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" |
103c1972 | 26 | #include "bio.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" |
606686ee | 32 | #include "rcu-string.h" |
8dabb742 | 33 | #include "dev-replace.h" |
53b381b3 | 34 | #include "raid56.h" |
5ac1d209 | 35 | #include "sysfs.h" |
fcebe456 | 36 | #include "qgroup.h" |
ebb8765b | 37 | #include "compression.h" |
557ea5dd | 38 | #include "tree-checker.h" |
fd708b81 | 39 | #include "ref-verify.h" |
aac0023c | 40 | #include "block-group.h" |
b0643e59 | 41 | #include "discard.h" |
f603bb94 | 42 | #include "space-info.h" |
b70f5097 | 43 | #include "zoned.h" |
139e8cd3 | 44 | #include "subpage.h" |
c7f13d42 | 45 | #include "fs.h" |
07e81dc9 | 46 | #include "accessors.h" |
a0231804 | 47 | #include "extent-tree.h" |
45c40c8f | 48 | #include "root-tree.h" |
59b818e0 | 49 | #include "defrag.h" |
c7a03b52 | 50 | #include "uuid-tree.h" |
67707479 | 51 | #include "relocation.h" |
2fc6822c | 52 | #include "scrub.h" |
c03b2207 | 53 | #include "super.h" |
eb60ceac | 54 | |
319e4d06 QW |
55 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
56 | BTRFS_HEADER_FLAG_RELOC |\ | |
57 | BTRFS_SUPER_FLAG_ERROR |\ | |
58 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
59 | BTRFS_SUPER_FLAG_METADUMP |\ |
60 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 61 | |
2ff7e61e JM |
62 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
63 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 64 | |
141386e1 JB |
65 | static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) |
66 | { | |
67 | if (fs_info->csum_shash) | |
68 | crypto_free_shash(fs_info->csum_shash); | |
69 | } | |
70 | ||
d352ac68 | 71 | /* |
2996e1f8 | 72 | * Compute the csum of a btree block and store the result to provided buffer. |
d352ac68 | 73 | */ |
c67b3892 | 74 | static void csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 75 | { |
d5178578 | 76 | struct btrfs_fs_info *fs_info = buf->fs_info; |
7280305e | 77 | const int num_pages = num_extent_pages(buf); |
a26663e7 | 78 | const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); |
d5178578 | 79 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
19c00ddc | 80 | char *kaddr; |
e9be5a30 | 81 | int i; |
d5178578 JT |
82 | |
83 | shash->tfm = fs_info->csum_shash; | |
84 | crypto_shash_init(shash); | |
a26663e7 | 85 | kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); |
e9be5a30 | 86 | crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, |
a26663e7 | 87 | first_page_part - BTRFS_CSUM_SIZE); |
19c00ddc | 88 | |
5ad9b471 | 89 | for (i = 1; i < num_pages && INLINE_EXTENT_BUFFER_PAGES > 1; i++) { |
e9be5a30 DS |
90 | kaddr = page_address(buf->pages[i]); |
91 | crypto_shash_update(shash, kaddr, PAGE_SIZE); | |
19c00ddc | 92 | } |
71a63551 | 93 | memset(result, 0, BTRFS_CSUM_SIZE); |
d5178578 | 94 | crypto_shash_final(shash, result); |
19c00ddc CM |
95 | } |
96 | ||
d352ac68 CM |
97 | /* |
98 | * we can't consider a given block up to date unless the transid of the | |
99 | * block matches the transid in the parent node's pointer. This is how we | |
100 | * detect blocks that either didn't get written at all or got written | |
101 | * in the wrong place. | |
102 | */ | |
d87e6575 | 103 | int btrfs_buffer_uptodate(struct extent_buffer *eb, u64 parent_transid, int atomic) |
1259ab75 | 104 | { |
d87e6575 | 105 | if (!extent_buffer_uptodate(eb)) |
1259ab75 CM |
106 | return 0; |
107 | ||
d87e6575 CH |
108 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) |
109 | return 1; | |
110 | ||
b9fab919 CM |
111 | if (atomic) |
112 | return -EAGAIN; | |
113 | ||
d87e6575 CH |
114 | if (!extent_buffer_uptodate(eb) || |
115 | btrfs_header_generation(eb) != parent_transid) { | |
116 | btrfs_err_rl(eb->fs_info, | |
8f0ed7d4 QW |
117 | "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", |
118 | eb->start, eb->read_mirror, | |
29549aec | 119 | parent_transid, btrfs_header_generation(eb)); |
d87e6575 | 120 | clear_extent_buffer_uptodate(eb); |
9e2aff90 | 121 | return 0; |
d87e6575 | 122 | } |
9e2aff90 | 123 | return 1; |
1259ab75 CM |
124 | } |
125 | ||
e7e16f48 JT |
126 | static bool btrfs_supported_super_csum(u16 csum_type) |
127 | { | |
128 | switch (csum_type) { | |
129 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 130 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 131 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 132 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
133 | return true; |
134 | default: | |
135 | return false; | |
136 | } | |
137 | } | |
138 | ||
1104a885 DS |
139 | /* |
140 | * Return 0 if the superblock checksum type matches the checksum value of that | |
141 | * algorithm. Pass the raw disk superblock data. | |
142 | */ | |
3d17adea QW |
143 | int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
144 | const struct btrfs_super_block *disk_sb) | |
1104a885 | 145 | { |
51bce6c9 | 146 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
147 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
148 | ||
149 | shash->tfm = fs_info->csum_shash; | |
1104a885 | 150 | |
51bce6c9 JT |
151 | /* |
152 | * The super_block structure does not span the whole | |
153 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
154 | * filled with zeros and is included in the checksum. | |
155 | */ | |
3d17adea | 156 | crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, |
fd08001f | 157 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); |
1104a885 | 158 | |
55fc29be | 159 | if (memcmp(disk_sb->csum, result, fs_info->csum_size)) |
51bce6c9 | 160 | return 1; |
1104a885 | 161 | |
e7e16f48 | 162 | return 0; |
1104a885 DS |
163 | } |
164 | ||
bacf60e5 CH |
165 | static int btrfs_repair_eb_io_failure(const struct extent_buffer *eb, |
166 | int mirror_num) | |
167 | { | |
168 | struct btrfs_fs_info *fs_info = eb->fs_info; | |
bacf60e5 CH |
169 | int i, num_pages = num_extent_pages(eb); |
170 | int ret = 0; | |
171 | ||
172 | if (sb_rdonly(fs_info->sb)) | |
173 | return -EROFS; | |
174 | ||
175 | for (i = 0; i < num_pages; i++) { | |
176 | struct page *p = eb->pages[i]; | |
917ac778 QW |
177 | u64 start = max_t(u64, eb->start, page_offset(p)); |
178 | u64 end = min_t(u64, eb->start + eb->len, page_offset(p) + PAGE_SIZE); | |
179 | u32 len = end - start; | |
bacf60e5 | 180 | |
917ac778 QW |
181 | ret = btrfs_repair_io_failure(fs_info, 0, start, len, |
182 | start, p, offset_in_page(start), mirror_num); | |
bacf60e5 CH |
183 | if (ret) |
184 | break; | |
bacf60e5 CH |
185 | } |
186 | ||
187 | return ret; | |
188 | } | |
189 | ||
d352ac68 CM |
190 | /* |
191 | * helper to read a given tree block, doing retries as required when | |
192 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 | 193 | * |
789d6a3a QW |
194 | * @check: expected tree parentness check, see the comments of the |
195 | * structure for details. | |
d352ac68 | 196 | */ |
6a2e9dc4 | 197 | int btrfs_read_extent_buffer(struct extent_buffer *eb, |
789d6a3a | 198 | struct btrfs_tree_parent_check *check) |
f188591e | 199 | { |
5ab12d1f | 200 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ea466794 | 201 | int failed = 0; |
f188591e CM |
202 | int ret; |
203 | int num_copies = 0; | |
204 | int mirror_num = 0; | |
ea466794 | 205 | int failed_mirror = 0; |
f188591e | 206 | |
789d6a3a QW |
207 | ASSERT(check); |
208 | ||
f188591e | 209 | while (1) { |
f8397d69 | 210 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
947a6299 QW |
211 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num, check); |
212 | if (!ret) | |
213 | break; | |
d397712b | 214 | |
0b246afa | 215 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 216 | eb->start, eb->len); |
4235298e | 217 | if (num_copies == 1) |
ea466794 | 218 | break; |
4235298e | 219 | |
5cf1ab56 JB |
220 | if (!failed_mirror) { |
221 | failed = 1; | |
222 | failed_mirror = eb->read_mirror; | |
223 | } | |
224 | ||
f188591e | 225 | mirror_num++; |
ea466794 JB |
226 | if (mirror_num == failed_mirror) |
227 | mirror_num++; | |
228 | ||
4235298e | 229 | if (mirror_num > num_copies) |
ea466794 | 230 | break; |
f188591e | 231 | } |
ea466794 | 232 | |
c0901581 | 233 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 234 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
235 | |
236 | return ret; | |
f188591e | 237 | } |
19c00ddc | 238 | |
31d89399 CH |
239 | /* |
240 | * Checksum a dirty tree block before IO. | |
241 | */ | |
242 | blk_status_t btree_csum_one_bio(struct btrfs_bio *bbio) | |
eca0f6f6 | 243 | { |
31d89399 | 244 | struct extent_buffer *eb = bbio->private; |
eca0f6f6 | 245 | struct btrfs_fs_info *fs_info = eb->fs_info; |
31d89399 | 246 | u64 found_start = btrfs_header_bytenr(eb); |
0124855f | 247 | u64 last_trans; |
eca0f6f6 QW |
248 | u8 result[BTRFS_CSUM_SIZE]; |
249 | int ret; | |
250 | ||
31d89399 CH |
251 | /* Btree blocks are always contiguous on disk. */ |
252 | if (WARN_ON_ONCE(bbio->file_offset != eb->start)) | |
253 | return BLK_STS_IOERR; | |
254 | if (WARN_ON_ONCE(bbio->bio.bi_iter.bi_size != eb->len)) | |
255 | return BLK_STS_IOERR; | |
256 | ||
aa6313e6 JT |
257 | /* |
258 | * If an extent_buffer is marked as EXTENT_BUFFER_ZONED_ZEROOUT, don't | |
259 | * checksum it but zero-out its content. This is done to preserve | |
260 | * ordering of I/O without unnecessarily writing out data. | |
261 | */ | |
cbf44cd9 | 262 | if (test_bit(EXTENT_BUFFER_ZONED_ZEROOUT, &eb->bflags)) { |
aa6313e6 | 263 | memzero_extent_buffer(eb, 0, eb->len); |
31d89399 CH |
264 | return BLK_STS_OK; |
265 | } | |
266 | ||
267 | if (WARN_ON_ONCE(found_start != eb->start)) | |
268 | return BLK_STS_IOERR; | |
269 | if (WARN_ON(!btrfs_page_test_uptodate(fs_info, eb->pages[0], eb->start, | |
270 | eb->len))) | |
271 | return BLK_STS_IOERR; | |
272 | ||
eca0f6f6 QW |
273 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, |
274 | offsetof(struct btrfs_header, fsid), | |
275 | BTRFS_FSID_SIZE) == 0); | |
276 | csum_tree_block(eb, result); | |
277 | ||
278 | if (btrfs_header_level(eb)) | |
279 | ret = btrfs_check_node(eb); | |
280 | else | |
85d8a826 | 281 | ret = btrfs_check_leaf(eb); |
eca0f6f6 | 282 | |
3777369f QW |
283 | if (ret < 0) |
284 | goto error; | |
285 | ||
286 | /* | |
287 | * Also check the generation, the eb reached here must be newer than | |
288 | * last committed. Or something seriously wrong happened. | |
289 | */ | |
0124855f FM |
290 | last_trans = btrfs_get_last_trans_committed(fs_info); |
291 | if (unlikely(btrfs_header_generation(eb) <= last_trans)) { | |
3777369f | 292 | ret = -EUCLEAN; |
eca0f6f6 | 293 | btrfs_err(fs_info, |
3777369f | 294 | "block=%llu bad generation, have %llu expect > %llu", |
0124855f | 295 | eb->start, btrfs_header_generation(eb), last_trans); |
3777369f | 296 | goto error; |
eca0f6f6 QW |
297 | } |
298 | write_extent_buffer(eb, result, 0, fs_info->csum_size); | |
31d89399 | 299 | return BLK_STS_OK; |
3777369f QW |
300 | |
301 | error: | |
302 | btrfs_print_tree(eb, 0); | |
303 | btrfs_err(fs_info, "block=%llu write time tree block corruption detected", | |
304 | eb->start); | |
16199ad9 FM |
305 | /* |
306 | * Be noisy if this is an extent buffer from a log tree. We don't abort | |
307 | * a transaction in case there's a bad log tree extent buffer, we just | |
308 | * fallback to a transaction commit. Still we want to know when there is | |
309 | * a bad log tree extent buffer, as that may signal a bug somewhere. | |
310 | */ | |
311 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG) || | |
312 | btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID); | |
deb6216f CH |
313 | return errno_to_blk_status(ret); |
314 | } | |
315 | ||
413fb1bc | 316 | static bool check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 317 | { |
b0c9b3b0 | 318 | struct btrfs_fs_info *fs_info = eb->fs_info; |
944d3f9f | 319 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; |
44880fdc | 320 | u8 fsid[BTRFS_FSID_SIZE]; |
2b82032c | 321 | |
9a8658e3 DS |
322 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
323 | BTRFS_FSID_SIZE); | |
67bc5ad0 | 324 | |
944d3f9f | 325 | /* |
f7361d8c AJ |
326 | * alloc_fsid_devices() copies the fsid into fs_devices::metadata_uuid. |
327 | * This is then overwritten by metadata_uuid if it is present in the | |
328 | * device_list_add(). The same true for a seed device as well. So use of | |
329 | * fs_devices::metadata_uuid is appropriate here. | |
944d3f9f | 330 | */ |
67bc5ad0 | 331 | if (memcmp(fsid, fs_info->fs_devices->metadata_uuid, BTRFS_FSID_SIZE) == 0) |
413fb1bc | 332 | return false; |
944d3f9f NB |
333 | |
334 | list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) | |
335 | if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) | |
413fb1bc | 336 | return false; |
944d3f9f | 337 | |
413fb1bc | 338 | return true; |
2b82032c YZ |
339 | } |
340 | ||
77bf40a2 | 341 | /* Do basic extent buffer checks at read time */ |
046b562b CH |
342 | int btrfs_validate_extent_buffer(struct extent_buffer *eb, |
343 | struct btrfs_tree_parent_check *check) | |
ce9adaa5 | 344 | { |
77bf40a2 | 345 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ce9adaa5 | 346 | u64 found_start; |
77bf40a2 QW |
347 | const u32 csum_size = fs_info->csum_size; |
348 | u8 found_level; | |
2996e1f8 | 349 | u8 result[BTRFS_CSUM_SIZE]; |
dfd29eed | 350 | const u8 *header_csum; |
77bf40a2 | 351 | int ret = 0; |
ea466794 | 352 | |
947a6299 QW |
353 | ASSERT(check); |
354 | ||
ce9adaa5 | 355 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 356 | if (found_start != eb->start) { |
8f0ed7d4 QW |
357 | btrfs_err_rl(fs_info, |
358 | "bad tree block start, mirror %u want %llu have %llu", | |
359 | eb->read_mirror, eb->start, found_start); | |
f188591e | 360 | ret = -EIO; |
77bf40a2 | 361 | goto out; |
ce9adaa5 | 362 | } |
b0c9b3b0 | 363 | if (check_tree_block_fsid(eb)) { |
8f0ed7d4 QW |
364 | btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", |
365 | eb->start, eb->read_mirror); | |
1259ab75 | 366 | ret = -EIO; |
77bf40a2 | 367 | goto out; |
1259ab75 | 368 | } |
ce9adaa5 | 369 | found_level = btrfs_header_level(eb); |
1c24c3ce | 370 | if (found_level >= BTRFS_MAX_LEVEL) { |
8f0ed7d4 QW |
371 | btrfs_err(fs_info, |
372 | "bad tree block level, mirror %u level %d on logical %llu", | |
373 | eb->read_mirror, btrfs_header_level(eb), eb->start); | |
1c24c3ce | 374 | ret = -EIO; |
77bf40a2 | 375 | goto out; |
1c24c3ce | 376 | } |
ce9adaa5 | 377 | |
c67b3892 | 378 | csum_tree_block(eb, result); |
dfd29eed DS |
379 | header_csum = page_address(eb->pages[0]) + |
380 | get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); | |
a826d6dc | 381 | |
dfd29eed | 382 | if (memcmp(result, header_csum, csum_size) != 0) { |
2996e1f8 | 383 | btrfs_warn_rl(fs_info, |
8f0ed7d4 QW |
384 | "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", |
385 | eb->start, eb->read_mirror, | |
dfd29eed | 386 | CSUM_FMT_VALUE(csum_size, header_csum), |
35be8851 JT |
387 | CSUM_FMT_VALUE(csum_size, result), |
388 | btrfs_header_level(eb)); | |
2996e1f8 | 389 | ret = -EUCLEAN; |
77bf40a2 | 390 | goto out; |
2996e1f8 JT |
391 | } |
392 | ||
947a6299 | 393 | if (found_level != check->level) { |
77177ed1 QW |
394 | btrfs_err(fs_info, |
395 | "level verify failed on logical %llu mirror %u wanted %u found %u", | |
396 | eb->start, eb->read_mirror, check->level, found_level); | |
947a6299 QW |
397 | ret = -EIO; |
398 | goto out; | |
399 | } | |
400 | if (unlikely(check->transid && | |
401 | btrfs_header_generation(eb) != check->transid)) { | |
402 | btrfs_err_rl(eb->fs_info, | |
403 | "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", | |
404 | eb->start, eb->read_mirror, check->transid, | |
405 | btrfs_header_generation(eb)); | |
406 | ret = -EIO; | |
407 | goto out; | |
408 | } | |
409 | if (check->has_first_key) { | |
410 | struct btrfs_key *expect_key = &check->first_key; | |
411 | struct btrfs_key found_key; | |
412 | ||
413 | if (found_level) | |
414 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
415 | else | |
416 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
417 | if (unlikely(btrfs_comp_cpu_keys(expect_key, &found_key))) { | |
418 | btrfs_err(fs_info, | |
419 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", | |
420 | eb->start, check->transid, | |
421 | expect_key->objectid, | |
422 | expect_key->type, expect_key->offset, | |
423 | found_key.objectid, found_key.type, | |
424 | found_key.offset); | |
425 | ret = -EUCLEAN; | |
426 | goto out; | |
427 | } | |
428 | } | |
429 | if (check->owner_root) { | |
430 | ret = btrfs_check_eb_owner(eb, check->owner_root); | |
431 | if (ret < 0) | |
432 | goto out; | |
433 | } | |
434 | ||
a826d6dc JB |
435 | /* |
436 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
437 | * that we don't try and read the other copies of this block, just | |
438 | * return -EIO. | |
439 | */ | |
85d8a826 | 440 | if (found_level == 0 && btrfs_check_leaf(eb)) { |
a826d6dc JB |
441 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
442 | ret = -EIO; | |
443 | } | |
ce9adaa5 | 444 | |
813fd1dc | 445 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
446 | ret = -EIO; |
447 | ||
aebcc159 | 448 | if (ret) |
75391f0d | 449 | btrfs_err(fs_info, |
8f0ed7d4 QW |
450 | "read time tree block corruption detected on logical %llu mirror %u", |
451 | eb->start, eb->read_mirror); | |
77bf40a2 QW |
452 | out: |
453 | return ret; | |
454 | } | |
455 | ||
3dd1462e | 456 | #ifdef CONFIG_MIGRATION |
8958b551 MWO |
457 | static int btree_migrate_folio(struct address_space *mapping, |
458 | struct folio *dst, struct folio *src, enum migrate_mode mode) | |
784b4e29 CM |
459 | { |
460 | /* | |
461 | * we can't safely write a btree page from here, | |
462 | * we haven't done the locking hook | |
463 | */ | |
8958b551 | 464 | if (folio_test_dirty(src)) |
784b4e29 CM |
465 | return -EAGAIN; |
466 | /* | |
467 | * Buffers may be managed in a filesystem specific way. | |
468 | * We must have no buffers or drop them. | |
469 | */ | |
8958b551 MWO |
470 | if (folio_get_private(src) && |
471 | !filemap_release_folio(src, GFP_KERNEL)) | |
784b4e29 | 472 | return -EAGAIN; |
54184650 | 473 | return migrate_folio(mapping, dst, src, mode); |
784b4e29 | 474 | } |
8958b551 MWO |
475 | #else |
476 | #define btree_migrate_folio NULL | |
3dd1462e | 477 | #endif |
784b4e29 | 478 | |
0da5468f CM |
479 | static int btree_writepages(struct address_space *mapping, |
480 | struct writeback_control *wbc) | |
481 | { | |
e2d84521 MX |
482 | struct btrfs_fs_info *fs_info; |
483 | int ret; | |
484 | ||
d8d5f3e1 | 485 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
486 | |
487 | if (wbc->for_kupdate) | |
488 | return 0; | |
489 | ||
e2d84521 | 490 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 491 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
492 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
493 | BTRFS_DIRTY_METADATA_THRESH, | |
494 | fs_info->dirty_metadata_batch); | |
e2d84521 | 495 | if (ret < 0) |
793955bc | 496 | return 0; |
793955bc | 497 | } |
0b32f4bb | 498 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
499 | } |
500 | ||
f913cff3 | 501 | static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) |
5f39d397 | 502 | { |
f913cff3 MWO |
503 | if (folio_test_writeback(folio) || folio_test_dirty(folio)) |
504 | return false; | |
0c4e538b | 505 | |
f913cff3 | 506 | return try_release_extent_buffer(&folio->page); |
d98237b3 CM |
507 | } |
508 | ||
895586eb MWO |
509 | static void btree_invalidate_folio(struct folio *folio, size_t offset, |
510 | size_t length) | |
d98237b3 | 511 | { |
d1310b2e | 512 | struct extent_io_tree *tree; |
895586eb MWO |
513 | tree = &BTRFS_I(folio->mapping->host)->io_tree; |
514 | extent_invalidate_folio(tree, folio, offset); | |
f913cff3 | 515 | btree_release_folio(folio, GFP_NOFS); |
895586eb MWO |
516 | if (folio_get_private(folio)) { |
517 | btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, | |
518 | "folio private not zero on folio %llu", | |
519 | (unsigned long long)folio_pos(folio)); | |
520 | folio_detach_private(folio); | |
9ad6b7bc | 521 | } |
d98237b3 CM |
522 | } |
523 | ||
bb146eb2 | 524 | #ifdef DEBUG |
0079c3b1 MWO |
525 | static bool btree_dirty_folio(struct address_space *mapping, |
526 | struct folio *folio) | |
527 | { | |
528 | struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); | |
5e0e8799 | 529 | struct btrfs_subpage_info *spi = fs_info->subpage_info; |
139e8cd3 | 530 | struct btrfs_subpage *subpage; |
0b32f4bb | 531 | struct extent_buffer *eb; |
139e8cd3 | 532 | int cur_bit = 0; |
0079c3b1 | 533 | u64 page_start = folio_pos(folio); |
139e8cd3 QW |
534 | |
535 | if (fs_info->sectorsize == PAGE_SIZE) { | |
0079c3b1 | 536 | eb = folio_get_private(folio); |
139e8cd3 QW |
537 | BUG_ON(!eb); |
538 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
539 | BUG_ON(!atomic_read(&eb->refs)); | |
49d0c642 | 540 | btrfs_assert_tree_write_locked(eb); |
0079c3b1 | 541 | return filemap_dirty_folio(mapping, folio); |
139e8cd3 | 542 | } |
5e0e8799 QW |
543 | |
544 | ASSERT(spi); | |
0079c3b1 | 545 | subpage = folio_get_private(folio); |
139e8cd3 | 546 | |
5e0e8799 QW |
547 | for (cur_bit = spi->dirty_offset; |
548 | cur_bit < spi->dirty_offset + spi->bitmap_nr_bits; | |
549 | cur_bit++) { | |
139e8cd3 QW |
550 | unsigned long flags; |
551 | u64 cur; | |
139e8cd3 QW |
552 | |
553 | spin_lock_irqsave(&subpage->lock, flags); | |
5e0e8799 | 554 | if (!test_bit(cur_bit, subpage->bitmaps)) { |
139e8cd3 | 555 | spin_unlock_irqrestore(&subpage->lock, flags); |
139e8cd3 QW |
556 | continue; |
557 | } | |
558 | spin_unlock_irqrestore(&subpage->lock, flags); | |
559 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 560 | |
139e8cd3 QW |
561 | eb = find_extent_buffer(fs_info, cur); |
562 | ASSERT(eb); | |
563 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
564 | ASSERT(atomic_read(&eb->refs)); | |
49d0c642 | 565 | btrfs_assert_tree_write_locked(eb); |
139e8cd3 QW |
566 | free_extent_buffer(eb); |
567 | ||
5e0e8799 | 568 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits) - 1; |
139e8cd3 | 569 | } |
0079c3b1 | 570 | return filemap_dirty_folio(mapping, folio); |
0b32f4bb | 571 | } |
0079c3b1 MWO |
572 | #else |
573 | #define btree_dirty_folio filemap_dirty_folio | |
574 | #endif | |
0b32f4bb | 575 | |
7f09410b | 576 | static const struct address_space_operations btree_aops = { |
0da5468f | 577 | .writepages = btree_writepages, |
f913cff3 | 578 | .release_folio = btree_release_folio, |
895586eb | 579 | .invalidate_folio = btree_invalidate_folio, |
8958b551 MWO |
580 | .migrate_folio = btree_migrate_folio, |
581 | .dirty_folio = btree_dirty_folio, | |
d98237b3 CM |
582 | }; |
583 | ||
2ff7e61e JM |
584 | struct extent_buffer *btrfs_find_create_tree_block( |
585 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
586 | u64 bytenr, u64 owner_root, |
587 | int level) | |
0999df54 | 588 | { |
0b246afa JM |
589 | if (btrfs_is_testing(fs_info)) |
590 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 591 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
592 | } |
593 | ||
581c1760 QW |
594 | /* |
595 | * Read tree block at logical address @bytenr and do variant basic but critical | |
596 | * verification. | |
597 | * | |
789d6a3a QW |
598 | * @check: expected tree parentness check, see comments of the |
599 | * structure for details. | |
581c1760 | 600 | */ |
2ff7e61e | 601 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
789d6a3a | 602 | struct btrfs_tree_parent_check *check) |
0999df54 CM |
603 | { |
604 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
605 | int ret; |
606 | ||
789d6a3a QW |
607 | ASSERT(check); |
608 | ||
609 | buf = btrfs_find_create_tree_block(fs_info, bytenr, check->owner_root, | |
610 | check->level); | |
c871b0f2 LB |
611 | if (IS_ERR(buf)) |
612 | return buf; | |
0999df54 | 613 | |
789d6a3a | 614 | ret = btrfs_read_extent_buffer(buf, check); |
0f0fe8f7 | 615 | if (ret) { |
537f38f0 | 616 | free_extent_buffer_stale(buf); |
64c043de | 617 | return ERR_PTR(ret); |
0f0fe8f7 | 618 | } |
789d6a3a | 619 | if (btrfs_check_eb_owner(buf, check->owner_root)) { |
88c602ab QW |
620 | free_extent_buffer_stale(buf); |
621 | return ERR_PTR(-EUCLEAN); | |
622 | } | |
5f39d397 | 623 | return buf; |
ce9adaa5 | 624 | |
eb60ceac CM |
625 | } |
626 | ||
da17066c | 627 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 628 | u64 objectid) |
d97e63b6 | 629 | { |
7c0260ee | 630 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
2e608bd1 JB |
631 | |
632 | memset(&root->root_key, 0, sizeof(root->root_key)); | |
633 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
634 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); | |
96dfcb46 | 635 | root->fs_info = fs_info; |
2e608bd1 | 636 | root->root_key.objectid = objectid; |
cfaa7295 | 637 | root->node = NULL; |
a28ec197 | 638 | root->commit_root = NULL; |
27cdeb70 | 639 | root->state = 0; |
