208 | if (verify) { | |
e4204ded | 209 | /* FIXME, this is not good */ |
63b10fc4 | 210 | if (memcmp_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE)) { |
e4204ded CM |
211 | u32 val; |
212 | u32 found = 0; | |
213 | memcpy(&found, result, BTRFS_CRC32_SIZE); | |
214 | ||
215 | read_extent_buffer(buf, &val, 0, BTRFS_CRC32_SIZE); | |
216 | printk("btrfs: %s checksum verify failed on %llu " | |
2dd3e67b | 217 | "wanted %X found %X level %d\n", |
19c00ddc | 218 | root->fs_info->sb->s_id, |
2dd3e67b | 219 | buf->start, val, found, btrfs_header_level(buf)); |
19c00ddc CM |
220 | return 1; |
221 | } | |
222 | } else { | |
223 | write_extent_buffer(buf, result, 0, BTRFS_CRC32_SIZE); | |
224 | } | |
225 | return 0; | |
226 | } | |
227 | ||
d352ac68 CM |
228 | /* |
229 | * we can't consider a given block up to date unless the transid of the | |
230 | * block matches the transid in the parent node's pointer. This is how we | |
231 | * detect blocks that either didn't get written at all or got written | |
232 | * in the wrong place. | |
233 | */ | |
1259ab75 CM |
234 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
235 | struct extent_buffer *eb, u64 parent_transid) | |
236 | { | |
237 | int ret; | |
238 | ||
239 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
240 | return 0; | |
241 | ||
242 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, GFP_NOFS); | |
243 | if (extent_buffer_uptodate(io_tree, eb) && | |
244 | btrfs_header_generation(eb) == parent_transid) { | |
245 | ret = 0; | |
246 | goto out; | |
247 | } | |
248 | printk("parent transid verify failed on %llu wanted %llu found %llu\n", | |
249 | (unsigned long long)eb->start, | |
250 | (unsigned long long)parent_transid, | |
251 | (unsigned long long)btrfs_header_generation(eb)); | |
252 | ret = 1; | |
1259ab75 | 253 | clear_extent_buffer_uptodate(io_tree, eb); |
33958dc6 | 254 | out: |
1259ab75 CM |
255 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
256 | GFP_NOFS); | |
257 | return ret; | |
1259ab75 CM |
258 | } |
259 | ||
d352ac68 CM |
260 | /* |
261 | * helper to read a given tree block, doing retries as required when | |
262 | * the checksums don't match and we have alternate mirrors to try. | |
263 | */ | |
f188591e CM |
264 | static int btree_read_extent_buffer_pages(struct btrfs_root *root, |
265 | struct extent_buffer *eb, | |
ca7a79ad | 266 | u64 start, u64 parent_transid) |
f188591e CM |
267 | { |
268 | struct extent_io_tree *io_tree; | |
269 | int ret; | |
270 | int num_copies = 0; | |
271 | int mirror_num = 0; | |
272 | ||
273 | io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; | |
274 | while (1) { | |
275 | ret = read_extent_buffer_pages(io_tree, eb, start, 1, | |
276 | btree_get_extent, mirror_num); | |
1259ab75 CM |
277 | if (!ret && |
278 | !verify_parent_transid(io_tree, eb, parent_transid)) | |
f188591e | 279 | return ret; |
a1b32a59 | 280 | printk("read extent buffer pages failed with ret %d mirror no %d\n", ret, mirror_num); |
f188591e CM |
281 | num_copies = btrfs_num_copies(&root->fs_info->mapping_tree, |
282 | eb->start, eb->len); | |
4235298e | 283 | if (num_copies == 1) |
f188591e | 284 | return ret; |
4235298e | 285 | |
f188591e | 286 | mirror_num++; |
4235298e | 287 | if (mirror_num > num_copies) |
f188591e | 288 | return ret; |
f188591e | 289 | } |
f188591e CM |
290 | return -EIO; |
291 | } | |
19c00ddc | 292 | |
d352ac68 CM |
293 | /* |
294 | * checksum a dirty tree block before IO. This has extra checks to make | |
295 | * sure we only fill in the checksum field in the first page of a multi-page block | |
296 | */ | |
19c00ddc CM |
297 | int csum_dirty_buffer(struct btrfs_root *root, struct page *page) |
298 | { | |
d1310b2e | 299 | struct extent_io_tree *tree; |
35ebb934 | 300 | u64 start = (u64)page->index << PAGE_CACHE_SHIFT; |
19c00ddc CM |
301 | u64 found_start; |
302 | int found_level; | |
303 | unsigned long len; | |
304 | struct extent_buffer *eb; | |
f188591e CM |
305 | int ret; |
306 | ||
d1310b2e | 307 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
19c00ddc CM |
308 | |
309 | if (page->private == EXTENT_PAGE_PRIVATE) | |
310 | goto out; | |
311 | if (!page->private) | |
312 | goto out; | |
313 | len = page->private >> 2; | |
314 | if (len == 0) { | |
315 | WARN_ON(1); | |
316 | } | |
317 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
ca7a79ad CM |
318 | ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, |
319 | btrfs_header_generation(eb)); | |
f188591e | 320 | BUG_ON(ret); |
19c00ddc CM |
321 | found_start = btrfs_header_bytenr(eb); |
322 | if (found_start != start) { | |
323 | printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n", | |
324 | start, found_start, len); | |
55c69072 CM |
325 | WARN_ON(1); |
326 | goto err; | |
327 | } | |
328 | if (eb->first_page != page) { | |
329 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
330 | page->index); | |
331 | WARN_ON(1); | |
332 | goto err; | |
333 | } | |
334 | if (!PageUptodate(page)) { | |
335 | printk("csum not up to date page %lu\n", page->index); | |
336 | WARN_ON(1); | |
337 | goto err; | |
19c00ddc CM |
338 | } |
339 | found_level = btrfs_header_level(eb); | |
4bef0848 | 340 | |
19c00ddc | 341 | csum_tree_block(root, eb, 0); |
55c69072 | 342 | err: |
19c00ddc CM |
343 | free_extent_buffer(eb); |
344 | out: | |
345 | return 0; | |
346 | } | |
347 | ||
2b82032c YZ |
348 | static int check_tree_block_fsid(struct btrfs_root *root, |
349 | struct extent_buffer *eb) | |
350 | { | |
351 | struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices; | |
352 | u8 fsid[BTRFS_UUID_SIZE]; | |
353 | int ret = 1; | |
354 | ||
355 | read_extent_buffer(eb, fsid, (unsigned long)btrfs_header_fsid(eb), | |
356 | BTRFS_FSID_SIZE); | |
357 | while (fs_devices) { | |
358 | if (!memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE)) { | |
359 | ret = 0; | |
360 | break; | |
361 | } | |
362 | fs_devices = fs_devices->seed; | |
363 | } | |
364 | return ret; | |
365 | } | |
366 | ||
ce9adaa5 CM |
367 | int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end, |
368 | struct extent_state *state) | |
369 | { | |
370 | struct extent_io_tree *tree; | |
371 | u64 found_start; | |
372 | int found_level; | |
373 | unsigned long len; | |
374 | struct extent_buffer *eb; | |
375 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; | |
f188591e | 376 | int ret = 0; |
ce9adaa5 CM |
377 | |
378 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
379 | if (page->private == EXTENT_PAGE_PRIVATE) | |
380 | goto out; | |
381 | if (!page->private) | |
382 | goto out; | |
383 | len = page->private >> 2; | |
384 | if (len == 0) { | |
385 | WARN_ON(1); | |
386 | } | |
387 | eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS); | |
f188591e | 388 | |
ce9adaa5 | 389 | found_start = btrfs_header_bytenr(eb); |
23a07867 | 390 | if (found_start != start) { |
a1b32a59 CM |
391 | printk("bad tree block start %llu %llu\n", |
392 | (unsigned long long)found_start, | |
393 | (unsigned long long)eb->start); | |
f188591e | 394 | ret = -EIO; |
ce9adaa5 CM |
395 | goto err; |
396 | } | |
397 | if (eb->first_page != page) { | |
398 | printk("bad first page %lu %lu\n", eb->first_page->index, | |
399 | page->index); | |
400 | WARN_ON(1); | |
f188591e | 401 | ret = -EIO; |
ce9adaa5 CM |
402 | goto err; |
403 | } | |
2b82032c | 404 | if (check_tree_block_fsid(root, eb)) { |
1259ab75 CM |
405 | printk("bad fsid on block %Lu\n", eb->start); |
406 | ret = -EIO; | |
407 | goto err; | |
408 | } | |
ce9adaa5 CM |
409 | found_level = btrfs_header_level(eb); |
410 | ||
411 | ret = csum_tree_block(root, eb, 1); | |
f188591e CM |
412 | if (ret) |
413 | ret = -EIO; | |
ce9adaa5 CM |
414 | |
415 | end = min_t(u64, eb->len, PAGE_CACHE_SIZE); | |
416 | end = eb->start + end - 1; | |
ce9adaa5 CM |
417 | err: |
418 | free_extent_buffer(eb); | |
419 | out: | |
f188591e | 420 | return ret; |
ce9adaa5 CM |
421 | } |
422 | ||
ce9adaa5 | 423 | static void end_workqueue_bio(struct bio *bio, int err) |
ce9adaa5 CM |
424 | { |
425 | struct end_io_wq *end_io_wq = bio->bi_private; | |
426 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 427 | |
ce9adaa5 | 428 | fs_info = end_io_wq->info; |
ce9adaa5 | 429 | end_io_wq->error = err; |
8b712842 CM |
430 | end_io_wq->work.func = end_workqueue_fn; |
431 | end_io_wq->work.flags = 0; | |
e6dcd2dc CM |
432 | if (bio->bi_rw & (1 << BIO_RW)) |
433 | btrfs_queue_worker(&fs_info->endio_write_workers, | |
434 | &end_io_wq->work); | |
435 | else | |
436 | btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); | |
ce9adaa5 CM |
437 | } |
438 | ||
22c59948 CM |
439 | int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, |
440 | int metadata) | |
0b86a832 | 441 | { |
ce9adaa5 | 442 | struct end_io_wq *end_io_wq; |
ce9adaa5 CM |
443 | end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS); |
444 | if (!end_io_wq) | |
445 | return -ENOMEM; | |
446 | ||
447 | end_io_wq->private = bio->bi_private; | |
448 | end_io_wq->end_io = bio->bi_end_io; | |
22c59948 | 449 | end_io_wq->info = info; |
ce9adaa5 CM |
450 | end_io_wq->error = 0; |
451 | end_io_wq->bio = bio; | |
22c59948 | 452 | end_io_wq->metadata = metadata; |
ce9adaa5 CM |
453 | |
454 | bio->bi_private = end_io_wq; | |
455 | bio->bi_end_io = end_workqueue_bio; | |
22c59948 CM |
456 | return 0; |
457 | } | |
458 | ||
b64a2851 | 459 | unsigned long btrfs_async_submit_limit(struct btrfs_fs_info *info) |
0986fe9e | 460 | { |
4854ddd0 CM |
461 | unsigned long limit = min_t(unsigned long, |
462 | info->workers.max_workers, | |
463 | info->fs_devices->open_devices); | |
464 | return 256 * limit; | |
465 | } | |
0986fe9e | 466 | |
4854ddd0 CM |
467 | int btrfs_congested_async(struct btrfs_fs_info *info, int iodone) |
468 | { | |
b64a2851 CM |
469 | return atomic_read(&info->nr_async_bios) > |
