Commit | Line | Data |
---|---|---|
c1d7c514 | 1 | // SPDX-License-Identifier: GPL-2.0 |
6cbd5570 CM |
2 | /* |
3 | * Copyright (C) 2007 Oracle. All rights reserved. | |
6cbd5570 CM |
4 | */ |
5 | ||
e20d96d6 | 6 | #include <linux/fs.h> |
d98237b3 | 7 | #include <linux/blkdev.h> |
fc7cbcd4 | 8 | #include <linux/radix-tree.h> |
35b7e476 | 9 | #include <linux/writeback.h> |
ce9adaa5 | 10 | #include <linux/workqueue.h> |
a74a4b97 | 11 | #include <linux/kthread.h> |
5a0e3ad6 | 12 | #include <linux/slab.h> |
784b4e29 | 13 | #include <linux/migrate.h> |
7a36ddec | 14 | #include <linux/ratelimit.h> |
6463fe58 | 15 | #include <linux/uuid.h> |
803b2f54 | 16 | #include <linux/semaphore.h> |
540adea3 | 17 | #include <linux/error-injection.h> |
9678c543 | 18 | #include <linux/crc32c.h> |
b89f6d1f | 19 | #include <linux/sched/mm.h> |
7e75bf3f | 20 | #include <asm/unaligned.h> |
6d97c6e3 | 21 | #include <crypto/hash.h> |
eb60ceac CM |
22 | #include "ctree.h" |
23 | #include "disk-io.h" | |
e089f05c | 24 | #include "transaction.h" |
0f7d52f4 | 25 | #include "btrfs_inode.h" |
0b86a832 | 26 | #include "volumes.h" |
db94535d | 27 | #include "print-tree.h" |
925baedd | 28 | #include "locking.h" |
e02119d5 | 29 | #include "tree-log.h" |
fa9c0d79 | 30 | #include "free-space-cache.h" |
70f6d82e | 31 | #include "free-space-tree.h" |
21adbd5c | 32 | #include "check-integrity.h" |
606686ee | 33 | #include "rcu-string.h" |
8dabb742 | 34 | #include "dev-replace.h" |
53b381b3 | 35 | #include "raid56.h" |
5ac1d209 | 36 | #include "sysfs.h" |
fcebe456 | 37 | #include "qgroup.h" |
ebb8765b | 38 | #include "compression.h" |
557ea5dd | 39 | #include "tree-checker.h" |
fd708b81 | 40 | #include "ref-verify.h" |
aac0023c | 41 | #include "block-group.h" |
b0643e59 | 42 | #include "discard.h" |
f603bb94 | 43 | #include "space-info.h" |
b70f5097 | 44 | #include "zoned.h" |
139e8cd3 | 45 | #include "subpage.h" |
eb60ceac | 46 | |
319e4d06 QW |
47 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
48 | BTRFS_HEADER_FLAG_RELOC |\ | |
49 | BTRFS_SUPER_FLAG_ERROR |\ | |
50 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
51 | BTRFS_SUPER_FLAG_METADUMP |\ |
52 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 53 | |
143bede5 | 54 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 55 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 56 | struct btrfs_fs_info *fs_info); |
143bede5 | 57 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
2ff7e61e | 58 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 59 | struct extent_io_tree *dirty_pages, |
60 | int mark); | |
2ff7e61e | 61 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 62 | struct extent_io_tree *pinned_extents); |
2ff7e61e JM |
63 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
64 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 65 | |
141386e1 JB |
66 | static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) |
67 | { | |
68 | if (fs_info->csum_shash) | |
69 | crypto_free_shash(fs_info->csum_shash); | |
70 | } | |
71 | ||
d352ac68 CM |
72 | /* |
73 | * async submit bios are used to offload expensive checksumming | |
74 | * onto the worker threads. They checksum file and metadata bios | |
75 | * just before they are sent down the IO stack. | |
76 | */ | |
44b8bd7e | 77 | struct async_submit_bio { |
8896a08d | 78 | struct inode *inode; |
44b8bd7e | 79 | struct bio *bio; |
a758781d | 80 | extent_submit_bio_start_t *submit_bio_start; |
44b8bd7e | 81 | int mirror_num; |
1941b64b QW |
82 | |
83 | /* Optional parameter for submit_bio_start used by direct io */ | |
84 | u64 dio_file_offset; | |
8b712842 | 85 | struct btrfs_work work; |
4e4cbee9 | 86 | blk_status_t status; |
44b8bd7e CM |
87 | }; |
88 | ||
d352ac68 | 89 | /* |
2996e1f8 | 90 | * Compute the csum of a btree block and store the result to provided buffer. |
d352ac68 | 91 | */ |
c67b3892 | 92 | static void csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 93 | { |
d5178578 | 94 | struct btrfs_fs_info *fs_info = buf->fs_info; |
7280305e | 95 | const int num_pages = num_extent_pages(buf); |
a26663e7 | 96 | const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); |
d5178578 | 97 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
19c00ddc | 98 | char *kaddr; |
e9be5a30 | 99 | int i; |
d5178578 JT |
100 | |
101 | shash->tfm = fs_info->csum_shash; | |
102 | crypto_shash_init(shash); | |
a26663e7 | 103 | kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); |
e9be5a30 | 104 | crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, |
a26663e7 | 105 | first_page_part - BTRFS_CSUM_SIZE); |
19c00ddc | 106 | |
e9be5a30 DS |
107 | for (i = 1; i < num_pages; i++) { |
108 | kaddr = page_address(buf->pages[i]); | |
109 | crypto_shash_update(shash, kaddr, PAGE_SIZE); | |
19c00ddc | 110 | } |
71a63551 | 111 | memset(result, 0, BTRFS_CSUM_SIZE); |
d5178578 | 112 | crypto_shash_final(shash, result); |
19c00ddc CM |
113 | } |
114 | ||
d352ac68 CM |
115 | /* |
116 | * we can't consider a given block up to date unless the transid of the | |
117 | * block matches the transid in the parent node's pointer. This is how we | |
118 | * detect blocks that either didn't get written at all or got written | |
119 | * in the wrong place. | |
120 | */ | |
1259ab75 | 121 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
122 | struct extent_buffer *eb, u64 parent_transid, |
123 | int atomic) | |
1259ab75 | 124 | { |
2ac55d41 | 125 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
126 | int ret; |
127 | ||
128 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
129 | return 0; | |
130 | ||
b9fab919 CM |
131 | if (atomic) |
132 | return -EAGAIN; | |
133 | ||
570eb97b | 134 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, &cached_state); |
0b32f4bb | 135 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
136 | btrfs_header_generation(eb) == parent_transid) { |
137 | ret = 0; | |
138 | goto out; | |
139 | } | |
94647322 | 140 | btrfs_err_rl(eb->fs_info, |
8f0ed7d4 QW |
141 | "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", |
142 | eb->start, eb->read_mirror, | |
29549aec | 143 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 144 | ret = 1; |
35b22c19 | 145 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 146 | out: |
570eb97b JB |
147 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
148 | &cached_state); | |
1259ab75 | 149 | return ret; |
1259ab75 CM |
150 | } |
151 | ||
e7e16f48 JT |
152 | static bool btrfs_supported_super_csum(u16 csum_type) |
153 | { | |
154 | switch (csum_type) { | |
155 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 156 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 157 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 158 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
159 | return true; |
160 | default: | |
161 | return false; | |
162 | } | |
163 | } | |
164 | ||
1104a885 DS |
165 | /* |
166 | * Return 0 if the superblock checksum type matches the checksum value of that | |
167 | * algorithm. Pass the raw disk superblock data. | |
168 | */ | |
ab8d0fc4 JM |
169 | static int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
170 | char *raw_disk_sb) | |
1104a885 DS |
171 | { |
172 | struct btrfs_super_block *disk_sb = | |
173 | (struct btrfs_super_block *)raw_disk_sb; | |
51bce6c9 | 174 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
175 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
176 | ||
177 | shash->tfm = fs_info->csum_shash; | |
1104a885 | 178 | |
51bce6c9 JT |
179 | /* |
180 | * The super_block structure does not span the whole | |
181 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
182 | * filled with zeros and is included in the checksum. | |
183 | */ | |
fd08001f EB |
184 | crypto_shash_digest(shash, raw_disk_sb + BTRFS_CSUM_SIZE, |
185 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); | |
1104a885 | 186 | |
55fc29be | 187 | if (memcmp(disk_sb->csum, result, fs_info->csum_size)) |
51bce6c9 | 188 | return 1; |
1104a885 | 189 | |
e7e16f48 | 190 | return 0; |
1104a885 DS |
191 | } |
192 | ||
e064d5e9 | 193 | int btrfs_verify_level_key(struct extent_buffer *eb, int level, |
448de471 | 194 | struct btrfs_key *first_key, u64 parent_transid) |
581c1760 | 195 | { |
e064d5e9 | 196 | struct btrfs_fs_info *fs_info = eb->fs_info; |
581c1760 QW |
197 | int found_level; |
198 | struct btrfs_key found_key; | |
199 | int ret; | |
200 | ||
201 | found_level = btrfs_header_level(eb); | |
202 | if (found_level != level) { | |
63489055 QW |
203 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
204 | KERN_ERR "BTRFS: tree level check failed\n"); | |
581c1760 QW |
205 | btrfs_err(fs_info, |
206 | "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", | |
207 | eb->start, level, found_level); | |
581c1760 QW |
208 | return -EIO; |
209 | } | |
210 | ||
211 | if (!first_key) | |
212 | return 0; | |
213 | ||
5d41be6f QW |
214 | /* |
215 | * For live tree block (new tree blocks in current transaction), | |
216 | * we need proper lock context to avoid race, which is impossible here. | |
217 | * So we only checks tree blocks which is read from disk, whose | |
218 | * generation <= fs_info->last_trans_committed. | |
219 | */ | |
220 | if (btrfs_header_generation(eb) > fs_info->last_trans_committed) | |
221 | return 0; | |
62fdaa52 QW |
222 | |
223 | /* We have @first_key, so this @eb must have at least one item */ | |
224 | if (btrfs_header_nritems(eb) == 0) { | |
225 | btrfs_err(fs_info, | |
226 | "invalid tree nritems, bytenr=%llu nritems=0 expect >0", | |
227 | eb->start); | |
228 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
229 | return -EUCLEAN; | |
230 | } | |
231 | ||
581c1760 QW |
232 | if (found_level) |
233 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
234 | else | |
235 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
236 | ret = btrfs_comp_cpu_keys(first_key, &found_key); | |
237 | ||
581c1760 | 238 | if (ret) { |
63489055 QW |
239 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
240 | KERN_ERR "BTRFS: tree first key check failed\n"); | |
581c1760 | 241 | btrfs_err(fs_info, |
ff76a864 LB |
242 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", |
243 | eb->start, parent_transid, first_key->objectid, | |
244 | first_key->type, first_key->offset, | |
245 | found_key.objectid, found_key.type, | |
246 | found_key.offset); | |
581c1760 | 247 | } |
581c1760 QW |
248 | return ret; |
249 | } | |
250 | ||
d352ac68 CM |
251 | /* |
252 | * helper to read a given tree block, doing retries as required when | |
253 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 QW |
254 | * |
255 | * @parent_transid: expected transid, skip check if 0 | |
256 | * @level: expected level, mandatory check | |
257 | * @first_key: expected key of first slot, skip check if NULL | |
d352ac68 | 258 | */ |
6a2e9dc4 FM |
259 | int btrfs_read_extent_buffer(struct extent_buffer *eb, |
260 | u64 parent_transid, int level, | |
261 | struct btrfs_key *first_key) | |
f188591e | 262 | { |
5ab12d1f | 263 | struct btrfs_fs_info *fs_info = eb->fs_info; |
f188591e | 264 | struct extent_io_tree *io_tree; |
ea466794 | 265 | int failed = 0; |
f188591e CM |
266 | int ret; |
267 | int num_copies = 0; | |
268 | int mirror_num = 0; | |
ea466794 | 269 | int failed_mirror = 0; |
f188591e | 270 | |
0b246afa | 271 | io_tree = &BTRFS_I(fs_info->btree_inode)->io_tree; |
f188591e | 272 | while (1) { |
f8397d69 | 273 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
c2ccfbc6 | 274 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num); |
256dd1bb | 275 | if (!ret) { |
581c1760 | 276 | if (verify_parent_transid(io_tree, eb, |
b9fab919 | 277 | parent_transid, 0)) |
256dd1bb | 278 | ret = -EIO; |
e064d5e9 | 279 | else if (btrfs_verify_level_key(eb, level, |
448de471 | 280 | first_key, parent_transid)) |
581c1760 QW |
281 | ret = -EUCLEAN; |
282 | else | |
283 | break; | |
256dd1bb | 284 | } |
d397712b | 285 | |
0b246afa | 286 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 287 | eb->start, eb->len); |
4235298e | 288 | if (num_copies == 1) |
ea466794 | 289 | break; |
4235298e | 290 | |
5cf1ab56 JB |
291 | if (!failed_mirror) { |
292 | failed = 1; | |
293 | failed_mirror = eb->read_mirror; | |
294 | } | |
295 | ||
f188591e | 296 | mirror_num++; |
ea466794 JB |
297 | if (mirror_num == failed_mirror) |
298 | mirror_num++; | |
299 | ||
4235298e | 300 | if (mirror_num > num_copies) |
ea466794 | 301 | break; |
f188591e | 302 | } |
ea466794 | 303 | |
c0901581 | 304 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 305 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
306 | |
307 | return ret; | |
f188591e | 308 | } |
19c00ddc | 309 | |
eca0f6f6 QW |
310 | static int csum_one_extent_buffer(struct extent_buffer *eb) |
311 | { | |
312 | struct btrfs_fs_info *fs_info = eb->fs_info; | |
313 | u8 result[BTRFS_CSUM_SIZE]; | |
314 | int ret; | |
315 | ||
316 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, | |
317 | offsetof(struct btrfs_header, fsid), | |
318 | BTRFS_FSID_SIZE) == 0); | |
319 | csum_tree_block(eb, result); | |
320 | ||
321 | if (btrfs_header_level(eb)) | |
322 | ret = btrfs_check_node(eb); | |
323 | else | |
324 | ret = btrfs_check_leaf_full(eb); | |
325 | ||
3777369f QW |
326 | if (ret < 0) |
327 | goto error; | |
328 | ||
329 | /* | |
330 | * Also check the generation, the eb reached here must be newer than | |
331 | * last committed. Or something seriously wrong happened. | |
332 | */ | |
333 | if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { | |
334 | ret = -EUCLEAN; | |
eca0f6f6 | 335 | btrfs_err(fs_info, |
3777369f QW |
336 | "block=%llu bad generation, have %llu expect > %llu", |
337 | eb->start, btrfs_header_generation(eb), | |
338 | fs_info->last_trans_committed); | |
339 | goto error; | |
eca0f6f6 QW |
340 | } |
341 | write_extent_buffer(eb, result, 0, fs_info->csum_size); | |
342 | ||
343 | return 0; | |
3777369f QW |
344 | |
345 | error: | |
346 | btrfs_print_tree(eb, 0); | |
347 | btrfs_err(fs_info, "block=%llu write time tree block corruption detected", | |
348 | eb->start); | |
349 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
350 | return ret; | |
eca0f6f6 QW |
351 | } |
352 | ||
353 | /* Checksum all dirty extent buffers in one bio_vec */ | |
354 | static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, | |
355 | struct bio_vec *bvec) | |
356 | { | |
357 | struct page *page = bvec->bv_page; | |
358 | u64 bvec_start = page_offset(page) + bvec->bv_offset; | |
359 | u64 cur; | |
360 | int ret = 0; | |
361 | ||
362 | for (cur = bvec_start; cur < bvec_start + bvec->bv_len; | |
363 | cur += fs_info->nodesize) { | |
364 | struct extent_buffer *eb; | |
365 | bool uptodate; | |
366 | ||
367 | eb = find_extent_buffer(fs_info, cur); | |
368 | uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, | |
369 | fs_info->nodesize); | |
370 | ||
01cd3909 | 371 | /* A dirty eb shouldn't disappear from buffer_radix */ |
eca0f6f6 QW |
372 | if (WARN_ON(!eb)) |
373 | return -EUCLEAN; | |
374 | ||
375 | if (WARN_ON(cur != btrfs_header_bytenr(eb))) { | |
376 | free_extent_buffer(eb); | |
377 | return -EUCLEAN; | |
378 | } | |
379 | if (WARN_ON(!uptodate)) { | |
380 | free_extent_buffer(eb); | |
381 | return -EUCLEAN; | |
382 | } | |
383 | ||
384 | ret = csum_one_extent_buffer(eb); | |
385 | free_extent_buffer(eb); | |
386 | if (ret < 0) | |
387 | return ret; | |
388 | } | |
389 | return ret; | |
390 | } | |
391 | ||
d352ac68 | 392 | /* |
ac303b69 QW |
393 | * Checksum a dirty tree block before IO. This has extra checks to make sure |
394 | * we only fill in the checksum field in the first page of a multi-page block. | |
395 | * For subpage extent buffers we need bvec to also read the offset in the page. | |
d352ac68 | 396 | */ |
ac303b69 | 397 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) |
19c00ddc | 398 | { |
ac303b69 | 399 | struct page *page = bvec->bv_page; |
4eee4fa4 | 400 | u64 start = page_offset(page); |
19c00ddc | 401 | u64 found_start; |
19c00ddc | 402 | struct extent_buffer *eb; |
eca0f6f6 | 403 | |
fbca46eb | 404 | if (fs_info->nodesize < PAGE_SIZE) |
eca0f6f6 | 405 | return csum_dirty_subpage_buffers(fs_info, bvec); |
f188591e | 406 | |
4f2de97a JB |
407 | eb = (struct extent_buffer *)page->private; |
408 | if (page != eb->pages[0]) | |
409 | return 0; | |
0f805531 | 410 | |
19c00ddc | 411 | found_start = btrfs_header_bytenr(eb); |
d3575156 NA |
412 | |
413 | if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { | |
414 | WARN_ON(found_start != 0); | |
415 | return 0; | |
416 | } | |
417 | ||
0f805531 AL |
418 | /* |
419 | * Please do not consolidate these warnings into a single if. | |
420 | * It is useful to know what went wrong. | |
421 | */ | |
422 | if (WARN_ON(found_start != start)) | |
423 | return -EUCLEAN; | |
424 | if (WARN_ON(!PageUptodate(page))) | |
425 | return -EUCLEAN; | |
426 | ||
eca0f6f6 | 427 | return csum_one_extent_buffer(eb); |
19c00ddc CM |
428 | } |
429 | ||
b0c9b3b0 | 430 | static int check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 431 | { |
b0c9b3b0 | 432 | struct btrfs_fs_info *fs_info = eb->fs_info; |
944d3f9f | 433 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; |
44880fdc | 434 | u8 fsid[BTRFS_FSID_SIZE]; |
944d3f9f | 435 | u8 *metadata_uuid; |
2b82032c | 436 | |
9a8658e3 DS |
437 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
438 | BTRFS_FSID_SIZE); | |
944d3f9f NB |
439 | /* |
440 | * Checking the incompat flag is only valid for the current fs. For | |
441 | * seed devices it's forbidden to have their uuid changed so reading | |
442 | * ->fsid in this case is fine | |
443 | */ | |
444 | if (btrfs_fs_incompat(fs_info, METADATA_UUID)) | |
445 | metadata_uuid = fs_devices->metadata_uuid; | |
446 | else | |
447 | metadata_uuid = fs_devices->fsid; | |
448 | ||
449 | if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) | |
450 | return 0; | |
451 | ||
452 | list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) | |
453 | if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) | |
454 | return 0; | |
455 | ||
456 | return 1; | |
2b82032c YZ |
457 | } |
458 | ||
77bf40a2 QW |
459 | /* Do basic extent buffer checks at read time */ |
460 | static int validate_extent_buffer(struct extent_buffer *eb) | |
ce9adaa5 | 461 | { |
77bf40a2 | 462 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ce9adaa5 | 463 | u64 found_start; |
77bf40a2 QW |
464 | const u32 csum_size = fs_info->csum_size; |
465 | u8 found_level; | |
2996e1f8 | 466 | u8 result[BTRFS_CSUM_SIZE]; |
dfd29eed | 467 | const u8 *header_csum; |
77bf40a2 | 468 | int ret = 0; |
ea466794 | 469 | |
ce9adaa5 | 470 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 471 | if (found_start != eb->start) { |
8f0ed7d4 QW |
472 | btrfs_err_rl(fs_info, |
473 | "bad tree block start, mirror %u want %llu have %llu", | |
474 | eb->read_mirror, eb->start, found_start); | |
f188591e | 475 | ret = -EIO; |
77bf40a2 | 476 | goto out; |
ce9adaa5 | 477 | } |
b0c9b3b0 | 478 | if (check_tree_block_fsid(eb)) { |
8f0ed7d4 QW |
479 | btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", |
480 | eb->start, eb->read_mirror); | |
1259ab75 | 481 | ret = -EIO; |
77bf40a2 | 482 | goto out; |
1259ab75 | 483 | } |
ce9adaa5 | 484 | found_level = btrfs_header_level(eb); |
1c24c3ce | 485 | if (found_level >= BTRFS_MAX_LEVEL) { |
8f0ed7d4 QW |
486 | btrfs_err(fs_info, |
487 | "bad tree block level, mirror %u level %d on logical %llu", | |
488 | eb->read_mirror, btrfs_header_level(eb), eb->start); | |
1c24c3ce | 489 | ret = -EIO; |
77bf40a2 | 490 | goto out; |
1c24c3ce | 491 | } |
ce9adaa5 | 492 | |
c67b3892 | 493 | csum_tree_block(eb, result); |
dfd29eed DS |
494 | header_csum = page_address(eb->pages[0]) + |
495 | get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); | |
a826d6dc | 496 | |
dfd29eed | 497 | if (memcmp(result, header_csum, csum_size) != 0) { |
2996e1f8 | 498 | btrfs_warn_rl(fs_info, |
8f0ed7d4 QW |
499 | "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", |
500 | eb->start, eb->read_mirror, | |
dfd29eed | 501 | CSUM_FMT_VALUE(csum_size, header_csum), |
35be8851 JT |
502 | CSUM_FMT_VALUE(csum_size, result), |
503 | btrfs_header_level(eb)); | |
2996e1f8 | 504 | ret = -EUCLEAN; |
77bf40a2 | 505 | goto out; |
2996e1f8 JT |
506 | } |
507 | ||
a826d6dc JB |
508 | /* |
509 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
510 | * that we don't try and read the other copies of this block, just | |
511 | * return -EIO. | |
512 | */ | |
1c4360ee | 513 | if (found_level == 0 && btrfs_check_leaf_full(eb)) { |
a826d6dc JB |
514 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
515 | ret = -EIO; | |
516 | } | |
ce9adaa5 | 517 | |
813fd1dc | 518 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
519 | ret = -EIO; |
520 | ||
0b32f4bb JB |
521 | if (!ret) |
522 | set_extent_buffer_uptodate(eb); | |
75391f0d QW |
523 | else |
524 | btrfs_err(fs_info, | |
8f0ed7d4 QW |
525 | "read time tree block corruption detected on logical %llu mirror %u", |
526 | eb->start, eb->read_mirror); | |
77bf40a2 QW |
527 | out: |
528 | return ret; | |
529 | } | |
530 | ||
371cdc07 QW |
531 | static int validate_subpage_buffer(struct page *page, u64 start, u64 end, |
532 | int mirror) | |
533 | { | |
534 | struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); | |
535 | struct extent_buffer *eb; | |
536 | bool reads_done; | |
537 | int ret = 0; | |
538 | ||
539 | /* | |
540 | * We don't allow bio merge for subpage metadata read, so we should | |
541 | * only get one eb for each endio hook. | |
542 | */ | |
543 | ASSERT(end == start + fs_info->nodesize - 1); | |
544 | ASSERT(PagePrivate(page)); | |
545 | ||
546 | eb = find_extent_buffer(fs_info, start); | |
547 | /* | |
548 | * When we are reading one tree block, eb must have been inserted into | |
549 | * the radix tree. If not, something is wrong. | |
550 | */ | |
551 | ASSERT(eb); | |
552 | ||
553 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
554 | /* Subpage read must finish in page read */ | |
555 | ASSERT(reads_done); | |
556 | ||
557 | eb->read_mirror = mirror; | |
558 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
559 | ret = -EIO; | |
560 | goto err; | |
561 | } | |
562 | ret = validate_extent_buffer(eb); | |
563 | if (ret < 0) | |
564 | goto err; | |
565 | ||
371cdc07 QW |
566 | set_extent_buffer_uptodate(eb); |
567 | ||
568 | free_extent_buffer(eb); | |
569 | return ret; | |
570 | err: | |
571 | /* | |
572 | * end_bio_extent_readpage decrements io_pages in case of error, | |
573 | * make sure it has something to decrement. | |
574 | */ | |
575 | atomic_inc(&eb->io_pages); | |
576 | clear_extent_buffer_uptodate(eb); | |
577 | free_extent_buffer(eb); | |
578 | return ret; | |
579 | } | |
580 | ||
c3a3b19b | 581 | int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, |
77bf40a2 QW |
582 | struct page *page, u64 start, u64 end, |
583 | int mirror) | |
584 | { | |
585 | struct extent_buffer *eb; | |
586 | int ret = 0; | |
587 | int reads_done; | |
588 | ||
589 | ASSERT(page->private); | |
371cdc07 | 590 | |
fbca46eb | 591 | if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) |
371cdc07 QW |
592 | return validate_subpage_buffer(page, start, end, mirror); |
593 | ||
77bf40a2 QW |
594 | eb = (struct extent_buffer *)page->private; |
595 | ||
596 | /* | |
597 | * The pending IO might have been the only thing that kept this buffer | |
598 | * in memory. Make sure we have a ref for all this other checks | |
599 | */ | |
600 | atomic_inc(&eb->refs); | |
601 | ||
602 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
603 | if (!reads_done) | |
604 | goto err; | |
605 | ||
606 | eb->read_mirror = mirror; | |
607 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
608 | ret = -EIO; | |
609 | goto err; | |
610 | } | |
611 | ret = validate_extent_buffer(eb); | |
ce9adaa5 | 612 | err: |
53b381b3 DW |
613 | if (ret) { |
614 | /* | |
615 | * our io error hook is going to dec the io pages | |
616 | * again, we have to make sure it has something | |
617 | * to decrement | |
618 | */ | |
619 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 620 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 621 | } |
0b32f4bb | 622 | free_extent_buffer(eb); |
77bf40a2 | 623 | |
f188591e | 624 | return ret; |
ce9adaa5 CM |
625 | } |
626 | ||
4a69a410 CM |
627 | static void run_one_async_start(struct btrfs_work *work) |
628 | { | |
4a69a410 | 629 | struct async_submit_bio *async; |
4e4cbee9 | 630 | blk_status_t ret; |
4a69a410 CM |
631 | |
632 | async = container_of(work, struct async_submit_bio, work); | |
1941b64b QW |
633 | ret = async->submit_bio_start(async->inode, async->bio, |
634 | async->dio_file_offset); | |
79787eaa | 635 | if (ret) |
4e4cbee9 | 636 | async->status = ret; |
4a69a410 CM |
637 | } |
638 | ||
06ea01b1 DS |
639 | /* |
640 | * In order to insert checksums into the metadata in large chunks, we wait | |
641 | * until bio submission time. All the pages in the bio are checksummed and | |
642 | * sums are attached onto the ordered extent record. | |
643 | * | |
644 | * At IO completion time the csums attached on the ordered extent record are | |
645 | * inserted into the tree. | |
646 | */ | |
4a69a410 | 647 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 | 648 | { |
917f32a2 CH |
649 | struct async_submit_bio *async = |
650 | container_of(work, struct async_submit_bio, work); | |
651 | struct inode *inode = async->inode; | |
652 | struct btrfs_bio *bbio = btrfs_bio(async->bio); | |
4854ddd0 | 653 | |
bb7ab3b9 | 654 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 | 655 | if (async->status) { |
917f32a2 | 656 | btrfs_bio_end_io(bbio, async->status); |
79787eaa JM |
657 | return; |
658 | } | |
659 | ||
ec39f769 CM |
660 | /* |
661 | * All of the bios that pass through here are from async helpers. | |
662 | * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. | |
663 | * This changes nothing when cgroups aren't in use. | |
664 | */ | |
665 | async->bio->bi_opf |= REQ_CGROUP_PUNT; | |
1a722d8f | 666 | btrfs_submit_bio(btrfs_sb(inode->i_sb), async->bio, async->mirror_num); |
4a69a410 CM |
667 | } |
668 | ||
669 | static void run_one_async_free(struct btrfs_work *work) | |
670 | { | |
671 | struct async_submit_bio *async; | |
