Commit | Line | Data |
---|---|---|
c1d7c514 | 1 | // SPDX-License-Identifier: GPL-2.0 |
6cbd5570 CM |
2 | /* |
3 | * Copyright (C) 2007 Oracle. All rights reserved. | |
6cbd5570 CM |
4 | */ |
5 | ||
e20d96d6 | 6 | #include <linux/fs.h> |
d98237b3 | 7 | #include <linux/blkdev.h> |
fc7cbcd4 | 8 | #include <linux/radix-tree.h> |
35b7e476 | 9 | #include <linux/writeback.h> |
ce9adaa5 | 10 | #include <linux/workqueue.h> |
a74a4b97 | 11 | #include <linux/kthread.h> |
5a0e3ad6 | 12 | #include <linux/slab.h> |
784b4e29 | 13 | #include <linux/migrate.h> |
7a36ddec | 14 | #include <linux/ratelimit.h> |
6463fe58 | 15 | #include <linux/uuid.h> |
803b2f54 | 16 | #include <linux/semaphore.h> |
540adea3 | 17 | #include <linux/error-injection.h> |
9678c543 | 18 | #include <linux/crc32c.h> |
b89f6d1f | 19 | #include <linux/sched/mm.h> |
7e75bf3f | 20 | #include <asm/unaligned.h> |
6d97c6e3 | 21 | #include <crypto/hash.h> |
eb60ceac CM |
22 | #include "ctree.h" |
23 | #include "disk-io.h" | |
e089f05c | 24 | #include "transaction.h" |
0f7d52f4 | 25 | #include "btrfs_inode.h" |
0b86a832 | 26 | #include "volumes.h" |
db94535d | 27 | #include "print-tree.h" |
925baedd | 28 | #include "locking.h" |
e02119d5 | 29 | #include "tree-log.h" |
fa9c0d79 | 30 | #include "free-space-cache.h" |
70f6d82e | 31 | #include "free-space-tree.h" |
21adbd5c | 32 | #include "check-integrity.h" |
606686ee | 33 | #include "rcu-string.h" |
8dabb742 | 34 | #include "dev-replace.h" |
53b381b3 | 35 | #include "raid56.h" |
5ac1d209 | 36 | #include "sysfs.h" |
fcebe456 | 37 | #include "qgroup.h" |
ebb8765b | 38 | #include "compression.h" |
557ea5dd | 39 | #include "tree-checker.h" |
fd708b81 | 40 | #include "ref-verify.h" |
aac0023c | 41 | #include "block-group.h" |
b0643e59 | 42 | #include "discard.h" |
f603bb94 | 43 | #include "space-info.h" |
b70f5097 | 44 | #include "zoned.h" |
139e8cd3 | 45 | #include "subpage.h" |
c7f13d42 | 46 | #include "fs.h" |
07e81dc9 | 47 | #include "accessors.h" |
a0231804 | 48 | #include "extent-tree.h" |
45c40c8f | 49 | #include "root-tree.h" |
59b818e0 | 50 | #include "defrag.h" |
c7a03b52 | 51 | #include "uuid-tree.h" |
67707479 | 52 | #include "relocation.h" |
2fc6822c | 53 | #include "scrub.h" |
c03b2207 | 54 | #include "super.h" |
eb60ceac | 55 | |
319e4d06 QW |
56 | #define BTRFS_SUPER_FLAG_SUPP (BTRFS_HEADER_FLAG_WRITTEN |\ |
57 | BTRFS_HEADER_FLAG_RELOC |\ | |
58 | BTRFS_SUPER_FLAG_ERROR |\ | |
59 | BTRFS_SUPER_FLAG_SEEDING |\ | |
e2731e55 AJ |
60 | BTRFS_SUPER_FLAG_METADUMP |\ |
61 | BTRFS_SUPER_FLAG_METADUMP_V2) | |
319e4d06 | 62 | |
143bede5 | 63 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root); |
acce952b | 64 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 65 | struct btrfs_fs_info *fs_info); |
143bede5 | 66 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root); |
2ff7e61e | 67 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 68 | struct extent_io_tree *dirty_pages, |
69 | int mark); | |
2ff7e61e | 70 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
acce952b | 71 | struct extent_io_tree *pinned_extents); |
2ff7e61e JM |
72 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info); |
73 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info); | |
ce9adaa5 | 74 | |
141386e1 JB |
75 | static void btrfs_free_csum_hash(struct btrfs_fs_info *fs_info) |
76 | { | |
77 | if (fs_info->csum_shash) | |
78 | crypto_free_shash(fs_info->csum_shash); | |
79 | } | |
80 | ||
d352ac68 CM |
81 | /* |
82 | * async submit bios are used to offload expensive checksumming | |
83 | * onto the worker threads. They checksum file and metadata bios | |
84 | * just before they are sent down the IO stack. | |
85 | */ | |
44b8bd7e | 86 | struct async_submit_bio { |
da67daab | 87 | struct btrfs_inode *inode; |
44b8bd7e | 88 | struct bio *bio; |
ab2072b2 | 89 | enum btrfs_wq_submit_cmd submit_cmd; |
44b8bd7e | 90 | int mirror_num; |
1941b64b | 91 | |
ab2072b2 | 92 | /* Optional parameter for used by direct io */ |
1941b64b | 93 | u64 dio_file_offset; |
8b712842 | 94 | struct btrfs_work work; |
4e4cbee9 | 95 | blk_status_t status; |
44b8bd7e CM |
96 | }; |
97 | ||
d352ac68 | 98 | /* |
2996e1f8 | 99 | * Compute the csum of a btree block and store the result to provided buffer. |
d352ac68 | 100 | */ |
c67b3892 | 101 | static void csum_tree_block(struct extent_buffer *buf, u8 *result) |
19c00ddc | 102 | { |
d5178578 | 103 | struct btrfs_fs_info *fs_info = buf->fs_info; |
7280305e | 104 | const int num_pages = num_extent_pages(buf); |
a26663e7 | 105 | const int first_page_part = min_t(u32, PAGE_SIZE, fs_info->nodesize); |
d5178578 | 106 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
19c00ddc | 107 | char *kaddr; |
e9be5a30 | 108 | int i; |
d5178578 JT |
109 | |
110 | shash->tfm = fs_info->csum_shash; | |
111 | crypto_shash_init(shash); | |
a26663e7 | 112 | kaddr = page_address(buf->pages[0]) + offset_in_page(buf->start); |
e9be5a30 | 113 | crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE, |
a26663e7 | 114 | first_page_part - BTRFS_CSUM_SIZE); |
19c00ddc | 115 | |
e9be5a30 DS |
116 | for (i = 1; i < num_pages; i++) { |
117 | kaddr = page_address(buf->pages[i]); | |
118 | crypto_shash_update(shash, kaddr, PAGE_SIZE); | |
19c00ddc | 119 | } |
71a63551 | 120 | memset(result, 0, BTRFS_CSUM_SIZE); |
d5178578 | 121 | crypto_shash_final(shash, result); |
19c00ddc CM |
122 | } |
123 | ||
d352ac68 CM |
124 | /* |
125 | * we can't consider a given block up to date unless the transid of the | |
126 | * block matches the transid in the parent node's pointer. This is how we | |
127 | * detect blocks that either didn't get written at all or got written | |
128 | * in the wrong place. | |
129 | */ | |
1259ab75 | 130 | static int verify_parent_transid(struct extent_io_tree *io_tree, |
b9fab919 CM |
131 | struct extent_buffer *eb, u64 parent_transid, |
132 | int atomic) | |
1259ab75 | 133 | { |
2ac55d41 | 134 | struct extent_state *cached_state = NULL; |
1259ab75 CM |
135 | int ret; |
136 | ||
137 | if (!parent_transid || btrfs_header_generation(eb) == parent_transid) | |
138 | return 0; | |
139 | ||
b9fab919 CM |
140 | if (atomic) |
141 | return -EAGAIN; | |
142 | ||
570eb97b | 143 | lock_extent(io_tree, eb->start, eb->start + eb->len - 1, &cached_state); |
0b32f4bb | 144 | if (extent_buffer_uptodate(eb) && |
1259ab75 CM |
145 | btrfs_header_generation(eb) == parent_transid) { |
146 | ret = 0; | |
147 | goto out; | |
148 | } | |
94647322 | 149 | btrfs_err_rl(eb->fs_info, |
8f0ed7d4 QW |
150 | "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", |
151 | eb->start, eb->read_mirror, | |
29549aec | 152 | parent_transid, btrfs_header_generation(eb)); |
1259ab75 | 153 | ret = 1; |
35b22c19 | 154 | clear_extent_buffer_uptodate(eb); |
33958dc6 | 155 | out: |
570eb97b JB |
156 | unlock_extent(io_tree, eb->start, eb->start + eb->len - 1, |
157 | &cached_state); | |
1259ab75 | 158 | return ret; |
1259ab75 CM |
159 | } |
160 | ||
e7e16f48 JT |
161 | static bool btrfs_supported_super_csum(u16 csum_type) |
162 | { | |
163 | switch (csum_type) { | |
164 | case BTRFS_CSUM_TYPE_CRC32: | |
3951e7f0 | 165 | case BTRFS_CSUM_TYPE_XXHASH: |
3831bf00 | 166 | case BTRFS_CSUM_TYPE_SHA256: |
352ae07b | 167 | case BTRFS_CSUM_TYPE_BLAKE2: |
e7e16f48 JT |
168 | return true; |
169 | default: | |
170 | return false; | |
171 | } | |
172 | } | |
173 | ||
1104a885 DS |
174 | /* |
175 | * Return 0 if the superblock checksum type matches the checksum value of that | |
176 | * algorithm. Pass the raw disk superblock data. | |
177 | */ | |
3d17adea QW |
178 | int btrfs_check_super_csum(struct btrfs_fs_info *fs_info, |
179 | const struct btrfs_super_block *disk_sb) | |
1104a885 | 180 | { |
51bce6c9 | 181 | char result[BTRFS_CSUM_SIZE]; |
d5178578 JT |
182 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
183 | ||
184 | shash->tfm = fs_info->csum_shash; | |
1104a885 | 185 | |
51bce6c9 JT |
186 | /* |
187 | * The super_block structure does not span the whole | |
188 | * BTRFS_SUPER_INFO_SIZE range, we expect that the unused space is | |
189 | * filled with zeros and is included in the checksum. | |
190 | */ | |
3d17adea | 191 | crypto_shash_digest(shash, (const u8 *)disk_sb + BTRFS_CSUM_SIZE, |
fd08001f | 192 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, result); |
1104a885 | 193 | |
55fc29be | 194 | if (memcmp(disk_sb->csum, result, fs_info->csum_size)) |
51bce6c9 | 195 | return 1; |
1104a885 | 196 | |
e7e16f48 | 197 | return 0; |
1104a885 DS |
198 | } |
199 | ||
e064d5e9 | 200 | int btrfs_verify_level_key(struct extent_buffer *eb, int level, |
448de471 | 201 | struct btrfs_key *first_key, u64 parent_transid) |
581c1760 | 202 | { |
e064d5e9 | 203 | struct btrfs_fs_info *fs_info = eb->fs_info; |
581c1760 QW |
204 | int found_level; |
205 | struct btrfs_key found_key; | |
206 | int ret; | |
207 | ||
208 | found_level = btrfs_header_level(eb); | |
209 | if (found_level != level) { | |
63489055 QW |
210 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
211 | KERN_ERR "BTRFS: tree level check failed\n"); | |
581c1760 QW |
212 | btrfs_err(fs_info, |
213 | "tree level mismatch detected, bytenr=%llu level expected=%u has=%u", | |
214 | eb->start, level, found_level); | |
581c1760 QW |
215 | return -EIO; |
216 | } | |
217 | ||
218 | if (!first_key) | |
219 | return 0; | |
220 | ||
5d41be6f QW |
221 | /* |
222 | * For live tree block (new tree blocks in current transaction), | |
223 | * we need proper lock context to avoid race, which is impossible here. | |
224 | * So we only checks tree blocks which is read from disk, whose | |
225 | * generation <= fs_info->last_trans_committed. | |
226 | */ | |
227 | if (btrfs_header_generation(eb) > fs_info->last_trans_committed) | |
228 | return 0; | |
62fdaa52 QW |
229 | |
230 | /* We have @first_key, so this @eb must have at least one item */ | |
231 | if (btrfs_header_nritems(eb) == 0) { | |
232 | btrfs_err(fs_info, | |
233 | "invalid tree nritems, bytenr=%llu nritems=0 expect >0", | |
234 | eb->start); | |
235 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
236 | return -EUCLEAN; | |
237 | } | |
238 | ||
581c1760 QW |
239 | if (found_level) |
240 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
241 | else | |
242 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
243 | ret = btrfs_comp_cpu_keys(first_key, &found_key); | |
244 | ||
581c1760 | 245 | if (ret) { |
63489055 QW |
246 | WARN(IS_ENABLED(CONFIG_BTRFS_DEBUG), |
247 | KERN_ERR "BTRFS: tree first key check failed\n"); | |
581c1760 | 248 | btrfs_err(fs_info, |
ff76a864 LB |
249 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", |
250 | eb->start, parent_transid, first_key->objectid, | |
251 | first_key->type, first_key->offset, | |
252 | found_key.objectid, found_key.type, | |
253 | found_key.offset); | |
581c1760 | 254 | } |
581c1760 QW |
255 | return ret; |
256 | } | |
257 | ||
d352ac68 CM |
258 | /* |
259 | * helper to read a given tree block, doing retries as required when | |
260 | * the checksums don't match and we have alternate mirrors to try. | |
581c1760 | 261 | * |
789d6a3a QW |
262 | * @check: expected tree parentness check, see the comments of the |
263 | * structure for details. | |
d352ac68 | 264 | */ |
6a2e9dc4 | 265 | int btrfs_read_extent_buffer(struct extent_buffer *eb, |
789d6a3a | 266 | struct btrfs_tree_parent_check *check) |
f188591e | 267 | { |
5ab12d1f | 268 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ea466794 | 269 | int failed = 0; |
f188591e CM |
270 | int ret; |
271 | int num_copies = 0; | |
272 | int mirror_num = 0; | |
ea466794 | 273 | int failed_mirror = 0; |
f188591e | 274 | |
789d6a3a QW |
275 | ASSERT(check); |
276 | ||
f188591e | 277 | while (1) { |
f8397d69 | 278 | clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
947a6299 QW |
279 | ret = read_extent_buffer_pages(eb, WAIT_COMPLETE, mirror_num, check); |
280 | if (!ret) | |
281 | break; | |
d397712b | 282 | |
0b246afa | 283 | num_copies = btrfs_num_copies(fs_info, |
f188591e | 284 | eb->start, eb->len); |
4235298e | 285 | if (num_copies == 1) |
ea466794 | 286 | break; |
4235298e | 287 | |
5cf1ab56 JB |
288 | if (!failed_mirror) { |
289 | failed = 1; | |
290 | failed_mirror = eb->read_mirror; | |
291 | } | |
292 | ||
f188591e | 293 | mirror_num++; |
ea466794 JB |
294 | if (mirror_num == failed_mirror) |
295 | mirror_num++; | |
296 | ||
4235298e | 297 | if (mirror_num > num_copies) |
ea466794 | 298 | break; |
f188591e | 299 | } |
ea466794 | 300 | |
c0901581 | 301 | if (failed && !ret && failed_mirror) |
20a1fbf9 | 302 | btrfs_repair_eb_io_failure(eb, failed_mirror); |
ea466794 JB |
303 | |
304 | return ret; | |
f188591e | 305 | } |
19c00ddc | 306 | |
eca0f6f6 QW |
307 | static int csum_one_extent_buffer(struct extent_buffer *eb) |
308 | { | |
309 | struct btrfs_fs_info *fs_info = eb->fs_info; | |
310 | u8 result[BTRFS_CSUM_SIZE]; | |
311 | int ret; | |
312 | ||
313 | ASSERT(memcmp_extent_buffer(eb, fs_info->fs_devices->metadata_uuid, | |
314 | offsetof(struct btrfs_header, fsid), | |
315 | BTRFS_FSID_SIZE) == 0); | |
316 | csum_tree_block(eb, result); | |
317 | ||
318 | if (btrfs_header_level(eb)) | |
319 | ret = btrfs_check_node(eb); | |
320 | else | |
321 | ret = btrfs_check_leaf_full(eb); | |
322 | ||
3777369f QW |
323 | if (ret < 0) |
324 | goto error; | |
325 | ||
326 | /* | |
327 | * Also check the generation, the eb reached here must be newer than | |
328 | * last committed. Or something seriously wrong happened. | |
329 | */ | |
330 | if (unlikely(btrfs_header_generation(eb) <= fs_info->last_trans_committed)) { | |
331 | ret = -EUCLEAN; | |
eca0f6f6 | 332 | btrfs_err(fs_info, |
3777369f QW |
333 | "block=%llu bad generation, have %llu expect > %llu", |
334 | eb->start, btrfs_header_generation(eb), | |
335 | fs_info->last_trans_committed); | |
336 | goto error; | |
eca0f6f6 QW |
337 | } |
338 | write_extent_buffer(eb, result, 0, fs_info->csum_size); | |
339 | ||
340 | return 0; | |
3777369f QW |
341 | |
342 | error: | |
343 | btrfs_print_tree(eb, 0); | |
344 | btrfs_err(fs_info, "block=%llu write time tree block corruption detected", | |
345 | eb->start); | |
346 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
347 | return ret; | |
eca0f6f6 QW |
348 | } |
349 | ||
350 | /* Checksum all dirty extent buffers in one bio_vec */ | |
351 | static int csum_dirty_subpage_buffers(struct btrfs_fs_info *fs_info, | |
352 | struct bio_vec *bvec) | |
353 | { | |
354 | struct page *page = bvec->bv_page; | |
355 | u64 bvec_start = page_offset(page) + bvec->bv_offset; | |
356 | u64 cur; | |
357 | int ret = 0; | |
358 | ||
359 | for (cur = bvec_start; cur < bvec_start + bvec->bv_len; | |
360 | cur += fs_info->nodesize) { | |
361 | struct extent_buffer *eb; | |
362 | bool uptodate; | |
363 | ||
364 | eb = find_extent_buffer(fs_info, cur); | |
365 | uptodate = btrfs_subpage_test_uptodate(fs_info, page, cur, | |
366 | fs_info->nodesize); | |
367 | ||
01cd3909 | 368 | /* A dirty eb shouldn't disappear from buffer_radix */ |
eca0f6f6 QW |
369 | if (WARN_ON(!eb)) |
370 | return -EUCLEAN; | |
371 | ||
372 | if (WARN_ON(cur != btrfs_header_bytenr(eb))) { | |
373 | free_extent_buffer(eb); | |
374 | return -EUCLEAN; | |
375 | } | |
376 | if (WARN_ON(!uptodate)) { | |
377 | free_extent_buffer(eb); | |
378 | return -EUCLEAN; | |
379 | } | |
380 | ||
381 | ret = csum_one_extent_buffer(eb); | |
382 | free_extent_buffer(eb); | |
383 | if (ret < 0) | |
384 | return ret; | |
385 | } | |
386 | return ret; | |
387 | } | |
388 | ||
d352ac68 | 389 | /* |
ac303b69 QW |
390 | * Checksum a dirty tree block before IO. This has extra checks to make sure |
391 | * we only fill in the checksum field in the first page of a multi-page block. | |
392 | * For subpage extent buffers we need bvec to also read the offset in the page. | |
d352ac68 | 393 | */ |
ac303b69 | 394 | static int csum_dirty_buffer(struct btrfs_fs_info *fs_info, struct bio_vec *bvec) |
19c00ddc | 395 | { |
ac303b69 | 396 | struct page *page = bvec->bv_page; |
4eee4fa4 | 397 | u64 start = page_offset(page); |
19c00ddc | 398 | u64 found_start; |
19c00ddc | 399 | struct extent_buffer *eb; |
eca0f6f6 | 400 | |
fbca46eb | 401 | if (fs_info->nodesize < PAGE_SIZE) |
eca0f6f6 | 402 | return csum_dirty_subpage_buffers(fs_info, bvec); |
f188591e | 403 | |
4f2de97a JB |
404 | eb = (struct extent_buffer *)page->private; |
405 | if (page != eb->pages[0]) | |
406 | return 0; | |
0f805531 | 407 | |
19c00ddc | 408 | found_start = btrfs_header_bytenr(eb); |
d3575156 NA |
409 | |
410 | if (test_bit(EXTENT_BUFFER_NO_CHECK, &eb->bflags)) { | |
411 | WARN_ON(found_start != 0); | |
412 | return 0; | |
413 | } | |
414 | ||
0f805531 AL |
415 | /* |
416 | * Please do not consolidate these warnings into a single if. | |
417 | * It is useful to know what went wrong. | |
418 | */ | |
419 | if (WARN_ON(found_start != start)) | |
420 | return -EUCLEAN; | |
421 | if (WARN_ON(!PageUptodate(page))) | |
422 | return -EUCLEAN; | |
423 | ||
eca0f6f6 | 424 | return csum_one_extent_buffer(eb); |
19c00ddc CM |
425 | } |
426 | ||
b0c9b3b0 | 427 | static int check_tree_block_fsid(struct extent_buffer *eb) |
2b82032c | 428 | { |
b0c9b3b0 | 429 | struct btrfs_fs_info *fs_info = eb->fs_info; |
944d3f9f | 430 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices, *seed_devs; |
44880fdc | 431 | u8 fsid[BTRFS_FSID_SIZE]; |
944d3f9f | 432 | u8 *metadata_uuid; |
2b82032c | 433 | |
9a8658e3 DS |
434 | read_extent_buffer(eb, fsid, offsetof(struct btrfs_header, fsid), |
435 | BTRFS_FSID_SIZE); | |
944d3f9f NB |
436 | /* |
437 | * Checking the incompat flag is only valid for the current fs. For | |
438 | * seed devices it's forbidden to have their uuid changed so reading | |
439 | * ->fsid in this case is fine | |
440 | */ | |
441 | if (btrfs_fs_incompat(fs_info, METADATA_UUID)) | |
442 | metadata_uuid = fs_devices->metadata_uuid; | |
443 | else | |
444 | metadata_uuid = fs_devices->fsid; | |
445 | ||
446 | if (!memcmp(fsid, metadata_uuid, BTRFS_FSID_SIZE)) | |
447 | return 0; | |
448 | ||
449 | list_for_each_entry(seed_devs, &fs_devices->seed_list, seed_list) | |
450 | if (!memcmp(fsid, seed_devs->fsid, BTRFS_FSID_SIZE)) | |
451 | return 0; | |
452 | ||
453 | return 1; | |
2b82032c YZ |
454 | } |
455 | ||
77bf40a2 | 456 | /* Do basic extent buffer checks at read time */ |
947a6299 QW |
457 | static int validate_extent_buffer(struct extent_buffer *eb, |
458 | struct btrfs_tree_parent_check *check) | |
ce9adaa5 | 459 | { |
77bf40a2 | 460 | struct btrfs_fs_info *fs_info = eb->fs_info; |
ce9adaa5 | 461 | u64 found_start; |
77bf40a2 QW |
462 | const u32 csum_size = fs_info->csum_size; |
463 | u8 found_level; | |
2996e1f8 | 464 | u8 result[BTRFS_CSUM_SIZE]; |
dfd29eed | 465 | const u8 *header_csum; |
77bf40a2 | 466 | int ret = 0; |
ea466794 | 467 | |
947a6299 QW |
468 | ASSERT(check); |
469 | ||
ce9adaa5 | 470 | found_start = btrfs_header_bytenr(eb); |
727011e0 | 471 | if (found_start != eb->start) { |
8f0ed7d4 QW |
472 | btrfs_err_rl(fs_info, |
473 | "bad tree block start, mirror %u want %llu have %llu", | |
474 | eb->read_mirror, eb->start, found_start); | |
f188591e | 475 | ret = -EIO; |
77bf40a2 | 476 | goto out; |
ce9adaa5 | 477 | } |
b0c9b3b0 | 478 | if (check_tree_block_fsid(eb)) { |
8f0ed7d4 QW |
479 | btrfs_err_rl(fs_info, "bad fsid on logical %llu mirror %u", |
480 | eb->start, eb->read_mirror); | |
1259ab75 | 481 | ret = -EIO; |
77bf40a2 | 482 | goto out; |
1259ab75 | 483 | } |
ce9adaa5 | 484 | found_level = btrfs_header_level(eb); |
1c24c3ce | 485 | if (found_level >= BTRFS_MAX_LEVEL) { |
8f0ed7d4 QW |
486 | btrfs_err(fs_info, |
487 | "bad tree block level, mirror %u level %d on logical %llu", | |
488 | eb->read_mirror, btrfs_header_level(eb), eb->start); | |
1c24c3ce | 489 | ret = -EIO; |
77bf40a2 | 490 | goto out; |
1c24c3ce | 491 | } |
ce9adaa5 | 492 | |
c67b3892 | 493 | csum_tree_block(eb, result); |
dfd29eed DS |
494 | header_csum = page_address(eb->pages[0]) + |
495 | get_eb_offset_in_page(eb, offsetof(struct btrfs_header, csum)); | |
a826d6dc | 496 | |
dfd29eed | 497 | if (memcmp(result, header_csum, csum_size) != 0) { |
2996e1f8 | 498 | btrfs_warn_rl(fs_info, |
8f0ed7d4 QW |
499 | "checksum verify failed on logical %llu mirror %u wanted " CSUM_FMT " found " CSUM_FMT " level %d", |
500 | eb->start, eb->read_mirror, | |
dfd29eed | 501 | CSUM_FMT_VALUE(csum_size, header_csum), |
35be8851 JT |
502 | CSUM_FMT_VALUE(csum_size, result), |
503 | btrfs_header_level(eb)); | |
2996e1f8 | 504 | ret = -EUCLEAN; |
77bf40a2 | 505 | goto out; |
2996e1f8 JT |
506 | } |
507 | ||
947a6299 QW |
508 | if (found_level != check->level) { |
509 | ret = -EIO; | |
510 | goto out; | |
511 | } | |
512 | if (unlikely(check->transid && | |
513 | btrfs_header_generation(eb) != check->transid)) { | |
514 | btrfs_err_rl(eb->fs_info, | |
515 | "parent transid verify failed on logical %llu mirror %u wanted %llu found %llu", | |
516 | eb->start, eb->read_mirror, check->transid, | |
517 | btrfs_header_generation(eb)); | |
518 | ret = -EIO; | |
519 | goto out; | |
520 | } | |
521 | if (check->has_first_key) { | |
522 | struct btrfs_key *expect_key = &check->first_key; | |
523 | struct btrfs_key found_key; | |
524 | ||
525 | if (found_level) | |
526 | btrfs_node_key_to_cpu(eb, &found_key, 0); | |
527 | else | |
528 | btrfs_item_key_to_cpu(eb, &found_key, 0); | |
529 | if (unlikely(btrfs_comp_cpu_keys(expect_key, &found_key))) { | |
530 | btrfs_err(fs_info, | |
531 | "tree first key mismatch detected, bytenr=%llu parent_transid=%llu key expected=(%llu,%u,%llu) has=(%llu,%u,%llu)", | |
532 | eb->start, check->transid, | |
533 | expect_key->objectid, | |
534 | expect_key->type, expect_key->offset, | |
535 | found_key.objectid, found_key.type, | |
536 | found_key.offset); | |
537 | ret = -EUCLEAN; | |
538 | goto out; | |
539 | } | |
540 | } | |
541 | if (check->owner_root) { | |
542 | ret = btrfs_check_eb_owner(eb, check->owner_root); | |
543 | if (ret < 0) | |
544 | goto out; | |
545 | } | |
546 | ||
a826d6dc JB |
547 | /* |
548 | * If this is a leaf block and it is corrupt, set the corrupt bit so | |
549 | * that we don't try and read the other copies of this block, just | |
550 | * return -EIO. | |
551 | */ | |
1c4360ee | 552 | if (found_level == 0 && btrfs_check_leaf_full(eb)) { |
a826d6dc JB |
553 | set_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags); |
554 | ret = -EIO; | |
555 | } | |
ce9adaa5 | 556 | |
813fd1dc | 557 | if (found_level > 0 && btrfs_check_node(eb)) |
053ab70f LB |
558 | ret = -EIO; |
559 | ||
0b32f4bb JB |
560 | if (!ret) |
561 | set_extent_buffer_uptodate(eb); | |
75391f0d QW |
562 | else |
563 | btrfs_err(fs_info, | |
8f0ed7d4 QW |
564 | "read time tree block corruption detected on logical %llu mirror %u", |
565 | eb->start, eb->read_mirror); | |
77bf40a2 QW |
566 | out: |
567 | return ret; | |
568 | } | |
569 | ||
371cdc07 | 570 | static int validate_subpage_buffer(struct page *page, u64 start, u64 end, |
947a6299 | 571 | int mirror, struct btrfs_tree_parent_check *check) |
371cdc07 QW |
572 | { |
573 | struct btrfs_fs_info *fs_info = btrfs_sb(page->mapping->host->i_sb); | |
574 | struct extent_buffer *eb; | |
575 | bool reads_done; | |
576 | int ret = 0; | |
577 | ||
947a6299 QW |
578 | ASSERT(check); |
579 | ||
371cdc07 QW |
580 | /* |
581 | * We don't allow bio merge for subpage metadata read, so we should | |
582 | * only get one eb for each endio hook. | |
583 | */ | |
584 | ASSERT(end == start + fs_info->nodesize - 1); | |
585 | ASSERT(PagePrivate(page)); | |
586 | ||
587 | eb = find_extent_buffer(fs_info, start); | |
588 | /* | |
589 | * When we are reading one tree block, eb must have been inserted into | |
590 | * the radix tree. If not, something is wrong. | |
591 | */ | |
592 | ASSERT(eb); | |
593 | ||
594 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
595 | /* Subpage read must finish in page read */ | |
596 | ASSERT(reads_done); | |
597 | ||
598 | eb->read_mirror = mirror; | |
599 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
600 | ret = -EIO; | |
601 | goto err; | |
602 | } | |
947a6299 | 603 | ret = validate_extent_buffer(eb, check); |
371cdc07 QW |
604 | if (ret < 0) |
605 | goto err; | |
606 | ||
371cdc07 QW |
607 | set_extent_buffer_uptodate(eb); |
608 | ||
609 | free_extent_buffer(eb); | |
610 | return ret; | |
611 | err: | |
612 | /* | |
613 | * end_bio_extent_readpage decrements io_pages in case of error, | |
614 | * make sure it has something to decrement. | |
615 | */ | |
616 | atomic_inc(&eb->io_pages); | |
617 | clear_extent_buffer_uptodate(eb); | |
618 | free_extent_buffer(eb); | |
619 | return ret; | |
620 | } | |
621 | ||
c3a3b19b | 622 | int btrfs_validate_metadata_buffer(struct btrfs_bio *bbio, |
77bf40a2 QW |
623 | struct page *page, u64 start, u64 end, |
624 | int mirror) | |
625 | { | |
626 | struct extent_buffer *eb; | |
627 | int ret = 0; | |
628 | int reads_done; | |
629 | ||
630 | ASSERT(page->private); | |
371cdc07 | 631 | |
fbca46eb | 632 | if (btrfs_sb(page->mapping->host->i_sb)->nodesize < PAGE_SIZE) |
947a6299 QW |
633 | return validate_subpage_buffer(page, start, end, mirror, |
634 | &bbio->parent_check); | |
371cdc07 | 635 | |
77bf40a2 QW |
636 | eb = (struct extent_buffer *)page->private; |
637 | ||
638 | /* | |
639 | * The pending IO might have been the only thing that kept this buffer | |
640 | * in memory. Make sure we have a ref for all this other checks | |
641 | */ | |
642 | atomic_inc(&eb->refs); | |
643 | ||
644 | reads_done = atomic_dec_and_test(&eb->io_pages); | |
645 | if (!reads_done) | |
646 | goto err; | |
647 | ||
648 | eb->read_mirror = mirror; | |
649 | if (test_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags)) { | |
650 | ret = -EIO; | |
651 | goto err; | |
652 | } | |
947a6299 | 653 | ret = validate_extent_buffer(eb, &bbio->parent_check); |
ce9adaa5 | 654 | err: |
53b381b3 DW |
655 | if (ret) { |
656 | /* | |
657 | * our io error hook is going to dec the io pages | |
658 | * again, we have to make sure it has something | |
659 | * to decrement | |
660 | */ | |
661 | atomic_inc(&eb->io_pages); | |
0b32f4bb | 662 | clear_extent_buffer_uptodate(eb); |
53b381b3 | 663 | } |
0b32f4bb | 664 | free_extent_buffer(eb); |
77bf40a2 | 665 | |
f188591e | 666 | return ret; |
ce9adaa5 CM |
667 | } |
668 | ||
4a69a410 CM |
669 | static void run_one_async_start(struct btrfs_work *work) |
670 | { | |
4a69a410 | 671 | struct async_submit_bio *async; |
4e4cbee9 | 672 | blk_status_t ret; |
4a69a410 CM |
673 | |
674 | async = container_of(work, struct async_submit_bio, work); | |
ab2072b2 DS |
675 | switch (async->submit_cmd) { |
676 | case WQ_SUBMIT_METADATA: | |
ad65ecf3 | 677 | ret = btree_submit_bio_start(async->bio); |
ab2072b2 DS |
678 | break; |
679 | case WQ_SUBMIT_DATA: | |
882681ac | 680 | ret = btrfs_submit_bio_start(async->inode, async->bio); |
ab2072b2 DS |
681 | break; |
682 | case WQ_SUBMIT_DATA_DIO: | |
bfa17066 | 683 | ret = btrfs_submit_bio_start_direct_io(async->inode, |
da67daab | 684 | async->bio, async->dio_file_offset); |
ab2072b2 DS |
685 | break; |
686 | } | |
79787eaa | 687 | if (ret) |
4e4cbee9 | 688 | async->status = ret; |
4a69a410 CM |
689 | } |
690 | ||
06ea01b1 DS |
691 | /* |
692 | * In order to insert checksums into the metadata in large chunks, we wait | |
693 | * until bio submission time. All the pages in the bio are checksummed and | |
694 | * sums are attached onto the ordered extent record. | |
695 | * | |
696 | * At IO completion time the csums attached on the ordered extent record are | |
697 | * inserted into the tree. | |
698 | */ | |
4a69a410 | 699 | static void run_one_async_done(struct btrfs_work *work) |
8b712842 | 700 | { |
917f32a2 CH |
701 | struct async_submit_bio *async = |
702 | container_of(work, struct async_submit_bio, work); | |
da67daab | 703 | struct btrfs_inode *inode = async->inode; |
917f32a2 | 704 | struct btrfs_bio *bbio = btrfs_bio(async->bio); |
4854ddd0 | 705 | |
bb7ab3b9 | 706 | /* If an error occurred we just want to clean up the bio and move on */ |
4e4cbee9 | 707 | if (async->status) { |