470 | btrfs_async_submit_limit(info); | |
0986fe9e CM |
471 | } |
472 | ||
4a69a410 CM |
473 | static void run_one_async_start(struct btrfs_work *work) |
474 | { | |
475 | struct btrfs_fs_info *fs_info; | |
476 | struct async_submit_bio *async; | |
477 | ||
478 | async = container_of(work, struct async_submit_bio, work); | |
479 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
480 | async->submit_bio_start(async->inode, async->rw, async->bio, | |
481 | async->mirror_num, async->bio_flags); | |
482 | } | |
483 | ||
484 | static void run_one_async_done(struct btrfs_work *work) | |
8b712842 CM |
485 | { |
486 | struct btrfs_fs_info *fs_info; | |
487 | struct async_submit_bio *async; | |
4854ddd0 | 488 | int limit; |
8b712842 CM |
489 | |
490 | async = container_of(work, struct async_submit_bio, work); | |
491 | fs_info = BTRFS_I(async->inode)->root->fs_info; | |
4854ddd0 | 492 | |
b64a2851 | 493 | limit = btrfs_async_submit_limit(fs_info); |
4854ddd0 CM |
494 | limit = limit * 2 / 3; |
495 | ||
8b712842 | 496 | atomic_dec(&fs_info->nr_async_submits); |
0986fe9e | 497 | |
b64a2851 CM |
498 | if (atomic_read(&fs_info->nr_async_submits) < limit && |
499 | waitqueue_active(&fs_info->async_submit_wait)) | |
4854ddd0 CM |
500 | wake_up(&fs_info->async_submit_wait); |
501 | ||
4a69a410 | 502 | async->submit_bio_done(async->inode, async->rw, async->bio, |
c8b97818 | 503 | async->mirror_num, async->bio_flags); |
4a69a410 CM |
504 | } |
505 | ||
506 | static void run_one_async_free(struct btrfs_work *work) | |
507 | { | |
508 | struct async_submit_bio *async; | |
509 | ||
510 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
511 | kfree(async); |
512 | } | |
513 | ||
44b8bd7e CM |
514 | int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode, |
515 | int rw, struct bio *bio, int mirror_num, | |
c8b97818 | 516 | unsigned long bio_flags, |
4a69a410 CM |
517 | extent_submit_bio_hook_t *submit_bio_start, |
518 | extent_submit_bio_hook_t *submit_bio_done) | |
44b8bd7e CM |
519 | { |
520 | struct async_submit_bio *async; | |
b64a2851 | 521 | int limit = btrfs_async_submit_limit(fs_info); |
44b8bd7e CM |
522 | |
523 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
524 | if (!async) | |
525 | return -ENOMEM; | |
526 | ||
527 | async->inode = inode; | |
528 | async->rw = rw; | |
529 | async->bio = bio; | |
530 | async->mirror_num = mirror_num; | |
4a69a410 CM |
531 | async->submit_bio_start = submit_bio_start; |
532 | async->submit_bio_done = submit_bio_done; | |
533 | ||
534 | async->work.func = run_one_async_start; | |
535 | async->work.ordered_func = run_one_async_done; | |
536 | async->work.ordered_free = run_one_async_free; | |
537 | ||
8b712842 | 538 | async->work.flags = 0; |
c8b97818 | 539 | async->bio_flags = bio_flags; |
8c8bee1d CM |
540 | |
541 | while(atomic_read(&fs_info->async_submit_draining) && | |
542 | atomic_read(&fs_info->nr_async_submits)) { | |
543 | wait_event(fs_info->async_submit_wait, | |
544 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
545 | } | |
546 | ||
cb03c743 | 547 | atomic_inc(&fs_info->nr_async_submits); |
8b712842 | 548 | btrfs_queue_worker(&fs_info->workers, &async->work); |
4854ddd0 | 549 | |
9473f16c CM |
550 | if (atomic_read(&fs_info->nr_async_submits) > limit) { |
551 | wait_event_timeout(fs_info->async_submit_wait, | |
4854ddd0 CM |
552 | (atomic_read(&fs_info->nr_async_submits) < limit), |
553 | HZ/10); | |
9473f16c CM |
554 | |
555 | wait_event_timeout(fs_info->async_submit_wait, | |
556 | (atomic_read(&fs_info->nr_async_bios) < limit), | |
557 | HZ/10); | |
558 | } | |
771ed689 CM |
559 | |
560 | while(atomic_read(&fs_info->async_submit_draining) && | |
561 | atomic_read(&fs_info->nr_async_submits)) { | |
562 | wait_event(fs_info->async_submit_wait, | |
563 | (atomic_read(&fs_info->nr_async_submits) == 0)); | |
564 | } | |
565 | ||
44b8bd7e CM |
566 | return 0; |
567 | } | |
568 | ||
ce3ed71a CM |
569 | static int btree_csum_one_bio(struct bio *bio) |
570 | { | |
571 | struct bio_vec *bvec = bio->bi_io_vec; | |
572 | int bio_index = 0; | |
573 | struct btrfs_root *root; | |
574 | ||
575 | WARN_ON(bio->bi_vcnt <= 0); | |
576 | while(bio_index < bio->bi_vcnt) { | |
577 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; | |
578 | csum_dirty_buffer(root, bvec->bv_page); | |
579 | bio_index++; | |
580 | bvec++; | |
581 | } | |
582 | return 0; | |
583 | } | |
584 | ||
4a69a410 CM |
585 | static int __btree_submit_bio_start(struct inode *inode, int rw, |
586 | struct bio *bio, int mirror_num, | |
587 | unsigned long bio_flags) | |
22c59948 | 588 | { |
8b712842 CM |
589 | /* |
590 | * when we're called for a write, we're already in the async | |
5443be45 | 591 | * submission context. Just jump into btrfs_map_bio |
8b712842 | 592 | */ |
4a69a410 CM |
593 | btree_csum_one_bio(bio); |
594 | return 0; | |
595 | } | |
22c59948 | 596 | |
4a69a410 CM |
597 | static int __btree_submit_bio_done(struct inode *inode, int rw, struct bio *bio, |
598 | int mirror_num, unsigned long bio_flags) | |
599 | { | |
8b712842 | 600 | /* |
4a69a410 CM |
601 | * when we're called for a write, we're already in the async |
602 | * submission context. Just jump into btrfs_map_bio | |
8b712842 | 603 | */ |
8b712842 | 604 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num, 1); |
0b86a832 CM |
605 | } |
606 | ||
44b8bd7e | 607 | static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio, |
c8b97818 | 608 | int mirror_num, unsigned long bio_flags) |
44b8bd7e | 609 | { |
8b712842 CM |
610 | /* |
611 | * kthread helpers are used to submit writes so that checksumming | |
612 | * can happen in parallel across all CPUs | |
613 | */ | |
44b8bd7e | 614 | if (!(rw & (1 << BIO_RW))) { |
4a69a410 CM |
615 | int ret; |
616 | /* | |
617 | * called for a read, do the setup so that checksum validation | |
618 | * can happen in the async kernel threads | |
619 | */ | |
620 | ret = btrfs_bio_wq_end_io(BTRFS_I(inode)->root->fs_info, | |
621 | bio, 1); | |
622 | BUG_ON(ret); | |
623 | ||
624 | return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, | |
6f3577bd | 625 | mirror_num, 0); |
44b8bd7e CM |
626 | } |
627 | return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info, | |
c8b97818 | 628 | inode, rw, bio, mirror_num, 0, |
4a69a410 CM |
629 | __btree_submit_bio_start, |
630 | __btree_submit_bio_done); | |
44b8bd7e CM |
631 | } |
632 | ||
0da5468f CM |
633 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
634 | { | |
d1310b2e CM |
635 | struct extent_io_tree *tree; |
636 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5443be45 CM |
637 | |
638 | if (current->flags & PF_MEMALLOC) { | |
639 | redirty_page_for_writepage(wbc, page); | |
640 | unlock_page(page); | |
641 | return 0; | |
642 | } | |
5f39d397 CM |
643 | return extent_write_full_page(tree, page, btree_get_extent, wbc); |
644 | } | |
0da5468f CM |
645 | |
646 | static int btree_writepages(struct address_space *mapping, | |
647 | struct writeback_control *wbc) | |
648 | { | |
d1310b2e CM |
649 | struct extent_io_tree *tree; |
650 | tree = &BTRFS_I(mapping->host)->io_tree; | |
d8d5f3e1 | 651 | if (wbc->sync_mode == WB_SYNC_NONE) { |
793955bc CM |
652 | u64 num_dirty; |
653 | u64 start = 0; | |
24ab9cd8 | 654 | unsigned long thresh = 32 * 1024 * 1024; |
448d640b CM |
655 | |
656 | if (wbc->for_kupdate) | |
657 | return 0; | |
658 | ||
1832a6d5 CM |
659 | num_dirty = count_range_bits(tree, &start, (u64)-1, |
660 | thresh, EXTENT_DIRTY); | |
793955bc CM |
661 | if (num_dirty < thresh) { |
662 | return 0; | |
663 | } | |
664 | } | |
0da5468f CM |
665 | return extent_writepages(tree, mapping, btree_get_extent, wbc); |
666 | } | |
667 | ||
5f39d397 CM |
668 | int btree_readpage(struct file *file, struct page *page) |
669 | { | |
d1310b2e CM |
670 | struct extent_io_tree *tree; |
671 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
672 | return extent_read_full_page(tree, page, btree_get_extent); |
673 | } | |
22b0ebda | 674 | |
70dec807 | 675 | static int btree_releasepage(struct page *page, gfp_t gfp_flags) |
5f39d397 | 676 | { |
d1310b2e CM |
677 | struct extent_io_tree *tree; |
678 | struct extent_map_tree *map; | |
5f39d397 | 679 | int ret; |
d98237b3 | 680 | |
98509cfc CM |
681 | if (PageWriteback(page) || PageDirty(page)) |
682 | return 0; | |
683 | ||
d1310b2e CM |
684 | tree = &BTRFS_I(page->mapping->host)->io_tree; |
685 | map = &BTRFS_I(page->mapping->host)->extent_tree; | |
6af118ce | 686 | |
7b13b7b1 | 687 | ret = try_release_extent_state(map, tree, page, gfp_flags); |
6af118ce CM |
688 | if (!ret) { |
689 | return 0; | |
690 | } | |
691 | ||
692 | ret = try_release_extent_buffer(tree, page); | |
5f39d397 CM |
693 | if (ret == 1) { |
694 | ClearPagePrivate(page); | |
695 | set_page_private(page, 0); | |
696 | page_cache_release(page); | |
697 | } | |
6af118ce | 698 | |
d98237b3 CM |
699 | return ret; |
700 | } | |
701 | ||
5f39d397 | 702 | static void btree_invalidatepage(struct page *page, unsigned long offset) |
d98237b3 | 703 | { |
d1310b2e CM |
704 | struct extent_io_tree *tree; |
705 | tree = &BTRFS_I(page->mapping->host)->io_tree; | |
5f39d397 CM |
706 | extent_invalidatepage(tree, page, offset); |
707 | btree_releasepage(page, GFP_NOFS); | |
9ad6b7bc | 708 | if (PagePrivate(page)) { |
6af118ce CM |
709 | printk("warning page private not zero on page %Lu\n", |
710 | page_offset(page)); | |
9ad6b7bc CM |
711 | ClearPagePrivate(page); |
712 | set_page_private(page, 0); | |
713 | page_cache_release(page); | |
714 | } | |
d98237b3 CM |
715 | } |
716 | ||
5f39d397 | 717 | #if 0 |
d98237b3 | 718 | static int btree_writepage(struct page *page, struct writeback_control *wbc) |
ed2ff2cb | 719 | { |
87cbda5c | 720 | struct buffer_head *bh; |
0f7d52f4 | 721 | struct btrfs_root *root = BTRFS_I(page->mapping->host)->root; |
87cbda5c | 722 | struct buffer_head *head; |