672 | ||
673 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
674 | kfree(async); |
675 | } | |
676 | ||
ea1f0ced CH |
677 | /* |
678 | * Submit bio to an async queue. | |
679 | * | |
680 | * Retrun: | |
681 | * - true if the work has been succesfuly submitted | |
682 | * - false in case of error | |
683 | */ | |
684 | bool btrfs_wq_submit_bio(struct inode *inode, struct bio *bio, int mirror_num, | |
685 | u64 dio_file_offset, | |
686 | extent_submit_bio_start_t *submit_bio_start) | |
44b8bd7e | 687 | { |
8896a08d | 688 | struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info; |
44b8bd7e CM |
689 | struct async_submit_bio *async; |
690 | ||
691 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
692 | if (!async) | |
ea1f0ced | 693 | return false; |
44b8bd7e | 694 | |
8896a08d | 695 | async->inode = inode; |
44b8bd7e CM |
696 | async->bio = bio; |
697 | async->mirror_num = mirror_num; | |
4a69a410 | 698 | async->submit_bio_start = submit_bio_start; |
4a69a410 | 699 | |
a0cac0ec OS |
700 | btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, |
701 | run_one_async_free); | |
4a69a410 | 702 | |
1941b64b | 703 | async->dio_file_offset = dio_file_offset; |
8c8bee1d | 704 | |
4e4cbee9 | 705 | async->status = 0; |
79787eaa | 706 | |
67f055c7 | 707 | if (op_is_sync(bio->bi_opf)) |
a31b4a43 CH |
708 | btrfs_queue_work(fs_info->hipri_workers, &async->work); |
709 | else | |
710 | btrfs_queue_work(fs_info->workers, &async->work); | |
ea1f0ced | 711 | return true; |
44b8bd7e CM |
712 | } |
713 | ||
4e4cbee9 | 714 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 715 | { |
2c30c71b | 716 | struct bio_vec *bvec; |
ce3ed71a | 717 | struct btrfs_root *root; |
2b070cfe | 718 | int ret = 0; |
6dc4f100 | 719 | struct bvec_iter_all iter_all; |
ce3ed71a | 720 | |
c09abff8 | 721 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2b070cfe | 722 | bio_for_each_segment_all(bvec, bio, iter_all) { |
ce3ed71a | 723 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
ac303b69 | 724 | ret = csum_dirty_buffer(root->fs_info, bvec); |
79787eaa JM |
725 | if (ret) |
726 | break; | |
ce3ed71a | 727 | } |
2c30c71b | 728 | |
4e4cbee9 | 729 | return errno_to_blk_status(ret); |
ce3ed71a CM |
730 | } |
731 | ||
8896a08d | 732 | static blk_status_t btree_submit_bio_start(struct inode *inode, struct bio *bio, |
1941b64b | 733 | u64 dio_file_offset) |
22c59948 | 734 | { |
8b712842 CM |
735 | /* |
736 | * when we're called for a write, we're already in the async | |
1a722d8f | 737 | * submission context. Just jump into btrfs_submit_bio. |
8b712842 | 738 | */ |
79787eaa | 739 | return btree_csum_one_bio(bio); |
4a69a410 | 740 | } |
22c59948 | 741 | |
f4dcfb30 | 742 | static bool should_async_write(struct btrfs_fs_info *fs_info, |
9b4e675a | 743 | struct btrfs_inode *bi) |
de0022b9 | 744 | { |
4eef29ef | 745 | if (btrfs_is_zoned(fs_info)) |
f4dcfb30 | 746 | return false; |
6300463b | 747 | if (atomic_read(&bi->sync_writers)) |
f4dcfb30 | 748 | return false; |
9b4e675a | 749 | if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) |
f4dcfb30 JT |
750 | return false; |
751 | return true; | |
de0022b9 JB |
752 | } |
753 | ||
94d9e11b | 754 | void btrfs_submit_metadata_bio(struct inode *inode, struct bio *bio, int mirror_num) |
44b8bd7e | 755 | { |
0b246afa | 756 | struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb); |
917f32a2 | 757 | struct btrfs_bio *bbio = btrfs_bio(bio); |
4e4cbee9 | 758 | blk_status_t ret; |
cad321ad | 759 | |
08a6f464 CH |
760 | bio->bi_opf |= REQ_META; |
761 | ||
cfe94440 | 762 | if (btrfs_op(bio) != BTRFS_MAP_WRITE) { |
1a722d8f CH |
763 | btrfs_submit_bio(fs_info, bio, mirror_num); |
764 | return; | |
44b8bd7e | 765 | } |
d313d7a3 | 766 | |
ea1f0ced CH |
767 | /* |
768 | * Kthread helpers are used to submit writes so that checksumming can | |
769 | * happen in parallel across all CPUs. | |
770 | */ | |
771 | if (should_async_write(fs_info, BTRFS_I(inode)) && | |
772 | btrfs_wq_submit_bio(inode, bio, mirror_num, 0, btree_submit_bio_start)) | |
773 | return; | |
774 | ||
775 | ret = btree_csum_one_bio(bio); | |
94d9e11b | 776 | if (ret) { |
917f32a2 | 777 | btrfs_bio_end_io(bbio, ret); |
ea1f0ced | 778 | return; |
94d9e11b | 779 | } |
ea1f0ced CH |
780 | |
781 | btrfs_submit_bio(fs_info, bio, mirror_num); | |
44b8bd7e CM |
782 | } |
783 | ||
3dd1462e | 784 | #ifdef CONFIG_MIGRATION |
8958b551 MWO |
785 | static int btree_migrate_folio(struct address_space *mapping, |
786 | struct folio *dst, struct folio *src, enum migrate_mode mode) | |
784b4e29 CM |
787 | { |
788 | /* | |
789 | * we can't safely write a btree page from here, | |
790 | * we haven't done the locking hook | |
791 | */ | |
8958b551 | 792 | if (folio_test_dirty(src)) |
784b4e29 CM |
793 | return -EAGAIN; |
794 | /* | |
795 | * Buffers may be managed in a filesystem specific way. | |
796 | * We must have no buffers or drop them. | |
797 | */ | |
8958b551 MWO |
798 | if (folio_get_private(src) && |
799 | !filemap_release_folio(src, GFP_KERNEL)) | |
784b4e29 | 800 | return -EAGAIN; |
54184650 | 801 | return migrate_folio(mapping, dst, src, mode); |
784b4e29 | 802 | } |
8958b551 MWO |
803 | #else |
804 | #define btree_migrate_folio NULL | |
3dd1462e | 805 | #endif |
784b4e29 | 806 | |
0da5468f CM |
807 | static int btree_writepages(struct address_space *mapping, |
808 | struct writeback_control *wbc) | |
809 | { | |
e2d84521 MX |
810 | struct btrfs_fs_info *fs_info; |
811 | int ret; | |
812 | ||
d8d5f3e1 | 813 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
814 | |
815 | if (wbc->for_kupdate) | |
816 | return 0; | |
817 | ||
e2d84521 | 818 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 819 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
820 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
821 | BTRFS_DIRTY_METADATA_THRESH, | |
822 | fs_info->dirty_metadata_batch); | |
e2d84521 | 823 | if (ret < 0) |
793955bc | 824 | return 0; |
793955bc | 825 | } |
0b32f4bb | 826 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
827 | } |
828 | ||
f913cff3 | 829 | static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) |
5f39d397 | 830 | { |
f913cff3 MWO |
831 | if (folio_test_writeback(folio) || folio_test_dirty(folio)) |
832 | return false; | |
0c4e538b | 833 | |
f913cff3 | 834 | return try_release_extent_buffer(&folio->page); |
d98237b3 CM |
835 | } |
836 | ||
895586eb MWO |
837 | static void btree_invalidate_folio(struct folio *folio, size_t offset, |
838 | size_t length) | |
d98237b3 | 839 | { |
d1310b2e | 840 | struct extent_io_tree *tree; |
895586eb MWO |
841 | tree = &BTRFS_I(folio->mapping->host)->io_tree; |
842 | extent_invalidate_folio(tree, folio, offset); | |
f913cff3 | 843 | btree_release_folio(folio, GFP_NOFS); |
895586eb MWO |
844 | if (folio_get_private(folio)) { |
845 | btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, | |
846 | "folio private not zero on folio %llu", | |
847 | (unsigned long long)folio_pos(folio)); | |
848 | folio_detach_private(folio); | |
9ad6b7bc | 849 | } |
d98237b3 CM |
850 | } |
851 | ||
bb146eb2 | 852 | #ifdef DEBUG |
0079c3b1 MWO |
853 | static bool btree_dirty_folio(struct address_space *mapping, |
854 | struct folio *folio) | |
855 | { | |
856 | struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); | |
139e8cd3 | 857 | struct btrfs_subpage *subpage; |
0b32f4bb | 858 | struct extent_buffer *eb; |
139e8cd3 | 859 | int cur_bit = 0; |
0079c3b1 | 860 | u64 page_start = folio_pos(folio); |
139e8cd3 QW |
861 | |
862 | if (fs_info->sectorsize == PAGE_SIZE) { | |
0079c3b1 | 863 | eb = folio_get_private(folio); |
139e8cd3 QW |
864 | BUG_ON(!eb); |
865 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
866 | BUG_ON(!atomic_read(&eb->refs)); | |
49d0c642 | 867 | btrfs_assert_tree_write_locked(eb); |
0079c3b1 | 868 | return filemap_dirty_folio(mapping, folio); |
139e8cd3 | 869 | } |
0079c3b1 | 870 | subpage = folio_get_private(folio); |
139e8cd3 QW |
871 | |
872 | ASSERT(subpage->dirty_bitmap); | |
873 | while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { | |
874 | unsigned long flags; | |
875 | u64 cur; | |
876 | u16 tmp = (1 << cur_bit); | |
877 | ||
878 | spin_lock_irqsave(&subpage->lock, flags); | |
879 | if (!(tmp & subpage->dirty_bitmap)) { | |
880 | spin_unlock_irqrestore(&subpage->lock, flags); | |
881 | cur_bit++; | |
882 | continue; | |
883 | } | |
884 | spin_unlock_irqrestore(&subpage->lock, flags); | |
885 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 886 | |
139e8cd3 QW |
887 | eb = find_extent_buffer(fs_info, cur); |
888 | ASSERT(eb); | |
889 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
890 | ASSERT(atomic_read(&eb->refs)); | |
49d0c642 | 891 | btrfs_assert_tree_write_locked(eb); |
139e8cd3 QW |
892 | free_extent_buffer(eb); |
893 | ||
894 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); | |
895 | } | |
0079c3b1 | 896 | return filemap_dirty_folio(mapping, folio); |
0b32f4bb | 897 | } |
0079c3b1 MWO |
898 | #else |
899 | #define btree_dirty_folio filemap_dirty_folio | |
900 | #endif | |
0b32f4bb | 901 | |
7f09410b | 902 | static const struct address_space_operations btree_aops = { |
0da5468f | 903 | .writepages = btree_writepages, |
f913cff3 | 904 | .release_folio = btree_release_folio, |
895586eb | 905 | .invalidate_folio = btree_invalidate_folio, |
8958b551 MWO |
906 | .migrate_folio = btree_migrate_folio, |
907 | .dirty_folio = btree_dirty_folio, | |
d98237b3 CM |
908 | }; |
909 | ||
2ff7e61e JM |
910 | struct extent_buffer *btrfs_find_create_tree_block( |
911 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
912 | u64 bytenr, u64 owner_root, |
913 | int level) | |
0999df54 | 914 | { |
0b246afa JM |
915 | if (btrfs_is_testing(fs_info)) |
916 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 917 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
918 | } |
919 | ||
581c1760 QW |
920 | /* |
921 | * Read tree block at logical address @bytenr and do variant basic but critical | |
922 | * verification. | |
923 | * | |
1b7ec85e | 924 | * @owner_root: the objectid of the root owner for this block. |
581c1760 QW |
925 | * @parent_transid: expected transid of this tree block, skip check if 0 |
926 | * @level: expected level, mandatory check | |
927 | * @first_key: expected key in slot 0, skip check if NULL | |
928 | */ | |
2ff7e61e | 929 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
1b7ec85e JB |
930 | u64 owner_root, u64 parent_transid, |
931 | int level, struct btrfs_key *first_key) | |
0999df54 CM |
932 | { |
933 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
934 | int ret; |
935 | ||
3fbaf258 | 936 | buf = btrfs_find_create_tree_block(fs_info, bytenr, owner_root, level); |
c871b0f2 LB |
937 | if (IS_ERR(buf)) |
938 | return buf; | |
0999df54 | 939 | |
6a2e9dc4 | 940 | ret = btrfs_read_extent_buffer(buf, parent_transid, level, first_key); |
0f0fe8f7 | 941 | if (ret) { |
537f38f0 | 942 | free_extent_buffer_stale(buf); |
64c043de | 943 | return ERR_PTR(ret); |
0f0fe8f7 | 944 | } |
88c602ab QW |
945 | if (btrfs_check_eb_owner(buf, owner_root)) { |
946 | free_extent_buffer_stale(buf); | |
947 | return ERR_PTR(-EUCLEAN); | |
948 | } | |
5f39d397 | 949 | return buf; |
ce9adaa5 | 950 | |
eb60ceac CM |
951 | } |
952 | ||
6a884d7d | 953 | void btrfs_clean_tree_block(struct extent_buffer *buf) |
ed2ff2cb | 954 | { |
6a884d7d | 955 | struct btrfs_fs_info *fs_info = buf->fs_info; |
55c69072 | 956 | if (btrfs_header_generation(buf) == |
e2d84521 | 957 | fs_info->running_transaction->transid) { |
49d0c642 | 958 | btrfs_assert_tree_write_locked(buf); |
b4ce94de | 959 | |
b9473439 | 960 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
961 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
962 | -buf->len, | |
963 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
964 | clear_extent_buffer_dirty(buf); |
965 | } | |
925baedd | 966 | } |
5f39d397 CM |
967 | } |
968 | ||
da17066c | 969 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 970 | u64 objectid) |
d97e63b6 | 971 | { |
7c0260ee | 972 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
2e608bd1 JB |
973 | |
974 | memset(&root->root_key, 0, sizeof(root->root_key)); | |
975 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
976 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); | |
96dfcb46 | 977 | root->fs_info = fs_info; |
2e608bd1 | 978 | root->root_key.objectid = objectid; |
cfaa7295 | 979 | root->node = NULL; |
a28ec197 | 980 | root->commit_root = NULL; |
27cdeb70 | 981 | root->state = 0; |
abed4aaa | 982 | RB_CLEAR_NODE(&root->rb_node); |
0b86a832 | 983 | |
0f7d52f4 | 984 | root->last_trans = 0; |
6b8fad57 | 985 | root->free_objectid = 0; |
eb73c1b7 | 986 | root->nr_delalloc_inodes = 0; |
199c2a9c | 987 | root->nr_ordered_extents = 0; |
6bef4d31 | 988 | root->inode_tree = RB_ROOT; |
088aea3b | 989 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
2e608bd1 JB |
990 | |
991 | btrfs_init_root_block_rsv(root); | |
0b86a832 CM |
992 | |
993 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 994 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
995 | INIT_LIST_HEAD(&root->delalloc_inodes); |
996 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
997 | INIT_LIST_HEAD(&root->ordered_extents); |
998 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 999 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
2ab28f32 JB |
1000 | INIT_LIST_HEAD(&root->logged_list[0]); |
1001 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1002 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1003 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1004 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1005 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1006 | spin_lock_init(&root->log_extents_lock[0]); |
1007 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1008 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1009 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1010 | mutex_init(&root->log_mutex); |
31f3d255 | 1011 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1012 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 1013 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
1014 | init_waitqueue_head(&root->log_writer_wait); |
1015 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1016 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1017 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1018 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1019 | atomic_set(&root->log_commit[0], 0); |
1020 | atomic_set(&root->log_commit[1], 0); | |
1021 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1022 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1023 | refcount_set(&root->refs, 1); |
8ecebf4d | 1024 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 1025 | atomic_set(&root->nr_swapfiles, 0); |
7237f183 | 1026 | root->log_transid = 0; |
d1433deb | 1027 | root->log_transid_committed = -1; |
257c62e1 | 1028 | root->last_log_commit = 0; |
2e608bd1 | 1029 | root->anon_dev = 0; |
e289f03e | 1030 | if (!dummy) { |
43eb5f29 QW |
1031 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
1032 | IO_TREE_ROOT_DIRTY_LOG_PAGES, NULL); | |
e289f03e FM |
1033 | extent_io_tree_init(fs_info, &root->log_csum_range, |
1034 | IO_TREE_LOG_CSUM_RANGE, NULL); | |
1035 | } | |
017e5369 | 1036 | |
5f3ab90a | 1037 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 1038 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
1039 | #ifdef CONFIG_BTRFS_DEBUG |
1040 | INIT_LIST_HEAD(&root->leak_list); | |
fc7cbcd4 | 1041 | spin_lock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 1042 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); |
fc7cbcd4 | 1043 | spin_unlock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 1044 | #endif |
3768f368 CM |
1045 | } |
1046 | ||
74e4d827 | 1047 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 1048 | u64 objectid, gfp_t flags) |
6f07e42e | 1049 | { |
74e4d827 | 1050 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 1051 | if (root) |
96dfcb46 | 1052 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
1053 | return root; |
1054 | } | |
1055 | ||
06ea65a3 JB |
1056 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1057 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1058 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1059 | { |
1060 | struct btrfs_root *root; | |
1061 | ||
7c0260ee JM |
1062 | if (!fs_info) |
1063 | return ERR_PTR(-EINVAL); | |
1064 | ||
96dfcb46 | 1065 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
1066 | if (!root) |
1067 | return ERR_PTR(-ENOMEM); | |
da17066c | 1068 | |
b9ef22de | 1069 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 1070 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1071 | |
1072 | return root; | |
1073 | } | |
1074 | #endif | |
1075 | ||
abed4aaa JB |
1076 | static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) |
1077 | { | |
1078 | const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); | |
1079 | const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); | |
1080 | ||
1081 | return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); | |
1082 | } | |
1083 | ||
1084 | static int global_root_key_cmp(const void *k, const struct rb_node *node) | |
1085 | { | |
1086 | const struct btrfs_key *key = k; | |
1087 | const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); | |
1088 | ||
1089 | return btrfs_comp_cpu_keys(key, &root->root_key); | |
1090 | } | |
1091 | ||
1092 | int btrfs_global_root_insert(struct btrfs_root *root) | |
1093 | { | |
1094 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1095 | struct rb_node *tmp; | |
1096 | ||
1097 | write_lock(&fs_info->global_root_lock); | |
1098 | tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); | |
1099 | write_unlock(&fs_info->global_root_lock); | |
1100 | ASSERT(!tmp); | |
1101 | ||
1102 | return tmp ? -EEXIST : 0; | |
1103 | } | |
1104 | ||
1105 | void btrfs_global_root_delete(struct btrfs_root *root) | |
1106 | { | |
1107 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1108 | ||
1109 | write_lock(&fs_info->global_root_lock); | |
1110 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1111 | write_unlock(&fs_info->global_root_lock); | |
1112 | } | |
1113 | ||
1114 | struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, | |
1115 | struct btrfs_key *key) | |
1116 | { | |
1117 | struct rb_node *node; | |
1118 | struct btrfs_root *root = NULL; | |
1119 | ||
1120 | read_lock(&fs_info->global_root_lock); | |
1121 | node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); | |
1122 | if (node) | |
1123 | root = container_of(node, struct btrfs_root, rb_node); | |
1124 | read_unlock(&fs_info->global_root_lock); | |
1125 | ||
1126 | return root; | |
1127 | } | |
1128 | ||
f7238e50 JB |
1129 | static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) |
1130 | { | |
1131 | struct btrfs_block_group *block_group; | |
1132 | u64 ret; | |
1133 | ||
1134 | if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
1135 | return 0; | |
1136 | ||
1137 | if (bytenr) | |
1138 | block_group = btrfs_lookup_block_group(fs_info, bytenr); | |
1139 | else | |
1140 | block_group = btrfs_lookup_first_block_group(fs_info, bytenr); | |
1141 | ASSERT(block_group); | |
1142 | if (!block_group) | |
1143 | return 0; | |
1144 | ret = block_group->global_root_id; | |
1145 | btrfs_put_block_group(block_group); | |
1146 | ||
1147 | return ret; | |
1148 | } | |
1149 | ||
abed4aaa JB |
1150 | struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) |
1151 | { | |
1152 | struct btrfs_key key = { | |
1153 | .objectid = BTRFS_CSUM_TREE_OBJECTID, | |
1154 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 1155 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
1156 | }; |
1157 | ||
1158 | return btrfs_global_root(fs_info, &key); | |
1159 | } | |
1160 | ||
1161 | struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) | |
1162 | { | |
1163 | struct btrfs_key key = { | |
1164 | .objectid = BTRFS_EXTENT_TREE_OBJECTID, | |
1165 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 1166 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
1167 | }; |
1168 | ||
1169 | return btrfs_global_root(fs_info, &key); | |
1170 | } | |
1171 | ||
20897f5c | 1172 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
1173 | u64 objectid) |
1174 | { | |
9b7a2440 | 1175 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
1176 | struct extent_buffer *leaf; |
1177 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1178 | struct btrfs_root *root; | |
1179 | struct btrfs_key key; | |
b89f6d1f | 1180 | unsigned int nofs_flag; |
20897f5c | 1181 | int ret = 0; |
20897f5c | 1182 | |
b89f6d1f FM |
1183 | /* |
1184 | * We're holding a transaction handle, so use a NOFS memory allocation | |
1185 | * context to avoid deadlock if reclaim happens. | |
1186 | */ | |
1187 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 1188 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 1189 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
1190 | if (!root) |
1191 | return ERR_PTR(-ENOMEM); | |
1192 | ||
20897f5c AJ |
1193 | root->root_key.objectid = objectid; |
1194 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1195 | root->root_key.offset = 0; | |
1196 | ||
9631e4cc JB |
1197 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
1198 | BTRFS_NESTING_NORMAL); | |
20897f5c AJ |
1199 | if (IS_ERR(leaf)) { |
1200 | ret = PTR_ERR(leaf); | |
1dd05682 | 1201 | leaf = NULL; |
8a6a87cd | 1202 | goto fail_unlock; |
20897f5c AJ |
1203 | } |
1204 | ||
20897f5c | 1205 | root->node = leaf; |
20897f5c AJ |
1206 | btrfs_mark_buffer_dirty(leaf); |
1207 | ||
1208 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1209 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 1210 | |
f944d2cb DS |
1211 | btrfs_set_root_flags(&root->root_item, 0); |
1212 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
1213 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
1214 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1215 | btrfs_set_root_level(&root->root_item, 0); | |
1216 | btrfs_set_root_refs(&root->root_item, 1); | |
1217 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1218 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1219 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 1220 | if (is_fstree(objectid)) |
807fc790 AS |
1221 | generate_random_guid(root->root_item.uuid); |
1222 | else | |
1223 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 1224 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 1225 | |
8a6a87cd BB |
1226 | btrfs_tree_unlock(leaf); |
1227 | ||
20897f5c AJ |
1228 | key.objectid = objectid; |
1229 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1230 | key.offset = 0; | |
1231 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1232 | if (ret) | |
1233 | goto fail; | |
1234 | ||
1dd05682 TI |
1235 | return root; |
1236 | ||
8a6a87cd | 1237 | fail_unlock: |
8c38938c | 1238 | if (leaf) |
1dd05682 | 1239 | btrfs_tree_unlock(leaf); |
8a6a87cd | 1240 | fail: |
00246528 | 1241 | btrfs_put_root(root); |
20897f5c | 1242 | |
1dd05682 | 1243 | return ERR_PTR(ret); |
20897f5c AJ |
1244 | } |
1245 | ||
7237f183 YZ |
1246 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1247 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1248 | { |
1249 | struct btrfs_root *root; | |
e02119d5 | 1250 | |
96dfcb46 | 1251 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 1252 | if (!root) |
7237f183 | 1253 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1254 | |
e02119d5 CM |
1255 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
1256 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1257 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1258 | |
6ab6ebb7 NA |
1259 | return root; |
1260 | } | |
1261 | ||
1262 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
1263 | struct btrfs_root *root) | |
1264 | { | |
1265 | struct extent_buffer *leaf; | |
1266 | ||
7237f183 | 1267 | /* |
92a7cc42 | 1268 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 1269 | * |
92a7cc42 QW |
1270 | * Log trees are not exposed to user space thus can't be snapshotted, |
1271 | * and they go away before a real commit is actually done. | |
1272 | * | |
1273 | * They do store pointers to file data extents, and those reference | |
1274 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 1275 | */ |
e02119d5 | 1276 | |
4d75f8a9 | 1277 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
9631e4cc | 1278 | NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
1279 | if (IS_ERR(leaf)) |
1280 | return PTR_ERR(leaf); | |
e02119d5 | 1281 | |
7237f183 | 1282 | root->node = leaf; |
e02119d5 | 1283 | |
e02119d5 CM |
1284 | btrfs_mark_buffer_dirty(root->node); |
1285 | btrfs_tree_unlock(root->node); | |
6ab6ebb7 NA |
1286 | |
1287 | return 0; | |
7237f183 YZ |
1288 | } |
1289 | ||
1290 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1291 | struct btrfs_fs_info *fs_info) | |
1292 | { | |
1293 | struct btrfs_root *log_root; | |
1294 | ||
1295 | log_root = alloc_log_tree(trans, fs_info); | |
1296 | if (IS_ERR(log_root)) | |
1297 | return PTR_ERR(log_root); | |
6ab6ebb7 | 1298 | |
3ddebf27 NA |
1299 | if (!btrfs_is_zoned(fs_info)) { |
1300 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
1301 | ||
1302 | if (ret) { | |
1303 | btrfs_put_root(log_root); | |
1304 | return ret; | |
1305 | } | |
6ab6ebb7 NA |
1306 | } |
1307 | ||
7237f183 YZ |
1308 | WARN_ON(fs_info->log_root_tree); |
1309 | fs_info->log_root_tree = log_root; | |
1310 | return 0; | |
1311 | } | |
1312 | ||
1313 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1314 | struct btrfs_root *root) | |
1315 | { | |
0b246afa | 1316 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1317 | struct btrfs_root *log_root; |
1318 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 1319 | int ret; |
7237f183 | 1320 | |
0b246afa | 1321 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1322 | if (IS_ERR(log_root)) |
1323 | return PTR_ERR(log_root); | |
1324 | ||
6ab6ebb7 NA |
1325 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
1326 | if (ret) { | |
1327 | btrfs_put_root(log_root); | |
1328 | return ret; | |
1329 | } | |
1330 | ||
7237f183 YZ |
1331 | log_root->last_trans = trans->transid; |