917f32a2 | 708 | btrfs_bio_end_io(bbio, async->status); |
79787eaa JM |
709 | return; |
710 | } | |
711 | ||
ec39f769 CM |
712 | /* |
713 | * All of the bios that pass through here are from async helpers. | |
714 | * Use REQ_CGROUP_PUNT to issue them from the owning cgroup's context. | |
715 | * This changes nothing when cgroups aren't in use. | |
716 | */ | |
717 | async->bio->bi_opf |= REQ_CGROUP_PUNT; | |
da67daab | 718 | btrfs_submit_bio(inode->root->fs_info, async->bio, async->mirror_num); |
4a69a410 CM |
719 | } |
720 | ||
721 | static void run_one_async_free(struct btrfs_work *work) | |
722 | { | |
723 | struct async_submit_bio *async; | |
724 | ||
725 | async = container_of(work, struct async_submit_bio, work); | |
8b712842 CM |
726 | kfree(async); |
727 | } | |
728 | ||
ea1f0ced CH |
729 | /* |
730 | * Submit bio to an async queue. | |
731 | * | |
732 | * Retrun: | |
733 | * - true if the work has been succesfuly submitted | |
734 | * - false in case of error | |
735 | */ | |
5fcdadc2 | 736 | bool btrfs_wq_submit_bio(struct btrfs_inode *inode, struct bio *bio, int mirror_num, |
ab2072b2 | 737 | u64 dio_file_offset, enum btrfs_wq_submit_cmd cmd) |
44b8bd7e | 738 | { |
5fcdadc2 | 739 | struct btrfs_fs_info *fs_info = inode->root->fs_info; |
44b8bd7e CM |
740 | struct async_submit_bio *async; |
741 | ||
742 | async = kmalloc(sizeof(*async), GFP_NOFS); | |
743 | if (!async) | |
ea1f0ced | 744 | return false; |
44b8bd7e | 745 | |
5fcdadc2 | 746 | async->inode = inode; |
44b8bd7e CM |
747 | async->bio = bio; |
748 | async->mirror_num = mirror_num; | |
ab2072b2 | 749 | async->submit_cmd = cmd; |
4a69a410 | 750 | |
a0cac0ec OS |
751 | btrfs_init_work(&async->work, run_one_async_start, run_one_async_done, |
752 | run_one_async_free); | |
4a69a410 | 753 | |
1941b64b | 754 | async->dio_file_offset = dio_file_offset; |
8c8bee1d | 755 | |
4e4cbee9 | 756 | async->status = 0; |
79787eaa | 757 | |
67f055c7 | 758 | if (op_is_sync(bio->bi_opf)) |
a31b4a43 CH |
759 | btrfs_queue_work(fs_info->hipri_workers, &async->work); |
760 | else | |
761 | btrfs_queue_work(fs_info->workers, &async->work); | |
ea1f0ced | 762 | return true; |
44b8bd7e CM |
763 | } |
764 | ||
4e4cbee9 | 765 | static blk_status_t btree_csum_one_bio(struct bio *bio) |
ce3ed71a | 766 | { |
2c30c71b | 767 | struct bio_vec *bvec; |
ce3ed71a | 768 | struct btrfs_root *root; |
2b070cfe | 769 | int ret = 0; |
6dc4f100 | 770 | struct bvec_iter_all iter_all; |
ce3ed71a | 771 | |
c09abff8 | 772 | ASSERT(!bio_flagged(bio, BIO_CLONED)); |
2b070cfe | 773 | bio_for_each_segment_all(bvec, bio, iter_all) { |
ce3ed71a | 774 | root = BTRFS_I(bvec->bv_page->mapping->host)->root; |
ac303b69 | 775 | ret = csum_dirty_buffer(root->fs_info, bvec); |
79787eaa JM |
776 | if (ret) |
777 | break; | |
ce3ed71a | 778 | } |
2c30c71b | 779 | |
4e4cbee9 | 780 | return errno_to_blk_status(ret); |
ce3ed71a CM |
781 | } |
782 | ||
ad65ecf3 | 783 | blk_status_t btree_submit_bio_start(struct bio *bio) |
22c59948 | 784 | { |
8b712842 CM |
785 | /* |
786 | * when we're called for a write, we're already in the async | |
1a722d8f | 787 | * submission context. Just jump into btrfs_submit_bio. |
8b712842 | 788 | */ |
79787eaa | 789 | return btree_csum_one_bio(bio); |
4a69a410 | 790 | } |
22c59948 | 791 | |
f4dcfb30 | 792 | static bool should_async_write(struct btrfs_fs_info *fs_info, |
9b4e675a | 793 | struct btrfs_inode *bi) |
de0022b9 | 794 | { |
4eef29ef | 795 | if (btrfs_is_zoned(fs_info)) |
f4dcfb30 | 796 | return false; |
6300463b | 797 | if (atomic_read(&bi->sync_writers)) |
f4dcfb30 | 798 | return false; |
9b4e675a | 799 | if (test_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags)) |
f4dcfb30 JT |
800 | return false; |
801 | return true; | |
de0022b9 JB |
802 | } |
803 | ||
644094fd | 804 | void btrfs_submit_metadata_bio(struct btrfs_inode *inode, struct bio *bio, int mirror_num) |
44b8bd7e | 805 | { |
644094fd | 806 | struct btrfs_fs_info *fs_info = inode->root->fs_info; |
917f32a2 | 807 | struct btrfs_bio *bbio = btrfs_bio(bio); |
4e4cbee9 | 808 | blk_status_t ret; |
cad321ad | 809 | |
08a6f464 | 810 | bio->bi_opf |= REQ_META; |
947a6299 | 811 | bbio->is_metadata = 1; |
08a6f464 | 812 | |
cfe94440 | 813 | if (btrfs_op(bio) != BTRFS_MAP_WRITE) { |
1a722d8f CH |
814 | btrfs_submit_bio(fs_info, bio, mirror_num); |
815 | return; | |
44b8bd7e | 816 | } |
d313d7a3 | 817 | |
ea1f0ced CH |
818 | /* |
819 | * Kthread helpers are used to submit writes so that checksumming can | |
820 | * happen in parallel across all CPUs. | |
821 | */ | |
644094fd DS |
822 | if (should_async_write(fs_info, inode) && |
823 | btrfs_wq_submit_bio(inode, bio, mirror_num, 0, WQ_SUBMIT_METADATA)) | |
ea1f0ced CH |
824 | return; |
825 | ||
826 | ret = btree_csum_one_bio(bio); | |
94d9e11b | 827 | if (ret) { |
917f32a2 | 828 | btrfs_bio_end_io(bbio, ret); |
ea1f0ced | 829 | return; |
94d9e11b | 830 | } |
ea1f0ced CH |
831 | |
832 | btrfs_submit_bio(fs_info, bio, mirror_num); | |
44b8bd7e CM |
833 | } |
834 | ||
3dd1462e | 835 | #ifdef CONFIG_MIGRATION |
8958b551 MWO |
836 | static int btree_migrate_folio(struct address_space *mapping, |
837 | struct folio *dst, struct folio *src, enum migrate_mode mode) | |
784b4e29 CM |
838 | { |
839 | /* | |
840 | * we can't safely write a btree page from here, | |
841 | * we haven't done the locking hook | |
842 | */ | |
8958b551 | 843 | if (folio_test_dirty(src)) |
784b4e29 CM |
844 | return -EAGAIN; |
845 | /* | |
846 | * Buffers may be managed in a filesystem specific way. | |
847 | * We must have no buffers or drop them. | |
848 | */ | |
8958b551 MWO |
849 | if (folio_get_private(src) && |
850 | !filemap_release_folio(src, GFP_KERNEL)) | |
784b4e29 | 851 | return -EAGAIN; |
54184650 | 852 | return migrate_folio(mapping, dst, src, mode); |
784b4e29 | 853 | } |
8958b551 MWO |
854 | #else |
855 | #define btree_migrate_folio NULL | |
3dd1462e | 856 | #endif |
784b4e29 | 857 | |
0da5468f CM |
858 | static int btree_writepages(struct address_space *mapping, |
859 | struct writeback_control *wbc) | |
860 | { | |
e2d84521 MX |
861 | struct btrfs_fs_info *fs_info; |
862 | int ret; | |
863 | ||
d8d5f3e1 | 864 | if (wbc->sync_mode == WB_SYNC_NONE) { |
448d640b CM |
865 | |
866 | if (wbc->for_kupdate) | |
867 | return 0; | |
868 | ||
e2d84521 | 869 | fs_info = BTRFS_I(mapping->host)->root->fs_info; |
b9473439 | 870 | /* this is a bit racy, but that's ok */ |
d814a491 EL |
871 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
872 | BTRFS_DIRTY_METADATA_THRESH, | |
873 | fs_info->dirty_metadata_batch); | |
e2d84521 | 874 | if (ret < 0) |
793955bc | 875 | return 0; |
793955bc | 876 | } |
0b32f4bb | 877 | return btree_write_cache_pages(mapping, wbc); |
0da5468f CM |
878 | } |
879 | ||
f913cff3 | 880 | static bool btree_release_folio(struct folio *folio, gfp_t gfp_flags) |
5f39d397 | 881 | { |
f913cff3 MWO |
882 | if (folio_test_writeback(folio) || folio_test_dirty(folio)) |
883 | return false; | |
0c4e538b | 884 | |
f913cff3 | 885 | return try_release_extent_buffer(&folio->page); |
d98237b3 CM |
886 | } |
887 | ||
895586eb MWO |
888 | static void btree_invalidate_folio(struct folio *folio, size_t offset, |
889 | size_t length) | |
d98237b3 | 890 | { |
d1310b2e | 891 | struct extent_io_tree *tree; |
895586eb MWO |
892 | tree = &BTRFS_I(folio->mapping->host)->io_tree; |
893 | extent_invalidate_folio(tree, folio, offset); | |
f913cff3 | 894 | btree_release_folio(folio, GFP_NOFS); |
895586eb MWO |
895 | if (folio_get_private(folio)) { |
896 | btrfs_warn(BTRFS_I(folio->mapping->host)->root->fs_info, | |
897 | "folio private not zero on folio %llu", | |
898 | (unsigned long long)folio_pos(folio)); | |
899 | folio_detach_private(folio); | |
9ad6b7bc | 900 | } |
d98237b3 CM |
901 | } |
902 | ||
bb146eb2 | 903 | #ifdef DEBUG |
0079c3b1 MWO |
904 | static bool btree_dirty_folio(struct address_space *mapping, |
905 | struct folio *folio) | |
906 | { | |
907 | struct btrfs_fs_info *fs_info = btrfs_sb(mapping->host->i_sb); | |
139e8cd3 | 908 | struct btrfs_subpage *subpage; |
0b32f4bb | 909 | struct extent_buffer *eb; |
139e8cd3 | 910 | int cur_bit = 0; |
0079c3b1 | 911 | u64 page_start = folio_pos(folio); |
139e8cd3 QW |
912 | |
913 | if (fs_info->sectorsize == PAGE_SIZE) { | |
0079c3b1 | 914 | eb = folio_get_private(folio); |
139e8cd3 QW |
915 | BUG_ON(!eb); |
916 | BUG_ON(!test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
917 | BUG_ON(!atomic_read(&eb->refs)); | |
49d0c642 | 918 | btrfs_assert_tree_write_locked(eb); |
0079c3b1 | 919 | return filemap_dirty_folio(mapping, folio); |
139e8cd3 | 920 | } |
0079c3b1 | 921 | subpage = folio_get_private(folio); |
139e8cd3 QW |
922 | |
923 | ASSERT(subpage->dirty_bitmap); | |
924 | while (cur_bit < BTRFS_SUBPAGE_BITMAP_SIZE) { | |
925 | unsigned long flags; | |
926 | u64 cur; | |
927 | u16 tmp = (1 << cur_bit); | |
928 | ||
929 | spin_lock_irqsave(&subpage->lock, flags); | |
930 | if (!(tmp & subpage->dirty_bitmap)) { | |
931 | spin_unlock_irqrestore(&subpage->lock, flags); | |
932 | cur_bit++; | |
933 | continue; | |
934 | } | |
935 | spin_unlock_irqrestore(&subpage->lock, flags); | |
936 | cur = page_start + cur_bit * fs_info->sectorsize; | |
0b32f4bb | 937 | |
139e8cd3 QW |
938 | eb = find_extent_buffer(fs_info, cur); |
939 | ASSERT(eb); | |
940 | ASSERT(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)); | |
941 | ASSERT(atomic_read(&eb->refs)); | |
49d0c642 | 942 | btrfs_assert_tree_write_locked(eb); |
139e8cd3 QW |
943 | free_extent_buffer(eb); |
944 | ||
945 | cur_bit += (fs_info->nodesize >> fs_info->sectorsize_bits); | |
946 | } | |
0079c3b1 | 947 | return filemap_dirty_folio(mapping, folio); |
0b32f4bb | 948 | } |
0079c3b1 MWO |
949 | #else |
950 | #define btree_dirty_folio filemap_dirty_folio | |
951 | #endif | |
0b32f4bb | 952 | |
7f09410b | 953 | static const struct address_space_operations btree_aops = { |
0da5468f | 954 | .writepages = btree_writepages, |
f913cff3 | 955 | .release_folio = btree_release_folio, |
895586eb | 956 | .invalidate_folio = btree_invalidate_folio, |
8958b551 MWO |
957 | .migrate_folio = btree_migrate_folio, |
958 | .dirty_folio = btree_dirty_folio, | |
d98237b3 CM |
959 | }; |
960 | ||
2ff7e61e JM |
961 | struct extent_buffer *btrfs_find_create_tree_block( |
962 | struct btrfs_fs_info *fs_info, | |
3fbaf258 JB |
963 | u64 bytenr, u64 owner_root, |
964 | int level) | |
0999df54 | 965 | { |
0b246afa JM |
966 | if (btrfs_is_testing(fs_info)) |
967 | return alloc_test_extent_buffer(fs_info, bytenr); | |
3fbaf258 | 968 | return alloc_extent_buffer(fs_info, bytenr, owner_root, level); |
0999df54 CM |
969 | } |
970 | ||
581c1760 QW |
971 | /* |
972 | * Read tree block at logical address @bytenr and do variant basic but critical | |
973 | * verification. | |
974 | * | |
789d6a3a QW |
975 | * @check: expected tree parentness check, see comments of the |
976 | * structure for details. | |
581c1760 | 977 | */ |
2ff7e61e | 978 | struct extent_buffer *read_tree_block(struct btrfs_fs_info *fs_info, u64 bytenr, |
789d6a3a | 979 | struct btrfs_tree_parent_check *check) |
0999df54 CM |
980 | { |
981 | struct extent_buffer *buf = NULL; | |
0999df54 CM |
982 | int ret; |
983 | ||
789d6a3a QW |
984 | ASSERT(check); |
985 | ||
986 | buf = btrfs_find_create_tree_block(fs_info, bytenr, check->owner_root, | |
987 | check->level); | |
c871b0f2 LB |
988 | if (IS_ERR(buf)) |
989 | return buf; | |
0999df54 | 990 | |
789d6a3a | 991 | ret = btrfs_read_extent_buffer(buf, check); |
0f0fe8f7 | 992 | if (ret) { |
537f38f0 | 993 | free_extent_buffer_stale(buf); |
64c043de | 994 | return ERR_PTR(ret); |
0f0fe8f7 | 995 | } |
789d6a3a | 996 | if (btrfs_check_eb_owner(buf, check->owner_root)) { |
88c602ab QW |
997 | free_extent_buffer_stale(buf); |
998 | return ERR_PTR(-EUCLEAN); | |
999 | } | |
5f39d397 | 1000 | return buf; |
ce9adaa5 | 1001 | |
eb60ceac CM |
1002 | } |
1003 | ||
6a884d7d | 1004 | void btrfs_clean_tree_block(struct extent_buffer *buf) |
ed2ff2cb | 1005 | { |
6a884d7d | 1006 | struct btrfs_fs_info *fs_info = buf->fs_info; |
55c69072 | 1007 | if (btrfs_header_generation(buf) == |
e2d84521 | 1008 | fs_info->running_transaction->transid) { |
49d0c642 | 1009 | btrfs_assert_tree_write_locked(buf); |
b4ce94de | 1010 | |
b9473439 | 1011 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)) { |
104b4e51 NB |
1012 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
1013 | -buf->len, | |
1014 | fs_info->dirty_metadata_batch); | |
ed7b63eb JB |
1015 | clear_extent_buffer_dirty(buf); |
1016 | } | |
925baedd | 1017 | } |
5f39d397 CM |
1018 | } |
1019 | ||
da17066c | 1020 | static void __setup_root(struct btrfs_root *root, struct btrfs_fs_info *fs_info, |
143bede5 | 1021 | u64 objectid) |
d97e63b6 | 1022 | { |
7c0260ee | 1023 | bool dummy = test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state); |
2e608bd1 JB |
1024 | |
1025 | memset(&root->root_key, 0, sizeof(root->root_key)); | |
1026 | memset(&root->root_item, 0, sizeof(root->root_item)); | |
1027 | memset(&root->defrag_progress, 0, sizeof(root->defrag_progress)); | |
96dfcb46 | 1028 | root->fs_info = fs_info; |
2e608bd1 | 1029 | root->root_key.objectid = objectid; |
cfaa7295 | 1030 | root->node = NULL; |
a28ec197 | 1031 | root->commit_root = NULL; |
27cdeb70 | 1032 | root->state = 0; |
abed4aaa | 1033 | RB_CLEAR_NODE(&root->rb_node); |
0b86a832 | 1034 | |
0f7d52f4 | 1035 | root->last_trans = 0; |
6b8fad57 | 1036 | root->free_objectid = 0; |
eb73c1b7 | 1037 | root->nr_delalloc_inodes = 0; |
199c2a9c | 1038 | root->nr_ordered_extents = 0; |
6bef4d31 | 1039 | root->inode_tree = RB_ROOT; |
088aea3b | 1040 | INIT_RADIX_TREE(&root->delayed_nodes_tree, GFP_ATOMIC); |
2e608bd1 JB |
1041 | |
1042 | btrfs_init_root_block_rsv(root); | |
0b86a832 CM |
1043 | |
1044 | INIT_LIST_HEAD(&root->dirty_list); | |
5d4f98a2 | 1045 | INIT_LIST_HEAD(&root->root_list); |
eb73c1b7 MX |
1046 | INIT_LIST_HEAD(&root->delalloc_inodes); |
1047 | INIT_LIST_HEAD(&root->delalloc_root); | |
199c2a9c MX |
1048 | INIT_LIST_HEAD(&root->ordered_extents); |
1049 | INIT_LIST_HEAD(&root->ordered_root); | |
d2311e69 | 1050 | INIT_LIST_HEAD(&root->reloc_dirty_list); |
2ab28f32 JB |
1051 | INIT_LIST_HEAD(&root->logged_list[0]); |
1052 | INIT_LIST_HEAD(&root->logged_list[1]); | |
5d4f98a2 | 1053 | spin_lock_init(&root->inode_lock); |
eb73c1b7 | 1054 | spin_lock_init(&root->delalloc_lock); |
199c2a9c | 1055 | spin_lock_init(&root->ordered_extent_lock); |
f0486c68 | 1056 | spin_lock_init(&root->accounting_lock); |
2ab28f32 JB |
1057 | spin_lock_init(&root->log_extents_lock[0]); |
1058 | spin_lock_init(&root->log_extents_lock[1]); | |
8287475a | 1059 | spin_lock_init(&root->qgroup_meta_rsv_lock); |
a2135011 | 1060 | mutex_init(&root->objectid_mutex); |
e02119d5 | 1061 | mutex_init(&root->log_mutex); |
31f3d255 | 1062 | mutex_init(&root->ordered_extent_mutex); |
573bfb72 | 1063 | mutex_init(&root->delalloc_mutex); |
c53e9653 | 1064 | init_waitqueue_head(&root->qgroup_flush_wait); |
7237f183 YZ |
1065 | init_waitqueue_head(&root->log_writer_wait); |
1066 | init_waitqueue_head(&root->log_commit_wait[0]); | |
1067 | init_waitqueue_head(&root->log_commit_wait[1]); | |
8b050d35 MX |
1068 | INIT_LIST_HEAD(&root->log_ctxs[0]); |
1069 | INIT_LIST_HEAD(&root->log_ctxs[1]); | |
7237f183 YZ |
1070 | atomic_set(&root->log_commit[0], 0); |
1071 | atomic_set(&root->log_commit[1], 0); | |
1072 | atomic_set(&root->log_writers, 0); | |
2ecb7923 | 1073 | atomic_set(&root->log_batch, 0); |
0700cea7 | 1074 | refcount_set(&root->refs, 1); |
8ecebf4d | 1075 | atomic_set(&root->snapshot_force_cow, 0); |
eede2bf3 | 1076 | atomic_set(&root->nr_swapfiles, 0); |
7237f183 | 1077 | root->log_transid = 0; |
d1433deb | 1078 | root->log_transid_committed = -1; |
257c62e1 | 1079 | root->last_log_commit = 0; |
2e608bd1 | 1080 | root->anon_dev = 0; |
e289f03e | 1081 | if (!dummy) { |
43eb5f29 | 1082 | extent_io_tree_init(fs_info, &root->dirty_log_pages, |
35da5a7e | 1083 | IO_TREE_ROOT_DIRTY_LOG_PAGES); |
e289f03e | 1084 | extent_io_tree_init(fs_info, &root->log_csum_range, |
35da5a7e | 1085 | IO_TREE_LOG_CSUM_RANGE); |
e289f03e | 1086 | } |
017e5369 | 1087 | |
5f3ab90a | 1088 | spin_lock_init(&root->root_item_lock); |
370a11b8 | 1089 | btrfs_qgroup_init_swapped_blocks(&root->swapped_blocks); |
bd647ce3 JB |
1090 | #ifdef CONFIG_BTRFS_DEBUG |
1091 | INIT_LIST_HEAD(&root->leak_list); | |
fc7cbcd4 | 1092 | spin_lock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 1093 | list_add_tail(&root->leak_list, &fs_info->allocated_roots); |
fc7cbcd4 | 1094 | spin_unlock(&fs_info->fs_roots_radix_lock); |
bd647ce3 | 1095 | #endif |
3768f368 CM |
1096 | } |
1097 | ||
74e4d827 | 1098 | static struct btrfs_root *btrfs_alloc_root(struct btrfs_fs_info *fs_info, |
96dfcb46 | 1099 | u64 objectid, gfp_t flags) |
6f07e42e | 1100 | { |
74e4d827 | 1101 | struct btrfs_root *root = kzalloc(sizeof(*root), flags); |
6f07e42e | 1102 | if (root) |
96dfcb46 | 1103 | __setup_root(root, fs_info, objectid); |
6f07e42e AV |
1104 | return root; |
1105 | } | |
1106 | ||
06ea65a3 JB |
1107 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
1108 | /* Should only be used by the testing infrastructure */ | |
da17066c | 1109 | struct btrfs_root *btrfs_alloc_dummy_root(struct btrfs_fs_info *fs_info) |
06ea65a3 JB |
1110 | { |
1111 | struct btrfs_root *root; | |
1112 | ||
7c0260ee JM |
1113 | if (!fs_info) |
1114 | return ERR_PTR(-EINVAL); | |
1115 | ||
96dfcb46 | 1116 | root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, GFP_KERNEL); |
06ea65a3 JB |
1117 | if (!root) |
1118 | return ERR_PTR(-ENOMEM); | |
da17066c | 1119 | |
b9ef22de | 1120 | /* We don't use the stripesize in selftest, set it as sectorsize */ |
faa2dbf0 | 1121 | root->alloc_bytenr = 0; |
06ea65a3 JB |
1122 | |
1123 | return root; | |
1124 | } | |
1125 | #endif | |
1126 | ||
abed4aaa JB |
1127 | static int global_root_cmp(struct rb_node *a_node, const struct rb_node *b_node) |
1128 | { | |
1129 | const struct btrfs_root *a = rb_entry(a_node, struct btrfs_root, rb_node); | |
1130 | const struct btrfs_root *b = rb_entry(b_node, struct btrfs_root, rb_node); | |
1131 | ||
1132 | return btrfs_comp_cpu_keys(&a->root_key, &b->root_key); | |
1133 | } | |
1134 | ||
1135 | static int global_root_key_cmp(const void *k, const struct rb_node *node) | |
1136 | { | |
1137 | const struct btrfs_key *key = k; | |
1138 | const struct btrfs_root *root = rb_entry(node, struct btrfs_root, rb_node); | |
1139 | ||
1140 | return btrfs_comp_cpu_keys(key, &root->root_key); | |
1141 | } | |
1142 | ||
1143 | int btrfs_global_root_insert(struct btrfs_root *root) | |
1144 | { | |
1145 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1146 | struct rb_node *tmp; | |
1147 | ||
1148 | write_lock(&fs_info->global_root_lock); | |
1149 | tmp = rb_find_add(&root->rb_node, &fs_info->global_root_tree, global_root_cmp); | |
1150 | write_unlock(&fs_info->global_root_lock); | |
1151 | ASSERT(!tmp); | |
1152 | ||
1153 | return tmp ? -EEXIST : 0; | |
1154 | } | |
1155 | ||
1156 | void btrfs_global_root_delete(struct btrfs_root *root) | |
1157 | { | |
1158 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1159 | ||
1160 | write_lock(&fs_info->global_root_lock); | |
1161 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1162 | write_unlock(&fs_info->global_root_lock); | |
1163 | } | |
1164 | ||
1165 | struct btrfs_root *btrfs_global_root(struct btrfs_fs_info *fs_info, | |
1166 | struct btrfs_key *key) | |
1167 | { | |
1168 | struct rb_node *node; | |
1169 | struct btrfs_root *root = NULL; | |
1170 | ||
1171 | read_lock(&fs_info->global_root_lock); | |
1172 | node = rb_find(key, &fs_info->global_root_tree, global_root_key_cmp); | |
1173 | if (node) | |
1174 | root = container_of(node, struct btrfs_root, rb_node); | |
1175 | read_unlock(&fs_info->global_root_lock); | |
1176 | ||
1177 | return root; | |
1178 | } | |
1179 | ||
f7238e50 JB |
1180 | static u64 btrfs_global_root_id(struct btrfs_fs_info *fs_info, u64 bytenr) |
1181 | { | |
1182 | struct btrfs_block_group *block_group; | |
1183 | u64 ret; | |
1184 | ||
1185 | if (!btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) | |
1186 | return 0; | |
1187 | ||
1188 | if (bytenr) | |
1189 | block_group = btrfs_lookup_block_group(fs_info, bytenr); | |
1190 | else | |
1191 | block_group = btrfs_lookup_first_block_group(fs_info, bytenr); | |
1192 | ASSERT(block_group); | |
1193 | if (!block_group) | |
1194 | return 0; | |
1195 | ret = block_group->global_root_id; | |
1196 | btrfs_put_block_group(block_group); | |
1197 | ||
1198 | return ret; | |
1199 | } | |
1200 | ||
abed4aaa JB |
1201 | struct btrfs_root *btrfs_csum_root(struct btrfs_fs_info *fs_info, u64 bytenr) |
1202 | { | |
1203 | struct btrfs_key key = { | |
1204 | .objectid = BTRFS_CSUM_TREE_OBJECTID, | |
1205 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 1206 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
1207 | }; |
1208 | ||
1209 | return btrfs_global_root(fs_info, &key); | |
1210 | } | |
1211 | ||
1212 | struct btrfs_root *btrfs_extent_root(struct btrfs_fs_info *fs_info, u64 bytenr) | |
1213 | { | |
1214 | struct btrfs_key key = { | |
1215 | .objectid = BTRFS_EXTENT_TREE_OBJECTID, | |
1216 | .type = BTRFS_ROOT_ITEM_KEY, | |
f7238e50 | 1217 | .offset = btrfs_global_root_id(fs_info, bytenr), |
abed4aaa JB |
1218 | }; |
1219 | ||
1220 | return btrfs_global_root(fs_info, &key); | |
1221 | } | |
1222 | ||
51129b33 JB |
1223 | struct btrfs_root *btrfs_block_group_root(struct btrfs_fs_info *fs_info) |
1224 | { | |
1225 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) | |
1226 | return fs_info->block_group_root; | |
1227 | return btrfs_extent_root(fs_info, 0); | |
1228 | } | |
1229 | ||
20897f5c | 1230 | struct btrfs_root *btrfs_create_tree(struct btrfs_trans_handle *trans, |
20897f5c AJ |
1231 | u64 objectid) |
1232 | { | |
9b7a2440 | 1233 | struct btrfs_fs_info *fs_info = trans->fs_info; |
20897f5c AJ |
1234 | struct extent_buffer *leaf; |
1235 | struct btrfs_root *tree_root = fs_info->tree_root; | |
1236 | struct btrfs_root *root; | |
1237 | struct btrfs_key key; | |
b89f6d1f | 1238 | unsigned int nofs_flag; |
20897f5c | 1239 | int ret = 0; |
20897f5c | 1240 | |
b89f6d1f FM |
1241 | /* |
1242 | * We're holding a transaction handle, so use a NOFS memory allocation | |
1243 | * context to avoid deadlock if reclaim happens. | |
1244 | */ | |
1245 | nofs_flag = memalloc_nofs_save(); | |
96dfcb46 | 1246 | root = btrfs_alloc_root(fs_info, objectid, GFP_KERNEL); |
b89f6d1f | 1247 | memalloc_nofs_restore(nofs_flag); |
20897f5c AJ |
1248 | if (!root) |
1249 | return ERR_PTR(-ENOMEM); | |
1250 | ||
20897f5c AJ |
1251 | root->root_key.objectid = objectid; |
1252 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1253 | root->root_key.offset = 0; | |
1254 | ||
9631e4cc JB |
1255 | leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0, |
1256 | BTRFS_NESTING_NORMAL); | |
20897f5c AJ |
1257 | if (IS_ERR(leaf)) { |
1258 | ret = PTR_ERR(leaf); | |
1dd05682 | 1259 | leaf = NULL; |
c1b07854 | 1260 | goto fail; |
20897f5c AJ |
1261 | } |
1262 | ||
20897f5c | 1263 | root->node = leaf; |
20897f5c AJ |
1264 | btrfs_mark_buffer_dirty(leaf); |
1265 | ||
1266 | root->commit_root = btrfs_root_node(root); | |
27cdeb70 | 1267 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
20897f5c | 1268 | |
f944d2cb DS |
1269 | btrfs_set_root_flags(&root->root_item, 0); |
1270 | btrfs_set_root_limit(&root->root_item, 0); | |
20897f5c AJ |
1271 | btrfs_set_root_bytenr(&root->root_item, leaf->start); |
1272 | btrfs_set_root_generation(&root->root_item, trans->transid); | |
1273 | btrfs_set_root_level(&root->root_item, 0); | |
1274 | btrfs_set_root_refs(&root->root_item, 1); | |
1275 | btrfs_set_root_used(&root->root_item, leaf->len); | |
1276 | btrfs_set_root_last_snapshot(&root->root_item, 0); | |
1277 | btrfs_set_root_dirid(&root->root_item, 0); | |
33d85fda | 1278 | if (is_fstree(objectid)) |
807fc790 AS |
1279 | generate_random_guid(root->root_item.uuid); |
1280 | else | |
1281 | export_guid(root->root_item.uuid, &guid_null); | |
c8422684 | 1282 | btrfs_set_root_drop_level(&root->root_item, 0); |
20897f5c | 1283 | |
8a6a87cd BB |
1284 | btrfs_tree_unlock(leaf); |
1285 | ||
20897f5c AJ |
1286 | key.objectid = objectid; |
1287 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1288 | key.offset = 0; | |
1289 | ret = btrfs_insert_root(trans, tree_root, &key, &root->root_item); | |
1290 | if (ret) | |
1291 | goto fail; | |
1292 | ||
1dd05682 TI |
1293 | return root; |
1294 | ||
8a6a87cd | 1295 | fail: |
00246528 | 1296 | btrfs_put_root(root); |
20897f5c | 1297 | |
1dd05682 | 1298 | return ERR_PTR(ret); |
20897f5c AJ |
1299 | } |
1300 | ||
7237f183 YZ |
1301 | static struct btrfs_root *alloc_log_tree(struct btrfs_trans_handle *trans, |
1302 | struct btrfs_fs_info *fs_info) | |
0f7d52f4 CM |
1303 | { |
1304 | struct btrfs_root *root; | |
e02119d5 | 1305 | |
96dfcb46 | 1306 | root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, GFP_NOFS); |
e02119d5 | 1307 | if (!root) |
7237f183 | 1308 | return ERR_PTR(-ENOMEM); |
e02119d5 | 1309 | |
e02119d5 CM |
1310 | root->root_key.objectid = BTRFS_TREE_LOG_OBJECTID; |
1311 | root->root_key.type = BTRFS_ROOT_ITEM_KEY; | |
1312 | root->root_key.offset = BTRFS_TREE_LOG_OBJECTID; | |
27cdeb70 | 1313 | |
6ab6ebb7 NA |
1314 | return root; |
1315 | } | |
1316 | ||
1317 | int btrfs_alloc_log_tree_node(struct btrfs_trans_handle *trans, | |
1318 | struct btrfs_root *root) | |
1319 | { | |
1320 | struct extent_buffer *leaf; | |
1321 | ||
7237f183 | 1322 | /* |
92a7cc42 | 1323 | * DON'T set SHAREABLE bit for log trees. |
27cdeb70 | 1324 | * |
92a7cc42 QW |
1325 | * Log trees are not exposed to user space thus can't be snapshotted, |
1326 | * and they go away before a real commit is actually done. | |
1327 | * | |
1328 | * They do store pointers to file data extents, and those reference | |
1329 | * counts still get updated (along with back refs to the log tree). | |
7237f183 | 1330 | */ |
e02119d5 | 1331 | |
4d75f8a9 | 1332 | leaf = btrfs_alloc_tree_block(trans, root, 0, BTRFS_TREE_LOG_OBJECTID, |
9631e4cc | 1333 | NULL, 0, 0, 0, BTRFS_NESTING_NORMAL); |
6ab6ebb7 NA |
1334 | if (IS_ERR(leaf)) |
1335 | return PTR_ERR(leaf); | |
e02119d5 | 1336 | |
7237f183 | 1337 | root->node = leaf; |
e02119d5 | 1338 | |
e02119d5 CM |
1339 | btrfs_mark_buffer_dirty(root->node); |
1340 | btrfs_tree_unlock(root->node); | |
6ab6ebb7 NA |
1341 | |
1342 | return 0; | |
7237f183 YZ |
1343 | } |
1344 | ||
1345 | int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans, | |
1346 | struct btrfs_fs_info *fs_info) | |
1347 | { | |
1348 | struct btrfs_root *log_root; | |
1349 | ||
1350 | log_root = alloc_log_tree(trans, fs_info); | |
1351 | if (IS_ERR(log_root)) | |
1352 | return PTR_ERR(log_root); | |
6ab6ebb7 | 1353 | |
3ddebf27 NA |
1354 | if (!btrfs_is_zoned(fs_info)) { |
1355 | int ret = btrfs_alloc_log_tree_node(trans, log_root); | |
1356 | ||
1357 | if (ret) { | |
1358 | btrfs_put_root(log_root); | |
1359 | return ret; | |
1360 | } | |
6ab6ebb7 NA |
1361 | } |
1362 | ||
7237f183 YZ |
1363 | WARN_ON(fs_info->log_root_tree); |
1364 | fs_info->log_root_tree = log_root; | |
1365 | return 0; | |
1366 | } | |
1367 | ||
1368 | int btrfs_add_log_tree(struct btrfs_trans_handle *trans, | |
1369 | struct btrfs_root *root) | |
1370 | { | |
0b246afa | 1371 | struct btrfs_fs_info *fs_info = root->fs_info; |
7237f183 YZ |
1372 | struct btrfs_root *log_root; |
1373 | struct btrfs_inode_item *inode_item; | |
6ab6ebb7 | 1374 | int ret; |
7237f183 | 1375 | |
0b246afa | 1376 | log_root = alloc_log_tree(trans, fs_info); |
7237f183 YZ |
1377 | if (IS_ERR(log_root)) |
1378 | return PTR_ERR(log_root); | |
1379 | ||
6ab6ebb7 NA |
1380 | ret = btrfs_alloc_log_tree_node(trans, log_root); |
1381 | if (ret) { | |
1382 | btrfs_put_root(log_root); | |
1383 | return ret; | |
1384 | } | |
1385 | ||
7237f183 YZ |
1386 | log_root->last_trans = trans->transid; |
1387 | log_root->root_key.offset = root->root_key.objectid; | |
1388 | ||
1389 | inode_item = &log_root->root_item.inode; | |