87cbda5c CM |
723 | if (!page_has_buffers(page)) { |
724 | create_empty_buffers(page, root->fs_info->sb->s_blocksize, | |
725 | (1 << BH_Dirty)|(1 << BH_Uptodate)); | |
726 | } | |
727 | head = page_buffers(page); | |
728 | bh = head; | |
729 | do { | |
730 | if (buffer_dirty(bh)) | |
731 | csum_tree_block(root, bh, 0); | |
732 | bh = bh->b_this_page; | |
733 | } while (bh != head); | |
d98237b3 | 734 | return block_write_full_page(page, btree_get_block, wbc); |
ed2ff2cb | 735 | } |
5f39d397 | 736 | #endif |
eb60ceac | 737 | |
d98237b3 CM |
738 | static struct address_space_operations btree_aops = { |
739 | .readpage = btree_readpage, | |
740 | .writepage = btree_writepage, | |
0da5468f | 741 | .writepages = btree_writepages, |
5f39d397 CM |
742 | .releasepage = btree_releasepage, |
743 | .invalidatepage = btree_invalidatepage, | |
d98237b3 CM |
744 | .sync_page = block_sync_page, |
745 | }; | |
746 | ||
ca7a79ad CM |
747 | int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, |
748 | u64 parent_transid) | |
090d1875 | 749 | { |
5f39d397 CM |
750 | struct extent_buffer *buf = NULL; |
751 | struct inode *btree_inode = root->fs_info->btree_inode; | |
de428b63 | 752 | int ret = 0; |
090d1875 | 753 | |
db94535d | 754 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); |
5f39d397 | 755 | if (!buf) |
090d1875 | 756 | return 0; |
d1310b2e | 757 | read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, |
f188591e | 758 | buf, 0, 0, btree_get_extent, 0); |
5f39d397 | 759 | free_extent_buffer(buf); |
de428b63 | 760 | return ret; |
090d1875 CM |
761 | } |
762 | ||
0999df54 CM |
763 | struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root, |
764 | u64 bytenr, u32 blocksize) | |
765 | { | |
766 | struct inode *btree_inode = root->fs_info->btree_inode; | |
767 | struct extent_buffer *eb; | |
768 | eb = find_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
769 | bytenr, blocksize, GFP_NOFS); | |
770 | return eb; | |
771 | } | |
772 | ||
773 | struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root, | |
774 | u64 bytenr, u32 blocksize) | |
775 | { | |
776 | struct inode *btree_inode = root->fs_info->btree_inode; | |
777 | struct extent_buffer *eb; | |
778 | ||
779 | eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->io_tree, | |
780 | bytenr, blocksize, NULL, GFP_NOFS); | |
781 | return eb; | |
782 | } | |
783 | ||
784 | ||
e02119d5 CM |
785 | int btrfs_write_tree_block(struct extent_buffer *buf) |
786 | { | |
787 | return btrfs_fdatawrite_range(buf->first_page->mapping, buf->start, | |
24ab9cd8 | 788 | buf->start + buf->len - 1, WB_SYNC_ALL); |
e02119d5 CM |
789 | } |
790 | ||
791 | int btrfs_wait_tree_block_writeback(struct extent_buffer *buf) | |
792 | { | |
793 | return btrfs_wait_on_page_writeback_range(buf->first_page->mapping, | |
794 | buf->start, buf->start + buf->len -1); | |
795 | } | |
796 | ||
0999df54 | 797 | struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr, |
ca7a79ad | 798 | u32 blocksize, u64 parent_transid) |
0999df54 CM |
799 | { |
800 | struct extent_buffer *buf = NULL; | |
801 | struct inode *btree_inode = root->fs_info->btree_inode; | |
802 | struct extent_io_tree *io_tree; | |
803 | int ret; | |
804 | ||
805 | io_tree = &BTRFS_I(btree_inode)->io_tree; | |
806 | ||
807 | buf = btrfs_find_create_tree_block(root, bytenr, blocksize); | |
808 | if (!buf) | |
809 | return NULL; | |
0999df54 | 810 | |
ca7a79ad | 811 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
812 | |
813 | if (ret == 0) { | |
814 | buf->flags |= EXTENT_UPTODATE; | |
a1b32a59 CM |
815 | } else { |
816 | WARN_ON(1); | |
ce9adaa5 | 817 | } |
5f39d397 | 818 | return buf; |
ce9adaa5 | 819 | |
eb60ceac CM |
820 | } |
821 | ||
e089f05c | 822 | int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root, |
5f39d397 | 823 | struct extent_buffer *buf) |
ed2ff2cb | 824 | { |
5f39d397 | 825 | struct inode *btree_inode = root->fs_info->btree_inode; |
55c69072 | 826 | if (btrfs_header_generation(buf) == |
925baedd CM |
827 | root->fs_info->running_transaction->transid) { |
828 | WARN_ON(!btrfs_tree_locked(buf)); | |
d1310b2e | 829 | clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, |
55c69072 | 830 | buf); |
925baedd | 831 | } |
5f39d397 CM |
832 | return 0; |
833 | } | |
834 | ||
db94535d | 835 | static int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize, |
87ee04eb | 836 | u32 stripesize, struct btrfs_root *root, |
9f5fae2f | 837 | struct btrfs_fs_info *fs_info, |
e20d96d6 | 838 | u64 objectid) |
d97e63b6 | 839 | { |
cfaa7295 | 840 | root->node = NULL; |
a28ec197 | 841 | root->commit_root = NULL; |
31153d81 | 842 | root->ref_tree = NULL; |
db94535d CM |
843 | root->sectorsize = sectorsize; |
844 | root->nodesize = nodesize; | |
845 | root->leafsize = leafsize; | |
87ee04eb | 846 | root->stripesize = stripesize; |
123abc88 | 847 | root->ref_cows = 0; |
0b86a832 CM |
848 | root->track_dirty = 0; |
849 | ||
9f5fae2f | 850 | root->fs_info = fs_info; |
0f7d52f4 CM |
851 | root->objectid = objectid; |
852 | root->last_trans = 0; | |
1b05da2e CM |
853 | root->highest_inode = 0; |
854 | root->last_inode_alloc = 0; | |
58176a96 | 855 | root->name = NULL; |
4313b399 | 856 | root->in_sysfs = 0; |
0b86a832 CM |
857 | |
858 | INIT_LIST_HEAD(&root->dirty_list); | |
7b128766 | 859 | INIT_LIST_HEAD(&root->orphan_list); |
bcc63abb | 860 | INIT_LIST_HEAD(&root->dead_list); |
925baedd | 861 | spin_lock_init(&root->node_lock); |
bcc63abb | 862 | spin_lock_init(&root->list_lock); |
a2135011 | 863 | mutex_init(&root->objectid_mutex); |
e02119d5 | 864 | mutex_init(&root->log_mutex); |
d0c803c4 CM |
865 | extent_io_tree_init(&root->dirty_log_pages, |
866 | fs_info->btree_inode->i_mapping, GFP_NOFS); | |
017e5369 CM |
867 | |
868 | btrfs_leaf_ref_tree_init(&root->ref_tree_struct); | |
869 | root->ref_tree = &root->ref_tree_struct; | |
870 | ||
3768f368 CM |
871 | memset(&root->root_key, 0, sizeof(root->root_key)); |
872 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
6702ed49 | 873 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); |
58176a96 | 874 | memset(&root->root_kobj, 0, sizeof(root->root_kobj)); |
3f157a2f | 875 | root->defrag_trans_start = fs_info->generation; |
58176a96 | 876 | init_completion(&root->kobj_unregister); |
6702ed49 CM |
877 | root->defrag_running = 0; |
878 | root->defrag_level = 0; | |
4d775673 | 879 | root->root_key.objectid = objectid; |
3768f368 CM |
880 | return 0; |
881 | } | |
882 | ||
db94535d | 883 | static int find_and_setup_root(struct btrfs_root *tree_root, |
9f5fae2f CM |
884 | struct btrfs_fs_info *fs_info, |
885 | u64 objectid, | |
e20d96d6 | 886 | struct btrfs_root *root) |
3768f368 CM |
887 | { |
888 | int ret; | |
db94535d | 889 | u32 blocksize; |
84234f3a | 890 | u64 generation; |
3768f368 | 891 | |
db94535d | 892 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
893 | tree_root->sectorsize, tree_root->stripesize, |
894 | root, fs_info, objectid); | |
3768f368 CM |
895 | ret = btrfs_find_last_root(tree_root, objectid, |
896 | &root->root_item, &root->root_key); | |
897 | BUG_ON(ret); | |
898 | ||
84234f3a | 899 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
900 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
901 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 902 | blocksize, generation); |
3768f368 | 903 | BUG_ON(!root->node); |
d97e63b6 CM |
904 | return 0; |
905 | } | |
906 | ||
e02119d5 CM |
907 | int btrfs_free_log_root_tree(struct btrfs_trans_handle *trans, |
908 | struct btrfs_fs_info *fs_info) | |
909 | { | |
910 | struct extent_buffer *eb; | |
d0c803c4 CM |
911 | struct btrfs_root *log_root_tree = fs_info->log_root_tree; |
912 | u64 start = 0; | |
913 | u64 end = 0; | |
e02119d5 CM |
914 | int ret; |
915 | ||
d0c803c4 | 916 | if (!log_root_tree) |
e02119d5 CM |
917 | return 0; |
918 | ||
d0c803c4 CM |
919 | while(1) { |
920 | ret = find_first_extent_bit(&log_root_tree->dirty_log_pages, | |
921 | 0, &start, &end, EXTENT_DIRTY); | |
922 | if (ret) | |
923 | break; | |
924 | ||
925 | clear_extent_dirty(&log_root_tree->dirty_log_pages, | |
926 | start, end, GFP_NOFS); | |
927 | } | |
e02119d5 CM |
928 | eb = fs_info->log_root_tree->node; |
929 | ||
930 | WARN_ON(btrfs_header_level(eb) != 0); | |
931 | WARN_ON(btrfs_header_nritems(eb) != 0); | |
932 | ||
d00aff00 CM |
933 | ret = btrfs_free_reserved_extent(fs_info->tree_root, |
934 | eb->start, eb->len); | |
e02119d5 CM |
935 | BUG_ON(ret); |
936 | ||
937 | free_extent_buffer(eb); | |
938 | kfree(fs_info->log_root_tree); | |
939 | fs_info->log_root_tree = NULL; | |
940 | return 0; | |
941 | } | |
942 | ||
943 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
944 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
945 | { |
946 | struct btrfs_root *root; | |
947 | struct btrfs_root *tree_root = fs_info->tree_root; | |
e02119d5 CM |
948 | |
949 | root = kzalloc(sizeof(*root), GFP_NOFS); | |
950 | if (!root) | |
951 | return -ENOMEM; | |
952 | ||
953 | __setup_root(tree_root->nodesize, tree_root->leafsize, | |
954 | tree_root->sectorsize, tree_root->stripesize, | |
955 | root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
956 | ||
957 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; | |
958 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
959 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
960 | root->ref_cows = 0; | |
961 | ||
962 | root->node = btrfs_alloc_free_block(trans, root, root->leafsize, | |
31840ae1 ZY |
963 | 0, BTRFS_TREE_LOG_OBJECTID, |
964 | trans->transid, 0, 0, 0); | |
e02119d5 CM |
965 | |
966 | btrfs_set_header_nritems(root->node, 0); | |
967 | btrfs_set_header_level(root->node, 0); | |
968 | btrfs_set_header_bytenr(root->node, root->node->start); | |
969 | btrfs_set_header_generation(root->node, trans->transid); | |
970 | btrfs_set_header_owner(root->node, BTRFS_TREE_LOG_OBJECTID); | |