1332 | log_root->root_key.offset = root->root_key.objectid; | |
1333 | ||
1334 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1335 | btrfs_set_stack_inode_generation(inode_item, 1); |
1336 | btrfs_set_stack_inode_size(inode_item, 3); | |
1337 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1338 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1339 | fs_info->nodesize); |
3cae210f | 1340 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1341 | |
5d4f98a2 | 1342 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1343 | |
1344 | WARN_ON(root->log_root); | |
1345 | root->log_root = log_root; | |
1346 | root->log_transid = 0; | |
d1433deb | 1347 | root->log_transid_committed = -1; |
257c62e1 | 1348 | root->last_log_commit = 0; |
e02119d5 CM |
1349 | return 0; |
1350 | } | |
1351 | ||
49d11bea JB |
1352 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1353 | struct btrfs_path *path, | |
1354 | struct btrfs_key *key) | |
e02119d5 CM |
1355 | { |
1356 | struct btrfs_root *root; | |
1357 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
84234f3a | 1358 | u64 generation; |
cb517eab | 1359 | int ret; |
581c1760 | 1360 | int level; |
0f7d52f4 | 1361 | |
96dfcb46 | 1362 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1363 | if (!root) |
1364 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1365 | |
cb517eab MX |
1366 | ret = btrfs_find_root(tree_root, key, path, |
1367 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1368 | if (ret) { |
13a8a7c8 YZ |
1369 | if (ret > 0) |
1370 | ret = -ENOENT; | |
49d11bea | 1371 | goto fail; |
0f7d52f4 | 1372 | } |
13a8a7c8 | 1373 | |
84234f3a | 1374 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1375 | level = btrfs_root_level(&root->root_item); |
2ff7e61e JM |
1376 | root->node = read_tree_block(fs_info, |
1377 | btrfs_root_bytenr(&root->root_item), | |
1b7ec85e | 1378 | key->objectid, generation, level, NULL); |
64c043de LB |
1379 | if (IS_ERR(root->node)) { |
1380 | ret = PTR_ERR(root->node); | |
8c38938c | 1381 | root->node = NULL; |
49d11bea | 1382 | goto fail; |
4eb150d6 QW |
1383 | } |
1384 | if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
cb517eab | 1385 | ret = -EIO; |
49d11bea | 1386 | goto fail; |
416bc658 | 1387 | } |
88c602ab QW |
1388 | |
1389 | /* | |
1390 | * For real fs, and not log/reloc trees, root owner must | |
1391 | * match its root node owner | |
1392 | */ | |
1393 | if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && | |
1394 | root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && | |
1395 | root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && | |
1396 | root->root_key.objectid != btrfs_header_owner(root->node)) { | |
1397 | btrfs_crit(fs_info, | |
1398 | "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", | |
1399 | root->root_key.objectid, root->node->start, | |
1400 | btrfs_header_owner(root->node), | |
1401 | root->root_key.objectid); | |
1402 | ret = -EUCLEAN; | |
1403 | goto fail; | |
1404 | } | |
5d4f98a2 | 1405 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1406 | return root; |
49d11bea | 1407 | fail: |
00246528 | 1408 | btrfs_put_root(root); |
49d11bea JB |
1409 | return ERR_PTR(ret); |
1410 | } | |
1411 | ||
1412 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1413 | struct btrfs_key *key) | |
1414 | { | |
1415 | struct btrfs_root *root; | |
1416 | struct btrfs_path *path; | |
1417 | ||
1418 | path = btrfs_alloc_path(); | |
1419 | if (!path) | |
1420 | return ERR_PTR(-ENOMEM); | |
1421 | root = read_tree_root_path(tree_root, path, key); | |
1422 | btrfs_free_path(path); | |
1423 | ||
1424 | return root; | |
cb517eab MX |
1425 | } |
1426 | ||
2dfb1e43 QW |
1427 | /* |
1428 | * Initialize subvolume root in-memory structure | |
1429 | * | |
1430 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1431 | */ | |
1432 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1433 | { |
1434 | int ret; | |
dcc3eb96 | 1435 | unsigned int nofs_flag; |
cb517eab | 1436 | |
dcc3eb96 NB |
1437 | /* |
1438 | * We might be called under a transaction (e.g. indirect backref | |
1439 | * resolution) which could deadlock if it triggers memory reclaim | |
1440 | */ | |
1441 | nofs_flag = memalloc_nofs_save(); | |
1442 | ret = btrfs_drew_lock_init(&root->snapshot_lock); | |
1443 | memalloc_nofs_restore(nofs_flag); | |
1444 | if (ret) | |
8257b2dc | 1445 | goto fail; |
8257b2dc | 1446 | |
aeb935a4 | 1447 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
37f00a6d | 1448 | !btrfs_is_data_reloc_root(root)) { |
92a7cc42 | 1449 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1450 | btrfs_check_and_init_root_item(&root->root_item); |
1451 | } | |
1452 | ||
851fd730 QW |
1453 | /* |
1454 | * Don't assign anonymous block device to roots that are not exposed to | |
1455 | * userspace, the id pool is limited to 1M | |
1456 | */ | |
1457 | if (is_fstree(root->root_key.objectid) && | |
1458 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1459 | if (!anon_dev) { |
1460 | ret = get_anon_bdev(&root->anon_dev); | |
1461 | if (ret) | |
1462 | goto fail; | |
1463 | } else { | |
1464 | root->anon_dev = anon_dev; | |
1465 | } | |
851fd730 | 1466 | } |
f32e48e9 CR |
1467 | |
1468 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1469 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1470 | if (ret) { |
1471 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1472 | goto fail; |
f32e48e9 CR |
1473 | } |
1474 | ||
6b8fad57 | 1475 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1476 | |
1477 | mutex_unlock(&root->objectid_mutex); | |
1478 | ||
cb517eab MX |
1479 | return 0; |
1480 | fail: | |
84db5ccf | 1481 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1482 | return ret; |
1483 | } | |
1484 | ||
a98db0f3 JB |
1485 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1486 | u64 root_id) | |
cb517eab MX |
1487 | { |
1488 | struct btrfs_root *root; | |
1489 | ||
fc7cbcd4 DS |
1490 | spin_lock(&fs_info->fs_roots_radix_lock); |
1491 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1492 | (unsigned long)root_id); | |
bc44d7c4 | 1493 | if (root) |
00246528 | 1494 | root = btrfs_grab_root(root); |
fc7cbcd4 | 1495 | spin_unlock(&fs_info->fs_roots_radix_lock); |
cb517eab MX |
1496 | return root; |
1497 | } | |
1498 | ||
49d11bea JB |
1499 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1500 | u64 objectid) | |
1501 | { | |
abed4aaa JB |
1502 | struct btrfs_key key = { |
1503 | .objectid = objectid, | |
1504 | .type = BTRFS_ROOT_ITEM_KEY, | |
1505 | .offset = 0, | |
1506 | }; | |
1507 | ||
49d11bea JB |
1508 | if (objectid == BTRFS_ROOT_TREE_OBJECTID) |
1509 | return btrfs_grab_root(fs_info->tree_root); | |
1510 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
abed4aaa | 1511 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1512 | if (objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1513 | return btrfs_grab_root(fs_info->chunk_root); | |
1514 | if (objectid == BTRFS_DEV_TREE_OBJECTID) | |
1515 | return btrfs_grab_root(fs_info->dev_root); | |
1516 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
abed4aaa | 1517 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1518 | if (objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1519 | return btrfs_grab_root(fs_info->quota_root) ? | |
1520 | fs_info->quota_root : ERR_PTR(-ENOENT); | |
1521 | if (objectid == BTRFS_UUID_TREE_OBJECTID) | |
1522 | return btrfs_grab_root(fs_info->uuid_root) ? | |
1523 | fs_info->uuid_root : ERR_PTR(-ENOENT); | |
14033b08 QW |
1524 | if (objectid == BTRFS_BLOCK_GROUP_TREE_OBJECTID) |
1525 | return btrfs_grab_root(fs_info->block_group_root) ? | |
1526 | fs_info->block_group_root : ERR_PTR(-ENOENT); | |
abed4aaa JB |
1527 | if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { |
1528 | struct btrfs_root *root = btrfs_global_root(fs_info, &key); | |
1529 | ||
1530 | return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); | |
1531 | } | |
49d11bea JB |
1532 | return NULL; |
1533 | } | |
1534 | ||
cb517eab MX |
1535 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1536 | struct btrfs_root *root) | |
1537 | { | |
1538 | int ret; | |
1539 | ||
fc7cbcd4 DS |
1540 | ret = radix_tree_preload(GFP_NOFS); |
1541 | if (ret) | |
1542 | return ret; | |
1543 | ||
1544 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1545 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1546 | (unsigned long)root->root_key.objectid, | |
1547 | root); | |
af01d2e5 | 1548 | if (ret == 0) { |
00246528 | 1549 | btrfs_grab_root(root); |
fc7cbcd4 | 1550 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1551 | } |
fc7cbcd4 DS |
1552 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1553 | radix_tree_preload_end(); | |
cb517eab MX |
1554 | |
1555 | return ret; | |
1556 | } | |
1557 | ||
bd647ce3 JB |
1558 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1559 | { | |
1560 | #ifdef CONFIG_BTRFS_DEBUG | |
1561 | struct btrfs_root *root; | |
1562 | ||
1563 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1564 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1565 | ||
bd647ce3 JB |
1566 | root = list_first_entry(&fs_info->allocated_roots, |
1567 | struct btrfs_root, leak_list); | |
457f1864 | 1568 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1569 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 JB |
1570 | refcount_read(&root->refs)); |
1571 | while (refcount_read(&root->refs) > 1) | |
00246528 JB |
1572 | btrfs_put_root(root); |
1573 | btrfs_put_root(root); | |
bd647ce3 JB |
1574 | } |
1575 | #endif | |
1576 | } | |
1577 | ||
abed4aaa JB |
1578 | static void free_global_roots(struct btrfs_fs_info *fs_info) |
1579 | { | |
1580 | struct btrfs_root *root; | |
1581 | struct rb_node *node; | |
1582 | ||
1583 | while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { | |
1584 | root = rb_entry(node, struct btrfs_root, rb_node); | |
1585 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1586 | btrfs_put_root(root); | |
1587 | } | |
1588 | } | |
1589 | ||
0d4b0463 JB |
1590 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1591 | { | |
141386e1 JB |
1592 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1593 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1594 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1595 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1596 | btrfs_free_csum_hash(fs_info); | |
1597 | btrfs_free_stripe_hash_table(fs_info); | |
1598 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1599 | kfree(fs_info->balance_ctl); |
1600 | kfree(fs_info->delayed_root); | |
abed4aaa | 1601 | free_global_roots(fs_info); |
00246528 JB |
1602 | btrfs_put_root(fs_info->tree_root); |
1603 | btrfs_put_root(fs_info->chunk_root); | |
1604 | btrfs_put_root(fs_info->dev_root); | |
00246528 JB |
1605 | btrfs_put_root(fs_info->quota_root); |
1606 | btrfs_put_root(fs_info->uuid_root); | |
00246528 | 1607 | btrfs_put_root(fs_info->fs_root); |
aeb935a4 | 1608 | btrfs_put_root(fs_info->data_reloc_root); |
9c54e80d | 1609 | btrfs_put_root(fs_info->block_group_root); |
bd647ce3 | 1610 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1611 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1612 | kfree(fs_info->super_copy); |
1613 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1614 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1615 | kvfree(fs_info); |
1616 | } | |
1617 | ||
1618 | ||
2dfb1e43 QW |
1619 | /* |
1620 | * Get an in-memory reference of a root structure. | |
1621 | * | |
1622 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1623 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1624 | * found, then read it from disk and add it to cached fs roots. | |
1625 | * | |
1626 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1627 | * | |
1628 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1629 | * same root objectid. | |
1630 | * | |
1631 | * @objectid: root id | |
1632 | * @anon_dev: preallocated anonymous block device number for new roots, | |
1633 | * pass 0 for new allocation. | |
1634 | * @check_ref: whether to check root item references, If true, return -ENOENT | |
1635 | * for orphan roots | |
1636 | */ | |
1637 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
1638 | u64 objectid, dev_t anon_dev, | |
1639 | bool check_ref) | |
5eda7b5e CM |
1640 | { |
1641 | struct btrfs_root *root; | |
381cf658 | 1642 | struct btrfs_path *path; |
1d4c08e0 | 1643 | struct btrfs_key key; |
5eda7b5e CM |
1644 | int ret; |
1645 | ||
49d11bea JB |
1646 | root = btrfs_get_global_root(fs_info, objectid); |
1647 | if (root) | |
1648 | return root; | |
4df27c4d | 1649 | again: |
56e9357a | 1650 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1651 | if (root) { |
2dfb1e43 QW |
1652 | /* Shouldn't get preallocated anon_dev for cached roots */ |
1653 | ASSERT(!anon_dev); | |
bc44d7c4 | 1654 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1655 | btrfs_put_root(root); |
48475471 | 1656 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1657 | } |
5eda7b5e | 1658 | return root; |
48475471 | 1659 | } |
5eda7b5e | 1660 | |
56e9357a DS |
1661 | key.objectid = objectid; |
1662 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1663 | key.offset = (u64)-1; | |
1664 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1665 | if (IS_ERR(root)) |
1666 | return root; | |
3394e160 | 1667 | |
c00869f1 | 1668 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1669 | ret = -ENOENT; |
581bb050 | 1670 | goto fail; |
35a30d7c | 1671 | } |
581bb050 | 1672 | |
2dfb1e43 | 1673 | ret = btrfs_init_fs_root(root, anon_dev); |
ac08aedf CM |
1674 | if (ret) |
1675 | goto fail; | |
3394e160 | 1676 | |
381cf658 DS |
1677 | path = btrfs_alloc_path(); |
1678 | if (!path) { | |
1679 | ret = -ENOMEM; | |
1680 | goto fail; | |
1681 | } | |
1d4c08e0 DS |
1682 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1683 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1684 | key.offset = objectid; |
1d4c08e0 DS |
1685 | |
1686 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1687 | btrfs_free_path(path); |
d68fc57b YZ |
1688 | if (ret < 0) |
1689 | goto fail; | |
1690 | if (ret == 0) | |
27cdeb70 | 1691 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1692 | |
cb517eab | 1693 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1694 | if (ret) { |
168a2f77 JJB |
1695 | if (ret == -EEXIST) { |
1696 | btrfs_put_root(root); | |
4df27c4d | 1697 | goto again; |
168a2f77 | 1698 | } |
4df27c4d | 1699 | goto fail; |
0f7d52f4 | 1700 | } |
edbd8d4e | 1701 | return root; |
4df27c4d | 1702 | fail: |
33fab972 FM |
1703 | /* |
1704 | * If our caller provided us an anonymous device, then it's his | |
143823cf | 1705 | * responsibility to free it in case we fail. So we have to set our |
33fab972 FM |
1706 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() |
1707 | * and once again by our caller. | |
1708 | */ | |
1709 | if (anon_dev) | |
1710 | root->anon_dev = 0; | |
8c38938c | 1711 | btrfs_put_root(root); |
4df27c4d | 1712 | return ERR_PTR(ret); |
edbd8d4e CM |
1713 | } |
1714 | ||
2dfb1e43 QW |
1715 | /* |
1716 | * Get in-memory reference of a root structure | |
1717 | * | |
1718 | * @objectid: tree objectid | |
1719 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1720 | * one reference | |
1721 | */ | |
1722 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1723 | u64 objectid, bool check_ref) | |
1724 | { | |
1725 | return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); | |
1726 | } | |
1727 | ||
1728 | /* | |
1729 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1730 | * the anonymous block device id | |
1731 | * | |
1732 | * @objectid: tree objectid | |
1733 | * @anon_dev: if zero, allocate a new anonymous block device or use the | |
1734 | * parameter value | |
1735 | */ | |
1736 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
1737 | u64 objectid, dev_t anon_dev) | |
1738 | { | |
1739 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1740 | } | |
1741 | ||
49d11bea JB |
1742 | /* |
1743 | * btrfs_get_fs_root_commit_root - return a root for the given objectid | |
1744 | * @fs_info: the fs_info | |
1745 | * @objectid: the objectid we need to lookup | |
1746 | * | |
1747 | * This is exclusively used for backref walking, and exists specifically because | |
1748 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1749 | * creation time, which means we may have to read the tree_root in order to look | |
1750 | * up a fs root that is not in memory. If the root is not in memory we will | |
1751 | * read the tree root commit root and look up the fs root from there. This is a | |
1752 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1753 | * have the most uptodate information, it'll simply be discarded once the | |
1754 | * backref code is finished using the root. | |
1755 | */ | |
1756 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1757 | struct btrfs_path *path, | |
1758 | u64 objectid) | |
1759 | { | |
1760 | struct btrfs_root *root; | |
1761 | struct btrfs_key key; | |
1762 | ||
1763 | ASSERT(path->search_commit_root && path->skip_locking); | |
1764 | ||
1765 | /* | |
1766 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1767 | * since this is called via the backref walking code we won't be looking | |
1768 | * up a root that doesn't exist, unless there's corruption. So if root | |
1769 | * != NULL just return it. | |
1770 | */ | |
1771 | root = btrfs_get_global_root(fs_info, objectid); | |
1772 | if (root) | |
1773 | return root; | |
1774 | ||
1775 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1776 | if (root) | |
1777 | return root; | |
1778 | ||
1779 | key.objectid = objectid; | |
1780 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1781 | key.offset = (u64)-1; | |
1782 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1783 | btrfs_release_path(path); | |
1784 | ||
1785 | return root; | |
1786 | } | |
1787 | ||
a74a4b97 CM |
1788 | static int cleaner_kthread(void *arg) |
1789 | { | |
0d031dc4 | 1790 | struct btrfs_fs_info *fs_info = arg; |
d0278245 | 1791 | int again; |
a74a4b97 | 1792 | |
d6fd0ae2 | 1793 | while (1) { |
d0278245 | 1794 | again = 0; |
a74a4b97 | 1795 | |
fd340d0f JB |
1796 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1797 | ||
d0278245 | 1798 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1799 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1800 | goto sleep; |
1801 | ||
90c711ab ZB |
1802 | /* |
1803 | * Do not do anything if we might cause open_ctree() to block | |
1804 | * before we have finished mounting the filesystem. | |
1805 | */ | |
0b246afa | 1806 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1807 | goto sleep; |
1808 | ||
0b246afa | 1809 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1810 | goto sleep; |
1811 | ||
dc7f370c MX |
1812 | /* |
1813 | * Avoid the problem that we change the status of the fs | |
1814 | * during the above check and trylock. | |
1815 | */ | |
2ff7e61e | 1816 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1817 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1818 | goto sleep; |
76dda93c | 1819 | } |
a74a4b97 | 1820 | |
2ff7e61e | 1821 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1822 | |
33c44184 | 1823 | again = btrfs_clean_one_deleted_snapshot(fs_info); |
0b246afa | 1824 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1825 | |
1826 | /* | |
05323cd1 MX |
1827 | * The defragger has dealt with the R/O remount and umount, |
1828 | * needn't do anything special here. | |
d0278245 | 1829 | */ |
0b246afa | 1830 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1831 | |
1832 | /* | |
f3372065 | 1833 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
1834 | * with relocation (btrfs_relocate_chunk) and relocation |
1835 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 1836 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
1837 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
1838 | * unused block groups. | |
1839 | */ | |
0b246afa | 1840 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
1841 | |
1842 | /* | |
1843 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
1844 | * all unused block_groups. This possibly gives us some more free | |
1845 | * space. | |
1846 | */ | |
1847 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 1848 | sleep: |
a0a1db70 | 1849 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1850 | if (kthread_should_park()) |
1851 | kthread_parkme(); | |
1852 | if (kthread_should_stop()) | |
1853 | return 0; | |
838fe188 | 1854 | if (!again) { |
a74a4b97 | 1855 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1856 | schedule(); |
a74a4b97 CM |
1857 | __set_current_state(TASK_RUNNING); |
1858 | } | |
da288d28 | 1859 | } |
a74a4b97 CM |
1860 | } |
1861 | ||
1862 | static int transaction_kthread(void *arg) | |
1863 | { | |
1864 | struct btrfs_root *root = arg; | |
0b246afa | 1865 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1866 | struct btrfs_trans_handle *trans; |
1867 | struct btrfs_transaction *cur; | |
8929ecfa | 1868 | u64 transid; |
643900be | 1869 | time64_t delta; |
a74a4b97 | 1870 | unsigned long delay; |
914b2007 | 1871 | bool cannot_commit; |
a74a4b97 CM |
1872 | |
1873 | do { | |
914b2007 | 1874 | cannot_commit = false; |
ba1bc00f | 1875 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 1876 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 1877 | |
0b246afa JM |
1878 | spin_lock(&fs_info->trans_lock); |
1879 | cur = fs_info->running_transaction; | |
a74a4b97 | 1880 | if (!cur) { |
0b246afa | 1881 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1882 | goto sleep; |
1883 | } | |
31153d81 | 1884 | |
643900be | 1885 | delta = ktime_get_seconds() - cur->start_time; |
fdfbf020 JB |
1886 | if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && |
1887 | cur->state < TRANS_STATE_COMMIT_START && | |
643900be | 1888 | delta < fs_info->commit_interval) { |
0b246afa | 1889 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
1890 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
1891 | delay = min(delay, | |
1892 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
1893 | goto sleep; |
1894 | } | |
8929ecfa | 1895 | transid = cur->transid; |
0b246afa | 1896 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1897 | |
79787eaa | 1898 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1899 | trans = btrfs_attach_transaction(root); |
914b2007 | 1900 | if (IS_ERR(trans)) { |
354aa0fb MX |
1901 | if (PTR_ERR(trans) != -ENOENT) |
1902 | cannot_commit = true; | |
79787eaa | 1903 | goto sleep; |
914b2007 | 1904 | } |
8929ecfa | 1905 | if (transid == trans->transid) { |
3a45bb20 | 1906 | btrfs_commit_transaction(trans); |
8929ecfa | 1907 | } else { |
3a45bb20 | 1908 | btrfs_end_transaction(trans); |
8929ecfa | 1909 | } |
a74a4b97 | 1910 | sleep: |
0b246afa JM |
1911 | wake_up_process(fs_info->cleaner_kthread); |
1912 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1913 | |
84961539 | 1914 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 1915 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1916 | if (!kthread_should_stop() && |
0b246afa | 1917 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1918 | cannot_commit)) |
bc5511d0 | 1919 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1920 | } while (!kthread_should_stop()); |
1921 | return 0; | |
1922 | } | |
1923 | ||
af31f5e5 | 1924 | /* |
01f0f9da NB |
1925 | * This will find the highest generation in the array of root backups. The |
1926 | * index of the highest array is returned, or -EINVAL if we can't find | |
1927 | * anything. | |
af31f5e5 CM |
1928 | * |
1929 | * We check to make sure the array is valid by comparing the | |
1930 | * generation of the latest root in the array with the generation | |
1931 | * in the super block. If they don't match we pitch it. | |
1932 | */ | |
01f0f9da | 1933 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 1934 | { |
01f0f9da | 1935 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 1936 | u64 cur; |
af31f5e5 CM |
1937 | struct btrfs_root_backup *root_backup; |
1938 | int i; | |
1939 | ||
1940 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1941 | root_backup = info->super_copy->super_roots + i; | |
1942 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1943 | if (cur == newest_gen) | |
01f0f9da | 1944 | return i; |
af31f5e5 CM |
1945 | } |
1946 | ||
01f0f9da | 1947 | return -EINVAL; |
af31f5e5 CM |
1948 | } |
1949 | ||
af31f5e5 CM |
1950 | /* |
1951 | * copy all the root pointers into the super backup array. | |
1952 | * this will bump the backup pointer by one when it is | |
1953 | * done | |
1954 | */ | |
1955 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1956 | { | |
6ef108dd | 1957 | const int next_backup = info->backup_root_index; |
af31f5e5 | 1958 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
1959 | |
1960 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1961 | ||
1962 | /* | |
1963 | * make sure all of our padding and empty slots get zero filled | |
1964 | * regardless of which ones we use today | |
1965 | */ | |
1966 | memset(root_backup, 0, sizeof(*root_backup)); | |
1967 | ||
1968 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1969 | ||
1970 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1971 | btrfs_set_backup_tree_root_gen(root_backup, | |
1972 | btrfs_header_generation(info->tree_root->node)); | |
1973 | ||
1974 | btrfs_set_backup_tree_root_level(root_backup, | |
1975 | btrfs_header_level(info->tree_root->node)); | |
1976 | ||
1977 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1978 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1979 | btrfs_header_generation(info->chunk_root->node)); | |
1980 | btrfs_set_backup_chunk_root_level(root_backup, | |
1981 | btrfs_header_level(info->chunk_root->node)); | |
1982 | ||
1c56ab99 | 1983 | if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { |
9c54e80d | 1984 | struct btrfs_root *extent_root = btrfs_extent_root(info, 0); |
f7238e50 | 1985 | struct btrfs_root *csum_root = btrfs_csum_root(info, 0); |
9c54e80d JB |
1986 | |
1987 | btrfs_set_backup_extent_root(root_backup, | |
1988 | extent_root->node->start); | |
1989 | btrfs_set_backup_extent_root_gen(root_backup, | |
1990 | btrfs_header_generation(extent_root->node)); | |