3cae210f QW |
1390 | btrfs_set_stack_inode_generation(inode_item, 1); |
1391 | btrfs_set_stack_inode_size(inode_item, 3); | |
1392 | btrfs_set_stack_inode_nlink(inode_item, 1); | |
da17066c | 1393 | btrfs_set_stack_inode_nbytes(inode_item, |
0b246afa | 1394 | fs_info->nodesize); |
3cae210f | 1395 | btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755); |
7237f183 | 1396 | |
5d4f98a2 | 1397 | btrfs_set_root_node(&log_root->root_item, log_root->node); |
7237f183 YZ |
1398 | |
1399 | WARN_ON(root->log_root); | |
1400 | root->log_root = log_root; | |
1401 | root->log_transid = 0; | |
d1433deb | 1402 | root->log_transid_committed = -1; |
257c62e1 | 1403 | root->last_log_commit = 0; |
e02119d5 CM |
1404 | return 0; |
1405 | } | |
1406 | ||
49d11bea JB |
1407 | static struct btrfs_root *read_tree_root_path(struct btrfs_root *tree_root, |
1408 | struct btrfs_path *path, | |
1409 | struct btrfs_key *key) | |
e02119d5 CM |
1410 | { |
1411 | struct btrfs_root *root; | |
789d6a3a | 1412 | struct btrfs_tree_parent_check check = { 0 }; |
e02119d5 | 1413 | struct btrfs_fs_info *fs_info = tree_root->fs_info; |
84234f3a | 1414 | u64 generation; |
cb517eab | 1415 | int ret; |
581c1760 | 1416 | int level; |
0f7d52f4 | 1417 | |
96dfcb46 | 1418 | root = btrfs_alloc_root(fs_info, key->objectid, GFP_NOFS); |
49d11bea JB |
1419 | if (!root) |
1420 | return ERR_PTR(-ENOMEM); | |
0f7d52f4 | 1421 | |
cb517eab MX |
1422 | ret = btrfs_find_root(tree_root, key, path, |
1423 | &root->root_item, &root->root_key); | |
0f7d52f4 | 1424 | if (ret) { |
13a8a7c8 YZ |
1425 | if (ret > 0) |
1426 | ret = -ENOENT; | |
49d11bea | 1427 | goto fail; |
0f7d52f4 | 1428 | } |
13a8a7c8 | 1429 | |
84234f3a | 1430 | generation = btrfs_root_generation(&root->root_item); |
581c1760 | 1431 | level = btrfs_root_level(&root->root_item); |
789d6a3a QW |
1432 | check.level = level; |
1433 | check.transid = generation; | |
1434 | check.owner_root = key->objectid; | |
1435 | root->node = read_tree_block(fs_info, btrfs_root_bytenr(&root->root_item), | |
1436 | &check); | |
64c043de LB |
1437 | if (IS_ERR(root->node)) { |
1438 | ret = PTR_ERR(root->node); | |
8c38938c | 1439 | root->node = NULL; |
49d11bea | 1440 | goto fail; |
4eb150d6 QW |
1441 | } |
1442 | if (!btrfs_buffer_uptodate(root->node, generation, 0)) { | |
cb517eab | 1443 | ret = -EIO; |
49d11bea | 1444 | goto fail; |
416bc658 | 1445 | } |
88c602ab QW |
1446 | |
1447 | /* | |
1448 | * For real fs, and not log/reloc trees, root owner must | |
1449 | * match its root node owner | |
1450 | */ | |
1451 | if (!test_bit(BTRFS_FS_STATE_DUMMY_FS_INFO, &fs_info->fs_state) && | |
1452 | root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && | |
1453 | root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && | |
1454 | root->root_key.objectid != btrfs_header_owner(root->node)) { | |
1455 | btrfs_crit(fs_info, | |
1456 | "root=%llu block=%llu, tree root owner mismatch, have %llu expect %llu", | |
1457 | root->root_key.objectid, root->node->start, | |
1458 | btrfs_header_owner(root->node), | |
1459 | root->root_key.objectid); | |
1460 | ret = -EUCLEAN; | |
1461 | goto fail; | |
1462 | } | |
5d4f98a2 | 1463 | root->commit_root = btrfs_root_node(root); |
cb517eab | 1464 | return root; |
49d11bea | 1465 | fail: |
00246528 | 1466 | btrfs_put_root(root); |
49d11bea JB |
1467 | return ERR_PTR(ret); |
1468 | } | |
1469 | ||
1470 | struct btrfs_root *btrfs_read_tree_root(struct btrfs_root *tree_root, | |
1471 | struct btrfs_key *key) | |
1472 | { | |
1473 | struct btrfs_root *root; | |
1474 | struct btrfs_path *path; | |
1475 | ||
1476 | path = btrfs_alloc_path(); | |
1477 | if (!path) | |
1478 | return ERR_PTR(-ENOMEM); | |
1479 | root = read_tree_root_path(tree_root, path, key); | |
1480 | btrfs_free_path(path); | |
1481 | ||
1482 | return root; | |
cb517eab MX |
1483 | } |
1484 | ||
2dfb1e43 QW |
1485 | /* |
1486 | * Initialize subvolume root in-memory structure | |
1487 | * | |
1488 | * @anon_dev: anonymous device to attach to the root, if zero, allocate new | |
1489 | */ | |
1490 | static int btrfs_init_fs_root(struct btrfs_root *root, dev_t anon_dev) | |
cb517eab MX |
1491 | { |
1492 | int ret; | |
dcc3eb96 | 1493 | unsigned int nofs_flag; |
cb517eab | 1494 | |
dcc3eb96 NB |
1495 | /* |
1496 | * We might be called under a transaction (e.g. indirect backref | |
1497 | * resolution) which could deadlock if it triggers memory reclaim | |
1498 | */ | |
1499 | nofs_flag = memalloc_nofs_save(); | |
1500 | ret = btrfs_drew_lock_init(&root->snapshot_lock); | |
1501 | memalloc_nofs_restore(nofs_flag); | |
1502 | if (ret) | |
8257b2dc | 1503 | goto fail; |
8257b2dc | 1504 | |
aeb935a4 | 1505 | if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID && |
37f00a6d | 1506 | !btrfs_is_data_reloc_root(root)) { |
92a7cc42 | 1507 | set_bit(BTRFS_ROOT_SHAREABLE, &root->state); |
f39e4571 JB |
1508 | btrfs_check_and_init_root_item(&root->root_item); |
1509 | } | |
1510 | ||
851fd730 QW |
1511 | /* |
1512 | * Don't assign anonymous block device to roots that are not exposed to | |
1513 | * userspace, the id pool is limited to 1M | |
1514 | */ | |
1515 | if (is_fstree(root->root_key.objectid) && | |
1516 | btrfs_root_refs(&root->root_item) > 0) { | |
2dfb1e43 QW |
1517 | if (!anon_dev) { |
1518 | ret = get_anon_bdev(&root->anon_dev); | |
1519 | if (ret) | |
1520 | goto fail; | |
1521 | } else { | |
1522 | root->anon_dev = anon_dev; | |
1523 | } | |
851fd730 | 1524 | } |
f32e48e9 CR |
1525 | |
1526 | mutex_lock(&root->objectid_mutex); | |
453e4873 | 1527 | ret = btrfs_init_root_free_objectid(root); |
f32e48e9 CR |
1528 | if (ret) { |
1529 | mutex_unlock(&root->objectid_mutex); | |
876d2cf1 | 1530 | goto fail; |
f32e48e9 CR |
1531 | } |
1532 | ||
6b8fad57 | 1533 | ASSERT(root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
f32e48e9 CR |
1534 | |
1535 | mutex_unlock(&root->objectid_mutex); | |
1536 | ||
cb517eab MX |
1537 | return 0; |
1538 | fail: | |
84db5ccf | 1539 | /* The caller is responsible to call btrfs_free_fs_root */ |
cb517eab MX |
1540 | return ret; |
1541 | } | |
1542 | ||
a98db0f3 JB |
1543 | static struct btrfs_root *btrfs_lookup_fs_root(struct btrfs_fs_info *fs_info, |
1544 | u64 root_id) | |
cb517eab MX |
1545 | { |
1546 | struct btrfs_root *root; | |
1547 | ||
fc7cbcd4 DS |
1548 | spin_lock(&fs_info->fs_roots_radix_lock); |
1549 | root = radix_tree_lookup(&fs_info->fs_roots_radix, | |
1550 | (unsigned long)root_id); | |
bc44d7c4 | 1551 | if (root) |
00246528 | 1552 | root = btrfs_grab_root(root); |
fc7cbcd4 | 1553 | spin_unlock(&fs_info->fs_roots_radix_lock); |
cb517eab MX |
1554 | return root; |
1555 | } | |
1556 | ||
49d11bea JB |
1557 | static struct btrfs_root *btrfs_get_global_root(struct btrfs_fs_info *fs_info, |
1558 | u64 objectid) | |
1559 | { | |
abed4aaa JB |
1560 | struct btrfs_key key = { |
1561 | .objectid = objectid, | |
1562 | .type = BTRFS_ROOT_ITEM_KEY, | |
1563 | .offset = 0, | |
1564 | }; | |
1565 | ||
49d11bea JB |
1566 | if (objectid == BTRFS_ROOT_TREE_OBJECTID) |
1567 | return btrfs_grab_root(fs_info->tree_root); | |
1568 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
abed4aaa | 1569 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1570 | if (objectid == BTRFS_CHUNK_TREE_OBJECTID) |
1571 | return btrfs_grab_root(fs_info->chunk_root); | |
1572 | if (objectid == BTRFS_DEV_TREE_OBJECTID) | |
1573 | return btrfs_grab_root(fs_info->dev_root); | |
1574 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
abed4aaa | 1575 | return btrfs_grab_root(btrfs_global_root(fs_info, &key)); |
49d11bea JB |
1576 | if (objectid == BTRFS_QUOTA_TREE_OBJECTID) |
1577 | return btrfs_grab_root(fs_info->quota_root) ? | |
1578 | fs_info->quota_root : ERR_PTR(-ENOENT); | |
1579 | if (objectid == BTRFS_UUID_TREE_OBJECTID) | |
1580 | return btrfs_grab_root(fs_info->uuid_root) ? | |
1581 | fs_info->uuid_root : ERR_PTR(-ENOENT); | |
14033b08 QW |
1582 | if (objectid == BTRFS_BLOCK_GROUP_TREE_OBJECTID) |
1583 | return btrfs_grab_root(fs_info->block_group_root) ? | |
1584 | fs_info->block_group_root : ERR_PTR(-ENOENT); | |
abed4aaa JB |
1585 | if (objectid == BTRFS_FREE_SPACE_TREE_OBJECTID) { |
1586 | struct btrfs_root *root = btrfs_global_root(fs_info, &key); | |
1587 | ||
1588 | return btrfs_grab_root(root) ? root : ERR_PTR(-ENOENT); | |
1589 | } | |
49d11bea JB |
1590 | return NULL; |
1591 | } | |
1592 | ||
cb517eab MX |
1593 | int btrfs_insert_fs_root(struct btrfs_fs_info *fs_info, |
1594 | struct btrfs_root *root) | |
1595 | { | |
1596 | int ret; | |
1597 | ||
fc7cbcd4 DS |
1598 | ret = radix_tree_preload(GFP_NOFS); |
1599 | if (ret) | |
1600 | return ret; | |
1601 | ||
1602 | spin_lock(&fs_info->fs_roots_radix_lock); | |
1603 | ret = radix_tree_insert(&fs_info->fs_roots_radix, | |
1604 | (unsigned long)root->root_key.objectid, | |
1605 | root); | |
af01d2e5 | 1606 | if (ret == 0) { |
00246528 | 1607 | btrfs_grab_root(root); |
fc7cbcd4 | 1608 | set_bit(BTRFS_ROOT_IN_RADIX, &root->state); |
af01d2e5 | 1609 | } |
fc7cbcd4 DS |
1610 | spin_unlock(&fs_info->fs_roots_radix_lock); |
1611 | radix_tree_preload_end(); | |
cb517eab MX |
1612 | |
1613 | return ret; | |
1614 | } | |
1615 | ||
bd647ce3 JB |
1616 | void btrfs_check_leaked_roots(struct btrfs_fs_info *fs_info) |
1617 | { | |
1618 | #ifdef CONFIG_BTRFS_DEBUG | |
1619 | struct btrfs_root *root; | |
1620 | ||
1621 | while (!list_empty(&fs_info->allocated_roots)) { | |
457f1864 JB |
1622 | char buf[BTRFS_ROOT_NAME_BUF_LEN]; |
1623 | ||
bd647ce3 JB |
1624 | root = list_first_entry(&fs_info->allocated_roots, |
1625 | struct btrfs_root, leak_list); | |
457f1864 | 1626 | btrfs_err(fs_info, "leaked root %s refcount %d", |
71008734 | 1627 | btrfs_root_name(&root->root_key, buf), |
bd647ce3 JB |
1628 | refcount_read(&root->refs)); |
1629 | while (refcount_read(&root->refs) > 1) | |
00246528 JB |
1630 | btrfs_put_root(root); |
1631 | btrfs_put_root(root); | |
bd647ce3 JB |
1632 | } |
1633 | #endif | |
1634 | } | |
1635 | ||
abed4aaa JB |
1636 | static void free_global_roots(struct btrfs_fs_info *fs_info) |
1637 | { | |
1638 | struct btrfs_root *root; | |
1639 | struct rb_node *node; | |
1640 | ||
1641 | while ((node = rb_first_postorder(&fs_info->global_root_tree)) != NULL) { | |
1642 | root = rb_entry(node, struct btrfs_root, rb_node); | |
1643 | rb_erase(&root->rb_node, &fs_info->global_root_tree); | |
1644 | btrfs_put_root(root); | |
1645 | } | |
1646 | } | |
1647 | ||
0d4b0463 JB |
1648 | void btrfs_free_fs_info(struct btrfs_fs_info *fs_info) |
1649 | { | |
141386e1 JB |
1650 | percpu_counter_destroy(&fs_info->dirty_metadata_bytes); |
1651 | percpu_counter_destroy(&fs_info->delalloc_bytes); | |
5deb17e1 | 1652 | percpu_counter_destroy(&fs_info->ordered_bytes); |
141386e1 JB |
1653 | percpu_counter_destroy(&fs_info->dev_replace.bio_counter); |
1654 | btrfs_free_csum_hash(fs_info); | |
1655 | btrfs_free_stripe_hash_table(fs_info); | |
1656 | btrfs_free_ref_cache(fs_info); | |
0d4b0463 JB |
1657 | kfree(fs_info->balance_ctl); |
1658 | kfree(fs_info->delayed_root); | |
abed4aaa | 1659 | free_global_roots(fs_info); |
00246528 JB |
1660 | btrfs_put_root(fs_info->tree_root); |
1661 | btrfs_put_root(fs_info->chunk_root); | |
1662 | btrfs_put_root(fs_info->dev_root); | |
00246528 JB |
1663 | btrfs_put_root(fs_info->quota_root); |
1664 | btrfs_put_root(fs_info->uuid_root); | |
00246528 | 1665 | btrfs_put_root(fs_info->fs_root); |
aeb935a4 | 1666 | btrfs_put_root(fs_info->data_reloc_root); |
9c54e80d | 1667 | btrfs_put_root(fs_info->block_group_root); |
bd647ce3 | 1668 | btrfs_check_leaked_roots(fs_info); |
3fd63727 | 1669 | btrfs_extent_buffer_leak_debug_check(fs_info); |
0d4b0463 JB |
1670 | kfree(fs_info->super_copy); |
1671 | kfree(fs_info->super_for_commit); | |
8481dd80 | 1672 | kfree(fs_info->subpage_info); |
0d4b0463 JB |
1673 | kvfree(fs_info); |
1674 | } | |
1675 | ||
1676 | ||
2dfb1e43 QW |
1677 | /* |
1678 | * Get an in-memory reference of a root structure. | |
1679 | * | |
1680 | * For essential trees like root/extent tree, we grab it from fs_info directly. | |
1681 | * For subvolume trees, we check the cached filesystem roots first. If not | |
1682 | * found, then read it from disk and add it to cached fs roots. | |
1683 | * | |
1684 | * Caller should release the root by calling btrfs_put_root() after the usage. | |
1685 | * | |
1686 | * NOTE: Reloc and log trees can't be read by this function as they share the | |
1687 | * same root objectid. | |
1688 | * | |
1689 | * @objectid: root id | |
1690 | * @anon_dev: preallocated anonymous block device number for new roots, | |
1691 | * pass 0 for new allocation. | |
1692 | * @check_ref: whether to check root item references, If true, return -ENOENT | |
1693 | * for orphan roots | |
1694 | */ | |
1695 | static struct btrfs_root *btrfs_get_root_ref(struct btrfs_fs_info *fs_info, | |
1696 | u64 objectid, dev_t anon_dev, | |
1697 | bool check_ref) | |
5eda7b5e CM |
1698 | { |
1699 | struct btrfs_root *root; | |
381cf658 | 1700 | struct btrfs_path *path; |
1d4c08e0 | 1701 | struct btrfs_key key; |
5eda7b5e CM |
1702 | int ret; |
1703 | ||
49d11bea JB |
1704 | root = btrfs_get_global_root(fs_info, objectid); |
1705 | if (root) | |
1706 | return root; | |
4df27c4d | 1707 | again: |
56e9357a | 1708 | root = btrfs_lookup_fs_root(fs_info, objectid); |
48475471 | 1709 | if (root) { |
2dfb1e43 QW |
1710 | /* Shouldn't get preallocated anon_dev for cached roots */ |
1711 | ASSERT(!anon_dev); | |
bc44d7c4 | 1712 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
00246528 | 1713 | btrfs_put_root(root); |
48475471 | 1714 | return ERR_PTR(-ENOENT); |
bc44d7c4 | 1715 | } |
5eda7b5e | 1716 | return root; |
48475471 | 1717 | } |
5eda7b5e | 1718 | |
56e9357a DS |
1719 | key.objectid = objectid; |
1720 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1721 | key.offset = (u64)-1; | |
1722 | root = btrfs_read_tree_root(fs_info->tree_root, &key); | |
5eda7b5e CM |
1723 | if (IS_ERR(root)) |
1724 | return root; | |
3394e160 | 1725 | |
c00869f1 | 1726 | if (check_ref && btrfs_root_refs(&root->root_item) == 0) { |
cb517eab | 1727 | ret = -ENOENT; |
581bb050 | 1728 | goto fail; |
35a30d7c | 1729 | } |
581bb050 | 1730 | |
2dfb1e43 | 1731 | ret = btrfs_init_fs_root(root, anon_dev); |
ac08aedf CM |
1732 | if (ret) |
1733 | goto fail; | |
3394e160 | 1734 | |
381cf658 DS |
1735 | path = btrfs_alloc_path(); |
1736 | if (!path) { | |
1737 | ret = -ENOMEM; | |
1738 | goto fail; | |
1739 | } | |
1d4c08e0 DS |
1740 | key.objectid = BTRFS_ORPHAN_OBJECTID; |
1741 | key.type = BTRFS_ORPHAN_ITEM_KEY; | |
56e9357a | 1742 | key.offset = objectid; |
1d4c08e0 DS |
1743 | |
1744 | ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0); | |
381cf658 | 1745 | btrfs_free_path(path); |
d68fc57b YZ |
1746 | if (ret < 0) |
1747 | goto fail; | |
1748 | if (ret == 0) | |
27cdeb70 | 1749 | set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); |
d68fc57b | 1750 | |
cb517eab | 1751 | ret = btrfs_insert_fs_root(fs_info, root); |
0f7d52f4 | 1752 | if (ret) { |
168a2f77 JJB |
1753 | if (ret == -EEXIST) { |
1754 | btrfs_put_root(root); | |
4df27c4d | 1755 | goto again; |
168a2f77 | 1756 | } |
4df27c4d | 1757 | goto fail; |
0f7d52f4 | 1758 | } |
edbd8d4e | 1759 | return root; |
4df27c4d | 1760 | fail: |
33fab972 FM |
1761 | /* |
1762 | * If our caller provided us an anonymous device, then it's his | |
143823cf | 1763 | * responsibility to free it in case we fail. So we have to set our |
33fab972 FM |
1764 | * root's anon_dev to 0 to avoid a double free, once by btrfs_put_root() |
1765 | * and once again by our caller. | |
1766 | */ | |
1767 | if (anon_dev) | |
1768 | root->anon_dev = 0; | |
8c38938c | 1769 | btrfs_put_root(root); |
4df27c4d | 1770 | return ERR_PTR(ret); |
edbd8d4e CM |
1771 | } |
1772 | ||
2dfb1e43 QW |
1773 | /* |
1774 | * Get in-memory reference of a root structure | |
1775 | * | |
1776 | * @objectid: tree objectid | |
1777 | * @check_ref: if set, verify that the tree exists and the item has at least | |
1778 | * one reference | |
1779 | */ | |
1780 | struct btrfs_root *btrfs_get_fs_root(struct btrfs_fs_info *fs_info, | |
1781 | u64 objectid, bool check_ref) | |
1782 | { | |
1783 | return btrfs_get_root_ref(fs_info, objectid, 0, check_ref); | |
1784 | } | |
1785 | ||
1786 | /* | |
1787 | * Get in-memory reference of a root structure, created as new, optionally pass | |
1788 | * the anonymous block device id | |
1789 | * | |
1790 | * @objectid: tree objectid | |
1791 | * @anon_dev: if zero, allocate a new anonymous block device or use the | |
1792 | * parameter value | |
1793 | */ | |
1794 | struct btrfs_root *btrfs_get_new_fs_root(struct btrfs_fs_info *fs_info, | |
1795 | u64 objectid, dev_t anon_dev) | |
1796 | { | |
1797 | return btrfs_get_root_ref(fs_info, objectid, anon_dev, true); | |
1798 | } | |
1799 | ||
49d11bea JB |
1800 | /* |
1801 | * btrfs_get_fs_root_commit_root - return a root for the given objectid | |
1802 | * @fs_info: the fs_info | |
1803 | * @objectid: the objectid we need to lookup | |
1804 | * | |
1805 | * This is exclusively used for backref walking, and exists specifically because | |
1806 | * of how qgroups does lookups. Qgroups will do a backref lookup at delayed ref | |
1807 | * creation time, which means we may have to read the tree_root in order to look | |
1808 | * up a fs root that is not in memory. If the root is not in memory we will | |
1809 | * read the tree root commit root and look up the fs root from there. This is a | |
1810 | * temporary root, it will not be inserted into the radix tree as it doesn't | |
1811 | * have the most uptodate information, it'll simply be discarded once the | |
1812 | * backref code is finished using the root. | |
1813 | */ | |
1814 | struct btrfs_root *btrfs_get_fs_root_commit_root(struct btrfs_fs_info *fs_info, | |
1815 | struct btrfs_path *path, | |
1816 | u64 objectid) | |
1817 | { | |
1818 | struct btrfs_root *root; | |
1819 | struct btrfs_key key; | |
1820 | ||
1821 | ASSERT(path->search_commit_root && path->skip_locking); | |
1822 | ||
1823 | /* | |
1824 | * This can return -ENOENT if we ask for a root that doesn't exist, but | |
1825 | * since this is called via the backref walking code we won't be looking | |
1826 | * up a root that doesn't exist, unless there's corruption. So if root | |
1827 | * != NULL just return it. | |
1828 | */ | |
1829 | root = btrfs_get_global_root(fs_info, objectid); | |
1830 | if (root) | |
1831 | return root; | |
1832 | ||
1833 | root = btrfs_lookup_fs_root(fs_info, objectid); | |
1834 | if (root) | |
1835 | return root; | |
1836 | ||
1837 | key.objectid = objectid; | |
1838 | key.type = BTRFS_ROOT_ITEM_KEY; | |
1839 | key.offset = (u64)-1; | |
1840 | root = read_tree_root_path(fs_info->tree_root, path, &key); | |
1841 | btrfs_release_path(path); | |
1842 | ||
1843 | return root; | |
1844 | } | |
1845 | ||
a74a4b97 CM |
1846 | static int cleaner_kthread(void *arg) |
1847 | { | |
0d031dc4 | 1848 | struct btrfs_fs_info *fs_info = arg; |
d0278245 | 1849 | int again; |
a74a4b97 | 1850 | |
d6fd0ae2 | 1851 | while (1) { |
d0278245 | 1852 | again = 0; |
a74a4b97 | 1853 | |
fd340d0f JB |
1854 | set_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
1855 | ||
d0278245 | 1856 | /* Make the cleaner go to sleep early. */ |
2ff7e61e | 1857 | if (btrfs_need_cleaner_sleep(fs_info)) |
d0278245 MX |
1858 | goto sleep; |
1859 | ||
90c711ab ZB |
1860 | /* |
1861 | * Do not do anything if we might cause open_ctree() to block | |
1862 | * before we have finished mounting the filesystem. | |
1863 | */ | |
0b246afa | 1864 | if (!test_bit(BTRFS_FS_OPEN, &fs_info->flags)) |
90c711ab ZB |
1865 | goto sleep; |
1866 | ||
0b246afa | 1867 | if (!mutex_trylock(&fs_info->cleaner_mutex)) |
d0278245 MX |
1868 | goto sleep; |
1869 | ||
dc7f370c MX |
1870 | /* |
1871 | * Avoid the problem that we change the status of the fs | |
1872 | * during the above check and trylock. | |
1873 | */ | |
2ff7e61e | 1874 | if (btrfs_need_cleaner_sleep(fs_info)) { |
0b246afa | 1875 | mutex_unlock(&fs_info->cleaner_mutex); |
dc7f370c | 1876 | goto sleep; |
76dda93c | 1877 | } |
a74a4b97 | 1878 | |
2ff7e61e | 1879 | btrfs_run_delayed_iputs(fs_info); |
c2d6cb16 | 1880 | |
33c44184 | 1881 | again = btrfs_clean_one_deleted_snapshot(fs_info); |
0b246afa | 1882 | mutex_unlock(&fs_info->cleaner_mutex); |
d0278245 MX |
1883 | |
1884 | /* | |
05323cd1 MX |
1885 | * The defragger has dealt with the R/O remount and umount, |
1886 | * needn't do anything special here. | |
d0278245 | 1887 | */ |
0b246afa | 1888 | btrfs_run_defrag_inodes(fs_info); |
67c5e7d4 FM |
1889 | |
1890 | /* | |
f3372065 | 1891 | * Acquires fs_info->reclaim_bgs_lock to avoid racing |
67c5e7d4 FM |
1892 | * with relocation (btrfs_relocate_chunk) and relocation |
1893 | * acquires fs_info->cleaner_mutex (btrfs_relocate_block_group) | |
f3372065 | 1894 | * after acquiring fs_info->reclaim_bgs_lock. So we |
67c5e7d4 FM |
1895 | * can't hold, nor need to, fs_info->cleaner_mutex when deleting |
1896 | * unused block groups. | |
1897 | */ | |
0b246afa | 1898 | btrfs_delete_unused_bgs(fs_info); |
18bb8bbf JT |
1899 | |
1900 | /* | |
1901 | * Reclaim block groups in the reclaim_bgs list after we deleted | |
1902 | * all unused block_groups. This possibly gives us some more free | |
1903 | * space. | |
1904 | */ | |
1905 | btrfs_reclaim_bgs(fs_info); | |
d0278245 | 1906 | sleep: |
a0a1db70 | 1907 | clear_and_wake_up_bit(BTRFS_FS_CLEANER_RUNNING, &fs_info->flags); |
d6fd0ae2 OS |
1908 | if (kthread_should_park()) |
1909 | kthread_parkme(); | |
1910 | if (kthread_should_stop()) | |
1911 | return 0; | |
838fe188 | 1912 | if (!again) { |
a74a4b97 | 1913 | set_current_state(TASK_INTERRUPTIBLE); |
d6fd0ae2 | 1914 | schedule(); |
a74a4b97 CM |
1915 | __set_current_state(TASK_RUNNING); |
1916 | } | |
da288d28 | 1917 | } |
a74a4b97 CM |
1918 | } |
1919 | ||
1920 | static int transaction_kthread(void *arg) | |
1921 | { | |
1922 | struct btrfs_root *root = arg; | |
0b246afa | 1923 | struct btrfs_fs_info *fs_info = root->fs_info; |
a74a4b97 CM |
1924 | struct btrfs_trans_handle *trans; |
1925 | struct btrfs_transaction *cur; | |
8929ecfa | 1926 | u64 transid; |
643900be | 1927 | time64_t delta; |
a74a4b97 | 1928 | unsigned long delay; |
914b2007 | 1929 | bool cannot_commit; |
a74a4b97 CM |
1930 | |
1931 | do { | |
914b2007 | 1932 | cannot_commit = false; |
ba1bc00f | 1933 | delay = msecs_to_jiffies(fs_info->commit_interval * 1000); |
0b246afa | 1934 | mutex_lock(&fs_info->transaction_kthread_mutex); |
a74a4b97 | 1935 | |
0b246afa JM |
1936 | spin_lock(&fs_info->trans_lock); |
1937 | cur = fs_info->running_transaction; | |
a74a4b97 | 1938 | if (!cur) { |
0b246afa | 1939 | spin_unlock(&fs_info->trans_lock); |
a74a4b97 CM |
1940 | goto sleep; |
1941 | } | |
31153d81 | 1942 | |
643900be | 1943 | delta = ktime_get_seconds() - cur->start_time; |
fdfbf020 JB |
1944 | if (!test_and_clear_bit(BTRFS_FS_COMMIT_TRANS, &fs_info->flags) && |
1945 | cur->state < TRANS_STATE_COMMIT_START && | |
643900be | 1946 | delta < fs_info->commit_interval) { |
0b246afa | 1947 | spin_unlock(&fs_info->trans_lock); |
fb8a7e94 NB |
1948 | delay -= msecs_to_jiffies((delta - 1) * 1000); |
1949 | delay = min(delay, | |
1950 | msecs_to_jiffies(fs_info->commit_interval * 1000)); | |
a74a4b97 CM |
1951 | goto sleep; |
1952 | } | |
8929ecfa | 1953 | transid = cur->transid; |
0b246afa | 1954 | spin_unlock(&fs_info->trans_lock); |
56bec294 | 1955 | |
79787eaa | 1956 | /* If the file system is aborted, this will always fail. */ |
354aa0fb | 1957 | trans = btrfs_attach_transaction(root); |
914b2007 | 1958 | if (IS_ERR(trans)) { |
354aa0fb MX |
1959 | if (PTR_ERR(trans) != -ENOENT) |
1960 | cannot_commit = true; | |
79787eaa | 1961 | goto sleep; |
914b2007 | 1962 | } |
8929ecfa | 1963 | if (transid == trans->transid) { |
3a45bb20 | 1964 | btrfs_commit_transaction(trans); |
8929ecfa | 1965 | } else { |
3a45bb20 | 1966 | btrfs_end_transaction(trans); |
8929ecfa | 1967 | } |
a74a4b97 | 1968 | sleep: |
0b246afa JM |
1969 | wake_up_process(fs_info->cleaner_kthread); |
1970 | mutex_unlock(&fs_info->transaction_kthread_mutex); | |
a74a4b97 | 1971 | |
84961539 | 1972 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 1973 | btrfs_cleanup_transaction(fs_info); |
ce63f891 | 1974 | if (!kthread_should_stop() && |
0b246afa | 1975 | (!btrfs_transaction_blocked(fs_info) || |
ce63f891 | 1976 | cannot_commit)) |
bc5511d0 | 1977 | schedule_timeout_interruptible(delay); |
a74a4b97 CM |
1978 | } while (!kthread_should_stop()); |
1979 | return 0; | |
1980 | } | |
1981 | ||
af31f5e5 | 1982 | /* |
01f0f9da NB |
1983 | * This will find the highest generation in the array of root backups. The |
1984 | * index of the highest array is returned, or -EINVAL if we can't find | |
1985 | * anything. | |
af31f5e5 CM |
1986 | * |
1987 | * We check to make sure the array is valid by comparing the | |
1988 | * generation of the latest root in the array with the generation | |
1989 | * in the super block. If they don't match we pitch it. | |
1990 | */ | |
01f0f9da | 1991 | static int find_newest_super_backup(struct btrfs_fs_info *info) |
af31f5e5 | 1992 | { |
01f0f9da | 1993 | const u64 newest_gen = btrfs_super_generation(info->super_copy); |
af31f5e5 | 1994 | u64 cur; |
af31f5e5 CM |
1995 | struct btrfs_root_backup *root_backup; |
1996 | int i; | |
1997 | ||
1998 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
1999 | root_backup = info->super_copy->super_roots + i; | |
2000 | cur = btrfs_backup_tree_root_gen(root_backup); | |
2001 | if (cur == newest_gen) | |
01f0f9da | 2002 | return i; |
af31f5e5 CM |
2003 | } |
2004 | ||
01f0f9da | 2005 | return -EINVAL; |
af31f5e5 CM |
2006 | } |
2007 | ||
af31f5e5 CM |
2008 | /* |
2009 | * copy all the root pointers into the super backup array. | |
2010 | * this will bump the backup pointer by one when it is | |
2011 | * done | |
2012 | */ | |
2013 | static void backup_super_roots(struct btrfs_fs_info *info) | |
2014 | { | |
6ef108dd | 2015 | const int next_backup = info->backup_root_index; |
af31f5e5 | 2016 | struct btrfs_root_backup *root_backup; |
af31f5e5 CM |
2017 | |
2018 | root_backup = info->super_for_commit->super_roots + next_backup; | |
2019 | ||
2020 | /* | |
2021 | * make sure all of our padding and empty slots get zero filled | |
2022 | * regardless of which ones we use today | |
2023 | */ | |
2024 | memset(root_backup, 0, sizeof(*root_backup)); | |
2025 | ||
2026 | info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS; | |
2027 | ||
2028 | btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start); | |
2029 | btrfs_set_backup_tree_root_gen(root_backup, | |
2030 | btrfs_header_generation(info->tree_root->node)); | |
2031 | ||
2032 | btrfs_set_backup_tree_root_level(root_backup, | |
2033 | btrfs_header_level(info->tree_root->node)); | |
2034 | ||
2035 | btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start); | |
2036 | btrfs_set_backup_chunk_root_gen(root_backup, | |
2037 | btrfs_header_generation(info->chunk_root->node)); | |
2038 | btrfs_set_backup_chunk_root_level(root_backup, | |
2039 | btrfs_header_level(info->chunk_root->node)); | |
2040 | ||
1c56ab99 | 2041 | if (!btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE)) { |
9c54e80d | 2042 | struct btrfs_root *extent_root = btrfs_extent_root(info, 0); |
f7238e50 | 2043 | struct btrfs_root *csum_root = btrfs_csum_root(info, 0); |
9c54e80d JB |
2044 | |
2045 | btrfs_set_backup_extent_root(root_backup, | |
2046 | extent_root->node->start); | |
2047 | btrfs_set_backup_extent_root_gen(root_backup, | |
2048 | btrfs_header_generation(extent_root->node)); | |
2049 | btrfs_set_backup_extent_root_level(root_backup, | |
2050 | btrfs_header_level(extent_root->node)); | |
f7238e50 JB |
2051 | |
2052 | btrfs_set_backup_csum_root(root_backup, csum_root->node->start); | |
2053 | btrfs_set_backup_csum_root_gen(root_backup, | |
2054 | btrfs_header_generation(csum_root->node)); | |
2055 | btrfs_set_backup_csum_root_level(root_backup, | |
2056 | btrfs_header_level(csum_root->node)); | |
