]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
19 | #include <linux/blkdev.h> | |
20 | #include <linux/module.h> | |
21 | #include <linux/buffer_head.h> | |
22 | #include <linux/fs.h> | |
23 | #include <linux/pagemap.h> | |
24 | #include <linux/highmem.h> | |
25 | #include <linux/time.h> | |
26 | #include <linux/init.h> | |
27 | #include <linux/seq_file.h> | |
28 | #include <linux/string.h> | |
29 | #include <linux/backing-dev.h> | |
30 | #include <linux/mount.h> | |
31 | #include <linux/mpage.h> | |
32 | #include <linux/swap.h> | |
33 | #include <linux/writeback.h> | |
34 | #include <linux/statfs.h> | |
35 | #include <linux/compat.h> | |
36 | #include <linux/parser.h> | |
37 | #include <linux/ctype.h> | |
38 | #include <linux/namei.h> | |
39 | #include <linux/miscdevice.h> | |
40 | #include <linux/magic.h> | |
41 | #include <linux/slab.h> | |
42 | #include <linux/cleancache.h> | |
43 | #include <linux/ratelimit.h> | |
44 | #include <linux/btrfs.h> | |
45 | #include "compat.h" | |
46 | #include "delayed-inode.h" | |
47 | #include "ctree.h" | |
48 | #include "disk-io.h" | |
49 | #include "transaction.h" | |
50 | #include "btrfs_inode.h" | |
51 | #include "print-tree.h" | |
52 | #include "xattr.h" | |
53 | #include "volumes.h" | |
54 | #include "export.h" | |
55 | #include "compression.h" | |
56 | #include "rcu-string.h" | |
57 | #include "dev-replace.h" | |
58 | #include "free-space-cache.h" | |
59 | #include "tests/btrfs-tests.h" | |
60 | ||
61 | #define CREATE_TRACE_POINTS | |
62 | #include <trace/events/btrfs.h> | |
63 | ||
64 | static const struct super_operations btrfs_super_ops; | |
65 | static struct file_system_type btrfs_fs_type; | |
66 | ||
67 | static const char *btrfs_decode_error(int errno) | |
68 | { | |
69 | char *errstr = "unknown"; | |
70 | ||
71 | switch (errno) { | |
72 | case -EIO: | |
73 | errstr = "IO failure"; | |
74 | break; | |
75 | case -ENOMEM: | |
76 | errstr = "Out of memory"; | |
77 | break; | |
78 | case -EROFS: | |
79 | errstr = "Readonly filesystem"; | |
80 | break; | |
81 | case -EEXIST: | |
82 | errstr = "Object already exists"; | |
83 | break; | |
84 | case -ENOSPC: | |
85 | errstr = "No space left"; | |
86 | break; | |
87 | case -ENOENT: | |
88 | errstr = "No such entry"; | |
89 | break; | |
90 | } | |
91 | ||
92 | return errstr; | |
93 | } | |
94 | ||
95 | static void save_error_info(struct btrfs_fs_info *fs_info) | |
96 | { | |
97 | /* | |
98 | * today we only save the error info into ram. Long term we'll | |
99 | * also send it down to the disk | |
100 | */ | |
101 | set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); | |
102 | } | |
103 | ||
104 | /* btrfs handle error by forcing the filesystem readonly */ | |
105 | static void btrfs_handle_error(struct btrfs_fs_info *fs_info) | |
106 | { | |
107 | struct super_block *sb = fs_info->sb; | |
108 | ||
109 | if (sb->s_flags & MS_RDONLY) | |
110 | return; | |
111 | ||
112 | if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { | |
113 | sb->s_flags |= MS_RDONLY; | |
114 | btrfs_info(fs_info, "forced readonly"); | |
115 | /* | |
116 | * Note that a running device replace operation is not | |
117 | * canceled here although there is no way to update | |
118 | * the progress. It would add the risk of a deadlock, | |
119 | * therefore the canceling is ommited. The only penalty | |
120 | * is that some I/O remains active until the procedure | |
121 | * completes. The next time when the filesystem is | |
122 | * mounted writeable again, the device replace | |
123 | * operation continues. | |
124 | */ | |
125 | } | |
126 | } | |
127 | ||
128 | #ifdef CONFIG_PRINTK | |
129 | /* | |
130 | * __btrfs_std_error decodes expected errors from the caller and | |
131 | * invokes the approciate error response. | |
132 | */ | |
133 | void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function, | |
134 | unsigned int line, int errno, const char *fmt, ...) | |
135 | { | |
136 | struct super_block *sb = fs_info->sb; | |
137 | const char *errstr; | |
138 | ||
139 | /* | |
140 | * Special case: if the error is EROFS, and we're already | |
141 | * under MS_RDONLY, then it is safe here. | |
142 | */ | |
143 | if (errno == -EROFS && (sb->s_flags & MS_RDONLY)) | |