1991 | btrfs_set_backup_extent_root_level(root_backup, | |
1992 | btrfs_header_level(extent_root->node)); | |
f7238e50 JB |
1993 | |
1994 | btrfs_set_backup_csum_root(root_backup, csum_root->node->start); | |
1995 | btrfs_set_backup_csum_root_gen(root_backup, | |
1996 | btrfs_header_generation(csum_root->node)); | |
1997 | btrfs_set_backup_csum_root_level(root_backup, | |
1998 | btrfs_header_level(csum_root->node)); | |
9c54e80d | 1999 | } |
af31f5e5 | 2000 | |
7c7e82a7 CM |
2001 | /* |
2002 | * we might commit during log recovery, which happens before we set | |
2003 | * the fs_root. Make sure it is valid before we fill it in. | |
2004 | */ | |
2005 | if (info->fs_root && info->fs_root->node) { | |
2006 | btrfs_set_backup_fs_root(root_backup, | |
2007 | info->fs_root->node->start); | |
2008 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2009 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2010 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2011 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2012 | } |
af31f5e5 CM |
2013 | |
2014 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2015 | btrfs_set_backup_dev_root_gen(root_backup, | |
2016 | btrfs_header_generation(info->dev_root->node)); | |
2017 | btrfs_set_backup_dev_root_level(root_backup, | |
2018 | btrfs_header_level(info->dev_root->node)); | |
2019 | ||
af31f5e5 CM |
2020 | btrfs_set_backup_total_bytes(root_backup, |
2021 | btrfs_super_total_bytes(info->super_copy)); | |
2022 | btrfs_set_backup_bytes_used(root_backup, | |
2023 | btrfs_super_bytes_used(info->super_copy)); | |
2024 | btrfs_set_backup_num_devices(root_backup, | |
2025 | btrfs_super_num_devices(info->super_copy)); | |
2026 | ||
2027 | /* | |
2028 | * if we don't copy this out to the super_copy, it won't get remembered | |
2029 | * for the next commit | |
2030 | */ | |
2031 | memcpy(&info->super_copy->super_roots, | |
2032 | &info->super_for_commit->super_roots, | |
2033 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2034 | } | |
2035 | ||
bd2336b2 NB |
2036 | /* |
2037 | * read_backup_root - Reads a backup root based on the passed priority. Prio 0 | |
2038 | * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
2039 | * | |
2040 | * fs_info - filesystem whose backup roots need to be read | |
2041 | * priority - priority of backup root required | |
2042 | * | |
2043 | * Returns backup root index on success and -EINVAL otherwise. | |
2044 | */ | |
2045 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
2046 | { | |
2047 | int backup_index = find_newest_super_backup(fs_info); | |
2048 | struct btrfs_super_block *super = fs_info->super_copy; | |
2049 | struct btrfs_root_backup *root_backup; | |
2050 | ||
2051 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
2052 | if (priority == 0) | |
2053 | return backup_index; | |
2054 | ||
2055 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
2056 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2057 | } else { | |
2058 | return -EINVAL; | |
2059 | } | |
2060 | ||
2061 | root_backup = super->super_roots + backup_index; | |
2062 | ||
2063 | btrfs_set_super_generation(super, | |
2064 | btrfs_backup_tree_root_gen(root_backup)); | |
2065 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2066 | btrfs_set_super_root_level(super, | |
2067 | btrfs_backup_tree_root_level(root_backup)); | |
2068 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2069 | ||
2070 | /* | |
2071 | * Fixme: the total bytes and num_devices need to match or we should | |
2072 | * need a fsck | |
2073 | */ | |
2074 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2075 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2076 | ||
2077 | return backup_index; | |
2078 | } | |
2079 | ||
7abadb64 LB |
2080 | /* helper to cleanup workers */ |
2081 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2082 | { | |
dc6e3209 | 2083 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2084 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
a31b4a43 | 2085 | btrfs_destroy_workqueue(fs_info->hipri_workers); |
5cdc7ad3 | 2086 | btrfs_destroy_workqueue(fs_info->workers); |
d7b9416f CH |
2087 | if (fs_info->endio_workers) |
2088 | destroy_workqueue(fs_info->endio_workers); | |
d34e123d CH |
2089 | if (fs_info->endio_raid56_workers) |
2090 | destroy_workqueue(fs_info->endio_raid56_workers); | |
385de0ef CH |
2091 | if (fs_info->rmw_workers) |
2092 | destroy_workqueue(fs_info->rmw_workers); | |
fed8a72d CH |
2093 | if (fs_info->compressed_write_workers) |
2094 | destroy_workqueue(fs_info->compressed_write_workers); | |
fccb5d86 QW |
2095 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2096 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 2097 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2098 | btrfs_destroy_workqueue(fs_info->caching_workers); |
a44903ab | 2099 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2100 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
2101 | if (fs_info->discard_ctl.discard_workers) |
2102 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
2103 | /* |
2104 | * Now that all other work queues are destroyed, we can safely destroy | |
2105 | * the queues used for metadata I/O, since tasks from those other work | |
2106 | * queues can do metadata I/O operations. | |
2107 | */ | |
d7b9416f CH |
2108 | if (fs_info->endio_meta_workers) |
2109 | destroy_workqueue(fs_info->endio_meta_workers); | |
7abadb64 LB |
2110 | } |
2111 | ||
2e9f5954 R |
2112 | static void free_root_extent_buffers(struct btrfs_root *root) |
2113 | { | |
2114 | if (root) { | |
2115 | free_extent_buffer(root->node); | |
2116 | free_extent_buffer(root->commit_root); | |
2117 | root->node = NULL; | |
2118 | root->commit_root = NULL; | |
2119 | } | |
2120 | } | |
2121 | ||
abed4aaa JB |
2122 | static void free_global_root_pointers(struct btrfs_fs_info *fs_info) |
2123 | { | |
2124 | struct btrfs_root *root, *tmp; | |
2125 | ||
2126 | rbtree_postorder_for_each_entry_safe(root, tmp, | |
2127 | &fs_info->global_root_tree, | |
2128 | rb_node) | |
2129 | free_root_extent_buffers(root); | |
2130 | } | |
2131 | ||
af31f5e5 | 2132 | /* helper to cleanup tree roots */ |
4273eaff | 2133 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 2134 | { |
2e9f5954 | 2135 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2136 | |
abed4aaa | 2137 | free_global_root_pointers(info); |
2e9f5954 | 2138 | free_root_extent_buffers(info->dev_root); |
2e9f5954 R |
2139 | free_root_extent_buffers(info->quota_root); |
2140 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 2141 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 2142 | free_root_extent_buffers(info->data_reloc_root); |
9c54e80d | 2143 | free_root_extent_buffers(info->block_group_root); |
4273eaff | 2144 | if (free_chunk_root) |
2e9f5954 | 2145 | free_root_extent_buffers(info->chunk_root); |
af31f5e5 CM |
2146 | } |
2147 | ||
8c38938c JB |
2148 | void btrfs_put_root(struct btrfs_root *root) |
2149 | { | |
2150 | if (!root) | |
2151 | return; | |
2152 | ||
2153 | if (refcount_dec_and_test(&root->refs)) { | |
2154 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 2155 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
2156 | if (root->anon_dev) |
2157 | free_anon_bdev(root->anon_dev); | |
2158 | btrfs_drew_lock_destroy(&root->snapshot_lock); | |
923eb523 | 2159 | free_root_extent_buffers(root); |
8c38938c | 2160 | #ifdef CONFIG_BTRFS_DEBUG |
fc7cbcd4 | 2161 | spin_lock(&root->fs_info->fs_roots_radix_lock); |
8c38938c | 2162 | list_del_init(&root->leak_list); |
fc7cbcd4 | 2163 | spin_unlock(&root->fs_info->fs_roots_radix_lock); |
8c38938c JB |
2164 | #endif |
2165 | kfree(root); | |
2166 | } | |
2167 | } | |
2168 | ||
faa2dbf0 | 2169 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 | 2170 | { |
fc7cbcd4 DS |
2171 | int ret; |
2172 | struct btrfs_root *gang[8]; | |
2173 | int i; | |
171f6537 JB |
2174 | |
2175 | while (!list_empty(&fs_info->dead_roots)) { | |
fc7cbcd4 DS |
2176 | gang[0] = list_entry(fs_info->dead_roots.next, |
2177 | struct btrfs_root, root_list); | |
2178 | list_del(&gang[0]->root_list); | |
171f6537 | 2179 | |
fc7cbcd4 DS |
2180 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
2181 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); | |
2182 | btrfs_put_root(gang[0]); | |
171f6537 JB |
2183 | } |
2184 | ||
fc7cbcd4 DS |
2185 | while (1) { |
2186 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2187 | (void **)gang, 0, | |
2188 | ARRAY_SIZE(gang)); | |
2189 | if (!ret) | |
2190 | break; | |
2191 | for (i = 0; i < ret; i++) | |
2192 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); | |
171f6537 JB |
2193 | } |
2194 | } | |
af31f5e5 | 2195 | |
638aa7ed ES |
2196 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2197 | { | |
2198 | mutex_init(&fs_info->scrub_lock); | |
2199 | atomic_set(&fs_info->scrubs_running, 0); | |
2200 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2201 | atomic_set(&fs_info->scrubs_paused, 0); | |
2202 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2203 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 2204 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
2205 | } |
2206 | ||
779a65a4 ES |
2207 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2208 | { | |
2209 | spin_lock_init(&fs_info->balance_lock); | |
2210 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2211 | atomic_set(&fs_info->balance_pause_req, 0); |
2212 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2213 | fs_info->balance_ctl = NULL; | |
2214 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 2215 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
2216 | } |
2217 | ||
6bccf3ab | 2218 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2219 | { |
2ff7e61e | 2220 | struct inode *inode = fs_info->btree_inode; |
e256927b JB |
2221 | unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, |
2222 | fs_info->tree_root); | |
2ff7e61e JM |
2223 | |
2224 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2225 | set_nlink(inode, 1); | |
f37938e0 ES |
2226 | /* |
2227 | * we set the i_size on the btree inode to the max possible int. | |
2228 | * the real end of the address space is determined by all of | |
2229 | * the devices in the system | |
2230 | */ | |
2ff7e61e JM |
2231 | inode->i_size = OFFSET_MAX; |
2232 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2233 | |
2ff7e61e | 2234 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 2235 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
efb0645b | 2236 | IO_TREE_BTREE_INODE_IO, NULL); |
2ff7e61e | 2237 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 2238 | |
5c8fd99f | 2239 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
adac5584 FM |
2240 | BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; |
2241 | BTRFS_I(inode)->location.type = 0; | |
2242 | BTRFS_I(inode)->location.offset = 0; | |
2ff7e61e | 2243 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); |
e256927b | 2244 | __insert_inode_hash(inode, hash); |
f37938e0 ES |
2245 | } |
2246 | ||
ad618368 ES |
2247 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2248 | { | |
ad618368 | 2249 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 2250 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 2251 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
2252 | } |
2253 | ||
f9e92e40 ES |
2254 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2255 | { | |
2256 | spin_lock_init(&fs_info->qgroup_lock); | |
2257 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2258 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
2259 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
2260 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2261 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2262 | fs_info->qgroup_rescan_running = false; |
011b46c3 | 2263 | fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; |
f9e92e40 ES |
2264 | mutex_init(&fs_info->qgroup_rescan_lock); |
2265 | } | |
2266 | ||
d21deec5 | 2267 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) |
2a458198 | 2268 | { |
f7b885be | 2269 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2270 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2271 | |
2272 | fs_info->workers = | |
a31b4a43 CH |
2273 | btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); |
2274 | fs_info->hipri_workers = | |
2275 | btrfs_alloc_workqueue(fs_info, "worker-high", | |
cb001095 | 2276 | flags | WQ_HIGHPRI, max_active, 16); |
2a458198 ES |
2277 | |
2278 | fs_info->delalloc_workers = | |
cb001095 JM |
2279 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2280 | flags, max_active, 2); | |
2a458198 ES |
2281 | |
2282 | fs_info->flush_workers = | |
cb001095 JM |
2283 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2284 | flags, max_active, 0); | |
2a458198 ES |
2285 | |
2286 | fs_info->caching_workers = | |
cb001095 | 2287 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 2288 | |
2a458198 | 2289 | fs_info->fixup_workers = |
cb001095 | 2290 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 | 2291 | |
2a458198 | 2292 | fs_info->endio_workers = |
d7b9416f | 2293 | alloc_workqueue("btrfs-endio", flags, max_active); |
2a458198 | 2294 | fs_info->endio_meta_workers = |
d7b9416f | 2295 | alloc_workqueue("btrfs-endio-meta", flags, max_active); |
2a458198 | 2296 | fs_info->endio_raid56_workers = |
d34e123d | 2297 | alloc_workqueue("btrfs-endio-raid56", flags, max_active); |
385de0ef | 2298 | fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); |
2a458198 | 2299 | fs_info->endio_write_workers = |
cb001095 JM |
2300 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2301 | max_active, 2); | |
fed8a72d CH |
2302 | fs_info->compressed_write_workers = |
2303 | alloc_workqueue("btrfs-compressed-write", flags, max_active); | |
2a458198 | 2304 | fs_info->endio_freespace_worker = |
cb001095 JM |
2305 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2306 | max_active, 0); | |
2a458198 | 2307 | fs_info->delayed_workers = |
cb001095 JM |
2308 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2309 | max_active, 0); | |
2a458198 | 2310 | fs_info->qgroup_rescan_workers = |
cb001095 | 2311 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
b0643e59 DZ |
2312 | fs_info->discard_ctl.discard_workers = |
2313 | alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); | |
2a458198 | 2314 | |
a31b4a43 CH |
2315 | if (!(fs_info->workers && fs_info->hipri_workers && |
2316 | fs_info->delalloc_workers && fs_info->flush_workers && | |
2a458198 | 2317 | fs_info->endio_workers && fs_info->endio_meta_workers && |
fed8a72d | 2318 | fs_info->compressed_write_workers && |
2a458198 ES |
2319 | fs_info->endio_write_workers && fs_info->endio_raid56_workers && |
2320 | fs_info->endio_freespace_worker && fs_info->rmw_workers && | |
f26c9238 QW |
2321 | fs_info->caching_workers && fs_info->fixup_workers && |
2322 | fs_info->delayed_workers && fs_info->qgroup_rescan_workers && | |
b0643e59 | 2323 | fs_info->discard_ctl.discard_workers)) { |
2a458198 ES |
2324 | return -ENOMEM; |
2325 | } | |
2326 | ||
2327 | return 0; | |
2328 | } | |
2329 | ||
6d97c6e3 JT |
2330 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2331 | { | |
2332 | struct crypto_shash *csum_shash; | |
b4e967be | 2333 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2334 | |
b4e967be | 2335 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2336 | |
2337 | if (IS_ERR(csum_shash)) { | |
2338 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2339 | csum_driver); |
6d97c6e3 JT |
2340 | return PTR_ERR(csum_shash); |
2341 | } | |
2342 | ||
2343 | fs_info->csum_shash = csum_shash; | |
2344 | ||
c8a5f8ca DS |
2345 | btrfs_info(fs_info, "using %s (%s) checksum algorithm", |
2346 | btrfs_super_csum_name(csum_type), | |
2347 | crypto_shash_driver_name(csum_shash)); | |
6d97c6e3 JT |
2348 | return 0; |
2349 | } | |
2350 | ||
63443bf5 ES |
2351 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2352 | struct btrfs_fs_devices *fs_devices) | |
2353 | { | |
2354 | int ret; | |
63443bf5 ES |
2355 | struct btrfs_root *log_tree_root; |
2356 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2357 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2358 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2359 | |
2360 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2361 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2362 | return -EIO; |
2363 | } | |
2364 | ||
96dfcb46 JB |
2365 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2366 | GFP_KERNEL); | |
63443bf5 ES |
2367 | if (!log_tree_root) |
2368 | return -ENOMEM; | |
2369 | ||
2ff7e61e | 2370 | log_tree_root->node = read_tree_block(fs_info, bytenr, |
1b7ec85e JB |
2371 | BTRFS_TREE_LOG_OBJECTID, |
2372 | fs_info->generation + 1, level, | |
2373 | NULL); | |
64c043de | 2374 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2375 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2376 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2377 | log_tree_root->node = NULL; |
00246528 | 2378 | btrfs_put_root(log_tree_root); |
0eeff236 | 2379 | return ret; |
4eb150d6 QW |
2380 | } |
2381 | if (!extent_buffer_uptodate(log_tree_root->node)) { | |
f14d104d | 2382 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2383 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2384 | return -EIO; |
2385 | } | |
4eb150d6 | 2386 | |
63443bf5 ES |
2387 | /* returns with log_tree_root freed on success */ |
2388 | ret = btrfs_recover_log_trees(log_tree_root); | |
2389 | if (ret) { | |
0b246afa JM |
2390 | btrfs_handle_fs_error(fs_info, ret, |
2391 | "Failed to recover log tree"); | |
00246528 | 2392 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2393 | return ret; |
2394 | } | |
2395 | ||
bc98a42c | 2396 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2397 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2398 | if (ret) |
2399 | return ret; | |
2400 | } | |
2401 | ||
2402 | return 0; | |
2403 | } | |
2404 | ||
abed4aaa JB |
2405 | static int load_global_roots_objectid(struct btrfs_root *tree_root, |
2406 | struct btrfs_path *path, u64 objectid, | |
2407 | const char *name) | |
2408 | { | |
2409 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
2410 | struct btrfs_root *root; | |
f7238e50 | 2411 | u64 max_global_id = 0; |
abed4aaa JB |
2412 | int ret; |
2413 | struct btrfs_key key = { | |
2414 | .objectid = objectid, | |
2415 | .type = BTRFS_ROOT_ITEM_KEY, | |
2416 | .offset = 0, | |
2417 | }; | |
2418 | bool found = false; | |
2419 | ||
2420 | /* If we have IGNOREDATACSUMS skip loading these roots. */ | |
2421 | if (objectid == BTRFS_CSUM_TREE_OBJECTID && | |
2422 | btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2423 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2424 | return 0; | |
2425 | } | |
2426 | ||
2427 | while (1) { | |
2428 | ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); | |
2429 | if (ret < 0) | |
2430 | break; | |
2431 | ||
2432 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
2433 | ret = btrfs_next_leaf(tree_root, path); | |
2434 | if (ret) { | |
2435 | if (ret > 0) | |
2436 | ret = 0; | |
2437 | break; | |
2438 | } | |
2439 | } | |
2440 | ret = 0; | |
2441 | ||
2442 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
2443 | if (key.objectid != objectid) | |
2444 | break; | |
2445 | btrfs_release_path(path); | |
2446 | ||
f7238e50 JB |
2447 | /* |
2448 | * Just worry about this for extent tree, it'll be the same for | |
2449 | * everybody. | |
2450 | */ | |
2451 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
2452 | max_global_id = max(max_global_id, key.offset); | |
2453 | ||
abed4aaa JB |
2454 | found = true; |
2455 | root = read_tree_root_path(tree_root, path, &key); | |
2456 | if (IS_ERR(root)) { | |
2457 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2458 | ret = PTR_ERR(root); | |
2459 | break; | |
2460 | } | |
2461 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2462 | ret = btrfs_global_root_insert(root); | |
2463 | if (ret) { | |
2464 | btrfs_put_root(root); | |
2465 | break; | |
2466 | } | |
2467 | key.offset++; | |
2468 | } | |
2469 | btrfs_release_path(path); | |
2470 | ||
f7238e50 JB |
2471 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) |
2472 | fs_info->nr_global_roots = max_global_id + 1; | |
2473 | ||
abed4aaa JB |
2474 | if (!found || ret) { |
2475 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
2476 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2477 | ||
2478 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2479 | ret = ret ? ret : -ENOENT; | |
2480 | else | |
2481 | ret = 0; | |
2482 | btrfs_err(fs_info, "failed to load root %s", name); | |
2483 | } | |
2484 | return ret; | |
2485 | } | |
2486 | ||
2487 | static int load_global_roots(struct btrfs_root *tree_root) | |
2488 | { | |
2489 | struct btrfs_path *path; | |
2490 | int ret = 0; | |
2491 | ||
2492 | path = btrfs_alloc_path(); | |
2493 | if (!path) | |
2494 | return -ENOMEM; | |
2495 | ||
2496 | ret = load_global_roots_objectid(tree_root, path, | |
2497 | BTRFS_EXTENT_TREE_OBJECTID, "extent"); | |
2498 | if (ret) | |
2499 | goto out; | |
2500 | ret = load_global_roots_objectid(tree_root, path, | |
2501 | BTRFS_CSUM_TREE_OBJECTID, "csum"); | |
2502 | if (ret) | |
2503 | goto out; | |
2504 | if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) | |
2505 | goto out; | |
2506 | ret = load_global_roots_objectid(tree_root, path, | |
2507 | BTRFS_FREE_SPACE_TREE_OBJECTID, | |
2508 | "free space"); | |
2509 | out: | |
2510 | btrfs_free_path(path); | |
2511 | return ret; | |
2512 | } | |
2513 | ||
6bccf3ab | 2514 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2515 | { |
6bccf3ab | 2516 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2517 | struct btrfs_root *root; |
4bbcaa64 ES |
2518 | struct btrfs_key location; |
2519 | int ret; | |
2520 | ||
6bccf3ab JM |
2521 | BUG_ON(!fs_info->tree_root); |
2522 | ||
abed4aaa JB |
2523 | ret = load_global_roots(tree_root); |
2524 | if (ret) | |
2525 | return ret; | |
2526 | ||
4bbcaa64 ES |
2527 | location.type = BTRFS_ROOT_ITEM_KEY; |
2528 | location.offset = 0; | |
2529 | ||
1c56ab99 | 2530 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { |
14033b08 QW |
2531 | location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; |
2532 | root = btrfs_read_tree_root(tree_root, &location); | |
2533 | if (IS_ERR(root)) { | |
2534 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2535 | ret = PTR_ERR(root); | |
2536 | goto out; | |
2537 | } | |
2538 | } else { | |
2539 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2540 | fs_info->block_group_root = root; | |
2541 | } | |
2542 | } | |
2543 | ||
2544 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2545 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2546 | if (IS_ERR(root)) { |
42437a63 JB |
2547 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2548 | ret = PTR_ERR(root); | |
2549 | goto out; | |
2550 | } | |
2551 | } else { | |
2552 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2553 | fs_info->dev_root = root; | |
f50f4353 | 2554 | } |
820a49da JB |
2555 | /* Initialize fs_info for all devices in any case */ |
2556 | btrfs_init_devices_late(fs_info); | |
4bbcaa64 | 2557 | |
aeb935a4 QW |
2558 | /* |
2559 | * This tree can share blocks with some other fs tree during relocation | |
2560 | * and we need a proper setup by btrfs_get_fs_root | |
2561 | */ | |
56e9357a DS |
2562 | root = btrfs_get_fs_root(tree_root->fs_info, |
2563 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2564 | if (IS_ERR(root)) { |
42437a63 JB |
2565 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2566 | ret = PTR_ERR(root); | |
2567 | goto out; | |
2568 | } | |
2569 | } else { | |
2570 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2571 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2572 | } |
aeb935a4 | 2573 | |
4bbcaa64 | 2574 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2575 | root = btrfs_read_tree_root(tree_root, &location); |
2576 | if (!IS_ERR(root)) { | |
2577 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2578 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2579 | fs_info->quota_root = root; |
4bbcaa64 ES |
2580 | } |
2581 | ||
2582 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2583 | root = btrfs_read_tree_root(tree_root, &location); |
2584 | if (IS_ERR(root)) { | |
42437a63 JB |
2585 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2586 | ret = PTR_ERR(root); | |
2587 | if (ret != -ENOENT) | |
2588 | goto out; | |
2589 | } | |
4bbcaa64 | 2590 | } else { |
a4f3d2c4 DS |
2591 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2592 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2593 | } |
2594 | ||
2595 | return 0; | |
f50f4353 LB |
2596 | out: |
2597 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2598 | location.objectid, ret); | |
2599 | return ret; | |
4bbcaa64 ES |
2600 | } |
2601 | ||
069ec957 QW |
2602 | /* |
2603 | * Real super block validation | |
2604 | * NOTE: super csum type and incompat features will not be checked here. | |
2605 | * | |
2606 | * @sb: super block to check | |
2607 | * @mirror_num: the super block number to check its bytenr: | |
2608 | * 0 the primary (1st) sb | |
2609 | * 1, 2 2nd and 3rd backup copy | |
2610 | * -1 skip bytenr check | |
2611 | */ | |
a05d3c91 QW |
2612 | int btrfs_validate_super(struct btrfs_fs_info *fs_info, |
2613 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2614 | { |
21a852b0 QW |
2615 | u64 nodesize = btrfs_super_nodesize(sb); |
2616 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2617 | int ret = 0; | |
2618 | ||
2619 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2620 | btrfs_err(fs_info, "no valid FS found"); | |
2621 | ret = -EINVAL; | |
2622 | } | |
2623 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2624 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2625 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2626 | ret = -EINVAL; | |
2627 | } | |
2628 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2629 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2630 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2631 | ret = -EINVAL; | |
2632 | } | |