abed4aaa | 640 | RB_CLEAR_NODE(&root->rb_node); |
0b86a832 | 641 | |
0f7d52f4 | 642 | root->last_trans = 0; |
6b8fad57 | 643 | root->free_objectid = 0; |
eb73c1b7 | 644 | root->nr_delalloc_inodes = 0; |
199c2a9c | 645 | root->nr_ordered_extents = 0; |
6bef4d31 | 646 | root->inode_tree = RB_ROOT; |
088aea3b | 647 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
2e608bd1 JB |
648 | |
649 | btrfs_init_root_block_rsv(root); | |
0b86a832 CM |
650 | |
651 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 652 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
653 | INIT_LIST_HEAD(&root->delalloc_inodes); |
654 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
655 | INIT_LIST_HEAD(&root->ordered_extents); |
656 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 657 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
5d4f98a2 | 658 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 659 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 660 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 661 | spin_lock_init(&root->accounting_lock); |
8287475a | 662 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 663 | mutex_init(&root->objectid_mutex); |
e02119d5 | 664 | mutex_init(&root->log_mutex); |
31f3d255 | 665 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 666 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 667 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
668 | init_waitqueue_head(&root->log_writer_wait); |
669 | init_waitqueue_head(&root->log_commit_wait[0]); | |
670 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
671 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
672 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
673 | atomic_set(&root->log_commit[0], 0); |
674 | atomic_set(&root->log_commit[1], 0); | |
675 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 676 | atomic_set(&root->log_batch, 0); |
0700cea7 | 677 | refcount_set(&root->refs, 1); |
8ecebf4d | 678 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 679 | atomic_set(&root->nr_swapfiles, 0); |
6008859b | 680 | btrfs_set_root_log_transid(root, 0); |
d1433deb | 681 | root->log_transid_committed = -1; |
f9850787 | 682 | btrfs_set_root_last_log_commit(root, 0); |
2e608bd1 | 683 | root->anon_dev = 0; |
e289f03e | 684 | if (!dummy) { |
43eb5f29 | 685 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
35da5a7e | 686 | IO_TREE_ROOT_DIRTY_LOG_PAGES); |
e289f03e | 687 | extent_io_tree_init(fs_info, &root->log_csum_range, |
35da5a7e | 688 | IO_TREE_LOG_CSUM_RANGE); |
e289f03e | 689 | } |
017e5369 | 690 | |
5f3ab90a | 691 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 692 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
693 | #ifdef CONFIG_BTRFS_DEBUG |
694 | INIT_LIST_HEAD(&root->leak_list); | |
fc7cbcd4 | 695 | spin_lock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 696 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); |
fc7cbcd4 | 697 | spin_unlock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 698 | #endif |
3768f368 CM |
699 | } |
700 | ||
74e4d827 | 701 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 702 | u64 objectid, gfp_t flags) |
6f07e42e | 703 | { |
74e4d827 | 704 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 705 | if (root) |
96dfcb46 | 706 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
707 | return root; |
708 | } | |
709 | ||
06ea65a3 JB |
710 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
711 | /* Should only be used by the testing infrastructure */ | |
da17066c | 712 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
713 | { |
714 | struct btrfs_root *root; | |
715 | ||
7c0260ee JM |
716 | if (!fs_info) |
717 | return ERR_PTR(-EINVAL); | |
718 | ||
96dfcb46 | 719 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
720 | if (!root) |
721 | return ERR_PTR(-ENOMEM); | |
da17066c | 722 | |
b9ef22de | 723 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 724 | root->alloc_bytenr = 0; |
06ea65a3 JB |
725 | |
726 | return root; | |
727 | } | |
728 | #endif | |
729 | ||
abed4aaa JB |
730 | static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) |
731 | { | |
732 | const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); | |
733 | const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); | |
734 | ||
735 | return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); | |
736 | } | |
737 | ||
738 | static int global_root_key_cmp(const void *k, const struct rb_node *node) | |
739 | { | |
740 | const struct btrfs_key *key = k; | |
741 | const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); | |
742 | ||
743 | return btrfs_comp_cpu_keys(key, &root->root_key); | |
744 | } | |
745 | ||
746 | int btrfs_global_root_insert(struct btrfs_root *root) | |
747 | { | |
748 | struct btrfs_fs_info *fs_info = root->fs_info; | |
749 | struct rb_node *tmp; | |
745806fb | 750 | int ret = 0; |
abed4aaa JB |
751 | |
752 | write_lock(&fs_info->global_root_lock); | |
753 | tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); | |
754 | write_unlock(&fs_info->global_root_lock); | |
abed4aaa | 755 | |
745806fb QW |
756 | if (tmp) { |
757 | ret = -EEXIST; | |
758 | btrfs_warn(fs_info, "global root %llu %llu already exists", | |
759 | root->root_key.objectid, root->root_key.offset); | |
760 | } | |
761 | return ret; | |
abed4aaa JB |
762 | } |
763 | ||
764 | void btrfs_global_root_delete(struct btrfs_root *root) | |
765 | { | |
766 | struct btrfs_fs_info *fs_info = root->fs_info; | |
767 | ||
768 | write_lock(&fs_info->global_root_lock); | |
769 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
770 | write_unlock(&fs_info->global_root_lock); | |
771 | } | |
772 | ||
773 | struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, | |
774 | struct btrfs_key *key) | |
775 | { | |
776 | struct rb_node *node; | |
777 | struct btrfs_root *root = NULL; | |
778 | ||
779 | read_lock(&fs_info->global_root_lock); | |
780 | node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); | |
781 | if (node) | |
782 | root = container_of(node, struct btrfs_root, rb_node); | |
783 | read_unlock(&fs_info->global_root_lock); | |
784 | ||
785 | return root; | |
786 | } | |
787 | ||
f7238e50 JB |
788 | static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) |
789 | { | |
790 | struct btrfs_block_group *block_group; | |
791 | u64 ret; | |
792 | ||
793 | if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
794 | return 0; | |
795 | ||
796 | if (bytenr) | |
797 | block_group = btrfs_lookup_block_group(fs_info, bytenr); | |
798 | else | |
799 | block_group = btrfs_lookup_first_block_group(fs_info, bytenr); | |
800 | ASSERT(block_group); | |
801 | if (!block_group) | |
802 | return 0; | |
803 | ret = block_group->global_root_id; | |
804 | btrfs_put_block_group(block_group); | |
805 | ||
806 | return ret; | |
807 | } | |
808 | ||
abed4aaa JB |
809 | struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) |
810 | { | |
811 | struct btrfs_key key = { | |
812 | .objectid = BTRFS_CSUM_TREE_OBJECTID, | |
813 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 814 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
815 | }; |
816 | ||
817 | return btrfs_global_root(fs_info, &key); | |
818 | } | |
819 | ||
820 | struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) | |
821 | { | |
822 | struct btrfs_key key = { | |
823 | .objectid = BTRFS_EXTENT_TREE_OBJECTID, | |
824 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 825 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
826 | }; |
827 | ||
828 | return btrfs_global_root(fs_info, &key); | |
829 | } | |
830 | ||
51129b33 JB |
831 | struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) |
832 | { | |
833 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) | |
834 | return fs_info->block_group_root; | |
835 | return btrfs_extent_root(fs_info, 0); | |
836 | } | |
837 | ||
20897f5c | 838 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
839 | u64 objectid) |
840 | { | |
9b7a2440 | 841 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
842 | struct extent_buffer *leaf; |
843 | struct btrfs_root *tree_root = fs_info->tree_root; | |
844 | struct btrfs_root *root; | |
845 | struct btrfs_key key; | |
b89f6d1f | 846 | unsigned int nofs_flag; |
20897f5c | 847 | int ret = 0; |
20897f5c | 848 | |
b89f6d1f FM |
849 | /* |
850 | * We're holding a transaction handle, so use a NOFS memory allocation | |
851 | * context to avoid deadlock if reclaim happens. | |
852 | */ | |
853 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 854 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 855 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
856 | if (!root) |
857 | return ERR_PTR(-ENOMEM); | |
858 | ||
20897f5c AJ |
859 | root->root_key.objectid = objectid; |
860 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
861 | root->root_key.offset = 0; | |
862 | ||
9631e4cc | 863 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
60ea105a | 864 | 0, BTRFS_NESTING_NORMAL); |
20897f5c AJ |
865 | if (IS_ERR(leaf)) { |
866 | ret = PTR_ERR(leaf); | |
1dd05682 | 867 | leaf = NULL; |
c1b07854 | 868 | goto fail; |
20897f5c AJ |
869 | } |
870 | ||
20897f5c | 871 | root->node = leaf; |
50564b65 | 872 | btrfs_mark_buffer_dirty(trans, leaf); |
20897f5c AJ |
873 | |
874 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 875 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 876 | |
f944d2cb DS |
877 | btrfs_set_root_flags(&root->root_item, 0); |
878 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
879 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
880 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
881 | btrfs_set_root_level(&root->root_item, 0); | |
882 | btrfs_set_root_refs(&root->root_item, 1); | |
883 | btrfs_set_root_used(&root->root_item, leaf->len); | |
884 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
885 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 886 | if (is_fstree(objectid)) |
807fc790 AS |
887 | generate_random_guid(root->root_item.uuid); |
888 | else | |
889 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 890 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 891 | |
8a6a87cd BB |
892 | btrfs_tree_unlock(leaf); |
893 | ||
20897f5c AJ |
894 | key.objectid = objectid; |
895 | key.type = BTRFS_ROOT_ITEM_KEY; | |
896 | key.offset = 0; | |
897 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
898 | if (ret) | |
899 | goto fail; | |
900 | ||
1dd05682 TI |
901 | return root; |
902 | ||
8a6a87cd | 903 | fail: |
00246528 | 904 | btrfs_put_root(root); |
20897f5c | 905 | |
1dd05682 | 906 | return ERR_PTR(ret); |
20897f5c AJ |
907 | } |
908 | ||
7237f183 YZ |
909 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
910 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
911 | { |
912 | struct btrfs_root *root; | |
e02119d5 | 913 | |
96dfcb46 | 914 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 915 | if (!root) |
7237f183 | 916 | return ERR_PTR(-ENOMEM); |
e02119d5 | 917 | |
e02119d5 CM |
918 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
919 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
920 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 921 | |
6ab6ebb7 NA |
922 | return root; |
923 | } | |
924 | ||
925 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
926 | struct btrfs_root *root) | |
927 | { | |
928 | struct extent_buffer *leaf; | |
929 | ||
7237f183 | 930 | /* |
92a7cc42 | 931 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 932 | * |
92a7cc42 QW |
933 | * Log trees are not exposed to user space thus can't be snapshotted, |
934 | * and they go away before a real commit is actually done. | |
935 | * | |
936 | * They do store pointers to file data extents, and those reference | |
937 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 938 | */ |
e02119d5 | 939 | |
4d75f8a9 | 940 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
60ea105a | 941 | NULL, 0, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
942 | if (IS_ERR(leaf)) |
943 | return PTR_ERR(leaf); | |
e02119d5 | 944 | |
7237f183 | 945 | root->node = leaf; |
e02119d5 | 946 | |
50564b65 | 947 | btrfs_mark_buffer_dirty(trans, root->node); |
e02119d5 | 948 | btrfs_tree_unlock(root->node); |
6ab6ebb7 NA |
949 | |
950 | return 0; | |
7237f183 YZ |
951 | } |
952 | ||
953 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
954 | struct btrfs_fs_info *fs_info) | |
955 | { | |
956 | struct btrfs_root *log_root; | |
957 | ||
958 | log_root = alloc_log_tree(trans, fs_info); | |
959 | if (IS_ERR(log_root)) | |
960 | return PTR_ERR(log_root); | |
6ab6ebb7 | 961 | |
3ddebf27 NA |
962 | if (!btrfs_is_zoned(fs_info)) { |
963 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
964 | ||
965 | if (ret) { | |
966 | btrfs_put_root(log_root); | |
967 | return ret; | |
968 | } | |
6ab6ebb7 NA |
969 | } |
970 | ||
7237f183 YZ |
971 | WARN_ON(fs_info->log_root_tree); |
972 | fs_info->log_root_tree = log_root; | |
973 | return 0; | |
974 | } | |
975 | ||
976 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
977 | struct btrfs_root *root) | |
978 | { | |
0b246afa | 979 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
980 | struct btrfs_root *log_root; |
981 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 982 | int ret; |
7237f183 | 983 | |
0b246afa | 984 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
985 | if (IS_ERR(log_root)) |
986 | return PTR_ERR(log_root); | |
987 | ||
6ab6ebb7 NA |
988 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
989 | if (ret) { | |
990 | btrfs_put_root(log_root); | |
991 | return ret; | |
992 | } | |
993 | ||
7237f183 YZ |
994 | log_root->last_trans = trans->transid; |
995 | log_root->root_key.offset = root->root_key.objectid; | |
996 | ||
997 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
998 | btrfs_set_stack_inode_generation(inode_item, 1); |
999 | btrfs_set_stack_inode_size(inode_item, 3); | |
1000 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1001 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1002 | fs_info->nodesize); |
3cae210f | 1003 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1004 | |
5d4f98a2 | 1005 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1006 | |
1007 | WARN_ON(root->log_root); | |
1008 | root->log_root = log_root; | |
6008859b | 1009 | btrfs_set_root_log_transid(root, 0); |
d1433deb | 1010 | root->log_transid_committed = -1; |
f9850787 | 1011 | btrfs_set_root_last_log_commit(root, 0); |
e02119d5 CM |
1012 | return 0; |
1013 | } | |
1014 | ||
49d11bea JB |
1015 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1016 | struct btrfs_path *path, | |
1017 | struct btrfs_key *key) | |
e02119d5 CM |
1018 | { |
1019 | struct btrfs_root *root; | |
789d6a3a | 1020 | struct btrfs_tree_parent_check check = { 0 }; |
e02119d5 | 1021 | struct btrfs_fs_info *fs_info = tree_root->fs_info; |
84234f3a | 1022 | u64 generation; |
cb517eab | 1023 | int ret; |
581c1760 | 1024 | int level; |
0f7d52f4 | 1025 | |
96dfcb46 | 1026 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1027 | if (!root) |
1028 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1029 | |
cb517eab MX |
1030 | ret = btrfs_find_root(tree_root, key, path, |
1031 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1032 | if (ret) { |
13a8a7c8 YZ |
1033 | if (ret > 0) |
1034 | ret = -ENOENT; | |
49d11bea | 1035 | goto fail; |
0f7d52f4 | 1036 | } |
13a8a7c8 | 1037 | |
84234f3a | 1038 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1039 | level = btrfs_root_level(&root->root_item); |
789d6a3a QW |
1040 | check.level = level; |
1041 | check.transid = generation; | |
1042 | check.owner_root = key->objectid; | |
1043 | root->node = read_tree_block(fs_info, btrfs_root_bytenr(&root->root_item), | |
1044 | &check); | |
64c043de LB |
1045 | if (IS_ERR(root->node)) { |
1046 | ret = PTR_ERR(root->node); | |
8c38938c | 1047 | root->node = NULL; |
49d11bea | 1048 | goto fail; |
4eb150d6 QW |
1049 | } |
1050 | if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
cb517eab | 1051 | ret = -EIO; |
49d11bea | 1052 | goto fail; |
416bc658 | 1053 | } |
88c602ab QW |
1054 | |
1055 | /* | |
1056 | * For real fs, and not log/reloc trees, root owner must | |
1057 | * match its root node owner | |
1058 | */ | |
1059 | if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && | |
1060 | root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && | |
1061 | root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && | |
1062 | root->root_key.objectid != btrfs_header_owner(root->node)) { | |
1063 | btrfs_crit(fs_info, | |
1064 | "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", | |
1065 | root->root_key.objectid, root->node->start, | |
1066 | btrfs_header_owner(root->node), | |
1067 | root->root_key.objectid); | |
1068 | ret = -EUCLEAN; | |
1069 | goto fail; | |
1070 | } | |
5d4f98a2 | 1071 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1072 | return root; |
49d11bea | 1073 | fail: |
00246528 | 1074 | btrfs_put_root(root); |
49d11bea JB |
1075 | return ERR_PTR(ret); |
1076 | } | |
1077 | ||
1078 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1079 | struct btrfs_key *key) | |
1080 | { | |
1081 | struct btrfs_root *root; | |
1082 | struct btrfs_path *path; | |
1083 | ||
1084 | path = btrfs_alloc_path(); | |
1085 | if (!path) | |
1086 | return ERR_PTR(-ENOMEM); | |
1087 | root = read_tree_root_path(tree_root, path, key); | |
1088 | btrfs_free_path(path); | |
1089 | ||
1090 | return root; | |
cb517eab MX |
1091 | } |
1092 | ||
2dfb1e43 QW |
1093 | /* |
1094 | * Initialize subvolume root in-memory structure | |
1095 | * | |
1096 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1097 | */ | |
1098 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1099 | { |
1100 | int ret; | |
1101 | ||
0b548539 | 1102 | btrfs_drew_lock_init(&root->snapshot_lock); |
8257b2dc | 1103 | |
aeb935a4 | 1104 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
6ebcd021 QW |
1105 | !btrfs_is_data_reloc_root(root) && |
1106 | is_fstree(root->root_key.objectid)) { | |
92a7cc42 | 1107 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1108 | btrfs_check_and_init_root_item(&root->root_item); |
1109 | } | |
1110 | ||
851fd730 QW |
1111 | /* |
1112 | * Don't assign anonymous block device to roots that are not exposed to | |
1113 | * userspace, the id pool is limited to 1M | |
1114 | */ | |
1115 | if (is_fstree(root->root_key.objectid) && | |
1116 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1117 | if (!anon_dev) { |
1118 | ret = get_anon_bdev(&root->anon_dev); | |
1119 | if (ret) | |
1120 | goto fail; | |
1121 | } else { | |
1122 | root->anon_dev = anon_dev; | |
1123 | } | |
851fd730 | 1124 | } |
f32e48e9 CR |
1125 | |
1126 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1127 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1128 | if (ret) { |
1129 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1130 | goto fail; |
f32e48e9 CR |
1131 | } |
1132 | ||
6b8fad57 | 1133 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1134 | |
1135 | mutex_unlock(&root->objectid_mutex); | |
1136 | ||
cb517eab MX |
1137 | return 0; |
1138 | fail: | |
84db5ccf | 1139 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1140 | return ret; |
1141 | } | |
1142 | ||
a98db0f3 JB |
1143 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1144 | u64 root_id) | |
cb517eab MX |
1145 | { |
1146 | struct btrfs_root *root; | |
1147 | ||
fc7cbcd4 DS |
1148 | spin_lock(&fs_info->fs_roots_radix_lock); |
1149 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1150 | (unsigned long)root_id); | |
25ac047c | 1151 | root = btrfs_grab_root(root); |
fc7cbcd4 | 1152 | spin_unlock(&fs_info->fs_roots_radix_lock); |
cb517eab MX |
1153 | return root; |
1154 | } | |
1155 | ||
49d11bea JB |
1156 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1157 | u64 objectid) | |
1158 | { | |
abed4aaa JB |
1159 | struct btrfs_key key = { |
1160 | .objectid = objectid, | |
1161 | .type = BTRFS_ROOT_ITEM_KEY, | |
1162 | .offset = 0, | |
1163 | }; | |
1164 | ||
e91909aa CH |
1165 | switch (objectid) { |
1166 | case BTRFS_ROOT_TREE_OBJECTID: | |
49d11bea | 1167 | return btrfs_grab_root(fs_info->tree_root); |
e91909aa | 1168 | case BTRFS_EXTENT_TREE_OBJECTID: |
abed4aaa | 1169 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
e91909aa | 1170 | case BTRFS_CHUNK_TREE_OBJECTID: |
49d11bea | 1171 | return btrfs_grab_root(fs_info->chunk_root); |
e91909aa | 1172 | case BTRFS_DEV_TREE_OBJECTID: |
49d11bea | 1173 | return btrfs_grab_root(fs_info->dev_root); |
e91909aa | 1174 | case BTRFS_CSUM_TREE_OBJECTID: |
abed4aaa | 1175 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
e91909aa | 1176 | case BTRFS_QUOTA_TREE_OBJECTID: |
85724171 | 1177 | return btrfs_grab_root(fs_info->quota_root); |
e91909aa | 1178 | case BTRFS_UUID_TREE_OBJECTID: |
85724171 | 1179 | return btrfs_grab_root(fs_info->uuid_root); |
e91909aa | 1180 | case BTRFS_BLOCK_GROUP_TREE_OBJECTID: |
85724171 | 1181 | return btrfs_grab_root(fs_info->block_group_root); |
e91909aa | 1182 | case BTRFS_FREE_SPACE_TREE_OBJECTID: |
85724171 | 1183 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
51502090 JT |
1184 | case BTRFS_RAID_STRIPE_TREE_OBJECTID: |
1185 | return btrfs_grab_root(fs_info->stripe_root); | |
e91909aa CH |
1186 | default: |
1187 | return NULL; | |
1188 | } | |
49d11bea JB |
1189 | } |
1190 | ||
cb517eab MX |
1191 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1192 | struct btrfs_root *root) | |
1193 | { | |
1194 | int ret; | |
1195 | ||
fc7cbcd4 DS |
1196 | ret = radix_tree_preload(GFP_NOFS); |
1197 | if (ret) | |
1198 | return ret; | |
1199 | ||
1200 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1201 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1202 | (unsigned long)root->root_key.objectid, | |
1203 | root); | |
af01d2e5 | 1204 | if (ret == 0) { |
00246528 | 1205 | btrfs_grab_root(root); |
fc7cbcd4 | 1206 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1207 | } |
fc7cbcd4 DS |
1208 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1209 | radix_tree_preload_end(); | |
cb517eab MX |
1210 | |
1211 | return ret; | |
1212 | } | |
1213 | ||
bd647ce3 JB |
1214 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1215 | { | |
1216 | #ifdef CONFIG_BTRFS_DEBUG | |
1217 | struct btrfs_root *root; | |
1218 | ||
1219 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1220 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1221 | ||
bd647ce3 JB |
1222 | root = list_first_entry(&fs_info->allocated_roots, |
1223 | struct btrfs_root, leak_list); | |
457f1864 | 1224 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1225 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 JB |
1226 | refcount_read(&root->refs)); |
1227 | while (refcount_read(&root->refs) > 1) | |
00246528 JB |
1228 | btrfs_put_root(root); |
1229 | btrfs_put_root(root); | |
bd647ce3 JB |
1230 | } |
1231 | #endif | |
1232 | } | |
1233 | ||
abed4aaa JB |
1234 | static void free_global_roots(struct btrfs_fs_info *fs_info) |
1235 | { | |
1236 | struct btrfs_root *root; | |
1237 | struct rb_node *node; | |
1238 | ||
1239 | while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { | |
1240 | root = rb_entry(node, struct btrfs_root, rb_node); | |
1241 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1242 | btrfs_put_root(root); | |
1243 | } | |
1244 | } | |
1245 | ||
0d4b0463 JB |
1246 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1247 | { | |
141386e1 JB |
1248 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1249 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1250 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1251 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1252 | btrfs_free_csum_hash(fs_info); | |
1253 | btrfs_free_stripe_hash_table(fs_info); | |
1254 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1255 | kfree(fs_info->balance_ctl); |
1256 | kfree(fs_info->delayed_root); | |
abed4aaa | 1257 | free_global_roots(fs_info); |
00246528 JB |
1258 | btrfs_put_root(fs_info->tree_root); |
1259 | btrfs_put_root(fs_info->chunk_root); | |
1260 | btrfs_put_root(fs_info->dev_root); | |
00246528 JB |
1261 | btrfs_put_root(fs_info->quota_root); |
1262 | btrfs_put_root(fs_info->uuid_root); | |
00246528 | 1263 | btrfs_put_root(fs_info->fs_root); |
aeb935a4 | 1264 | btrfs_put_root(fs_info->data_reloc_root); |
9c54e80d | 1265 | btrfs_put_root(fs_info->block_group_root); |
51502090 | 1266 | btrfs_put_root(fs_info->stripe_root); |
bd647ce3 | 1267 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1268 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1269 | kfree(fs_info->super_copy); |
1270 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1271 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1272 | kvfree(fs_info); |
1273 | } | |
1274 | ||
1275 | ||
2dfb1e43 QW |
1276 | /* |
1277 | * Get an in-memory reference of a root structure. | |
1278 | * | |