9c54e80d | 2057 | } |
af31f5e5 | 2058 | |
7c7e82a7 CM |
2059 | /* |
2060 | * we might commit during log recovery, which happens before we set | |
2061 | * the fs_root. Make sure it is valid before we fill it in. | |
2062 | */ | |
2063 | if (info->fs_root && info->fs_root->node) { | |
2064 | btrfs_set_backup_fs_root(root_backup, | |
2065 | info->fs_root->node->start); | |
2066 | btrfs_set_backup_fs_root_gen(root_backup, | |
af31f5e5 | 2067 | btrfs_header_generation(info->fs_root->node)); |
7c7e82a7 | 2068 | btrfs_set_backup_fs_root_level(root_backup, |
af31f5e5 | 2069 | btrfs_header_level(info->fs_root->node)); |
7c7e82a7 | 2070 | } |
af31f5e5 CM |
2071 | |
2072 | btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start); | |
2073 | btrfs_set_backup_dev_root_gen(root_backup, | |
2074 | btrfs_header_generation(info->dev_root->node)); | |
2075 | btrfs_set_backup_dev_root_level(root_backup, | |
2076 | btrfs_header_level(info->dev_root->node)); | |
2077 | ||
af31f5e5 CM |
2078 | btrfs_set_backup_total_bytes(root_backup, |
2079 | btrfs_super_total_bytes(info->super_copy)); | |
2080 | btrfs_set_backup_bytes_used(root_backup, | |
2081 | btrfs_super_bytes_used(info->super_copy)); | |
2082 | btrfs_set_backup_num_devices(root_backup, | |
2083 | btrfs_super_num_devices(info->super_copy)); | |
2084 | ||
2085 | /* | |
2086 | * if we don't copy this out to the super_copy, it won't get remembered | |
2087 | * for the next commit | |
2088 | */ | |
2089 | memcpy(&info->super_copy->super_roots, | |
2090 | &info->super_for_commit->super_roots, | |
2091 | sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS); | |
2092 | } | |
2093 | ||
bd2336b2 NB |
2094 | /* |
2095 | * read_backup_root - Reads a backup root based on the passed priority. Prio 0 | |
2096 | * is the newest, prio 1/2/3 are 2nd newest/3rd newest/4th (oldest) backup roots | |
2097 | * | |
2098 | * fs_info - filesystem whose backup roots need to be read | |
2099 | * priority - priority of backup root required | |
2100 | * | |
2101 | * Returns backup root index on success and -EINVAL otherwise. | |
2102 | */ | |
2103 | static int read_backup_root(struct btrfs_fs_info *fs_info, u8 priority) | |
2104 | { | |
2105 | int backup_index = find_newest_super_backup(fs_info); | |
2106 | struct btrfs_super_block *super = fs_info->super_copy; | |
2107 | struct btrfs_root_backup *root_backup; | |
2108 | ||
2109 | if (priority < BTRFS_NUM_BACKUP_ROOTS && backup_index >= 0) { | |
2110 | if (priority == 0) | |
2111 | return backup_index; | |
2112 | ||
2113 | backup_index = backup_index + BTRFS_NUM_BACKUP_ROOTS - priority; | |
2114 | backup_index %= BTRFS_NUM_BACKUP_ROOTS; | |
2115 | } else { | |
2116 | return -EINVAL; | |
2117 | } | |
2118 | ||
2119 | root_backup = super->super_roots + backup_index; | |
2120 | ||
2121 | btrfs_set_super_generation(super, | |
2122 | btrfs_backup_tree_root_gen(root_backup)); | |
2123 | btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup)); | |
2124 | btrfs_set_super_root_level(super, | |
2125 | btrfs_backup_tree_root_level(root_backup)); | |
2126 | btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup)); | |
2127 | ||
2128 | /* | |
2129 | * Fixme: the total bytes and num_devices need to match or we should | |
2130 | * need a fsck | |
2131 | */ | |
2132 | btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup)); | |
2133 | btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup)); | |
2134 | ||
2135 | return backup_index; | |
2136 | } | |
2137 | ||
7abadb64 LB |
2138 | /* helper to cleanup workers */ |
2139 | static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) | |
2140 | { | |
dc6e3209 | 2141 | btrfs_destroy_workqueue(fs_info->fixup_workers); |
afe3d242 | 2142 | btrfs_destroy_workqueue(fs_info->delalloc_workers); |
a31b4a43 | 2143 | btrfs_destroy_workqueue(fs_info->hipri_workers); |
5cdc7ad3 | 2144 | btrfs_destroy_workqueue(fs_info->workers); |
d7b9416f CH |
2145 | if (fs_info->endio_workers) |
2146 | destroy_workqueue(fs_info->endio_workers); | |
385de0ef CH |
2147 | if (fs_info->rmw_workers) |
2148 | destroy_workqueue(fs_info->rmw_workers); | |
fed8a72d CH |
2149 | if (fs_info->compressed_write_workers) |
2150 | destroy_workqueue(fs_info->compressed_write_workers); | |
fccb5d86 QW |
2151 | btrfs_destroy_workqueue(fs_info->endio_write_workers); |
2152 | btrfs_destroy_workqueue(fs_info->endio_freespace_worker); | |
5b3bc44e | 2153 | btrfs_destroy_workqueue(fs_info->delayed_workers); |
e66f0bb1 | 2154 | btrfs_destroy_workqueue(fs_info->caching_workers); |
a44903ab | 2155 | btrfs_destroy_workqueue(fs_info->flush_workers); |
fc97fab0 | 2156 | btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); |
b0643e59 DZ |
2157 | if (fs_info->discard_ctl.discard_workers) |
2158 | destroy_workqueue(fs_info->discard_ctl.discard_workers); | |
a9b9477d FM |
2159 | /* |
2160 | * Now that all other work queues are destroyed, we can safely destroy | |
2161 | * the queues used for metadata I/O, since tasks from those other work | |
2162 | * queues can do metadata I/O operations. | |
2163 | */ | |
d7b9416f CH |
2164 | if (fs_info->endio_meta_workers) |
2165 | destroy_workqueue(fs_info->endio_meta_workers); | |
7abadb64 LB |
2166 | } |
2167 | ||
2e9f5954 R |
2168 | static void free_root_extent_buffers(struct btrfs_root *root) |
2169 | { | |
2170 | if (root) { | |
2171 | free_extent_buffer(root->node); | |
2172 | free_extent_buffer(root->commit_root); | |
2173 | root->node = NULL; | |
2174 | root->commit_root = NULL; | |
2175 | } | |
2176 | } | |
2177 | ||
abed4aaa JB |
2178 | static void free_global_root_pointers(struct btrfs_fs_info *fs_info) |
2179 | { | |
2180 | struct btrfs_root *root, *tmp; | |
2181 | ||
2182 | rbtree_postorder_for_each_entry_safe(root, tmp, | |
2183 | &fs_info->global_root_tree, | |
2184 | rb_node) | |
2185 | free_root_extent_buffers(root); | |
2186 | } | |
2187 | ||
af31f5e5 | 2188 | /* helper to cleanup tree roots */ |
4273eaff | 2189 | static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
af31f5e5 | 2190 | { |
2e9f5954 | 2191 | free_root_extent_buffers(info->tree_root); |
655b09fe | 2192 | |
abed4aaa | 2193 | free_global_root_pointers(info); |
2e9f5954 | 2194 | free_root_extent_buffers(info->dev_root); |
2e9f5954 R |
2195 | free_root_extent_buffers(info->quota_root); |
2196 | free_root_extent_buffers(info->uuid_root); | |
8c38938c | 2197 | free_root_extent_buffers(info->fs_root); |
aeb935a4 | 2198 | free_root_extent_buffers(info->data_reloc_root); |
9c54e80d | 2199 | free_root_extent_buffers(info->block_group_root); |
4273eaff | 2200 | if (free_chunk_root) |
2e9f5954 | 2201 | free_root_extent_buffers(info->chunk_root); |
af31f5e5 CM |
2202 | } |
2203 | ||
8c38938c JB |
2204 | void btrfs_put_root(struct btrfs_root *root) |
2205 | { | |
2206 | if (!root) | |
2207 | return; | |
2208 | ||
2209 | if (refcount_dec_and_test(&root->refs)) { | |
2210 | WARN_ON(!RB_EMPTY_ROOT(&root->inode_tree)); | |
1dae7e0e | 2211 | WARN_ON(test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state)); |
8c38938c JB |
2212 | if (root->anon_dev) |
2213 | free_anon_bdev(root->anon_dev); | |
2214 | btrfs_drew_lock_destroy(&root->snapshot_lock); | |
923eb523 | 2215 | free_root_extent_buffers(root); |
8c38938c | 2216 | #ifdef CONFIG_BTRFS_DEBUG |
fc7cbcd4 | 2217 | spin_lock(&root->fs_info->fs_roots_radix_lock); |
8c38938c | 2218 | list_del_init(&root->leak_list); |
fc7cbcd4 | 2219 | spin_unlock(&root->fs_info->fs_roots_radix_lock); |
8c38938c JB |
2220 | #endif |
2221 | kfree(root); | |
2222 | } | |
2223 | } | |
2224 | ||
faa2dbf0 | 2225 | void btrfs_free_fs_roots(struct btrfs_fs_info *fs_info) |
171f6537 | 2226 | { |
fc7cbcd4 DS |
2227 | int ret; |
2228 | struct btrfs_root *gang[8]; | |
2229 | int i; | |
171f6537 JB |
2230 | |
2231 | while (!list_empty(&fs_info->dead_roots)) { | |
fc7cbcd4 DS |
2232 | gang[0] = list_entry(fs_info->dead_roots.next, |
2233 | struct btrfs_root, root_list); | |
2234 | list_del(&gang[0]->root_list); | |
171f6537 | 2235 | |
fc7cbcd4 DS |
2236 | if (test_bit(BTRFS_ROOT_IN_RADIX, &gang[0]->state)) |
2237 | btrfs_drop_and_free_fs_root(fs_info, gang[0]); | |
2238 | btrfs_put_root(gang[0]); | |
171f6537 JB |
2239 | } |
2240 | ||
fc7cbcd4 DS |
2241 | while (1) { |
2242 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
2243 | (void **)gang, 0, | |
2244 | ARRAY_SIZE(gang)); | |
2245 | if (!ret) | |
2246 | break; | |
2247 | for (i = 0; i < ret; i++) | |
2248 | btrfs_drop_and_free_fs_root(fs_info, gang[i]); | |
171f6537 JB |
2249 | } |
2250 | } | |
af31f5e5 | 2251 | |
638aa7ed ES |
2252 | static void btrfs_init_scrub(struct btrfs_fs_info *fs_info) |
2253 | { | |
2254 | mutex_init(&fs_info->scrub_lock); | |
2255 | atomic_set(&fs_info->scrubs_running, 0); | |
2256 | atomic_set(&fs_info->scrub_pause_req, 0); | |
2257 | atomic_set(&fs_info->scrubs_paused, 0); | |
2258 | atomic_set(&fs_info->scrub_cancel_req, 0); | |
2259 | init_waitqueue_head(&fs_info->scrub_pause_wait); | |
ff09c4ca | 2260 | refcount_set(&fs_info->scrub_workers_refcnt, 0); |
638aa7ed ES |
2261 | } |
2262 | ||
779a65a4 ES |
2263 | static void btrfs_init_balance(struct btrfs_fs_info *fs_info) |
2264 | { | |
2265 | spin_lock_init(&fs_info->balance_lock); | |
2266 | mutex_init(&fs_info->balance_mutex); | |
779a65a4 ES |
2267 | atomic_set(&fs_info->balance_pause_req, 0); |
2268 | atomic_set(&fs_info->balance_cancel_req, 0); | |
2269 | fs_info->balance_ctl = NULL; | |
2270 | init_waitqueue_head(&fs_info->balance_wait_q); | |
907d2710 | 2271 | atomic_set(&fs_info->reloc_cancel_req, 0); |
779a65a4 ES |
2272 | } |
2273 | ||
6bccf3ab | 2274 | static void btrfs_init_btree_inode(struct btrfs_fs_info *fs_info) |
f37938e0 | 2275 | { |
2ff7e61e | 2276 | struct inode *inode = fs_info->btree_inode; |
e256927b JB |
2277 | unsigned long hash = btrfs_inode_hash(BTRFS_BTREE_INODE_OBJECTID, |
2278 | fs_info->tree_root); | |
2ff7e61e JM |
2279 | |
2280 | inode->i_ino = BTRFS_BTREE_INODE_OBJECTID; | |
2281 | set_nlink(inode, 1); | |
f37938e0 ES |
2282 | /* |
2283 | * we set the i_size on the btree inode to the max possible int. | |
2284 | * the real end of the address space is determined by all of | |
2285 | * the devices in the system | |
2286 | */ | |
2ff7e61e JM |
2287 | inode->i_size = OFFSET_MAX; |
2288 | inode->i_mapping->a_ops = &btree_aops; | |
f37938e0 | 2289 | |
2ff7e61e | 2290 | RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node); |
43eb5f29 | 2291 | extent_io_tree_init(fs_info, &BTRFS_I(inode)->io_tree, |
35da5a7e | 2292 | IO_TREE_BTREE_INODE_IO); |
2ff7e61e | 2293 | extent_map_tree_init(&BTRFS_I(inode)->extent_tree); |
f37938e0 | 2294 | |
5c8fd99f | 2295 | BTRFS_I(inode)->root = btrfs_grab_root(fs_info->tree_root); |
adac5584 FM |
2296 | BTRFS_I(inode)->location.objectid = BTRFS_BTREE_INODE_OBJECTID; |
2297 | BTRFS_I(inode)->location.type = 0; | |
2298 | BTRFS_I(inode)->location.offset = 0; | |
2ff7e61e | 2299 | set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags); |
e256927b | 2300 | __insert_inode_hash(inode, hash); |
f37938e0 ES |
2301 | } |
2302 | ||
ad618368 ES |
2303 | static void btrfs_init_dev_replace_locks(struct btrfs_fs_info *fs_info) |
2304 | { | |
ad618368 | 2305 | mutex_init(&fs_info->dev_replace.lock_finishing_cancel_unmount); |
129827e3 | 2306 | init_rwsem(&fs_info->dev_replace.rwsem); |
7f8d236a | 2307 | init_waitqueue_head(&fs_info->dev_replace.replace_wait); |
ad618368 ES |
2308 | } |
2309 | ||
f9e92e40 ES |
2310 | static void btrfs_init_qgroup(struct btrfs_fs_info *fs_info) |
2311 | { | |
2312 | spin_lock_init(&fs_info->qgroup_lock); | |
2313 | mutex_init(&fs_info->qgroup_ioctl_lock); | |
2314 | fs_info->qgroup_tree = RB_ROOT; | |
f9e92e40 ES |
2315 | INIT_LIST_HEAD(&fs_info->dirty_qgroups); |
2316 | fs_info->qgroup_seq = 1; | |
f9e92e40 | 2317 | fs_info->qgroup_ulist = NULL; |
d2c609b8 | 2318 | fs_info->qgroup_rescan_running = false; |
011b46c3 | 2319 | fs_info->qgroup_drop_subtree_thres = BTRFS_MAX_LEVEL; |
f9e92e40 ES |
2320 | mutex_init(&fs_info->qgroup_rescan_lock); |
2321 | } | |
2322 | ||
d21deec5 | 2323 | static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info) |
2a458198 | 2324 | { |
f7b885be | 2325 | u32 max_active = fs_info->thread_pool_size; |
6f011058 | 2326 | unsigned int flags = WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND; |
2a458198 ES |
2327 | |
2328 | fs_info->workers = | |
a31b4a43 CH |
2329 | btrfs_alloc_workqueue(fs_info, "worker", flags, max_active, 16); |
2330 | fs_info->hipri_workers = | |
2331 | btrfs_alloc_workqueue(fs_info, "worker-high", | |
cb001095 | 2332 | flags | WQ_HIGHPRI, max_active, 16); |
2a458198 ES |
2333 | |
2334 | fs_info->delalloc_workers = | |
cb001095 JM |
2335 | btrfs_alloc_workqueue(fs_info, "delalloc", |
2336 | flags, max_active, 2); | |
2a458198 ES |
2337 | |
2338 | fs_info->flush_workers = | |
cb001095 JM |
2339 | btrfs_alloc_workqueue(fs_info, "flush_delalloc", |
2340 | flags, max_active, 0); | |
2a458198 ES |
2341 | |
2342 | fs_info->caching_workers = | |
cb001095 | 2343 | btrfs_alloc_workqueue(fs_info, "cache", flags, max_active, 0); |
2a458198 | 2344 | |
2a458198 | 2345 | fs_info->fixup_workers = |
cb001095 | 2346 | btrfs_alloc_workqueue(fs_info, "fixup", flags, 1, 0); |
2a458198 | 2347 | |
2a458198 | 2348 | fs_info->endio_workers = |
d7b9416f | 2349 | alloc_workqueue("btrfs-endio", flags, max_active); |
2a458198 | 2350 | fs_info->endio_meta_workers = |
d7b9416f | 2351 | alloc_workqueue("btrfs-endio-meta", flags, max_active); |
385de0ef | 2352 | fs_info->rmw_workers = alloc_workqueue("btrfs-rmw", flags, max_active); |
2a458198 | 2353 | fs_info->endio_write_workers = |
cb001095 JM |
2354 | btrfs_alloc_workqueue(fs_info, "endio-write", flags, |
2355 | max_active, 2); | |
fed8a72d CH |
2356 | fs_info->compressed_write_workers = |
2357 | alloc_workqueue("btrfs-compressed-write", flags, max_active); | |
2a458198 | 2358 | fs_info->endio_freespace_worker = |
cb001095 JM |
2359 | btrfs_alloc_workqueue(fs_info, "freespace-write", flags, |
2360 | max_active, 0); | |
2a458198 | 2361 | fs_info->delayed_workers = |
cb001095 JM |
2362 | btrfs_alloc_workqueue(fs_info, "delayed-meta", flags, |
2363 | max_active, 0); | |
2a458198 | 2364 | fs_info->qgroup_rescan_workers = |
cb001095 | 2365 | btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); |
b0643e59 DZ |
2366 | fs_info->discard_ctl.discard_workers = |
2367 | alloc_workqueue("btrfs_discard", WQ_UNBOUND | WQ_FREEZABLE, 1); | |
2a458198 | 2368 | |
a31b4a43 CH |
2369 | if (!(fs_info->workers && fs_info->hipri_workers && |
2370 | fs_info->delalloc_workers && fs_info->flush_workers && | |
2a458198 | 2371 | fs_info->endio_workers && fs_info->endio_meta_workers && |
fed8a72d | 2372 | fs_info->compressed_write_workers && |
1a1a2851 | 2373 | fs_info->endio_write_workers && |
2a458198 | 2374 | fs_info->endio_freespace_worker && fs_info->rmw_workers && |
f26c9238 QW |
2375 | fs_info->caching_workers && fs_info->fixup_workers && |
2376 | fs_info->delayed_workers && fs_info->qgroup_rescan_workers && | |
b0643e59 | 2377 | fs_info->discard_ctl.discard_workers)) { |
2a458198 ES |
2378 | return -ENOMEM; |
2379 | } | |
2380 | ||
2381 | return 0; | |
2382 | } | |
2383 | ||
6d97c6e3 JT |
2384 | static int btrfs_init_csum_hash(struct btrfs_fs_info *fs_info, u16 csum_type) |
2385 | { | |
2386 | struct crypto_shash *csum_shash; | |
b4e967be | 2387 | const char *csum_driver = btrfs_super_csum_driver(csum_type); |
6d97c6e3 | 2388 | |
b4e967be | 2389 | csum_shash = crypto_alloc_shash(csum_driver, 0, 0); |
6d97c6e3 JT |
2390 | |
2391 | if (IS_ERR(csum_shash)) { | |
2392 | btrfs_err(fs_info, "error allocating %s hash for checksum", | |
b4e967be | 2393 | csum_driver); |
6d97c6e3 JT |
2394 | return PTR_ERR(csum_shash); |
2395 | } | |
2396 | ||
2397 | fs_info->csum_shash = csum_shash; | |
2398 | ||
c8a5f8ca DS |
2399 | btrfs_info(fs_info, "using %s (%s) checksum algorithm", |
2400 | btrfs_super_csum_name(csum_type), | |
2401 | crypto_shash_driver_name(csum_shash)); | |
6d97c6e3 JT |
2402 | return 0; |
2403 | } | |
2404 | ||
63443bf5 ES |
2405 | static int btrfs_replay_log(struct btrfs_fs_info *fs_info, |
2406 | struct btrfs_fs_devices *fs_devices) | |
2407 | { | |
2408 | int ret; | |
789d6a3a | 2409 | struct btrfs_tree_parent_check check = { 0 }; |
63443bf5 ES |
2410 | struct btrfs_root *log_tree_root; |
2411 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
2412 | u64 bytenr = btrfs_super_log_root(disk_super); | |
581c1760 | 2413 | int level = btrfs_super_log_root_level(disk_super); |
63443bf5 ES |
2414 | |
2415 | if (fs_devices->rw_devices == 0) { | |
f14d104d | 2416 | btrfs_warn(fs_info, "log replay required on RO media"); |
63443bf5 ES |
2417 | return -EIO; |
2418 | } | |
2419 | ||
96dfcb46 JB |
2420 | log_tree_root = btrfs_alloc_root(fs_info, BTRFS_TREE_LOG_OBJECTID, |
2421 | GFP_KERNEL); | |
63443bf5 ES |
2422 | if (!log_tree_root) |
2423 | return -ENOMEM; | |
2424 | ||
789d6a3a QW |
2425 | check.level = level; |
2426 | check.transid = fs_info->generation + 1; | |
2427 | check.owner_root = BTRFS_TREE_LOG_OBJECTID; | |
2428 | log_tree_root->node = read_tree_block(fs_info, bytenr, &check); | |
64c043de | 2429 | if (IS_ERR(log_tree_root->node)) { |
f14d104d | 2430 | btrfs_warn(fs_info, "failed to read log tree"); |
0eeff236 | 2431 | ret = PTR_ERR(log_tree_root->node); |
8c38938c | 2432 | log_tree_root->node = NULL; |
00246528 | 2433 | btrfs_put_root(log_tree_root); |
0eeff236 | 2434 | return ret; |
4eb150d6 QW |
2435 | } |
2436 | if (!extent_buffer_uptodate(log_tree_root->node)) { | |
f14d104d | 2437 | btrfs_err(fs_info, "failed to read log tree"); |
00246528 | 2438 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2439 | return -EIO; |
2440 | } | |
4eb150d6 | 2441 | |
63443bf5 ES |
2442 | /* returns with log_tree_root freed on success */ |
2443 | ret = btrfs_recover_log_trees(log_tree_root); | |
2444 | if (ret) { | |
0b246afa JM |
2445 | btrfs_handle_fs_error(fs_info, ret, |
2446 | "Failed to recover log tree"); | |
00246528 | 2447 | btrfs_put_root(log_tree_root); |
63443bf5 ES |
2448 | return ret; |
2449 | } | |
2450 | ||
bc98a42c | 2451 | if (sb_rdonly(fs_info->sb)) { |
6bccf3ab | 2452 | ret = btrfs_commit_super(fs_info); |
63443bf5 ES |
2453 | if (ret) |
2454 | return ret; | |
2455 | } | |
2456 | ||
2457 | return 0; | |
2458 | } | |
2459 | ||
abed4aaa JB |
2460 | static int load_global_roots_objectid(struct btrfs_root *tree_root, |
2461 | struct btrfs_path *path, u64 objectid, | |
2462 | const char *name) | |
2463 | { | |
2464 | struct btrfs_fs_info *fs_info = tree_root->fs_info; | |
2465 | struct btrfs_root *root; | |
f7238e50 | 2466 | u64 max_global_id = 0; |
abed4aaa JB |
2467 | int ret; |
2468 | struct btrfs_key key = { | |
2469 | .objectid = objectid, | |
2470 | .type = BTRFS_ROOT_ITEM_KEY, | |
2471 | .offset = 0, | |
2472 | }; | |
2473 | bool found = false; | |
2474 | ||
2475 | /* If we have IGNOREDATACSUMS skip loading these roots. */ | |
2476 | if (objectid == BTRFS_CSUM_TREE_OBJECTID && | |
2477 | btrfs_test_opt(fs_info, IGNOREDATACSUMS)) { | |
2478 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2479 | return 0; | |
2480 | } | |
2481 | ||
2482 | while (1) { | |
2483 | ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); | |
2484 | if (ret < 0) | |
2485 | break; | |
2486 | ||
2487 | if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { | |
2488 | ret = btrfs_next_leaf(tree_root, path); | |
2489 | if (ret) { | |
2490 | if (ret > 0) | |
2491 | ret = 0; | |
2492 | break; | |
2493 | } | |
2494 | } | |
2495 | ret = 0; | |
2496 | ||
2497 | btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]); | |
2498 | if (key.objectid != objectid) | |
2499 | break; | |
2500 | btrfs_release_path(path); | |
2501 | ||
f7238e50 JB |
2502 | /* |
2503 | * Just worry about this for extent tree, it'll be the same for | |
2504 | * everybody. | |
2505 | */ | |
2506 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) | |
2507 | max_global_id = max(max_global_id, key.offset); | |
2508 | ||
abed4aaa JB |
2509 | found = true; |
2510 | root = read_tree_root_path(tree_root, path, &key); | |
2511 | if (IS_ERR(root)) { | |
2512 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2513 | ret = PTR_ERR(root); | |
2514 | break; | |
2515 | } | |
2516 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2517 | ret = btrfs_global_root_insert(root); | |
2518 | if (ret) { | |
2519 | btrfs_put_root(root); | |
2520 | break; | |
2521 | } | |
2522 | key.offset++; | |
2523 | } | |
2524 | btrfs_release_path(path); | |
2525 | ||
f7238e50 JB |
2526 | if (objectid == BTRFS_EXTENT_TREE_OBJECTID) |
2527 | fs_info->nr_global_roots = max_global_id + 1; | |
2528 | ||
abed4aaa JB |
2529 | if (!found || ret) { |
2530 | if (objectid == BTRFS_CSUM_TREE_OBJECTID) | |
2531 | set_bit(BTRFS_FS_STATE_NO_CSUMS, &fs_info->fs_state); | |
2532 | ||
2533 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) | |
2534 | ret = ret ? ret : -ENOENT; | |
2535 | else | |
2536 | ret = 0; | |
2537 | btrfs_err(fs_info, "failed to load root %s", name); | |
2538 | } | |
2539 | return ret; | |
2540 | } | |
2541 | ||
2542 | static int load_global_roots(struct btrfs_root *tree_root) | |
2543 | { | |
2544 | struct btrfs_path *path; | |
2545 | int ret = 0; | |
2546 | ||
2547 | path = btrfs_alloc_path(); | |
2548 | if (!path) | |
2549 | return -ENOMEM; | |
2550 | ||
2551 | ret = load_global_roots_objectid(tree_root, path, | |
2552 | BTRFS_EXTENT_TREE_OBJECTID, "extent"); | |
2553 | if (ret) | |
2554 | goto out; | |
2555 | ret = load_global_roots_objectid(tree_root, path, | |
2556 | BTRFS_CSUM_TREE_OBJECTID, "csum"); | |
2557 | if (ret) | |
2558 | goto out; | |
2559 | if (!btrfs_fs_compat_ro(tree_root->fs_info, FREE_SPACE_TREE)) | |
2560 | goto out; | |
2561 | ret = load_global_roots_objectid(tree_root, path, | |
2562 | BTRFS_FREE_SPACE_TREE_OBJECTID, | |
2563 | "free space"); | |
2564 | out: | |
2565 | btrfs_free_path(path); | |
2566 | return ret; | |
2567 | } | |
2568 | ||
6bccf3ab | 2569 | static int btrfs_read_roots(struct btrfs_fs_info *fs_info) |
4bbcaa64 | 2570 | { |
6bccf3ab | 2571 | struct btrfs_root *tree_root = fs_info->tree_root; |
a4f3d2c4 | 2572 | struct btrfs_root *root; |
4bbcaa64 ES |
2573 | struct btrfs_key location; |
2574 | int ret; | |
2575 | ||
6bccf3ab JM |
2576 | BUG_ON(!fs_info->tree_root); |
2577 | ||
abed4aaa JB |
2578 | ret = load_global_roots(tree_root); |
2579 | if (ret) | |
2580 | return ret; | |
2581 | ||
4bbcaa64 ES |
2582 | location.type = BTRFS_ROOT_ITEM_KEY; |
2583 | location.offset = 0; | |
2584 | ||
1c56ab99 | 2585 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE)) { |
14033b08 QW |
2586 | location.objectid = BTRFS_BLOCK_GROUP_TREE_OBJECTID; |
2587 | root = btrfs_read_tree_root(tree_root, &location); | |
2588 | if (IS_ERR(root)) { | |
2589 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { | |
2590 | ret = PTR_ERR(root); | |
2591 | goto out; | |
2592 | } | |
2593 | } else { | |
2594 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2595 | fs_info->block_group_root = root; | |
2596 | } | |
2597 | } | |
2598 | ||
2599 | location.objectid = BTRFS_DEV_TREE_OBJECTID; | |
a4f3d2c4 | 2600 | root = btrfs_read_tree_root(tree_root, &location); |
f50f4353 | 2601 | if (IS_ERR(root)) { |
42437a63 JB |
2602 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2603 | ret = PTR_ERR(root); | |
2604 | goto out; | |
2605 | } | |
2606 | } else { | |
2607 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2608 | fs_info->dev_root = root; | |
f50f4353 | 2609 | } |
820a49da | 2610 | /* Initialize fs_info for all devices in any case */ |
a8d1b164 JT |
2611 | ret = btrfs_init_devices_late(fs_info); |
2612 | if (ret) | |
2613 | goto out; | |
4bbcaa64 | 2614 | |
aeb935a4 QW |
2615 | /* |
2616 | * This tree can share blocks with some other fs tree during relocation | |
2617 | * and we need a proper setup by btrfs_get_fs_root | |
2618 | */ | |
56e9357a DS |
2619 | root = btrfs_get_fs_root(tree_root->fs_info, |
2620 | BTRFS_DATA_RELOC_TREE_OBJECTID, true); | |
aeb935a4 | 2621 | if (IS_ERR(root)) { |
42437a63 JB |
2622 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2623 | ret = PTR_ERR(root); | |
2624 | goto out; | |
2625 | } | |
2626 | } else { | |
2627 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
2628 | fs_info->data_reloc_root = root; | |
aeb935a4 | 2629 | } |
aeb935a4 | 2630 | |
4bbcaa64 | 2631 | location.objectid = BTRFS_QUOTA_TREE_OBJECTID; |
a4f3d2c4 DS |
2632 | root = btrfs_read_tree_root(tree_root, &location); |
2633 | if (!IS_ERR(root)) { | |
2634 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); | |
afcdd129 | 2635 | set_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags); |
a4f3d2c4 | 2636 | fs_info->quota_root = root; |
4bbcaa64 ES |
2637 | } |
2638 | ||
2639 | location.objectid = BTRFS_UUID_TREE_OBJECTID; | |
a4f3d2c4 DS |
2640 | root = btrfs_read_tree_root(tree_root, &location); |
2641 | if (IS_ERR(root)) { | |
42437a63 JB |
2642 | if (!btrfs_test_opt(fs_info, IGNOREBADROOTS)) { |
2643 | ret = PTR_ERR(root); | |
2644 | if (ret != -ENOENT) | |
2645 | goto out; | |
2646 | } | |
4bbcaa64 | 2647 | } else { |
a4f3d2c4 DS |
2648 | set_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state); |
2649 | fs_info->uuid_root = root; | |
4bbcaa64 ES |
2650 | } |
2651 | ||
2652 | return 0; | |
f50f4353 LB |
2653 | out: |
2654 | btrfs_warn(fs_info, "failed to read root (objectid=%llu): %d", | |
2655 | location.objectid, ret); | |
2656 | return ret; | |
4bbcaa64 ES |
2657 | } |
2658 | ||
069ec957 QW |
2659 | /* |
2660 | * Real super block validation | |
2661 | * NOTE: super csum type and incompat features will not be checked here. | |
2662 | * | |
2663 | * @sb: super block to check | |
2664 | * @mirror_num: the super block number to check its bytenr: | |
2665 | * 0 the primary (1st) sb | |
2666 | * 1, 2 2nd and 3rd backup copy | |
2667 | * -1 skip bytenr check | |
2668 | */ | |
a05d3c91 QW |
2669 | int btrfs_validate_super(struct btrfs_fs_info *fs_info, |
2670 | struct btrfs_super_block *sb, int mirror_num) | |
21a852b0 | 2671 | { |
21a852b0 QW |
2672 | u64 nodesize = btrfs_super_nodesize(sb); |
2673 | u64 sectorsize = btrfs_super_sectorsize(sb); | |
2674 | int ret = 0; | |
2675 | ||
2676 | if (btrfs_super_magic(sb) != BTRFS_MAGIC) { | |
2677 | btrfs_err(fs_info, "no valid FS found"); | |
2678 | ret = -EINVAL; | |
2679 | } | |
2680 | if (btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP) { | |
2681 | btrfs_err(fs_info, "unrecognized or unsupported super flag: %llu", | |
2682 | btrfs_super_flags(sb) & ~BTRFS_SUPER_FLAG_SUPP); | |
2683 | ret = -EINVAL; | |
2684 | } | |
2685 | if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2686 | btrfs_err(fs_info, "tree_root level too big: %d >= %d", | |
2687 | btrfs_super_root_level(sb), BTRFS_MAX_LEVEL); | |
2688 | ret = -EINVAL; | |
2689 | } | |
2690 | if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2691 | btrfs_err(fs_info, "chunk_root level too big: %d >= %d", | |
2692 | btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL); | |
2693 | ret = -EINVAL; | |
2694 | } | |
2695 | if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) { | |
2696 | btrfs_err(fs_info, "log_root level too big: %d >= %d", | |
2697 | btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL); | |
2698 | ret = -EINVAL; | |
2699 | } | |
2700 | ||
2701 | /* | |
2702 | * Check sectorsize and nodesize first, other check will need it. | |
2703 | * Check all possible sectorsize(4K, 8K, 16K, 32K, 64K) here. | |
2704 | */ | |
2705 | if (!is_power_of_2(sectorsize) || sectorsize < 4096 || | |
2706 | sectorsize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2707 | btrfs_err(fs_info, "invalid sectorsize %llu", sectorsize); | |
2708 | ret = -EINVAL; | |
2709 | } | |
0bb3eb3e QW |
2710 | |
2711 | /* | |
1a42daab QW |
2712 | * We only support at most two sectorsizes: 4K and PAGE_SIZE. |
2713 | * | |
2714 | * We can support 16K sectorsize with 64K page size without problem, | |
2715 | * but such sectorsize/pagesize combination doesn't make much sense. | |
2716 | * 4K will be our future standard, PAGE_SIZE is supported from the very | |