2633 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2634 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2635 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2636 | ret = -EINVAL; | |
2637 | } | |
2638 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2639 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2640 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2641 | ret = -EINVAL; | |
2642 | } | |
2643 | ||
2644 | /* | |
2645 | * Check sectorsize and nodesize first, other check will need it. | |
2646 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2647 | */ | |
2648 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2649 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2650 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2651 | ret = -EINVAL; | |
2652 | } | |
0bb3eb3e QW |
2653 | |
2654 | /* | |
1a42daab QW |
2655 | * We only support at most two sectorsizes: 4K and PAGE_SIZE. |
2656 | * | |
2657 | * We can support 16K sectorsize with 64K page size without problem, | |
2658 | * but such sectorsize/pagesize combination doesn't make much sense. | |
2659 | * 4K will be our future standard, PAGE_SIZE is supported from the very | |
2660 | * beginning. | |
0bb3eb3e | 2661 | */ |
1a42daab | 2662 | if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { |
21a852b0 | 2663 | btrfs_err(fs_info, |
0bb3eb3e | 2664 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2665 | sectorsize, PAGE_SIZE); |
2666 | ret = -EINVAL; | |
2667 | } | |
0bb3eb3e | 2668 | |
21a852b0 QW |
2669 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2670 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2671 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2672 | ret = -EINVAL; | |
2673 | } | |
2674 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2675 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2676 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2677 | ret = -EINVAL; | |
2678 | } | |
2679 | ||
2680 | /* Root alignment check */ | |
2681 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2682 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2683 | btrfs_super_root(sb)); | |
2684 | ret = -EINVAL; | |
2685 | } | |
2686 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2687 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2688 | btrfs_super_chunk_root(sb)); | |
2689 | ret = -EINVAL; | |
2690 | } | |
2691 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2692 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2693 | btrfs_super_log_root(sb)); | |
2694 | ret = -EINVAL; | |
2695 | } | |
2696 | ||
aefd7f70 NB |
2697 | if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, |
2698 | BTRFS_FSID_SIZE)) { | |
2699 | btrfs_err(fs_info, | |
2700 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
2701 | fs_info->super_copy->fsid, fs_info->fs_devices->fsid); | |
2702 | ret = -EINVAL; | |
2703 | } | |
2704 | ||
2705 | if (btrfs_fs_incompat(fs_info, METADATA_UUID) && | |
2706 | memcmp(fs_info->fs_devices->metadata_uuid, | |
2707 | fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { | |
2708 | btrfs_err(fs_info, | |
2709 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
2710 | fs_info->super_copy->metadata_uuid, | |
2711 | fs_info->fs_devices->metadata_uuid); | |
2712 | ret = -EINVAL; | |
2713 | } | |
2714 | ||
1c56ab99 QW |
2715 | /* |
2716 | * Artificial requirement for block-group-tree to force newer features | |
2717 | * (free-space-tree, no-holes) so the test matrix is smaller. | |
2718 | */ | |
2719 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
2720 | (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || | |
2721 | !btrfs_fs_incompat(fs_info, NO_HOLES))) { | |
2722 | btrfs_err(fs_info, | |
2723 | "block-group-tree feature requires fres-space-tree and no-holes"); | |
2724 | ret = -EINVAL; | |
2725 | } | |
2726 | ||
de37aa51 | 2727 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
7239ff4b | 2728 | BTRFS_FSID_SIZE) != 0) { |
21a852b0 | 2729 | btrfs_err(fs_info, |
7239ff4b | 2730 | "dev_item UUID does not match metadata fsid: %pU != %pU", |
de37aa51 | 2731 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); |
21a852b0 QW |
2732 | ret = -EINVAL; |
2733 | } | |
2734 | ||
2735 | /* | |
2736 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2737 | * done later | |
2738 | */ | |
2739 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2740 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2741 | btrfs_super_bytes_used(sb)); | |
2742 | ret = -EINVAL; | |
2743 | } | |
2744 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2745 | btrfs_err(fs_info, "invalid stripesize %u", | |
2746 | btrfs_super_stripesize(sb)); | |
2747 | ret = -EINVAL; | |
2748 | } | |
2749 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2750 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2751 | btrfs_super_num_devices(sb)); | |
2752 | if (btrfs_super_num_devices(sb) == 0) { | |
2753 | btrfs_err(fs_info, "number of devices is 0"); | |
2754 | ret = -EINVAL; | |
2755 | } | |
2756 | ||
069ec957 QW |
2757 | if (mirror_num >= 0 && |
2758 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2759 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2760 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2761 | ret = -EINVAL; | |
2762 | } | |
2763 | ||
2764 | /* | |
2765 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2766 | * and one chunk | |
2767 | */ | |
2768 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2769 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2770 | btrfs_super_sys_array_size(sb), | |
2771 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2772 | ret = -EINVAL; | |
2773 | } | |
2774 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2775 | + sizeof(struct btrfs_chunk)) { | |
2776 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2777 | btrfs_super_sys_array_size(sb), | |
2778 | sizeof(struct btrfs_disk_key) | |
2779 | + sizeof(struct btrfs_chunk)); | |
2780 | ret = -EINVAL; | |
2781 | } | |
2782 | ||
2783 | /* | |
2784 | * The generation is a global counter, we'll trust it more than the others | |
2785 | * but it's still possible that it's the one that's wrong. | |
2786 | */ | |
2787 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2788 | btrfs_warn(fs_info, | |
2789 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2790 | btrfs_super_generation(sb), | |
2791 | btrfs_super_chunk_root_generation(sb)); | |
2792 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2793 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2794 | btrfs_warn(fs_info, | |
2795 | "suspicious: generation < cache_generation: %llu < %llu", | |
2796 | btrfs_super_generation(sb), | |
2797 | btrfs_super_cache_generation(sb)); | |
2798 | ||
2799 | return ret; | |
2800 | } | |
2801 | ||
069ec957 QW |
2802 | /* |
2803 | * Validation of super block at mount time. | |
2804 | * Some checks already done early at mount time, like csum type and incompat | |
2805 | * flags will be skipped. | |
2806 | */ | |
2807 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2808 | { | |
a05d3c91 | 2809 | return btrfs_validate_super(fs_info, fs_info->super_copy, 0); |
069ec957 QW |
2810 | } |
2811 | ||
75cb857d QW |
2812 | /* |
2813 | * Validation of super block at write time. | |
2814 | * Some checks like bytenr check will be skipped as their values will be | |
2815 | * overwritten soon. | |
2816 | * Extra checks like csum type and incompat flags will be done here. | |
2817 | */ | |
2818 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2819 | struct btrfs_super_block *sb) | |
2820 | { | |
2821 | int ret; | |
2822 | ||
a05d3c91 | 2823 | ret = btrfs_validate_super(fs_info, sb, -1); |
75cb857d QW |
2824 | if (ret < 0) |
2825 | goto out; | |
e7e16f48 | 2826 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2827 | ret = -EUCLEAN; |
2828 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2829 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2830 | goto out; | |
2831 | } | |
2832 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2833 | ret = -EUCLEAN; | |
2834 | btrfs_err(fs_info, | |
2835 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2836 | btrfs_super_incompat_flags(sb), | |
2837 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2838 | goto out; | |
2839 | } | |
2840 | out: | |
2841 | if (ret < 0) | |
2842 | btrfs_err(fs_info, | |
2843 | "super block corruption detected before writing it to disk"); | |
2844 | return ret; | |
2845 | } | |
2846 | ||
bd676446 JB |
2847 | static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) |
2848 | { | |
2849 | int ret = 0; | |
2850 | ||
2851 | root->node = read_tree_block(root->fs_info, bytenr, | |
2852 | root->root_key.objectid, gen, level, NULL); | |
2853 | if (IS_ERR(root->node)) { | |
2854 | ret = PTR_ERR(root->node); | |
2855 | root->node = NULL; | |
4eb150d6 QW |
2856 | return ret; |
2857 | } | |
2858 | if (!extent_buffer_uptodate(root->node)) { | |
bd676446 JB |
2859 | free_extent_buffer(root->node); |
2860 | root->node = NULL; | |
4eb150d6 | 2861 | return -EIO; |
bd676446 JB |
2862 | } |
2863 | ||
bd676446 JB |
2864 | btrfs_set_root_node(&root->root_item, root->node); |
2865 | root->commit_root = btrfs_root_node(root); | |
2866 | btrfs_set_root_refs(&root->root_item, 1); | |
2867 | return ret; | |
2868 | } | |
2869 | ||
2870 | static int load_important_roots(struct btrfs_fs_info *fs_info) | |
2871 | { | |
2872 | struct btrfs_super_block *sb = fs_info->super_copy; | |
2873 | u64 gen, bytenr; | |
2874 | int level, ret; | |
2875 | ||
2876 | bytenr = btrfs_super_root(sb); | |
2877 | gen = btrfs_super_generation(sb); | |
2878 | level = btrfs_super_root_level(sb); | |
2879 | ret = load_super_root(fs_info->tree_root, bytenr, gen, level); | |
9c54e80d | 2880 | if (ret) { |
bd676446 | 2881 | btrfs_warn(fs_info, "couldn't read tree root"); |
9c54e80d JB |
2882 | return ret; |
2883 | } | |
14033b08 | 2884 | return 0; |
bd676446 JB |
2885 | } |
2886 | ||
6ef108dd | 2887 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2888 | { |
6ef108dd | 2889 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2890 | struct btrfs_super_block *sb = fs_info->super_copy; |
2891 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2892 | bool handle_error = false; | |
2893 | int ret = 0; | |
2894 | int i; | |
2895 | ||
2896 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
b8522a1e NB |
2897 | if (handle_error) { |
2898 | if (!IS_ERR(tree_root->node)) | |
2899 | free_extent_buffer(tree_root->node); | |
2900 | tree_root->node = NULL; | |
2901 | ||
2902 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2903 | break; | |
2904 | ||
2905 | free_root_pointers(fs_info, 0); | |
2906 | ||
2907 | /* | |
2908 | * Don't use the log in recovery mode, it won't be | |
2909 | * valid | |
2910 | */ | |
2911 | btrfs_set_super_log_root(sb, 0); | |
2912 | ||
2913 | /* We can't trust the free space cache either */ | |
2914 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2915 | ||
2916 | ret = read_backup_root(fs_info, i); | |
6ef108dd | 2917 | backup_index = ret; |
b8522a1e NB |
2918 | if (ret < 0) |
2919 | return ret; | |
2920 | } | |
b8522a1e | 2921 | |
bd676446 JB |
2922 | ret = load_important_roots(fs_info); |
2923 | if (ret) { | |
217f5004 | 2924 | handle_error = true; |
b8522a1e NB |
2925 | continue; |
2926 | } | |
2927 | ||
336a0d8d NB |
2928 | /* |
2929 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2930 | * hasn't been fully initialised and we are the only user | |
2931 | */ | |
453e4873 | 2932 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 2933 | if (ret < 0) { |
b8522a1e NB |
2934 | handle_error = true; |
2935 | continue; | |
2936 | } | |
2937 | ||
6b8fad57 | 2938 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
2939 | |
2940 | ret = btrfs_read_roots(fs_info); | |
2941 | if (ret < 0) { | |
2942 | handle_error = true; | |
2943 | continue; | |
2944 | } | |
2945 | ||
2946 | /* All successful */ | |
bd676446 JB |
2947 | fs_info->generation = btrfs_header_generation(tree_root->node); |
2948 | fs_info->last_trans_committed = fs_info->generation; | |
d96b3424 | 2949 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
2950 | |
2951 | /* Always begin writing backup roots after the one being used */ | |
2952 | if (backup_index < 0) { | |
2953 | fs_info->backup_root_index = 0; | |
2954 | } else { | |
2955 | fs_info->backup_root_index = backup_index + 1; | |
2956 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2957 | } | |
b8522a1e NB |
2958 | break; |
2959 | } | |
2960 | ||
2961 | return ret; | |
2962 | } | |
2963 | ||
8260edba | 2964 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 2965 | { |
fc7cbcd4 | 2966 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
01cd3909 | 2967 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2968 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2969 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2970 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2971 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2972 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2973 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2974 | spin_lock_init(&fs_info->trans_lock); |
fc7cbcd4 | 2975 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2976 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2977 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 2978 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2979 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2980 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 2981 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 2982 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
c2707a25 | 2983 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 2984 | rwlock_init(&fs_info->tree_mod_log_lock); |
abed4aaa | 2985 | rwlock_init(&fs_info->global_root_lock); |
d7c15171 | 2986 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 2987 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 2988 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2989 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 2990 | mutex_init(&fs_info->zoned_meta_io_lock); |
5f0addf7 | 2991 | mutex_init(&fs_info->zoned_data_reloc_io_lock); |
de98ced9 | 2992 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2993 | |
e1489b4f | 2994 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); |
5a9ba670 | 2995 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); |
8b53779e | 2996 | btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); |
5f4403e1 | 2997 | btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); |
3e738c53 IA |
2998 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, |
2999 | BTRFS_LOCKDEP_TRANS_COMMIT_START); | |
3000 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, | |
3001 | BTRFS_LOCKDEP_TRANS_UNBLOCKED); | |
3002 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, | |
3003 | BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); | |
3004 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, | |
3005 | BTRFS_LOCKDEP_TRANS_COMPLETED); | |
e1489b4f | 3006 | |
0b86a832 | 3007 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 3008 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 3009 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 3010 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 3011 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 3012 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
3013 | #ifdef CONFIG_BTRFS_DEBUG |
3014 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
3015 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
3016 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 3017 | #endif |
c8bf1b67 | 3018 | extent_map_tree_init(&fs_info->mapping_tree); |
66d8f3dd MX |
3019 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
3020 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
3021 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
3022 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
3023 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
3024 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
3025 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
3026 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
3027 | BTRFS_BLOCK_RSV_DELREFS); | |
3028 | ||
771ed689 | 3029 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 3030 | atomic_set(&fs_info->defrag_running, 0); |
034f784d | 3031 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 3032 | atomic64_set(&fs_info->tree_mod_seq, 0); |
abed4aaa | 3033 | fs_info->global_root_tree = RB_ROOT; |
95ac567a | 3034 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 3035 | fs_info->metadata_ratio = 0; |
4cb5300b | 3036 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 3037 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 3038 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 3039 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 3040 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
fd708b81 | 3041 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 3042 | |
b34b086c CM |
3043 | fs_info->thread_pool_size = min_t(unsigned long, |
3044 | num_online_cpus() + 2, 8); | |
0afbaf8c | 3045 | |
199c2a9c MX |
3046 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
3047 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 3048 | |
638aa7ed | 3049 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
3050 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3051 | fs_info->check_integrity_print_mask = 0; | |
3052 | #endif | |
779a65a4 | 3053 | btrfs_init_balance(fs_info); |
57056740 | 3054 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 3055 | |
16b0c258 | 3056 | rwlock_init(&fs_info->block_group_cache_lock); |
08dddb29 | 3057 | fs_info->block_group_cache_tree = RB_ROOT_CACHED; |
0f9dd46c | 3058 | |
fe119a6e NB |
3059 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
3060 | IO_TREE_FS_EXCLUDED_EXTENTS, NULL); | |
39279cc3 | 3061 | |
5a3f23d5 | 3062 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 3063 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 3064 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
3065 | mutex_init(&fs_info->transaction_kthread_mutex); |
3066 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 3067 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 3068 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 3069 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 3070 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 3071 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 3072 | |
ad618368 | 3073 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 3074 | btrfs_init_qgroup(fs_info); |
b0643e59 | 3075 | btrfs_discard_init(fs_info); |
416ac51d | 3076 | |
fa9c0d79 CM |
3077 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
3078 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
3079 | ||
e6dcd2dc | 3080 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 3081 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 3082 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 3083 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 3084 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 3085 | |
da17066c JM |
3086 | /* Usable values until the real ones are cached from the superblock */ |
3087 | fs_info->nodesize = 4096; | |
3088 | fs_info->sectorsize = 4096; | |
ab108d99 | 3089 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
3090 | fs_info->stripesize = 4096; |
3091 | ||
f7b12a62 NA |
3092 | fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; |
3093 | ||
eede2bf3 OS |
3094 | spin_lock_init(&fs_info->swapfile_pins_lock); |
3095 | fs_info->swapfile_pins = RB_ROOT; | |
3096 | ||
18bb8bbf JT |
3097 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
3098 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
3099 | } |
3100 | ||
3101 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
3102 | { | |
3103 | int ret; | |
3104 | ||
3105 | fs_info->sb = sb; | |
3106 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; | |
3107 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 3108 | |
5deb17e1 | 3109 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 3110 | if (ret) |
c75e8394 | 3111 | return ret; |
ae18c37a JB |
3112 | |
3113 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
3114 | if (ret) | |
c75e8394 | 3115 | return ret; |
ae18c37a JB |
3116 | |
3117 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
3118 | (1 + ilog2(nr_cpu_ids)); | |
3119 | ||
3120 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
3121 | if (ret) | |
c75e8394 | 3122 | return ret; |
ae18c37a JB |
3123 | |
3124 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
3125 | GFP_KERNEL); | |
3126 | if (ret) | |
c75e8394 | 3127 | return ret; |
ae18c37a JB |
3128 | |
3129 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
3130 | GFP_KERNEL); | |
c75e8394 JB |
3131 | if (!fs_info->delayed_root) |
3132 | return -ENOMEM; | |
ae18c37a JB |
3133 | btrfs_init_delayed_root(fs_info->delayed_root); |
3134 | ||
a0a1db70 FM |
3135 | if (sb_rdonly(sb)) |
3136 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
3137 | ||
c75e8394 | 3138 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
3139 | } |
3140 | ||
97f4dd09 NB |
3141 | static int btrfs_uuid_rescan_kthread(void *data) |
3142 | { | |
0d031dc4 | 3143 | struct btrfs_fs_info *fs_info = data; |
97f4dd09 NB |
3144 | int ret; |
3145 | ||
3146 | /* | |
3147 | * 1st step is to iterate through the existing UUID tree and | |
3148 | * to delete all entries that contain outdated data. | |
3149 | * 2nd step is to add all missing entries to the UUID tree. | |
3150 | */ | |
3151 | ret = btrfs_uuid_tree_iterate(fs_info); | |
3152 | if (ret < 0) { | |
c94bec2c JB |
3153 | if (ret != -EINTR) |
3154 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
3155 | ret); | |
97f4dd09 NB |
3156 | up(&fs_info->uuid_tree_rescan_sem); |
3157 | return ret; | |
3158 | } | |
3159 | return btrfs_uuid_scan_kthread(data); | |
3160 | } | |
3161 | ||
3162 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
3163 | { | |
3164 | struct task_struct *task; | |
3165 | ||
3166 | down(&fs_info->uuid_tree_rescan_sem); | |
3167 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
3168 | if (IS_ERR(task)) { | |
3169 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
3170 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
3171 | up(&fs_info->uuid_tree_rescan_sem); | |
3172 | return PTR_ERR(task); | |
3173 | } | |
3174 | ||
3175 | return 0; | |
3176 | } | |
3177 | ||
8cd29088 BB |
3178 | /* |
3179 | * Some options only have meaning at mount time and shouldn't persist across | |
3180 | * remounts, or be displayed. Clear these at the end of mount and remount | |
3181 | * code paths. | |
3182 | */ | |
3183 | void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) | |
3184 | { | |
3185 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
8b228324 | 3186 | btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); |
8cd29088 BB |
3187 | } |
3188 | ||
44c0ca21 BB |
3189 | /* |
3190 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
3191 | * and btrfs_remount when remounting from read-only to read-write. | |
3192 | */ | |
3193 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
3194 | { | |
3195 | int ret; | |
94846229 | 3196 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
8b228324 BB |
3197 | bool clear_free_space_tree = false; |
3198 | ||
3199 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
3200 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3201 | clear_free_space_tree = true; | |
3202 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3203 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3204 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3205 | clear_free_space_tree = true; | |
3206 | } | |
3207 | ||
3208 | if (clear_free_space_tree) { | |
3209 | btrfs_info(fs_info, "clearing free space tree"); | |
3210 | ret = btrfs_clear_free_space_tree(fs_info); | |
3211 | if (ret) { | |
3212 | btrfs_warn(fs_info, | |
3213 | "failed to clear free space tree: %d", ret); | |
3214 | goto out; | |
3215 | } | |
3216 | } | |
44c0ca21 | 3217 | |
8d488a8c FM |
3218 | /* |
3219 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3220 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
fc7cbcd4 | 3221 | * them into the fs_info->fs_roots_radix tree. This must be done before |
8d488a8c FM |
3222 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it |
3223 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3224 | * item before the root's tree is deleted - this means that if we unmount | |
3225 | * or crash before the deletion completes, on the next mount we will not | |
3226 | * delete what remains of the tree because the orphan item does not | |
3227 | * exists anymore, which is what tells us we have a pending deletion. | |
3228 | */ | |
3229 | ret = btrfs_find_orphan_roots(fs_info); | |
3230 | if (ret) | |
3231 | goto out; | |
3232 | ||
44c0ca21 BB |
3233 | ret = btrfs_cleanup_fs_roots(fs_info); |
3234 | if (ret) | |
3235 | goto out; | |
3236 | ||
8f1c21d7 BB |
3237 | down_read(&fs_info->cleanup_work_sem); |
3238 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3239 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3240 | up_read(&fs_info->cleanup_work_sem); | |
3241 | goto out; | |
3242 | } | |
3243 | up_read(&fs_info->cleanup_work_sem); | |
3244 | ||
44c0ca21 | 3245 | mutex_lock(&fs_info->cleaner_mutex); |
7eefae6b | 3246 | ret = btrfs_recover_relocation(fs_info); |
44c0ca21 BB |
3247 | mutex_unlock(&fs_info->cleaner_mutex); |
3248 | if (ret < 0) { | |
3249 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3250 | goto out; | |
3251 | } | |
3252 | ||
5011139a BB |
3253 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3254 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3255 | btrfs_info(fs_info, "creating free space tree"); | |
3256 | ret = btrfs_create_free_space_tree(fs_info); | |
3257 | if (ret) { | |
3258 | btrfs_warn(fs_info, | |
3259 | "failed to create free space tree: %d", ret); | |
3260 | goto out; | |
3261 | } | |
3262 | } | |
3263 | ||
94846229 BB |
3264 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3265 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3266 | if (ret) | |
3267 | goto out; | |
3268 | } | |
3269 | ||
44c0ca21 BB |