971 | ||
972 | write_extent_buffer(root->node, root->fs_info->fsid, | |
973 | (unsigned long)btrfs_header_fsid(root->node), | |
974 | BTRFS_FSID_SIZE); | |
975 | btrfs_mark_buffer_dirty(root->node); | |
976 | btrfs_tree_unlock(root->node); | |
977 | fs_info->log_root_tree = root; | |
978 | return 0; | |
979 | } | |
980 | ||
981 | struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_root *tree_root, | |
982 | struct btrfs_key *location) | |
983 | { | |
984 | struct btrfs_root *root; | |
985 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
0f7d52f4 | 986 | struct btrfs_path *path; |
5f39d397 | 987 | struct extent_buffer *l; |
1b05da2e | 988 | u64 highest_inode; |
84234f3a | 989 | u64 generation; |
db94535d | 990 | u32 blocksize; |
0f7d52f4 CM |
991 | int ret = 0; |
992 | ||
5eda7b5e | 993 | root = kzalloc(sizeof(*root), GFP_NOFS); |
0cf6c620 | 994 | if (!root) |
0f7d52f4 | 995 | return ERR_PTR(-ENOMEM); |
0f7d52f4 | 996 | if (location->offset == (u64)-1) { |
db94535d | 997 | ret = find_and_setup_root(tree_root, fs_info, |
0f7d52f4 CM |
998 | location->objectid, root); |
999 | if (ret) { | |
0f7d52f4 CM |
1000 | kfree(root); |
1001 | return ERR_PTR(ret); | |
1002 | } | |
1003 | goto insert; | |
1004 | } | |
1005 | ||
db94535d | 1006 | __setup_root(tree_root->nodesize, tree_root->leafsize, |
87ee04eb CM |
1007 | tree_root->sectorsize, tree_root->stripesize, |
1008 | root, fs_info, location->objectid); | |
0f7d52f4 CM |
1009 | |
1010 | path = btrfs_alloc_path(); | |
1011 | BUG_ON(!path); | |
1012 | ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0); | |
1013 | if (ret != 0) { | |
0f7d52f4 CM |
1014 | if (ret > 0) |
1015 | ret = -ENOENT; | |
1016 | goto out; | |
1017 | } | |
5f39d397 CM |
1018 | l = path->nodes[0]; |
1019 | read_extent_buffer(l, &root->root_item, | |
1020 | btrfs_item_ptr_offset(l, path->slots[0]), | |
0f7d52f4 | 1021 | sizeof(root->root_item)); |
44b36eb2 | 1022 | memcpy(&root->root_key, location, sizeof(*location)); |
0f7d52f4 CM |
1023 | ret = 0; |
1024 | out: | |
1025 | btrfs_release_path(root, path); | |
1026 | btrfs_free_path(path); | |
1027 | if (ret) { | |
1028 | kfree(root); | |
1029 | return ERR_PTR(ret); | |
1030 | } | |
84234f3a | 1031 | generation = btrfs_root_generation(&root->root_item); |
db94535d CM |
1032 | blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item)); |
1033 | root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item), | |
84234f3a | 1034 | blocksize, generation); |
0f7d52f4 CM |
1035 | BUG_ON(!root->node); |
1036 | insert: | |
e02119d5 CM |
1037 | if (location->objectid != BTRFS_TREE_LOG_OBJECTID) { |
1038 | root->ref_cows = 1; | |
1039 | ret = btrfs_find_highest_inode(root, &highest_inode); | |
1040 | if (ret == 0) { | |
1041 | root->highest_inode = highest_inode; | |
1042 | root->last_inode_alloc = highest_inode; | |
1043 | } | |
5eda7b5e CM |
1044 | } |
1045 | return root; | |
1046 | } | |
1047 | ||
dc17ff8f CM |
1048 | struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1049 | u64 root_objectid) | |
1050 | { | |
1051 | struct btrfs_root *root; | |
1052 | ||
1053 | if (root_objectid == BTRFS_ROOT_TREE_OBJECTID) | |
1054 | return fs_info->tree_root; | |
1055 | if (root_objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1056 | return fs_info->extent_root; | |
1057 | ||
1058 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1059 | (unsigned long)root_objectid); | |
1060 | return root; | |
1061 | } | |
1062 | ||
edbd8d4e CM |
1063 | struct btrfs_root *btrfs_read_fs_root_no_name(struct btrfs_fs_info *fs_info, |
1064 | struct btrfs_key *location) | |
5eda7b5e CM |
1065 | { |
1066 | struct btrfs_root *root; | |
1067 | int ret; | |
1068 | ||
edbd8d4e CM |
1069 | if (location->objectid == BTRFS_ROOT_TREE_OBJECTID) |
1070 | return fs_info->tree_root; | |
1071 | if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
1072 | return fs_info->extent_root; | |
8f18cf13 CM |
1073 | if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1074 | return fs_info->chunk_root; | |
1075 | if (location->objectid == BTRFS_DEV_TREE_OBJECTID) | |
1076 | return fs_info->dev_root; | |
edbd8d4e | 1077 | |
5eda7b5e CM |
1078 | root = radix_tree_lookup(&fs_info->fs_roots_radix, |
1079 | (unsigned long)location->objectid); | |
1080 | if (root) | |
1081 | return root; | |
1082 | ||
e02119d5 | 1083 | root = btrfs_read_fs_root_no_radix(fs_info->tree_root, location); |
5eda7b5e CM |
1084 | if (IS_ERR(root)) |
1085 | return root; | |
2619ba1f CM |
1086 | ret = radix_tree_insert(&fs_info->fs_roots_radix, |
1087 | (unsigned long)root->root_key.objectid, | |
0f7d52f4 CM |
1088 | root); |
1089 | if (ret) { | |
5f39d397 | 1090 | free_extent_buffer(root->node); |
0f7d52f4 CM |
1091 | kfree(root); |
1092 | return ERR_PTR(ret); | |
1093 | } | |
c146afad YZ |
1094 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { |
1095 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1096 | root->root_key.objectid, root); | |
1097 | BUG_ON(ret); | |
1098 | btrfs_orphan_cleanup(root); | |
1099 | } | |
edbd8d4e CM |
1100 | return root; |
1101 | } | |
1102 | ||
1103 | struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info, | |
1104 | struct btrfs_key *location, | |
1105 | const char *name, int namelen) | |
1106 | { | |
1107 | struct btrfs_root *root; | |
1108 | int ret; | |
1109 | ||
1110 | root = btrfs_read_fs_root_no_name(fs_info, location); | |
1111 | if (!root) | |
1112 | return NULL; | |
58176a96 | 1113 | |
4313b399 CM |
1114 | if (root->in_sysfs) |
1115 | return root; | |
1116 | ||
58176a96 JB |
1117 | ret = btrfs_set_root_name(root, name, namelen); |
1118 | if (ret) { | |
5f39d397 | 1119 | free_extent_buffer(root->node); |
58176a96 JB |
1120 | kfree(root); |
1121 | return ERR_PTR(ret); | |
1122 | } | |
1123 | ||
1124 | ret = btrfs_sysfs_add_root(root); | |
1125 | if (ret) { | |
5f39d397 | 1126 | free_extent_buffer(root->node); |
58176a96 JB |
1127 | kfree(root->name); |
1128 | kfree(root); | |
1129 | return ERR_PTR(ret); | |
1130 | } | |
4313b399 | 1131 | root->in_sysfs = 1; |
0f7d52f4 CM |
1132 | return root; |
1133 | } | |
19c00ddc CM |
1134 | #if 0 |
1135 | static int add_hasher(struct btrfs_fs_info *info, char *type) { | |
1136 | struct btrfs_hasher *hasher; | |
1137 | ||
1138 | hasher = kmalloc(sizeof(*hasher), GFP_NOFS); | |
1139 | if (!hasher) | |
1140 | return -ENOMEM; | |
1141 | hasher->hash_tfm = crypto_alloc_hash(type, 0, CRYPTO_ALG_ASYNC); | |
1142 | if (!hasher->hash_tfm) { | |
1143 | kfree(hasher); | |
1144 | return -EINVAL; | |
1145 | } | |
1146 | spin_lock(&info->hash_lock); | |
1147 | list_add(&hasher->list, &info->hashers); | |
1148 | spin_unlock(&info->hash_lock); | |
1149 | return 0; | |
1150 | } | |
1151 | #endif | |
04160088 CM |
1152 | |
1153 | static int btrfs_congested_fn(void *congested_data, int bdi_bits) | |
1154 | { | |
1155 | struct btrfs_fs_info *info = (struct btrfs_fs_info *)congested_data; | |
1156 | int ret = 0; | |
1157 | struct list_head *cur; | |
1158 | struct btrfs_device *device; | |
1159 | struct backing_dev_info *bdi; | |
6f3577bd | 1160 | #if 0 |
cb03c743 | 1161 | if ((bdi_bits & (1 << BDI_write_congested)) && |
777e6bd7 | 1162 | btrfs_congested_async(info, 0)) |
cb03c743 | 1163 | return 1; |
6f3577bd | 1164 | #endif |
04160088 CM |
1165 | list_for_each(cur, &info->fs_devices->devices) { |
1166 | device = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1167 | if (!device->bdev) |
1168 | continue; | |
04160088 CM |
1169 | bdi = blk_get_backing_dev_info(device->bdev); |
1170 | if (bdi && bdi_congested(bdi, bdi_bits)) { | |
1171 | ret = 1; | |
1172 | break; | |
1173 | } | |
1174 | } | |
1175 | return ret; | |
1176 | } | |
1177 | ||
38b66988 CM |
1178 | /* |
1179 | * this unplugs every device on the box, and it is only used when page | |
1180 | * is null | |
1181 | */ | |
1182 | static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page) | |
1183 | { | |
1184 | struct list_head *cur; | |
1185 | struct btrfs_device *device; | |
1186 | struct btrfs_fs_info *info; | |
1187 | ||
1188 | info = (struct btrfs_fs_info *)bdi->unplug_io_data; | |
1189 | list_for_each(cur, &info->fs_devices->devices) { | |
1190 | device = list_entry(cur, struct btrfs_device, dev_list); | |
1191 | bdi = blk_get_backing_dev_info(device->bdev); | |
1192 | if (bdi->unplug_io_fn) { | |
1193 | bdi->unplug_io_fn(bdi, page); | |
1194 | } | |
1195 | } | |
1196 | } | |
1197 | ||
04160088 CM |
1198 | void btrfs_unplug_io_fn(struct backing_dev_info *bdi, struct page *page) |
1199 | { | |
38b66988 | 1200 | struct inode *inode; |
f2d8d74d CM |
1201 | struct extent_map_tree *em_tree; |
1202 | struct extent_map *em; | |
bcbfce8a | 1203 | struct address_space *mapping; |
38b66988 CM |
1204 | u64 offset; |
1205 | ||
bcbfce8a | 1206 | /* the generic O_DIRECT read code does this */ |
38b66988 CM |
1207 | if (!page) { |
1208 | __unplug_io_fn(bdi, page); | |
1209 | return; | |
1210 | } | |
1211 | ||
bcbfce8a CM |
1212 | /* |
1213 | * page->mapping may change at any time. Get a consistent copy | |
1214 | * and use that for everything below | |
1215 | */ | |
1216 | smp_mb(); | |
1217 | mapping = page->mapping; | |
1218 | if (!mapping) | |
1219 | return; | |
1220 | ||
1221 | inode = mapping->host; | |
240d5d48 CM |
1222 | |
1223 | /* | |
1224 | * don't do the expensive searching for a small number of | |
1225 | * devices | |
1226 | */ | |
1227 | if (BTRFS_I(inode)->root->fs_info->fs_devices->open_devices <= 2) { | |
1228 | __unplug_io_fn(bdi, page); | |
1229 | return; | |
1230 | } | |
1231 | ||
38b66988 | 1232 | offset = page_offset(page); |
04160088 | 1233 | |
f2d8d74d CM |
1234 | em_tree = &BTRFS_I(inode)->extent_tree; |
1235 | spin_lock(&em_tree->lock); | |
1236 | em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE); | |
1237 | spin_unlock(&em_tree->lock); | |
89642229 CM |