1279 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1280 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1281 | * found, then read it from disk and add it to cached fs roots. | |
1282 | * | |
1283 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1284 | * | |
1285 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1286 | * same root objectid. | |
1287 | * | |
1288 | * @objectid: root id | |
1289 | * @anon_dev: preallocated anonymous block device number for new roots, | |
1290 | * pass 0 for new allocation. | |
1291 | * @check_ref: whether to check root item references, If true, return -ENOENT | |
1292 | * for orphan roots | |
1293 | */ | |
1294 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
1295 | u64 objectid, dev_t anon_dev, | |
1296 | bool check_ref) | |
5eda7b5e CM |
1297 | { |
1298 | struct btrfs_root *root; | |
381cf658 | 1299 | struct btrfs_path *path; |
1d4c08e0 | 1300 | struct btrfs_key key; |
5eda7b5e CM |
1301 | int ret; |
1302 | ||
49d11bea JB |
1303 | root = btrfs_get_global_root(fs_info, objectid); |
1304 | if (root) | |
1305 | return root; | |
773e722a QW |
1306 | |
1307 | /* | |
1308 | * If we're called for non-subvolume trees, and above function didn't | |
1309 | * find one, do not try to read it from disk. | |
1310 | * | |
1311 | * This is namely for free-space-tree and quota tree, which can change | |
1312 | * at runtime and should only be grabbed from fs_info. | |
1313 | */ | |
1314 | if (!is_fstree(objectid) && objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) | |
1315 | return ERR_PTR(-ENOENT); | |
4df27c4d | 1316 | again: |
56e9357a | 1317 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1318 | if (root) { |
2dfb1e43 QW |
1319 | /* Shouldn't get preallocated anon_dev for cached roots */ |
1320 | ASSERT(!anon_dev); | |
bc44d7c4 | 1321 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1322 | btrfs_put_root(root); |
48475471 | 1323 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1324 | } |
5eda7b5e | 1325 | return root; |
48475471 | 1326 | } |
5eda7b5e | 1327 | |
56e9357a DS |
1328 | key.objectid = objectid; |
1329 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1330 | key.offset = (u64)-1; | |
1331 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1332 | if (IS_ERR(root)) |
1333 | return root; | |
3394e160 | 1334 | |
c00869f1 | 1335 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1336 | ret = -ENOENT; |
581bb050 | 1337 | goto fail; |
35a30d7c | 1338 | } |
581bb050 | 1339 | |
2dfb1e43 | 1340 | ret = btrfs_init_fs_root(root, anon_dev); |
ac08aedf CM |
1341 | if (ret) |
1342 | goto fail; | |
3394e160 | 1343 | |
381cf658 DS |
1344 | path = btrfs_alloc_path(); |
1345 | if (!path) { | |
1346 | ret = -ENOMEM; | |
1347 | goto fail; | |
1348 | } | |
1d4c08e0 DS |
1349 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1350 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1351 | key.offset = objectid; |
1d4c08e0 DS |
1352 | |
1353 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1354 | btrfs_free_path(path); |
d68fc57b YZ |
1355 | if (ret < 0) |
1356 | goto fail; | |
1357 | if (ret == 0) | |
27cdeb70 | 1358 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1359 | |
cb517eab | 1360 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1361 | if (ret) { |
168a2f77 JJB |
1362 | if (ret == -EEXIST) { |
1363 | btrfs_put_root(root); | |
4df27c4d | 1364 | goto again; |
168a2f77 | 1365 | } |
4df27c4d | 1366 | goto fail; |
0f7d52f4 | 1367 | } |
edbd8d4e | 1368 | return root; |
4df27c4d | 1369 | fail: |
33fab972 FM |
1370 | /* |
1371 | * If our caller provided us an anonymous device, then it's his | |
143823cf | 1372 | * responsibility to free it in case we fail. So we have to set our |
33fab972 FM |
1373 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() |
1374 | * and once again by our caller. | |
1375 | */ | |
1376 | if (anon_dev) | |
1377 | root->anon_dev = 0; | |
8c38938c | 1378 | btrfs_put_root(root); |
4df27c4d | 1379 | return ERR_PTR(ret); |
edbd8d4e CM |
1380 | } |
1381 | ||
2dfb1e43 QW |
1382 | /* |
1383 | * Get in-memory reference of a root structure | |
1384 | * | |
1385 | * @objectid: tree objectid | |
1386 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1387 | * one reference | |
1388 | */ | |
1389 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1390 | u64 objectid, bool check_ref) | |
1391 | { | |
1392 | return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); | |
1393 | } | |
1394 | ||
1395 | /* | |
1396 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1397 | * the anonymous block device id | |
1398 | * | |
1399 | * @objectid: tree objectid | |
1400 | * @anon_dev: if zero, allocate a new anonymous block device or use the | |
1401 | * parameter value | |
1402 | */ | |
1403 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
1404 | u64 objectid, dev_t anon_dev) | |
1405 | { | |
1406 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1407 | } | |
1408 | ||
49d11bea | 1409 | /* |
9580503b DS |
1410 | * Return a root for the given objectid. |
1411 | * | |
49d11bea JB |
1412 | * @fs_info: the fs_info |
1413 | * @objectid: the objectid we need to lookup | |
1414 | * | |
1415 | * This is exclusively used for backref walking, and exists specifically because | |
1416 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1417 | * creation time, which means we may have to read the tree_root in order to look | |
1418 | * up a fs root that is not in memory. If the root is not in memory we will | |
1419 | * read the tree root commit root and look up the fs root from there. This is a | |
1420 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1421 | * have the most uptodate information, it'll simply be discarded once the | |
1422 | * backref code is finished using the root. | |
1423 | */ | |
1424 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1425 | struct btrfs_path *path, | |
1426 | u64 objectid) | |
1427 | { | |
1428 | struct btrfs_root *root; | |
1429 | struct btrfs_key key; | |
1430 | ||
1431 | ASSERT(path->search_commit_root && path->skip_locking); | |
1432 | ||
1433 | /* | |
1434 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1435 | * since this is called via the backref walking code we won't be looking | |
1436 | * up a root that doesn't exist, unless there's corruption. So if root | |
1437 | * != NULL just return it. | |
1438 | */ | |
1439 | root = btrfs_get_global_root(fs_info, objectid); | |
1440 | if (root) | |
1441 | return root; | |
1442 | ||
1443 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1444 | if (root) | |
1445 | return root; | |
1446 | ||
1447 | key.objectid = objectid; | |
1448 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1449 | key.offset = (u64)-1; | |
1450 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1451 | btrfs_release_path(path); | |
1452 | ||
1453 | return root; | |
1454 | } | |
1455 | ||
a74a4b97 CM |
1456 | static int cleaner_kthread(void *arg) |
1457 | { | |
0d031dc4 | 1458 | struct btrfs_fs_info *fs_info = arg; |
d0278245 | 1459 | int again; |
a74a4b97 | 1460 | |
d6fd0ae2 | 1461 | while (1) { |
d0278245 | 1462 | again = 0; |
a74a4b97 | 1463 | |
fd340d0f JB |
1464 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1465 | ||
d0278245 | 1466 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1467 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1468 | goto sleep; |
1469 | ||
90c711ab ZB |
1470 | /* |
1471 | * Do not do anything if we might cause open_ctree() to block | |
1472 | * before we have finished mounting the filesystem. | |
1473 | */ | |
0b246afa | 1474 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1475 | goto sleep; |
1476 | ||
0b246afa | 1477 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1478 | goto sleep; |
1479 | ||
dc7f370c MX |
1480 | /* |
1481 | * Avoid the problem that we change the status of the fs | |
1482 | * during the above check and trylock. | |
1483 | */ | |
2ff7e61e | 1484 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1485 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1486 | goto sleep; |
76dda93c | 1487 | } |
a74a4b97 | 1488 | |
b7625f46 QW |
1489 | if (test_and_clear_bit(BTRFS_FS_FEATURE_CHANGED, &fs_info->flags)) |
1490 | btrfs_sysfs_feature_update(fs_info); | |
1491 | ||
2ff7e61e | 1492 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1493 | |
33c44184 | 1494 | again = btrfs_clean_one_deleted_snapshot(fs_info); |
0b246afa | 1495 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1496 | |
1497 | /* | |
05323cd1 MX |
1498 | * The defragger has dealt with the R/O remount and umount, |
1499 | * needn't do anything special here. | |
d0278245 | 1500 | */ |
0b246afa | 1501 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1502 | |
1503 | /* | |
f3372065 | 1504 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
1505 | * with relocation (btrfs_relocate_chunk) and relocation |
1506 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 1507 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
1508 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
1509 | * unused block groups. | |
1510 | */ | |
0b246afa | 1511 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
1512 | |
1513 | /* | |
1514 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
1515 | * all unused block_groups. This possibly gives us some more free | |
1516 | * space. | |
1517 | */ | |
1518 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 1519 | sleep: |
a0a1db70 | 1520 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1521 | if (kthread_should_park()) |
1522 | kthread_parkme(); | |
1523 | if (kthread_should_stop()) | |
1524 | return 0; | |
838fe188 | 1525 | if (!again) { |
a74a4b97 | 1526 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1527 | schedule(); |
a74a4b97 CM |
1528 | __set_current_state(TASK_RUNNING); |
1529 | } | |
da288d28 | 1530 | } |
a74a4b97 CM |
1531 | } |
1532 | ||
1533 | static int transaction_kthread(void *arg) | |
1534 | { | |
1535 | struct btrfs_root *root = arg; | |
0b246afa | 1536 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1537 | struct btrfs_trans_handle *trans; |
1538 | struct btrfs_transaction *cur; | |
8929ecfa | 1539 | u64 transid; |
643900be | 1540 | time64_t delta; |
a74a4b97 | 1541 | unsigned long delay; |
914b2007 | 1542 | bool cannot_commit; |
a74a4b97 CM |
1543 | |
1544 | do { | |
914b2007 | 1545 | cannot_commit = false; |
ba1bc00f | 1546 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 1547 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 1548 | |
0b246afa JM |
1549 | spin_lock(&fs_info->trans_lock); |
1550 | cur = fs_info->running_transaction; | |
a74a4b97 | 1551 | if (!cur) { |
0b246afa | 1552 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1553 | goto sleep; |
1554 | } | |
31153d81 | 1555 | |
643900be | 1556 | delta = ktime_get_seconds() - cur->start_time; |
fdfbf020 | 1557 | if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && |
77d20c68 | 1558 | cur->state < TRANS_STATE_COMMIT_PREP && |
643900be | 1559 | delta < fs_info->commit_interval) { |
0b246afa | 1560 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
1561 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
1562 | delay = min(delay, | |
1563 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
1564 | goto sleep; |
1565 | } | |
8929ecfa | 1566 | transid = cur->transid; |
0b246afa | 1567 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1568 | |
79787eaa | 1569 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1570 | trans = btrfs_attach_transaction(root); |
914b2007 | 1571 | if (IS_ERR(trans)) { |
354aa0fb MX |
1572 | if (PTR_ERR(trans) != -ENOENT) |
1573 | cannot_commit = true; | |
79787eaa | 1574 | goto sleep; |
914b2007 | 1575 | } |
8929ecfa | 1576 | if (transid == trans->transid) { |
3a45bb20 | 1577 | btrfs_commit_transaction(trans); |
8929ecfa | 1578 | } else { |
3a45bb20 | 1579 | btrfs_end_transaction(trans); |
8929ecfa | 1580 | } |
a74a4b97 | 1581 | sleep: |
0b246afa JM |
1582 | wake_up_process(fs_info->cleaner_kthread); |
1583 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1584 | |
84961539 | 1585 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 1586 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1587 | if (!kthread_should_stop() && |
0b246afa | 1588 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1589 | cannot_commit)) |
bc5511d0 | 1590 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1591 | } while (!kthread_should_stop()); |
1592 | return 0; | |
1593 | } | |
1594 | ||
af31f5e5 | 1595 | /* |
01f0f9da NB |
1596 | * This will find the highest generation in the array of root backups. The |
1597 | * index of the highest array is returned, or -EINVAL if we can't find | |
1598 | * anything. | |
af31f5e5 CM |
1599 | * |
1600 | * We check to make sure the array is valid by comparing the | |
1601 | * generation of the latest root in the array with the generation | |
1602 | * in the super block. If they don't match we pitch it. | |
1603 | */ | |
01f0f9da | 1604 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 1605 | { |
01f0f9da | 1606 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 1607 | u64 cur; |
af31f5e5 CM |
1608 | struct btrfs_root_backup *root_backup; |
1609 | int i; | |
1610 | ||
1611 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1612 | root_backup = info->super_copy->super_roots + i; | |
1613 | cur = btrfs_backup_tree_root_gen(root_backup); | |
1614 | if (cur == newest_gen) | |
01f0f9da | 1615 | return i; |
af31f5e5 CM |
1616 | } |
1617 | ||
01f0f9da | 1618 | return -EINVAL; |
af31f5e5 CM |
1619 | } |
1620 | ||
af31f5e5 CM |
1621 | /* |
1622 | * copy all the root pointers into the super backup array. | |
1623 | * this will bump the backup pointer by one when it is | |
1624 | * done | |
1625 | */ | |
1626 | static void backup_super_roots(struct btrfs_fs_info *info) | |
1627 | { | |
6ef108dd | 1628 | const int next_backup = info->backup_root_index; |
af31f5e5 | 1629 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
1630 | |
1631 | root_backup = info->super_for_commit->super_roots + next_backup; | |
1632 | ||
1633 | /* | |
1634 | * make sure all of our padding and empty slots get zero filled | |
1635 | * regardless of which ones we use today | |
1636 | */ | |
1637 | memset(root_backup, 0, sizeof(*root_backup)); | |
1638 | ||
1639 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
1640 | ||
1641 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
1642 | btrfs_set_backup_tree_root_gen(root_backup, | |
1643 | btrfs_header_generation(info->tree_root->node)); | |
1644 | ||
1645 | btrfs_set_backup_tree_root_level(root_backup, | |
1646 | btrfs_header_level(info->tree_root->node)); | |
1647 | ||
1648 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
1649 | btrfs_set_backup_chunk_root_gen(root_backup, | |
1650 | btrfs_header_generation(info->chunk_root->node)); | |
1651 | btrfs_set_backup_chunk_root_level(root_backup, | |
1652 | btrfs_header_level(info->chunk_root->node)); | |
1653 | ||
1c56ab99 | 1654 | if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { |
9c54e80d | 1655 | struct btrfs_root *extent_root = btrfs_extent_root(info, 0); |
f7238e50 | 1656 | struct btrfs_root *csum_root = btrfs_csum_root(info, 0); |
9c54e80d JB |
1657 | |
1658 | btrfs_set_backup_extent_root(root_backup, | |
1659 | extent_root->node->start); | |
1660 | btrfs_set_backup_extent_root_gen(root_backup, | |
1661 | btrfs_header_generation(extent_root->node)); | |
1662 | btrfs_set_backup_extent_root_level(root_backup, | |
1663 | btrfs_header_level(extent_root->node)); | |
f7238e50 JB |
1664 | |
1665 | btrfs_set_backup_csum_root(root_backup, csum_root->node->start); | |
1666 | btrfs_set_backup_csum_root_gen(root_backup, | |
1667 | btrfs_header_generation(csum_root->node)); | |
1668 | btrfs_set_backup_csum_root_level(root_backup, | |
1669 | btrfs_header_level(csum_root->node)); | |
9c54e80d | 1670 | } |
af31f5e5 | 1671 | |
7c7e82a7 CM |
1672 | /* |
1673 | * we might commit during log recovery, which happens before we set | |
1674 | * the fs_root. Make sure it is valid before we fill it in. | |
1675 | */ | |
1676 | if (info->fs_root && info->fs_root->node) { | |
1677 | btrfs_set_backup_fs_root(root_backup, | |
1678 | info->fs_root->node->start); | |
1679 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 1680 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 1681 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 1682 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 1683 | } |
af31f5e5 CM |
1684 | |
1685 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
1686 | btrfs_set_backup_dev_root_gen(root_backup, | |
1687 | btrfs_header_generation(info->dev_root->node)); | |
1688 | btrfs_set_backup_dev_root_level(root_backup, | |
1689 | btrfs_header_level(info->dev_root->node)); | |
1690 | ||
af31f5e5 CM |
1691 | btrfs_set_backup_total_bytes(root_backup, |
1692 | btrfs_super_total_bytes(info->super_copy)); | |
1693 | btrfs_set_backup_bytes_used(root_backup, | |
1694 | btrfs_super_bytes_used(info->super_copy)); | |
1695 | btrfs_set_backup_num_devices(root_backup, | |
1696 | btrfs_super_num_devices(info->super_copy)); | |
1697 | ||
1698 | /* | |
1699 | * if we don't copy this out to the super_copy, it won't get remembered | |
1700 | * for the next commit | |
1701 | */ | |
1702 | memcpy(&info->super_copy->super_roots, | |
1703 | &info->super_for_commit->super_roots, | |
1704 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
1705 | } | |
1706 | ||
bd2336b2 | 1707 | /* |
9580503b DS |
1708 | * Reads a backup root based on the passed priority. Prio 0 is the newest, prio |
1709 | * 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
bd2336b2 | 1710 | * |
9580503b DS |
1711 | * @fs_info: filesystem whose backup roots need to be read |
1712 | * @priority: priority of backup root required | |
bd2336b2 NB |
1713 | * |
1714 | * Returns backup root index on success and -EINVAL otherwise. | |
1715 | */ | |
1716 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
1717 | { | |
1718 | int backup_index = find_newest_super_backup(fs_info); | |
1719 | struct btrfs_super_block *super = fs_info->super_copy; | |
1720 | struct btrfs_root_backup *root_backup; | |
1721 | ||
1722 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
1723 | if (priority == 0) | |
1724 | return backup_index; | |
1725 | ||
1726 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
1727 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
1728 | } else { | |
1729 | return -EINVAL; | |
1730 | } | |
1731 | ||
1732 | root_backup = super->super_roots + backup_index; | |
1733 | ||
1734 | btrfs_set_super_generation(super, | |
1735 | btrfs_backup_tree_root_gen(root_backup)); | |
1736 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
1737 | btrfs_set_super_root_level(super, | |
1738 | btrfs_backup_tree_root_level(root_backup)); | |
1739 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
1740 | ||
1741 | /* | |
1742 | * Fixme: the total bytes and num_devices need to match or we should | |
1743 | * need a fsck | |
1744 | */ | |
1745 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
1746 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
1747 | ||
1748 | return backup_index; | |
1749 | } | |
1750 | ||
7abadb64 LB |
1751 | /* helper to cleanup workers */ |
1752 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
1753 | { | |
dc6e3209 | 1754 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 1755 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
5cdc7ad3 | 1756 | btrfs_destroy_workqueue(fs_info->workers); |
d7b9416f CH |
1757 | if (fs_info->endio_workers) |
1758 | destroy_workqueue(fs_info->endio_workers); | |
385de0ef CH |
1759 | if (fs_info->rmw_workers) |
1760 | destroy_workqueue(fs_info->rmw_workers); | |
fed8a72d CH |
1761 | if (fs_info->compressed_write_workers) |
1762 | destroy_workqueue(fs_info->compressed_write_workers); | |
fccb5d86 QW |
1763 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
1764 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 1765 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 1766 | btrfs_destroy_workqueue(fs_info->caching_workers); |
a44903ab | 1767 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 1768 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
1769 | if (fs_info->discard_ctl.discard_workers) |
1770 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
1771 | /* |
1772 | * Now that all other work queues are destroyed, we can safely destroy | |
1773 | * the queues used for metadata I/O, since tasks from those other work | |
1774 | * queues can do metadata I/O operations. | |
1775 | */ | |
d7b9416f CH |
1776 | if (fs_info->endio_meta_workers) |
1777 | destroy_workqueue(fs_info->endio_meta_workers); | |
7abadb64 LB |
1778 | } |
1779 | ||
2e9f5954 R |
1780 | static void free_root_extent_buffers(struct btrfs_root *root) |
1781 | { | |
1782 | if (root) { | |
1783 | free_extent_buffer(root->node); | |
1784 | free_extent_buffer(root->commit_root); | |
1785 | root->node = NULL; | |
1786 | root->commit_root = NULL; | |
1787 | } | |
1788 | } | |
1789 | ||
abed4aaa JB |
1790 | static void free_global_root_pointers(struct btrfs_fs_info *fs_info) |
1791 | { | |
1792 | struct btrfs_root *root, *tmp; | |
1793 | ||
1794 | rbtree_postorder_for_each_entry_safe(root, tmp, | |
1795 | &fs_info->global_root_tree, | |
1796 | rb_node) | |
1797 | free_root_extent_buffers(root); | |
1798 | } | |
1799 | ||
af31f5e5 | 1800 | /* helper to cleanup tree roots */ |
4273eaff | 1801 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 1802 | { |
2e9f5954 | 1803 | free_root_extent_buffers(info->tree_root); |
655b09fe | 1804 | |
abed4aaa | 1805 | free_global_root_pointers(info); |
2e9f5954 | 1806 | free_root_extent_buffers(info->dev_root); |
2e9f5954 R |
1807 | free_root_extent_buffers(info->quota_root); |
1808 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 1809 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 1810 | free_root_extent_buffers(info->data_reloc_root); |
9c54e80d | 1811 | free_root_extent_buffers(info->block_group_root); |
51502090 | 1812 | free_root_extent_buffers(info->stripe_root); |
4273eaff | 1813 | if (free_chunk_root) |
2e9f5954 | 1814 | free_root_extent_buffers(info->chunk_root); |
af31f5e5 CM |
1815 | } |
1816 | ||
8c38938c JB |
1817 | void btrfs_put_root(struct btrfs_root *root) |
1818 | { | |
1819 | if (!root) | |
1820 | return; | |
1821 | ||
1822 | if (refcount_dec_and_test(&root->refs)) { | |
1823 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 1824 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
1825 | if (root->anon_dev) |
1826 | free_anon_bdev(root->anon_dev); | |
923eb523 | 1827 | free_root_extent_buffers(root); |
8c38938c | 1828 | #ifdef CONFIG_BTRFS_DEBUG |
fc7cbcd4 | 1829 | spin_lock(&root->fs_info->fs_roots_radix_lock); |
8c38938c | 1830 | list_del_init(&root->leak_list); |
fc7cbcd4 | 1831 | spin_unlock(&root->fs_info->fs_roots_radix_lock); |
8c38938c JB |
1832 | #endif |
1833 | kfree(root); | |
1834 | } | |
1835 | } | |
1836 | ||
faa2dbf0 | 1837 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 | 1838 | { |
fc7cbcd4 DS |
1839 | int ret; |
1840 | struct btrfs_root *gang[8]; | |
1841 | int i; | |
171f6537 JB |
1842 | |
1843 | while (!list_empty(&fs_info->dead_roots)) { | |
fc7cbcd4 DS |
1844 | gang[0] = list_entry(fs_info->dead_roots.next, |
1845 | struct btrfs_root, root_list); | |
1846 | list_del(&gang[0]->root_list); | |
171f6537 | 1847 | |
fc7cbcd4 DS |
1848 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
1849 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); | |
1850 | btrfs_put_root(gang[0]); | |
171f6537 JB |
1851 | } |
1852 | ||
fc7cbcd4 DS |
1853 | while (1) { |
1854 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
1855 | (void **)gang, 0, | |
1856 | ARRAY_SIZE(gang)); | |
1857 | if (!ret) | |
1858 | break; | |
1859 | for (i = 0; i < ret; i++) | |
1860 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); | |
171f6537 JB |
1861 | } |
1862 | } | |
af31f5e5 | 1863 | |
638aa7ed ES |
1864 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
1865 | { | |
1866 | mutex_init(&fs_info->scrub_lock); | |
1867 | atomic_set(&fs_info->scrubs_running, 0); | |
1868 | atomic_set(&fs_info->scrub_pause_req, 0); | |
1869 | atomic_set(&fs_info->scrubs_paused, 0); | |
1870 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
1871 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 1872 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
1873 | } |
1874 | ||
779a65a4 ES |
1875 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
1876 | { | |
1877 | spin_lock_init(&fs_info->balance_lock); | |
1878 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
1879 | atomic_set(&fs_info->balance_pause_req, 0); |
1880 | atomic_set(&fs_info->balance_cancel_req, 0); | |
1881 | fs_info->balance_ctl = NULL; | |