3270 | ret = btrfs_resume_balance_async(fs_info); |
3271 | if (ret) | |
3272 | goto out; | |
3273 | ||
3274 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3275 | if (ret) { | |
3276 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3277 | goto out; | |
3278 | } | |
3279 | ||
3280 | btrfs_qgroup_rescan_resume(fs_info); | |
3281 | ||
3282 | if (!fs_info->uuid_root) { | |
3283 | btrfs_info(fs_info, "creating UUID tree"); | |
3284 | ret = btrfs_create_uuid_tree(fs_info); | |
3285 | if (ret) { | |
3286 | btrfs_warn(fs_info, | |
3287 | "failed to create the UUID tree %d", ret); | |
3288 | goto out; | |
3289 | } | |
3290 | } | |
3291 | ||
3292 | out: | |
3293 | return ret; | |
3294 | } | |
3295 | ||
ae18c37a JB |
3296 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3297 | char *options) | |
3298 | { | |
3299 | u32 sectorsize; | |
3300 | u32 nodesize; | |
3301 | u32 stripesize; | |
3302 | u64 generation; | |
3303 | u64 features; | |
3304 | u16 csum_type; | |
ae18c37a JB |
3305 | struct btrfs_super_block *disk_super; |
3306 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3307 | struct btrfs_root *tree_root; | |
3308 | struct btrfs_root *chunk_root; | |
3309 | int ret; | |
3310 | int err = -EINVAL; | |
ae18c37a JB |
3311 | int level; |
3312 | ||
8260edba | 3313 | ret = init_mount_fs_info(fs_info, sb); |
53b381b3 | 3314 | if (ret) { |
83c8266a | 3315 | err = ret; |
ae18c37a | 3316 | goto fail; |
53b381b3 DW |
3317 | } |
3318 | ||
ae18c37a JB |
3319 | /* These need to be init'ed before we start creating inodes and such. */ |
3320 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3321 | GFP_KERNEL); | |
3322 | fs_info->tree_root = tree_root; | |
3323 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3324 | GFP_KERNEL); | |
3325 | fs_info->chunk_root = chunk_root; | |
3326 | if (!tree_root || !chunk_root) { | |
3327 | err = -ENOMEM; | |
c75e8394 | 3328 | goto fail; |
ae18c37a JB |
3329 | } |
3330 | ||
3331 | fs_info->btree_inode = new_inode(sb); | |
3332 | if (!fs_info->btree_inode) { | |
3333 | err = -ENOMEM; | |
c75e8394 | 3334 | goto fail; |
ae18c37a JB |
3335 | } |
3336 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
3337 | btrfs_init_btree_inode(fs_info); | |
3338 | ||
d24fa5c1 | 3339 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3340 | |
3341 | /* | |
3342 | * Read super block and check the signature bytes only | |
3343 | */ | |
d24fa5c1 | 3344 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d JT |
3345 | if (IS_ERR(disk_super)) { |
3346 | err = PTR_ERR(disk_super); | |
16cdcec7 | 3347 | goto fail_alloc; |
20b45077 | 3348 | } |
39279cc3 | 3349 | |
8dc3f22c | 3350 | /* |
260db43c | 3351 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3352 | * corrupted, we'll find out |
3353 | */ | |
8f32380d | 3354 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3355 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3356 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3357 | csum_type); |
8dc3f22c | 3358 | err = -EINVAL; |
8f32380d | 3359 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3360 | goto fail_alloc; |
3361 | } | |
3362 | ||
83c68bbc SY |
3363 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3364 | ||
6d97c6e3 JT |
3365 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3366 | if (ret) { | |
3367 | err = ret; | |
8f32380d | 3368 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3369 | goto fail_alloc; |
3370 | } | |
3371 | ||
1104a885 DS |
3372 | /* |
3373 | * We want to check superblock checksum, the type is stored inside. | |
3374 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3375 | */ | |
8f32380d | 3376 | if (btrfs_check_super_csum(fs_info, (u8 *)disk_super)) { |
05135f59 | 3377 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 3378 | err = -EINVAL; |
8f32380d | 3379 | btrfs_release_disk_super(disk_super); |
141386e1 | 3380 | goto fail_alloc; |
1104a885 DS |
3381 | } |
3382 | ||
3383 | /* | |
3384 | * super_copy is zeroed at allocation time and we never touch the | |
3385 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3386 | * the whole block of INFO_SIZE | |
3387 | */ | |
8f32380d JT |
3388 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3389 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3390 | |
fbc6feae NB |
3391 | disk_super = fs_info->super_copy; |
3392 | ||
0b86a832 | 3393 | |
fbc6feae NB |
3394 | features = btrfs_super_flags(disk_super); |
3395 | if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { | |
3396 | features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; | |
3397 | btrfs_set_super_flags(disk_super, features); | |
3398 | btrfs_info(fs_info, | |
3399 | "found metadata UUID change in progress flag, clearing"); | |
3400 | } | |
3401 | ||
3402 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
3403 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3404 | |
069ec957 | 3405 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3406 | if (ret) { |
05135f59 | 3407 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 | 3408 | err = -EINVAL; |
141386e1 | 3409 | goto fail_alloc; |
1104a885 DS |
3410 | } |
3411 | ||
0f7d52f4 | 3412 | if (!btrfs_super_root(disk_super)) |
141386e1 | 3413 | goto fail_alloc; |
0f7d52f4 | 3414 | |
acce952b | 3415 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
3416 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
3417 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 3418 | |
75e7cb7f LB |
3419 | /* |
3420 | * In the long term, we'll store the compression type in the super | |
3421 | * block, and it'll be used for per file compression control. | |
3422 | */ | |
3423 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
3424 | ||
6f93e834 AJ |
3425 | |
3426 | /* Set up fs_info before parsing mount options */ | |
3427 | nodesize = btrfs_super_nodesize(disk_super); | |
3428 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3429 | stripesize = sectorsize; | |
3430 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3431 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3432 | ||
3433 | fs_info->nodesize = nodesize; | |
3434 | fs_info->sectorsize = sectorsize; | |
3435 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3436 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3437 | fs_info->stripesize = stripesize; | |
3438 | ||
2ff7e61e | 3439 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
3440 | if (ret) { |
3441 | err = ret; | |
141386e1 | 3442 | goto fail_alloc; |
2b82032c | 3443 | } |
dfe25020 | 3444 | |
f2b636e8 JB |
3445 | features = btrfs_super_incompat_flags(disk_super) & |
3446 | ~BTRFS_FEATURE_INCOMPAT_SUPP; | |
3447 | if (features) { | |
05135f59 | 3448 | btrfs_err(fs_info, |
d5321a0f | 3449 | "cannot mount because of unsupported optional features (0x%llx)", |
05135f59 | 3450 | features); |
f2b636e8 | 3451 | err = -EINVAL; |
141386e1 | 3452 | goto fail_alloc; |
f2b636e8 JB |
3453 | } |
3454 | ||
5d4f98a2 | 3455 | features = btrfs_super_incompat_flags(disk_super); |
a6fa6fae | 3456 | features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; |
0b246afa | 3457 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) |
a6fa6fae | 3458 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; |
5c1aab1d NT |
3459 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) |
3460 | features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
727011e0 | 3461 | |
e26b04c4 NB |
3462 | /* |
3463 | * Flag our filesystem as having big metadata blocks if they are bigger | |
3464 | * than the page size. | |
3465 | */ | |
6b769dac | 3466 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) |
e26b04c4 | 3467 | features |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; |
3173a18f | 3468 | |
bc3f116f CM |
3469 | /* |
3470 | * mixed block groups end up with duplicate but slightly offset | |
3471 | * extent buffers for the same range. It leads to corruptions | |
3472 | */ | |
3473 | if ((features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
707e8a07 | 3474 | (sectorsize != nodesize)) { |
05135f59 DS |
3475 | btrfs_err(fs_info, |
3476 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3477 | nodesize, sectorsize); | |
141386e1 | 3478 | goto fail_alloc; |
bc3f116f CM |
3479 | } |
3480 | ||
ceda0864 MX |
3481 | /* |
3482 | * Needn't use the lock because there is no other task which will | |
3483 | * update the flag. | |
3484 | */ | |
a6fa6fae | 3485 | btrfs_set_super_incompat_flags(disk_super, features); |
5d4f98a2 | 3486 | |
f2b636e8 JB |
3487 | features = btrfs_super_compat_ro_flags(disk_super) & |
3488 | ~BTRFS_FEATURE_COMPAT_RO_SUPP; | |
bc98a42c | 3489 | if (!sb_rdonly(sb) && features) { |
05135f59 | 3490 | btrfs_err(fs_info, |
d5321a0f | 3491 | "cannot mount read-write because of unsupported optional features (0x%llx)", |
c1c9ff7c | 3492 | features); |
f2b636e8 | 3493 | err = -EINVAL; |
141386e1 | 3494 | goto fail_alloc; |
f2b636e8 | 3495 | } |
dc4d3168 QW |
3496 | /* |
3497 | * We have unsupported RO compat features, although RO mounted, we | |
3498 | * should not cause any metadata write, including log replay. | |
3499 | * Or we could screw up whatever the new feature requires. | |
3500 | */ | |
3501 | if (unlikely(features && btrfs_super_log_root(disk_super) && | |
3502 | !btrfs_test_opt(fs_info, NOLOGREPLAY))) { | |
3503 | btrfs_err(fs_info, | |
3504 | "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", | |
3505 | features); | |
3506 | err = -EINVAL; | |
3507 | goto fail_alloc; | |
3508 | } | |
3509 | ||
61d92c32 | 3510 | |
8481dd80 QW |
3511 | if (sectorsize < PAGE_SIZE) { |
3512 | struct btrfs_subpage_info *subpage_info; | |
3513 | ||
9f73f1ae QW |
3514 | /* |
3515 | * V1 space cache has some hardcoded PAGE_SIZE usage, and is | |
3516 | * going to be deprecated. | |
3517 | * | |
3518 | * Force to use v2 cache for subpage case. | |
3519 | */ | |
3520 | btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); | |
3521 | btrfs_set_and_info(fs_info, FREE_SPACE_TREE, | |
3522 | "forcing free space tree for sector size %u with page size %lu", | |
3523 | sectorsize, PAGE_SIZE); | |
3524 | ||
95ea0486 QW |
3525 | btrfs_warn(fs_info, |
3526 | "read-write for sector size %u with page size %lu is experimental", | |
3527 | sectorsize, PAGE_SIZE); | |
8481dd80 QW |
3528 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
3529 | if (!subpage_info) | |
3530 | goto fail_alloc; | |
3531 | btrfs_init_subpage_info(subpage_info, sectorsize); | |
3532 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3533 | } |
0bb3eb3e | 3534 | |
d21deec5 | 3535 | ret = btrfs_init_workqueues(fs_info); |
2a458198 ES |
3536 | if (ret) { |
3537 | err = ret; | |
0dc3b84a JB |
3538 | goto fail_sb_buffer; |
3539 | } | |
4543df7e | 3540 | |
9e11ceee JK |
3541 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3542 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3543 | |
a061fc8d CM |
3544 | sb->s_blocksize = sectorsize; |
3545 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3546 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3547 | |
925baedd | 3548 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3549 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3550 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3551 | if (ret) { |
05135f59 | 3552 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3553 | goto fail_sb_buffer; |
84eed90f | 3554 | } |
0b86a832 | 3555 | |
84234f3a | 3556 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3557 | level = btrfs_super_chunk_root_level(disk_super); |
bd676446 JB |
3558 | ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), |
3559 | generation, level); | |
3560 | if (ret) { | |
05135f59 | 3561 | btrfs_err(fs_info, "failed to read chunk root"); |
af31f5e5 | 3562 | goto fail_tree_roots; |
83121942 | 3563 | } |
0b86a832 | 3564 | |
e17cade2 | 3565 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3566 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3567 | BTRFS_UUID_SIZE); | |
e17cade2 | 3568 | |
5b4aacef | 3569 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3570 | if (ret) { |
05135f59 | 3571 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3572 | goto fail_tree_roots; |
2b82032c | 3573 | } |
0b86a832 | 3574 | |
8dabb742 | 3575 | /* |
bacce86a AJ |
3576 | * At this point we know all the devices that make this filesystem, |
3577 | * including the seed devices but we don't know yet if the replace | |
3578 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3579 | * filesystem but skip the replace target device which is checked |
bacce86a | 3580 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3581 | */ |
bacce86a | 3582 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3583 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3584 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3585 | goto fail_tree_roots; |
3586 | } | |
3587 | ||
b8522a1e | 3588 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3589 | if (ret) |
b8522a1e | 3590 | goto fail_tree_roots; |
8929ecfa | 3591 | |
73651042 NA |
3592 | /* |
3593 | * Get zone type information of zoned block devices. This will also | |
3594 | * handle emulation of a zoned filesystem if a regular device has the | |
3595 | * zoned incompat feature flag set. | |
3596 | */ | |
3597 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3598 | if (ret) { | |
3599 | btrfs_err(fs_info, | |
3600 | "zoned: failed to read device zone info: %d", | |
3601 | ret); | |
3602 | goto fail_block_groups; | |
3603 | } | |
3604 | ||
75ec1db8 JB |
3605 | /* |
3606 | * If we have a uuid root and we're not being told to rescan we need to | |
3607 | * check the generation here so we can set the | |
3608 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3609 | * transaction during a balance or the log replay without updating the | |
3610 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3611 | * even though it was perfectly fine. | |
3612 | */ | |
3613 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3614 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3615 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3616 | ||
cf90d884 QW |
3617 | ret = btrfs_verify_dev_extents(fs_info); |
3618 | if (ret) { | |
3619 | btrfs_err(fs_info, | |
3620 | "failed to verify dev extents against chunks: %d", | |
3621 | ret); | |
3622 | goto fail_block_groups; | |
3623 | } | |
68310a5e ID |
3624 | ret = btrfs_recover_balance(fs_info); |
3625 | if (ret) { | |
05135f59 | 3626 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3627 | goto fail_block_groups; |
3628 | } | |
3629 | ||
733f4fbb SB |
3630 | ret = btrfs_init_dev_stats(fs_info); |
3631 | if (ret) { | |
05135f59 | 3632 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3633 | goto fail_block_groups; |
3634 | } | |
3635 | ||
8dabb742 SB |
3636 | ret = btrfs_init_dev_replace(fs_info); |
3637 | if (ret) { | |
05135f59 | 3638 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3639 | goto fail_block_groups; |
3640 | } | |
3641 | ||
b70f5097 NA |
3642 | ret = btrfs_check_zoned_mode(fs_info); |
3643 | if (ret) { | |
3644 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3645 | ret); | |
3646 | goto fail_block_groups; | |
3647 | } | |
3648 | ||
c6761a9e | 3649 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3650 | if (ret) { |
05135f59 DS |
3651 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3652 | ret); | |
b7c35e81 AJ |
3653 | goto fail_block_groups; |
3654 | } | |
3655 | ||
96f3136e | 3656 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3657 | if (ret) { |
05135f59 | 3658 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3659 | goto fail_fsdev_sysfs; |
c59021f8 | 3660 | } |
3661 | ||
c59021f8 | 3662 | ret = btrfs_init_space_info(fs_info); |
3663 | if (ret) { | |
05135f59 | 3664 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3665 | goto fail_sysfs; |
c59021f8 | 3666 | } |
3667 | ||
5b4aacef | 3668 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3669 | if (ret) { |
05135f59 | 3670 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3671 | goto fail_sysfs; |
1b1d1f66 | 3672 | } |
4330e183 | 3673 | |
16beac87 NA |
3674 | btrfs_free_zone_cache(fs_info); |
3675 | ||
5c78a5e7 AJ |
3676 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3677 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3678 | btrfs_warn(fs_info, |
52042d8e | 3679 | "writable mount is not allowed due to too many missing devices"); |
2365dd3c | 3680 | goto fail_sysfs; |
292fd7fc | 3681 | } |
9078a3e1 | 3682 | |
33c44184 | 3683 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, |
a74a4b97 | 3684 | "btrfs-cleaner"); |
57506d50 | 3685 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3686 | goto fail_sysfs; |
a74a4b97 CM |
3687 | |
3688 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3689 | tree_root, | |
3690 | "btrfs-transaction"); | |
57506d50 | 3691 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3692 | goto fail_cleaner; |
a74a4b97 | 3693 | |
583b7231 | 3694 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3695 | !fs_info->fs_devices->rotating) { |
583b7231 | 3696 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3697 | } |
3698 | ||
572d9ab7 | 3699 | /* |
01327610 | 3700 | * Mount does not set all options immediately, we can do it now and do |
572d9ab7 DS |
3701 | * not have to wait for transaction commit |
3702 | */ | |
3703 | btrfs_apply_pending_changes(fs_info); | |
3818aea2 | 3704 | |
21adbd5c | 3705 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3706 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3707 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3708 | btrfs_test_opt(fs_info, |
cbeaae4f | 3709 | CHECK_INTEGRITY_DATA) ? 1 : 0, |
21adbd5c SB |
3710 | fs_info->check_integrity_print_mask); |
3711 | if (ret) | |
05135f59 DS |
3712 | btrfs_warn(fs_info, |
3713 | "failed to initialize integrity check module: %d", | |
3714 | ret); | |
21adbd5c SB |
3715 | } |
3716 | #endif | |
bcef60f2 AJ |
3717 | ret = btrfs_read_qgroup_config(fs_info); |
3718 | if (ret) | |
3719 | goto fail_trans_kthread; | |
21adbd5c | 3720 | |
fd708b81 JB |
3721 | if (btrfs_build_ref_tree(fs_info)) |
3722 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3723 | ||
96da0919 QW |
3724 | /* do not make disk changes in broken FS or nologreplay is given */ |
3725 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3726 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3727 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3728 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3729 | if (ret) { |
63443bf5 | 3730 | err = ret; |
28c16cbb | 3731 | goto fail_qgroup; |
79787eaa | 3732 | } |
e02119d5 | 3733 | } |
1a40e23b | 3734 | |
56e9357a | 3735 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 DC |
3736 | if (IS_ERR(fs_info->fs_root)) { |
3737 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3738 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
315bf8ef | 3739 | fs_info->fs_root = NULL; |
bcef60f2 | 3740 | goto fail_qgroup; |
3140c9a3 | 3741 | } |
c289811c | 3742 | |
bc98a42c | 3743 | if (sb_rdonly(sb)) |
8cd29088 | 3744 | goto clear_oneshot; |
59641015 | 3745 | |
44c0ca21 | 3746 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3747 | if (ret) { |
6bccf3ab | 3748 | close_ctree(fs_info); |
2b6ba629 | 3749 | return ret; |
e3acc2a6 | 3750 | } |
b0643e59 | 3751 | btrfs_discard_resume(fs_info); |
b382a324 | 3752 | |
44c0ca21 BB |
3753 | if (fs_info->uuid_root && |
3754 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3755 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3756 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3757 | ret = btrfs_check_uuid_tree(fs_info); |
3758 | if (ret) { | |
05135f59 DS |
3759 | btrfs_warn(fs_info, |
3760 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3761 | close_ctree(fs_info); |
70f80175 SB |
3762 | return ret; |
3763 | } | |
f7a81ea4 | 3764 | } |
94846229 | 3765 | |
afcdd129 | 3766 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3767 | |
b4be6aef JB |
3768 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3769 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3770 | wake_up_process(fs_info->cleaner_kthread); | |
3771 | ||
8cd29088 BB |
3772 | clear_oneshot: |
3773 | btrfs_clear_oneshot_options(fs_info); | |
ad2b2c80 | 3774 | return 0; |
39279cc3 | 3775 | |
bcef60f2 AJ |
3776 | fail_qgroup: |
3777 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3778 | fail_trans_kthread: |
3779 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3780 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3781 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3782 | fail_cleaner: |
a74a4b97 | 3783 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3784 | |
3785 | /* | |
3786 | * make sure we're done with the btree inode before we stop our | |
3787 | * kthreads | |
3788 | */ | |
3789 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3790 | |
2365dd3c | 3791 | fail_sysfs: |
6618a59b | 3792 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3793 | |
b7c35e81 AJ |
3794 | fail_fsdev_sysfs: |
3795 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3796 | ||
1b1d1f66 | 3797 | fail_block_groups: |
54067ae9 | 3798 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3799 | |
3800 | fail_tree_roots: | |
9e3aa805 JB |
3801 | if (fs_info->data_reloc_root) |
3802 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3803 | free_root_pointers(fs_info, true); |
2b8195bb | 3804 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3805 | |
39279cc3 | 3806 | fail_sb_buffer: |
7abadb64 | 3807 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3808 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3809 | fail_alloc: |
586e46e2 ID |
3810 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3811 | ||
4543df7e | 3812 | iput(fs_info->btree_inode); |
7e662854 | 3813 | fail: |
586e46e2 | 3814 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3815 | return err; |
eb60ceac | 3816 | } |
663faf9f | 3817 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3818 | |
314b6dd0 | 3819 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3820 | { |
314b6dd0 JT |
3821 | struct btrfs_device *device = bio->bi_private; |
3822 | struct bio_vec *bvec; | |
3823 | struct bvec_iter_all iter_all; | |
3824 | struct page *page; | |
3825 | ||
3826 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3827 | page = bvec->bv_page; | |
3828 | ||
3829 | if (bio->bi_status) { | |
3830 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3831 | "lost page write due to IO error on %s (%d)", | |
3832 | rcu_str_deref(device->name), | |
3833 | blk_status_to_errno(bio->bi_status)); | |
3834 | ClearPageUptodate(page); | |
3835 | SetPageError(page); | |
3836 | btrfs_dev_stat_inc_and_print(device, | |
3837 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3838 | } else { | |
3839 | SetPageUptodate(page); | |
3840 | } | |
3841 | ||
3842 | put_page(page); | |
3843 | unlock_page(page); | |
f2984462 | 3844 | } |
314b6dd0 JT |
3845 | |
3846 | bio_put(bio); | |
f2984462 CM |
3847 | } |
3848 | ||
8f32380d | 3849 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
a05d3c91 | 3850 | int copy_num, bool drop_cache) |
29c36d72 | 3851 | { |
29c36d72 | 3852 | struct btrfs_super_block *super; |
8f32380d | 3853 | struct page *page; |
12659251 | 3854 | u64 bytenr, bytenr_orig; |
8f32380d | 3855 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
3856 | int ret; |
3857 | ||
3858 | bytenr_orig = btrfs_sb_offset(copy_num); | |
3859 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
3860 | if (ret == -ENOENT) | |
3861 | return ERR_PTR(-EINVAL); | |
3862 | else if (ret) | |
3863 | return ERR_PTR(ret); | |
29c36d72 | 3864 | |
cda00eba | 3865 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) |
8f32380d | 3866 | return ERR_PTR(-EINVAL); |
29c36d72 | 3867 | |
a05d3c91 QW |
3868 | if (drop_cache) { |
3869 | /* This should only be called with the primary sb. */ | |
3870 | ASSERT(copy_num == 0); | |
3871 | ||
3872 | /* | |
3873 | * Drop the page of the primary superblock, so later read will | |
3874 | * always read from the device. | |
3875 | */ | |
3876 | invalidate_inode_pages2_range(mapping, | |
3877 | bytenr >> PAGE_SHIFT, | |
3878 | (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); | |
3879 | } | |
3880 | ||
8f32380d JT |
3881 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3882 | if (IS_ERR(page)) | |
3883 | return ERR_CAST(page); | |
29c36d72 | 3884 | |
8f32380d | 3885 | super = page_address(page); |
96c2e067 AJ |
3886 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
3887 | btrfs_release_disk_super(super); | |
3888 | return ERR_PTR(-ENODATA); | |
3889 | } | |
3890 | ||
12659251 | 3891 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
3892 | btrfs_release_disk_super(super); |
3893 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
3894 | } |
3895 | ||
8f32380d | 3896 | return super; |
29c36d72 AJ |
3897 | } |
3898 | ||
3899 | ||
8f32380d | 3900 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 3901 | { |
8f32380d | 3902 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
3903 | int i; |
3904 | u64 transid = 0; | |
a512bbf8 YZ |
3905 | |
3906 | /* we would like to check all the supers, but that would make | |
3907 | * a btrfs mount succeed after a mkfs from a different FS. | |
3908 | * So, we need to add a special mount option to scan for | |
3909 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3910 | */ | |
3911 | for (i = 0; i < 1; i++) { | |
a05d3c91 | 3912 | super = btrfs_read_dev_one_super(bdev, i, false); |
8f32380d | 3913 | if (IS_ERR(super)) |
a512bbf8 YZ |
3914 | continue; |
3915 | ||
a512bbf8 | 3916 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
3917 | if (latest) |
3918 | btrfs_release_disk_super(super); | |
3919 | ||
3920 | latest = super; | |
a512bbf8 | 3921 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
3922 | } |
3923 | } | |
92fc03fb | 3924 | |
8f32380d | 3925 | return super; |
a512bbf8 YZ |
3926 | } |
3927 | ||
4eedeb75 | 3928 | /* |
abbb3b8e | 3929 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 3930 | * pages we use for writing are locked. |
4eedeb75 | 3931 | * |
abbb3b8e DS |
3932 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3933 | * the expected device size at commit time. Note that max_mirrors must be | |