2717 | * beginning. | |
0bb3eb3e | 2718 | */ |
1a42daab | 2719 | if (sectorsize > PAGE_SIZE || (sectorsize != SZ_4K && sectorsize != PAGE_SIZE)) { |
21a852b0 | 2720 | btrfs_err(fs_info, |
0bb3eb3e | 2721 | "sectorsize %llu not yet supported for page size %lu", |
21a852b0 QW |
2722 | sectorsize, PAGE_SIZE); |
2723 | ret = -EINVAL; | |
2724 | } | |
0bb3eb3e | 2725 | |
21a852b0 QW |
2726 | if (!is_power_of_2(nodesize) || nodesize < sectorsize || |
2727 | nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) { | |
2728 | btrfs_err(fs_info, "invalid nodesize %llu", nodesize); | |
2729 | ret = -EINVAL; | |
2730 | } | |
2731 | if (nodesize != le32_to_cpu(sb->__unused_leafsize)) { | |
2732 | btrfs_err(fs_info, "invalid leafsize %u, should be %llu", | |
2733 | le32_to_cpu(sb->__unused_leafsize), nodesize); | |
2734 | ret = -EINVAL; | |
2735 | } | |
2736 | ||
2737 | /* Root alignment check */ | |
2738 | if (!IS_ALIGNED(btrfs_super_root(sb), sectorsize)) { | |
2739 | btrfs_warn(fs_info, "tree_root block unaligned: %llu", | |
2740 | btrfs_super_root(sb)); | |
2741 | ret = -EINVAL; | |
2742 | } | |
2743 | if (!IS_ALIGNED(btrfs_super_chunk_root(sb), sectorsize)) { | |
2744 | btrfs_warn(fs_info, "chunk_root block unaligned: %llu", | |
2745 | btrfs_super_chunk_root(sb)); | |
2746 | ret = -EINVAL; | |
2747 | } | |
2748 | if (!IS_ALIGNED(btrfs_super_log_root(sb), sectorsize)) { | |
2749 | btrfs_warn(fs_info, "log_root block unaligned: %llu", | |
2750 | btrfs_super_log_root(sb)); | |
2751 | ret = -EINVAL; | |
2752 | } | |
2753 | ||
aefd7f70 NB |
2754 | if (memcmp(fs_info->fs_devices->fsid, fs_info->super_copy->fsid, |
2755 | BTRFS_FSID_SIZE)) { | |
2756 | btrfs_err(fs_info, | |
2757 | "superblock fsid doesn't match fsid of fs_devices: %pU != %pU", | |
2758 | fs_info->super_copy->fsid, fs_info->fs_devices->fsid); | |
2759 | ret = -EINVAL; | |
2760 | } | |
2761 | ||
2762 | if (btrfs_fs_incompat(fs_info, METADATA_UUID) && | |
2763 | memcmp(fs_info->fs_devices->metadata_uuid, | |
2764 | fs_info->super_copy->metadata_uuid, BTRFS_FSID_SIZE)) { | |
2765 | btrfs_err(fs_info, | |
2766 | "superblock metadata_uuid doesn't match metadata uuid of fs_devices: %pU != %pU", | |
2767 | fs_info->super_copy->metadata_uuid, | |
2768 | fs_info->fs_devices->metadata_uuid); | |
2769 | ret = -EINVAL; | |
2770 | } | |
2771 | ||
1c56ab99 QW |
2772 | /* |
2773 | * Artificial requirement for block-group-tree to force newer features | |
2774 | * (free-space-tree, no-holes) so the test matrix is smaller. | |
2775 | */ | |
2776 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
2777 | (!btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID) || | |
2778 | !btrfs_fs_incompat(fs_info, NO_HOLES))) { | |
2779 | btrfs_err(fs_info, | |
2780 | "block-group-tree feature requires fres-space-tree and no-holes"); | |
2781 | ret = -EINVAL; | |
2782 | } | |
2783 | ||
de37aa51 | 2784 | if (memcmp(fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid, |
7239ff4b | 2785 | BTRFS_FSID_SIZE) != 0) { |
21a852b0 | 2786 | btrfs_err(fs_info, |
7239ff4b | 2787 | "dev_item UUID does not match metadata fsid: %pU != %pU", |
de37aa51 | 2788 | fs_info->fs_devices->metadata_uuid, sb->dev_item.fsid); |
21a852b0 QW |
2789 | ret = -EINVAL; |
2790 | } | |
2791 | ||
2792 | /* | |
2793 | * Hint to catch really bogus numbers, bitflips or so, more exact checks are | |
2794 | * done later | |
2795 | */ | |
2796 | if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) { | |
2797 | btrfs_err(fs_info, "bytes_used is too small %llu", | |
2798 | btrfs_super_bytes_used(sb)); | |
2799 | ret = -EINVAL; | |
2800 | } | |
2801 | if (!is_power_of_2(btrfs_super_stripesize(sb))) { | |
2802 | btrfs_err(fs_info, "invalid stripesize %u", | |
2803 | btrfs_super_stripesize(sb)); | |
2804 | ret = -EINVAL; | |
2805 | } | |
2806 | if (btrfs_super_num_devices(sb) > (1UL << 31)) | |
2807 | btrfs_warn(fs_info, "suspicious number of devices: %llu", | |
2808 | btrfs_super_num_devices(sb)); | |
2809 | if (btrfs_super_num_devices(sb) == 0) { | |
2810 | btrfs_err(fs_info, "number of devices is 0"); | |
2811 | ret = -EINVAL; | |
2812 | } | |
2813 | ||
069ec957 QW |
2814 | if (mirror_num >= 0 && |
2815 | btrfs_super_bytenr(sb) != btrfs_sb_offset(mirror_num)) { | |
21a852b0 QW |
2816 | btrfs_err(fs_info, "super offset mismatch %llu != %u", |
2817 | btrfs_super_bytenr(sb), BTRFS_SUPER_INFO_OFFSET); | |
2818 | ret = -EINVAL; | |
2819 | } | |
2820 | ||
2821 | /* | |
2822 | * Obvious sys_chunk_array corruptions, it must hold at least one key | |
2823 | * and one chunk | |
2824 | */ | |
2825 | if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) { | |
2826 | btrfs_err(fs_info, "system chunk array too big %u > %u", | |
2827 | btrfs_super_sys_array_size(sb), | |
2828 | BTRFS_SYSTEM_CHUNK_ARRAY_SIZE); | |
2829 | ret = -EINVAL; | |
2830 | } | |
2831 | if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key) | |
2832 | + sizeof(struct btrfs_chunk)) { | |
2833 | btrfs_err(fs_info, "system chunk array too small %u < %zu", | |
2834 | btrfs_super_sys_array_size(sb), | |
2835 | sizeof(struct btrfs_disk_key) | |
2836 | + sizeof(struct btrfs_chunk)); | |
2837 | ret = -EINVAL; | |
2838 | } | |
2839 | ||
2840 | /* | |
2841 | * The generation is a global counter, we'll trust it more than the others | |
2842 | * but it's still possible that it's the one that's wrong. | |
2843 | */ | |
2844 | if (btrfs_super_generation(sb) < btrfs_super_chunk_root_generation(sb)) | |
2845 | btrfs_warn(fs_info, | |
2846 | "suspicious: generation < chunk_root_generation: %llu < %llu", | |
2847 | btrfs_super_generation(sb), | |
2848 | btrfs_super_chunk_root_generation(sb)); | |
2849 | if (btrfs_super_generation(sb) < btrfs_super_cache_generation(sb) | |
2850 | && btrfs_super_cache_generation(sb) != (u64)-1) | |
2851 | btrfs_warn(fs_info, | |
2852 | "suspicious: generation < cache_generation: %llu < %llu", | |
2853 | btrfs_super_generation(sb), | |
2854 | btrfs_super_cache_generation(sb)); | |
2855 | ||
2856 | return ret; | |
2857 | } | |
2858 | ||
069ec957 QW |
2859 | /* |
2860 | * Validation of super block at mount time. | |
2861 | * Some checks already done early at mount time, like csum type and incompat | |
2862 | * flags will be skipped. | |
2863 | */ | |
2864 | static int btrfs_validate_mount_super(struct btrfs_fs_info *fs_info) | |
2865 | { | |
a05d3c91 | 2866 | return btrfs_validate_super(fs_info, fs_info->super_copy, 0); |
069ec957 QW |
2867 | } |
2868 | ||
75cb857d QW |
2869 | /* |
2870 | * Validation of super block at write time. | |
2871 | * Some checks like bytenr check will be skipped as their values will be | |
2872 | * overwritten soon. | |
2873 | * Extra checks like csum type and incompat flags will be done here. | |
2874 | */ | |
2875 | static int btrfs_validate_write_super(struct btrfs_fs_info *fs_info, | |
2876 | struct btrfs_super_block *sb) | |
2877 | { | |
2878 | int ret; | |
2879 | ||
a05d3c91 | 2880 | ret = btrfs_validate_super(fs_info, sb, -1); |
75cb857d QW |
2881 | if (ret < 0) |
2882 | goto out; | |
e7e16f48 | 2883 | if (!btrfs_supported_super_csum(btrfs_super_csum_type(sb))) { |
75cb857d QW |
2884 | ret = -EUCLEAN; |
2885 | btrfs_err(fs_info, "invalid csum type, has %u want %u", | |
2886 | btrfs_super_csum_type(sb), BTRFS_CSUM_TYPE_CRC32); | |
2887 | goto out; | |
2888 | } | |
2889 | if (btrfs_super_incompat_flags(sb) & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
2890 | ret = -EUCLEAN; | |
2891 | btrfs_err(fs_info, | |
2892 | "invalid incompat flags, has 0x%llx valid mask 0x%llx", | |
2893 | btrfs_super_incompat_flags(sb), | |
2894 | (unsigned long long)BTRFS_FEATURE_INCOMPAT_SUPP); | |
2895 | goto out; | |
2896 | } | |
2897 | out: | |
2898 | if (ret < 0) | |
2899 | btrfs_err(fs_info, | |
2900 | "super block corruption detected before writing it to disk"); | |
2901 | return ret; | |
2902 | } | |
2903 | ||
bd676446 JB |
2904 | static int load_super_root(struct btrfs_root *root, u64 bytenr, u64 gen, int level) |
2905 | { | |
789d6a3a QW |
2906 | struct btrfs_tree_parent_check check = { |
2907 | .level = level, | |
2908 | .transid = gen, | |
2909 | .owner_root = root->root_key.objectid | |
2910 | }; | |
bd676446 JB |
2911 | int ret = 0; |
2912 | ||
789d6a3a | 2913 | root->node = read_tree_block(root->fs_info, bytenr, &check); |
bd676446 JB |
2914 | if (IS_ERR(root->node)) { |
2915 | ret = PTR_ERR(root->node); | |
2916 | root->node = NULL; | |
4eb150d6 QW |
2917 | return ret; |
2918 | } | |
2919 | if (!extent_buffer_uptodate(root->node)) { | |
bd676446 JB |
2920 | free_extent_buffer(root->node); |
2921 | root->node = NULL; | |
4eb150d6 | 2922 | return -EIO; |
bd676446 JB |
2923 | } |
2924 | ||
bd676446 JB |
2925 | btrfs_set_root_node(&root->root_item, root->node); |
2926 | root->commit_root = btrfs_root_node(root); | |
2927 | btrfs_set_root_refs(&root->root_item, 1); | |
2928 | return ret; | |
2929 | } | |
2930 | ||
2931 | static int load_important_roots(struct btrfs_fs_info *fs_info) | |
2932 | { | |
2933 | struct btrfs_super_block *sb = fs_info->super_copy; | |
2934 | u64 gen, bytenr; | |
2935 | int level, ret; | |
2936 | ||
2937 | bytenr = btrfs_super_root(sb); | |
2938 | gen = btrfs_super_generation(sb); | |
2939 | level = btrfs_super_root_level(sb); | |
2940 | ret = load_super_root(fs_info->tree_root, bytenr, gen, level); | |
9c54e80d | 2941 | if (ret) { |
bd676446 | 2942 | btrfs_warn(fs_info, "couldn't read tree root"); |
9c54e80d JB |
2943 | return ret; |
2944 | } | |
14033b08 | 2945 | return 0; |
bd676446 JB |
2946 | } |
2947 | ||
6ef108dd | 2948 | static int __cold init_tree_roots(struct btrfs_fs_info *fs_info) |
b8522a1e | 2949 | { |
6ef108dd | 2950 | int backup_index = find_newest_super_backup(fs_info); |
b8522a1e NB |
2951 | struct btrfs_super_block *sb = fs_info->super_copy; |
2952 | struct btrfs_root *tree_root = fs_info->tree_root; | |
2953 | bool handle_error = false; | |
2954 | int ret = 0; | |
2955 | int i; | |
2956 | ||
2957 | for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) { | |
b8522a1e NB |
2958 | if (handle_error) { |
2959 | if (!IS_ERR(tree_root->node)) | |
2960 | free_extent_buffer(tree_root->node); | |
2961 | tree_root->node = NULL; | |
2962 | ||
2963 | if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) | |
2964 | break; | |
2965 | ||
2966 | free_root_pointers(fs_info, 0); | |
2967 | ||
2968 | /* | |
2969 | * Don't use the log in recovery mode, it won't be | |
2970 | * valid | |
2971 | */ | |
2972 | btrfs_set_super_log_root(sb, 0); | |
2973 | ||
2974 | /* We can't trust the free space cache either */ | |
2975 | btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE); | |
2976 | ||
2977 | ret = read_backup_root(fs_info, i); | |
6ef108dd | 2978 | backup_index = ret; |
b8522a1e NB |
2979 | if (ret < 0) |
2980 | return ret; | |
2981 | } | |
b8522a1e | 2982 | |
bd676446 JB |
2983 | ret = load_important_roots(fs_info); |
2984 | if (ret) { | |
217f5004 | 2985 | handle_error = true; |
b8522a1e NB |
2986 | continue; |
2987 | } | |
2988 | ||
336a0d8d NB |
2989 | /* |
2990 | * No need to hold btrfs_root::objectid_mutex since the fs | |
2991 | * hasn't been fully initialised and we are the only user | |
2992 | */ | |
453e4873 | 2993 | ret = btrfs_init_root_free_objectid(tree_root); |
b8522a1e | 2994 | if (ret < 0) { |
b8522a1e NB |
2995 | handle_error = true; |
2996 | continue; | |
2997 | } | |
2998 | ||
6b8fad57 | 2999 | ASSERT(tree_root->free_objectid <= BTRFS_LAST_FREE_OBJECTID); |
b8522a1e NB |
3000 | |
3001 | ret = btrfs_read_roots(fs_info); | |
3002 | if (ret < 0) { | |
3003 | handle_error = true; | |
3004 | continue; | |
3005 | } | |
3006 | ||
3007 | /* All successful */ | |
bd676446 JB |
3008 | fs_info->generation = btrfs_header_generation(tree_root->node); |
3009 | fs_info->last_trans_committed = fs_info->generation; | |
d96b3424 | 3010 | fs_info->last_reloc_trans = 0; |
6ef108dd NB |
3011 | |
3012 | /* Always begin writing backup roots after the one being used */ | |
3013 | if (backup_index < 0) { | |
3014 | fs_info->backup_root_index = 0; | |
3015 | } else { | |
3016 | fs_info->backup_root_index = backup_index + 1; | |
3017 | fs_info->backup_root_index %= BTRFS_NUM_BACKUP_ROOTS; | |
3018 | } | |
b8522a1e NB |
3019 | break; |
3020 | } | |
3021 | ||
3022 | return ret; | |
3023 | } | |
3024 | ||
8260edba | 3025 | void btrfs_init_fs_info(struct btrfs_fs_info *fs_info) |
2e635a27 | 3026 | { |
fc7cbcd4 | 3027 | INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_ATOMIC); |
01cd3909 | 3028 | INIT_RADIX_TREE(&fs_info->buffer_radix, GFP_ATOMIC); |
8fd17795 | 3029 | INIT_LIST_HEAD(&fs_info->trans_list); |
facda1e7 | 3030 | INIT_LIST_HEAD(&fs_info->dead_roots); |
24bbcf04 | 3031 | INIT_LIST_HEAD(&fs_info->delayed_iputs); |
eb73c1b7 | 3032 | INIT_LIST_HEAD(&fs_info->delalloc_roots); |
11833d66 | 3033 | INIT_LIST_HEAD(&fs_info->caching_block_groups); |
eb73c1b7 | 3034 | spin_lock_init(&fs_info->delalloc_root_lock); |
a4abeea4 | 3035 | spin_lock_init(&fs_info->trans_lock); |
fc7cbcd4 | 3036 | spin_lock_init(&fs_info->fs_roots_radix_lock); |
24bbcf04 | 3037 | spin_lock_init(&fs_info->delayed_iput_lock); |
4cb5300b | 3038 | spin_lock_init(&fs_info->defrag_inodes_lock); |
ceda0864 | 3039 | spin_lock_init(&fs_info->super_lock); |
f28491e0 | 3040 | spin_lock_init(&fs_info->buffer_lock); |
47ab2a6c | 3041 | spin_lock_init(&fs_info->unused_bgs_lock); |
40ab3be1 | 3042 | spin_lock_init(&fs_info->treelog_bg_lock); |
afba2bc0 | 3043 | spin_lock_init(&fs_info->zone_active_bgs_lock); |
c2707a25 | 3044 | spin_lock_init(&fs_info->relocation_bg_lock); |
f29021b2 | 3045 | rwlock_init(&fs_info->tree_mod_log_lock); |
abed4aaa | 3046 | rwlock_init(&fs_info->global_root_lock); |
d7c15171 | 3047 | mutex_init(&fs_info->unused_bg_unpin_mutex); |
f3372065 | 3048 | mutex_init(&fs_info->reclaim_bgs_lock); |
7585717f | 3049 | mutex_init(&fs_info->reloc_mutex); |
573bfb72 | 3050 | mutex_init(&fs_info->delalloc_root_mutex); |
0bc09ca1 | 3051 | mutex_init(&fs_info->zoned_meta_io_lock); |
5f0addf7 | 3052 | mutex_init(&fs_info->zoned_data_reloc_io_lock); |
de98ced9 | 3053 | seqlock_init(&fs_info->profiles_lock); |
19c00ddc | 3054 | |
e1489b4f | 3055 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_writers); |
5a9ba670 | 3056 | btrfs_lockdep_init_map(fs_info, btrfs_trans_num_extwriters); |
8b53779e | 3057 | btrfs_lockdep_init_map(fs_info, btrfs_trans_pending_ordered); |
5f4403e1 | 3058 | btrfs_lockdep_init_map(fs_info, btrfs_ordered_extent); |
3e738c53 IA |
3059 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_commit_start, |
3060 | BTRFS_LOCKDEP_TRANS_COMMIT_START); | |
3061 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_unblocked, | |
3062 | BTRFS_LOCKDEP_TRANS_UNBLOCKED); | |
3063 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_super_committed, | |
3064 | BTRFS_LOCKDEP_TRANS_SUPER_COMMITTED); | |
3065 | btrfs_state_lockdep_init_map(fs_info, btrfs_trans_completed, | |
3066 | BTRFS_LOCKDEP_TRANS_COMPLETED); | |
e1489b4f | 3067 | |
0b86a832 | 3068 | INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots); |
6324fbf3 | 3069 | INIT_LIST_HEAD(&fs_info->space_info); |
f29021b2 | 3070 | INIT_LIST_HEAD(&fs_info->tree_mod_seq_list); |
47ab2a6c | 3071 | INIT_LIST_HEAD(&fs_info->unused_bgs); |
18bb8bbf | 3072 | INIT_LIST_HEAD(&fs_info->reclaim_bgs); |
afba2bc0 | 3073 | INIT_LIST_HEAD(&fs_info->zone_active_bgs); |
bd647ce3 JB |
3074 | #ifdef CONFIG_BTRFS_DEBUG |
3075 | INIT_LIST_HEAD(&fs_info->allocated_roots); | |
3fd63727 JB |
3076 | INIT_LIST_HEAD(&fs_info->allocated_ebs); |
3077 | spin_lock_init(&fs_info->eb_leak_lock); | |
bd647ce3 | 3078 | #endif |
c8bf1b67 | 3079 | extent_map_tree_init(&fs_info->mapping_tree); |
66d8f3dd MX |
3080 | btrfs_init_block_rsv(&fs_info->global_block_rsv, |
3081 | BTRFS_BLOCK_RSV_GLOBAL); | |
66d8f3dd MX |
3082 | btrfs_init_block_rsv(&fs_info->trans_block_rsv, BTRFS_BLOCK_RSV_TRANS); |
3083 | btrfs_init_block_rsv(&fs_info->chunk_block_rsv, BTRFS_BLOCK_RSV_CHUNK); | |
3084 | btrfs_init_block_rsv(&fs_info->empty_block_rsv, BTRFS_BLOCK_RSV_EMPTY); | |
3085 | btrfs_init_block_rsv(&fs_info->delayed_block_rsv, | |
3086 | BTRFS_BLOCK_RSV_DELOPS); | |
ba2c4d4e JB |
3087 | btrfs_init_block_rsv(&fs_info->delayed_refs_rsv, |
3088 | BTRFS_BLOCK_RSV_DELREFS); | |
3089 | ||
771ed689 | 3090 | atomic_set(&fs_info->async_delalloc_pages, 0); |
4cb5300b | 3091 | atomic_set(&fs_info->defrag_running, 0); |
034f784d | 3092 | atomic_set(&fs_info->nr_delayed_iputs, 0); |
fc36ed7e | 3093 | atomic64_set(&fs_info->tree_mod_seq, 0); |
abed4aaa | 3094 | fs_info->global_root_tree = RB_ROOT; |
95ac567a | 3095 | fs_info->max_inline = BTRFS_DEFAULT_MAX_INLINE; |
9ed74f2d | 3096 | fs_info->metadata_ratio = 0; |
4cb5300b | 3097 | fs_info->defrag_inodes = RB_ROOT; |
a5ed45f8 | 3098 | atomic64_set(&fs_info->free_chunk_space, 0); |
f29021b2 | 3099 | fs_info->tree_mod_log = RB_ROOT; |
8b87dc17 | 3100 | fs_info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; |
f8c269d7 | 3101 | fs_info->avg_delayed_ref_runtime = NSEC_PER_SEC >> 6; /* div by 64 */ |
fd708b81 | 3102 | btrfs_init_ref_verify(fs_info); |
c8b97818 | 3103 | |
b34b086c CM |
3104 | fs_info->thread_pool_size = min_t(unsigned long, |
3105 | num_online_cpus() + 2, 8); | |
0afbaf8c | 3106 | |
199c2a9c MX |
3107 | INIT_LIST_HEAD(&fs_info->ordered_roots); |
3108 | spin_lock_init(&fs_info->ordered_root_lock); | |
69fe2d75 | 3109 | |
638aa7ed | 3110 | btrfs_init_scrub(fs_info); |
21adbd5c SB |
3111 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
3112 | fs_info->check_integrity_print_mask = 0; | |
3113 | #endif | |
779a65a4 | 3114 | btrfs_init_balance(fs_info); |
57056740 | 3115 | btrfs_init_async_reclaim_work(fs_info); |
a2de733c | 3116 | |
16b0c258 | 3117 | rwlock_init(&fs_info->block_group_cache_lock); |
08dddb29 | 3118 | fs_info->block_group_cache_tree = RB_ROOT_CACHED; |
0f9dd46c | 3119 | |
fe119a6e | 3120 | extent_io_tree_init(fs_info, &fs_info->excluded_extents, |
35da5a7e | 3121 | IO_TREE_FS_EXCLUDED_EXTENTS); |
39279cc3 | 3122 | |
5a3f23d5 | 3123 | mutex_init(&fs_info->ordered_operations_mutex); |
e02119d5 | 3124 | mutex_init(&fs_info->tree_log_mutex); |
925baedd | 3125 | mutex_init(&fs_info->chunk_mutex); |
a74a4b97 CM |
3126 | mutex_init(&fs_info->transaction_kthread_mutex); |
3127 | mutex_init(&fs_info->cleaner_mutex); | |
1bbc621e | 3128 | mutex_init(&fs_info->ro_block_group_mutex); |
9e351cc8 | 3129 | init_rwsem(&fs_info->commit_root_sem); |
c71bf099 | 3130 | init_rwsem(&fs_info->cleanup_work_sem); |
76dda93c | 3131 | init_rwsem(&fs_info->subvol_sem); |
803b2f54 | 3132 | sema_init(&fs_info->uuid_tree_rescan_sem, 1); |
fa9c0d79 | 3133 | |
ad618368 | 3134 | btrfs_init_dev_replace_locks(fs_info); |
f9e92e40 | 3135 | btrfs_init_qgroup(fs_info); |
b0643e59 | 3136 | btrfs_discard_init(fs_info); |
416ac51d | 3137 | |
fa9c0d79 CM |
3138 | btrfs_init_free_cluster(&fs_info->meta_alloc_cluster); |
3139 | btrfs_init_free_cluster(&fs_info->data_alloc_cluster); | |
3140 | ||
e6dcd2dc | 3141 | init_waitqueue_head(&fs_info->transaction_throttle); |
f9295749 | 3142 | init_waitqueue_head(&fs_info->transaction_wait); |
bb9c12c9 | 3143 | init_waitqueue_head(&fs_info->transaction_blocked_wait); |
4854ddd0 | 3144 | init_waitqueue_head(&fs_info->async_submit_wait); |
034f784d | 3145 | init_waitqueue_head(&fs_info->delayed_iputs_wait); |
3768f368 | 3146 | |
da17066c JM |
3147 | /* Usable values until the real ones are cached from the superblock */ |
3148 | fs_info->nodesize = 4096; | |
3149 | fs_info->sectorsize = 4096; | |
ab108d99 | 3150 | fs_info->sectorsize_bits = ilog2(4096); |
da17066c JM |
3151 | fs_info->stripesize = 4096; |
3152 | ||
f7b12a62 NA |
3153 | fs_info->max_extent_size = BTRFS_MAX_EXTENT_SIZE; |
3154 | ||
eede2bf3 OS |
3155 | spin_lock_init(&fs_info->swapfile_pins_lock); |
3156 | fs_info->swapfile_pins = RB_ROOT; | |
3157 | ||
18bb8bbf JT |
3158 | fs_info->bg_reclaim_threshold = BTRFS_DEFAULT_RECLAIM_THRESH; |
3159 | INIT_WORK(&fs_info->reclaim_bgs_work, btrfs_reclaim_bgs_work); | |
8260edba JB |
3160 | } |
3161 | ||
3162 | static int init_mount_fs_info(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
3163 | { | |
3164 | int ret; | |
3165 | ||
3166 | fs_info->sb = sb; | |
3167 | sb->s_blocksize = BTRFS_BDEV_BLOCKSIZE; | |
3168 | sb->s_blocksize_bits = blksize_bits(BTRFS_BDEV_BLOCKSIZE); | |
9e967495 | 3169 | |
5deb17e1 | 3170 | ret = percpu_counter_init(&fs_info->ordered_bytes, 0, GFP_KERNEL); |
ae18c37a | 3171 | if (ret) |
c75e8394 | 3172 | return ret; |
ae18c37a JB |
3173 | |
3174 | ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL); | |
3175 | if (ret) | |
c75e8394 | 3176 | return ret; |
ae18c37a JB |
3177 | |
3178 | fs_info->dirty_metadata_batch = PAGE_SIZE * | |
3179 | (1 + ilog2(nr_cpu_ids)); | |
3180 | ||
3181 | ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL); | |
3182 | if (ret) | |
c75e8394 | 3183 | return ret; |
ae18c37a JB |
3184 | |
3185 | ret = percpu_counter_init(&fs_info->dev_replace.bio_counter, 0, | |
3186 | GFP_KERNEL); | |
3187 | if (ret) | |
c75e8394 | 3188 | return ret; |
ae18c37a JB |
3189 | |
3190 | fs_info->delayed_root = kmalloc(sizeof(struct btrfs_delayed_root), | |
3191 | GFP_KERNEL); | |
c75e8394 JB |
3192 | if (!fs_info->delayed_root) |
3193 | return -ENOMEM; | |
ae18c37a JB |
3194 | btrfs_init_delayed_root(fs_info->delayed_root); |
3195 | ||
a0a1db70 FM |
3196 | if (sb_rdonly(sb)) |
3197 | set_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); | |
3198 | ||
c75e8394 | 3199 | return btrfs_alloc_stripe_hash_table(fs_info); |
ae18c37a JB |
3200 | } |
3201 | ||
97f4dd09 NB |
3202 | static int btrfs_uuid_rescan_kthread(void *data) |
3203 | { | |
0d031dc4 | 3204 | struct btrfs_fs_info *fs_info = data; |
97f4dd09 NB |
3205 | int ret; |
3206 | ||
3207 | /* | |
3208 | * 1st step is to iterate through the existing UUID tree and | |
3209 | * to delete all entries that contain outdated data. | |
3210 | * 2nd step is to add all missing entries to the UUID tree. | |
3211 | */ | |
3212 | ret = btrfs_uuid_tree_iterate(fs_info); | |
3213 | if (ret < 0) { | |
c94bec2c JB |
3214 | if (ret != -EINTR) |
3215 | btrfs_warn(fs_info, "iterating uuid_tree failed %d", | |
3216 | ret); | |
97f4dd09 NB |
3217 | up(&fs_info->uuid_tree_rescan_sem); |
3218 | return ret; | |
3219 | } | |
3220 | return btrfs_uuid_scan_kthread(data); | |
3221 | } | |
3222 | ||
3223 | static int btrfs_check_uuid_tree(struct btrfs_fs_info *fs_info) | |
3224 | { | |
3225 | struct task_struct *task; | |
3226 | ||
3227 | down(&fs_info->uuid_tree_rescan_sem); | |
3228 | task = kthread_run(btrfs_uuid_rescan_kthread, fs_info, "btrfs-uuid"); | |
3229 | if (IS_ERR(task)) { | |
3230 | /* fs_info->update_uuid_tree_gen remains 0 in all error case */ | |
3231 | btrfs_warn(fs_info, "failed to start uuid_rescan task"); | |
3232 | up(&fs_info->uuid_tree_rescan_sem); | |
3233 | return PTR_ERR(task); | |
3234 | } | |
3235 | ||
3236 | return 0; | |
3237 | } | |
3238 | ||
8cd29088 BB |
3239 | /* |
3240 | * Some options only have meaning at mount time and shouldn't persist across | |
3241 | * remounts, or be displayed. Clear these at the end of mount and remount | |
3242 | * code paths. | |
3243 | */ | |
3244 | void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) | |
3245 | { | |
3246 | btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); | |
8b228324 | 3247 | btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); |
8cd29088 BB |
3248 | } |
3249 | ||
44c0ca21 BB |
3250 | /* |
3251 | * Mounting logic specific to read-write file systems. Shared by open_ctree | |
3252 | * and btrfs_remount when remounting from read-only to read-write. | |
3253 | */ | |
3254 | int btrfs_start_pre_rw_mount(struct btrfs_fs_info *fs_info) | |
3255 | { | |
3256 | int ret; | |
94846229 | 3257 | const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); |
8b228324 BB |
3258 | bool clear_free_space_tree = false; |
3259 | ||
3260 | if (btrfs_test_opt(fs_info, CLEAR_CACHE) && | |
3261 | btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3262 | clear_free_space_tree = true; | |
3263 | } else if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && | |
3264 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE_VALID)) { | |
3265 | btrfs_warn(fs_info, "free space tree is invalid"); | |
3266 | clear_free_space_tree = true; | |
3267 | } | |
3268 | ||
3269 | if (clear_free_space_tree) { | |
3270 | btrfs_info(fs_info, "clearing free space tree"); | |
3271 | ret = btrfs_clear_free_space_tree(fs_info); | |
3272 | if (ret) { | |
3273 | btrfs_warn(fs_info, | |
3274 | "failed to clear free space tree: %d", ret); | |
3275 | goto out; | |
3276 | } | |
3277 | } | |
44c0ca21 | 3278 | |
8d488a8c FM |
3279 | /* |
3280 | * btrfs_find_orphan_roots() is responsible for finding all the dead | |
3281 | * roots (with 0 refs), flag them with BTRFS_ROOT_DEAD_TREE and load | |
fc7cbcd4 | 3282 | * them into the fs_info->fs_roots_radix tree. This must be done before |
8d488a8c FM |
3283 | * calling btrfs_orphan_cleanup() on the tree root. If we don't do it |
3284 | * first, then btrfs_orphan_cleanup() will delete a dead root's orphan | |
3285 | * item before the root's tree is deleted - this means that if we unmount | |
3286 | * or crash before the deletion completes, on the next mount we will not | |
3287 | * delete what remains of the tree because the orphan item does not | |
3288 | * exists anymore, which is what tells us we have a pending deletion. | |
3289 | */ | |
3290 | ret = btrfs_find_orphan_roots(fs_info); | |
3291 | if (ret) | |
3292 | goto out; | |
3293 | ||
44c0ca21 BB |
3294 | ret = btrfs_cleanup_fs_roots(fs_info); |
3295 | if (ret) | |
3296 | goto out; | |
3297 | ||
8f1c21d7 BB |
3298 | down_read(&fs_info->cleanup_work_sem); |
3299 | if ((ret = btrfs_orphan_cleanup(fs_info->fs_root)) || | |
3300 | (ret = btrfs_orphan_cleanup(fs_info->tree_root))) { | |
3301 | up_read(&fs_info->cleanup_work_sem); | |
3302 | goto out; | |
3303 | } | |
3304 | up_read(&fs_info->cleanup_work_sem); | |
3305 | ||
44c0ca21 | 3306 | mutex_lock(&fs_info->cleaner_mutex); |
7eefae6b | 3307 | ret = btrfs_recover_relocation(fs_info); |
44c0ca21 BB |
3308 | mutex_unlock(&fs_info->cleaner_mutex); |
3309 | if (ret < 0) { | |
3310 | btrfs_warn(fs_info, "failed to recover relocation: %d", ret); | |
3311 | goto out; | |
3312 | } | |
3313 | ||
5011139a BB |
3314 | if (btrfs_test_opt(fs_info, FREE_SPACE_TREE) && |
3315 | !btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { | |
3316 | btrfs_info(fs_info, "creating free space tree"); | |
3317 | ret = btrfs_create_free_space_tree(fs_info); | |
3318 | if (ret) { | |
3319 | btrfs_warn(fs_info, | |
3320 | "failed to create free space tree: %d", ret); | |
3321 | goto out; | |
3322 | } | |
3323 | } | |
3324 | ||
94846229 BB |
3325 | if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) { |
3326 | ret = btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); | |
3327 | if (ret) | |
3328 | goto out; | |
3329 | } | |
3330 | ||
44c0ca21 BB |
3331 | ret = btrfs_resume_balance_async(fs_info); |
3332 | if (ret) | |
3333 | goto out; | |
3334 | ||
3335 | ret = btrfs_resume_dev_replace_async(fs_info); | |
3336 | if (ret) { | |
3337 | btrfs_warn(fs_info, "failed to resume dev_replace"); | |
3338 | goto out; | |
3339 | } | |
3340 | ||
3341 | btrfs_qgroup_rescan_resume(fs_info); | |
3342 | ||
3343 | if (!fs_info->uuid_root) { | |
3344 | btrfs_info(fs_info, "creating UUID tree"); | |
3345 | ret = btrfs_create_uuid_tree(fs_info); | |
3346 | if (ret) { | |
3347 | btrfs_warn(fs_info, | |
3348 | "failed to create the UUID tree %d", ret); | |
3349 | goto out; | |
3350 | } | |
3351 | } | |
3352 | ||
3353 | out: | |
3354 | return ret; | |
3355 | } | |
3356 | ||
d7f67ac9 QW |
3357 | /* |
3358 | * Do various sanity and dependency checks of different features. | |
3359 | * | |
3360 | * This is the place for less strict checks (like for subpage or artificial | |
3361 | * feature dependencies). | |
3362 | * | |
3363 | * For strict checks or possible corruption detection, see | |
3364 | * btrfs_validate_super(). | |
3365 | * | |