1882 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 1883 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
1884 | } |
1885 | ||
dcb2137c | 1886 | static int btrfs_init_btree_inode(struct super_block *sb) |
f37938e0 | 1887 | { |
dcb2137c | 1888 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); |
e256927b JB |
1889 | unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, |
1890 | fs_info->tree_root); | |
dcb2137c CH |
1891 | struct inode *inode; |
1892 | ||
1893 | inode = new_inode(sb); | |
1894 | if (!inode) | |
1895 | return -ENOMEM; | |
2ff7e61e JM |
1896 | |
1897 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
1898 | set_nlink(inode, 1); | |
f37938e0 ES |
1899 | /* |
1900 | * we set the i_size on the btree inode to the max possible int. | |
1901 | * the real end of the address space is determined by all of | |
1902 | * the devices in the system | |
1903 | */ | |
2ff7e61e JM |
1904 | inode->i_size = OFFSET_MAX; |
1905 | inode->i_mapping->a_ops = &btree_aops; | |
dcb2137c | 1906 | mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS); |
f37938e0 | 1907 | |
2ff7e61e | 1908 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 1909 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
35da5a7e | 1910 | IO_TREE_BTREE_INODE_IO); |
2ff7e61e | 1911 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 1912 | |
5c8fd99f | 1913 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
adac5584 FM |
1914 | BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; |
1915 | BTRFS_I(inode)->location.type = 0; | |
1916 | BTRFS_I(inode)->location.offset = 0; | |
2ff7e61e | 1917 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); |
e256927b | 1918 | __insert_inode_hash(inode, hash); |
dcb2137c CH |
1919 | fs_info->btree_inode = inode; |
1920 | ||
1921 | return 0; | |
f37938e0 ES |
1922 | } |
1923 | ||
ad618368 ES |
1924 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
1925 | { | |
ad618368 | 1926 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 1927 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 1928 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
1929 | } |
1930 | ||
f9e92e40 ES |
1931 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
1932 | { | |
1933 | spin_lock_init(&fs_info->qgroup_lock); | |
1934 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
1935 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
1936 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
1937 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 1938 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 1939 | fs_info->qgroup_rescan_running = false; |
011b46c3 | 1940 | fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; |
f9e92e40 ES |
1941 | mutex_init(&fs_info->qgroup_rescan_lock); |
1942 | } | |
1943 | ||
d21deec5 | 1944 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) |
2a458198 | 1945 | { |
f7b885be | 1946 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 1947 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
58e814fc | 1948 | unsigned int ordered_flags = WQ_MEM_RECLAIM | WQ_FREEZABLE; |
2a458198 ES |
1949 | |
1950 | fs_info->workers = | |
a31b4a43 | 1951 | btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); |
2a458198 ES |
1952 | |
1953 | fs_info->delalloc_workers = | |
cb001095 JM |
1954 | btrfs_alloc_workqueue(fs_info, "delalloc", |
1955 | flags, max_active, 2); | |
2a458198 ES |
1956 | |
1957 | fs_info->flush_workers = | |
cb001095 JM |
1958 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
1959 | flags, max_active, 0); | |
2a458198 ES |
1960 | |
1961 | fs_info->caching_workers = | |
cb001095 | 1962 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 1963 | |
2a458198 | 1964 | fs_info->fixup_workers = |
58e814fc | 1965 | btrfs_alloc_ordered_workqueue(fs_info, "fixup", ordered_flags); |
2a458198 | 1966 | |
2a458198 | 1967 | fs_info->endio_workers = |
d7b9416f | 1968 | alloc_workqueue("btrfs-endio", flags, max_active); |
2a458198 | 1969 | fs_info->endio_meta_workers = |
d7b9416f | 1970 | alloc_workqueue("btrfs-endio-meta", flags, max_active); |
385de0ef | 1971 | fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); |
2a458198 | 1972 | fs_info->endio_write_workers = |
cb001095 JM |
1973 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
1974 | max_active, 2); | |
fed8a72d CH |
1975 | fs_info->compressed_write_workers = |
1976 | alloc_workqueue("btrfs-compressed-write", flags, max_active); | |
2a458198 | 1977 | fs_info->endio_freespace_worker = |
cb001095 JM |
1978 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
1979 | max_active, 0); | |
2a458198 | 1980 | fs_info->delayed_workers = |
cb001095 JM |
1981 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
1982 | max_active, 0); | |
2a458198 | 1983 | fs_info->qgroup_rescan_workers = |
58e814fc TH |
1984 | btrfs_alloc_ordered_workqueue(fs_info, "qgroup-rescan", |
1985 | ordered_flags); | |
b0643e59 | 1986 | fs_info->discard_ctl.discard_workers = |
58e814fc | 1987 | alloc_ordered_workqueue("btrfs_discard", WQ_FREEZABLE); |
2a458198 | 1988 | |
8bfec2e4 | 1989 | if (!(fs_info->workers && |
a31b4a43 | 1990 | fs_info->delalloc_workers && fs_info->flush_workers && |
2a458198 | 1991 | fs_info->endio_workers && fs_info->endio_meta_workers && |
fed8a72d | 1992 | fs_info->compressed_write_workers && |
1a1a2851 | 1993 | fs_info->endio_write_workers && |
2a458198 | 1994 | fs_info->endio_freespace_worker && fs_info->rmw_workers && |
f26c9238 QW |
1995 | fs_info->caching_workers && fs_info->fixup_workers && |
1996 | fs_info->delayed_workers && fs_info->qgroup_rescan_workers && | |
b0643e59 | 1997 | fs_info->discard_ctl.discard_workers)) { |
2a458198 ES |
1998 | return -ENOMEM; |
1999 | } | |
2000 | ||
2001 | return 0; | |
2002 | } | |
2003 | ||
6d97c6e3 JT |
2004 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2005 | { | |
2006 | struct crypto_shash *csum_shash; | |
b4e967be | 2007 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2008 | |
b4e967be | 2009 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2010 | |
2011 | if (IS_ERR(csum_shash)) { | |
2012 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2013 | csum_driver); |
6d97c6e3 JT |
2014 | return PTR_ERR(csum_shash); |
2015 | } | |
2016 | ||
2017 | fs_info->csum_shash = csum_shash; | |
2018 | ||
68d99ab0 CH |
2019 | /* |
2020 | * Check if the checksum implementation is a fast accelerated one. | |
2021 | * As-is this is a bit of a hack and should be replaced once the csum | |
2022 | * implementations provide that information themselves. | |
2023 | */ | |
2024 | switch (csum_type) { | |
2025 | case BTRFS_CSUM_TYPE_CRC32: | |
2026 | if (!strstr(crypto_shash_driver_name(csum_shash), "generic")) | |
2027 | set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); | |
2028 | break; | |
efcfcbc6 DS |
2029 | case BTRFS_CSUM_TYPE_XXHASH: |
2030 | set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags); | |
2031 | break; | |
68d99ab0 CH |
2032 | default: |
2033 | break; | |
2034 | } | |
2035 | ||
c8a5f8ca DS |
2036 | btrfs_info(fs_info, "using %s (%s) checksum algorithm", |
2037 | btrfs_super_csum_name(csum_type), | |
2038 | crypto_shash_driver_name(csum_shash)); | |
6d97c6e3 JT |
2039 | return 0; |
2040 | } | |
2041 | ||
63443bf5 ES |
2042 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2043 | struct btrfs_fs_devices *fs_devices) | |
2044 | { | |
2045 | int ret; | |
789d6a3a | 2046 | struct btrfs_tree_parent_check check = { 0 }; |
63443bf5 ES |
2047 | struct btrfs_root *log_tree_root; |
2048 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2049 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2050 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2051 | |
2052 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2053 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2054 | return -EIO; |
2055 | } | |
2056 | ||
96dfcb46 JB |
2057 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2058 | GFP_KERNEL); | |
63443bf5 ES |
2059 | if (!log_tree_root) |
2060 | return -ENOMEM; | |
2061 | ||
789d6a3a QW |
2062 | check.level = level; |
2063 | check.transid = fs_info->generation + 1; | |
2064 | check.owner_root = BTRFS_TREE_LOG_OBJECTID; | |
2065 | log_tree_root->node = read_tree_block(fs_info, bytenr, &check); | |
64c043de | 2066 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2067 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2068 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2069 | log_tree_root->node = NULL; |
00246528 | 2070 | btrfs_put_root(log_tree_root); |
0eeff236 | 2071 | return ret; |
4eb150d6 QW |
2072 | } |
2073 | if (!extent_buffer_uptodate(log_tree_root->node)) { | |
f14d104d | 2074 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2075 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2076 | return -EIO; |
2077 | } | |
4eb150d6 | 2078 | |
63443bf5 ES |
2079 | /* returns with log_tree_root freed on success */ |
2080 | ret = btrfs_recover_log_trees(log_tree_root); | |
2081 | if (ret) { | |
0b246afa JM |
2082 | btrfs_handle_fs_error(fs_info, ret, |
2083 | "Failed to recover log tree"); | |
00246528 | 2084 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2085 | return ret; |
2086 | } | |
2087 | ||
bc98a42c | 2088 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2089 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2090 | if (ret) |
2091 | return ret; | |
2092 | } | |
2093 | ||
2094 | return 0; | |
2095 | } | |
2096 | ||
abed4aaa JB |
2097 | static int load_global_roots_objectid(struct btrfs_root *tree_root, |
2098 | struct btrfs_path *path, u64 objectid, | |
2099 | const char *name) | |
2100 | { | |
2101 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
2102 | struct btrfs_root *root; | |
f7238e50 | 2103 | u64 max_global_id = 0; |
abed4aaa JB |
2104 | int ret; |
2105 | struct btrfs_key key = { | |
2106 | .objectid = objectid, | |
2107 | .type = BTRFS_ROOT_ITEM_KEY, | |
2108 | .offset = 0, | |
2109 | }; | |
2110 | bool found = false; | |
2111 | ||
2112 | /* If we have IGNOREDATACSUMS skip loading these roots. */ | |
2113 | if (objectid == BTRFS_CSUM_TREE_OBJECTID && | |
2114 | btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2115 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2116 | return 0; | |
2117 | } | |
2118 | ||
2119 | while (1) { | |
2120 | ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); | |
2121 | if (ret < 0) | |
2122 | break; | |
2123 | ||
2124 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
2125 | ret = btrfs_next_leaf(tree_root, path); | |
2126 | if (ret) { | |
2127 | if (ret > 0) | |
2128 | ret = 0; | |
2129 | break; | |
2130 | } | |
2131 | } | |
2132 | ret = 0; | |
2133 | ||
2134 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
2135 | if (key.objectid != objectid) | |
2136 | break; | |
2137 | btrfs_release_path(path); | |
2138 | ||
f7238e50 JB |
2139 | /* |
2140 | * Just worry about this for extent tree, it'll be the same for | |
2141 | * everybody. | |
2142 | */ | |
2143 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
2144 | max_global_id = max(max_global_id, key.offset); | |
2145 | ||
abed4aaa JB |
2146 | found = true; |
2147 | root = read_tree_root_path(tree_root, path, &key); | |
2148 | if (IS_ERR(root)) { | |
2149 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2150 | ret = PTR_ERR(root); | |
2151 | break; | |
2152 | } | |
2153 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2154 | ret = btrfs_global_root_insert(root); | |
2155 | if (ret) { | |
2156 | btrfs_put_root(root); | |
2157 | break; | |
2158 | } | |
2159 | key.offset++; | |
2160 | } | |
2161 | btrfs_release_path(path); | |
2162 | ||
f7238e50 JB |
2163 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) |
2164 | fs_info->nr_global_roots = max_global_id + 1; | |
2165 | ||
abed4aaa JB |
2166 | if (!found || ret) { |
2167 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
2168 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2169 | ||
2170 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2171 | ret = ret ? ret : -ENOENT; | |
2172 | else | |
2173 | ret = 0; | |
2174 | btrfs_err(fs_info, "failed to load root %s", name); | |
2175 | } | |
2176 | return ret; | |
2177 | } | |
2178 | ||
2179 | static int load_global_roots(struct btrfs_root *tree_root) | |
2180 | { | |
2181 | struct btrfs_path *path; | |
2182 | int ret = 0; | |
2183 | ||
2184 | path = btrfs_alloc_path(); | |
2185 | if (!path) | |
2186 | return -ENOMEM; | |
2187 | ||
2188 | ret = load_global_roots_objectid(tree_root, path, | |
2189 | BTRFS_EXTENT_TREE_OBJECTID, "extent"); | |
2190 | if (ret) | |
2191 | goto out; | |
2192 | ret = load_global_roots_objectid(tree_root, path, | |
2193 | BTRFS_CSUM_TREE_OBJECTID, "csum"); | |
2194 | if (ret) | |
2195 | goto out; | |
2196 | if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) | |
2197 | goto out; | |
2198 | ret = load_global_roots_objectid(tree_root, path, | |
2199 | BTRFS_FREE_SPACE_TREE_OBJECTID, | |
2200 | "free space"); | |
2201 | out: | |
2202 | btrfs_free_path(path); | |
2203 | return ret; | |
2204 | } | |
2205 | ||
6bccf3ab | 2206 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2207 | { |
6bccf3ab | 2208 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2209 | struct btrfs_root *root; |
4bbcaa64 ES |
2210 | struct btrfs_key location; |
2211 | int ret; | |
2212 | ||
6bccf3ab JM |
2213 | BUG_ON(!fs_info->tree_root); |
2214 | ||
abed4aaa JB |
2215 | ret = load_global_roots(tree_root); |
2216 | if (ret) | |
2217 | return ret; | |
2218 | ||
4bbcaa64 ES |
2219 | location.type = BTRFS_ROOT_ITEM_KEY; |
2220 | location.offset = 0; | |
2221 | ||
1c56ab99 | 2222 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { |
14033b08 QW |
2223 | location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; |
2224 | root = btrfs_read_tree_root(tree_root, &location); | |
2225 | if (IS_ERR(root)) { | |
2226 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2227 | ret = PTR_ERR(root); | |
2228 | goto out; | |
2229 | } | |
2230 | } else { | |
2231 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2232 | fs_info->block_group_root = root; | |
2233 | } | |
2234 | } | |
2235 | ||
2236 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2237 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2238 | if (IS_ERR(root)) { |
42437a63 JB |
2239 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2240 | ret = PTR_ERR(root); | |
2241 | goto out; | |
2242 | } | |
2243 | } else { | |
2244 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2245 | fs_info->dev_root = root; | |
f50f4353 | 2246 | } |
820a49da | 2247 | /* Initialize fs_info for all devices in any case */ |
a8d1b164 JT |
2248 | ret = btrfs_init_devices_late(fs_info); |
2249 | if (ret) | |
2250 | goto out; | |
4bbcaa64 | 2251 | |
aeb935a4 QW |
2252 | /* |
2253 | * This tree can share blocks with some other fs tree during relocation | |
2254 | * and we need a proper setup by btrfs_get_fs_root | |
2255 | */ | |
56e9357a DS |
2256 | root = btrfs_get_fs_root(tree_root->fs_info, |
2257 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2258 | if (IS_ERR(root)) { |
42437a63 JB |
2259 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2260 | ret = PTR_ERR(root); | |
2261 | goto out; | |
2262 | } | |
2263 | } else { | |
2264 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2265 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2266 | } |
aeb935a4 | 2267 | |
4bbcaa64 | 2268 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2269 | root = btrfs_read_tree_root(tree_root, &location); |
2270 | if (!IS_ERR(root)) { | |
2271 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
a4f3d2c4 | 2272 | fs_info->quota_root = root; |
4bbcaa64 ES |
2273 | } |
2274 | ||
2275 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2276 | root = btrfs_read_tree_root(tree_root, &location); |
2277 | if (IS_ERR(root)) { | |
42437a63 JB |
2278 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2279 | ret = PTR_ERR(root); | |
2280 | if (ret != -ENOENT) | |
2281 | goto out; | |
2282 | } | |
4bbcaa64 | 2283 | } else { |
a4f3d2c4 DS |
2284 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2285 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2286 | } |
2287 | ||
51502090 JT |
2288 | if (btrfs_fs_incompat(fs_info, RAID_STRIPE_TREE)) { |
2289 | location.objectid = BTRFS_RAID_STRIPE_TREE_OBJECTID; | |
2290 | root = btrfs_read_tree_root(tree_root, &location); | |
2291 | if (IS_ERR(root)) { | |
2292 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2293 | ret = PTR_ERR(root); | |
2294 | goto out; | |
2295 | } | |
2296 | } else { | |
2297 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2298 | fs_info->stripe_root = root; | |
2299 | } | |
2300 | } | |
2301 | ||
4bbcaa64 | 2302 | return 0; |
f50f4353 LB |
2303 | out: |
2304 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2305 | location.objectid, ret); | |
2306 | return ret; | |
4bbcaa64 ES |
2307 | } |
2308 | ||
069ec957 QW |
2309 | /* |
2310 | * Real super block validation | |
2311 | * NOTE: super csum type and incompat features will not be checked here. | |
2312 | * | |
2313 | * @sb: super block to check | |
2314 | * @mirror_num: the super block number to check its bytenr: | |
2315 | * 0 the primary (1st) sb | |
2316 | * 1, 2 2nd and 3rd backup copy | |
2317 | * -1 skip bytenr check | |
2318 | */ | |
a05d3c91 QW |
2319 | int btrfs_validate_super(struct btrfs_fs_info *fs_info, |
2320 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2321 | { |
21a852b0 QW |
2322 | u64 nodesize = btrfs_super_nodesize(sb); |
2323 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2324 | int ret = 0; | |
2325 | ||
2326 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2327 | btrfs_err(fs_info, "no valid FS found"); | |
2328 | ret = -EINVAL; | |
2329 | } | |
2330 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2331 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2332 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2333 | ret = -EINVAL; | |
2334 | } | |
2335 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2336 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2337 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2338 | ret = -EINVAL; | |
2339 | } | |
2340 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2341 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2342 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2343 | ret = -EINVAL; | |
2344 | } | |
2345 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2346 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2347 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2348 | ret = -EINVAL; | |
2349 | } | |
2350 | ||
2351 | /* | |
2352 | * Check sectorsize and nodesize first, other check will need it. | |
2353 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2354 | */ | |
2355 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2356 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2357 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2358 | ret = -EINVAL; | |
2359 | } | |
0bb3eb3e QW |
2360 | |
2361 | /* | |
1a42daab QW |
2362 | * We only support at most two sectorsizes: 4K and PAGE_SIZE. |
2363 | * | |
2364 | * We can support 16K sectorsize with 64K page size without problem, | |
2365 | * but such sectorsize/pagesize combination doesn't make much sense. | |
2366 | * 4K will be our future standard, PAGE_SIZE is supported from the very | |
2367 | * beginning. | |
0bb3eb3e | 2368 | */ |
1a42daab | 2369 | if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { |
21a852b0 | 2370 | btrfs_err(fs_info, |
0bb3eb3e | 2371 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2372 | sectorsize, PAGE_SIZE); |
2373 | ret = -EINVAL; | |
2374 | } | |
0bb3eb3e | 2375 | |
21a852b0 QW |
2376 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2377 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2378 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2379 | ret = -EINVAL; | |
2380 | } | |
2381 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2382 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2383 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2384 | ret = -EINVAL; | |
2385 | } | |
2386 | ||
2387 | /* Root alignment check */ | |
2388 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2389 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2390 | btrfs_super_root(sb)); | |
2391 | ret = -EINVAL; | |
2392 | } | |
2393 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2394 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2395 | btrfs_super_chunk_root(sb)); | |
2396 | ret = -EINVAL; | |
2397 | } | |
2398 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2399 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2400 | btrfs_super_log_root(sb)); | |
2401 | ret = -EINVAL; | |
2402 | } | |
2403 | ||
a5b8a5f9 AJ |
2404 | if (!fs_info->fs_devices->temp_fsid && |
2405 | memcmp(fs_info->fs_devices->fsid, sb->fsid, BTRFS_FSID_SIZE) != 0) { | |
aefd7f70 NB |
2406 | btrfs_err(fs_info, |
2407 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
d167aa76 | 2408 | sb->fsid, fs_info->fs_devices->fsid); |
aefd7f70 NB |
2409 | ret = -EINVAL; |
2410 | } | |
2411 | ||
6bfe3959 AJ |
2412 | if (memcmp(fs_info->fs_devices->metadata_uuid, btrfs_sb_fsid_ptr(sb), |
2413 | BTRFS_FSID_SIZE) != 0) { | |
aefd7f70 NB |
2414 | btrfs_err(fs_info, |
2415 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
6bfe3959 | 2416 | btrfs_sb_fsid_ptr(sb), fs_info->fs_devices->metadata_uuid); |
aefd7f70 NB |
2417 | ret = -EINVAL; |
2418 | } | |
2419 | ||
25984a5a AJ |
2420 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
2421 | BTRFS_FSID_SIZE) != 0) { | |
2422 | btrfs_err(fs_info, | |
2423 | "dev_item UUID does not match metadata fsid: %pU != %pU", | |
2424 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); | |
2425 | ret = -EINVAL; | |
2426 | } | |
2427 | ||
1c56ab99 QW |
2428 | /* |
2429 | * Artificial requirement for block-group-tree to force newer features | |
2430 | * (free-space-tree, no-holes) so the test matrix is smaller. | |
2431 | */ | |
2432 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
2433 | (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || | |
2434 | !btrfs_fs_incompat(fs_info, NO_HOLES))) { | |
2435 | btrfs_err(fs_info, | |
2436 | "block-group-tree feature requires fres-space-tree and no-holes"); | |
2437 | ret = -EINVAL; | |
2438 | } | |
2439 | ||
21a852b0 QW |
2440 | /* |
2441 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2442 | * done later | |
2443 | */ | |
2444 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2445 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2446 | btrfs_super_bytes_used(sb)); | |
2447 | ret = -EINVAL; | |
2448 | } | |
2449 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2450 | btrfs_err(fs_info, "invalid stripesize %u", | |
2451 | btrfs_super_stripesize(sb)); | |
2452 | ret = -EINVAL; | |
2453 | } | |
2454 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2455 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2456 | btrfs_super_num_devices(sb)); | |
2457 | if (btrfs_super_num_devices(sb) == 0) { | |
2458 | btrfs_err(fs_info, "number of devices is 0"); | |
2459 | ret = -EINVAL; | |
2460 | } | |
2461 | ||
069ec957 QW |
2462 | if (mirror_num >= 0 && |
2463 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2464 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2465 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2466 | ret = -EINVAL; | |
2467 | } | |
2468 | ||
2469 | /* | |
2470 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2471 | * and one chunk | |
2472 | */ | |
2473 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2474 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2475 | btrfs_super_sys_array_size(sb), | |
2476 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2477 | ret = -EINVAL; | |
2478 | } | |
2479 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2480 | + sizeof(struct btrfs_chunk)) { | |
2481 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2482 | btrfs_super_sys_array_size(sb), | |
2483 | sizeof(struct btrfs_disk_key) | |
2484 | + sizeof(struct btrfs_chunk)); | |
2485 | ret = -EINVAL; | |
2486 | } | |
2487 | ||
2488 | /* | |
2489 | * The generation is a global counter, we'll trust it more than the others | |
2490 | * but it's still possible that it's the one that's wrong. | |
2491 | */ | |
2492 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2493 | btrfs_warn(fs_info, | |
2494 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2495 | btrfs_super_generation(sb), | |
2496 | btrfs_super_chunk_root_generation(sb)); | |
2497 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2498 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2499 | btrfs_warn(fs_info, | |
2500 | "suspicious: generation < cache_generation: %llu < %llu", | |
2501 | btrfs_super_generation(sb), | |
2502 | btrfs_super_cache_generation(sb)); | |
2503 | ||
2504 | return ret; | |
2505 | } | |
2506 | ||
069ec957 QW |