1238 | if (!em) { |
1239 | __unplug_io_fn(bdi, page); | |
f2d8d74d | 1240 | return; |
89642229 | 1241 | } |
f2d8d74d | 1242 | |
89642229 CM |
1243 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) { |
1244 | free_extent_map(em); | |
1245 | __unplug_io_fn(bdi, page); | |
1246 | return; | |
1247 | } | |
f2d8d74d CM |
1248 | offset = offset - em->start; |
1249 | btrfs_unplug_page(&BTRFS_I(inode)->root->fs_info->mapping_tree, | |
1250 | em->block_start + offset, page); | |
1251 | free_extent_map(em); | |
04160088 CM |
1252 | } |
1253 | ||
1254 | static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi) | |
1255 | { | |
1256 | bdi_init(bdi); | |
4575c9cc | 1257 | bdi->ra_pages = default_backing_dev_info.ra_pages; |
04160088 CM |
1258 | bdi->state = 0; |
1259 | bdi->capabilities = default_backing_dev_info.capabilities; | |
1260 | bdi->unplug_io_fn = btrfs_unplug_io_fn; | |
1261 | bdi->unplug_io_data = info; | |
1262 | bdi->congested_fn = btrfs_congested_fn; | |
1263 | bdi->congested_data = info; | |
1264 | return 0; | |
1265 | } | |
1266 | ||
ce9adaa5 CM |
1267 | static int bio_ready_for_csum(struct bio *bio) |
1268 | { | |
1269 | u64 length = 0; | |
1270 | u64 buf_len = 0; | |
1271 | u64 start = 0; | |
1272 | struct page *page; | |
1273 | struct extent_io_tree *io_tree = NULL; | |
1274 | struct btrfs_fs_info *info = NULL; | |
1275 | struct bio_vec *bvec; | |
1276 | int i; | |
1277 | int ret; | |
1278 | ||
1279 | bio_for_each_segment(bvec, bio, i) { | |
1280 | page = bvec->bv_page; | |
1281 | if (page->private == EXTENT_PAGE_PRIVATE) { | |
1282 | length += bvec->bv_len; | |
1283 | continue; | |
1284 | } | |
1285 | if (!page->private) { | |
1286 | length += bvec->bv_len; | |
1287 | continue; | |
1288 | } | |
1289 | length = bvec->bv_len; | |
1290 | buf_len = page->private >> 2; | |
1291 | start = page_offset(page) + bvec->bv_offset; | |
1292 | io_tree = &BTRFS_I(page->mapping->host)->io_tree; | |
1293 | info = BTRFS_I(page->mapping->host)->root->fs_info; | |
1294 | } | |
1295 | /* are we fully contained in this bio? */ | |
1296 | if (buf_len <= length) | |
1297 | return 1; | |
1298 | ||
1299 | ret = extent_range_uptodate(io_tree, start + length, | |
1300 | start + buf_len - 1); | |
1301 | if (ret == 1) | |
1302 | return ret; | |
1303 | return ret; | |
1304 | } | |
1305 | ||
8b712842 CM |
1306 | /* |
1307 | * called by the kthread helper functions to finally call the bio end_io | |
1308 | * functions. This is where read checksum verification actually happens | |
1309 | */ | |
1310 | static void end_workqueue_fn(struct btrfs_work *work) | |
ce9adaa5 | 1311 | { |
ce9adaa5 | 1312 | struct bio *bio; |
8b712842 CM |
1313 | struct end_io_wq *end_io_wq; |
1314 | struct btrfs_fs_info *fs_info; | |
ce9adaa5 | 1315 | int error; |
ce9adaa5 | 1316 | |
8b712842 CM |
1317 | end_io_wq = container_of(work, struct end_io_wq, work); |
1318 | bio = end_io_wq->bio; | |
1319 | fs_info = end_io_wq->info; | |
ce9adaa5 | 1320 | |
8b712842 CM |
1321 | /* metadata bios are special because the whole tree block must |
1322 | * be checksummed at once. This makes sure the entire block is in | |
1323 | * ram and up to date before trying to verify things. For | |
1324 | * blocksize <= pagesize, it is basically a noop | |
1325 | */ | |
1326 | if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { | |
1327 | btrfs_queue_worker(&fs_info->endio_workers, | |
1328 | &end_io_wq->work); | |
1329 | return; | |
1330 | } | |
1331 | error = end_io_wq->error; | |
1332 | bio->bi_private = end_io_wq->private; | |
1333 | bio->bi_end_io = end_io_wq->end_io; | |
1334 | kfree(end_io_wq); | |
8b712842 | 1335 | bio_endio(bio, error); |
44b8bd7e CM |
1336 | } |
1337 | ||
a74a4b97 CM |
1338 | static int cleaner_kthread(void *arg) |
1339 | { | |
1340 | struct btrfs_root *root = arg; | |
1341 | ||
1342 | do { | |
1343 | smp_mb(); | |
1344 | if (root->fs_info->closing) | |
1345 | break; | |
1346 | ||
1347 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1348 | mutex_lock(&root->fs_info->cleaner_mutex); | |
a74a4b97 | 1349 | btrfs_clean_old_snapshots(root); |
a74a4b97 CM |
1350 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1351 | ||
1352 | if (freezing(current)) { | |
1353 | refrigerator(); | |
1354 | } else { | |
1355 | smp_mb(); | |
1356 | if (root->fs_info->closing) | |
1357 | break; | |
1358 | set_current_state(TASK_INTERRUPTIBLE); | |
1359 | schedule(); | |
1360 | __set_current_state(TASK_RUNNING); | |
1361 | } | |
1362 | } while (!kthread_should_stop()); | |
1363 | return 0; | |
1364 | } | |
1365 | ||
1366 | static int transaction_kthread(void *arg) | |
1367 | { | |
1368 | struct btrfs_root *root = arg; | |
1369 | struct btrfs_trans_handle *trans; | |
1370 | struct btrfs_transaction *cur; | |
1371 | unsigned long now; | |
1372 | unsigned long delay; | |
1373 | int ret; | |
1374 | ||
1375 | do { | |
1376 | smp_mb(); | |
1377 | if (root->fs_info->closing) | |
1378 | break; | |
1379 | ||
1380 | delay = HZ * 30; | |
1381 | vfs_check_frozen(root->fs_info->sb, SB_FREEZE_WRITE); | |
1382 | mutex_lock(&root->fs_info->transaction_kthread_mutex); | |
1383 | ||
ab78c84d CM |
1384 | if (root->fs_info->total_ref_cache_size > 20 * 1024 * 1024) { |
1385 | printk("btrfs: total reference cache size %Lu\n", | |
1386 | root->fs_info->total_ref_cache_size); | |
1387 | } | |
31153d81 | 1388 | |
a74a4b97 CM |
1389 | mutex_lock(&root->fs_info->trans_mutex); |
1390 | cur = root->fs_info->running_transaction; | |
1391 | if (!cur) { | |
1392 | mutex_unlock(&root->fs_info->trans_mutex); | |
1393 | goto sleep; | |
1394 | } | |
31153d81 | 1395 | |
a74a4b97 CM |
1396 | now = get_seconds(); |
1397 | if (now < cur->start_time || now - cur->start_time < 30) { | |
1398 | mutex_unlock(&root->fs_info->trans_mutex); | |
1399 | delay = HZ * 5; | |
1400 | goto sleep; | |
1401 | } | |
1402 | mutex_unlock(&root->fs_info->trans_mutex); | |
a74a4b97 CM |
1403 | trans = btrfs_start_transaction(root, 1); |
1404 | ret = btrfs_commit_transaction(trans, root); | |
1405 | sleep: | |
1406 | wake_up_process(root->fs_info->cleaner_kthread); | |
1407 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
1408 | ||
1409 | if (freezing(current)) { | |
1410 | refrigerator(); | |
1411 | } else { | |
1412 | if (root->fs_info->closing) | |
1413 | break; | |
1414 | set_current_state(TASK_INTERRUPTIBLE); | |
1415 | schedule_timeout(delay); | |
1416 | __set_current_state(TASK_RUNNING); | |
1417 | } | |
1418 | } while (!kthread_should_stop()); | |
1419 | return 0; | |
1420 | } | |
1421 | ||
8a4b83cc | 1422 | struct btrfs_root *open_ctree(struct super_block *sb, |
dfe25020 CM |
1423 | struct btrfs_fs_devices *fs_devices, |
1424 | char *options) | |
2e635a27 | 1425 | { |
db94535d CM |
1426 | u32 sectorsize; |
1427 | u32 nodesize; | |
1428 | u32 leafsize; | |
1429 | u32 blocksize; | |
87ee04eb | 1430 | u32 stripesize; |
84234f3a | 1431 | u64 generation; |
3de4586c | 1432 | struct btrfs_key location; |
a061fc8d | 1433 | struct buffer_head *bh; |
e02119d5 | 1434 | struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1435 | GFP_NOFS); |
e02119d5 | 1436 | struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), |
e20d96d6 | 1437 | GFP_NOFS); |
8790d502 | 1438 | struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), |
e20d96d6 | 1439 | GFP_NOFS); |
e02119d5 | 1440 | struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1441 | GFP_NOFS); |
e02119d5 | 1442 | struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), |
0b86a832 | 1443 | GFP_NOFS); |
e02119d5 CM |
1444 | struct btrfs_root *log_tree_root; |
1445 | ||
eb60ceac | 1446 | int ret; |
e58ca020 | 1447 | int err = -EINVAL; |
4543df7e | 1448 | |
2c90e5d6 | 1449 | struct btrfs_super_block *disk_super; |
8790d502 | 1450 | |
0463bb4e JM |
1451 | if (!extent_root || !tree_root || !fs_info || |
1452 | !chunk_root || !dev_root) { | |
39279cc3 CM |
1453 | err = -ENOMEM; |
1454 | goto fail; | |
1455 | } | |
0f7d52f4 | 1456 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS); |
8fd17795 | 1457 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 1458 | INIT_LIST_HEAD(&fs_info->dead_roots); |
19c00ddc | 1459 | INIT_LIST_HEAD(&fs_info->hashers); |
ea8c2819 | 1460 | INIT_LIST_HEAD(&fs_info->delalloc_inodes); |
19c00ddc | 1461 | spin_lock_init(&fs_info->hash_lock); |
1832a6d5 | 1462 | spin_lock_init(&fs_info->delalloc_lock); |
cee36a03 | 1463 | spin_lock_init(&fs_info->new_trans_lock); |
31153d81 | 1464 | spin_lock_init(&fs_info->ref_cache_lock); |
19c00ddc | 1465 | |
58176a96 | 1466 | init_completion(&fs_info->kobj_unregister); |
9f5fae2f CM |
1467 | fs_info->tree_root = tree_root; |
1468 | fs_info->extent_root = extent_root; | |
0b86a832 CM |
1469 | fs_info->chunk_root = chunk_root; |
1470 | fs_info->dev_root = dev_root; | |
8a4b83cc | 1471 | fs_info->fs_devices = fs_devices; |
0b86a832 | 1472 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 1473 | INIT_LIST_HEAD(&fs_info->space_info); |
0b86a832 | 1474 | btrfs_mapping_init(&fs_info->mapping_tree); |
cb03c743 | 1475 | atomic_set(&fs_info->nr_async_submits, 0); |
771ed689 | 1476 | atomic_set(&fs_info->async_delalloc_pages, 0); |
8c8bee1d | 1477 | atomic_set(&fs_info->async_submit_draining, 0); |
0986fe9e | 1478 | atomic_set(&fs_info->nr_async_bios, 0); |
a2135011 | 1479 | atomic_set(&fs_info->throttles, 0); |
ab78c84d | 1480 | atomic_set(&fs_info->throttle_gen, 0); |
e20d96d6 | 1481 | fs_info->sb = sb; |
c59f8951 | 1482 | fs_info->max_extent = (u64)-1; |
6f568d35 | 1483 | fs_info->max_inline = 8192 * 1024; |
04160088 | 1484 | setup_bdi(fs_info, &fs_info->bdi); |
d98237b3 CM |
1485 | fs_info->btree_inode = new_inode(sb); |
1486 | fs_info->btree_inode->i_ino = 1; | |
2c90e5d6 | 1487 | fs_info->btree_inode->i_nlink = 1; |
c8b97818 | 1488 | |
4543df7e | 1489 | fs_info->thread_pool_size = min(num_online_cpus() + 2, 8); |