3366 | * This should be called after btrfs_parse_options(), as some mount options | |
3367 | * (space cache related) can modify on-disk format like free space tree and | |
3368 | * screw up certain feature dependencies. | |
3369 | */ | |
3370 | int btrfs_check_features(struct btrfs_fs_info *fs_info, struct super_block *sb) | |
3371 | { | |
3372 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
3373 | u64 incompat = btrfs_super_incompat_flags(disk_super); | |
3374 | const u64 compat_ro = btrfs_super_compat_ro_flags(disk_super); | |
3375 | const u64 compat_ro_unsupp = (compat_ro & ~BTRFS_FEATURE_COMPAT_RO_SUPP); | |
3376 | ||
3377 | if (incompat & ~BTRFS_FEATURE_INCOMPAT_SUPP) { | |
3378 | btrfs_err(fs_info, | |
3379 | "cannot mount because of unknown incompat features (0x%llx)", | |
3380 | incompat); | |
3381 | return -EINVAL; | |
3382 | } | |
3383 | ||
3384 | /* Runtime limitation for mixed block groups. */ | |
3385 | if ((incompat & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS) && | |
3386 | (fs_info->sectorsize != fs_info->nodesize)) { | |
3387 | btrfs_err(fs_info, | |
3388 | "unequal nodesize/sectorsize (%u != %u) are not allowed for mixed block groups", | |
3389 | fs_info->nodesize, fs_info->sectorsize); | |
3390 | return -EINVAL; | |
3391 | } | |
3392 | ||
3393 | /* Mixed backref is an always-enabled feature. */ | |
3394 | incompat |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF; | |
3395 | ||
3396 | /* Set compression related flags just in case. */ | |
3397 | if (fs_info->compress_type == BTRFS_COMPRESS_LZO) | |
3398 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO; | |
3399 | else if (fs_info->compress_type == BTRFS_COMPRESS_ZSTD) | |
3400 | incompat |= BTRFS_FEATURE_INCOMPAT_COMPRESS_ZSTD; | |
3401 | ||
3402 | /* | |
3403 | * An ancient flag, which should really be marked deprecated. | |
3404 | * Such runtime limitation doesn't really need a incompat flag. | |
3405 | */ | |
3406 | if (btrfs_super_nodesize(disk_super) > PAGE_SIZE) | |
3407 | incompat |= BTRFS_FEATURE_INCOMPAT_BIG_METADATA; | |
3408 | ||
3409 | if (compat_ro_unsupp && !sb_rdonly(sb)) { | |
3410 | btrfs_err(fs_info, | |
3411 | "cannot mount read-write because of unknown compat_ro features (0x%llx)", | |
3412 | compat_ro); | |
3413 | return -EINVAL; | |
3414 | } | |
3415 | ||
3416 | /* | |
3417 | * We have unsupported RO compat features, although RO mounted, we | |
3418 | * should not cause any metadata writes, including log replay. | |
3419 | * Or we could screw up whatever the new feature requires. | |
3420 | */ | |
3421 | if (compat_ro_unsupp && btrfs_super_log_root(disk_super) && | |
3422 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { | |
3423 | btrfs_err(fs_info, | |
3424 | "cannot replay dirty log with unsupported compat_ro features (0x%llx), try rescue=nologreplay", | |
3425 | compat_ro); | |
3426 | return -EINVAL; | |
3427 | } | |
3428 | ||
3429 | /* | |
3430 | * Artificial limitations for block group tree, to force | |
3431 | * block-group-tree to rely on no-holes and free-space-tree. | |
3432 | */ | |
3433 | if (btrfs_fs_compat_ro(fs_info, BLOCK_GROUP_TREE) && | |
3434 | (!btrfs_fs_incompat(fs_info, NO_HOLES) || | |
3435 | !btrfs_test_opt(fs_info, FREE_SPACE_TREE))) { | |
3436 | btrfs_err(fs_info, | |
3437 | "block-group-tree feature requires no-holes and free-space-tree features"); | |
3438 | return -EINVAL; | |
3439 | } | |
3440 | ||
3441 | /* | |
3442 | * Subpage runtime limitation on v1 cache. | |
3443 | * | |
3444 | * V1 space cache still has some hard codeed PAGE_SIZE usage, while | |
3445 | * we're already defaulting to v2 cache, no need to bother v1 as it's | |
3446 | * going to be deprecated anyway. | |
3447 | */ | |
3448 | if (fs_info->sectorsize < PAGE_SIZE && btrfs_test_opt(fs_info, SPACE_CACHE)) { | |
3449 | btrfs_warn(fs_info, | |
3450 | "v1 space cache is not supported for page size %lu with sectorsize %u", | |
3451 | PAGE_SIZE, fs_info->sectorsize); | |
3452 | return -EINVAL; | |
3453 | } | |
3454 | ||
3455 | /* This can be called by remount, we need to protect the super block. */ | |
3456 | spin_lock(&fs_info->super_lock); | |
3457 | btrfs_set_super_incompat_flags(disk_super, incompat); | |
3458 | spin_unlock(&fs_info->super_lock); | |
3459 | ||
3460 | return 0; | |
3461 | } | |
3462 | ||
ae18c37a JB |
3463 | int __cold open_ctree(struct super_block *sb, struct btrfs_fs_devices *fs_devices, |
3464 | char *options) | |
3465 | { | |
3466 | u32 sectorsize; | |
3467 | u32 nodesize; | |
3468 | u32 stripesize; | |
3469 | u64 generation; | |
3470 | u64 features; | |
3471 | u16 csum_type; | |
ae18c37a JB |
3472 | struct btrfs_super_block *disk_super; |
3473 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
3474 | struct btrfs_root *tree_root; | |
3475 | struct btrfs_root *chunk_root; | |
3476 | int ret; | |
3477 | int err = -EINVAL; | |
ae18c37a JB |
3478 | int level; |
3479 | ||
8260edba | 3480 | ret = init_mount_fs_info(fs_info, sb); |
53b381b3 | 3481 | if (ret) { |
83c8266a | 3482 | err = ret; |
ae18c37a | 3483 | goto fail; |
53b381b3 DW |
3484 | } |
3485 | ||
ae18c37a JB |
3486 | /* These need to be init'ed before we start creating inodes and such. */ |
3487 | tree_root = btrfs_alloc_root(fs_info, BTRFS_ROOT_TREE_OBJECTID, | |
3488 | GFP_KERNEL); | |
3489 | fs_info->tree_root = tree_root; | |
3490 | chunk_root = btrfs_alloc_root(fs_info, BTRFS_CHUNK_TREE_OBJECTID, | |
3491 | GFP_KERNEL); | |
3492 | fs_info->chunk_root = chunk_root; | |
3493 | if (!tree_root || !chunk_root) { | |
3494 | err = -ENOMEM; | |
c75e8394 | 3495 | goto fail; |
ae18c37a JB |
3496 | } |
3497 | ||
3498 | fs_info->btree_inode = new_inode(sb); | |
3499 | if (!fs_info->btree_inode) { | |
3500 | err = -ENOMEM; | |
c75e8394 | 3501 | goto fail; |
ae18c37a JB |
3502 | } |
3503 | mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS); | |
3504 | btrfs_init_btree_inode(fs_info); | |
3505 | ||
d24fa5c1 | 3506 | invalidate_bdev(fs_devices->latest_dev->bdev); |
1104a885 DS |
3507 | |
3508 | /* | |
3509 | * Read super block and check the signature bytes only | |
3510 | */ | |
d24fa5c1 | 3511 | disk_super = btrfs_read_dev_super(fs_devices->latest_dev->bdev); |
8f32380d JT |
3512 | if (IS_ERR(disk_super)) { |
3513 | err = PTR_ERR(disk_super); | |
16cdcec7 | 3514 | goto fail_alloc; |
20b45077 | 3515 | } |
39279cc3 | 3516 | |
8dc3f22c | 3517 | /* |
260db43c | 3518 | * Verify the type first, if that or the checksum value are |
8dc3f22c JT |
3519 | * corrupted, we'll find out |
3520 | */ | |
8f32380d | 3521 | csum_type = btrfs_super_csum_type(disk_super); |
51bce6c9 | 3522 | if (!btrfs_supported_super_csum(csum_type)) { |
8dc3f22c | 3523 | btrfs_err(fs_info, "unsupported checksum algorithm: %u", |
51bce6c9 | 3524 | csum_type); |
8dc3f22c | 3525 | err = -EINVAL; |
8f32380d | 3526 | btrfs_release_disk_super(disk_super); |
8dc3f22c JT |
3527 | goto fail_alloc; |
3528 | } | |
3529 | ||
83c68bbc SY |
3530 | fs_info->csum_size = btrfs_super_csum_size(disk_super); |
3531 | ||
6d97c6e3 JT |
3532 | ret = btrfs_init_csum_hash(fs_info, csum_type); |
3533 | if (ret) { | |
3534 | err = ret; | |
8f32380d | 3535 | btrfs_release_disk_super(disk_super); |
6d97c6e3 JT |
3536 | goto fail_alloc; |
3537 | } | |
3538 | ||
1104a885 DS |
3539 | /* |
3540 | * We want to check superblock checksum, the type is stored inside. | |
3541 | * Pass the whole disk block of size BTRFS_SUPER_INFO_SIZE (4k). | |
3542 | */ | |
3d17adea | 3543 | if (btrfs_check_super_csum(fs_info, disk_super)) { |
05135f59 | 3544 | btrfs_err(fs_info, "superblock checksum mismatch"); |
1104a885 | 3545 | err = -EINVAL; |
8f32380d | 3546 | btrfs_release_disk_super(disk_super); |
141386e1 | 3547 | goto fail_alloc; |
1104a885 DS |
3548 | } |
3549 | ||
3550 | /* | |
3551 | * super_copy is zeroed at allocation time and we never touch the | |
3552 | * following bytes up to INFO_SIZE, the checksum is calculated from | |
3553 | * the whole block of INFO_SIZE | |
3554 | */ | |
8f32380d JT |
3555 | memcpy(fs_info->super_copy, disk_super, sizeof(*fs_info->super_copy)); |
3556 | btrfs_release_disk_super(disk_super); | |
5f39d397 | 3557 | |
fbc6feae NB |
3558 | disk_super = fs_info->super_copy; |
3559 | ||
0b86a832 | 3560 | |
fbc6feae NB |
3561 | features = btrfs_super_flags(disk_super); |
3562 | if (features & BTRFS_SUPER_FLAG_CHANGING_FSID_V2) { | |
3563 | features &= ~BTRFS_SUPER_FLAG_CHANGING_FSID_V2; | |
3564 | btrfs_set_super_flags(disk_super, features); | |
3565 | btrfs_info(fs_info, | |
3566 | "found metadata UUID change in progress flag, clearing"); | |
3567 | } | |
3568 | ||
3569 | memcpy(fs_info->super_for_commit, fs_info->super_copy, | |
3570 | sizeof(*fs_info->super_for_commit)); | |
de37aa51 | 3571 | |
069ec957 | 3572 | ret = btrfs_validate_mount_super(fs_info); |
1104a885 | 3573 | if (ret) { |
05135f59 | 3574 | btrfs_err(fs_info, "superblock contains fatal errors"); |
1104a885 | 3575 | err = -EINVAL; |
141386e1 | 3576 | goto fail_alloc; |
1104a885 DS |
3577 | } |
3578 | ||
0f7d52f4 | 3579 | if (!btrfs_super_root(disk_super)) |
141386e1 | 3580 | goto fail_alloc; |
0f7d52f4 | 3581 | |
acce952b | 3582 | /* check FS state, whether FS is broken. */ |
87533c47 MX |
3583 | if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_ERROR) |
3584 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
acce952b | 3585 | |
75e7cb7f LB |
3586 | /* |
3587 | * In the long term, we'll store the compression type in the super | |
3588 | * block, and it'll be used for per file compression control. | |
3589 | */ | |
3590 | fs_info->compress_type = BTRFS_COMPRESS_ZLIB; | |
3591 | ||
6f93e834 AJ |
3592 | |
3593 | /* Set up fs_info before parsing mount options */ | |
3594 | nodesize = btrfs_super_nodesize(disk_super); | |
3595 | sectorsize = btrfs_super_sectorsize(disk_super); | |
3596 | stripesize = sectorsize; | |
3597 | fs_info->dirty_metadata_batch = nodesize * (1 + ilog2(nr_cpu_ids)); | |
3598 | fs_info->delalloc_batch = sectorsize * 512 * (1 + ilog2(nr_cpu_ids)); | |
3599 | ||
3600 | fs_info->nodesize = nodesize; | |
3601 | fs_info->sectorsize = sectorsize; | |
3602 | fs_info->sectorsize_bits = ilog2(sectorsize); | |
3603 | fs_info->csums_per_leaf = BTRFS_MAX_ITEM_SIZE(fs_info) / fs_info->csum_size; | |
3604 | fs_info->stripesize = stripesize; | |
3605 | ||
2ff7e61e | 3606 | ret = btrfs_parse_options(fs_info, options, sb->s_flags); |
2b82032c YZ |
3607 | if (ret) { |
3608 | err = ret; | |
141386e1 | 3609 | goto fail_alloc; |
2b82032c | 3610 | } |
dfe25020 | 3611 | |
d7f67ac9 QW |
3612 | ret = btrfs_check_features(fs_info, sb); |
3613 | if (ret < 0) { | |
3614 | err = ret; | |
dc4d3168 QW |
3615 | goto fail_alloc; |
3616 | } | |
3617 | ||
8481dd80 QW |
3618 | if (sectorsize < PAGE_SIZE) { |
3619 | struct btrfs_subpage_info *subpage_info; | |
3620 | ||
9f73f1ae QW |
3621 | /* |
3622 | * V1 space cache has some hardcoded PAGE_SIZE usage, and is | |
3623 | * going to be deprecated. | |
3624 | * | |
3625 | * Force to use v2 cache for subpage case. | |
3626 | */ | |
3627 | btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); | |
3628 | btrfs_set_and_info(fs_info, FREE_SPACE_TREE, | |
3629 | "forcing free space tree for sector size %u with page size %lu", | |
3630 | sectorsize, PAGE_SIZE); | |
3631 | ||
95ea0486 QW |
3632 | btrfs_warn(fs_info, |
3633 | "read-write for sector size %u with page size %lu is experimental", | |
3634 | sectorsize, PAGE_SIZE); | |
8481dd80 QW |
3635 | subpage_info = kzalloc(sizeof(*subpage_info), GFP_KERNEL); |
3636 | if (!subpage_info) | |
3637 | goto fail_alloc; | |
3638 | btrfs_init_subpage_info(subpage_info, sectorsize); | |
3639 | fs_info->subpage_info = subpage_info; | |
c8050b3b | 3640 | } |
0bb3eb3e | 3641 | |
d21deec5 | 3642 | ret = btrfs_init_workqueues(fs_info); |
2a458198 ES |
3643 | if (ret) { |
3644 | err = ret; | |
0dc3b84a JB |
3645 | goto fail_sb_buffer; |
3646 | } | |
4543df7e | 3647 | |
9e11ceee JK |
3648 | sb->s_bdi->ra_pages *= btrfs_super_num_devices(disk_super); |
3649 | sb->s_bdi->ra_pages = max(sb->s_bdi->ra_pages, SZ_4M / PAGE_SIZE); | |
4575c9cc | 3650 | |
a061fc8d CM |
3651 | sb->s_blocksize = sectorsize; |
3652 | sb->s_blocksize_bits = blksize_bits(sectorsize); | |
de37aa51 | 3653 | memcpy(&sb->s_uuid, fs_info->fs_devices->fsid, BTRFS_FSID_SIZE); |
db94535d | 3654 | |
925baedd | 3655 | mutex_lock(&fs_info->chunk_mutex); |
6bccf3ab | 3656 | ret = btrfs_read_sys_array(fs_info); |
925baedd | 3657 | mutex_unlock(&fs_info->chunk_mutex); |
84eed90f | 3658 | if (ret) { |
05135f59 | 3659 | btrfs_err(fs_info, "failed to read the system array: %d", ret); |
5d4f98a2 | 3660 | goto fail_sb_buffer; |
84eed90f | 3661 | } |
0b86a832 | 3662 | |
84234f3a | 3663 | generation = btrfs_super_chunk_root_generation(disk_super); |
581c1760 | 3664 | level = btrfs_super_chunk_root_level(disk_super); |
bd676446 JB |
3665 | ret = load_super_root(chunk_root, btrfs_super_chunk_root(disk_super), |
3666 | generation, level); | |
3667 | if (ret) { | |
05135f59 | 3668 | btrfs_err(fs_info, "failed to read chunk root"); |
af31f5e5 | 3669 | goto fail_tree_roots; |
83121942 | 3670 | } |
0b86a832 | 3671 | |
e17cade2 | 3672 | read_extent_buffer(chunk_root->node, fs_info->chunk_tree_uuid, |
c4ac7541 DS |
3673 | offsetof(struct btrfs_header, chunk_tree_uuid), |
3674 | BTRFS_UUID_SIZE); | |
e17cade2 | 3675 | |
5b4aacef | 3676 | ret = btrfs_read_chunk_tree(fs_info); |
2b82032c | 3677 | if (ret) { |
05135f59 | 3678 | btrfs_err(fs_info, "failed to read chunk tree: %d", ret); |
af31f5e5 | 3679 | goto fail_tree_roots; |
2b82032c | 3680 | } |
0b86a832 | 3681 | |
8dabb742 | 3682 | /* |
bacce86a AJ |
3683 | * At this point we know all the devices that make this filesystem, |
3684 | * including the seed devices but we don't know yet if the replace | |
3685 | * target is required. So free devices that are not part of this | |
1a9fd417 | 3686 | * filesystem but skip the replace target device which is checked |
bacce86a | 3687 | * below in btrfs_init_dev_replace(). |
8dabb742 | 3688 | */ |
bacce86a | 3689 | btrfs_free_extra_devids(fs_devices); |
d24fa5c1 | 3690 | if (!fs_devices->latest_dev->bdev) { |
05135f59 | 3691 | btrfs_err(fs_info, "failed to read devices"); |
a6b0d5c8 CM |
3692 | goto fail_tree_roots; |
3693 | } | |
3694 | ||
b8522a1e | 3695 | ret = init_tree_roots(fs_info); |
4bbcaa64 | 3696 | if (ret) |
b8522a1e | 3697 | goto fail_tree_roots; |
8929ecfa | 3698 | |
73651042 NA |
3699 | /* |
3700 | * Get zone type information of zoned block devices. This will also | |
3701 | * handle emulation of a zoned filesystem if a regular device has the | |
3702 | * zoned incompat feature flag set. | |
3703 | */ | |
3704 | ret = btrfs_get_dev_zone_info_all_devices(fs_info); | |
3705 | if (ret) { | |
3706 | btrfs_err(fs_info, | |
3707 | "zoned: failed to read device zone info: %d", | |
3708 | ret); | |
3709 | goto fail_block_groups; | |
3710 | } | |
3711 | ||
75ec1db8 JB |
3712 | /* |
3713 | * If we have a uuid root and we're not being told to rescan we need to | |
3714 | * check the generation here so we can set the | |
3715 | * BTRFS_FS_UPDATE_UUID_TREE_GEN bit. Otherwise we could commit the | |
3716 | * transaction during a balance or the log replay without updating the | |
3717 | * uuid generation, and then if we crash we would rescan the uuid tree, | |
3718 | * even though it was perfectly fine. | |
3719 | */ | |
3720 | if (fs_info->uuid_root && !btrfs_test_opt(fs_info, RESCAN_UUID_TREE) && | |
3721 | fs_info->generation == btrfs_super_uuid_tree_generation(disk_super)) | |
3722 | set_bit(BTRFS_FS_UPDATE_UUID_TREE_GEN, &fs_info->flags); | |
3723 | ||
cf90d884 QW |
3724 | ret = btrfs_verify_dev_extents(fs_info); |
3725 | if (ret) { | |
3726 | btrfs_err(fs_info, | |
3727 | "failed to verify dev extents against chunks: %d", | |
3728 | ret); | |
3729 | goto fail_block_groups; | |
3730 | } | |
68310a5e ID |
3731 | ret = btrfs_recover_balance(fs_info); |
3732 | if (ret) { | |
05135f59 | 3733 | btrfs_err(fs_info, "failed to recover balance: %d", ret); |
68310a5e ID |
3734 | goto fail_block_groups; |
3735 | } | |
3736 | ||
733f4fbb SB |
3737 | ret = btrfs_init_dev_stats(fs_info); |
3738 | if (ret) { | |
05135f59 | 3739 | btrfs_err(fs_info, "failed to init dev_stats: %d", ret); |
733f4fbb SB |
3740 | goto fail_block_groups; |
3741 | } | |
3742 | ||
8dabb742 SB |
3743 | ret = btrfs_init_dev_replace(fs_info); |
3744 | if (ret) { | |
05135f59 | 3745 | btrfs_err(fs_info, "failed to init dev_replace: %d", ret); |
8dabb742 SB |
3746 | goto fail_block_groups; |
3747 | } | |
3748 | ||
b70f5097 NA |
3749 | ret = btrfs_check_zoned_mode(fs_info); |
3750 | if (ret) { | |
3751 | btrfs_err(fs_info, "failed to initialize zoned mode: %d", | |
3752 | ret); | |
3753 | goto fail_block_groups; | |
3754 | } | |
3755 | ||
c6761a9e | 3756 | ret = btrfs_sysfs_add_fsid(fs_devices); |
b7c35e81 | 3757 | if (ret) { |
05135f59 DS |
3758 | btrfs_err(fs_info, "failed to init sysfs fsid interface: %d", |
3759 | ret); | |
b7c35e81 AJ |
3760 | goto fail_block_groups; |
3761 | } | |
3762 | ||
96f3136e | 3763 | ret = btrfs_sysfs_add_mounted(fs_info); |
c59021f8 | 3764 | if (ret) { |
05135f59 | 3765 | btrfs_err(fs_info, "failed to init sysfs interface: %d", ret); |
b7c35e81 | 3766 | goto fail_fsdev_sysfs; |
c59021f8 | 3767 | } |
3768 | ||
c59021f8 | 3769 | ret = btrfs_init_space_info(fs_info); |
3770 | if (ret) { | |
05135f59 | 3771 | btrfs_err(fs_info, "failed to initialize space info: %d", ret); |
2365dd3c | 3772 | goto fail_sysfs; |
c59021f8 | 3773 | } |
3774 | ||
5b4aacef | 3775 | ret = btrfs_read_block_groups(fs_info); |
1b1d1f66 | 3776 | if (ret) { |
05135f59 | 3777 | btrfs_err(fs_info, "failed to read block groups: %d", ret); |
2365dd3c | 3778 | goto fail_sysfs; |
1b1d1f66 | 3779 | } |
4330e183 | 3780 | |
16beac87 NA |
3781 | btrfs_free_zone_cache(fs_info); |
3782 | ||
5c78a5e7 AJ |
3783 | if (!sb_rdonly(sb) && fs_info->fs_devices->missing_devices && |
3784 | !btrfs_check_rw_degradable(fs_info, NULL)) { | |
05135f59 | 3785 | btrfs_warn(fs_info, |
52042d8e | 3786 | "writable mount is not allowed due to too many missing devices"); |
2365dd3c | 3787 | goto fail_sysfs; |
292fd7fc | 3788 | } |
9078a3e1 | 3789 | |
33c44184 | 3790 | fs_info->cleaner_kthread = kthread_run(cleaner_kthread, fs_info, |
a74a4b97 | 3791 | "btrfs-cleaner"); |
57506d50 | 3792 | if (IS_ERR(fs_info->cleaner_kthread)) |
2365dd3c | 3793 | goto fail_sysfs; |
a74a4b97 CM |
3794 | |
3795 | fs_info->transaction_kthread = kthread_run(transaction_kthread, | |
3796 | tree_root, | |
3797 | "btrfs-transaction"); | |
57506d50 | 3798 | if (IS_ERR(fs_info->transaction_kthread)) |
3f157a2f | 3799 | goto fail_cleaner; |
a74a4b97 | 3800 | |
583b7231 | 3801 | if (!btrfs_test_opt(fs_info, NOSSD) && |
c289811c | 3802 | !fs_info->fs_devices->rotating) { |
583b7231 | 3803 | btrfs_set_and_info(fs_info, SSD, "enabling ssd optimizations"); |
c289811c CM |
3804 | } |
3805 | ||
63a7cb13 DS |
3806 | /* |
3807 | * For devices supporting discard turn on discard=async automatically, | |
3808 | * unless it's already set or disabled. This could be turned off by | |
3809 | * nodiscard for the same mount. | |
3810 | */ | |
3811 | if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || | |
3812 | btrfs_test_opt(fs_info, DISCARD_ASYNC) || | |
3813 | btrfs_test_opt(fs_info, NODISCARD)) && | |
3814 | fs_info->fs_devices->discardable) { | |
3815 | btrfs_set_and_info(fs_info, DISCARD_ASYNC, | |
3816 | "auto enabling async discard"); | |
3817 | btrfs_clear_opt(fs_info->mount_opt, NODISCARD); | |
3818 | } | |
3819 | ||
21adbd5c | 3820 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 3821 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) { |
2ff7e61e | 3822 | ret = btrfsic_mount(fs_info, fs_devices, |
0b246afa | 3823 | btrfs_test_opt(fs_info, |
cbeaae4f | 3824 | CHECK_INTEGRITY_DATA) ? 1 : 0, |
21adbd5c SB |
3825 | fs_info->check_integrity_print_mask); |
3826 | if (ret) | |
05135f59 DS |
3827 | btrfs_warn(fs_info, |
3828 | "failed to initialize integrity check module: %d", | |
3829 | ret); | |
21adbd5c SB |
3830 | } |
3831 | #endif | |
bcef60f2 AJ |
3832 | ret = btrfs_read_qgroup_config(fs_info); |
3833 | if (ret) | |
3834 | goto fail_trans_kthread; | |
21adbd5c | 3835 | |
fd708b81 JB |
3836 | if (btrfs_build_ref_tree(fs_info)) |
3837 | btrfs_err(fs_info, "couldn't build ref tree"); | |
3838 | ||
96da0919 QW |
3839 | /* do not make disk changes in broken FS or nologreplay is given */ |
3840 | if (btrfs_super_log_root(disk_super) != 0 && | |
0b246afa | 3841 | !btrfs_test_opt(fs_info, NOLOGREPLAY)) { |
e8294f2f | 3842 | btrfs_info(fs_info, "start tree-log replay"); |
63443bf5 | 3843 | ret = btrfs_replay_log(fs_info, fs_devices); |
79787eaa | 3844 | if (ret) { |
63443bf5 | 3845 | err = ret; |
28c16cbb | 3846 | goto fail_qgroup; |
79787eaa | 3847 | } |
e02119d5 | 3848 | } |
1a40e23b | 3849 | |
56e9357a | 3850 | fs_info->fs_root = btrfs_get_fs_root(fs_info, BTRFS_FS_TREE_OBJECTID, true); |
3140c9a3 DC |
3851 | if (IS_ERR(fs_info->fs_root)) { |
3852 | err = PTR_ERR(fs_info->fs_root); | |
f50f4353 | 3853 | btrfs_warn(fs_info, "failed to read fs tree: %d", err); |
315bf8ef | 3854 | fs_info->fs_root = NULL; |
bcef60f2 | 3855 | goto fail_qgroup; |
3140c9a3 | 3856 | } |
c289811c | 3857 | |
bc98a42c | 3858 | if (sb_rdonly(sb)) |
8cd29088 | 3859 | goto clear_oneshot; |
59641015 | 3860 | |
44c0ca21 | 3861 | ret = btrfs_start_pre_rw_mount(fs_info); |
2b6ba629 | 3862 | if (ret) { |
6bccf3ab | 3863 | close_ctree(fs_info); |
2b6ba629 | 3864 | return ret; |
e3acc2a6 | 3865 | } |
b0643e59 | 3866 | btrfs_discard_resume(fs_info); |
b382a324 | 3867 | |
44c0ca21 BB |
3868 | if (fs_info->uuid_root && |
3869 | (btrfs_test_opt(fs_info, RESCAN_UUID_TREE) || | |
3870 | fs_info->generation != btrfs_super_uuid_tree_generation(disk_super))) { | |
05135f59 | 3871 | btrfs_info(fs_info, "checking UUID tree"); |
70f80175 SB |
3872 | ret = btrfs_check_uuid_tree(fs_info); |
3873 | if (ret) { | |
05135f59 DS |
3874 | btrfs_warn(fs_info, |
3875 | "failed to check the UUID tree: %d", ret); | |
6bccf3ab | 3876 | close_ctree(fs_info); |
70f80175 SB |
3877 | return ret; |
3878 | } | |
f7a81ea4 | 3879 | } |
94846229 | 3880 | |
afcdd129 | 3881 | set_bit(BTRFS_FS_OPEN, &fs_info->flags); |
47ab2a6c | 3882 | |
b4be6aef JB |
3883 | /* Kick the cleaner thread so it'll start deleting snapshots. */ |
3884 | if (test_bit(BTRFS_FS_UNFINISHED_DROPS, &fs_info->flags)) | |
3885 | wake_up_process(fs_info->cleaner_kthread); | |
3886 | ||
8cd29088 BB |
3887 | clear_oneshot: |
3888 | btrfs_clear_oneshot_options(fs_info); | |
ad2b2c80 | 3889 | return 0; |
39279cc3 | 3890 | |
bcef60f2 AJ |
3891 | fail_qgroup: |
3892 | btrfs_free_qgroup_config(fs_info); | |
7c2ca468 CM |
3893 | fail_trans_kthread: |
3894 | kthread_stop(fs_info->transaction_kthread); | |
2ff7e61e | 3895 | btrfs_cleanup_transaction(fs_info); |
faa2dbf0 | 3896 | btrfs_free_fs_roots(fs_info); |
3f157a2f | 3897 | fail_cleaner: |
a74a4b97 | 3898 | kthread_stop(fs_info->cleaner_kthread); |
7c2ca468 CM |
3899 | |
3900 | /* | |
3901 | * make sure we're done with the btree inode before we stop our | |
3902 | * kthreads | |
3903 | */ | |
3904 | filemap_write_and_wait(fs_info->btree_inode->i_mapping); | |
7c2ca468 | 3905 | |
2365dd3c | 3906 | fail_sysfs: |
6618a59b | 3907 | btrfs_sysfs_remove_mounted(fs_info); |
2365dd3c | 3908 | |
b7c35e81 AJ |
3909 | fail_fsdev_sysfs: |
3910 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); | |
3911 | ||
1b1d1f66 | 3912 | fail_block_groups: |
54067ae9 | 3913 | btrfs_put_block_group_cache(fs_info); |
af31f5e5 CM |
3914 | |
3915 | fail_tree_roots: | |
9e3aa805 JB |
3916 | if (fs_info->data_reloc_root) |
3917 | btrfs_drop_and_free_fs_root(fs_info, fs_info->data_reloc_root); | |
4273eaff | 3918 | free_root_pointers(fs_info, true); |
2b8195bb | 3919 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
af31f5e5 | 3920 | |
39279cc3 | 3921 | fail_sb_buffer: |
7abadb64 | 3922 | btrfs_stop_all_workers(fs_info); |
5cdd7db6 | 3923 | btrfs_free_block_groups(fs_info); |
16cdcec7 | 3924 | fail_alloc: |
586e46e2 ID |
3925 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
3926 | ||
4543df7e | 3927 | iput(fs_info->btree_inode); |
7e662854 | 3928 | fail: |
586e46e2 | 3929 | btrfs_close_devices(fs_info->fs_devices); |
ad2b2c80 | 3930 | return err; |
eb60ceac | 3931 | } |
663faf9f | 3932 | ALLOW_ERROR_INJECTION(open_ctree, ERRNO); |
eb60ceac | 3933 | |
314b6dd0 | 3934 | static void btrfs_end_super_write(struct bio *bio) |
f2984462 | 3935 | { |
314b6dd0 JT |
3936 | struct btrfs_device *device = bio->bi_private; |
3937 | struct bio_vec *bvec; | |
3938 | struct bvec_iter_all iter_all; | |
3939 | struct page *page; | |
3940 | ||
3941 | bio_for_each_segment_all(bvec, bio, iter_all) { | |
3942 | page = bvec->bv_page; | |
3943 | ||
3944 | if (bio->bi_status) { | |
3945 | btrfs_warn_rl_in_rcu(device->fs_info, | |
3946 | "lost page write due to IO error on %s (%d)", | |
cb3e217b | 3947 | btrfs_dev_name(device), |
314b6dd0 JT |
3948 | blk_status_to_errno(bio->bi_status)); |
3949 | ClearPageUptodate(page); | |
3950 | SetPageError(page); | |
3951 | btrfs_dev_stat_inc_and_print(device, | |
3952 | BTRFS_DEV_STAT_WRITE_ERRS); | |
3953 | } else { | |
3954 | SetPageUptodate(page); | |
3955 | } | |
3956 | ||
3957 | put_page(page); | |
3958 | unlock_page(page); | |
f2984462 | 3959 | } |
314b6dd0 JT |
3960 | |
3961 | bio_put(bio); | |
f2984462 CM |
3962 | } |
3963 | ||
8f32380d | 3964 | struct btrfs_super_block *btrfs_read_dev_one_super(struct block_device *bdev, |
a05d3c91 | 3965 | int copy_num, bool drop_cache) |
29c36d72 | 3966 | { |
29c36d72 | 3967 | struct btrfs_super_block *super; |
8f32380d | 3968 | struct page *page; |
12659251 | 3969 | u64 bytenr, bytenr_orig; |
8f32380d | 3970 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
12659251 NA |
3971 | int ret; |
3972 | ||
3973 | bytenr_orig = btrfs_sb_offset(copy_num); | |
3974 | ret = btrfs_sb_log_location_bdev(bdev, copy_num, READ, &bytenr); | |
3975 | if (ret == -ENOENT) | |
3976 | return ERR_PTR(-EINVAL); | |
3977 | else if (ret) | |
3978 | return ERR_PTR(ret); | |
29c36d72 | 3979 | |
cda00eba | 3980 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= bdev_nr_bytes(bdev)) |
8f32380d | 3981 | return ERR_PTR(-EINVAL); |
29c36d72 | 3982 | |
a05d3c91 QW |
3983 | if (drop_cache) { |
3984 | /* This should only be called with the primary sb. */ | |
3985 | ASSERT(copy_num == 0); | |
3986 | ||
3987 | /* | |
3988 | * Drop the page of the primary superblock, so later read will | |
3989 | * always read from the device. | |
3990 | */ | |
3991 | invalidate_inode_pages2_range(mapping, | |
3992 | bytenr >> PAGE_SHIFT, | |
3993 | (bytenr + BTRFS_SUPER_INFO_SIZE) >> PAGE_SHIFT); | |
3994 | } | |
3995 | ||
8f32380d JT |
3996 | page = read_cache_page_gfp(mapping, bytenr >> PAGE_SHIFT, GFP_NOFS); |
3997 | if (IS_ERR(page)) | |
3998 | return ERR_CAST(page); | |
29c36d72 | 3999 | |
8f32380d | 4000 | super = page_address(page); |
96c2e067 AJ |
4001 | if (btrfs_super_magic(super) != BTRFS_MAGIC) { |
4002 | btrfs_release_disk_super(super); | |
4003 | return ERR_PTR(-ENODATA); | |
4004 | } | |
4005 | ||
12659251 | 4006 | if (btrfs_super_bytenr(super) != bytenr_orig) { |
8f32380d JT |
4007 | btrfs_release_disk_super(super); |
4008 | return ERR_PTR(-EINVAL); | |
29c36d72 AJ |
4009 | } |
4010 | ||
8f32380d | 4011 | return super; |
29c36d72 AJ |
4012 | } |
4013 | ||
4014 | ||
8f32380d | 4015 | struct btrfs_super_block *btrfs_read_dev_super(struct block_device *bdev) |
a512bbf8 | 4016 | { |
8f32380d | 4017 | struct btrfs_super_block *super, *latest = NULL; |
a512bbf8 YZ |
4018 | int i; |
4019 | u64 transid = 0; | |
a512bbf8 YZ |
4020 | |
4021 | /* we would like to check all the supers, but that would make | |
4022 | * a btrfs mount succeed after a mkfs from a different FS. | |
4023 | * So, we need to add a special mount option to scan for | |
4024 | * later supers, using BTRFS_SUPER_MIRROR_MAX instead | |
4025 | */ | |
4026 | for (i = 0; i < 1; i++) { | |
a05d3c91 | 4027 | super = btrfs_read_dev_one_super(bdev, i, false); |
8f32380d | 4028 | if (IS_ERR(super)) |