2507 | /* |
2508 | * Validation of super block at mount time. | |
2509 | * Some checks already done early at mount time, like csum type and incompat | |
2510 | * flags will be skipped. | |
2511 | */ | |
2512 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2513 | { | |
a05d3c91 | 2514 | return btrfs_validate_super(fs_info, fs_info->super_copy, 0); |
069ec957 QW |
2515 | } |
2516 | ||
75cb857d QW |
2517 | /* |
2518 | * Validation of super block at write time. | |
2519 | * Some checks like bytenr check will be skipped as their values will be | |
2520 | * overwritten soon. | |
2521 | * Extra checks like csum type and incompat flags will be done here. | |
2522 | */ | |
2523 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2524 | struct btrfs_super_block *sb) | |
2525 | { | |
2526 | int ret; | |
2527 | ||
a05d3c91 | 2528 | ret = btrfs_validate_super(fs_info, sb, -1); |
75cb857d QW |
2529 | if (ret < 0) |
2530 | goto out; | |
e7e16f48 | 2531 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2532 | ret = -EUCLEAN; |
2533 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2534 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2535 | goto out; | |
2536 | } | |
2537 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2538 | ret = -EUCLEAN; | |
2539 | btrfs_err(fs_info, | |
2540 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2541 | btrfs_super_incompat_flags(sb), | |
2542 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2543 | goto out; | |
2544 | } | |
2545 | out: | |
2546 | if (ret < 0) | |
2547 | btrfs_err(fs_info, | |
2548 | "super block corruption detected before writing it to disk"); | |
2549 | return ret; | |
2550 | } | |
2551 | ||
bd676446 JB |
2552 | static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) |
2553 | { | |
789d6a3a QW |
2554 | struct btrfs_tree_parent_check check = { |
2555 | .level = level, | |
2556 | .transid = gen, | |
2557 | .owner_root = root->root_key.objectid | |
2558 | }; | |
bd676446 JB |
2559 | int ret = 0; |
2560 | ||
789d6a3a | 2561 | root->node = read_tree_block(root->fs_info, bytenr, &check); |
bd676446 JB |
2562 | if (IS_ERR(root->node)) { |
2563 | ret = PTR_ERR(root->node); | |
2564 | root->node = NULL; | |
4eb150d6 QW |
2565 | return ret; |
2566 | } | |
2567 | if (!extent_buffer_uptodate(root->node)) { | |
bd676446 JB |
2568 | free_extent_buffer(root->node); |
2569 | root->node = NULL; | |
4eb150d6 | 2570 | return -EIO; |
bd676446 JB |
2571 | } |
2572 | ||
bd676446 JB |
2573 | btrfs_set_root_node(&root->root_item, root->node); |
2574 | root->commit_root = btrfs_root_node(root); | |
2575 | btrfs_set_root_refs(&root->root_item, 1); | |
2576 | return ret; | |
2577 | } | |
2578 | ||
2579 | static int load_important_roots(struct btrfs_fs_info *fs_info) | |
2580 | { | |
2581 | struct btrfs_super_block *sb = fs_info->super_copy; | |
2582 | u64 gen, bytenr; | |
2583 | int level, ret; | |
2584 | ||
2585 | bytenr = btrfs_super_root(sb); | |
2586 | gen = btrfs_super_generation(sb); | |
2587 | level = btrfs_super_root_level(sb); | |
2588 | ret = load_super_root(fs_info->tree_root, bytenr, gen, level); | |
9c54e80d | 2589 | if (ret) { |
bd676446 | 2590 | btrfs_warn(fs_info, "couldn't read tree root"); |
9c54e80d JB |
2591 | return ret; |
2592 | } | |
14033b08 | 2593 | return 0; |
bd676446 JB |
2594 | } |
2595 | ||
6ef108dd | 2596 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2597 | { |
6ef108dd | 2598 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2599 | struct btrfs_super_block *sb = fs_info->super_copy; |
2600 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2601 | bool handle_error = false; | |
2602 | int ret = 0; | |
2603 | int i; | |
2604 | ||
2605 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
b8522a1e NB |
2606 | if (handle_error) { |
2607 | if (!IS_ERR(tree_root->node)) | |
2608 | free_extent_buffer(tree_root->node); | |
2609 | tree_root->node = NULL; | |
2610 | ||
2611 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2612 | break; | |
2613 | ||
2614 | free_root_pointers(fs_info, 0); | |
2615 | ||
2616 | /* | |
2617 | * Don't use the log in recovery mode, it won't be | |
2618 | * valid | |
2619 | */ | |
2620 | btrfs_set_super_log_root(sb, 0); | |
2621 | ||
2622 | /* We can't trust the free space cache either */ | |
2623 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2624 | ||
745806fb | 2625 | btrfs_warn(fs_info, "try to load backup roots slot %d", i); |
b8522a1e | 2626 | ret = read_backup_root(fs_info, i); |
6ef108dd | 2627 | backup_index = ret; |
b8522a1e NB |
2628 | if (ret < 0) |
2629 | return ret; | |
2630 | } | |
b8522a1e | 2631 | |
bd676446 JB |
2632 | ret = load_important_roots(fs_info); |
2633 | if (ret) { | |
217f5004 | 2634 | handle_error = true; |
b8522a1e NB |
2635 | continue; |
2636 | } | |
2637 | ||
336a0d8d NB |
2638 | /* |
2639 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2640 | * hasn't been fully initialised and we are the only user | |
2641 | */ | |
453e4873 | 2642 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 2643 | if (ret < 0) { |
b8522a1e NB |
2644 | handle_error = true; |
2645 | continue; | |
2646 | } | |
2647 | ||
6b8fad57 | 2648 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
2649 | |
2650 | ret = btrfs_read_roots(fs_info); | |
2651 | if (ret < 0) { | |
2652 | handle_error = true; | |
2653 | continue; | |
2654 | } | |
2655 | ||
2656 | /* All successful */ | |
bd676446 | 2657 | fs_info->generation = btrfs_header_generation(tree_root->node); |
0124855f | 2658 | btrfs_set_last_trans_committed(fs_info, fs_info->generation); |
d96b3424 | 2659 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
2660 | |
2661 | /* Always begin writing backup roots after the one being used */ | |
2662 | if (backup_index < 0) { | |
2663 | fs_info->backup_root_index = 0; | |
2664 | } else { | |
2665 | fs_info->backup_root_index = backup_index + 1; | |
2666 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2667 | } | |
b8522a1e NB |
2668 | break; |
2669 | } | |
2670 | ||
2671 | return ret; | |
2672 | } | |
2673 | ||
8260edba | 2674 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 2675 | { |
fc7cbcd4 | 2676 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
01cd3909 | 2677 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 2678 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 2679 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 2680 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 2681 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 2682 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 2683 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 2684 | spin_lock_init(&fs_info->trans_lock); |
fc7cbcd4 | 2685 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 2686 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 2687 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 2688 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 2689 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 2690 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 2691 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 2692 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
c2707a25 | 2693 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 2694 | rwlock_init(&fs_info->tree_mod_log_lock); |
abed4aaa | 2695 | rwlock_init(&fs_info->global_root_lock); |
d7c15171 | 2696 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 2697 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 2698 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 2699 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 2700 | mutex_init(&fs_info->zoned_meta_io_lock); |
5f0addf7 | 2701 | mutex_init(&fs_info->zoned_data_reloc_io_lock); |
de98ced9 | 2702 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 2703 | |
e1489b4f | 2704 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); |
5a9ba670 | 2705 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); |
8b53779e | 2706 | btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); |
5f4403e1 | 2707 | btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); |
77d20c68 JB |
2708 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_prep, |
2709 | BTRFS_LOCKDEP_TRANS_COMMIT_PREP); | |
3e738c53 IA |
2710 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, |
2711 | BTRFS_LOCKDEP_TRANS_UNBLOCKED); | |
2712 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, | |
2713 | BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); | |
2714 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, | |
2715 | BTRFS_LOCKDEP_TRANS_COMPLETED); | |
e1489b4f | 2716 | |
0b86a832 | 2717 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 2718 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 2719 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 2720 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 2721 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 2722 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
2723 | #ifdef CONFIG_BTRFS_DEBUG |
2724 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
2725 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
2726 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 2727 | #endif |
7dc66abb FM |
2728 | fs_info->mapping_tree = RB_ROOT_CACHED; |
2729 | rwlock_init(&fs_info->mapping_tree_lock); | |
66d8f3dd MX |
2730 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
2731 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
2732 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
2733 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
2734 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
2735 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
2736 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
2737 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
2738 | BTRFS_BLOCK_RSV_DELREFS); | |
2739 | ||
771ed689 | 2740 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 2741 | atomic_set(&fs_info->defrag_running, 0); |
034f784d | 2742 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 2743 | atomic64_set(&fs_info->tree_mod_seq, 0); |
abed4aaa | 2744 | fs_info->global_root_tree = RB_ROOT; |
95ac567a | 2745 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 2746 | fs_info->metadata_ratio = 0; |
4cb5300b | 2747 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 2748 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 2749 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 2750 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
fd708b81 | 2751 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 2752 | |
b34b086c CM |
2753 | fs_info->thread_pool_size = min_t(unsigned long, |
2754 | num_online_cpus() + 2, 8); | |
0afbaf8c | 2755 | |
199c2a9c MX |
2756 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
2757 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 2758 | |
638aa7ed | 2759 | btrfs_init_scrub(fs_info); |
779a65a4 | 2760 | btrfs_init_balance(fs_info); |
57056740 | 2761 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 2762 | |
16b0c258 | 2763 | rwlock_init(&fs_info->block_group_cache_lock); |
08dddb29 | 2764 | fs_info->block_group_cache_tree = RB_ROOT_CACHED; |
0f9dd46c | 2765 | |
fe119a6e | 2766 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
35da5a7e | 2767 | IO_TREE_FS_EXCLUDED_EXTENTS); |
39279cc3 | 2768 | |
5a3f23d5 | 2769 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 2770 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 2771 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
2772 | mutex_init(&fs_info->transaction_kthread_mutex); |
2773 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 2774 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 2775 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 2776 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 2777 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 2778 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 2779 | |
ad618368 | 2780 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 2781 | btrfs_init_qgroup(fs_info); |
b0643e59 | 2782 | btrfs_discard_init(fs_info); |
416ac51d | 2783 | |
fa9c0d79 CM |
2784 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
2785 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
2786 | ||
e6dcd2dc | 2787 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 2788 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 2789 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 2790 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 2791 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 2792 | |
da17066c JM |
2793 | /* Usable values until the real ones are cached from the superblock */ |
2794 | fs_info->nodesize = 4096; | |
2795 | fs_info->sectorsize = 4096; | |
ab108d99 | 2796 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
2797 | fs_info->stripesize = 4096; |
2798 | ||
f7b12a62 NA |
2799 | fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; |
2800 | ||
eede2bf3 OS |
2801 | spin_lock_init(&fs_info->swapfile_pins_lock); |
2802 | fs_info->swapfile_pins = RB_ROOT; | |
2803 | ||
18bb8bbf JT |
2804 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
2805 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
2806 | } |
2807 | ||
2808 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
2809 | { | |
2810 | int ret; | |
2811 | ||
2812 | fs_info->sb = sb; | |
2813 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; | |
2814 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 2815 | |
5deb17e1 | 2816 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 2817 | if (ret) |
c75e8394 | 2818 | return ret; |
ae18c37a JB |
2819 | |
2820 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
2821 | if (ret) | |
c75e8394 | 2822 | return ret; |
ae18c37a JB |
2823 | |
2824 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
2825 | (1 + ilog2(nr_cpu_ids)); | |
2826 | ||
2827 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
2828 | if (ret) | |
c75e8394 | 2829 | return ret; |
ae18c37a JB |
2830 | |
2831 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
2832 | GFP_KERNEL); | |
2833 | if (ret) | |
c75e8394 | 2834 | return ret; |
ae18c37a JB |
2835 | |
2836 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
2837 | GFP_KERNEL); | |
c75e8394 JB |
2838 | if (!fs_info->delayed_root) |
2839 | return -ENOMEM; | |
ae18c37a JB |
2840 | btrfs_init_delayed_root(fs_info->delayed_root); |
2841 | ||
a0a1db70 FM |
2842 | if (sb_rdonly(sb)) |
2843 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
2844 | ||
c75e8394 | 2845 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
2846 | } |
2847 | ||
97f4dd09 NB |
2848 | static int btrfs_uuid_rescan_kthread(void *data) |
2849 | { | |
0d031dc4 | 2850 | struct btrfs_fs_info *fs_info = data; |
97f4dd09 NB |
2851 | int ret; |
2852 | ||
2853 | /* | |
2854 | * 1st step is to iterate through the existing UUID tree and | |
2855 | * to delete all entries that contain outdated data. | |
2856 | * 2nd step is to add all missing entries to the UUID tree. | |
2857 | */ | |
2858 | ret = btrfs_uuid_tree_iterate(fs_info); | |
2859 | if (ret < 0) { | |
c94bec2c JB |
2860 | if (ret != -EINTR) |
2861 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
2862 | ret); | |
97f4dd09 NB |
2863 | up(&fs_info->uuid_tree_rescan_sem); |
2864 | return ret; | |
2865 | } | |
2866 | return btrfs_uuid_scan_kthread(data); | |
2867 | } | |
2868 | ||
2869 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
2870 | { | |
2871 | struct task_struct *task; | |
2872 | ||
2873 | down(&fs_info->uuid_tree_rescan_sem); | |
2874 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
2875 | if (IS_ERR(task)) { | |
2876 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
2877 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
2878 | up(&fs_info->uuid_tree_rescan_sem); | |
2879 | return PTR_ERR(task); | |
2880 | } | |
2881 | ||
2882 | return 0; | |
2883 | } | |
2884 | ||
504b1596 FM |
2885 | static int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
2886 | { | |
2887 | u64 root_objectid = 0; | |
2888 | struct btrfs_root *gang[8]; | |
2889 | int i = 0; | |
2890 | int err = 0; | |
2891 | unsigned int ret = 0; | |
2892 | ||
2893 | while (1) { | |
2894 | spin_lock(&fs_info->fs_roots_radix_lock); | |
2895 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2896 | (void **)gang, root_objectid, | |
2897 | ARRAY_SIZE(gang)); | |
2898 | if (!ret) { | |
2899 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
2900 | break; | |
2901 | } | |
2902 | root_objectid = gang[ret - 1]->root_key.objectid + 1; | |
2903 | ||
2904 | for (i = 0; i < ret; i++) { | |
2905 | /* Avoid to grab roots in dead_roots. */ | |
2906 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
2907 | gang[i] = NULL; | |
2908 | continue; | |
2909 | } | |
2910 | /* Grab all the search result for later use. */ | |
2911 | gang[i] = btrfs_grab_root(gang[i]); | |
2912 | } | |
2913 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
2914 | ||
2915 | for (i = 0; i < ret; i++) { | |
2916 | if (!gang[i]) | |
2917 | continue; | |
2918 | root_objectid = gang[i]->root_key.objectid; | |
2919 | err = btrfs_orphan_cleanup(gang[i]); | |
2920 | if (err) | |
2921 | goto out; | |
2922 | btrfs_put_root(gang[i]); | |
2923 | } | |
2924 | root_objectid++; | |
2925 | } | |
2926 | out: | |
2927 | /* Release the uncleaned roots due to error. */ | |
2928 | for (; i < ret; i++) { | |
2929 | if (gang[i]) | |
2930 | btrfs_put_root(gang[i]); | |
2931 | } | |
2932 | return err; | |
2933 | } | |
2934 | ||
8cd29088 BB |
2935 | /* |
2936 | * Some options only have meaning at mount time and shouldn't persist across | |
2937 | * remounts, or be displayed. Clear these at the end of mount and remount | |
2938 | * code paths. | |
2939 | */ | |
2940 | void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) | |
2941 | { | |
2942 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
8b228324 | 2943 | btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); |
8cd29088 BB |
2944 | } |
2945 | ||
44c0ca21 BB |
2946 | /* |
2947 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
2948 | * and btrfs_remount when remounting from read-only to read-write. | |
2949 | */ | |
2950 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
2951 | { | |
2952 | int ret; | |
94846229 | 2953 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
1d6a4fc8 | 2954 | bool rebuild_free_space_tree = false; |
8b228324 BB |
2955 | |
2956 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
2957 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
1d6a4fc8 | 2958 | rebuild_free_space_tree = true; |
8b228324 BB |
2959 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && |
2960 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
2961 | btrfs_warn(fs_info, "free space tree is invalid"); | |
1d6a4fc8 | 2962 | rebuild_free_space_tree = true; |
8b228324 BB |
2963 | } |
2964 | ||
1d6a4fc8 QW |
2965 | if (rebuild_free_space_tree) { |
2966 | btrfs_info(fs_info, "rebuilding free space tree"); | |
2967 | ret = btrfs_rebuild_free_space_tree(fs_info); | |
8b228324 BB |
2968 | if (ret) { |
2969 | btrfs_warn(fs_info, | |
1d6a4fc8 QW |
2970 | "failed to rebuild free space tree: %d", ret); |
2971 | goto out; | |
2972 | } | |
2973 | } | |
2974 | ||
2975 | if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
2976 | !btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { | |
2977 | btrfs_info(fs_info, "disabling free space tree"); | |
2978 | ret = btrfs_delete_free_space_tree(fs_info); | |
2979 | if (ret) { | |
2980 | btrfs_warn(fs_info, | |
2981 | "failed to disable free space tree: %d", ret); | |
8b228324 BB |
2982 | goto out; |
2983 | } | |
2984 | } | |
44c0ca21 | 2985 | |
8d488a8c FM |
2986 | /* |
2987 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
2988 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
fc7cbcd4 | 2989 | * them into the fs_info->fs_roots_radix tree. This must be done before |
8d488a8c FM |
2990 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it |
2991 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
2992 | * item before the root's tree is deleted - this means that if we unmount | |
2993 | * or crash before the deletion completes, on the next mount we will not | |
2994 | * delete what remains of the tree because the orphan item does not | |
2995 | * exists anymore, which is what tells us we have a pending deletion. | |
2996 | */ | |
2997 | ret = btrfs_find_orphan_roots(fs_info); | |
2998 | if (ret) | |
2999 | goto out; | |
3000 | ||
44c0ca21 BB |
3001 | ret = btrfs_cleanup_fs_roots(fs_info); |
3002 | if (ret) | |
3003 | goto out; | |
3004 | ||
8f1c21d7 BB |
3005 | down_read(&fs_info->cleanup_work_sem); |
3006 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3007 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3008 | up_read(&fs_info->cleanup_work_sem); | |
3009 | goto out; | |
3010 | } | |
3011 | up_read(&fs_info->cleanup_work_sem); | |
3012 | ||
44c0ca21 | 3013 | mutex_lock(&fs_info->cleaner_mutex); |
7eefae6b | 3014 | ret = btrfs_recover_relocation(fs_info); |
44c0ca21 BB |
3015 | mutex_unlock(&fs_info->cleaner_mutex); |
3016 | if (ret < 0) { | |
3017 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3018 | goto out; | |
3019 | } | |
3020 | ||
5011139a BB |
3021 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3022 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3023 | btrfs_info(fs_info, "creating free space tree"); | |
3024 | ret = btrfs_create_free_space_tree(fs_info); | |
3025 | if (ret) { | |
3026 | btrfs_warn(fs_info, | |
3027 | "failed to create free space tree: %d", ret); | |
3028 | goto out; | |
3029 | } | |
3030 | } | |
3031 | ||
94846229 BB |
3032 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3033 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3034 | if (ret) | |
3035 | goto out; | |
3036 | } | |
3037 | ||
44c0ca21 BB |
3038 | ret = btrfs_resume_balance_async(fs_info); |
3039 | if (ret) | |
3040 | goto out; | |
3041 | ||
3042 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3043 | if (ret) { | |
3044 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3045 | goto out; | |
3046 | } | |
3047 | ||
3048 | btrfs_qgroup_rescan_resume(fs_info); | |
3049 | ||
3050 | if (!fs_info->uuid_root) { | |
3051 | btrfs_info(fs_info, "creating UUID tree"); | |
3052 | ret = btrfs_create_uuid_tree(fs_info); | |
3053 | if (ret) { | |
3054 | btrfs_warn(fs_info, | |
3055 | "failed to create the UUID tree %d", ret); | |
3056 | goto out; | |
3057 | } | |
3058 | } | |
3059 | ||
3060 | out: | |
3061 | return ret; | |
3062 | } | |
3063 | ||
d7f67ac9 QW |
3064 | /* |
3065 | * Do various sanity and dependency checks of different features. | |
3066 | * | |
2ba48b20 QW |
3067 | * @is_rw_mount: If the mount is read-write. |
3068 | * | |
d7f67ac9 QW |
3069 | * This is the place for less strict checks (like for subpage or artificial |
3070 | * feature dependencies). | |
3071 | * | |
3072 | * For strict checks or possible corruption detection, see | |
3073 | * btrfs_validate_super(). | |
3074 | * | |
3075 | * This should be called after btrfs_parse_options(), as some mount options | |
3076 | * (space cache related) can modify on-disk format like free space tree and | |
3077 | * screw up certain feature dependencies. | |
3078 | */ | |
2ba48b20 | 3079 | int btrfs_check_features(struct btrfs_fs_info *fs_info, bool is_rw_mount) |
d7f67ac9 QW |
3080 | { |
3081 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
3082 | u64 incompat = btrfs_super_incompat_flags(disk_super); | |
3083 | const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); | |
3084 | const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); | |
3085 | ||
3086 | if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
3087 | btrfs_err(fs_info, | |
3088 | "cannot mount because of unknown incompat features (0x%llx)", | |
3089 | incompat); | |
3090 | return -EINVAL; | |
3091 | } | |
3092 | ||
3093 | /* Runtime limitation for mixed block groups. */ | |
3094 | if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
3095 | (fs_info->sectorsize != fs_info->nodesize)) { | |
3096 | btrfs_err(fs_info, | |
3097 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3098 | fs_info->nodesize, fs_info->sectorsize); | |
3099 | return -EINVAL; | |
3100 | } | |
3101 | ||
3102 | /* Mixed backref is an always-enabled feature. */ | |
3103 | incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; | |
3104 | ||
3105 | /* Set compression related flags just in case. */ | |
3106 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) | |
3107 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
3108 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) | |
3109 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
3110 | ||
3111 | /* | |
3112 | * An ancient flag, which should really be marked deprecated. | |
3113 | * Such runtime limitation doesn't really need a incompat flag. | |
3114 | */ | |