0afbaf8c | 1490 | |
3eaa2885 CM |
1491 | INIT_LIST_HEAD(&fs_info->ordered_extents); |
1492 | spin_lock_init(&fs_info->ordered_extent_lock); | |
1493 | ||
a061fc8d CM |
1494 | sb->s_blocksize = 4096; |
1495 | sb->s_blocksize_bits = blksize_bits(4096); | |
1496 | ||
0afbaf8c CM |
1497 | /* |
1498 | * we set the i_size on the btree inode to the max possible int. | |
1499 | * the real end of the address space is determined by all of | |
1500 | * the devices in the system | |
1501 | */ | |
1502 | fs_info->btree_inode->i_size = OFFSET_MAX; | |
d98237b3 | 1503 | fs_info->btree_inode->i_mapping->a_ops = &btree_aops; |
04160088 CM |
1504 | fs_info->btree_inode->i_mapping->backing_dev_info = &fs_info->bdi; |
1505 | ||
d1310b2e | 1506 | extent_io_tree_init(&BTRFS_I(fs_info->btree_inode)->io_tree, |
5f39d397 CM |
1507 | fs_info->btree_inode->i_mapping, |
1508 | GFP_NOFS); | |
d1310b2e CM |
1509 | extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree, |
1510 | GFP_NOFS); | |
1511 | ||
1512 | BTRFS_I(fs_info->btree_inode)->io_tree.ops = &btree_extent_io_ops; | |
0da5468f | 1513 | |
0f9dd46c JB |
1514 | spin_lock_init(&fs_info->block_group_cache_lock); |
1515 | fs_info->block_group_cache_tree.rb_node = NULL; | |
1516 | ||
d1310b2e | 1517 | extent_io_tree_init(&fs_info->pinned_extents, |
1a5bc167 | 1518 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1519 | extent_io_tree_init(&fs_info->pending_del, |
1a5bc167 | 1520 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
d1310b2e | 1521 | extent_io_tree_init(&fs_info->extent_ins, |
1a5bc167 | 1522 | fs_info->btree_inode->i_mapping, GFP_NOFS); |
e66f709b | 1523 | fs_info->do_barriers = 1; |
e18e4809 | 1524 | |
1a40e23b ZY |
1525 | INIT_LIST_HEAD(&fs_info->dead_reloc_roots); |
1526 | btrfs_leaf_ref_tree_init(&fs_info->reloc_ref_tree); | |
e4657689 ZY |
1527 | btrfs_leaf_ref_tree_init(&fs_info->shared_ref_tree); |
1528 | ||
0f7d52f4 CM |
1529 | BTRFS_I(fs_info->btree_inode)->root = tree_root; |
1530 | memset(&BTRFS_I(fs_info->btree_inode)->location, 0, | |
1531 | sizeof(struct btrfs_key)); | |
22b0ebda | 1532 | insert_inode_hash(fs_info->btree_inode); |
39279cc3 | 1533 | |
79154b1b | 1534 | mutex_init(&fs_info->trans_mutex); |
e02119d5 | 1535 | mutex_init(&fs_info->tree_log_mutex); |
a2135011 | 1536 | mutex_init(&fs_info->drop_mutex); |
25179201 JB |
1537 | mutex_init(&fs_info->extent_ins_mutex); |
1538 | mutex_init(&fs_info->pinned_mutex); | |
925baedd | 1539 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
1540 | mutex_init(&fs_info->transaction_kthread_mutex); |
1541 | mutex_init(&fs_info->cleaner_mutex); | |
7d9eb12c | 1542 | mutex_init(&fs_info->volume_mutex); |
1a40e23b | 1543 | mutex_init(&fs_info->tree_reloc_mutex); |
e6dcd2dc | 1544 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 1545 | init_waitqueue_head(&fs_info->transaction_wait); |
4854ddd0 | 1546 | init_waitqueue_head(&fs_info->async_submit_wait); |
e02119d5 CM |
1547 | init_waitqueue_head(&fs_info->tree_log_wait); |
1548 | atomic_set(&fs_info->tree_log_commit, 0); | |
1549 | atomic_set(&fs_info->tree_log_writers, 0); | |
1550 | fs_info->tree_log_transid = 0; | |
3768f368 | 1551 | |
19c00ddc CM |
1552 | #if 0 |
1553 | ret = add_hasher(fs_info, "crc32c"); | |
1554 | if (ret) { | |
1555 | printk("btrfs: failed hash setup, modprobe cryptomgr?\n"); | |
1556 | err = -ENOMEM; | |
1557 | goto fail_iput; | |
1558 | } | |
1559 | #endif | |
0b86a832 | 1560 | __setup_root(4096, 4096, 4096, 4096, tree_root, |
2c90e5d6 | 1561 | fs_info, BTRFS_ROOT_TREE_OBJECTID); |
7eccb903 | 1562 | |
d98237b3 | 1563 | |
a061fc8d CM |
1564 | bh = __bread(fs_devices->latest_bdev, |
1565 | BTRFS_SUPER_INFO_OFFSET / 4096, 4096); | |
1566 | if (!bh) | |
39279cc3 | 1567 | goto fail_iput; |
39279cc3 | 1568 | |
a061fc8d CM |
1569 | memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy)); |
1570 | brelse(bh); | |
5f39d397 | 1571 | |
a061fc8d | 1572 | memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE); |
0b86a832 | 1573 | |
5f39d397 | 1574 | disk_super = &fs_info->super_copy; |
0f7d52f4 | 1575 | if (!btrfs_super_root(disk_super)) |
39279cc3 | 1576 | goto fail_sb_buffer; |
0f7d52f4 | 1577 | |
2b82032c YZ |
1578 | ret = btrfs_parse_options(tree_root, options); |
1579 | if (ret) { | |
1580 | err = ret; | |
edf24abe | 1581 | goto fail_sb_buffer; |
2b82032c | 1582 | } |
dfe25020 | 1583 | |
4543df7e CM |
1584 | /* |
1585 | * we need to start all the end_io workers up front because the | |
1586 | * queue work function gets called at interrupt time, and so it | |
1587 | * cannot dynamically grow. | |
1588 | */ | |
5443be45 CM |
1589 | btrfs_init_workers(&fs_info->workers, "worker", |
1590 | fs_info->thread_pool_size); | |
c8b97818 | 1591 | |
771ed689 CM |
1592 | btrfs_init_workers(&fs_info->delalloc_workers, "delalloc", |
1593 | fs_info->thread_pool_size); | |
1594 | ||
5443be45 | 1595 | btrfs_init_workers(&fs_info->submit_workers, "submit", |
b720d209 CM |
1596 | min_t(u64, fs_devices->num_devices, |
1597 | fs_info->thread_pool_size)); | |
61b49440 CM |
1598 | |
1599 | /* a higher idle thresh on the submit workers makes it much more | |
1600 | * likely that bios will be send down in a sane order to the | |
1601 | * devices | |
1602 | */ | |
1603 | fs_info->submit_workers.idle_thresh = 64; | |
53863232 | 1604 | |
771ed689 | 1605 | fs_info->workers.idle_thresh = 16; |
4a69a410 | 1606 | fs_info->workers.ordered = 1; |
61b49440 | 1607 | |
771ed689 CM |
1608 | fs_info->delalloc_workers.idle_thresh = 2; |
1609 | fs_info->delalloc_workers.ordered = 1; | |
1610 | ||
5443be45 CM |
1611 | btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); |
1612 | btrfs_init_workers(&fs_info->endio_workers, "endio", | |
1613 | fs_info->thread_pool_size); | |
1614 | btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", | |
e6dcd2dc | 1615 | fs_info->thread_pool_size); |
61b49440 CM |
1616 | |
1617 | /* | |
1618 | * endios are largely parallel and should have a very | |
1619 | * low idle thresh | |
1620 | */ | |
1621 | fs_info->endio_workers.idle_thresh = 4; | |
3a5f1d45 | 1622 | fs_info->endio_write_workers.idle_thresh = 64; |
61b49440 | 1623 | |
4543df7e | 1624 | btrfs_start_workers(&fs_info->workers, 1); |
1cc127b5 | 1625 | btrfs_start_workers(&fs_info->submit_workers, 1); |
771ed689 | 1626 | btrfs_start_workers(&fs_info->delalloc_workers, 1); |
247e743c | 1627 | btrfs_start_workers(&fs_info->fixup_workers, 1); |
4543df7e | 1628 | btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); |
e6dcd2dc CM |
1629 | btrfs_start_workers(&fs_info->endio_write_workers, |
1630 | fs_info->thread_pool_size); | |
4543df7e | 1631 | |
4575c9cc | 1632 | fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super); |
c8b97818 CM |
1633 | fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages, |
1634 | 4 * 1024 * 1024 / PAGE_CACHE_SIZE); | |
4575c9cc | 1635 | |
db94535d CM |
1636 | nodesize = btrfs_super_nodesize(disk_super); |
1637 | leafsize = btrfs_super_leafsize(disk_super); | |
1638 | sectorsize = btrfs_super_sectorsize(disk_super); | |
87ee04eb | 1639 | stripesize = btrfs_super_stripesize(disk_super); |
db94535d CM |
1640 | tree_root->nodesize = nodesize; |
1641 | tree_root->leafsize = leafsize; | |
1642 | tree_root->sectorsize = sectorsize; | |
87ee04eb | 1643 | tree_root->stripesize = stripesize; |
a061fc8d CM |
1644 | |
1645 | sb->s_blocksize = sectorsize; | |
1646 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
db94535d | 1647 | |
39279cc3 CM |
1648 | if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC, |
1649 | sizeof(disk_super->magic))) { | |
1650 | printk("btrfs: valid FS not found on %s\n", sb->s_id); | |
1651 | goto fail_sb_buffer; | |
1652 | } | |
19c00ddc | 1653 | |
925baedd | 1654 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1655 | ret = btrfs_read_sys_array(tree_root); |
925baedd | 1656 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f CM |
1657 | if (ret) { |
1658 | printk("btrfs: failed to read the system array on %s\n", | |
1659 | sb->s_id); | |
1660 | goto fail_sys_array; | |
1661 | } | |
0b86a832 CM |
1662 | |
1663 | blocksize = btrfs_level_size(tree_root, | |
1664 | btrfs_super_chunk_root_level(disk_super)); | |
84234f3a | 1665 | generation = btrfs_super_chunk_root_generation(disk_super); |
0b86a832 CM |
1666 | |
1667 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1668 | chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID); | |
1669 | ||
1670 | chunk_root->node = read_tree_block(chunk_root, | |
1671 | btrfs_super_chunk_root(disk_super), | |
84234f3a | 1672 | blocksize, generation); |
0b86a832 CM |
1673 | BUG_ON(!chunk_root->node); |
1674 | ||
e17cade2 CM |
1675 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
1676 | (unsigned long)btrfs_header_chunk_tree_uuid(chunk_root->node), | |
1677 | BTRFS_UUID_SIZE); | |
1678 | ||
925baedd | 1679 | mutex_lock(&fs_info->chunk_mutex); |
0b86a832 | 1680 | ret = btrfs_read_chunk_tree(chunk_root); |
925baedd | 1681 | mutex_unlock(&fs_info->chunk_mutex); |
2b82032c YZ |
1682 | if (ret) { |
1683 | printk("btrfs: failed to read chunk tree on %s\n", sb->s_id); | |
1684 | goto fail_chunk_root; | |
1685 | } | |
0b86a832 | 1686 | |
dfe25020 CM |
1687 | btrfs_close_extra_devices(fs_devices); |
1688 | ||
db94535d CM |
1689 | blocksize = btrfs_level_size(tree_root, |
1690 | btrfs_super_root_level(disk_super)); | |
84234f3a | 1691 | generation = btrfs_super_generation(disk_super); |
0b86a832 | 1692 | |
e20d96d6 | 1693 | tree_root->node = read_tree_block(tree_root, |
db94535d | 1694 | btrfs_super_root(disk_super), |
84234f3a | 1695 | blocksize, generation); |
39279cc3 | 1696 | if (!tree_root->node) |
2b82032c | 1697 | goto fail_chunk_root; |
3768f368 | 1698 | |
db94535d CM |
1699 | |
1700 | ret = find_and_setup_root(tree_root, fs_info, | |
e20d96d6 | 1701 | BTRFS_EXTENT_TREE_OBJECTID, extent_root); |