a512bbf8 YZ |
4029 | continue; |
4030 | ||
a512bbf8 | 4031 | if (!latest || btrfs_super_generation(super) > transid) { |
8f32380d JT |
4032 | if (latest) |
4033 | btrfs_release_disk_super(super); | |
4034 | ||
4035 | latest = super; | |
a512bbf8 | 4036 | transid = btrfs_super_generation(super); |
a512bbf8 YZ |
4037 | } |
4038 | } | |
92fc03fb | 4039 | |
8f32380d | 4040 | return super; |
a512bbf8 YZ |
4041 | } |
4042 | ||
4eedeb75 | 4043 | /* |
abbb3b8e | 4044 | * Write superblock @sb to the @device. Do not wait for completion, all the |
314b6dd0 | 4045 | * pages we use for writing are locked. |
4eedeb75 | 4046 | * |
abbb3b8e DS |
4047 | * Write @max_mirrors copies of the superblock, where 0 means default that fit |
4048 | * the expected device size at commit time. Note that max_mirrors must be | |
4049 | * same for write and wait phases. | |
4eedeb75 | 4050 | * |
314b6dd0 | 4051 | * Return number of errors when page is not found or submission fails. |
4eedeb75 | 4052 | */ |
a512bbf8 | 4053 | static int write_dev_supers(struct btrfs_device *device, |
abbb3b8e | 4054 | struct btrfs_super_block *sb, int max_mirrors) |
a512bbf8 | 4055 | { |
d5178578 | 4056 | struct btrfs_fs_info *fs_info = device->fs_info; |
314b6dd0 | 4057 | struct address_space *mapping = device->bdev->bd_inode->i_mapping; |
d5178578 | 4058 | SHASH_DESC_ON_STACK(shash, fs_info->csum_shash); |
a512bbf8 | 4059 | int i; |
a512bbf8 | 4060 | int errors = 0; |
12659251 NA |
4061 | int ret; |
4062 | u64 bytenr, bytenr_orig; | |
a512bbf8 YZ |
4063 | |
4064 | if (max_mirrors == 0) | |
4065 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
4066 | ||
d5178578 JT |
4067 | shash->tfm = fs_info->csum_shash; |
4068 | ||
a512bbf8 | 4069 | for (i = 0; i < max_mirrors; i++) { |
314b6dd0 JT |
4070 | struct page *page; |
4071 | struct bio *bio; | |
4072 | struct btrfs_super_block *disk_super; | |
4073 | ||
12659251 NA |
4074 | bytenr_orig = btrfs_sb_offset(i); |
4075 | ret = btrfs_sb_log_location(device, i, WRITE, &bytenr); | |
4076 | if (ret == -ENOENT) { | |
4077 | continue; | |
4078 | } else if (ret < 0) { | |
4079 | btrfs_err(device->fs_info, | |
4080 | "couldn't get super block location for mirror %d", | |
4081 | i); | |
4082 | errors++; | |
4083 | continue; | |
4084 | } | |
935e5cc9 MX |
4085 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4086 | device->commit_total_bytes) | |
a512bbf8 YZ |
4087 | break; |
4088 | ||
12659251 | 4089 | btrfs_set_super_bytenr(sb, bytenr_orig); |
4eedeb75 | 4090 | |
fd08001f EB |
4091 | crypto_shash_digest(shash, (const char *)sb + BTRFS_CSUM_SIZE, |
4092 | BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, | |
4093 | sb->csum); | |
4eedeb75 | 4094 | |
314b6dd0 JT |
4095 | page = find_or_create_page(mapping, bytenr >> PAGE_SHIFT, |
4096 | GFP_NOFS); | |
4097 | if (!page) { | |
abbb3b8e | 4098 | btrfs_err(device->fs_info, |
314b6dd0 | 4099 | "couldn't get super block page for bytenr %llu", |
abbb3b8e DS |
4100 | bytenr); |
4101 | errors++; | |
4eedeb75 | 4102 | continue; |
abbb3b8e | 4103 | } |
634554dc | 4104 | |
314b6dd0 JT |
4105 | /* Bump the refcount for wait_dev_supers() */ |
4106 | get_page(page); | |
a512bbf8 | 4107 | |
314b6dd0 JT |
4108 | disk_super = page_address(page); |
4109 | memcpy(disk_super, sb, BTRFS_SUPER_INFO_SIZE); | |
4eedeb75 | 4110 | |
314b6dd0 JT |
4111 | /* |
4112 | * Directly use bios here instead of relying on the page cache | |
4113 | * to do I/O, so we don't lose the ability to do integrity | |
4114 | * checking. | |
4115 | */ | |
07888c66 CH |
4116 | bio = bio_alloc(device->bdev, 1, |
4117 | REQ_OP_WRITE | REQ_SYNC | REQ_META | REQ_PRIO, | |
4118 | GFP_NOFS); | |
314b6dd0 JT |
4119 | bio->bi_iter.bi_sector = bytenr >> SECTOR_SHIFT; |
4120 | bio->bi_private = device; | |
4121 | bio->bi_end_io = btrfs_end_super_write; | |
4122 | __bio_add_page(bio, page, BTRFS_SUPER_INFO_SIZE, | |
4123 | offset_in_page(bytenr)); | |
a512bbf8 | 4124 | |
387125fc | 4125 | /* |
314b6dd0 JT |
4126 | * We FUA only the first super block. The others we allow to |
4127 | * go down lazy and there's a short window where the on-disk | |
4128 | * copies might still contain the older version. | |
387125fc | 4129 | */ |
1b9e619c | 4130 | if (i == 0 && !btrfs_test_opt(device->fs_info, NOBARRIER)) |
314b6dd0 JT |
4131 | bio->bi_opf |= REQ_FUA; |
4132 | ||
58ff51f1 CH |
4133 | btrfsic_check_bio(bio); |
4134 | submit_bio(bio); | |
8376d9e1 NA |
4135 | |
4136 | if (btrfs_advance_sb_log(device, i)) | |
4137 | errors++; | |
a512bbf8 YZ |
4138 | } |
4139 | return errors < i ? 0 : -1; | |
4140 | } | |
4141 | ||
abbb3b8e DS |
4142 | /* |
4143 | * Wait for write completion of superblocks done by write_dev_supers, | |
4144 | * @max_mirrors same for write and wait phases. | |
4145 | * | |
314b6dd0 | 4146 | * Return number of errors when page is not found or not marked up to |
abbb3b8e DS |
4147 | * date. |
4148 | */ | |
4149 | static int wait_dev_supers(struct btrfs_device *device, int max_mirrors) | |
4150 | { | |
abbb3b8e DS |
4151 | int i; |
4152 | int errors = 0; | |
b6a535fa | 4153 | bool primary_failed = false; |
12659251 | 4154 | int ret; |
abbb3b8e DS |
4155 | u64 bytenr; |
4156 | ||
4157 | if (max_mirrors == 0) | |
4158 | max_mirrors = BTRFS_SUPER_MIRROR_MAX; | |
4159 | ||
4160 | for (i = 0; i < max_mirrors; i++) { | |
314b6dd0 JT |
4161 | struct page *page; |
4162 | ||
12659251 NA |
4163 | ret = btrfs_sb_log_location(device, i, READ, &bytenr); |
4164 | if (ret == -ENOENT) { | |
4165 | break; | |
4166 | } else if (ret < 0) { | |
4167 | errors++; | |
4168 | if (i == 0) | |
4169 | primary_failed = true; | |
4170 | continue; | |
4171 | } | |
abbb3b8e DS |
4172 | if (bytenr + BTRFS_SUPER_INFO_SIZE >= |
4173 | device->commit_total_bytes) | |
4174 | break; | |
4175 | ||
314b6dd0 JT |
4176 | page = find_get_page(device->bdev->bd_inode->i_mapping, |
4177 | bytenr >> PAGE_SHIFT); | |
4178 | if (!page) { | |
abbb3b8e | 4179 | errors++; |
b6a535fa HM |
4180 | if (i == 0) |
4181 | primary_failed = true; | |
abbb3b8e DS |
4182 | continue; |
4183 | } | |
314b6dd0 JT |
4184 | /* Page is submitted locked and unlocked once the IO completes */ |
4185 | wait_on_page_locked(page); | |
4186 | if (PageError(page)) { | |
abbb3b8e | 4187 | errors++; |
b6a535fa HM |
4188 | if (i == 0) |
4189 | primary_failed = true; | |
4190 | } | |
abbb3b8e | 4191 | |
314b6dd0 JT |
4192 | /* Drop our reference */ |
4193 | put_page(page); | |
abbb3b8e | 4194 | |
314b6dd0 JT |
4195 | /* Drop the reference from the writing run */ |
4196 | put_page(page); | |
abbb3b8e DS |
4197 | } |
4198 | ||
b6a535fa HM |
4199 | /* log error, force error return */ |
4200 | if (primary_failed) { | |
4201 | btrfs_err(device->fs_info, "error writing primary super block to device %llu", | |
4202 | device->devid); | |
4203 | return -1; | |
4204 | } | |
4205 | ||
abbb3b8e DS |
4206 | return errors < i ? 0 : -1; |
4207 | } | |
4208 | ||
387125fc CM |
4209 | /* |
4210 | * endio for the write_dev_flush, this will wake anyone waiting | |
4211 | * for the barrier when it is done | |
4212 | */ | |
4246a0b6 | 4213 | static void btrfs_end_empty_barrier(struct bio *bio) |
387125fc | 4214 | { |
f9e69aa9 | 4215 | bio_uninit(bio); |
e0ae9994 | 4216 | complete(bio->bi_private); |
387125fc CM |
4217 | } |
4218 | ||
4219 | /* | |
4fc6441a AJ |
4220 | * Submit a flush request to the device if it supports it. Error handling is |
4221 | * done in the waiting counterpart. | |
387125fc | 4222 | */ |
4fc6441a | 4223 | static void write_dev_flush(struct btrfs_device *device) |
387125fc | 4224 | { |
f9e69aa9 | 4225 | struct bio *bio = &device->flush_bio; |
387125fc | 4226 | |
a91cf0ff WY |
4227 | #ifndef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
4228 | /* | |
4229 | * When a disk has write caching disabled, we skip submission of a bio | |
4230 | * with flush and sync requests before writing the superblock, since | |
4231 | * it's not needed. However when the integrity checker is enabled, this | |
4232 | * results in reports that there are metadata blocks referred by a | |
4233 | * superblock that were not properly flushed. So don't skip the bio | |
4234 | * submission only when the integrity checker is enabled for the sake | |
4235 | * of simplicity, since this is a debug tool and not meant for use in | |
4236 | * non-debug builds. | |
4237 | */ | |
08e688fd | 4238 | if (!bdev_write_cache(device->bdev)) |
4fc6441a | 4239 | return; |
a91cf0ff | 4240 | #endif |
387125fc | 4241 | |
f9e69aa9 CH |
4242 | bio_init(bio, device->bdev, NULL, 0, |
4243 | REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH); | |
387125fc | 4244 | bio->bi_end_io = btrfs_end_empty_barrier; |
387125fc CM |
4245 | init_completion(&device->flush_wait); |
4246 | bio->bi_private = &device->flush_wait; | |
387125fc | 4247 | |
58ff51f1 CH |
4248 | btrfsic_check_bio(bio); |
4249 | submit_bio(bio); | |
1c3063b6 | 4250 | set_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
4fc6441a | 4251 | } |
387125fc | 4252 | |
4fc6441a AJ |
4253 | /* |
4254 | * If the flush bio has been submitted by write_dev_flush, wait for it. | |
4255 | */ | |
8c27cb35 | 4256 | static blk_status_t wait_dev_flush(struct btrfs_device *device) |
4fc6441a | 4257 | { |
f9e69aa9 | 4258 | struct bio *bio = &device->flush_bio; |
387125fc | 4259 | |
1c3063b6 | 4260 | if (!test_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state)) |
58efbc9f | 4261 | return BLK_STS_OK; |
387125fc | 4262 | |
1c3063b6 | 4263 | clear_bit(BTRFS_DEV_STATE_FLUSH_SENT, &device->dev_state); |
2980d574 | 4264 | wait_for_completion_io(&device->flush_wait); |
387125fc | 4265 | |
8c27cb35 | 4266 | return bio->bi_status; |
387125fc | 4267 | } |
387125fc | 4268 | |
d10b82fe | 4269 | static int check_barrier_error(struct btrfs_fs_info *fs_info) |
401b41e5 | 4270 | { |
6528b99d | 4271 | if (!btrfs_check_rw_degradable(fs_info, NULL)) |
401b41e5 | 4272 | return -EIO; |
387125fc CM |
4273 | return 0; |
4274 | } | |
4275 | ||
4276 | /* | |
4277 | * send an empty flush down to each device in parallel, | |
4278 | * then wait for them | |
4279 | */ | |
4280 | static int barrier_all_devices(struct btrfs_fs_info *info) | |
4281 | { | |
4282 | struct list_head *head; | |
4283 | struct btrfs_device *dev; | |
5af3e8cc | 4284 | int errors_wait = 0; |
4e4cbee9 | 4285 | blk_status_t ret; |
387125fc | 4286 | |
1538e6c5 | 4287 | lockdep_assert_held(&info->fs_devices->device_list_mutex); |
387125fc CM |
4288 | /* send down all the barriers */ |
4289 | head = &info->fs_devices->devices; | |
1538e6c5 | 4290 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4291 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4292 | continue; |
cea7c8bf | 4293 | if (!dev->bdev) |
387125fc | 4294 | continue; |
e12c9621 | 4295 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4296 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4297 | continue; |
4298 | ||
4fc6441a | 4299 | write_dev_flush(dev); |
58efbc9f | 4300 | dev->last_flush_error = BLK_STS_OK; |
387125fc CM |
4301 | } |
4302 | ||
4303 | /* wait for all the barriers */ | |
1538e6c5 | 4304 | list_for_each_entry(dev, head, dev_list) { |
e6e674bd | 4305 | if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) |
f88ba6a2 | 4306 | continue; |
387125fc | 4307 | if (!dev->bdev) { |
5af3e8cc | 4308 | errors_wait++; |
387125fc CM |
4309 | continue; |
4310 | } | |
e12c9621 | 4311 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4312 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
387125fc CM |
4313 | continue; |
4314 | ||
4fc6441a | 4315 | ret = wait_dev_flush(dev); |
401b41e5 AJ |
4316 | if (ret) { |
4317 | dev->last_flush_error = ret; | |
66b4993e DS |
4318 | btrfs_dev_stat_inc_and_print(dev, |
4319 | BTRFS_DEV_STAT_FLUSH_ERRS); | |
5af3e8cc | 4320 | errors_wait++; |
401b41e5 AJ |
4321 | } |
4322 | } | |
4323 | ||
cea7c8bf | 4324 | if (errors_wait) { |
401b41e5 AJ |
4325 | /* |
4326 | * At some point we need the status of all disks | |
4327 | * to arrive at the volume status. So error checking | |
4328 | * is being pushed to a separate loop. | |
4329 | */ | |
d10b82fe | 4330 | return check_barrier_error(info); |
387125fc | 4331 | } |
387125fc CM |
4332 | return 0; |
4333 | } | |
4334 | ||
943c6e99 ZL |
4335 | int btrfs_get_num_tolerated_disk_barrier_failures(u64 flags) |
4336 | { | |
8789f4fe ZL |
4337 | int raid_type; |
4338 | int min_tolerated = INT_MAX; | |
943c6e99 | 4339 | |
8789f4fe ZL |
4340 | if ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) == 0 || |
4341 | (flags & BTRFS_AVAIL_ALLOC_BIT_SINGLE)) | |
8c3e3582 | 4342 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4343 | btrfs_raid_array[BTRFS_RAID_SINGLE]. |
4344 | tolerated_failures); | |
943c6e99 | 4345 | |
8789f4fe ZL |
4346 | for (raid_type = 0; raid_type < BTRFS_NR_RAID_TYPES; raid_type++) { |
4347 | if (raid_type == BTRFS_RAID_SINGLE) | |
4348 | continue; | |
41a6e891 | 4349 | if (!(flags & btrfs_raid_array[raid_type].bg_flag)) |
8789f4fe | 4350 | continue; |
8c3e3582 | 4351 | min_tolerated = min_t(int, min_tolerated, |
8789f4fe ZL |
4352 | btrfs_raid_array[raid_type]. |
4353 | tolerated_failures); | |
4354 | } | |
943c6e99 | 4355 | |
8789f4fe | 4356 | if (min_tolerated == INT_MAX) { |
ab8d0fc4 | 4357 | pr_warn("BTRFS: unknown raid flag: %llu", flags); |
8789f4fe ZL |
4358 | min_tolerated = 0; |
4359 | } | |
4360 | ||
4361 | return min_tolerated; | |
943c6e99 ZL |
4362 | } |
4363 | ||
eece6a9c | 4364 | int write_all_supers(struct btrfs_fs_info *fs_info, int max_mirrors) |
f2984462 | 4365 | { |
e5e9a520 | 4366 | struct list_head *head; |
f2984462 | 4367 | struct btrfs_device *dev; |
a061fc8d | 4368 | struct btrfs_super_block *sb; |
f2984462 | 4369 | struct btrfs_dev_item *dev_item; |
f2984462 CM |
4370 | int ret; |
4371 | int do_barriers; | |
a236aed1 CM |
4372 | int max_errors; |
4373 | int total_errors = 0; | |
a061fc8d | 4374 | u64 flags; |
f2984462 | 4375 | |
0b246afa | 4376 | do_barriers = !btrfs_test_opt(fs_info, NOBARRIER); |
fed3b381 LB |
4377 | |
4378 | /* | |
4379 | * max_mirrors == 0 indicates we're from commit_transaction, | |
4380 | * not from fsync where the tree roots in fs_info have not | |
4381 | * been consistent on disk. | |
4382 | */ | |
4383 | if (max_mirrors == 0) | |
4384 | backup_super_roots(fs_info); | |
f2984462 | 4385 | |
0b246afa | 4386 | sb = fs_info->super_for_commit; |
a061fc8d | 4387 | dev_item = &sb->dev_item; |
e5e9a520 | 4388 | |
0b246afa JM |
4389 | mutex_lock(&fs_info->fs_devices->device_list_mutex); |
4390 | head = &fs_info->fs_devices->devices; | |
4391 | max_errors = btrfs_super_num_devices(fs_info->super_copy) - 1; | |
387125fc | 4392 | |
5af3e8cc | 4393 | if (do_barriers) { |
0b246afa | 4394 | ret = barrier_all_devices(fs_info); |
5af3e8cc SB |
4395 | if (ret) { |
4396 | mutex_unlock( | |
0b246afa JM |
4397 | &fs_info->fs_devices->device_list_mutex); |
4398 | btrfs_handle_fs_error(fs_info, ret, | |
4399 | "errors while submitting device barriers."); | |
5af3e8cc SB |
4400 | return ret; |
4401 | } | |
4402 | } | |
387125fc | 4403 | |
1538e6c5 | 4404 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4405 | if (!dev->bdev) { |
4406 | total_errors++; | |
4407 | continue; | |
4408 | } | |
e12c9621 | 4409 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4410 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4411 | continue; |
4412 | ||
2b82032c | 4413 | btrfs_set_stack_device_generation(dev_item, 0); |
a061fc8d CM |
4414 | btrfs_set_stack_device_type(dev_item, dev->type); |
4415 | btrfs_set_stack_device_id(dev_item, dev->devid); | |
7df69d3e | 4416 | btrfs_set_stack_device_total_bytes(dev_item, |
935e5cc9 | 4417 | dev->commit_total_bytes); |
ce7213c7 MX |
4418 | btrfs_set_stack_device_bytes_used(dev_item, |
4419 | dev->commit_bytes_used); | |
a061fc8d CM |
4420 | btrfs_set_stack_device_io_align(dev_item, dev->io_align); |
4421 | btrfs_set_stack_device_io_width(dev_item, dev->io_width); | |
4422 | btrfs_set_stack_device_sector_size(dev_item, dev->sector_size); | |
4423 | memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE); | |
7239ff4b NB |
4424 | memcpy(dev_item->fsid, dev->fs_devices->metadata_uuid, |
4425 | BTRFS_FSID_SIZE); | |
a512bbf8 | 4426 | |
a061fc8d CM |
4427 | flags = btrfs_super_flags(sb); |
4428 | btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN); | |
4429 | ||
75cb857d QW |
4430 | ret = btrfs_validate_write_super(fs_info, sb); |
4431 | if (ret < 0) { | |
4432 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
4433 | btrfs_handle_fs_error(fs_info, -EUCLEAN, | |
4434 | "unexpected superblock corruption detected"); | |
4435 | return -EUCLEAN; | |
4436 | } | |
4437 | ||
abbb3b8e | 4438 | ret = write_dev_supers(dev, sb, max_mirrors); |
a236aed1 CM |
4439 | if (ret) |
4440 | total_errors++; | |
f2984462 | 4441 | } |
a236aed1 | 4442 | if (total_errors > max_errors) { |
0b246afa JM |
4443 | btrfs_err(fs_info, "%d errors while writing supers", |
4444 | total_errors); | |
4445 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
79787eaa | 4446 | |
9d565ba4 | 4447 | /* FUA is masked off if unsupported and can't be the reason */ |
0b246afa JM |
4448 | btrfs_handle_fs_error(fs_info, -EIO, |
4449 | "%d errors while writing supers", | |
4450 | total_errors); | |
9d565ba4 | 4451 | return -EIO; |
a236aed1 | 4452 | } |
f2984462 | 4453 | |
a512bbf8 | 4454 | total_errors = 0; |
1538e6c5 | 4455 | list_for_each_entry(dev, head, dev_list) { |
dfe25020 CM |
4456 | if (!dev->bdev) |
4457 | continue; | |
e12c9621 | 4458 | if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) || |
ebbede42 | 4459 | !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) |
dfe25020 CM |
4460 | continue; |
4461 | ||
abbb3b8e | 4462 | ret = wait_dev_supers(dev, max_mirrors); |
a512bbf8 YZ |
4463 | if (ret) |
4464 | total_errors++; | |
f2984462 | 4465 | } |
0b246afa | 4466 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); |
a236aed1 | 4467 | if (total_errors > max_errors) { |
0b246afa JM |
4468 | btrfs_handle_fs_error(fs_info, -EIO, |
4469 | "%d errors while writing supers", | |
4470 | total_errors); | |
79787eaa | 4471 | return -EIO; |
a236aed1 | 4472 | } |
f2984462 CM |
4473 | return 0; |
4474 | } | |
4475 | ||
cb517eab MX |
4476 | /* Drop a fs root from the radix tree and free it. */ |
4477 | void btrfs_drop_and_free_fs_root(struct btrfs_fs_info *fs_info, | |
4478 | struct btrfs_root *root) | |
2619ba1f | 4479 | { |
4785e24f JB |
4480 | bool drop_ref = false; |
4481 | ||
fc7cbcd4 DS |
4482 | spin_lock(&fs_info->fs_roots_radix_lock); |
4483 | radix_tree_delete(&fs_info->fs_roots_radix, | |
4484 | (unsigned long)root->root_key.objectid); | |
4485 | if (test_and_clear_bit(BTRFS_ROOT_IN_RADIX, &root->state)) | |
4785e24f | 4486 | drop_ref = true; |
fc7cbcd4 | 4487 | spin_unlock(&fs_info->fs_roots_radix_lock); |
76dda93c | 4488 | |
84961539 | 4489 | if (BTRFS_FS_ERROR(fs_info)) { |
ef67963d | 4490 | ASSERT(root->log_root == NULL); |
1c1ea4f7 | 4491 | if (root->reloc_root) { |
00246528 | 4492 | btrfs_put_root(root->reloc_root); |
1c1ea4f7 LB |
4493 | root->reloc_root = NULL; |
4494 | } | |
4495 | } | |
3321719e | 4496 | |
4785e24f JB |
4497 | if (drop_ref) |
4498 | btrfs_put_root(root); | |
2619ba1f CM |
4499 | } |
4500 | ||
c146afad | 4501 | int btrfs_cleanup_fs_roots(struct btrfs_fs_info *fs_info) |
cfaa7295 | 4502 | { |
fc7cbcd4 DS |
4503 | u64 root_objectid = 0; |
4504 | struct btrfs_root *gang[8]; | |
4505 | int i = 0; | |
65d33fd7 | 4506 | int err = 0; |
fc7cbcd4 | 4507 | unsigned int ret = 0; |
e089f05c | 4508 | |
c146afad | 4509 | while (1) { |
fc7cbcd4 DS |
4510 | spin_lock(&fs_info->fs_roots_radix_lock); |
4511 | ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4512 | (void **)gang, root_objectid, | |
4513 | ARRAY_SIZE(gang)); | |
4514 | if (!ret) { | |
4515 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4516 | break; | |
65d33fd7 | 4517 | } |
fc7cbcd4 | 4518 | root_objectid = gang[ret - 1]->root_key.objectid + 1; |
65d33fd7 | 4519 | |
fc7cbcd4 DS |
4520 | for (i = 0; i < ret; i++) { |
4521 | /* Avoid to grab roots in dead_roots */ | |
4522 | if (btrfs_root_refs(&gang[i]->root_item) == 0) { | |
4523 | gang[i] = NULL; | |
4524 | continue; | |
4525 | } | |
4526 | /* grab all the search result for later use */ | |
4527 | gang[i] = btrfs_grab_root(gang[i]); | |
65d33fd7 | 4528 | } |
fc7cbcd4 | 4529 | spin_unlock(&fs_info->fs_roots_radix_lock); |
66b4ffd1 | 4530 | |
fc7cbcd4 DS |
4531 | for (i = 0; i < ret; i++) { |
4532 | if (!gang[i]) | |
65d33fd7 | 4533 | continue; |
fc7cbcd4 DS |
4534 | root_objectid = gang[i]->root_key.objectid; |
4535 | err = btrfs_orphan_cleanup(gang[i]); | |
66b4ffd1 | 4536 | if (err) |
fc7cbcd4 DS |
4537 | break; |
4538 | btrfs_put_root(gang[i]); | |
c146afad | 4539 | } |
fc7cbcd4 | 4540 | root_objectid++; |
c146afad | 4541 | } |
65d33fd7 | 4542 | |
fc7cbcd4 DS |
4543 | /* release the uncleaned roots due to error */ |
4544 | for (; i < ret; i++) { | |
4545 | if (gang[i]) | |
4546 | btrfs_put_root(gang[i]); | |
65d33fd7 QW |
4547 | } |
4548 | return err; | |
c146afad | 4549 | } |
a2135011 | 4550 | |
6bccf3ab | 4551 | int btrfs_commit_super(struct btrfs_fs_info *fs_info) |
c146afad | 4552 | { |
6bccf3ab | 4553 | struct btrfs_root *root = fs_info->tree_root; |
c146afad | 4554 | struct btrfs_trans_handle *trans; |
a74a4b97 | 4555 | |
0b246afa | 4556 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4557 | btrfs_run_delayed_iputs(fs_info); |
0b246afa JM |
4558 | mutex_unlock(&fs_info->cleaner_mutex); |
4559 | wake_up_process(fs_info->cleaner_kthread); | |
c71bf099 YZ |
4560 | |
4561 | /* wait until ongoing cleanup work done */ | |
0b246afa JM |
4562 | down_write(&fs_info->cleanup_work_sem); |
4563 | up_write(&fs_info->cleanup_work_sem); | |
c71bf099 | 4564 | |
7a7eaa40 | 4565 | trans = btrfs_join_transaction(root); |
3612b495 TI |
4566 | if (IS_ERR(trans)) |
4567 | return PTR_ERR(trans); | |
3a45bb20 | 4568 | return btrfs_commit_transaction(trans); |
c146afad YZ |
4569 | } |
4570 | ||
36c86a9e QW |
4571 | static void warn_about_uncommitted_trans(struct btrfs_fs_info *fs_info) |
4572 | { | |
4573 | struct btrfs_transaction *trans; | |
4574 | struct btrfs_transaction *tmp; | |
4575 | bool found = false; | |
4576 | ||
4577 | if (list_empty(&fs_info->trans_list)) | |
4578 | return; | |
4579 | ||
4580 | /* | |
4581 | * This function is only called at the very end of close_ctree(), | |
4582 | * thus no other running transaction, no need to take trans_lock. | |
4583 | */ | |
4584 | ASSERT(test_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags)); | |
4585 | list_for_each_entry_safe(trans, tmp, &fs_info->trans_list, list) { | |
4586 | struct extent_state *cached = NULL; | |
4587 | u64 dirty_bytes = 0; | |
4588 | u64 cur = 0; | |
4589 | u64 found_start; | |
4590 | u64 found_end; | |
4591 | ||
4592 | found = true; | |
4593 | while (!find_first_extent_bit(&trans->dirty_pages, cur, | |
4594 | &found_start, &found_end, EXTENT_DIRTY, &cached)) { | |
4595 | dirty_bytes += found_end + 1 - found_start; | |
4596 | cur = found_end + 1; | |
4597 | } | |
4598 | btrfs_warn(fs_info, | |
4599 | "transaction %llu (with %llu dirty metadata bytes) is not committed", | |
4600 | trans->transid, dirty_bytes); | |
4601 | btrfs_cleanup_one_transaction(trans, fs_info); | |
4602 | ||
4603 | if (trans == fs_info->running_transaction) | |
4604 | fs_info->running_transaction = NULL; | |
4605 | list_del_init(&trans->list); | |
4606 | ||
4607 | btrfs_put_transaction(trans); | |
4608 | trace_btrfs_transaction_commit(fs_info); | |
4609 | } | |
4610 | ASSERT(!found); | |
4611 | } | |
4612 | ||
b105e927 | 4613 | void __cold close_ctree(struct btrfs_fs_info *fs_info) |
c146afad | 4614 | { |
c146afad YZ |
4615 | int ret; |
4616 | ||
afcdd129 | 4617 | set_bit(BTRFS_FS_CLOSING_START, &fs_info->flags); |
31e70e52 | 4618 | |
8a1f1e3d FM |
4619 | /* |
4620 | * If we had UNFINISHED_DROPS we could still be processing them, so | |
4621 | * clear that bit and wake up relocation so it can stop. | |
4622 | * We must do this before stopping the block group reclaim task, because | |
4623 | * at btrfs_relocate_block_group() we wait for this bit, and after the | |
4624 | * wait we stop with -EINTR if btrfs_fs_closing() returns non-zero - we | |
4625 | * have just set BTRFS_FS_CLOSING_START, so btrfs_fs_closing() will | |
4626 | * return 1. | |
4627 | */ | |
4628 | btrfs_wake_unfinished_drop(fs_info); | |
4629 | ||
31e70e52 FM |
4630 | /* |
4631 | * We may have the reclaim task running and relocating a data block group, | |
4632 | * in which case it may create delayed iputs. So stop it before we park | |
4633 | * the cleaner kthread otherwise we can get new delayed iputs after | |
4634 | * parking the cleaner, and that can make the async reclaim task to hang | |
4635 | * if it's waiting for delayed iputs to complete, since the cleaner is | |
4636 | * parked and can not run delayed iputs - this will make us hang when | |
4637 | * trying to stop the async reclaim task. | |
4638 | */ | |
4639 | cancel_work_sync(&fs_info->reclaim_bgs_work); | |
d6fd0ae2 OS |
4640 | /* |
4641 | * We don't want the cleaner to start new transactions, add more delayed | |
4642 | * iputs, etc. while we're closing. We can't use kthread_stop() yet | |
4643 | * because that frees the task_struct, and the transaction kthread might | |
4644 | * still try to wake up the cleaner. | |
4645 | */ | |
4646 | kthread_park(fs_info->cleaner_kthread); | |
c146afad | 4647 | |
7343dd61 | 4648 | /* wait for the qgroup rescan worker to stop */ |
d06f23d6 | 4649 | btrfs_qgroup_wait_for_completion(fs_info, false); |
7343dd61 | 4650 | |
803b2f54 SB |
4651 | /* wait for the uuid_scan task to finish */ |
4652 | down(&fs_info->uuid_tree_rescan_sem); | |
4653 | /* avoid complains from lockdep et al., set sem back to initial state */ | |
4654 | up(&fs_info->uuid_tree_rescan_sem); | |
4655 | ||
837d5b6e | 4656 | /* pause restriper - we want to resume on mount */ |
aa1b8cd4 | 4657 | btrfs_pause_balance(fs_info); |
837d5b6e | 4658 | |
8dabb742 SB |
4659 | btrfs_dev_replace_suspend_for_unmount(fs_info); |
4660 | ||
aa1b8cd4 | 4661 | btrfs_scrub_cancel(fs_info); |
4cb5300b CM |
4662 | |
4663 | /* wait for any defraggers to finish */ | |
4664 | wait_event(fs_info->transaction_wait, | |
4665 | (atomic_read(&fs_info->defrag_running) == 0)); | |
4666 | ||
4667 | /* clear out the rbtree of defraggable inodes */ | |
26176e7c | 4668 | btrfs_cleanup_defrag_inodes(fs_info); |
4cb5300b | 4669 | |
a362bb86 FM |
4670 | /* |
4671 | * After we parked the cleaner kthread, ordered extents may have | |
4672 | * completed and created new delayed iputs. If one of the async reclaim | |
4673 | * tasks is running and in the RUN_DELAYED_IPUTS flush state, then we | |
4674 | * can hang forever trying to stop it, because if a delayed iput is | |
4675 | * added after it ran btrfs_run_delayed_iputs() and before it called | |
4676 | * btrfs_wait_on_delayed_iputs(), it will hang forever since there is | |
4677 | * no one else to run iputs. | |
4678 | * | |
4679 | * So wait for all ongoing ordered extents to complete and then run | |
4680 | * delayed iputs. This works because once we reach this point no one | |
4681 | * can either create new ordered extents nor create delayed iputs | |
4682 | * through some other means. | |
4683 | * | |
4684 | * Also note that btrfs_wait_ordered_roots() is not safe here, because | |
4685 | * it waits for BTRFS_ORDERED_COMPLETE to be set on an ordered extent, | |
4686 | * but the delayed iput for the respective inode is made only when doing | |
4687 | * the final btrfs_put_ordered_extent() (which must happen at | |
4688 | * btrfs_finish_ordered_io() when we are unmounting). | |
4689 | */ | |
4690 | btrfs_flush_workqueue(fs_info->endio_write_workers); | |
4691 | /* Ordered extents for free space inodes. */ | |
4692 | btrfs_flush_workqueue(fs_info->endio_freespace_worker); | |
4693 | btrfs_run_delayed_iputs(fs_info); | |
4694 | ||
21c7e756 | 4695 | cancel_work_sync(&fs_info->async_reclaim_work); |
57056740 | 4696 | cancel_work_sync(&fs_info->async_data_reclaim_work); |