3115 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) | |
3116 | incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3117 | ||
2ba48b20 | 3118 | if (compat_ro_unsupp && is_rw_mount) { |
d7f67ac9 QW |
3119 | btrfs_err(fs_info, |
3120 | "cannot mount read-write because of unknown compat_ro features (0x%llx)", | |
3121 | compat_ro); | |
3122 | return -EINVAL; | |
3123 | } | |
3124 | ||
3125 | /* | |
3126 | * We have unsupported RO compat features, although RO mounted, we | |
3127 | * should not cause any metadata writes, including log replay. | |
3128 | * Or we could screw up whatever the new feature requires. | |
3129 | */ | |
3130 | if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && | |
3131 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { | |
3132 | btrfs_err(fs_info, | |
3133 | "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", | |
3134 | compat_ro); | |
3135 | return -EINVAL; | |
3136 | } | |
3137 | ||
3138 | /* | |
3139 | * Artificial limitations for block group tree, to force | |
3140 | * block-group-tree to rely on no-holes and free-space-tree. | |
3141 | */ | |
3142 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
3143 | (!btrfs_fs_incompat(fs_info, NO_HOLES) || | |
3144 | !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { | |
3145 | btrfs_err(fs_info, | |
3146 | "block-group-tree feature requires no-holes and free-space-tree features"); | |
3147 | return -EINVAL; | |
3148 | } | |
3149 | ||
3150 | /* | |
3151 | * Subpage runtime limitation on v1 cache. | |
3152 | * | |
3153 | * V1 space cache still has some hard codeed PAGE_SIZE usage, while | |
3154 | * we're already defaulting to v2 cache, no need to bother v1 as it's | |
3155 | * going to be deprecated anyway. | |
3156 | */ | |
3157 | if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { | |
3158 | btrfs_warn(fs_info, | |
3159 | "v1 space cache is not supported for page size %lu with sectorsize %u", | |
3160 | PAGE_SIZE, fs_info->sectorsize); | |
3161 | return -EINVAL; | |
3162 | } | |
3163 | ||
3164 | /* This can be called by remount, we need to protect the super block. */ | |
3165 | spin_lock(&fs_info->super_lock); | |
3166 | btrfs_set_super_incompat_flags(disk_super, incompat); | |
3167 | spin_unlock(&fs_info->super_lock); | |
3168 | ||
3169 | return 0; | |
3170 | } | |
3171 | ||
ae18c37a JB |
3172 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3173 | char *options) | |
3174 | { | |
3175 | u32 sectorsize; | |
3176 | u32 nodesize; | |
3177 | u32 stripesize; | |
3178 | u64 generation; | |
ae18c37a | 3179 | u16 csum_type; |
ae18c37a JB |
3180 | struct btrfs_super_block *disk_super; |
3181 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3182 | struct btrfs_root *tree_root; | |
3183 | struct btrfs_root *chunk_root; | |
3184 | int ret; | |
ae18c37a JB |
3185 | int level; |
3186 | ||
8260edba | 3187 | ret = init_mount_fs_info(fs_info, sb); |
4871c33b | 3188 | if (ret) |
ae18c37a | 3189 | goto fail; |
53b381b3 | 3190 | |
ae18c37a JB |
3191 | /* These need to be init'ed before we start creating inodes and such. */ |
3192 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3193 | GFP_KERNEL); | |
3194 | fs_info->tree_root = tree_root; | |
3195 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3196 | GFP_KERNEL); | |
3197 | fs_info->chunk_root = chunk_root; | |
3198 | if (!tree_root || !chunk_root) { | |
4871c33b | 3199 | ret = -ENOMEM; |
c75e8394 | 3200 | goto fail; |
ae18c37a JB |
3201 | } |
3202 | ||
dcb2137c | 3203 | ret = btrfs_init_btree_inode(sb); |
4871c33b | 3204 | if (ret) |
c75e8394 | 3205 | goto fail; |
ae18c37a | 3206 | |
d24fa5c1 | 3207 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3208 | |
3209 | /* | |
3210 | * Read super block and check the signature bytes only | |
3211 | */ | |
d24fa5c1 | 3212 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d | 3213 | if (IS_ERR(disk_super)) { |
4871c33b | 3214 | ret = PTR_ERR(disk_super); |
16cdcec7 | 3215 | goto fail_alloc; |
20b45077 | 3216 | } |
39279cc3 | 3217 | |
2db31320 | 3218 | btrfs_info(fs_info, "first mount of filesystem %pU", disk_super->fsid); |
8dc3f22c | 3219 | /* |
260db43c | 3220 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3221 | * corrupted, we'll find out |
3222 | */ | |
8f32380d | 3223 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3224 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3225 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3226 | csum_type); |
4871c33b | 3227 | ret = -EINVAL; |
8f32380d | 3228 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3229 | goto fail_alloc; |
3230 | } | |
3231 | ||
83c68bbc SY |
3232 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3233 | ||
6d97c6e3 JT |
3234 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3235 | if (ret) { | |
8f32380d | 3236 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3237 | goto fail_alloc; |
3238 | } | |
3239 | ||
1104a885 DS |
3240 | /* |
3241 | * We want to check superblock checksum, the type is stored inside. | |
3242 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3243 | */ | |
3d17adea | 3244 | if (btrfs_check_super_csum(fs_info, disk_super)) { |
05135f59 | 3245 | btrfs_err(fs_info, "superblock checksum mismatch"); |
4871c33b | 3246 | ret = -EINVAL; |
8f32380d | 3247 | btrfs_release_disk_super(disk_super); |
141386e1 | 3248 | goto fail_alloc; |
1104a885 DS |
3249 | } |
3250 | ||
3251 | /* | |
3252 | * super_copy is zeroed at allocation time and we never touch the | |
3253 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3254 | * the whole block of INFO_SIZE | |
3255 | */ | |
8f32380d JT |
3256 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3257 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3258 | |
fbc6feae NB |
3259 | disk_super = fs_info->super_copy; |
3260 | ||
fbc6feae NB |
3261 | memcpy(fs_info->super_for_commit, fs_info->super_copy, |
3262 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3263 | |
069ec957 | 3264 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3265 | if (ret) { |
05135f59 | 3266 | btrfs_err(fs_info, "superblock contains fatal errors"); |
4871c33b | 3267 | ret = -EINVAL; |
141386e1 | 3268 | goto fail_alloc; |
1104a885 DS |
3269 | } |
3270 | ||
4871c33b QW |
3271 | if (!btrfs_super_root(disk_super)) { |
3272 | btrfs_err(fs_info, "invalid superblock tree root bytenr"); | |
3273 | ret = -EINVAL; | |
141386e1 | 3274 | goto fail_alloc; |
4871c33b | 3275 | } |
0f7d52f4 | 3276 | |
acce952b | 3277 | /* check FS state, whether FS is broken. */ |
87533c47 | 3278 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
ae3364e5 | 3279 | WRITE_ONCE(fs_info->fs_error, -EUCLEAN); |
acce952b | 3280 | |
75e7cb7f LB |
3281 | /* |
3282 | * In the long term, we'll store the compression type in the super | |
3283 | * block, and it'll be used for per file compression control. | |
3284 | */ | |
3285 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
3286 | ||
6f93e834 AJ |
3287 | |
3288 | /* Set up fs_info before parsing mount options */ | |
3289 | nodesize = btrfs_super_nodesize(disk_super); | |
3290 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3291 | stripesize = sectorsize; | |
3292 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3293 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3294 | ||
3295 | fs_info->nodesize = nodesize; | |
3296 | fs_info->sectorsize = sectorsize; | |
3297 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3298 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3299 | fs_info->stripesize = stripesize; | |
3300 | ||
2ff7e61e | 3301 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
4871c33b | 3302 | if (ret) |
141386e1 | 3303 | goto fail_alloc; |
dfe25020 | 3304 | |
2ba48b20 | 3305 | ret = btrfs_check_features(fs_info, !sb_rdonly(sb)); |
4871c33b | 3306 | if (ret < 0) |
dc4d3168 | 3307 | goto fail_alloc; |
dc4d3168 | 3308 | |
8481dd80 QW |
3309 | if (sectorsize < PAGE_SIZE) { |
3310 | struct btrfs_subpage_info *subpage_info; | |
3311 | ||
9f73f1ae QW |
3312 | /* |
3313 | * V1 space cache has some hardcoded PAGE_SIZE usage, and is | |
3314 | * going to be deprecated. | |
3315 | * | |
3316 | * Force to use v2 cache for subpage case. | |
3317 | */ | |
3318 | btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); | |
3319 | btrfs_set_and_info(fs_info, FREE_SPACE_TREE, | |
3320 | "forcing free space tree for sector size %u with page size %lu", | |
3321 | sectorsize, PAGE_SIZE); | |
3322 | ||
95ea0486 QW |
3323 | btrfs_warn(fs_info, |
3324 | "read-write for sector size %u with page size %lu is experimental", | |
3325 | sectorsize, PAGE_SIZE); | |
8481dd80 | 3326 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
4871c33b QW |
3327 | if (!subpage_info) { |
3328 | ret = -ENOMEM; | |
8481dd80 | 3329 | goto fail_alloc; |
4871c33b | 3330 | } |
8481dd80 QW |
3331 | btrfs_init_subpage_info(subpage_info, sectorsize); |
3332 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3333 | } |
0bb3eb3e | 3334 | |
d21deec5 | 3335 | ret = btrfs_init_workqueues(fs_info); |
4871c33b | 3336 | if (ret) |
0dc3b84a | 3337 | goto fail_sb_buffer; |
4543df7e | 3338 | |
9e11ceee JK |
3339 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3340 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3341 | |
a061fc8d CM |
3342 | sb->s_blocksize = sectorsize; |
3343 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3344 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3345 | |
925baedd | 3346 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3347 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3348 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3349 | if (ret) { |
05135f59 | 3350 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3351 | goto fail_sb_buffer; |
84eed90f | 3352 | } |
0b86a832 | 3353 | |
84234f3a | 3354 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3355 | level = btrfs_super_chunk_root_level(disk_super); |
bd676446 JB |
3356 | ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), |
3357 | generation, level); | |
3358 | if (ret) { | |
05135f59 | 3359 | btrfs_err(fs_info, "failed to read chunk root"); |
af31f5e5 | 3360 | goto fail_tree_roots; |
83121942 | 3361 | } |
0b86a832 | 3362 | |
e17cade2 | 3363 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3364 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3365 | BTRFS_UUID_SIZE); | |
e17cade2 | 3366 | |
5b4aacef | 3367 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3368 | if (ret) { |
05135f59 | 3369 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3370 | goto fail_tree_roots; |
2b82032c | 3371 | } |
0b86a832 | 3372 | |
8dabb742 | 3373 | /* |
bacce86a AJ |
3374 | * At this point we know all the devices that make this filesystem, |
3375 | * including the seed devices but we don't know yet if the replace | |
3376 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3377 | * filesystem but skip the replace target device which is checked |
bacce86a | 3378 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3379 | */ |
bacce86a | 3380 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3381 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3382 | btrfs_err(fs_info, "failed to read devices"); |
4871c33b | 3383 | ret = -EIO; |
a6b0d5c8 CM |
3384 | goto fail_tree_roots; |
3385 | } | |
3386 | ||
b8522a1e | 3387 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3388 | if (ret) |
b8522a1e | 3389 | goto fail_tree_roots; |
8929ecfa | 3390 | |
73651042 NA |
3391 | /* |
3392 | * Get zone type information of zoned block devices. This will also | |
3393 | * handle emulation of a zoned filesystem if a regular device has the | |
3394 | * zoned incompat feature flag set. | |
3395 | */ | |
3396 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3397 | if (ret) { | |
3398 | btrfs_err(fs_info, | |
4871c33b | 3399 | "zoned: failed to read device zone info: %d", ret); |
73651042 NA |
3400 | goto fail_block_groups; |
3401 | } | |
3402 | ||
75ec1db8 JB |
3403 | /* |
3404 | * If we have a uuid root and we're not being told to rescan we need to | |
3405 | * check the generation here so we can set the | |
3406 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3407 | * transaction during a balance or the log replay without updating the | |
3408 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3409 | * even though it was perfectly fine. | |
3410 | */ | |
3411 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3412 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3413 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3414 | ||
cf90d884 QW |
3415 | ret = btrfs_verify_dev_extents(fs_info); |
3416 | if (ret) { | |
3417 | btrfs_err(fs_info, | |
3418 | "failed to verify dev extents against chunks: %d", | |
3419 | ret); | |
3420 | goto fail_block_groups; | |
3421 | } | |
68310a5e ID |
3422 | ret = btrfs_recover_balance(fs_info); |
3423 | if (ret) { | |
05135f59 | 3424 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3425 | goto fail_block_groups; |
3426 | } | |
3427 | ||
733f4fbb SB |
3428 | ret = btrfs_init_dev_stats(fs_info); |
3429 | if (ret) { | |
05135f59 | 3430 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3431 | goto fail_block_groups; |
3432 | } | |
3433 | ||
8dabb742 SB |
3434 | ret = btrfs_init_dev_replace(fs_info); |
3435 | if (ret) { | |
05135f59 | 3436 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3437 | goto fail_block_groups; |
3438 | } | |
3439 | ||
b70f5097 NA |
3440 | ret = btrfs_check_zoned_mode(fs_info); |
3441 | if (ret) { | |
3442 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3443 | ret); | |
3444 | goto fail_block_groups; | |
3445 | } | |
3446 | ||
c6761a9e | 3447 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3448 | if (ret) { |
05135f59 DS |
3449 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3450 | ret); | |
b7c35e81 AJ |
3451 | goto fail_block_groups; |
3452 | } | |
3453 | ||
96f3136e | 3454 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3455 | if (ret) { |
05135f59 | 3456 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3457 | goto fail_fsdev_sysfs; |
c59021f8 | 3458 | } |
3459 | ||
c59021f8 | 3460 | ret = btrfs_init_space_info(fs_info); |
3461 | if (ret) { | |
05135f59 | 3462 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3463 | goto fail_sysfs; |
c59021f8 | 3464 | } |
3465 | ||
5b4aacef | 3466 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3467 | if (ret) { |
05135f59 | 3468 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3469 | goto fail_sysfs; |
1b1d1f66 | 3470 | } |
4330e183 | 3471 | |
16beac87 NA |
3472 | btrfs_free_zone_cache(fs_info); |
3473 | ||
a7e1ac7b NA |
3474 | btrfs_check_active_zone_reservation(fs_info); |
3475 | ||
5c78a5e7 AJ |
3476 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3477 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3478 | btrfs_warn(fs_info, |
52042d8e | 3479 | "writable mount is not allowed due to too many missing devices"); |
4871c33b | 3480 | ret = -EINVAL; |
2365dd3c | 3481 | goto fail_sysfs; |
292fd7fc | 3482 | } |
9078a3e1 | 3483 | |
33c44184 | 3484 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, |
a74a4b97 | 3485 | "btrfs-cleaner"); |
4871c33b QW |
3486 | if (IS_ERR(fs_info->cleaner_kthread)) { |
3487 | ret = PTR_ERR(fs_info->cleaner_kthread); | |
2365dd3c | 3488 | goto fail_sysfs; |
4871c33b | 3489 | } |
a74a4b97 CM |
3490 | |
3491 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3492 | tree_root, | |
3493 | "btrfs-transaction"); | |
4871c33b QW |
3494 | if (IS_ERR(fs_info->transaction_kthread)) { |
3495 | ret = PTR_ERR(fs_info->transaction_kthread); | |
3f157a2f | 3496 | goto fail_cleaner; |
4871c33b | 3497 | } |
a74a4b97 | 3498 | |
583b7231 | 3499 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3500 | !fs_info->fs_devices->rotating) { |
583b7231 | 3501 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3502 | } |
3503 | ||
63a7cb13 DS |
3504 | /* |
3505 | * For devices supporting discard turn on discard=async automatically, | |
3506 | * unless it's already set or disabled. This could be turned off by | |
3507 | * nodiscard for the same mount. | |
95ca6599 NA |
3508 | * |
3509 | * The zoned mode piggy backs on the discard functionality for | |
3510 | * resetting a zone. There is no reason to delay the zone reset as it is | |
3511 | * fast enough. So, do not enable async discard for zoned mode. | |
63a7cb13 DS |
3512 | */ |
3513 | if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || | |
3514 | btrfs_test_opt(fs_info, DISCARD_ASYNC) || | |
3515 | btrfs_test_opt(fs_info, NODISCARD)) && | |
95ca6599 NA |
3516 | fs_info->fs_devices->discardable && |
3517 | !btrfs_is_zoned(fs_info)) { | |
63a7cb13 DS |
3518 | btrfs_set_and_info(fs_info, DISCARD_ASYNC, |
3519 | "auto enabling async discard"); | |
63a7cb13 DS |
3520 | } |
3521 | ||
bcef60f2 AJ |
3522 | ret = btrfs_read_qgroup_config(fs_info); |
3523 | if (ret) | |
3524 | goto fail_trans_kthread; | |
21adbd5c | 3525 | |
fd708b81 JB |
3526 | if (btrfs_build_ref_tree(fs_info)) |
3527 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3528 | ||
96da0919 QW |
3529 | /* do not make disk changes in broken FS or nologreplay is given */ |
3530 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3531 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3532 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3533 | ret = btrfs_replay_log(fs_info, fs_devices); |
4871c33b | 3534 | if (ret) |
28c16cbb | 3535 | goto fail_qgroup; |
e02119d5 | 3536 | } |
1a40e23b | 3537 | |
56e9357a | 3538 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 | 3539 | if (IS_ERR(fs_info->fs_root)) { |
4871c33b QW |
3540 | ret = PTR_ERR(fs_info->fs_root); |
3541 | btrfs_warn(fs_info, "failed to read fs tree: %d", ret); | |
315bf8ef | 3542 | fs_info->fs_root = NULL; |
bcef60f2 | 3543 | goto fail_qgroup; |
3140c9a3 | 3544 | } |
c289811c | 3545 | |
bc98a42c | 3546 | if (sb_rdonly(sb)) |
8cd29088 | 3547 | goto clear_oneshot; |
59641015 | 3548 | |
44c0ca21 | 3549 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3550 | if (ret) { |
6bccf3ab | 3551 | close_ctree(fs_info); |
2b6ba629 | 3552 | return ret; |
e3acc2a6 | 3553 | } |
b0643e59 | 3554 | btrfs_discard_resume(fs_info); |
b382a324 | 3555 | |
44c0ca21 BB |
3556 | if (fs_info->uuid_root && |
3557 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3558 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3559 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3560 | ret = btrfs_check_uuid_tree(fs_info); |
3561 | if (ret) { | |
05135f59 DS |
3562 | btrfs_warn(fs_info, |
3563 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3564 | close_ctree(fs_info); |
70f80175 SB |
3565 | return ret; |
3566 | } | |
f7a81ea4 | 3567 | } |
94846229 | 3568 | |
afcdd129 | 3569 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3570 | |
b4be6aef JB |
3571 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3572 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3573 | wake_up_process(fs_info->cleaner_kthread); | |
3574 | ||
8cd29088 BB |
3575 | clear_oneshot: |
3576 | btrfs_clear_oneshot_options(fs_info); | |
ad2b2c80 | 3577 | return 0; |
39279cc3 | 3578 | |
bcef60f2 AJ |
3579 | fail_qgroup: |
3580 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3581 | fail_trans_kthread: |
3582 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3583 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3584 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3585 | fail_cleaner: |
a74a4b97 | 3586 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3587 | |
3588 | /* | |
3589 | * make sure we're done with the btree inode before we stop our | |
3590 | * kthreads | |
3591 | */ | |
3592 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3593 | |
2365dd3c | 3594 | fail_sysfs: |
6618a59b | 3595 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3596 | |
b7c35e81 AJ |
3597 | fail_fsdev_sysfs: |
3598 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3599 | ||
1b1d1f66 | 3600 | fail_block_groups: |
54067ae9 | 3601 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3602 | |
3603 | fail_tree_roots: | |
9e3aa805 JB |
3604 | if (fs_info->data_reloc_root) |
3605 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3606 | free_root_pointers(fs_info, true); |
2b8195bb | 3607 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3608 | |
39279cc3 | 3609 | fail_sb_buffer: |
7abadb64 | 3610 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3611 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3612 | fail_alloc: |
7dc66abb | 3613 | btrfs_mapping_tree_free(fs_info); |
586e46e2 | 3614 | |
4543df7e | 3615 | iput(fs_info->btree_inode); |
7e662854 | 3616 | fail: |
586e46e2 | 3617 | btrfs_close_devices(fs_info->fs_devices); |
4871c33b QW |
3618 | ASSERT(ret < 0); |
3619 | return ret; | |
eb60ceac | 3620 | } |
663faf9f | 3621 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3622 | |
314b6dd0 | 3623 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3624 | { |
314b6dd0 JT |
3625 | struct btrfs_device *device = bio->bi_private; |
3626 | struct bio_vec *bvec; | |
3627 | struct bvec_iter_all iter_all; | |
3628 | struct page *page; | |
3629 | ||
3630 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3631 | page = bvec->bv_page; | |
3632 | ||
3633 | if (bio->bi_status) { | |
3634 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3635 | "lost page write due to IO error on %s (%d)", | |
cb3e217b | 3636 | btrfs_dev_name(device), |
314b6dd0 JT |
3637 | blk_status_to_errno(bio->bi_status)); |
3638 | ClearPageUptodate(page); | |
3639 | SetPageError(page); | |
3640 | btrfs_dev_stat_inc_and_print(device, | |
3641 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3642 | } else { | |
3643 | SetPageUptodate(page); | |
3644 | } | |
3645 | ||
3646 | put_page(page); | |
3647 | unlock_page(page); | |
f2984462 | 3648 | } |
314b6dd0 JT |
3649 | |
3650 | bio_put(bio); | |
f2984462 CM |
3651 | } |
3652 | ||
8f32380d | 3653 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
a05d3c91 | 3654 | int copy_num, bool drop_cache) |
29c36d72 | 3655 | { |
29c36d72 | 3656 | struct btrfs_super_block *super; |
8f32380d | 3657 | struct page *page; |
12659251 | 3658 | u64 bytenr, bytenr_orig; |
8f32380d | 3659 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
3660 | int ret; |
3661 | ||
3662 | bytenr_orig = btrfs_sb_offset(copy_num); | |
3663 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
3664 | if (ret == -ENOENT) | |
3665 | return ERR_PTR(-EINVAL); | |
3666 | else if (ret) | |
3667 | return ERR_PTR(ret); | |
29c36d72 | 3668 | |
cda00eba | 3669 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) |
8f32380d | 3670 | return ERR_PTR(-EINVAL); |
29c36d72 | 3671 | |
a05d3c91 QW |
3672 | if (drop_cache) { |
3673 | /* This should only be called with the primary sb. */ | |
3674 | ASSERT(copy_num == 0); | |
3675 | ||
3676 | /* | |
3677 | * Drop the page of the primary superblock, so later read will | |
3678 | * always read from the device. | |
3679 | */ | |
3680 | invalidate_inode_pages2_range(mapping, | |
3681 | bytenr >> PAGE_SHIFT, | |
3682 | (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); | |
3683 | } | |
3684 | ||
8f32380d JT |
3685 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3686 | if (IS_ERR(page)) | |
3687 | return ERR_CAST(page); | |
29c36d72 | 3688 | |
8f32380d | 3689 | super = page_address(page); |
96c2e067 AJ |
3690 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
3691 | btrfs_release_disk_super(super); | |
3692 | return ERR_PTR(-ENODATA); | |
3693 | } | |
3694 | ||
12659251 | 3695 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
3696 | btrfs_release_disk_super(super); |
3697 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
3698 | } |
3699 | ||
8f32380d | 3700 | return super; |
29c36d72 AJ |
3701 | } |
3702 | ||
3703 | ||
8f32380d | 3704 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 3705 | { |
8f32380d | 3706 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
3707 | int i; |
3708 | u64 transid = 0; | |
a512bbf8 YZ |
3709 | |
3710 | /* we would like to check all the supers, but that would make | |
3711 | * a btrfs mount succeed after a mkfs from a different FS. | |
3712 | * So, we need to add a special mount option to scan for | |
3713 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
3714 | */ | |
3715 | for (i = 0; i < 1; i++) { | |
a05d3c91 | 3716 | super = btrfs_read_dev_one_super(bdev, i, false); |
8f32380d | 3717 | if (IS_ERR(super)) |
a512bbf8 YZ |
3718 | continue; |
3719 | ||
a512bbf8 | 3720 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
3721 | if (latest) |
3722 | btrfs_release_disk_super(super); | |
3723 | ||
3724 | latest = super; | |
a512bbf8 | 3725 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
3726 | } |
3727 | } | |
92fc03fb | 3728 | |
8f32380d | 3729 | return super; |
a512bbf8 YZ |
3730 | } |
3731 | ||
4eedeb75 | 3732 | /* |
abbb3b8e | 3733 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 3734 | * pages we use for writing are locked. |