3934 | * same for write and wait phases. | |
4eedeb75 | 3935 | * |
314b6dd0 | 3936 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 3937 | */ |
a512bbf8 | 3938 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3939 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3940 | { |
d5178578 | 3941 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 3942 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 3943 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 3944 | int i; |
a512bbf8 | 3945 | int errors = 0; |
12659251 NA |
3946 | int ret; |
3947 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
3948 | |
3949 | if (max_mirrors == 0) | |
3950 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3951 | ||
d5178578 JT |
3952 | shash->tfm = fs_info->csum_shash; |
3953 | ||
a512bbf8 | 3954 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
3955 | struct page *page; |
3956 | struct bio *bio; | |
3957 | struct btrfs_super_block *disk_super; | |
3958 | ||
12659251 NA |
3959 | bytenr_orig = btrfs_sb_offset(i); |
3960 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
3961 | if (ret == -ENOENT) { | |
3962 | continue; | |
3963 | } else if (ret < 0) { | |
3964 | btrfs_err(device->fs_info, | |
3965 | "couldn't get super block location for mirror %d", | |
3966 | i); | |
3967 | errors++; | |
3968 | continue; | |
3969 | } | |
935e5cc9 MX |
3970 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3971 | device->commit_total_bytes) | |
a512bbf8 YZ |
3972 | break; |
3973 | ||
12659251 | 3974 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 3975 | |
fd08001f EB |
3976 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
3977 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
3978 | sb->csum); | |
4eedeb75 | 3979 | |
314b6dd0 JT |
3980 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
3981 | GFP_NOFS); | |
3982 | if (!page) { | |
abbb3b8e | 3983 | btrfs_err(device->fs_info, |
314b6dd0 | 3984 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
3985 | bytenr); |
3986 | errors++; | |
4eedeb75 | 3987 | continue; |
abbb3b8e | 3988 | } |
634554dc | 3989 | |
314b6dd0 JT |
3990 | /* Bump the refcount for wait_dev_supers() */ |
3991 | get_page(page); | |
a512bbf8 | 3992 | |
314b6dd0 JT |
3993 | disk_super = page_address(page); |
3994 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 3995 | |
314b6dd0 JT |
3996 | /* |
3997 | * Directly use bios here instead of relying on the page cache | |
3998 | * to do I/O, so we don't lose the ability to do integrity | |
3999 | * checking. | |
4000 | */ | |
07888c66 CH |
4001 | bio = bio_alloc(device->bdev, 1, |
4002 | REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, | |
4003 | GFP_NOFS); | |
314b6dd0 JT |
4004 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; |
4005 | bio->bi_private = device; | |
4006 | bio->bi_end_io = btrfs_end_super_write; | |
4007 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
4008 | offset_in_page(bytenr)); | |
a512bbf8 | 4009 | |
387125fc | 4010 | /* |
314b6dd0 JT |
4011 | * We FUA only the first super block. The others we allow to |
4012 | * go down lazy and there's a short window where the on-disk | |
4013 | * copies might still contain the older version. | |
387125fc | 4014 | */ |
1b9e619c | 4015 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 JT |
4016 | bio->bi_opf |= REQ_FUA; |
4017 | ||
58ff51f1 CH |
4018 | btrfsic_check_bio(bio); |
4019 | submit_bio(bio); | |
8376d9e1 NA |
4020 | |
4021 | if (btrfs_advance_sb_log(device, i)) | |
4022 | errors++; | |
a512bbf8 YZ |
4023 | } |
4024 | return errors < i ? 0 : -1; | |
4025 | } | |
4026 | ||
abbb3b8e DS |
4027 | /* |
4028 | * Wait for write completion of superblocks done by write_dev_supers, | |
4029 | * @max_mirrors same for write and wait phases. | |
4030 | * | |
314b6dd0 | 4031 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
4032 | * date. |
4033 | */ | |
4034 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
4035 | { | |
abbb3b8e DS |
4036 | int i; |
4037 | int errors = 0; | |
b6a535fa | 4038 | bool primary_failed = false; |
12659251 | 4039 | int ret; |
abbb3b8e DS |
4040 | u64 bytenr; |
4041 | ||
4042 | if (max_mirrors == 0) | |
4043 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
4044 | ||
4045 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
4046 | struct page *page; |
4047 | ||
12659251 NA |
4048 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
4049 | if (ret == -ENOENT) { | |
4050 | break; | |
4051 | } else if (ret < 0) { | |
4052 | errors++; | |
4053 | if (i == 0) | |
4054 | primary_failed = true; | |
4055 | continue; | |
4056 | } | |
abbb3b8e DS |
4057 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4058 | device->commit_total_bytes) | |
4059 | break; | |
4060 | ||
314b6dd0 JT |
4061 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
4062 | bytenr >> PAGE_SHIFT); | |
4063 | if (!page) { | |
abbb3b8e | 4064 | errors++; |
b6a535fa HM |
4065 | if (i == 0) |
4066 | primary_failed = true; | |
abbb3b8e DS |
4067 | continue; |
4068 | } | |
314b6dd0 JT |
4069 | /* Page is submitted locked and unlocked once the IO completes */ |
4070 | wait_on_page_locked(page); | |
4071 | if (PageError(page)) { | |
abbb3b8e | 4072 | errors++; |
b6a535fa HM |
4073 | if (i == 0) |
4074 | primary_failed = true; | |
4075 | } | |
abbb3b8e | 4076 | |
314b6dd0 JT |
4077 | /* Drop our reference */ |
4078 | put_page(page); | |
abbb3b8e | 4079 | |
314b6dd0 JT |
4080 | /* Drop the reference from the writing run */ |
4081 | put_page(page); | |
abbb3b8e DS |
4082 | } |
4083 | ||
b6a535fa HM |
4084 | /* log error, force error return */ |
4085 | if (primary_failed) { | |
4086 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
4087 | device->devid); | |
4088 | return -1; | |
4089 | } | |
4090 | ||
abbb3b8e DS |
4091 | return errors < i ? 0 : -1; |
4092 | } | |
4093 | ||
387125fc CM |
4094 | /* |
4095 | * endio for the write_dev_flush, this will wake anyone waiting | |
4096 | * for the barrier when it is done | |
4097 | */ | |
4246a0b6 | 4098 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 4099 | { |
f9e69aa9 | 4100 | bio_uninit(bio); |
e0ae9994 | 4101 | complete(bio->bi_private); |
387125fc CM |
4102 | } |
4103 | ||
4104 | /* | |
4fc6441a AJ |
4105 | * Submit a flush request to the device if it supports it. Error handling is |
4106 | * done in the waiting counterpart. | |
387125fc | 4107 | */ |
4fc6441a | 4108 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 4109 | { |
f9e69aa9 | 4110 | struct bio *bio = &device->flush_bio; |
387125fc | 4111 | |
a91cf0ff WY |
4112 | #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4113 | /* | |
4114 | * When a disk has write caching disabled, we skip submission of a bio | |
4115 | * with flush and sync requests before writing the superblock, since | |
4116 | * it's not needed. However when the integrity checker is enabled, this | |
4117 | * results in reports that there are metadata blocks referred by a | |
4118 | * superblock that were not properly flushed. So don't skip the bio | |
4119 | * submission only when the integrity checker is enabled for the sake | |
4120 | * of simplicity, since this is a debug tool and not meant for use in | |
4121 | * non-debug builds. | |
4122 | */ | |
08e688fd | 4123 | if (!bdev_write_cache(device->bdev)) |
4fc6441a | 4124 | return; |
a91cf0ff | 4125 | #endif |
387125fc | 4126 | |
f9e69aa9 CH |
4127 | bio_init(bio, device->bdev, NULL, 0, |
4128 | REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); | |
387125fc | 4129 | bio->bi_end_io = btrfs_end_empty_barrier; |
387125fc CM |
4130 | init_completion(&device->flush_wait); |
4131 | bio->bi_private = &device->flush_wait; | |
387125fc | 4132 | |
58ff51f1 CH |
4133 | btrfsic_check_bio(bio); |
4134 | submit_bio(bio); | |
1c3063b6 | 4135 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 4136 | } |
387125fc | 4137 | |
4fc6441a AJ |
4138 | /* |
4139 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
4140 | */ | |
8c27cb35 | 4141 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 4142 | { |
f9e69aa9 | 4143 | struct bio *bio = &device->flush_bio; |
387125fc | 4144 | |
1c3063b6 | 4145 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 4146 | return BLK_STS_OK; |
387125fc | 4147 | |
1c3063b6 | 4148 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 4149 | wait_for_completion_io(&device->flush_wait); |
387125fc | 4150 | |
8c27cb35 | 4151 | return bio->bi_status; |
387125fc | 4152 | } |
387125fc | 4153 | |
d10b82fe | 4154 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 4155 | { |
6528b99d | 4156 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 4157 | return -EIO; |
387125fc CM |
4158 | return 0; |
4159 | } | |
4160 | ||
4161 | /* | |
4162 | * send an empty flush down to each device in parallel, | |
4163 | * then wait for them | |
4164 | */ | |
4165 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
4166 | { | |
4167 | struct list_head *head; | |
4168 | struct btrfs_device *dev; | |
5af3e8cc | 4169 | int errors_wait = 0; |
4e4cbee9 | 4170 | blk_status_t ret; |
387125fc | 4171 | |
1538e6c5 | 4172 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
4173 | /* send down all the barriers */ |
4174 | head = &info->fs_devices->devices; | |
1538e6c5 | 4175 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4176 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4177 | continue; |
cea7c8bf | 4178 | if (!dev->bdev) |
387125fc | 4179 | continue; |
e12c9621 | 4180 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4181 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4182 | continue; |
4183 | ||
4fc6441a | 4184 | write_dev_flush(dev); |
58efbc9f | 4185 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
4186 | } |
4187 | ||
4188 | /* wait for all the barriers */ | |
1538e6c5 | 4189 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4190 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4191 | continue; |
387125fc | 4192 | if (!dev->bdev) { |
5af3e8cc | 4193 | errors_wait++; |
387125fc CM |
4194 | continue; |
4195 | } | |
e12c9621 | 4196 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4197 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4198 | continue; |
4199 | ||
4fc6441a | 4200 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
4201 | if (ret) { |
4202 | dev->last_flush_error = ret; | |
66b4993e DS |
4203 | btrfs_dev_stat_inc_and_print(dev, |
4204 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 4205 | errors_wait++; |
401b41e5 AJ |
4206 | } |
4207 | } | |
4208 | ||
cea7c8bf | 4209 | if (errors_wait) { |
401b41e5 AJ |
4210 | /* |
4211 | * At some point we need the status of all disks | |
4212 | * to arrive at the volume status. So error checking | |
4213 | * is being pushed to a separate loop. | |
4214 | */ | |
d10b82fe | 4215 | return check_barrier_error(info); |
387125fc | 4216 | } |
387125fc CM |
4217 | return 0; |
4218 | } | |
4219 | ||
943c6e99 ZL |
4220 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
4221 | { | |
8789f4fe ZL |
4222 | int raid_type; |
4223 | int min_tolerated = INT_MAX; | |
943c6e99 | 4224 | |
8789f4fe ZL |
4225 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
4226 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 4227 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4228 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4229 | tolerated_failures); | |
943c6e99 | 4230 | |
8789f4fe ZL |
4231 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4232 | if (raid_type == BTRFS_RAID_SINGLE) | |
4233 | continue; | |
41a6e891 | 4234 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4235 | continue; |
8c3e3582 | 4236 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4237 | btrfs_raid_array[raid_type]. |
4238 | tolerated_failures); | |
4239 | } | |
943c6e99 | 4240 | |
8789f4fe | 4241 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4242 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4243 | min_tolerated = 0; |
4244 | } | |
4245 | ||
4246 | return min_tolerated; | |
943c6e99 ZL |
4247 | } |
4248 | ||
eece6a9c | 4249 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4250 | { |
e5e9a520 | 4251 | struct list_head *head; |
f2984462 | 4252 | struct btrfs_device *dev; |
a061fc8d | 4253 | struct btrfs_super_block *sb; |
f2984462 | 4254 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4255 | int ret; |
4256 | int do_barriers; | |
a236aed1 CM |
4257 | int max_errors; |
4258 | int total_errors = 0; | |
a061fc8d | 4259 | u64 flags; |
f2984462 | 4260 | |
0b246afa | 4261 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4262 | |
4263 | /* | |
4264 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4265 | * not from fsync where the tree roots in fs_info have not | |
4266 | * been consistent on disk. | |
4267 | */ | |
4268 | if (max_mirrors == 0) | |
4269 | backup_super_roots(fs_info); | |
f2984462 | 4270 | |
0b246afa | 4271 | sb = fs_info->super_for_commit; |
a061fc8d | 4272 | dev_item = &sb->dev_item; |
e5e9a520 | 4273 | |
0b246afa JM |
4274 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4275 | head = &fs_info->fs_devices->devices; | |
4276 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4277 | |
5af3e8cc | 4278 | if (do_barriers) { |
0b246afa | 4279 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4280 | if (ret) { |
4281 | mutex_unlock( | |
0b246afa JM |
4282 | &fs_info->fs_devices->device_list_mutex); |
4283 | btrfs_handle_fs_error(fs_info, ret, | |
4284 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4285 | return ret; |
4286 | } | |
4287 | } | |
387125fc | 4288 | |
1538e6c5 | 4289 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4290 | if (!dev->bdev) { |
4291 | total_errors++; | |
4292 | continue; | |
4293 | } | |
e12c9621 | 4294 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4295 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4296 | continue; |
4297 | ||
2b82032c | 4298 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4299 | btrfs_set_stack_device_type(dev_item, dev->type); |
4300 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4301 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4302 | dev->commit_total_bytes); |
ce7213c7 MX |
4303 | btrfs_set_stack_device_bytes_used(dev_item, |
4304 | dev->commit_bytes_used); | |
a061fc8d CM |
4305 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4306 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4307 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4308 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4309 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4310 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4311 | |
a061fc8d CM |
4312 | flags = btrfs_super_flags(sb); |
4313 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4314 | ||
75cb857d QW |
4315 | ret = btrfs_validate_write_super(fs_info, sb); |
4316 | if (ret < 0) { | |
4317 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4318 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4319 | "unexpected superblock corruption detected"); | |
4320 | return -EUCLEAN; | |
4321 | } | |
4322 | ||
abbb3b8e | 4323 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4324 | if (ret) |
4325 | total_errors++; | |
f2984462 | 4326 | } |
a236aed1 | 4327 | if (total_errors > max_errors) { |
0b246afa JM |
4328 | btrfs_err(fs_info, "%d errors while writing supers", |
4329 | total_errors); | |
4330 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4331 | |
9d565ba4 | 4332 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4333 | btrfs_handle_fs_error(fs_info, -EIO, |
4334 | "%d errors while writing supers", | |
4335 | total_errors); | |
9d565ba4 | 4336 | return -EIO; |
a236aed1 | 4337 | } |
f2984462 | 4338 | |
a512bbf8 | 4339 | total_errors = 0; |
1538e6c5 | 4340 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4341 | if (!dev->bdev) |
4342 | continue; | |
e12c9621 | 4343 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4344 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4345 | continue; |
4346 | ||
abbb3b8e | 4347 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4348 | if (ret) |
4349 | total_errors++; | |
f2984462 | 4350 | } |
0b246afa | 4351 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4352 | if (total_errors > max_errors) { |
0b246afa JM |
4353 | btrfs_handle_fs_error(fs_info, -EIO, |
4354 | "%d errors while writing supers", | |
4355 | total_errors); | |
79787eaa | 4356 | return -EIO; |
a236aed1 | 4357 | } |
f2984462 CM |
4358 | return 0; |
4359 | } | |
4360 | ||
cb517eab MX |
4361 | /* Drop a fs root from the radix tree and free it. */ |
4362 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4363 | struct btrfs_root *root) | |
2619ba1f | 4364 | { |
4785e24f JB |
4365 | bool drop_ref = false; |
4366 | ||
fc7cbcd4 DS |
4367 | spin_lock(&fs_info->fs_roots_radix_lock); |
4368 | radix_tree_delete(&fs_info->fs_roots_radix, | |
4369 | (unsigned long)root->root_key.objectid); | |
4370 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) | |
4785e24f | 4371 | drop_ref = true; |
fc7cbcd4 | 4372 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4373 | |
84961539 | 4374 | if (BTRFS_FS_ERROR(fs_info)) { |
ef67963d | 4375 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4376 | if (root->reloc_root) { |
00246528 | 4377 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4378 | root->reloc_root = NULL; |
4379 | } | |
4380 | } | |
3321719e | 4381 | |
4785e24f JB |
4382 | if (drop_ref) |
4383 | btrfs_put_root(root); | |
2619ba1f CM |
4384 | } |
4385 | ||
c146afad | 4386 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 4387 | { |
fc7cbcd4 DS |
4388 | u64 root_objectid = 0; |
4389 | struct btrfs_root *gang[8]; | |
4390 | int i = 0; | |
65d33fd7 | 4391 | int err = 0; |
fc7cbcd4 | 4392 | unsigned int ret = 0; |
e089f05c | 4393 | |
c146afad | 4394 | while (1) { |
fc7cbcd4 DS |
4395 | spin_lock(&fs_info->fs_roots_radix_lock); |
4396 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4397 | (void **)gang, root_objectid, | |
4398 | ARRAY_SIZE(gang)); | |
4399 | if (!ret) { | |
4400 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4401 | break; | |
65d33fd7 | 4402 | } |
fc7cbcd4 | 4403 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 4404 | |
fc7cbcd4 DS |
4405 | for (i = 0; i < ret; i++) { |
4406 | /* Avoid to grab roots in dead_roots */ | |
4407 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
4408 | gang[i] = NULL; | |
4409 | continue; | |
4410 | } | |
4411 | /* grab all the search result for later use */ | |
4412 | gang[i] = btrfs_grab_root(gang[i]); | |
65d33fd7 | 4413 | } |
fc7cbcd4 | 4414 | spin_unlock(&fs_info->fs_roots_radix_lock); |
66b4ffd1 | 4415 | |
fc7cbcd4 DS |
4416 | for (i = 0; i < ret; i++) { |
4417 | if (!gang[i]) | |
65d33fd7 | 4418 | continue; |
fc7cbcd4 DS |
4419 | root_objectid = gang[i]->root_key.objectid; |
4420 | err = btrfs_orphan_cleanup(gang[i]); | |
66b4ffd1 | 4421 | if (err) |
fc7cbcd4 DS |
4422 | break; |
4423 | btrfs_put_root(gang[i]); | |
c146afad | 4424 | } |
fc7cbcd4 | 4425 | root_objectid++; |
c146afad | 4426 | } |
65d33fd7 | 4427 | |
fc7cbcd4 DS |
4428 | /* release the uncleaned roots due to error */ |
4429 | for (; i < ret; i++) { | |
4430 | if (gang[i]) | |
4431 | btrfs_put_root(gang[i]); | |
65d33fd7 QW |
4432 | } |
4433 | return err; | |
c146afad | 4434 | } |
a2135011 | 4435 | |
6bccf3ab | 4436 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4437 | { |
6bccf3ab | 4438 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4439 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4440 | |
0b246afa | 4441 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4442 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4443 | mutex_unlock(&fs_info->cleaner_mutex); |
4444 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4445 | |
4446 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4447 | down_write(&fs_info->cleanup_work_sem); |
4448 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4449 | |
7a7eaa40 | 4450 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4451 | if (IS_ERR(trans)) |
4452 | return PTR_ERR(trans); | |
3a45bb20 | 4453 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4454 | } |
4455 | ||
36c86a9e QW |
4456 | static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) |
4457 | { | |
4458 | struct btrfs_transaction *trans; | |
4459 | struct btrfs_transaction *tmp; | |
4460 | bool found = false; | |
4461 | ||
4462 | if (list_empty(&fs_info->trans_list)) | |
4463 | return; | |
4464 | ||
4465 | /* | |
4466 | * This function is only called at the very end of close_ctree(), | |
4467 | * thus no other running transaction, no need to take trans_lock. | |
4468 | */ | |
4469 | ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); | |
4470 | list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { | |
4471 | struct extent_state *cached = NULL; | |
4472 | u64 dirty_bytes = 0; | |
4473 | u64 cur = 0; | |
4474 | u64 found_start; | |
4475 | u64 found_end; | |
4476 | ||
4477 | found = true; | |
4478 | while (!find_first_extent_bit(&trans->dirty_pages, cur, | |
4479 | &found_start, &found_end, EXTENT_DIRTY, &cached)) { | |
4480 | dirty_bytes += found_end + 1 - found_start; | |
4481 | cur = found_end + 1; | |
4482 | } | |
4483 | btrfs_warn(fs_info, | |
4484 | "transaction %llu (with %llu dirty metadata bytes) is not committed", | |
4485 | trans->transid, dirty_bytes); | |
4486 | btrfs_cleanup_one_transaction(trans, fs_info); | |
4487 | ||
4488 | if (trans == fs_info->running_transaction) | |
4489 | fs_info->running_transaction = NULL; | |
4490 | list_del_init(&trans->list); | |
4491 | ||
4492 | btrfs_put_transaction(trans); | |
4493 | trace_btrfs_transaction_commit(fs_info); | |
4494 | } | |
4495 | ASSERT(!found); | |
4496 | } | |
4497 | ||
b105e927 | 4498 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4499 | { |
c146afad YZ |
4500 | int ret; |
4501 | ||
afcdd129 | 4502 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
31e70e52 | 4503 | |
8a1f1e3d FM |
4504 | /* |
4505 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4506 | * clear that bit and wake up relocation so it can stop. | |
4507 | * We must do this before stopping the block group reclaim task, because | |
4508 | * at btrfs_relocate_block_group() we wait for this bit, and after the | |
4509 | * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we | |
4510 | * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will | |
4511 | * return 1. | |
4512 | */ | |
4513 | btrfs_wake_unfinished_drop(fs_info); | |
4514 | ||
31e70e52 FM |
4515 | /* |
4516 | * We may have the reclaim task running and relocating a data block group, | |
4517 | * in which case it may create delayed iputs. So stop it before we park | |
4518 | * the cleaner kthread otherwise we can get new delayed iputs after | |
4519 | * parking the cleaner, and that can make the async reclaim task to hang | |
4520 | * if it's waiting for delayed iputs to complete, since the cleaner is | |
4521 | * parked and can not run delayed iputs - this will make us hang when | |
4522 | * trying to stop the async reclaim task. | |
4523 | */ | |
4524 | cancel_work_sync(&fs_info->reclaim_bgs_work); | |
d6fd0ae2 OS |
4525 | /* |
4526 | * We don't want the cleaner to start new transactions, add more delayed | |
4527 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4528 | * because that frees the task_struct, and the transaction kthread might | |
4529 | * still try to wake up the cleaner. | |
4530 | */ | |
4531 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4532 | |
7343dd61 | 4533 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4534 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4535 | |
803b2f54 SB |
4536 | /* wait for the uuid_scan task to finish */ |
4537 | down(&fs_info->uuid_tree_rescan_sem); | |
4538 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4539 | up(&fs_info->uuid_tree_rescan_sem); | |
4540 | ||
837d5b6e | 4541 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4542 | btrfs_pause_balance(fs_info); |
837d5b6e | 4543 | |
8dabb742 SB |
4544 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4545 | ||
aa1b8cd4 | 4546 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4547 | |
4548 | /* wait for any defraggers to finish */ | |
4549 | wait_event(fs_info->transaction_wait, | |
4550 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4551 | ||
4552 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4553 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4554 | |
a362bb86 FM |
4555 | /* |
4556 | * After we parked the cleaner kthread, ordered extents may have | |
4557 | * completed and created new delayed iputs. If one of the async reclaim | |
4558 | * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we | |
4559 | * can hang forever trying to stop it, because if a delayed iput is | |
4560 | * added after it ran btrfs_run_delayed_iputs() and before it called | |
4561 | * btrfs_wait_on_delayed_iputs(), it will hang forever since there is | |
4562 | * no one else to run iputs. | |
4563 | * | |
4564 | * So wait for all ongoing ordered extents to complete and then run | |
4565 | * delayed iputs. This works because once we reach this point no one | |
4566 | * can either create new ordered extents nor create delayed iputs | |
4567 | * through some other means. | |
4568 | * | |
4569 | * Also note that btrfs_wait_ordered_roots() is not safe here, because | |
4570 | * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, | |
4571 | * but the delayed iput for the respective inode is made only when doing | |
4572 | * the final btrfs_put_ordered_extent() (which must happen at | |
4573 | * btrfs_finish_ordered_io() when we are unmounting). | |
4574 | */ | |
4575 | btrfs_flush_workqueue(fs_info->endio_write_workers); | |
4576 | /* Ordered extents for free space inodes. */ | |
4577 | btrfs_flush_workqueue(fs_info->endio_freespace_worker); | |
4578 | btrfs_run_delayed_iputs(fs_info); | |
4579 | ||
21c7e756 | 4580 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4581 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4582 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4583 | |
b0643e59 DZ |
4584 | /* Cancel or finish ongoing discard work */ |
4585 | btrfs_discard_cleanup(fs_info); | |
4586 | ||