576fa348 | 4697 | cancel_work_sync(&fs_info->preempt_reclaim_work); |
21c7e756 | 4698 | |
b0643e59 DZ |
4699 | /* Cancel or finish ongoing discard work */ |
4700 | btrfs_discard_cleanup(fs_info); | |
4701 | ||
bc98a42c | 4702 | if (!sb_rdonly(fs_info->sb)) { |
e44163e1 | 4703 | /* |
d6fd0ae2 OS |
4704 | * The cleaner kthread is stopped, so do one final pass over |
4705 | * unused block groups. | |
e44163e1 | 4706 | */ |
0b246afa | 4707 | btrfs_delete_unused_bgs(fs_info); |
e44163e1 | 4708 | |
f0cc2cd7 FM |
4709 | /* |
4710 | * There might be existing delayed inode workers still running | |
4711 | * and holding an empty delayed inode item. We must wait for | |
4712 | * them to complete first because they can create a transaction. | |
4713 | * This happens when someone calls btrfs_balance_delayed_items() | |
4714 | * and then a transaction commit runs the same delayed nodes | |
4715 | * before any delayed worker has done something with the nodes. | |
4716 | * We must wait for any worker here and not at transaction | |
4717 | * commit time since that could cause a deadlock. | |
4718 | * This is a very rare case. | |
4719 | */ | |
4720 | btrfs_flush_workqueue(fs_info->delayed_workers); | |
4721 | ||
6bccf3ab | 4722 | ret = btrfs_commit_super(fs_info); |
acce952b | 4723 | if (ret) |
04892340 | 4724 | btrfs_err(fs_info, "commit super ret %d", ret); |
acce952b | 4725 | } |
4726 | ||
84961539 | 4727 | if (BTRFS_FS_ERROR(fs_info)) |
2ff7e61e | 4728 | btrfs_error_commit_super(fs_info); |
0f7d52f4 | 4729 | |
e3029d9f AV |
4730 | kthread_stop(fs_info->transaction_kthread); |
4731 | kthread_stop(fs_info->cleaner_kthread); | |
8929ecfa | 4732 | |
e187831e | 4733 | ASSERT(list_empty(&fs_info->delayed_iputs)); |
afcdd129 | 4734 | set_bit(BTRFS_FS_CLOSING_DONE, &fs_info->flags); |
f25784b3 | 4735 | |
5958253c QW |
4736 | if (btrfs_check_quota_leak(fs_info)) { |
4737 | WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); | |
4738 | btrfs_err(fs_info, "qgroup reserved space leaked"); | |
4739 | } | |
4740 | ||
04892340 | 4741 | btrfs_free_qgroup_config(fs_info); |
fe816d0f | 4742 | ASSERT(list_empty(&fs_info->delalloc_roots)); |
bcef60f2 | 4743 | |
963d678b | 4744 | if (percpu_counter_sum(&fs_info->delalloc_bytes)) { |
04892340 | 4745 | btrfs_info(fs_info, "at unmount delalloc count %lld", |
963d678b | 4746 | percpu_counter_sum(&fs_info->delalloc_bytes)); |
b0c68f8b | 4747 | } |
bcc63abb | 4748 | |
5deb17e1 | 4749 | if (percpu_counter_sum(&fs_info->ordered_bytes)) |
4297ff84 | 4750 | btrfs_info(fs_info, "at unmount dio bytes count %lld", |
5deb17e1 | 4751 | percpu_counter_sum(&fs_info->ordered_bytes)); |
4297ff84 | 4752 | |
6618a59b | 4753 | btrfs_sysfs_remove_mounted(fs_info); |
b7c35e81 | 4754 | btrfs_sysfs_remove_fsid(fs_info->fs_devices); |
5ac1d209 | 4755 | |
1a4319cc LB |
4756 | btrfs_put_block_group_cache(fs_info); |
4757 | ||
de348ee0 WS |
4758 | /* |
4759 | * we must make sure there is not any read request to | |
4760 | * submit after we stopping all workers. | |
4761 | */ | |
4762 | invalidate_inode_pages2(fs_info->btree_inode->i_mapping); | |
96192499 JB |
4763 | btrfs_stop_all_workers(fs_info); |
4764 | ||
0a31daa4 | 4765 | /* We shouldn't have any transaction open at this point */ |
36c86a9e | 4766 | warn_about_uncommitted_trans(fs_info); |
0a31daa4 | 4767 | |
afcdd129 | 4768 | clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
4273eaff | 4769 | free_root_pointers(fs_info, true); |
8c38938c | 4770 | btrfs_free_fs_roots(fs_info); |
9ad6b7bc | 4771 | |
4e19443d JB |
4772 | /* |
4773 | * We must free the block groups after dropping the fs_roots as we could | |
4774 | * have had an IO error and have left over tree log blocks that aren't | |
4775 | * cleaned up until the fs roots are freed. This makes the block group | |
4776 | * accounting appear to be wrong because there's pending reserved bytes, | |
4777 | * so make sure we do the block group cleanup afterwards. | |
4778 | */ | |
4779 | btrfs_free_block_groups(fs_info); | |
4780 | ||
13e6c37b | 4781 | iput(fs_info->btree_inode); |
d6bfde87 | 4782 | |
21adbd5c | 4783 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
0b246afa | 4784 | if (btrfs_test_opt(fs_info, CHECK_INTEGRITY)) |
2ff7e61e | 4785 | btrfsic_unmount(fs_info->fs_devices); |
21adbd5c SB |
4786 | #endif |
4787 | ||
0b86a832 | 4788 | btrfs_mapping_tree_free(&fs_info->mapping_tree); |
68c94e55 | 4789 | btrfs_close_devices(fs_info->fs_devices); |
eb60ceac CM |
4790 | } |
4791 | ||
b9fab919 CM |
4792 | int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid, |
4793 | int atomic) | |
5f39d397 | 4794 | { |
1259ab75 | 4795 | int ret; |
727011e0 | 4796 | struct inode *btree_inode = buf->pages[0]->mapping->host; |
1259ab75 | 4797 | |
0b32f4bb | 4798 | ret = extent_buffer_uptodate(buf); |
1259ab75 CM |
4799 | if (!ret) |
4800 | return ret; | |
4801 | ||
4802 | ret = verify_parent_transid(&BTRFS_I(btree_inode)->io_tree, buf, | |
b9fab919 CM |
4803 | parent_transid, atomic); |
4804 | if (ret == -EAGAIN) | |
4805 | return ret; | |
1259ab75 | 4806 | return !ret; |
5f39d397 CM |
4807 | } |
4808 | ||
5f39d397 CM |
4809 | void btrfs_mark_buffer_dirty(struct extent_buffer *buf) |
4810 | { | |
2f4d60df | 4811 | struct btrfs_fs_info *fs_info = buf->fs_info; |
5f39d397 | 4812 | u64 transid = btrfs_header_generation(buf); |
b9473439 | 4813 | int was_dirty; |
b4ce94de | 4814 | |
06ea65a3 JB |
4815 | #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS |
4816 | /* | |
4817 | * This is a fast path so only do this check if we have sanity tests | |
52042d8e | 4818 | * enabled. Normal people shouldn't be using unmapped buffers as dirty |
06ea65a3 JB |
4819 | * outside of the sanity tests. |
4820 | */ | |
b0132a3b | 4821 | if (unlikely(test_bit(EXTENT_BUFFER_UNMAPPED, &buf->bflags))) |
06ea65a3 JB |
4822 | return; |
4823 | #endif | |
49d0c642 | 4824 | btrfs_assert_tree_write_locked(buf); |
0b246afa | 4825 | if (transid != fs_info->generation) |
5d163e0e | 4826 | WARN(1, KERN_CRIT "btrfs transid mismatch buffer %llu, found %llu running %llu\n", |
0b246afa | 4827 | buf->start, transid, fs_info->generation); |
0b32f4bb | 4828 | was_dirty = set_extent_buffer_dirty(buf); |
e2d84521 | 4829 | if (!was_dirty) |
104b4e51 NB |
4830 | percpu_counter_add_batch(&fs_info->dirty_metadata_bytes, |
4831 | buf->len, | |
4832 | fs_info->dirty_metadata_batch); | |
1f21ef0a | 4833 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY |
69fc6cbb QW |
4834 | /* |
4835 | * Since btrfs_mark_buffer_dirty() can be called with item pointer set | |
4836 | * but item data not updated. | |
4837 | * So here we should only check item pointers, not item data. | |
4838 | */ | |
4839 | if (btrfs_header_level(buf) == 0 && | |
cfdaad5e | 4840 | btrfs_check_leaf_relaxed(buf)) { |
a4f78750 | 4841 | btrfs_print_leaf(buf); |
1f21ef0a FM |
4842 | ASSERT(0); |
4843 | } | |
4844 | #endif | |
eb60ceac CM |
4845 | } |
4846 | ||
2ff7e61e | 4847 | static void __btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info, |
b53d3f5d | 4848 | int flush_delayed) |
16cdcec7 MX |
4849 | { |
4850 | /* | |
4851 | * looks as though older kernels can get into trouble with | |
4852 | * this code, they end up stuck in balance_dirty_pages forever | |
4853 | */ | |
e2d84521 | 4854 | int ret; |
16cdcec7 MX |
4855 | |
4856 | if (current->flags & PF_MEMALLOC) | |
4857 | return; | |
4858 | ||
b53d3f5d | 4859 | if (flush_delayed) |
2ff7e61e | 4860 | btrfs_balance_delayed_items(fs_info); |
16cdcec7 | 4861 | |
d814a491 EL |
4862 | ret = __percpu_counter_compare(&fs_info->dirty_metadata_bytes, |
4863 | BTRFS_DIRTY_METADATA_THRESH, | |
4864 | fs_info->dirty_metadata_batch); | |
e2d84521 | 4865 | if (ret > 0) { |
0b246afa | 4866 | balance_dirty_pages_ratelimited(fs_info->btree_inode->i_mapping); |
16cdcec7 | 4867 | } |
16cdcec7 MX |
4868 | } |
4869 | ||
2ff7e61e | 4870 | void btrfs_btree_balance_dirty(struct btrfs_fs_info *fs_info) |
35b7e476 | 4871 | { |
2ff7e61e | 4872 | __btrfs_btree_balance_dirty(fs_info, 1); |
b53d3f5d | 4873 | } |
585ad2c3 | 4874 | |
2ff7e61e | 4875 | void btrfs_btree_balance_dirty_nodelay(struct btrfs_fs_info *fs_info) |
b53d3f5d | 4876 | { |
2ff7e61e | 4877 | __btrfs_btree_balance_dirty(fs_info, 0); |
35b7e476 | 4878 | } |
6b80053d | 4879 | |
2ff7e61e | 4880 | static void btrfs_error_commit_super(struct btrfs_fs_info *fs_info) |
acce952b | 4881 | { |
fe816d0f NB |
4882 | /* cleanup FS via transaction */ |
4883 | btrfs_cleanup_transaction(fs_info); | |
4884 | ||
0b246afa | 4885 | mutex_lock(&fs_info->cleaner_mutex); |
2ff7e61e | 4886 | btrfs_run_delayed_iputs(fs_info); |
0b246afa | 4887 | mutex_unlock(&fs_info->cleaner_mutex); |
acce952b | 4888 | |
0b246afa JM |
4889 | down_write(&fs_info->cleanup_work_sem); |
4890 | up_write(&fs_info->cleanup_work_sem); | |
acce952b | 4891 | } |
4892 | ||
ef67963d JB |
4893 | static void btrfs_drop_all_logs(struct btrfs_fs_info *fs_info) |
4894 | { | |
fc7cbcd4 DS |
4895 | struct btrfs_root *gang[8]; |
4896 | u64 root_objectid = 0; | |
4897 | int ret; | |
4898 | ||
4899 | spin_lock(&fs_info->fs_roots_radix_lock); | |
4900 | while ((ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix, | |
4901 | (void **)gang, root_objectid, | |
4902 | ARRAY_SIZE(gang))) != 0) { | |
4903 | int i; | |
4904 | ||
4905 | for (i = 0; i < ret; i++) | |
4906 | gang[i] = btrfs_grab_root(gang[i]); | |
4907 | spin_unlock(&fs_info->fs_roots_radix_lock); | |
4908 | ||
4909 | for (i = 0; i < ret; i++) { | |
4910 | if (!gang[i]) | |
ef67963d | 4911 | continue; |
fc7cbcd4 DS |
4912 | root_objectid = gang[i]->root_key.objectid; |
4913 | btrfs_free_log(NULL, gang[i]); | |
4914 | btrfs_put_root(gang[i]); | |
ef67963d | 4915 | } |
fc7cbcd4 DS |
4916 | root_objectid++; |
4917 | spin_lock(&fs_info->fs_roots_radix_lock); | |
ef67963d | 4918 | } |
fc7cbcd4 | 4919 | spin_unlock(&fs_info->fs_roots_radix_lock); |
ef67963d JB |
4920 | btrfs_free_log_root_tree(NULL, fs_info); |
4921 | } | |
4922 | ||
143bede5 | 4923 | static void btrfs_destroy_ordered_extents(struct btrfs_root *root) |
acce952b | 4924 | { |
acce952b | 4925 | struct btrfs_ordered_extent *ordered; |
acce952b | 4926 | |
199c2a9c | 4927 | spin_lock(&root->ordered_extent_lock); |
779880ef JB |
4928 | /* |
4929 | * This will just short circuit the ordered completion stuff which will | |
4930 | * make sure the ordered extent gets properly cleaned up. | |
4931 | */ | |
199c2a9c | 4932 | list_for_each_entry(ordered, &root->ordered_extents, |
779880ef JB |
4933 | root_extent_list) |
4934 | set_bit(BTRFS_ORDERED_IOERR, &ordered->flags); | |
199c2a9c MX |
4935 | spin_unlock(&root->ordered_extent_lock); |
4936 | } | |
4937 | ||
4938 | static void btrfs_destroy_all_ordered_extents(struct btrfs_fs_info *fs_info) | |
4939 | { | |
4940 | struct btrfs_root *root; | |
4941 | struct list_head splice; | |
4942 | ||
4943 | INIT_LIST_HEAD(&splice); | |
4944 | ||
4945 | spin_lock(&fs_info->ordered_root_lock); | |
4946 | list_splice_init(&fs_info->ordered_roots, &splice); | |
4947 | while (!list_empty(&splice)) { | |
4948 | root = list_first_entry(&splice, struct btrfs_root, | |
4949 | ordered_root); | |
1de2cfde JB |
4950 | list_move_tail(&root->ordered_root, |
4951 | &fs_info->ordered_roots); | |
199c2a9c | 4952 | |
2a85d9ca | 4953 | spin_unlock(&fs_info->ordered_root_lock); |
199c2a9c MX |
4954 | btrfs_destroy_ordered_extents(root); |
4955 | ||
2a85d9ca LB |
4956 | cond_resched(); |
4957 | spin_lock(&fs_info->ordered_root_lock); | |
199c2a9c MX |
4958 | } |
4959 | spin_unlock(&fs_info->ordered_root_lock); | |
74d5d229 JB |
4960 | |
4961 | /* | |
4962 | * We need this here because if we've been flipped read-only we won't | |
4963 | * get sync() from the umount, so we need to make sure any ordered | |
4964 | * extents that haven't had their dirty pages IO start writeout yet | |
4965 | * actually get run and error out properly. | |
4966 | */ | |
4967 | btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); | |
acce952b | 4968 | } |
4969 | ||
35a3621b | 4970 | static int btrfs_destroy_delayed_refs(struct btrfs_transaction *trans, |
2ff7e61e | 4971 | struct btrfs_fs_info *fs_info) |
acce952b | 4972 | { |
4973 | struct rb_node *node; | |
4974 | struct btrfs_delayed_ref_root *delayed_refs; | |
4975 | struct btrfs_delayed_ref_node *ref; | |
4976 | int ret = 0; | |
4977 | ||
4978 | delayed_refs = &trans->delayed_refs; | |
4979 | ||
4980 | spin_lock(&delayed_refs->lock); | |
d7df2c79 | 4981 | if (atomic_read(&delayed_refs->num_entries) == 0) { |
cfece4db | 4982 | spin_unlock(&delayed_refs->lock); |
b79ce3dd | 4983 | btrfs_debug(fs_info, "delayed_refs has NO entry"); |
acce952b | 4984 | return ret; |
4985 | } | |
4986 | ||
5c9d028b | 4987 | while ((node = rb_first_cached(&delayed_refs->href_root)) != NULL) { |
d7df2c79 | 4988 | struct btrfs_delayed_ref_head *head; |
0e0adbcf | 4989 | struct rb_node *n; |
e78417d1 | 4990 | bool pin_bytes = false; |
acce952b | 4991 | |
d7df2c79 JB |
4992 | head = rb_entry(node, struct btrfs_delayed_ref_head, |
4993 | href_node); | |
3069bd26 | 4994 | if (btrfs_delayed_ref_lock(delayed_refs, head)) |
d7df2c79 | 4995 | continue; |
3069bd26 | 4996 | |
d7df2c79 | 4997 | spin_lock(&head->lock); |
e3d03965 | 4998 | while ((n = rb_first_cached(&head->ref_tree)) != NULL) { |
0e0adbcf JB |
4999 | ref = rb_entry(n, struct btrfs_delayed_ref_node, |
5000 | ref_node); | |
d7df2c79 | 5001 | ref->in_tree = 0; |
e3d03965 | 5002 | rb_erase_cached(&ref->ref_node, &head->ref_tree); |
0e0adbcf | 5003 | RB_CLEAR_NODE(&ref->ref_node); |
1d57ee94 WX |
5004 | if (!list_empty(&ref->add_list)) |
5005 | list_del(&ref->add_list); | |
d7df2c79 JB |
5006 | atomic_dec(&delayed_refs->num_entries); |
5007 | btrfs_put_delayed_ref(ref); | |
e78417d1 | 5008 | } |
d7df2c79 JB |
5009 | if (head->must_insert_reserved) |
5010 | pin_bytes = true; | |
5011 | btrfs_free_delayed_extent_op(head->extent_op); | |
fa781cea | 5012 | btrfs_delete_ref_head(delayed_refs, head); |
d7df2c79 JB |
5013 | spin_unlock(&head->lock); |
5014 | spin_unlock(&delayed_refs->lock); | |
5015 | mutex_unlock(&head->mutex); | |
acce952b | 5016 | |
f603bb94 NB |
5017 | if (pin_bytes) { |
5018 | struct btrfs_block_group *cache; | |
5019 | ||
5020 | cache = btrfs_lookup_block_group(fs_info, head->bytenr); | |
5021 | BUG_ON(!cache); | |
5022 | ||
5023 | spin_lock(&cache->space_info->lock); | |
5024 | spin_lock(&cache->lock); | |
5025 | cache->pinned += head->num_bytes; | |
5026 | btrfs_space_info_update_bytes_pinned(fs_info, | |
5027 | cache->space_info, head->num_bytes); | |
5028 | cache->reserved -= head->num_bytes; | |
5029 | cache->space_info->bytes_reserved -= head->num_bytes; | |
5030 | spin_unlock(&cache->lock); | |
5031 | spin_unlock(&cache->space_info->lock); | |
f603bb94 NB |
5032 | |
5033 | btrfs_put_block_group(cache); | |
5034 | ||
5035 | btrfs_error_unpin_extent_range(fs_info, head->bytenr, | |
5036 | head->bytenr + head->num_bytes - 1); | |
5037 | } | |
31890da0 | 5038 | btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); |
d278850e | 5039 | btrfs_put_delayed_ref_head(head); |
acce952b | 5040 | cond_resched(); |
5041 | spin_lock(&delayed_refs->lock); | |
5042 | } | |
81f7eb00 | 5043 | btrfs_qgroup_destroy_extent_records(trans); |
acce952b | 5044 | |
5045 | spin_unlock(&delayed_refs->lock); | |
5046 | ||
5047 | return ret; | |
5048 | } | |
5049 | ||
143bede5 | 5050 | static void btrfs_destroy_delalloc_inodes(struct btrfs_root *root) |
acce952b | 5051 | { |
5052 | struct btrfs_inode *btrfs_inode; | |
5053 | struct list_head splice; | |
5054 | ||
5055 | INIT_LIST_HEAD(&splice); | |
5056 | ||
eb73c1b7 MX |
5057 | spin_lock(&root->delalloc_lock); |
5058 | list_splice_init(&root->delalloc_inodes, &splice); | |
acce952b | 5059 | |
5060 | while (!list_empty(&splice)) { | |
fe816d0f | 5061 | struct inode *inode = NULL; |
eb73c1b7 MX |
5062 | btrfs_inode = list_first_entry(&splice, struct btrfs_inode, |
5063 | delalloc_inodes); | |
fe816d0f | 5064 | __btrfs_del_delalloc_inode(root, btrfs_inode); |
eb73c1b7 | 5065 | spin_unlock(&root->delalloc_lock); |
acce952b | 5066 | |
fe816d0f NB |
5067 | /* |
5068 | * Make sure we get a live inode and that it'll not disappear | |
5069 | * meanwhile. | |
5070 | */ | |
5071 | inode = igrab(&btrfs_inode->vfs_inode); | |
5072 | if (inode) { | |
5073 | invalidate_inode_pages2(inode->i_mapping); | |
5074 | iput(inode); | |
5075 | } | |
eb73c1b7 | 5076 | spin_lock(&root->delalloc_lock); |
acce952b | 5077 | } |
eb73c1b7 MX |
5078 | spin_unlock(&root->delalloc_lock); |
5079 | } | |
5080 | ||
5081 | static void btrfs_destroy_all_delalloc_inodes(struct btrfs_fs_info *fs_info) | |
5082 | { | |
5083 | struct btrfs_root *root; | |
5084 | struct list_head splice; | |
5085 | ||
5086 | INIT_LIST_HEAD(&splice); | |
5087 | ||
5088 | spin_lock(&fs_info->delalloc_root_lock); | |
5089 | list_splice_init(&fs_info->delalloc_roots, &splice); | |
5090 | while (!list_empty(&splice)) { | |
5091 | root = list_first_entry(&splice, struct btrfs_root, | |
5092 | delalloc_root); | |
00246528 | 5093 | root = btrfs_grab_root(root); |
eb73c1b7 MX |
5094 | BUG_ON(!root); |
5095 | spin_unlock(&fs_info->delalloc_root_lock); | |
5096 | ||
5097 | btrfs_destroy_delalloc_inodes(root); | |
00246528 | 5098 | btrfs_put_root(root); |
eb73c1b7 MX |
5099 | |
5100 | spin_lock(&fs_info->delalloc_root_lock); | |
5101 | } | |
5102 | spin_unlock(&fs_info->delalloc_root_lock); | |
acce952b | 5103 | } |
5104 | ||
2ff7e61e | 5105 | static int btrfs_destroy_marked_extents(struct btrfs_fs_info *fs_info, |
acce952b | 5106 | struct extent_io_tree *dirty_pages, |
5107 | int mark) | |
5108 | { | |
5109 | int ret; | |
acce952b | 5110 | struct extent_buffer *eb; |
5111 | u64 start = 0; | |
5112 | u64 end; | |
acce952b | 5113 | |
5114 | while (1) { | |
5115 | ret = find_first_extent_bit(dirty_pages, start, &start, &end, | |
e6138876 | 5116 | mark, NULL); |
acce952b | 5117 | if (ret) |
5118 | break; | |
5119 | ||
91166212 | 5120 | clear_extent_bits(dirty_pages, start, end, mark); |
acce952b | 5121 | while (start <= end) { |
0b246afa JM |
5122 | eb = find_extent_buffer(fs_info, start); |
5123 | start += fs_info->nodesize; | |
fd8b2b61 | 5124 | if (!eb) |
acce952b | 5125 | continue; |
fd8b2b61 | 5126 | wait_on_extent_buffer_writeback(eb); |
acce952b | 5127 | |
fd8b2b61 JB |
5128 | if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, |
5129 | &eb->bflags)) | |
5130 | clear_extent_buffer_dirty(eb); | |
5131 | free_extent_buffer_stale(eb); | |
acce952b | 5132 | } |
5133 | } | |
5134 | ||
5135 | return ret; | |
5136 | } | |
5137 | ||
2ff7e61e | 5138 | static int btrfs_destroy_pinned_extent(struct btrfs_fs_info *fs_info, |
fe119a6e | 5139 | struct extent_io_tree *unpin) |
acce952b | 5140 | { |
acce952b | 5141 | u64 start; |
5142 | u64 end; | |
5143 | int ret; | |
5144 | ||
acce952b | 5145 | while (1) { |
0e6ec385 FM |
5146 | struct extent_state *cached_state = NULL; |
5147 | ||
fcd5e742 LF |
5148 | /* |
5149 | * The btrfs_finish_extent_commit() may get the same range as | |
5150 | * ours between find_first_extent_bit and clear_extent_dirty. | |
5151 | * Hence, hold the unused_bg_unpin_mutex to avoid double unpin | |
5152 | * the same extent range. | |
5153 | */ | |
5154 | mutex_lock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5155 | ret = find_first_extent_bit(unpin, 0, &start, &end, |
0e6ec385 | 5156 | EXTENT_DIRTY, &cached_state); |
fcd5e742 LF |
5157 | if (ret) { |
5158 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); | |
acce952b | 5159 | break; |
fcd5e742 | 5160 | } |
acce952b | 5161 | |
0e6ec385 FM |
5162 | clear_extent_dirty(unpin, start, end, &cached_state); |
5163 | free_extent_state(cached_state); | |
2ff7e61e | 5164 | btrfs_error_unpin_extent_range(fs_info, start, end); |
fcd5e742 | 5165 | mutex_unlock(&fs_info->unused_bg_unpin_mutex); |
acce952b | 5166 | cond_resched(); |
5167 | } | |
5168 | ||
5169 | return 0; | |
5170 | } | |
5171 | ||
32da5386 | 5172 | static void btrfs_cleanup_bg_io(struct btrfs_block_group *cache) |
c79a1751 LB |
5173 | { |
5174 | struct inode *inode; | |
5175 | ||
5176 | inode = cache->io_ctl.inode; | |
5177 | if (inode) { | |
5178 | invalidate_inode_pages2(inode->i_mapping); | |
5179 | BTRFS_I(inode)->generation = 0; | |
5180 | cache->io_ctl.inode = NULL; | |
5181 | iput(inode); | |
5182 | } | |
bbc37d6e | 5183 | ASSERT(cache->io_ctl.pages == NULL); |
c79a1751 LB |
5184 | btrfs_put_block_group(cache); |
5185 | } | |
5186 | ||
5187 | void btrfs_cleanup_dirty_bgs(struct btrfs_transaction *cur_trans, | |
2ff7e61e | 5188 | struct btrfs_fs_info *fs_info) |
c79a1751 | 5189 | { |
32da5386 | 5190 | struct btrfs_block_group *cache; |
c79a1751 LB |
5191 | |
5192 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5193 | while (!list_empty(&cur_trans->dirty_bgs)) { | |
5194 | cache = list_first_entry(&cur_trans->dirty_bgs, | |
32da5386 | 5195 | struct btrfs_block_group, |
c79a1751 | 5196 | dirty_list); |
c79a1751 LB |
5197 | |
5198 | if (!list_empty(&cache->io_list)) { | |
5199 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5200 | list_del_init(&cache->io_list); | |
5201 | btrfs_cleanup_bg_io(cache); | |
5202 | spin_lock(&cur_trans->dirty_bgs_lock); | |
5203 | } | |
5204 | ||
5205 | list_del_init(&cache->dirty_list); | |
5206 | spin_lock(&cache->lock); | |
5207 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5208 | spin_unlock(&cache->lock); | |
5209 | ||
5210 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5211 | btrfs_put_block_group(cache); | |
ba2c4d4e | 5212 | btrfs_delayed_refs_rsv_release(fs_info, 1); |
c79a1751 LB |
5213 | spin_lock(&cur_trans->dirty_bgs_lock); |
5214 | } | |
5215 | spin_unlock(&cur_trans->dirty_bgs_lock); | |
5216 | ||
45ae2c18 NB |
5217 | /* |
5218 | * Refer to the definition of io_bgs member for details why it's safe | |
5219 | * to use it without any locking | |
5220 | */ | |
c79a1751 LB |
5221 | while (!list_empty(&cur_trans->io_bgs)) { |
5222 | cache = list_first_entry(&cur_trans->io_bgs, | |
32da5386 | 5223 | struct btrfs_block_group, |
c79a1751 | 5224 | io_list); |
c79a1751 LB |
5225 | |
5226 | list_del_init(&cache->io_list); | |
5227 | spin_lock(&cache->lock); | |
5228 | cache->disk_cache_state = BTRFS_DC_ERROR; | |
5229 | spin_unlock(&cache->lock); | |
5230 | btrfs_cleanup_bg_io(cache); | |
5231 | } | |
5232 | } | |
5233 | ||
49b25e05 | 5234 | void btrfs_cleanup_one_transaction(struct btrfs_transaction *cur_trans, |
2ff7e61e | 5235 | struct btrfs_fs_info *fs_info) |
49b25e05 | 5236 | { |
bbbf7243 NB |
5237 | struct btrfs_device *dev, *tmp; |
5238 | ||
2ff7e61e | 5239 | btrfs_cleanup_dirty_bgs(cur_trans, fs_info); |
c79a1751 LB |
5240 | ASSERT(list_empty(&cur_trans->dirty_bgs)); |
5241 | ASSERT(list_empty(&cur_trans->io_bgs)); | |
5242 | ||
bbbf7243 NB |
5243 | list_for_each_entry_safe(dev, tmp, &cur_trans->dev_update_list, |
5244 | post_commit_list) { | |
5245 | list_del_init(&dev->post_commit_list); | |
5246 | } | |
5247 | ||
2ff7e61e | 5248 | btrfs_destroy_delayed_refs(cur_trans, fs_info); |
49b25e05 | 5249 | |
4a9d8bde | 5250 | cur_trans->state = TRANS_STATE_COMMIT_START; |
0b246afa | 5251 | wake_up(&fs_info->transaction_blocked_wait); |
49b25e05 | 5252 | |
4a9d8bde | 5253 | cur_trans->state = TRANS_STATE_UNBLOCKED; |
0b246afa | 5254 | wake_up(&fs_info->transaction_wait); |
49b25e05 | 5255 | |
ccdf9b30 | 5256 | btrfs_destroy_delayed_inodes(fs_info); |
49b25e05 | 5257 | |
2ff7e61e | 5258 | btrfs_destroy_marked_extents(fs_info, &cur_trans->dirty_pages, |
49b25e05 | 5259 | EXTENT_DIRTY); |
fe119a6e | 5260 | btrfs_destroy_pinned_extent(fs_info, &cur_trans->pinned_extents); |
49b25e05 | 5261 | |
d3575156 NA |
5262 | btrfs_free_redirty_list(cur_trans); |
5263 | ||
4a9d8bde MX |
5264 | cur_trans->state =TRANS_STATE_COMPLETED; |
5265 | wake_up(&cur_trans->commit_wait); | |
49b25e05 JM |
5266 | } |
5267 | ||
2ff7e61e | 5268 | static int btrfs_cleanup_transaction(struct btrfs_fs_info *fs_info) |
acce952b | 5269 | { |
5270 | struct btrfs_transaction *t; | |
acce952b | 5271 | |
0b246afa | 5272 | mutex_lock(&fs_info->transaction_kthread_mutex); |
acce952b | 5273 | |
0b246afa JM |
5274 | spin_lock(&fs_info->trans_lock); |
5275 | while (!list_empty(&fs_info->trans_list)) { | |
5276 | t = list_first_entry(&fs_info->trans_list, | |
724e2315 JB |
5277 | struct btrfs_transaction, list); |
5278 | if (t->state >= TRANS_STATE_COMMIT_START) { | |
9b64f57d | 5279 | refcount_inc(&t->use_count); |
0b246afa | 5280 | spin_unlock(&fs_info->trans_lock); |
2ff7e61e | 5281 | btrfs_wait_for_commit(fs_info, t->transid); |
724e2315 | 5282 | btrfs_put_transaction(t); |
0b246afa | 5283 | spin_lock(&fs_info->trans_lock); |
724e2315 JB |
5284 | continue; |
5285 | } | |
0b246afa | 5286 | if (t == fs_info->running_transaction) { |
724e2315 | 5287 | t->state = TRANS_STATE_COMMIT_DOING; |
0b246afa | 5288 | spin_unlock(&fs_info->trans_lock); |
724e2315 JB |
5289 | /* |
5290 | * We wait for 0 num_writers since we don't hold a trans | |
5291 | * handle open currently for this transaction. | |
5292 | */ | |
5293 | wait_event(t->writer_wait, | |
5294 | atomic_read(&t->num_writers) == 0); | |
5295 | } else { | |
0b246afa | 5296 | spin_unlock(&fs_info->trans_lock); |
724e2315 | 5297 | } |
2ff7e61e | 5298 | btrfs_cleanup_one_transaction(t, fs_info); |
4a9d8bde | 5299 | |
0b246afa JM |
5300 | spin_lock(&fs_info->trans_lock); |
5301 | if (t == fs_info->running_transaction) | |
5302 | fs_info->running_transaction = NULL; | |
acce952b | 5303 | list_del_init(&t->list); |
0b246afa | 5304 | spin_unlock(&fs_info->trans_lock); |
acce952b | 5305 | |
724e2315 | 5306 | btrfs_put_transaction(t); |
2e4e97ab | 5307 | trace_btrfs_transaction_commit(fs_info); |
0b246afa | 5308 | spin_lock(&fs_info->trans_lock); |
724e2315 | 5309 | } |
0b246afa JM |
5310 | spin_unlock(&fs_info->trans_lock); |
5311 | btrfs_destroy_all_ordered_extents(fs_info); | |
ccdf9b30 JM |
5312 | btrfs_destroy_delayed_inodes(fs_info); |
5313 | btrfs_assert_delayed_root_empty(fs_info); | |
0b246afa | 5314 | btrfs_destroy_all_delalloc_inodes(fs_info); |
ef67963d | 5315 | btrfs_drop_all_logs(fs_info); |
0b246afa | 5316 | mutex_unlock(&fs_info->transaction_kthread_mutex); |
acce952b | 5317 | |
5318 | return 0; | |
5319 | } | |
ec7d6dfd | 5320 | |
453e4873 | 5321 | int btrfs_init_root_free_objectid(struct btrfs_root *root) |
ec7d6dfd NB |
5322 | { |
5323 | struct btrfs_path *path; | |
5324 | int ret; | |
5325 | struct extent_buffer *l; | |
5326 | struct btrfs_key search_key; | |
5327 | struct btrfs_key found_key; | |
5328 | int slot; | |
5329 | ||
5330 | path = btrfs_alloc_path(); | |
5331 | if (!path) | |
5332 | return -ENOMEM; | |
5333 | ||
5334 | search_key.objectid = BTRFS_LAST_FREE_OBJECTID; | |
5335 | search_key.type = -1; | |
5336 | search_key.offset = (u64)-1; | |
5337 | ret = btrfs_search_slot(NULL, root, &search_key, path, 0, 0); | |
5338 | if (ret < 0) | |
5339 | goto error; | |
5340 | BUG_ON(ret == 0); /* Corruption */ | |
5341 | if (path->slots[0] > 0) { | |
5342 | slot = path->slots[0] - 1; | |
5343 | l = path->nodes[0]; | |
5344 | btrfs_item_key_to_cpu(l, &found_key, slot); | |
23125104 NB |
5345 | root->free_objectid = max_t(u64, found_key.objectid + 1, |
5346 | BTRFS_FIRST_FREE_OBJECTID); | |
ec7d6dfd | 5347 | } else { |
23125104 | 5348 | root->free_objectid = BTRFS_FIRST_FREE_OBJECTID; |
ec7d6dfd NB |
5349 | } |
5350 | ret = 0; | |
5351 | error: | |
5352 | btrfs_free_path(path); | |
5353 | return ret; | |
5354 | } | |
5355 | ||
543068a2 | 5356 | int btrfs_get_free_objectid(struct btrfs_root *root, u64 *objectid) |
ec7d6dfd NB |
5357 | { |
5358 | int ret; | |
5359 | mutex_lock(&root->objectid_mutex); | |
5360 | ||
6b8fad57 | 5361 | if (unlikely(root->free_objectid >= BTRFS_LAST_FREE_OBJECTID)) { |
ec7d6dfd NB |
5362 | btrfs_warn(root->fs_info, |
5363 | "the objectid of root %llu reaches its highest value", | |
5364 | root->root_key.objectid); | |
5365 | ret = -ENOSPC; | |
5366 | goto out; | |
5367 | } | |
5368 | ||
23125104 | 5369 | *objectid = root->free_objectid++; |
ec7d6dfd NB |
5370 | ret = 0; |
5371 | out: | |
5372 | mutex_unlock(&root->objectid_mutex); | |
5373 | return ret; | |
5374 | } |