4eedeb75 | 3735 | * |
abbb3b8e DS |
3736 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
3737 | * the expected device size at commit time. Note that max_mirrors must be | |
3738 | * same for write and wait phases. | |
4eedeb75 | 3739 | * |
314b6dd0 | 3740 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 3741 | */ |
a512bbf8 | 3742 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 3743 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 3744 | { |
d5178578 | 3745 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 3746 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 3747 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 3748 | int i; |
a512bbf8 | 3749 | int errors = 0; |
12659251 NA |
3750 | int ret; |
3751 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
3752 | |
3753 | if (max_mirrors == 0) | |
3754 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3755 | ||
d5178578 JT |
3756 | shash->tfm = fs_info->csum_shash; |
3757 | ||
a512bbf8 | 3758 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
3759 | struct page *page; |
3760 | struct bio *bio; | |
3761 | struct btrfs_super_block *disk_super; | |
3762 | ||
12659251 NA |
3763 | bytenr_orig = btrfs_sb_offset(i); |
3764 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
3765 | if (ret == -ENOENT) { | |
3766 | continue; | |
3767 | } else if (ret < 0) { | |
3768 | btrfs_err(device->fs_info, | |
3769 | "couldn't get super block location for mirror %d", | |
3770 | i); | |
3771 | errors++; | |
3772 | continue; | |
3773 | } | |
935e5cc9 MX |
3774 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3775 | device->commit_total_bytes) | |
a512bbf8 YZ |
3776 | break; |
3777 | ||
12659251 | 3778 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 3779 | |
fd08001f EB |
3780 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
3781 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
3782 | sb->csum); | |
4eedeb75 | 3783 | |
314b6dd0 JT |
3784 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
3785 | GFP_NOFS); | |
3786 | if (!page) { | |
abbb3b8e | 3787 | btrfs_err(device->fs_info, |
314b6dd0 | 3788 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
3789 | bytenr); |
3790 | errors++; | |
4eedeb75 | 3791 | continue; |
abbb3b8e | 3792 | } |
634554dc | 3793 | |
314b6dd0 JT |
3794 | /* Bump the refcount for wait_dev_supers() */ |
3795 | get_page(page); | |
a512bbf8 | 3796 | |
314b6dd0 JT |
3797 | disk_super = page_address(page); |
3798 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 3799 | |
314b6dd0 JT |
3800 | /* |
3801 | * Directly use bios here instead of relying on the page cache | |
3802 | * to do I/O, so we don't lose the ability to do integrity | |
3803 | * checking. | |
3804 | */ | |
07888c66 CH |
3805 | bio = bio_alloc(device->bdev, 1, |
3806 | REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, | |
3807 | GFP_NOFS); | |
314b6dd0 JT |
3808 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; |
3809 | bio->bi_private = device; | |
3810 | bio->bi_end_io = btrfs_end_super_write; | |
3811 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
3812 | offset_in_page(bytenr)); | |
a512bbf8 | 3813 | |
387125fc | 3814 | /* |
314b6dd0 JT |
3815 | * We FUA only the first super block. The others we allow to |
3816 | * go down lazy and there's a short window where the on-disk | |
3817 | * copies might still contain the older version. | |
387125fc | 3818 | */ |
1b9e619c | 3819 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 | 3820 | bio->bi_opf |= REQ_FUA; |
58ff51f1 | 3821 | submit_bio(bio); |
8376d9e1 NA |
3822 | |
3823 | if (btrfs_advance_sb_log(device, i)) | |
3824 | errors++; | |
a512bbf8 YZ |
3825 | } |
3826 | return errors < i ? 0 : -1; | |
3827 | } | |
3828 | ||
abbb3b8e DS |
3829 | /* |
3830 | * Wait for write completion of superblocks done by write_dev_supers, | |
3831 | * @max_mirrors same for write and wait phases. | |
3832 | * | |
314b6dd0 | 3833 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
3834 | * date. |
3835 | */ | |
3836 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
3837 | { | |
abbb3b8e DS |
3838 | int i; |
3839 | int errors = 0; | |
b6a535fa | 3840 | bool primary_failed = false; |
12659251 | 3841 | int ret; |
abbb3b8e DS |
3842 | u64 bytenr; |
3843 | ||
3844 | if (max_mirrors == 0) | |
3845 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
3846 | ||
3847 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
3848 | struct page *page; |
3849 | ||
12659251 NA |
3850 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
3851 | if (ret == -ENOENT) { | |
3852 | break; | |
3853 | } else if (ret < 0) { | |
3854 | errors++; | |
3855 | if (i == 0) | |
3856 | primary_failed = true; | |
3857 | continue; | |
3858 | } | |
abbb3b8e DS |
3859 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
3860 | device->commit_total_bytes) | |
3861 | break; | |
3862 | ||
314b6dd0 JT |
3863 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
3864 | bytenr >> PAGE_SHIFT); | |
3865 | if (!page) { | |
abbb3b8e | 3866 | errors++; |
b6a535fa HM |
3867 | if (i == 0) |
3868 | primary_failed = true; | |
abbb3b8e DS |
3869 | continue; |
3870 | } | |
314b6dd0 JT |
3871 | /* Page is submitted locked and unlocked once the IO completes */ |
3872 | wait_on_page_locked(page); | |
3873 | if (PageError(page)) { | |
abbb3b8e | 3874 | errors++; |
b6a535fa HM |
3875 | if (i == 0) |
3876 | primary_failed = true; | |
3877 | } | |
abbb3b8e | 3878 | |
314b6dd0 JT |
3879 | /* Drop our reference */ |
3880 | put_page(page); | |
abbb3b8e | 3881 | |
314b6dd0 JT |
3882 | /* Drop the reference from the writing run */ |
3883 | put_page(page); | |
abbb3b8e DS |
3884 | } |
3885 | ||
b6a535fa HM |
3886 | /* log error, force error return */ |
3887 | if (primary_failed) { | |
3888 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
3889 | device->devid); | |
3890 | return -1; | |
3891 | } | |
3892 | ||
abbb3b8e DS |
3893 | return errors < i ? 0 : -1; |
3894 | } | |
3895 | ||
387125fc CM |
3896 | /* |
3897 | * endio for the write_dev_flush, this will wake anyone waiting | |
3898 | * for the barrier when it is done | |
3899 | */ | |
4246a0b6 | 3900 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 3901 | { |
f9e69aa9 | 3902 | bio_uninit(bio); |
e0ae9994 | 3903 | complete(bio->bi_private); |
387125fc CM |
3904 | } |
3905 | ||
3906 | /* | |
4fc6441a AJ |
3907 | * Submit a flush request to the device if it supports it. Error handling is |
3908 | * done in the waiting counterpart. | |
387125fc | 3909 | */ |
4fc6441a | 3910 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 3911 | { |
f9e69aa9 | 3912 | struct bio *bio = &device->flush_bio; |
387125fc | 3913 | |
bfd3ea94 AJ |
3914 | device->last_flush_error = BLK_STS_OK; |
3915 | ||
f9e69aa9 CH |
3916 | bio_init(bio, device->bdev, NULL, 0, |
3917 | REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); | |
387125fc | 3918 | bio->bi_end_io = btrfs_end_empty_barrier; |
387125fc CM |
3919 | init_completion(&device->flush_wait); |
3920 | bio->bi_private = &device->flush_wait; | |
58ff51f1 | 3921 | submit_bio(bio); |
1c3063b6 | 3922 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 3923 | } |
387125fc | 3924 | |
4fc6441a AJ |
3925 | /* |
3926 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
1b465784 | 3927 | * Return true for any error, and false otherwise. |
4fc6441a | 3928 | */ |
1b465784 | 3929 | static bool wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 3930 | { |
f9e69aa9 | 3931 | struct bio *bio = &device->flush_bio; |
387125fc | 3932 | |
7e812f20 | 3933 | if (!test_and_clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
1b465784 | 3934 | return false; |
387125fc | 3935 | |
2980d574 | 3936 | wait_for_completion_io(&device->flush_wait); |
387125fc | 3937 | |
bfd3ea94 AJ |
3938 | if (bio->bi_status) { |
3939 | device->last_flush_error = bio->bi_status; | |
3940 | btrfs_dev_stat_inc_and_print(device, BTRFS_DEV_STAT_FLUSH_ERRS); | |
1b465784 | 3941 | return true; |
bfd3ea94 AJ |
3942 | } |
3943 | ||
1b465784 | 3944 | return false; |
387125fc | 3945 | } |
387125fc | 3946 | |
387125fc CM |
3947 | /* |
3948 | * send an empty flush down to each device in parallel, | |
3949 | * then wait for them | |
3950 | */ | |
3951 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
3952 | { | |
3953 | struct list_head *head; | |
3954 | struct btrfs_device *dev; | |
5af3e8cc | 3955 | int errors_wait = 0; |
387125fc | 3956 | |
1538e6c5 | 3957 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
3958 | /* send down all the barriers */ |
3959 | head = &info->fs_devices->devices; | |
1538e6c5 | 3960 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3961 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3962 | continue; |
cea7c8bf | 3963 | if (!dev->bdev) |
387125fc | 3964 | continue; |
e12c9621 | 3965 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3966 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3967 | continue; |
3968 | ||
4fc6441a | 3969 | write_dev_flush(dev); |
387125fc CM |
3970 | } |
3971 | ||
3972 | /* wait for all the barriers */ | |
1538e6c5 | 3973 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 3974 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 3975 | continue; |
387125fc | 3976 | if (!dev->bdev) { |
5af3e8cc | 3977 | errors_wait++; |
387125fc CM |
3978 | continue; |
3979 | } | |
e12c9621 | 3980 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 3981 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
3982 | continue; |
3983 | ||
1b465784 | 3984 | if (wait_dev_flush(dev)) |
5af3e8cc | 3985 | errors_wait++; |
401b41e5 AJ |
3986 | } |
3987 | ||
de38a206 AJ |
3988 | /* |
3989 | * Checks last_flush_error of disks in order to determine the device | |
3990 | * state. | |
3991 | */ | |
3992 | if (errors_wait && !btrfs_check_rw_degradable(info, NULL)) | |
3993 | return -EIO; | |
3994 | ||
387125fc CM |
3995 | return 0; |
3996 | } | |
3997 | ||
943c6e99 ZL |
3998 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
3999 | { | |
8789f4fe ZL |
4000 | int raid_type; |
4001 | int min_tolerated = INT_MAX; | |
943c6e99 | 4002 | |
8789f4fe ZL |
4003 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
4004 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 4005 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4006 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4007 | tolerated_failures); | |
943c6e99 | 4008 | |
8789f4fe ZL |
4009 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4010 | if (raid_type == BTRFS_RAID_SINGLE) | |
4011 | continue; | |
41a6e891 | 4012 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4013 | continue; |
8c3e3582 | 4014 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4015 | btrfs_raid_array[raid_type]. |
4016 | tolerated_failures); | |
4017 | } | |
943c6e99 | 4018 | |
8789f4fe | 4019 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4020 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4021 | min_tolerated = 0; |
4022 | } | |
4023 | ||
4024 | return min_tolerated; | |
943c6e99 ZL |
4025 | } |
4026 | ||
eece6a9c | 4027 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4028 | { |
e5e9a520 | 4029 | struct list_head *head; |
f2984462 | 4030 | struct btrfs_device *dev; |
a061fc8d | 4031 | struct btrfs_super_block *sb; |
f2984462 | 4032 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4033 | int ret; |
4034 | int do_barriers; | |
a236aed1 CM |
4035 | int max_errors; |
4036 | int total_errors = 0; | |
a061fc8d | 4037 | u64 flags; |
f2984462 | 4038 | |
0b246afa | 4039 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4040 | |
4041 | /* | |
4042 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4043 | * not from fsync where the tree roots in fs_info have not | |
4044 | * been consistent on disk. | |
4045 | */ | |
4046 | if (max_mirrors == 0) | |
4047 | backup_super_roots(fs_info); | |
f2984462 | 4048 | |
0b246afa | 4049 | sb = fs_info->super_for_commit; |
a061fc8d | 4050 | dev_item = &sb->dev_item; |
e5e9a520 | 4051 | |
0b246afa JM |
4052 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4053 | head = &fs_info->fs_devices->devices; | |
4054 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4055 | |
5af3e8cc | 4056 | if (do_barriers) { |
0b246afa | 4057 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4058 | if (ret) { |
4059 | mutex_unlock( | |
0b246afa JM |
4060 | &fs_info->fs_devices->device_list_mutex); |
4061 | btrfs_handle_fs_error(fs_info, ret, | |
4062 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4063 | return ret; |
4064 | } | |
4065 | } | |
387125fc | 4066 | |
1538e6c5 | 4067 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4068 | if (!dev->bdev) { |
4069 | total_errors++; | |
4070 | continue; | |
4071 | } | |
e12c9621 | 4072 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4073 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4074 | continue; |
4075 | ||
2b82032c | 4076 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4077 | btrfs_set_stack_device_type(dev_item, dev->type); |
4078 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4079 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4080 | dev->commit_total_bytes); |
ce7213c7 MX |
4081 | btrfs_set_stack_device_bytes_used(dev_item, |
4082 | dev->commit_bytes_used); | |
a061fc8d CM |
4083 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4084 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4085 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4086 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4087 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4088 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4089 | |
a061fc8d CM |
4090 | flags = btrfs_super_flags(sb); |
4091 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4092 | ||
75cb857d QW |
4093 | ret = btrfs_validate_write_super(fs_info, sb); |
4094 | if (ret < 0) { | |
4095 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4096 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4097 | "unexpected superblock corruption detected"); | |
4098 | return -EUCLEAN; | |
4099 | } | |
4100 | ||
abbb3b8e | 4101 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4102 | if (ret) |
4103 | total_errors++; | |
f2984462 | 4104 | } |
a236aed1 | 4105 | if (total_errors > max_errors) { |
0b246afa JM |
4106 | btrfs_err(fs_info, "%d errors while writing supers", |
4107 | total_errors); | |
4108 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4109 | |
9d565ba4 | 4110 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4111 | btrfs_handle_fs_error(fs_info, -EIO, |
4112 | "%d errors while writing supers", | |
4113 | total_errors); | |
9d565ba4 | 4114 | return -EIO; |
a236aed1 | 4115 | } |
f2984462 | 4116 | |
a512bbf8 | 4117 | total_errors = 0; |
1538e6c5 | 4118 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4119 | if (!dev->bdev) |
4120 | continue; | |
e12c9621 | 4121 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4122 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4123 | continue; |
4124 | ||
abbb3b8e | 4125 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4126 | if (ret) |
4127 | total_errors++; | |
f2984462 | 4128 | } |
0b246afa | 4129 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4130 | if (total_errors > max_errors) { |
0b246afa JM |
4131 | btrfs_handle_fs_error(fs_info, -EIO, |
4132 | "%d errors while writing supers", | |
4133 | total_errors); | |
79787eaa | 4134 | return -EIO; |
a236aed1 | 4135 | } |
f2984462 CM |
4136 | return 0; |
4137 | } | |
4138 | ||
cb517eab MX |
4139 | /* Drop a fs root from the radix tree and free it. */ |
4140 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4141 | struct btrfs_root *root) | |
2619ba1f | 4142 | { |
4785e24f JB |
4143 | bool drop_ref = false; |
4144 | ||
fc7cbcd4 DS |
4145 | spin_lock(&fs_info->fs_roots_radix_lock); |
4146 | radix_tree_delete(&fs_info->fs_roots_radix, | |
4147 | (unsigned long)root->root_key.objectid); | |
4148 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) | |
4785e24f | 4149 | drop_ref = true; |
fc7cbcd4 | 4150 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4151 | |
84961539 | 4152 | if (BTRFS_FS_ERROR(fs_info)) { |
ef67963d | 4153 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4154 | if (root->reloc_root) { |
00246528 | 4155 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4156 | root->reloc_root = NULL; |
4157 | } | |
4158 | } | |
3321719e | 4159 | |
4785e24f JB |
4160 | if (drop_ref) |
4161 | btrfs_put_root(root); | |
2619ba1f CM |
4162 | } |
4163 | ||
6bccf3ab | 4164 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4165 | { |
6bccf3ab | 4166 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4167 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4168 | |
0b246afa | 4169 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4170 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4171 | mutex_unlock(&fs_info->cleaner_mutex); |
4172 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4173 | |
4174 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4175 | down_write(&fs_info->cleanup_work_sem); |
4176 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4177 | |
7a7eaa40 | 4178 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4179 | if (IS_ERR(trans)) |
4180 | return PTR_ERR(trans); | |
3a45bb20 | 4181 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4182 | } |
4183 | ||
36c86a9e QW |
4184 | static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) |
4185 | { | |
4186 | struct btrfs_transaction *trans; | |
4187 | struct btrfs_transaction *tmp; | |
4188 | bool found = false; | |
4189 | ||
4190 | if (list_empty(&fs_info->trans_list)) | |
4191 | return; | |
4192 | ||
4193 | /* | |
4194 | * This function is only called at the very end of close_ctree(), | |
4195 | * thus no other running transaction, no need to take trans_lock. | |
4196 | */ | |
4197 | ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); | |
4198 | list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { | |
4199 | struct extent_state *cached = NULL; | |
4200 | u64 dirty_bytes = 0; | |
4201 | u64 cur = 0; | |
4202 | u64 found_start; | |
4203 | u64 found_end; | |
4204 | ||
4205 | found = true; | |
e5860f82 | 4206 | while (find_first_extent_bit(&trans->dirty_pages, cur, |
36c86a9e QW |
4207 | &found_start, &found_end, EXTENT_DIRTY, &cached)) { |
4208 | dirty_bytes += found_end + 1 - found_start; | |
4209 | cur = found_end + 1; | |
4210 | } | |
4211 | btrfs_warn(fs_info, | |
4212 | "transaction %llu (with %llu dirty metadata bytes) is not committed", | |
4213 | trans->transid, dirty_bytes); | |
4214 | btrfs_cleanup_one_transaction(trans, fs_info); | |
4215 | ||
4216 | if (trans == fs_info->running_transaction) | |
4217 | fs_info->running_transaction = NULL; | |
4218 | list_del_init(&trans->list); | |
4219 | ||
4220 | btrfs_put_transaction(trans); | |
4221 | trace_btrfs_transaction_commit(fs_info); | |
4222 | } | |
4223 | ASSERT(!found); | |
4224 | } | |
4225 | ||
b105e927 | 4226 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4227 | { |
c146afad YZ |
4228 | int ret; |
4229 | ||
afcdd129 | 4230 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
31e70e52 | 4231 | |
8a1f1e3d FM |
4232 | /* |
4233 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4234 | * clear that bit and wake up relocation so it can stop. | |
4235 | * We must do this before stopping the block group reclaim task, because | |
4236 | * at btrfs_relocate_block_group() we wait for this bit, and after the | |
4237 | * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we | |
4238 | * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will | |
4239 | * return 1. | |
4240 | */ | |
4241 | btrfs_wake_unfinished_drop(fs_info); | |
4242 | ||
31e70e52 FM |
4243 | /* |
4244 | * We may have the reclaim task running and relocating a data block group, | |
4245 | * in which case it may create delayed iputs. So stop it before we park | |
4246 | * the cleaner kthread otherwise we can get new delayed iputs after | |
4247 | * parking the cleaner, and that can make the async reclaim task to hang | |
4248 | * if it's waiting for delayed iputs to complete, since the cleaner is | |
4249 | * parked and can not run delayed iputs - this will make us hang when | |
4250 | * trying to stop the async reclaim task. | |
4251 | */ | |
4252 | cancel_work_sync(&fs_info->reclaim_bgs_work); | |
d6fd0ae2 OS |
4253 | /* |
4254 | * We don't want the cleaner to start new transactions, add more delayed | |
4255 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4256 | * because that frees the task_struct, and the transaction kthread might | |
4257 | * still try to wake up the cleaner. | |
4258 | */ | |
4259 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4260 | |
7343dd61 | 4261 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4262 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4263 | |
803b2f54 SB |
4264 | /* wait for the uuid_scan task to finish */ |
4265 | down(&fs_info->uuid_tree_rescan_sem); | |
4266 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4267 | up(&fs_info->uuid_tree_rescan_sem); | |
4268 | ||
837d5b6e | 4269 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4270 | btrfs_pause_balance(fs_info); |
837d5b6e | 4271 | |
8dabb742 SB |
4272 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4273 | ||
aa1b8cd4 | 4274 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4275 | |
4276 | /* wait for any defraggers to finish */ | |
4277 | wait_event(fs_info->transaction_wait, | |
4278 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4279 | ||
4280 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4281 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4282 | |
a362bb86 FM |
4283 | /* |
4284 | * After we parked the cleaner kthread, ordered extents may have | |
4285 | * completed and created new delayed iputs. If one of the async reclaim | |
4286 | * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we | |
4287 | * can hang forever trying to stop it, because if a delayed iput is | |
4288 | * added after it ran btrfs_run_delayed_iputs() and before it called | |
4289 | * btrfs_wait_on_delayed_iputs(), it will hang forever since there is | |
4290 | * no one else to run iputs. | |
4291 | * | |
4292 | * So wait for all ongoing ordered extents to complete and then run | |
4293 | * delayed iputs. This works because once we reach this point no one | |
4294 | * can either create new ordered extents nor create delayed iputs | |
4295 | * through some other means. | |
4296 | * | |
4297 | * Also note that btrfs_wait_ordered_roots() is not safe here, because | |
4298 | * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, | |
4299 | * but the delayed iput for the respective inode is made only when doing | |
4300 | * the final btrfs_put_ordered_extent() (which must happen at | |
4301 | * btrfs_finish_ordered_io() when we are unmounting). | |
4302 | */ | |
4303 | btrfs_flush_workqueue(fs_info->endio_write_workers); | |
4304 | /* Ordered extents for free space inodes. */ | |
4305 | btrfs_flush_workqueue(fs_info->endio_freespace_worker); | |
4306 | btrfs_run_delayed_iputs(fs_info); | |
4307 | ||
21c7e756 | 4308 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4309 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4310 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4311 | |
b0643e59 DZ |
4312 | /* Cancel or finish ongoing discard work */ |
4313 | btrfs_discard_cleanup(fs_info); | |
4314 | ||
bc98a42c | 4315 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4316 | /* |
d6fd0ae2 OS |
4317 | * The cleaner kthread is stopped, so do one final pass over |
4318 | * unused block groups. | |
e44163e1 | 4319 | */ |
0b246afa | 4320 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4321 | |
f0cc2cd7 FM |
4322 | /* |
4323 | * There might be existing delayed inode workers still running | |
4324 | * and holding an empty delayed inode item. We must wait for | |
4325 | * them to complete first because they can create a transaction. | |
4326 | * This happens when someone calls btrfs_balance_delayed_items() | |
4327 | * and then a transaction commit runs the same delayed nodes | |
4328 | * before any delayed worker has done something with the nodes. | |
4329 | * We must wait for any worker here and not at transaction | |
4330 | * commit time since that could cause a deadlock. | |
4331 | * This is a very rare case. | |
4332 | */ | |
4333 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4334 | ||
6bccf3ab | 4335 | ret = btrfs_commit_super(fs_info); |
acce952b | 4336 | if (ret) |
04892340 | 4337 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4338 | } |
4339 | ||
84961539 | 4340 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 4341 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4342 | |
e3029d9f AV |
4343 | kthread_stop(fs_info->transaction_kthread); |