bc98a42c | 4587 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4588 | /* |
d6fd0ae2 OS |
4589 | * The cleaner kthread is stopped, so do one final pass over |
4590 | * unused block groups. | |
e44163e1 | 4591 | */ |
0b246afa | 4592 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4593 | |
f0cc2cd7 FM |
4594 | /* |
4595 | * There might be existing delayed inode workers still running | |
4596 | * and holding an empty delayed inode item. We must wait for | |
4597 | * them to complete first because they can create a transaction. | |
4598 | * This happens when someone calls btrfs_balance_delayed_items() | |
4599 | * and then a transaction commit runs the same delayed nodes | |
4600 | * before any delayed worker has done something with the nodes. | |
4601 | * We must wait for any worker here and not at transaction | |
4602 | * commit time since that could cause a deadlock. | |
4603 | * This is a very rare case. | |
4604 | */ | |
4605 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4606 | ||
6bccf3ab | 4607 | ret = btrfs_commit_super(fs_info); |
acce952b | 4608 | if (ret) |
04892340 | 4609 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4610 | } |
4611 | ||
84961539 | 4612 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 4613 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4614 | |
e3029d9f AV |
4615 | kthread_stop(fs_info->transaction_kthread); |
4616 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4617 | |
e187831e | 4618 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4619 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4620 | |
5958253c QW |
4621 | if (btrfs_check_quota_leak(fs_info)) { |
4622 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4623 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4624 | } | |
4625 | ||
04892340 | 4626 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4627 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4628 | |
963d678b | 4629 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4630 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4631 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4632 | } |
bcc63abb | 4633 | |
5deb17e1 | 4634 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4635 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4636 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4637 | |
6618a59b | 4638 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4639 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4640 | |
1a4319cc LB |
4641 | btrfs_put_block_group_cache(fs_info); |
4642 | ||
de348ee0 WS |
4643 | /* |
4644 | * we must make sure there is not any read request to | |
4645 | * submit after we stopping all workers. | |
4646 | */ | |
4647 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4648 | btrfs_stop_all_workers(fs_info); |
4649 | ||
0a31daa4 | 4650 | /* We shouldn't have any transaction open at this point */ |
36c86a9e | 4651 | warn_about_uncommitted_trans(fs_info); |
0a31daa4 | 4652 | |
afcdd129 | 4653 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4654 | free_root_pointers(fs_info, true); |
8c38938c | 4655 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4656 | |
4e19443d JB |
4657 | /* |
4658 | * We must free the block groups after dropping the fs_roots as we could | |
4659 | * have had an IO error and have left over tree log blocks that aren't | |
4660 | * cleaned up until the fs roots are freed. This makes the block group | |
4661 | * accounting appear to be wrong because there's pending reserved bytes, | |
4662 | * so make sure we do the block group cleanup afterwards. | |
4663 | */ | |
4664 | btrfs_free_block_groups(fs_info); | |
4665 | ||
13e6c37b | 4666 | iput(fs_info->btree_inode); |
d6bfde87 | 4667 | |
21adbd5c | 4668 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4669 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4670 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4671 | #endif |
4672 | ||
0b86a832 | 4673 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
68c94e55 | 4674 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4675 | } |
4676 | ||
b9fab919 CM |
4677 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4678 | int atomic) | |
5f39d397 | 4679 | { |
1259ab75 | 4680 | int ret; |
727011e0 | 4681 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4682 | |
0b32f4bb | 4683 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4684 | if (!ret) |
4685 | return ret; | |
4686 | ||
4687 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4688 | parent_transid, atomic); |
4689 | if (ret == -EAGAIN) | |
4690 | return ret; | |
1259ab75 | 4691 | return !ret; |
5f39d397 CM |
4692 | } |
4693 | ||
5f39d397 CM |
4694 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4695 | { | |
2f4d60df | 4696 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4697 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4698 | int was_dirty; |
b4ce94de | 4699 | |
06ea65a3 JB |
4700 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4701 | /* | |
4702 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4703 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4704 | * outside of the sanity tests. |
4705 | */ | |
b0132a3b | 4706 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4707 | return; |
4708 | #endif | |
49d0c642 | 4709 | btrfs_assert_tree_write_locked(buf); |
0b246afa | 4710 | if (transid != fs_info->generation) |
5d163e0e | 4711 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4712 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4713 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4714 | if (!was_dirty) |
104b4e51 NB |
4715 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4716 | buf->len, | |
4717 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4718 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4719 | /* |
4720 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4721 | * but item data not updated. | |
4722 | * So here we should only check item pointers, not item data. | |
4723 | */ | |
4724 | if (btrfs_header_level(buf) == 0 && | |
cfdaad5e | 4725 | btrfs_check_leaf_relaxed(buf)) { |
a4f78750 | 4726 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4727 | ASSERT(0); |
4728 | } | |
4729 | #endif | |
eb60ceac CM |
4730 | } |
4731 | ||
2ff7e61e | 4732 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4733 | int flush_delayed) |
16cdcec7 MX |
4734 | { |
4735 | /* | |
4736 | * looks as though older kernels can get into trouble with | |
4737 | * this code, they end up stuck in balance_dirty_pages forever | |
4738 | */ | |
e2d84521 | 4739 | int ret; |
16cdcec7 MX |
4740 | |
4741 | if (current->flags & PF_MEMALLOC) | |
4742 | return; | |
4743 | ||
b53d3f5d | 4744 | if (flush_delayed) |
2ff7e61e | 4745 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4746 | |
d814a491 EL |
4747 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4748 | BTRFS_DIRTY_METADATA_THRESH, | |
4749 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4750 | if (ret > 0) { |
0b246afa | 4751 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4752 | } |
16cdcec7 MX |
4753 | } |
4754 | ||
2ff7e61e | 4755 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4756 | { |
2ff7e61e | 4757 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4758 | } |
585ad2c3 | 4759 | |
2ff7e61e | 4760 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4761 | { |
2ff7e61e | 4762 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4763 | } |
6b80053d | 4764 | |
2ff7e61e | 4765 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4766 | { |
fe816d0f NB |
4767 | /* cleanup FS via transaction */ |
4768 | btrfs_cleanup_transaction(fs_info); | |
4769 | ||
0b246afa | 4770 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4771 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4772 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4773 | |
0b246afa JM |
4774 | down_write(&fs_info->cleanup_work_sem); |
4775 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4776 | } |
4777 | ||
ef67963d JB |
4778 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4779 | { | |
fc7cbcd4 DS |
4780 | struct btrfs_root *gang[8]; |
4781 | u64 root_objectid = 0; | |
4782 | int ret; | |
4783 | ||
4784 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4785 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4786 | (void **)gang, root_objectid, | |
4787 | ARRAY_SIZE(gang))) != 0) { | |
4788 | int i; | |
4789 | ||
4790 | for (i = 0; i < ret; i++) | |
4791 | gang[i] = btrfs_grab_root(gang[i]); | |
4792 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4793 | ||
4794 | for (i = 0; i < ret; i++) { | |
4795 | if (!gang[i]) | |
ef67963d | 4796 | continue; |
fc7cbcd4 DS |
4797 | root_objectid = gang[i]->root_key.objectid; |
4798 | btrfs_free_log(NULL, gang[i]); | |
4799 | btrfs_put_root(gang[i]); | |
ef67963d | 4800 | } |
fc7cbcd4 DS |
4801 | root_objectid++; |
4802 | spin_lock(&fs_info->fs_roots_radix_lock); | |
ef67963d | 4803 | } |
fc7cbcd4 | 4804 | spin_unlock(&fs_info->fs_roots_radix_lock); |
ef67963d JB |
4805 | btrfs_free_log_root_tree(NULL, fs_info); |
4806 | } | |
4807 | ||
143bede5 | 4808 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4809 | { |
acce952b | 4810 | struct btrfs_ordered_extent *ordered; |
acce952b | 4811 | |
199c2a9c | 4812 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4813 | /* |
4814 | * This will just short circuit the ordered completion stuff which will | |
4815 | * make sure the ordered extent gets properly cleaned up. | |
4816 | */ | |
199c2a9c | 4817 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4818 | root_extent_list) |
4819 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4820 | spin_unlock(&root->ordered_extent_lock); |
4821 | } | |
4822 | ||
4823 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4824 | { | |
4825 | struct btrfs_root *root; | |
4826 | struct list_head splice; | |
4827 | ||
4828 | INIT_LIST_HEAD(&splice); | |
4829 | ||
4830 | spin_lock(&fs_info->ordered_root_lock); | |
4831 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4832 | while (!list_empty(&splice)) { | |
4833 | root = list_first_entry(&splice, struct btrfs_root, | |
4834 | ordered_root); | |
1de2cfde JB |
4835 | list_move_tail(&root->ordered_root, |
4836 | &fs_info->ordered_roots); | |
199c2a9c | 4837 | |
2a85d9ca | 4838 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4839 | btrfs_destroy_ordered_extents(root); |
4840 | ||
2a85d9ca LB |
4841 | cond_resched(); |
4842 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4843 | } |
4844 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4845 | |
4846 | /* | |
4847 | * We need this here because if we've been flipped read-only we won't | |
4848 | * get sync() from the umount, so we need to make sure any ordered | |
4849 | * extents that haven't had their dirty pages IO start writeout yet | |
4850 | * actually get run and error out properly. | |
4851 | */ | |
4852 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4853 | } |
4854 | ||
35a3621b | 4855 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4856 | struct btrfs_fs_info *fs_info) |
acce952b | 4857 | { |
4858 | struct rb_node *node; | |
4859 | struct btrfs_delayed_ref_root *delayed_refs; | |
4860 | struct btrfs_delayed_ref_node *ref; | |
4861 | int ret = 0; | |
4862 | ||
4863 | delayed_refs = &trans->delayed_refs; | |
4864 | ||
4865 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4866 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4867 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4868 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
acce952b | 4869 | return ret; |
4870 | } | |
4871 | ||
5c9d028b | 4872 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4873 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4874 | struct rb_node *n; |
e78417d1 | 4875 | bool pin_bytes = false; |
acce952b | 4876 | |
d7df2c79 JB |
4877 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4878 | href_node); | |
3069bd26 | 4879 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4880 | continue; |
3069bd26 | 4881 | |
d7df2c79 | 4882 | spin_lock(&head->lock); |
e3d03965 | 4883 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4884 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4885 | ref_node); | |
d7df2c79 | 4886 | ref->in_tree = 0; |
e3d03965 | 4887 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4888 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4889 | if (!list_empty(&ref->add_list)) |
4890 | list_del(&ref->add_list); | |
d7df2c79 JB |
4891 | atomic_dec(&delayed_refs->num_entries); |
4892 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 4893 | } |
d7df2c79 JB |
4894 | if (head->must_insert_reserved) |
4895 | pin_bytes = true; | |
4896 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4897 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4898 | spin_unlock(&head->lock); |
4899 | spin_unlock(&delayed_refs->lock); | |
4900 | mutex_unlock(&head->mutex); | |
acce952b | 4901 | |
f603bb94 NB |
4902 | if (pin_bytes) { |
4903 | struct btrfs_block_group *cache; | |
4904 | ||
4905 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
4906 | BUG_ON(!cache); | |
4907 | ||
4908 | spin_lock(&cache->space_info->lock); | |
4909 | spin_lock(&cache->lock); | |
4910 | cache->pinned += head->num_bytes; | |
4911 | btrfs_space_info_update_bytes_pinned(fs_info, | |
4912 | cache->space_info, head->num_bytes); | |
4913 | cache->reserved -= head->num_bytes; | |
4914 | cache->space_info->bytes_reserved -= head->num_bytes; | |
4915 | spin_unlock(&cache->lock); | |
4916 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
4917 | |
4918 | btrfs_put_block_group(cache); | |
4919 | ||
4920 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
4921 | head->bytenr + head->num_bytes - 1); | |
4922 | } | |
31890da0 | 4923 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4924 | btrfs_put_delayed_ref_head(head); |
acce952b | 4925 | cond_resched(); |
4926 | spin_lock(&delayed_refs->lock); | |
4927 | } | |
81f7eb00 | 4928 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4929 | |
4930 | spin_unlock(&delayed_refs->lock); | |
4931 | ||
4932 | return ret; | |
4933 | } | |
4934 | ||
143bede5 | 4935 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4936 | { |
4937 | struct btrfs_inode *btrfs_inode; | |
4938 | struct list_head splice; | |
4939 | ||
4940 | INIT_LIST_HEAD(&splice); | |
4941 | ||
eb73c1b7 MX |
4942 | spin_lock(&root->delalloc_lock); |
4943 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4944 | |
4945 | while (!list_empty(&splice)) { | |
fe816d0f | 4946 | struct inode *inode = NULL; |
eb73c1b7 MX |
4947 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4948 | delalloc_inodes); | |
fe816d0f | 4949 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4950 | spin_unlock(&root->delalloc_lock); |
acce952b | 4951 | |
fe816d0f NB |
4952 | /* |
4953 | * Make sure we get a live inode and that it'll not disappear | |
4954 | * meanwhile. | |
4955 | */ | |
4956 | inode = igrab(&btrfs_inode->vfs_inode); | |
4957 | if (inode) { | |
4958 | invalidate_inode_pages2(inode->i_mapping); | |
4959 | iput(inode); | |
4960 | } | |
eb73c1b7 | 4961 | spin_lock(&root->delalloc_lock); |
acce952b | 4962 | } |
eb73c1b7 MX |
4963 | spin_unlock(&root->delalloc_lock); |
4964 | } | |
4965 | ||
4966 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4967 | { | |
4968 | struct btrfs_root *root; | |
4969 | struct list_head splice; | |
4970 | ||
4971 | INIT_LIST_HEAD(&splice); | |
4972 | ||
4973 | spin_lock(&fs_info->delalloc_root_lock); | |
4974 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4975 | while (!list_empty(&splice)) { | |
4976 | root = list_first_entry(&splice, struct btrfs_root, | |
4977 | delalloc_root); | |
00246528 | 4978 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
4979 | BUG_ON(!root); |
4980 | spin_unlock(&fs_info->delalloc_root_lock); | |
4981 | ||
4982 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 4983 | btrfs_put_root(root); |
eb73c1b7 MX |
4984 | |
4985 | spin_lock(&fs_info->delalloc_root_lock); | |
4986 | } | |
4987 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4988 | } |
4989 | ||
2ff7e61e | 4990 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 4991 | struct extent_io_tree *dirty_pages, |
4992 | int mark) | |
4993 | { | |
4994 | int ret; | |
acce952b | 4995 | struct extent_buffer *eb; |
4996 | u64 start = 0; | |
4997 | u64 end; | |
acce952b | 4998 | |
4999 | while (1) { | |
5000 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 5001 | mark, NULL); |
acce952b | 5002 | if (ret) |
5003 | break; | |
5004 | ||
91166212 | 5005 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 5006 | while (start <= end) { |
0b246afa JM |
5007 | eb = find_extent_buffer(fs_info, start); |
5008 | start += fs_info->nodesize; | |
fd8b2b61 | 5009 | if (!eb) |
acce952b | 5010 | continue; |
fd8b2b61 | 5011 | wait_on_extent_buffer_writeback(eb); |
acce952b | 5012 | |
fd8b2b61 JB |
5013 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
5014 | &eb->bflags)) | |
5015 | clear_extent_buffer_dirty(eb); | |
5016 | free_extent_buffer_stale(eb); | |
acce952b | 5017 | } |
5018 | } | |
5019 | ||
5020 | return ret; | |
5021 | } | |
5022 | ||
2ff7e61e | 5023 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
fe119a6e | 5024 | struct extent_io_tree *unpin) |
acce952b | 5025 | { |
acce952b | 5026 | u64 start; |
5027 | u64 end; | |
5028 | int ret; | |
5029 | ||
acce952b | 5030 | while (1) { |
0e6ec385 FM |
5031 | struct extent_state *cached_state = NULL; |
5032 | ||
fcd5e742 LF |
5033 | /* |
5034 | * The btrfs_finish_extent_commit() may get the same range as | |
5035 | * ours between find_first_extent_bit and clear_extent_dirty. | |
5036 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
5037 | * the same extent range. | |
5038 | */ | |
5039 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5040 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
0e6ec385 | 5041 | EXTENT_DIRTY, &cached_state); |
fcd5e742 LF |
5042 | if (ret) { |
5043 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5044 | break; |
fcd5e742 | 5045 | } |
acce952b | 5046 | |
0e6ec385 FM |
5047 | clear_extent_dirty(unpin, start, end, &cached_state); |
5048 | free_extent_state(cached_state); | |
2ff7e61e | 5049 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 5050 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 5051 | cond_resched(); |
5052 | } | |
5053 | ||
5054 | return 0; | |
5055 | } | |
5056 | ||
32da5386 | 5057 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
5058 | { |
5059 | struct inode *inode; | |
5060 | ||
5061 | inode = cache->io_ctl.inode; | |
5062 | if (inode) { | |
5063 | invalidate_inode_pages2(inode->i_mapping); | |
5064 | BTRFS_I(inode)->generation = 0; | |
5065 | cache->io_ctl.inode = NULL; | |
5066 | iput(inode); | |
5067 | } | |
bbc37d6e | 5068 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
5069 | btrfs_put_block_group(cache); |
5070 | } | |
5071 | ||
5072 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 5073 | struct btrfs_fs_info *fs_info) |
c79a1751 | 5074 | { |
32da5386 | 5075 | struct btrfs_block_group *cache; |
c79a1751 LB |
5076 | |
5077 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5078 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
5079 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 5080 | struct btrfs_block_group, |
c79a1751 | 5081 | dirty_list); |
c79a1751 LB |
5082 | |
5083 | if (!list_empty(&cache->io_list)) { | |
5084 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5085 | list_del_init(&cache->io_list); | |
5086 | btrfs_cleanup_bg_io(cache); | |
5087 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5088 | } | |
5089 | ||
5090 | list_del_init(&cache->dirty_list); | |
5091 | spin_lock(&cache->lock); | |
5092 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5093 | spin_unlock(&cache->lock); | |
5094 | ||
5095 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5096 | btrfs_put_block_group(cache); | |
ba2c4d4e | 5097 | btrfs_delayed_refs_rsv_release(fs_info, 1); |
c79a1751 LB |
5098 | spin_lock(&cur_trans->dirty_bgs_lock); |
5099 | } | |
5100 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5101 | ||
45ae2c18 NB |
5102 | /* |
5103 | * Refer to the definition of io_bgs member for details why it's safe | |
5104 | * to use it without any locking | |
5105 | */ | |
c79a1751 LB |
5106 | while (!list_empty(&cur_trans->io_bgs)) { |
5107 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 5108 | struct btrfs_block_group, |
c79a1751 | 5109 | io_list); |
c79a1751 LB |
5110 | |
5111 | list_del_init(&cache->io_list); | |
5112 | spin_lock(&cache->lock); | |
5113 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5114 | spin_unlock(&cache->lock); | |
5115 | btrfs_cleanup_bg_io(cache); | |
5116 | } | |
5117 | } | |
5118 | ||
49b25e05 | 5119 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 5120 | struct btrfs_fs_info *fs_info) |
49b25e05 | 5121 | { |
bbbf7243 NB |
5122 | struct btrfs_device *dev, *tmp; |
5123 | ||
2ff7e61e | 5124 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
5125 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
5126 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
5127 | ||
bbbf7243 NB |
5128 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
5129 | post_commit_list) { | |
5130 | list_del_init(&dev->post_commit_list); | |
5131 | } | |
5132 | ||
2ff7e61e | 5133 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 5134 | |
4a9d8bde | 5135 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 5136 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 5137 | |
4a9d8bde | 5138 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 5139 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 5140 | |
ccdf9b30 | 5141 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 5142 | |
2ff7e61e | 5143 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 5144 | EXTENT_DIRTY); |
fe119a6e | 5145 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 5146 | |
d3575156 NA |
5147 | btrfs_free_redirty_list(cur_trans); |
5148 | ||
4a9d8bde MX |
5149 | cur_trans->state =TRANS_STATE_COMPLETED; |
5150 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
5151 | } |
5152 | ||
2ff7e61e | 5153 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 5154 | { |
5155 | struct btrfs_transaction *t; | |
acce952b | 5156 | |
0b246afa | 5157 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 5158 | |
0b246afa JM |
5159 | spin_lock(&fs_info->trans_lock); |
5160 | while (!list_empty(&fs_info->trans_list)) { | |
5161 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
5162 | struct btrfs_transaction, list); |
5163 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 5164 | refcount_inc(&t->use_count); |
0b246afa | 5165 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 5166 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 5167 | btrfs_put_transaction(t); |
0b246afa | 5168 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
5169 | continue; |
5170 | } | |
0b246afa | 5171 | if (t == fs_info->running_transaction) { |
724e2315 | 5172 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 5173 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
5174 | /* |
5175 | * We wait for 0 num_writers since we don't hold a trans | |
5176 | * handle open currently for this transaction. | |
5177 | */ | |
5178 | wait_event(t->writer_wait, | |
5179 | atomic_read(&t->num_writers) == 0); | |
5180 | } else { | |
0b246afa | 5181 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 5182 | } |
2ff7e61e | 5183 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 5184 | |
0b246afa JM |
5185 | spin_lock(&fs_info->trans_lock); |
5186 | if (t == fs_info->running_transaction) | |
5187 | fs_info->running_transaction = NULL; | |
acce952b | 5188 | list_del_init(&t->list); |
0b246afa | 5189 | spin_unlock(&fs_info->trans_lock); |
acce952b | 5190 | |
724e2315 | 5191 | btrfs_put_transaction(t); |
2e4e97ab | 5192 | trace_btrfs_transaction_commit(fs_info); |
0b246afa | 5193 | spin_lock(&fs_info->trans_lock); |
724e2315 | 5194 | } |
0b246afa JM |
5195 | spin_unlock(&fs_info->trans_lock); |
5196 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
5197 | btrfs_destroy_delayed_inodes(fs_info); |
5198 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 5199 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 5200 | btrfs_drop_all_logs(fs_info); |
0b246afa | 5201 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 5202 | |
5203 | return 0; | |
5204 | } | |
ec7d6dfd | 5205 | |
453e4873 | 5206 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
5207 | { |
5208 | struct btrfs_path *path; | |
5209 | int ret; | |
5210 | struct extent_buffer *l; | |
5211 | struct btrfs_key search_key; | |
5212 | struct btrfs_key found_key; | |
5213 | int slot; | |
5214 | ||
5215 | path = btrfs_alloc_path(); | |
5216 | if (!path) | |
5217 | return -ENOMEM; | |
5218 | ||
5219 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
5220 | search_key.type = -1; | |
5221 | search_key.offset = (u64)-1; | |
5222 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
5223 | if (ret < 0) | |
5224 | goto error; | |
5225 | BUG_ON(ret == 0); /* Corruption */ | |
5226 | if (path->slots[0] > 0) { | |
5227 | slot = path->slots[0] - 1; | |
5228 | l = path->nodes[0]; | |
5229 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
5230 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
5231 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 5232 | } else { |
23125104 | 5233 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
5234 | } |
5235 | ret = 0; | |
5236 | error: | |
5237 | btrfs_free_path(path); | |
5238 | return ret; | |
5239 | } | |
5240 | ||
543068a2 | 5241 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
5242 | { |
5243 | int ret; | |
5244 | mutex_lock(&root->objectid_mutex); | |
5245 | ||
6b8fad57 | 5246 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
5247 | btrfs_warn(root->fs_info, |
5248 | "the objectid of root %llu reaches its highest value", | |
5249 | root->root_key.objectid); | |
5250 | ret = -ENOSPC; | |
5251 | goto out; | |
5252 | } | |
5253 | ||
23125104 | 5254 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
5255 | ret = 0; |
5256 | out: | |
5257 | mutex_unlock(&root->objectid_mutex); | |
5258 | return ret; | |
5259 | } |