0b86a832 | 1702 | if (ret) |
39279cc3 | 1703 | goto fail_tree_root; |
0b86a832 CM |
1704 | extent_root->track_dirty = 1; |
1705 | ||
1706 | ret = find_and_setup_root(tree_root, fs_info, | |
1707 | BTRFS_DEV_TREE_OBJECTID, dev_root); | |
1708 | dev_root->track_dirty = 1; | |
1709 | ||
1710 | if (ret) | |
1711 | goto fail_extent_root; | |
3768f368 | 1712 | |
9078a3e1 CM |
1713 | btrfs_read_block_groups(extent_root); |
1714 | ||
c146afad YZ |
1715 | fs_info->generation = generation + 1; |
1716 | fs_info->last_trans_committed = generation; | |
d18a2c44 CM |
1717 | fs_info->data_alloc_profile = (u64)-1; |
1718 | fs_info->metadata_alloc_profile = (u64)-1; | |
1719 | fs_info->system_alloc_profile = fs_info->metadata_alloc_profile; | |
a74a4b97 CM |
1720 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, |
1721 | "btrfs-cleaner"); | |
1722 | if (!fs_info->cleaner_kthread) | |
1723 | goto fail_extent_root; | |
1724 | ||
1725 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
1726 | tree_root, | |
1727 | "btrfs-transaction"); | |
1728 | if (!fs_info->transaction_kthread) | |
3f157a2f | 1729 | goto fail_cleaner; |
a74a4b97 | 1730 | |
c146afad | 1731 | if (sb->s_flags & MS_RDONLY) |
3de4586c | 1732 | goto read_fs_root; |
c146afad | 1733 | |
e02119d5 CM |
1734 | if (btrfs_super_log_root(disk_super) != 0) { |
1735 | u32 blocksize; | |
1736 | u64 bytenr = btrfs_super_log_root(disk_super); | |
1737 | ||
1738 | blocksize = | |
1739 | btrfs_level_size(tree_root, | |
1740 | btrfs_super_log_root_level(disk_super)); | |
d18a2c44 | 1741 | |
e02119d5 CM |
1742 | log_tree_root = kzalloc(sizeof(struct btrfs_root), |
1743 | GFP_NOFS); | |
1744 | ||
1745 | __setup_root(nodesize, leafsize, sectorsize, stripesize, | |
1746 | log_tree_root, fs_info, BTRFS_TREE_LOG_OBJECTID); | |
1747 | ||
1748 | log_tree_root->node = read_tree_block(tree_root, bytenr, | |
84234f3a YZ |
1749 | blocksize, |
1750 | generation + 1); | |
e02119d5 CM |
1751 | ret = btrfs_recover_log_trees(log_tree_root); |
1752 | BUG_ON(ret); | |
1753 | } | |
1a40e23b ZY |
1754 | |
1755 | ret = btrfs_cleanup_reloc_trees(tree_root); | |
1756 | BUG_ON(ret); | |
1757 | ||
3de4586c CM |
1758 | location.objectid = BTRFS_FS_TREE_OBJECTID; |
1759 | location.type = BTRFS_ROOT_ITEM_KEY; | |
1760 | location.offset = (u64)-1; | |
1761 | ||
1762 | read_fs_root: | |
1763 | fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); | |
1764 | if (!fs_info->fs_root) | |
1765 | goto fail_cleaner; | |
0f7d52f4 | 1766 | return tree_root; |
39279cc3 | 1767 | |
3f157a2f | 1768 | fail_cleaner: |
a74a4b97 | 1769 | kthread_stop(fs_info->cleaner_kthread); |
0b86a832 CM |
1770 | fail_extent_root: |
1771 | free_extent_buffer(extent_root->node); | |
39279cc3 | 1772 | fail_tree_root: |
5f39d397 | 1773 | free_extent_buffer(tree_root->node); |
2b82032c YZ |
1774 | fail_chunk_root: |
1775 | free_extent_buffer(chunk_root->node); | |
84eed90f | 1776 | fail_sys_array: |
39279cc3 | 1777 | fail_sb_buffer: |
247e743c | 1778 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 1779 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
1780 | btrfs_stop_workers(&fs_info->workers); |
1781 | btrfs_stop_workers(&fs_info->endio_workers); | |
e6dcd2dc | 1782 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 1783 | btrfs_stop_workers(&fs_info->submit_workers); |
4543df7e CM |
1784 | fail_iput: |
1785 | iput(fs_info->btree_inode); | |
39279cc3 | 1786 | fail: |
dfe25020 | 1787 | btrfs_close_devices(fs_info->fs_devices); |
84eed90f CM |
1788 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
1789 | ||
39279cc3 CM |
1790 | kfree(extent_root); |
1791 | kfree(tree_root); | |
2d2ae547 | 1792 | bdi_destroy(&fs_info->bdi); |
39279cc3 | 1793 | kfree(fs_info); |
83afeac4 JM |
1794 | kfree(chunk_root); |
1795 | kfree(dev_root); | |
39279cc3 | 1796 | return ERR_PTR(err); |
eb60ceac CM |
1797 | } |
1798 | ||
f2984462 CM |
1799 | static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate) |
1800 | { | |
1801 | char b[BDEVNAME_SIZE]; | |
1802 | ||
1803 | if (uptodate) { | |
1804 | set_buffer_uptodate(bh); | |
1805 | } else { | |
1806 | if (!buffer_eopnotsupp(bh) && printk_ratelimit()) { | |
1807 | printk(KERN_WARNING "lost page write due to " | |
1808 | "I/O error on %s\n", | |
1809 | bdevname(bh->b_bdev, b)); | |
1810 | } | |
1259ab75 CM |
1811 | /* note, we dont' set_buffer_write_io_error because we have |
1812 | * our own ways of dealing with the IO errors | |
1813 | */ | |
f2984462 CM |
1814 | clear_buffer_uptodate(bh); |
1815 | } | |
1816 | unlock_buffer(bh); | |
1817 | put_bh(bh); | |
1818 | } | |
1819 | ||
1820 | int write_all_supers(struct btrfs_root *root) | |
1821 | { | |
1822 | struct list_head *cur; | |
1823 | struct list_head *head = &root->fs_info->fs_devices->devices; | |
1824 | struct btrfs_device *dev; | |
a061fc8d | 1825 | struct btrfs_super_block *sb; |
f2984462 CM |
1826 | struct btrfs_dev_item *dev_item; |
1827 | struct buffer_head *bh; | |
1828 | int ret; | |
1829 | int do_barriers; | |
a236aed1 CM |
1830 | int max_errors; |
1831 | int total_errors = 0; | |
a061fc8d CM |
1832 | u32 crc; |
1833 | u64 flags; | |
f2984462 | 1834 | |
a236aed1 | 1835 | max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1; |
f2984462 CM |
1836 | do_barriers = !btrfs_test_opt(root, NOBARRIER); |
1837 | ||
a061fc8d CM |
1838 | sb = &root->fs_info->super_for_commit; |
1839 | dev_item = &sb->dev_item; | |
f2984462 CM |
1840 | list_for_each(cur, head) { |
1841 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1842 | if (!dev->bdev) { |
1843 | total_errors++; | |
1844 | continue; | |
1845 | } | |
2b82032c | 1846 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
1847 | continue; |
1848 | ||
2b82032c | 1849 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
1850 | btrfs_set_stack_device_type(dev_item, dev->type); |
1851 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
1852 | btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes); | |
1853 | btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used); | |
1854 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); | |
1855 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
1856 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
1857 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
2b82032c | 1858 | memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE); |
a061fc8d CM |
1859 | flags = btrfs_super_flags(sb); |
1860 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
1861 | ||
1862 | ||
1863 | crc = ~(u32)0; | |
1864 | crc = btrfs_csum_data(root, (char *)sb + BTRFS_CSUM_SIZE, crc, | |
1865 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE); | |
1866 | btrfs_csum_final(crc, sb->csum); | |
1867 | ||
1868 | bh = __getblk(dev->bdev, BTRFS_SUPER_INFO_OFFSET / 4096, | |
f2984462 CM |
1869 | BTRFS_SUPER_INFO_SIZE); |
1870 | ||
a061fc8d | 1871 | memcpy(bh->b_data, sb, BTRFS_SUPER_INFO_SIZE); |
f2984462 CM |
1872 | dev->pending_io = bh; |
1873 | ||
1874 | get_bh(bh); | |
1875 | set_buffer_uptodate(bh); | |
1876 | lock_buffer(bh); | |
1877 | bh->b_end_io = btrfs_end_buffer_write_sync; | |
1878 | ||
1879 | if (do_barriers && dev->barriers) { | |
1880 | ret = submit_bh(WRITE_BARRIER, bh); | |
1881 | if (ret == -EOPNOTSUPP) { | |
1882 | printk("btrfs: disabling barriers on dev %s\n", | |
1883 | dev->name); | |
1884 | set_buffer_uptodate(bh); | |
1885 | dev->barriers = 0; | |
1886 | get_bh(bh); | |
1887 | lock_buffer(bh); | |
1888 | ret = submit_bh(WRITE, bh); | |
1889 | } | |
1890 | } else { | |
1891 | ret = submit_bh(WRITE, bh); | |
1892 | } | |
a236aed1 CM |
1893 | if (ret) |
1894 | total_errors++; | |
f2984462 | 1895 | } |
a236aed1 CM |
1896 | if (total_errors > max_errors) { |
1897 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1898 | BUG(); | |
1899 | } | |
1900 | total_errors = 0; | |
f2984462 CM |
1901 | |
1902 | list_for_each(cur, head) { | |
1903 | dev = list_entry(cur, struct btrfs_device, dev_list); | |
dfe25020 CM |
1904 | if (!dev->bdev) |
1905 | continue; | |
2b82032c | 1906 | if (!dev->in_fs_metadata || !dev->writeable) |
dfe25020 CM |
1907 | continue; |
1908 | ||
f2984462 CM |
1909 | BUG_ON(!dev->pending_io); |
1910 | bh = dev->pending_io; | |
1911 | wait_on_buffer(bh); | |
1912 | if (!buffer_uptodate(dev->pending_io)) { | |
1913 | if (do_barriers && dev->barriers) { | |
1914 | printk("btrfs: disabling barriers on dev %s\n", | |
1915 | dev->name); | |
1916 | set_buffer_uptodate(bh); | |
1917 | get_bh(bh); | |
1918 | lock_buffer(bh); | |
1919 | dev->barriers = 0; | |
1920 | ret = submit_bh(WRITE, bh); | |
1921 | BUG_ON(ret); | |
1922 | wait_on_buffer(bh); | |
1259ab75 CM |
1923 | if (!buffer_uptodate(bh)) |
1924 | total_errors++; | |
f2984462 | 1925 | } else { |
a236aed1 | 1926 | total_errors++; |
f2984462 CM |
1927 | } |
1928 | ||
1929 | } | |
1930 | dev->pending_io = NULL; | |
1931 | brelse(bh); | |
1932 | } | |
a236aed1 CM |
1933 | if (total_errors > max_errors) { |
1934 | printk("btrfs: %d errors while writing supers\n", total_errors); | |
1935 | BUG(); | |
1936 | } | |
f2984462 CM |
1937 | return 0; |
1938 | } | |
1939 | ||
e089f05c | 1940 | int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root |
79154b1b | 1941 | *root) |
eb60ceac | 1942 | { |
e66f709b | 1943 | int ret; |
5f39d397 | 1944 | |
f2984462 | 1945 | ret = write_all_supers(root); |
5f39d397 | 1946 | return ret; |
cfaa7295 CM |
1947 | } |
1948 | ||
5eda7b5e | 1949 | int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root) |
2619ba1f CM |
1950 | { |
1951 | radix_tree_delete(&fs_info->fs_roots_radix, | |
1952 | (unsigned long)root->root_key.objectid); | |
b99aa6cb CM |
1953 | if (root->in_sysfs) |
1954 | btrfs_sysfs_del_root(root); | |
2619ba1f | 1955 | if (root->node) |
5f39d397 | 1956 | free_extent_buffer(root->node); |
2619ba1f | 1957 | if (root->commit_root) |