4344 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4345 | |
e187831e | 4346 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4347 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4348 | |
5958253c QW |
4349 | if (btrfs_check_quota_leak(fs_info)) { |
4350 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4351 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4352 | } | |
4353 | ||
04892340 | 4354 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4355 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4356 | |
963d678b | 4357 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4358 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4359 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4360 | } |
bcc63abb | 4361 | |
5deb17e1 | 4362 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4363 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4364 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4365 | |
6618a59b | 4366 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4367 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4368 | |
1a4319cc LB |
4369 | btrfs_put_block_group_cache(fs_info); |
4370 | ||
de348ee0 WS |
4371 | /* |
4372 | * we must make sure there is not any read request to | |
4373 | * submit after we stopping all workers. | |
4374 | */ | |
4375 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4376 | btrfs_stop_all_workers(fs_info); |
4377 | ||
0a31daa4 | 4378 | /* We shouldn't have any transaction open at this point */ |
36c86a9e | 4379 | warn_about_uncommitted_trans(fs_info); |
0a31daa4 | 4380 | |
afcdd129 | 4381 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4382 | free_root_pointers(fs_info, true); |
8c38938c | 4383 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4384 | |
4e19443d JB |
4385 | /* |
4386 | * We must free the block groups after dropping the fs_roots as we could | |
4387 | * have had an IO error and have left over tree log blocks that aren't | |
4388 | * cleaned up until the fs roots are freed. This makes the block group | |
4389 | * accounting appear to be wrong because there's pending reserved bytes, | |
4390 | * so make sure we do the block group cleanup afterwards. | |
4391 | */ | |
4392 | btrfs_free_block_groups(fs_info); | |
4393 | ||
13e6c37b | 4394 | iput(fs_info->btree_inode); |
d6bfde87 | 4395 | |
7dc66abb | 4396 | btrfs_mapping_tree_free(fs_info); |
68c94e55 | 4397 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4398 | } |
4399 | ||
50564b65 FM |
4400 | void btrfs_mark_buffer_dirty(struct btrfs_trans_handle *trans, |
4401 | struct extent_buffer *buf) | |
5f39d397 | 4402 | { |
2f4d60df | 4403 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4404 | u64 transid = btrfs_header_generation(buf); |
b4ce94de | 4405 | |
06ea65a3 JB |
4406 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4407 | /* | |
4408 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4409 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4410 | * outside of the sanity tests. |
4411 | */ | |
b0132a3b | 4412 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4413 | return; |
4414 | #endif | |
50564b65 FM |
4415 | /* This is an active transaction (its state < TRANS_STATE_UNBLOCKED). */ |
4416 | ASSERT(trans->transid == fs_info->generation); | |
49d0c642 | 4417 | btrfs_assert_tree_write_locked(buf); |
4ebe8d47 | 4418 | if (unlikely(transid != fs_info->generation)) { |
50564b65 | 4419 | btrfs_abort_transaction(trans, -EUCLEAN); |
20cbe460 FM |
4420 | btrfs_crit(fs_info, |
4421 | "dirty buffer transid mismatch, logical %llu found transid %llu running transid %llu", | |
4422 | buf->start, transid, fs_info->generation); | |
50564b65 | 4423 | } |
f18cc978 | 4424 | set_extent_buffer_dirty(buf); |
eb60ceac CM |
4425 | } |
4426 | ||
2ff7e61e | 4427 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4428 | int flush_delayed) |
16cdcec7 MX |
4429 | { |
4430 | /* | |
4431 | * looks as though older kernels can get into trouble with | |
4432 | * this code, they end up stuck in balance_dirty_pages forever | |
4433 | */ | |
e2d84521 | 4434 | int ret; |
16cdcec7 MX |
4435 | |
4436 | if (current->flags & PF_MEMALLOC) | |
4437 | return; | |
4438 | ||
b53d3f5d | 4439 | if (flush_delayed) |
2ff7e61e | 4440 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4441 | |
d814a491 EL |
4442 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4443 | BTRFS_DIRTY_METADATA_THRESH, | |
4444 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4445 | if (ret > 0) { |
0b246afa | 4446 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4447 | } |
16cdcec7 MX |
4448 | } |
4449 | ||
2ff7e61e | 4450 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4451 | { |
2ff7e61e | 4452 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4453 | } |
585ad2c3 | 4454 | |
2ff7e61e | 4455 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4456 | { |
2ff7e61e | 4457 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4458 | } |
6b80053d | 4459 | |
2ff7e61e | 4460 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4461 | { |
fe816d0f NB |
4462 | /* cleanup FS via transaction */ |
4463 | btrfs_cleanup_transaction(fs_info); | |
4464 | ||
0b246afa | 4465 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4466 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4467 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4468 | |
0b246afa JM |
4469 | down_write(&fs_info->cleanup_work_sem); |
4470 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4471 | } |
4472 | ||
ef67963d JB |
4473 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4474 | { | |
fc7cbcd4 DS |
4475 | struct btrfs_root *gang[8]; |
4476 | u64 root_objectid = 0; | |
4477 | int ret; | |
4478 | ||
4479 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4480 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4481 | (void **)gang, root_objectid, | |
4482 | ARRAY_SIZE(gang))) != 0) { | |
4483 | int i; | |
4484 | ||
4485 | for (i = 0; i < ret; i++) | |
4486 | gang[i] = btrfs_grab_root(gang[i]); | |
4487 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4488 | ||
4489 | for (i = 0; i < ret; i++) { | |
4490 | if (!gang[i]) | |
ef67963d | 4491 | continue; |
fc7cbcd4 DS |
4492 | root_objectid = gang[i]->root_key.objectid; |
4493 | btrfs_free_log(NULL, gang[i]); | |
4494 | btrfs_put_root(gang[i]); | |
ef67963d | 4495 | } |
fc7cbcd4 DS |
4496 | root_objectid++; |
4497 | spin_lock(&fs_info->fs_roots_radix_lock); | |
ef67963d | 4498 | } |
fc7cbcd4 | 4499 | spin_unlock(&fs_info->fs_roots_radix_lock); |
ef67963d JB |
4500 | btrfs_free_log_root_tree(NULL, fs_info); |
4501 | } | |
4502 | ||
143bede5 | 4503 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4504 | { |
acce952b | 4505 | struct btrfs_ordered_extent *ordered; |
acce952b | 4506 | |
199c2a9c | 4507 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4508 | /* |
4509 | * This will just short circuit the ordered completion stuff which will | |
4510 | * make sure the ordered extent gets properly cleaned up. | |
4511 | */ | |
199c2a9c | 4512 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4513 | root_extent_list) |
4514 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4515 | spin_unlock(&root->ordered_extent_lock); |
4516 | } | |
4517 | ||
4518 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4519 | { | |
4520 | struct btrfs_root *root; | |
84af994b | 4521 | LIST_HEAD(splice); |
199c2a9c MX |
4522 | |
4523 | spin_lock(&fs_info->ordered_root_lock); | |
4524 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4525 | while (!list_empty(&splice)) { | |
4526 | root = list_first_entry(&splice, struct btrfs_root, | |
4527 | ordered_root); | |
1de2cfde JB |
4528 | list_move_tail(&root->ordered_root, |
4529 | &fs_info->ordered_roots); | |
199c2a9c | 4530 | |
2a85d9ca | 4531 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4532 | btrfs_destroy_ordered_extents(root); |
4533 | ||
2a85d9ca LB |
4534 | cond_resched(); |
4535 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4536 | } |
4537 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4538 | |
4539 | /* | |
4540 | * We need this here because if we've been flipped read-only we won't | |
4541 | * get sync() from the umount, so we need to make sure any ordered | |
4542 | * extents that haven't had their dirty pages IO start writeout yet | |
4543 | * actually get run and error out properly. | |
4544 | */ | |
4545 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4546 | } |
4547 | ||
99f09ce3 FM |
4548 | static void btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
4549 | struct btrfs_fs_info *fs_info) | |
acce952b | 4550 | { |
4551 | struct rb_node *node; | |
4552 | struct btrfs_delayed_ref_root *delayed_refs; | |
4553 | struct btrfs_delayed_ref_node *ref; | |
acce952b | 4554 | |
4555 | delayed_refs = &trans->delayed_refs; | |
4556 | ||
4557 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4558 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4559 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4560 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
99f09ce3 | 4561 | return; |
acce952b | 4562 | } |
4563 | ||
5c9d028b | 4564 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4565 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4566 | struct rb_node *n; |
e78417d1 | 4567 | bool pin_bytes = false; |
acce952b | 4568 | |
d7df2c79 JB |
4569 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4570 | href_node); | |
3069bd26 | 4571 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4572 | continue; |
3069bd26 | 4573 | |
d7df2c79 | 4574 | spin_lock(&head->lock); |
e3d03965 | 4575 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4576 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
4577 | ref_node); | |
e3d03965 | 4578 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 4579 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
4580 | if (!list_empty(&ref->add_list)) |
4581 | list_del(&ref->add_list); | |
d7df2c79 JB |
4582 | atomic_dec(&delayed_refs->num_entries); |
4583 | btrfs_put_delayed_ref(ref); | |
adb86dbe | 4584 | btrfs_delayed_refs_rsv_release(fs_info, 1, 0); |
e78417d1 | 4585 | } |
d7df2c79 JB |
4586 | if (head->must_insert_reserved) |
4587 | pin_bytes = true; | |
4588 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 4589 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
4590 | spin_unlock(&head->lock); |
4591 | spin_unlock(&delayed_refs->lock); | |
4592 | mutex_unlock(&head->mutex); | |
acce952b | 4593 | |
f603bb94 NB |
4594 | if (pin_bytes) { |
4595 | struct btrfs_block_group *cache; | |
4596 | ||
4597 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
4598 | BUG_ON(!cache); | |
4599 | ||
4600 | spin_lock(&cache->space_info->lock); | |
4601 | spin_lock(&cache->lock); | |
4602 | cache->pinned += head->num_bytes; | |
4603 | btrfs_space_info_update_bytes_pinned(fs_info, | |
4604 | cache->space_info, head->num_bytes); | |
4605 | cache->reserved -= head->num_bytes; | |
4606 | cache->space_info->bytes_reserved -= head->num_bytes; | |
4607 | spin_unlock(&cache->lock); | |
4608 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
4609 | |
4610 | btrfs_put_block_group(cache); | |
4611 | ||
4612 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
4613 | head->bytenr + head->num_bytes - 1); | |
4614 | } | |
31890da0 | 4615 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 4616 | btrfs_put_delayed_ref_head(head); |
acce952b | 4617 | cond_resched(); |
4618 | spin_lock(&delayed_refs->lock); | |
4619 | } | |
81f7eb00 | 4620 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 4621 | |
4622 | spin_unlock(&delayed_refs->lock); | |
acce952b | 4623 | } |
4624 | ||
143bede5 | 4625 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 4626 | { |
4627 | struct btrfs_inode *btrfs_inode; | |
84af994b | 4628 | LIST_HEAD(splice); |
acce952b | 4629 | |
eb73c1b7 MX |
4630 | spin_lock(&root->delalloc_lock); |
4631 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 4632 | |
4633 | while (!list_empty(&splice)) { | |
fe816d0f | 4634 | struct inode *inode = NULL; |
eb73c1b7 MX |
4635 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
4636 | delalloc_inodes); | |
fe816d0f | 4637 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 4638 | spin_unlock(&root->delalloc_lock); |
acce952b | 4639 | |
fe816d0f NB |
4640 | /* |
4641 | * Make sure we get a live inode and that it'll not disappear | |
4642 | * meanwhile. | |
4643 | */ | |
4644 | inode = igrab(&btrfs_inode->vfs_inode); | |
4645 | if (inode) { | |
597441b3 JB |
4646 | unsigned int nofs_flag; |
4647 | ||
4648 | nofs_flag = memalloc_nofs_save(); | |
fe816d0f | 4649 | invalidate_inode_pages2(inode->i_mapping); |
597441b3 | 4650 | memalloc_nofs_restore(nofs_flag); |
fe816d0f NB |
4651 | iput(inode); |
4652 | } | |
eb73c1b7 | 4653 | spin_lock(&root->delalloc_lock); |
acce952b | 4654 | } |
eb73c1b7 MX |
4655 | spin_unlock(&root->delalloc_lock); |
4656 | } | |
4657 | ||
4658 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
4659 | { | |
4660 | struct btrfs_root *root; | |
84af994b | 4661 | LIST_HEAD(splice); |
eb73c1b7 MX |
4662 | |
4663 | spin_lock(&fs_info->delalloc_root_lock); | |
4664 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
4665 | while (!list_empty(&splice)) { | |
4666 | root = list_first_entry(&splice, struct btrfs_root, | |
4667 | delalloc_root); | |
00246528 | 4668 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
4669 | BUG_ON(!root); |
4670 | spin_unlock(&fs_info->delalloc_root_lock); | |
4671 | ||
4672 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 4673 | btrfs_put_root(root); |
eb73c1b7 MX |
4674 | |
4675 | spin_lock(&fs_info->delalloc_root_lock); | |
4676 | } | |
4677 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 4678 | } |
4679 | ||
aec5716c FM |
4680 | static void btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
4681 | struct extent_io_tree *dirty_pages, | |
4682 | int mark) | |
acce952b | 4683 | { |
acce952b | 4684 | struct extent_buffer *eb; |
4685 | u64 start = 0; | |
4686 | u64 end; | |
acce952b | 4687 | |
e5860f82 FM |
4688 | while (find_first_extent_bit(dirty_pages, start, &start, &end, |
4689 | mark, NULL)) { | |
91166212 | 4690 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 4691 | while (start <= end) { |
0b246afa JM |
4692 | eb = find_extent_buffer(fs_info, start); |
4693 | start += fs_info->nodesize; | |
fd8b2b61 | 4694 | if (!eb) |
acce952b | 4695 | continue; |
c4e54a65 JB |
4696 | |
4697 | btrfs_tree_lock(eb); | |
fd8b2b61 | 4698 | wait_on_extent_buffer_writeback(eb); |
190a8339 | 4699 | btrfs_clear_buffer_dirty(NULL, eb); |
c4e54a65 | 4700 | btrfs_tree_unlock(eb); |
acce952b | 4701 | |
fd8b2b61 | 4702 | free_extent_buffer_stale(eb); |
acce952b | 4703 | } |
4704 | } | |
acce952b | 4705 | } |
4706 | ||
46d81ebd FM |
4707 | static void btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
4708 | struct extent_io_tree *unpin) | |
acce952b | 4709 | { |
acce952b | 4710 | u64 start; |
4711 | u64 end; | |
acce952b | 4712 | |
acce952b | 4713 | while (1) { |
0e6ec385 FM |
4714 | struct extent_state *cached_state = NULL; |
4715 | ||
fcd5e742 LF |
4716 | /* |
4717 | * The btrfs_finish_extent_commit() may get the same range as | |
4718 | * ours between find_first_extent_bit and clear_extent_dirty. | |
4719 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
4720 | * the same extent range. | |
4721 | */ | |
4722 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
e5860f82 FM |
4723 | if (!find_first_extent_bit(unpin, 0, &start, &end, |
4724 | EXTENT_DIRTY, &cached_state)) { | |
fcd5e742 | 4725 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4726 | break; |
fcd5e742 | 4727 | } |
acce952b | 4728 | |
0e6ec385 FM |
4729 | clear_extent_dirty(unpin, start, end, &cached_state); |
4730 | free_extent_state(cached_state); | |
2ff7e61e | 4731 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 4732 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 4733 | cond_resched(); |
4734 | } | |
acce952b | 4735 | } |
4736 | ||
32da5386 | 4737 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
4738 | { |
4739 | struct inode *inode; | |
4740 | ||
4741 | inode = cache->io_ctl.inode; | |
4742 | if (inode) { | |
597441b3 JB |
4743 | unsigned int nofs_flag; |
4744 | ||
4745 | nofs_flag = memalloc_nofs_save(); | |
c79a1751 | 4746 | invalidate_inode_pages2(inode->i_mapping); |
597441b3 JB |
4747 | memalloc_nofs_restore(nofs_flag); |
4748 | ||
c79a1751 LB |
4749 | BTRFS_I(inode)->generation = 0; |
4750 | cache->io_ctl.inode = NULL; | |
4751 | iput(inode); | |
4752 | } | |
bbc37d6e | 4753 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
4754 | btrfs_put_block_group(cache); |
4755 | } | |
4756 | ||
4757 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 4758 | struct btrfs_fs_info *fs_info) |
c79a1751 | 4759 | { |
32da5386 | 4760 | struct btrfs_block_group *cache; |
c79a1751 LB |
4761 | |
4762 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4763 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
4764 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 4765 | struct btrfs_block_group, |
c79a1751 | 4766 | dirty_list); |
c79a1751 LB |
4767 | |
4768 | if (!list_empty(&cache->io_list)) { | |
4769 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4770 | list_del_init(&cache->io_list); | |
4771 | btrfs_cleanup_bg_io(cache); | |
4772 | spin_lock(&cur_trans->dirty_bgs_lock); | |
4773 | } | |
4774 | ||
4775 | list_del_init(&cache->dirty_list); | |
4776 | spin_lock(&cache->lock); | |
4777 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4778 | spin_unlock(&cache->lock); | |
4779 | ||
4780 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4781 | btrfs_put_block_group(cache); | |
f66e0209 | 4782 | btrfs_dec_delayed_refs_rsv_bg_updates(fs_info); |
c79a1751 LB |
4783 | spin_lock(&cur_trans->dirty_bgs_lock); |
4784 | } | |
4785 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
4786 | ||
45ae2c18 NB |
4787 | /* |
4788 | * Refer to the definition of io_bgs member for details why it's safe | |
4789 | * to use it without any locking | |
4790 | */ | |
c79a1751 LB |
4791 | while (!list_empty(&cur_trans->io_bgs)) { |
4792 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 4793 | struct btrfs_block_group, |
c79a1751 | 4794 | io_list); |
c79a1751 LB |
4795 | |
4796 | list_del_init(&cache->io_list); | |
4797 | spin_lock(&cache->lock); | |
4798 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
4799 | spin_unlock(&cache->lock); | |
4800 | btrfs_cleanup_bg_io(cache); | |
4801 | } | |
4802 | } | |
4803 | ||
b321a52c BB |
4804 | static void btrfs_free_all_qgroup_pertrans(struct btrfs_fs_info *fs_info) |
4805 | { | |
4806 | struct btrfs_root *gang[8]; | |
4807 | int i; | |
4808 | int ret; | |
4809 | ||
4810 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4811 | while (1) { | |
4812 | ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix, | |
4813 | (void **)gang, 0, | |
4814 | ARRAY_SIZE(gang), | |
4815 | BTRFS_ROOT_TRANS_TAG); | |
4816 | if (ret == 0) | |
4817 | break; | |
4818 | for (i = 0; i < ret; i++) { | |
4819 | struct btrfs_root *root = gang[i]; | |
4820 | ||
4821 | btrfs_qgroup_free_meta_all_pertrans(root); | |
4822 | radix_tree_tag_clear(&fs_info->fs_roots_radix, | |
4823 | (unsigned long)root->root_key.objectid, | |
4824 | BTRFS_ROOT_TRANS_TAG); | |
4825 | } | |
4826 | } | |
4827 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4828 | } | |
4829 | ||
49b25e05 | 4830 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 4831 | struct btrfs_fs_info *fs_info) |
49b25e05 | 4832 | { |
bbbf7243 NB |
4833 | struct btrfs_device *dev, *tmp; |
4834 | ||
2ff7e61e | 4835 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
4836 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
4837 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
4838 | ||
bbbf7243 NB |
4839 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
4840 | post_commit_list) { | |
4841 | list_del_init(&dev->post_commit_list); | |
4842 | } | |
4843 | ||
2ff7e61e | 4844 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 4845 | |
4a9d8bde | 4846 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 4847 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 4848 | |
4a9d8bde | 4849 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 4850 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 4851 | |
ccdf9b30 | 4852 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 4853 | |
2ff7e61e | 4854 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 4855 | EXTENT_DIRTY); |
fe119a6e | 4856 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 4857 | |
b321a52c BB |
4858 | btrfs_free_all_qgroup_pertrans(fs_info); |
4859 | ||
4a9d8bde MX |
4860 | cur_trans->state =TRANS_STATE_COMPLETED; |
4861 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
4862 | } |
4863 | ||
2ff7e61e | 4864 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 4865 | { |
4866 | struct btrfs_transaction *t; | |
acce952b | 4867 | |
0b246afa | 4868 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 4869 | |
0b246afa JM |
4870 | spin_lock(&fs_info->trans_lock); |
4871 | while (!list_empty(&fs_info->trans_list)) { | |
4872 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 | 4873 | struct btrfs_transaction, list); |
77d20c68 | 4874 | if (t->state >= TRANS_STATE_COMMIT_PREP) { |
9b64f57d | 4875 | refcount_inc(&t->use_count); |
0b246afa | 4876 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 4877 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 4878 | btrfs_put_transaction(t); |
0b246afa | 4879 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
4880 | continue; |
4881 | } | |
0b246afa | 4882 | if (t == fs_info->running_transaction) { |
724e2315 | 4883 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 4884 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
4885 | /* |
4886 | * We wait for 0 num_writers since we don't hold a trans | |
4887 | * handle open currently for this transaction. | |
4888 | */ | |
4889 | wait_event(t->writer_wait, | |
4890 | atomic_read(&t->num_writers) == 0); | |
4891 | } else { | |
0b246afa | 4892 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 4893 | } |
2ff7e61e | 4894 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 4895 | |
0b246afa JM |
4896 | spin_lock(&fs_info->trans_lock); |
4897 | if (t == fs_info->running_transaction) | |
4898 | fs_info->running_transaction = NULL; | |
acce952b | 4899 | list_del_init(&t->list); |
0b246afa | 4900 | spin_unlock(&fs_info->trans_lock); |
acce952b | 4901 | |
724e2315 | 4902 | btrfs_put_transaction(t); |
2e4e97ab | 4903 | trace_btrfs_transaction_commit(fs_info); |
0b246afa | 4904 | spin_lock(&fs_info->trans_lock); |
724e2315 | 4905 | } |
0b246afa JM |
4906 | spin_unlock(&fs_info->trans_lock); |
4907 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
4908 | btrfs_destroy_delayed_inodes(fs_info); |
4909 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 4910 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 4911 | btrfs_drop_all_logs(fs_info); |
0b246afa | 4912 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 4913 | |
4914 | return 0; | |
4915 | } | |
ec7d6dfd | 4916 | |
453e4873 | 4917 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
4918 | { |
4919 | struct btrfs_path *path; | |
4920 | int ret; | |
4921 | struct extent_buffer *l; | |
4922 | struct btrfs_key search_key; | |
4923 | struct btrfs_key found_key; | |
4924 | int slot; | |
4925 | ||
4926 | path = btrfs_alloc_path(); | |
4927 | if (!path) | |
4928 | return -ENOMEM; | |
4929 | ||
4930 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
4931 | search_key.type = -1; | |
4932 | search_key.offset = (u64)-1; | |
4933 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
4934 | if (ret < 0) | |
4935 | goto error; | |
4936 | BUG_ON(ret == 0); /* Corruption */ | |
4937 | if (path->slots[0] > 0) { | |
4938 | slot = path->slots[0] - 1; | |
4939 | l = path->nodes[0]; | |
4940 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
4941 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
4942 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 4943 | } else { |
23125104 | 4944 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
4945 | } |
4946 | ret = 0; | |
4947 | error: | |
4948 | btrfs_free_path(path); | |
4949 | return ret; | |
4950 | } | |
4951 | ||
543068a2 | 4952 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
4953 | { |
4954 | int ret; | |
4955 | mutex_lock(&root->objectid_mutex); | |
4956 | ||
6b8fad57 | 4957 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
4958 | btrfs_warn(root->fs_info, |
4959 | "the objectid of root %llu reaches its highest value", | |
4960 | root->root_key.objectid); | |
4961 | ret = -ENOSPC; | |
4962 | goto out; | |
4963 | } | |
4964 | ||
23125104 | 4965 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
4966 | ret = 0; |
4967 | out: | |
4968 | mutex_unlock(&root->objectid_mutex); | |
4969 | return ret; | |
4970 | } |