5f39d397 | 1958 | free_extent_buffer(root->commit_root); |
58176a96 JB |
1959 | if (root->name) |
1960 | kfree(root->name); | |
2619ba1f CM |
1961 | kfree(root); |
1962 | return 0; | |
1963 | } | |
1964 | ||
35b7e476 | 1965 | static int del_fs_roots(struct btrfs_fs_info *fs_info) |
0f7d52f4 CM |
1966 | { |
1967 | int ret; | |
1968 | struct btrfs_root *gang[8]; | |
1969 | int i; | |
1970 | ||
1971 | while(1) { | |
1972 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1973 | (void **)gang, 0, | |
1974 | ARRAY_SIZE(gang)); | |
1975 | if (!ret) | |
1976 | break; | |
2619ba1f | 1977 | for (i = 0; i < ret; i++) |
5eda7b5e | 1978 | btrfs_free_fs_root(fs_info, gang[i]); |
0f7d52f4 CM |
1979 | } |
1980 | return 0; | |
1981 | } | |
b4100d64 | 1982 | |
c146afad | 1983 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 1984 | { |
c146afad YZ |
1985 | u64 root_objectid = 0; |
1986 | struct btrfs_root *gang[8]; | |
1987 | int i; | |
3768f368 | 1988 | int ret; |
e089f05c | 1989 | |
c146afad YZ |
1990 | while (1) { |
1991 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1992 | (void **)gang, root_objectid, | |
1993 | ARRAY_SIZE(gang)); | |
1994 | if (!ret) | |
1995 | break; | |
1996 | for (i = 0; i < ret; i++) { | |
1997 | root_objectid = gang[i]->root_key.objectid; | |
1998 | ret = btrfs_find_dead_roots(fs_info->tree_root, | |
1999 | root_objectid, gang[i]); | |
2000 | BUG_ON(ret); | |
2001 | btrfs_orphan_cleanup(gang[i]); | |
2002 | } | |
2003 | root_objectid++; | |
2004 | } | |
2005 | return 0; | |
2006 | } | |
a2135011 | 2007 | |
c146afad YZ |
2008 | int btrfs_commit_super(struct btrfs_root *root) |
2009 | { | |
2010 | struct btrfs_trans_handle *trans; | |
2011 | int ret; | |
a74a4b97 | 2012 | |
c146afad | 2013 | mutex_lock(&root->fs_info->cleaner_mutex); |
a74a4b97 | 2014 | btrfs_clean_old_snapshots(root); |
c146afad | 2015 | mutex_unlock(&root->fs_info->cleaner_mutex); |
79154b1b | 2016 | trans = btrfs_start_transaction(root, 1); |
54aa1f4d | 2017 | ret = btrfs_commit_transaction(trans, root); |
c146afad YZ |
2018 | BUG_ON(ret); |
2019 | /* run commit again to drop the original snapshot */ | |
79154b1b CM |
2020 | trans = btrfs_start_transaction(root, 1); |
2021 | btrfs_commit_transaction(trans, root); | |
2022 | ret = btrfs_write_and_wait_transaction(NULL, root); | |
3768f368 | 2023 | BUG_ON(ret); |
d6bfde87 | 2024 | |
c146afad YZ |
2025 | ret = write_ctree_super(NULL, root); |
2026 | return ret; | |
2027 | } | |
2028 | ||
2029 | int close_ctree(struct btrfs_root *root) | |
2030 | { | |
2031 | struct btrfs_fs_info *fs_info = root->fs_info; | |
2032 | int ret; | |
2033 | ||
2034 | fs_info->closing = 1; | |
2035 | smp_mb(); | |
2036 | ||
2037 | kthread_stop(root->fs_info->transaction_kthread); | |
2038 | kthread_stop(root->fs_info->cleaner_kthread); | |
2039 | ||
2040 | if (!(fs_info->sb->s_flags & MS_RDONLY)) { | |
2041 | ret = btrfs_commit_super(root); | |
2042 | if (ret) { | |
2043 | printk("btrfs: commit super returns %d\n", ret); | |
2044 | } | |
2045 | } | |
0f7d52f4 | 2046 | |
b0c68f8b CM |
2047 | if (fs_info->delalloc_bytes) { |
2048 | printk("btrfs: at unmount delalloc count %Lu\n", | |
2049 | fs_info->delalloc_bytes); | |
2050 | } | |
31153d81 YZ |
2051 | if (fs_info->total_ref_cache_size) { |
2052 | printk("btrfs: at umount reference cache size %Lu\n", | |
2053 | fs_info->total_ref_cache_size); | |
2054 | } | |
bcc63abb | 2055 | |
0f7d52f4 | 2056 | if (fs_info->extent_root->node) |
5f39d397 | 2057 | free_extent_buffer(fs_info->extent_root->node); |
f510cfec | 2058 | |
0f7d52f4 | 2059 | if (fs_info->tree_root->node) |
5f39d397 | 2060 | free_extent_buffer(fs_info->tree_root->node); |
f510cfec | 2061 | |
0b86a832 CM |
2062 | if (root->fs_info->chunk_root->node); |
2063 | free_extent_buffer(root->fs_info->chunk_root->node); | |
2064 | ||
2065 | if (root->fs_info->dev_root->node); | |
2066 | free_extent_buffer(root->fs_info->dev_root->node); | |
2067 | ||
9078a3e1 | 2068 | btrfs_free_block_groups(root->fs_info); |
d10c5f31 | 2069 | |
c146afad | 2070 | del_fs_roots(fs_info); |
d10c5f31 | 2071 | |
c146afad | 2072 | iput(fs_info->btree_inode); |
9ad6b7bc | 2073 | |
247e743c | 2074 | btrfs_stop_workers(&fs_info->fixup_workers); |
771ed689 | 2075 | btrfs_stop_workers(&fs_info->delalloc_workers); |
8b712842 CM |
2076 | btrfs_stop_workers(&fs_info->workers); |
2077 | btrfs_stop_workers(&fs_info->endio_workers); | |
e6dcd2dc | 2078 | btrfs_stop_workers(&fs_info->endio_write_workers); |
1cc127b5 | 2079 | btrfs_stop_workers(&fs_info->submit_workers); |
d6bfde87 | 2080 | |
19c00ddc CM |
2081 | #if 0 |
2082 | while(!list_empty(&fs_info->hashers)) { | |
2083 | struct btrfs_hasher *hasher; | |
2084 | hasher = list_entry(fs_info->hashers.next, struct btrfs_hasher, | |
2085 | hashers); | |
2086 | list_del(&hasher->hashers); | |
2087 | crypto_free_hash(&fs_info->hash_tfm); | |
2088 | kfree(hasher); | |
2089 | } | |
2090 | #endif | |
dfe25020 | 2091 | btrfs_close_devices(fs_info->fs_devices); |
0b86a832 | 2092 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
b248a415 | 2093 | |
04160088 | 2094 | bdi_destroy(&fs_info->bdi); |
0b86a832 | 2095 | |
0f7d52f4 | 2096 | kfree(fs_info->extent_root); |
0f7d52f4 | 2097 | kfree(fs_info->tree_root); |
0b86a832 CM |
2098 | kfree(fs_info->chunk_root); |
2099 | kfree(fs_info->dev_root); | |
eb60ceac CM |
2100 | return 0; |
2101 | } | |
2102 | ||
1259ab75 | 2103 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid) |
5f39d397 | 2104 | { |
1259ab75 | 2105 | int ret; |
810191ff | 2106 | struct inode *btree_inode = buf->first_page->mapping->host; |
1259ab75 CM |
2107 | |
2108 | ret = extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, buf); | |
2109 | if (!ret) | |
2110 | return ret; | |
2111 | ||
2112 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
2113 | parent_transid); | |
2114 | return !ret; | |
5f39d397 CM |
2115 | } |
2116 | ||
2117 | int btrfs_set_buffer_uptodate(struct extent_buffer *buf) | |
ccd467d6 | 2118 | { |
810191ff | 2119 | struct inode *btree_inode = buf->first_page->mapping->host; |
d1310b2e | 2120 | return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->io_tree, |
5f39d397 CM |
2121 | buf); |
2122 | } | |
6702ed49 | 2123 | |
5f39d397 CM |
2124 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
2125 | { | |
810191ff | 2126 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
5f39d397 CM |
2127 | u64 transid = btrfs_header_generation(buf); |
2128 | struct inode *btree_inode = root->fs_info->btree_inode; | |
6702ed49 | 2129 | |
925baedd | 2130 | WARN_ON(!btrfs_tree_locked(buf)); |
ccd467d6 CM |
2131 | if (transid != root->fs_info->generation) { |
2132 | printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n", | |
db94535d | 2133 | (unsigned long long)buf->start, |
ccd467d6 CM |
2134 | transid, root->fs_info->generation); |
2135 | WARN_ON(1); | |
2136 | } | |
d1310b2e | 2137 | set_extent_buffer_dirty(&BTRFS_I(btree_inode)->io_tree, buf); |
eb60ceac CM |
2138 | } |
2139 | ||
d3c2fdcf | 2140 | void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr) |
35b7e476 | 2141 | { |
188de649 CM |
2142 | /* |
2143 | * looks as though older kernels can get into trouble with | |
2144 | * this code, they end up stuck in balance_dirty_pages forever | |
2145 | */ | |
d6bfde87 CM |
2146 | struct extent_io_tree *tree; |
2147 | u64 num_dirty; | |
2148 | u64 start = 0; | |
771ed689 | 2149 | unsigned long thresh = 32 * 1024 * 1024; |
d6bfde87 CM |
2150 | tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree; |
2151 | ||
b64a2851 | 2152 | if (current_is_pdflush() || current->flags & PF_MEMALLOC) |
d6bfde87 CM |
2153 | return; |
2154 | ||
2155 | num_dirty = count_range_bits(tree, &start, (u64)-1, | |
2156 | thresh, EXTENT_DIRTY); | |
2157 | if (num_dirty > thresh) { | |
2158 | balance_dirty_pages_ratelimited_nr( | |
d7fc640e | 2159 | root->fs_info->btree_inode->i_mapping, 1); |
d6bfde87 | 2160 | } |
188de649 | 2161 | return; |
35b7e476 | 2162 | } |
6b80053d | 2163 | |
ca7a79ad | 2164 | int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid) |
6b80053d | 2165 | { |
810191ff | 2166 | struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root; |
ce9adaa5 | 2167 | int ret; |
ca7a79ad | 2168 | ret = btree_read_extent_buffer_pages(root, buf, 0, parent_transid); |
ce9adaa5 CM |
2169 | if (ret == 0) { |
2170 | buf->flags |= EXTENT_UPTODATE; | |
2171 | } | |
2172 | return ret; | |
6b80053d | 2173 | } |
0da5468f | 2174 | |
4bef0848 CM |
2175 | int btree_lock_page_hook(struct page *page) |
2176 | { | |
2177 | struct inode *inode = page->mapping->host; | |
2178 | struct btrfs_root *root = BTRFS_I(inode)->root; | |
2179 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | |
2180 | struct extent_buffer *eb; | |
2181 | unsigned long len; | |
2182 | u64 bytenr = page_offset(page); | |
2183 | ||
2184 | if (page->private == EXTENT_PAGE_PRIVATE) | |
2185 | goto out; | |
2186 | ||
2187 | len = page->private >> 2; | |
2188 | eb = find_extent_buffer(io_tree, bytenr, len, GFP_NOFS); | |
2189 | if (!eb) | |
2190 | goto out; | |
2191 | ||
2192 | btrfs_tree_lock(eb); | |
2193 | spin_lock(&root->fs_info->hash_lock); | |
2194 | btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN); | |
2195 | spin_unlock(&root->fs_info->hash_lock); | |
2196 | btrfs_tree_unlock(eb); | |
2197 | free_extent_buffer(eb); | |
2198 | out: | |
2199 | lock_page(page); | |
2200 | return 0; | |
2201 | } | |
2202 | ||
d1310b2e | 2203 | static struct extent_io_ops btree_extent_io_ops = { |
4bef0848 | 2204 | .write_cache_pages_lock_hook = btree_lock_page_hook, |
ce9adaa5 | 2205 | .readpage_end_io_hook = btree_readpage_end_io_hook, |
0b86a832 | 2206 | .submit_bio_hook = btree_submit_bio_hook, |
239b14b3 CM |
2207 | /* note we're sharing with inode.c for the merge bio hook */ |
2208 | .merge_bio_hook = btrfs_merge_bio_hook, | |
0da5468f | 2209 | }; |