144 | return; | |
145 | ||
146 | errstr = btrfs_decode_error(errno); | |
147 | if (fmt) { | |
148 | struct va_format vaf; | |
149 | va_list args; | |
150 | ||
151 | va_start(args, fmt); | |
152 | vaf.fmt = fmt; | |
153 | vaf.va = &args; | |
154 | ||
155 | printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: errno=%d %s (%pV)\n", | |
156 | sb->s_id, function, line, errno, errstr, &vaf); | |
157 | va_end(args); | |
158 | } else { | |
159 | printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: errno=%d %s\n", | |
160 | sb->s_id, function, line, errno, errstr); | |
161 | } | |
162 | ||
163 | /* Don't go through full error handling during mount */ | |
164 | save_error_info(fs_info); | |
165 | if (sb->s_flags & MS_BORN) | |
166 | btrfs_handle_error(fs_info); | |
167 | } | |
168 | ||
169 | static const char * const logtypes[] = { | |
170 | "emergency", | |
171 | "alert", | |
172 | "critical", | |
173 | "error", | |
174 | "warning", | |
175 | "notice", | |
176 | "info", | |
177 | "debug", | |
178 | }; | |
179 | ||
180 | void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...) | |
181 | { | |
182 | struct super_block *sb = fs_info->sb; | |
183 | char lvl[4]; | |
184 | struct va_format vaf; | |
185 | va_list args; | |
186 | const char *type = logtypes[4]; | |
187 | int kern_level; | |
188 | ||
189 | va_start(args, fmt); | |
190 | ||
191 | kern_level = printk_get_level(fmt); | |
192 | if (kern_level) { | |
193 | size_t size = printk_skip_level(fmt) - fmt; | |
194 | memcpy(lvl, fmt, size); | |
195 | lvl[size] = '\0'; | |
196 | fmt += size; | |
197 | type = logtypes[kern_level - '0']; | |
198 | } else | |
199 | *lvl = '\0'; | |
200 | ||
201 | vaf.fmt = fmt; | |
202 | vaf.va = &args; | |
203 | ||
204 | printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf); | |
205 | ||
206 | va_end(args); | |
207 | } | |
208 | ||
209 | #else | |
210 | ||
211 | void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function, | |
212 | unsigned int line, int errno, const char *fmt, ...) | |
213 | { | |
214 | struct super_block *sb = fs_info->sb; | |
215 | ||
216 | /* | |
217 | * Special case: if the error is EROFS, and we're already | |
218 | * under MS_RDONLY, then it is safe here. | |
219 | */ | |
220 | if (errno == -EROFS && (sb->s_flags & MS_RDONLY)) | |
221 | return; | |
222 | ||
223 | /* Don't go through full error handling during mount */ | |
224 | if (sb->s_flags & MS_BORN) { | |
225 | save_error_info(fs_info); | |
226 | btrfs_handle_error(fs_info); | |
227 | } | |
228 | } | |
229 | #endif | |
230 | ||
231 | /* | |
232 | * We only mark the transaction aborted and then set the file system read-only. | |
233 | * This will prevent new transactions from starting or trying to join this | |
234 | * one. | |
235 | * | |
236 | * This means that error recovery at the call site is limited to freeing | |
237 | * any local memory allocations and passing the error code up without | |
238 | * further cleanup. The transaction should complete as it normally would | |
239 | * in the call path but will return -EIO. | |
240 | * | |
241 | * We'll complete the cleanup in btrfs_end_transaction and | |
242 | * btrfs_commit_transaction. | |
243 | */ | |
244 | void __btrfs_abort_transaction(struct btrfs_trans_handle *trans, | |
245 | struct btrfs_root *root, const char *function, | |
246 | unsigned int line, int errno) | |
247 | { | |
248 | /* | |
249 | * Report first abort since mount | |
250 | */ | |
251 | if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED, | |
252 | &root->fs_info->fs_state)) { | |
253 | WARN(1, KERN_DEBUG "btrfs: Transaction aborted (error %d)\n", | |
254 | errno); | |
255 | } | |
256 | trans->aborted = errno; | |
257 | /* Nothing used. The other threads that have joined this | |
258 | * transaction may be able to continue. */ | |
259 | if (!trans->blocks_used) { | |
260 | const char *errstr; | |
261 | ||
262 | errstr = btrfs_decode_error(errno); | |
263 | btrfs_warn(root->fs_info, | |
264 | "%s:%d: Aborting unused transaction(%s).", | |
265 | function, line, errstr); | |
266 | return; | |
267 | } | |
268 | ACCESS_ONCE(trans->transaction->aborted) = errno; | |
269 | /* Wake up anybody who may be waiting on this transaction */ | |
270 | wake_up(&root->fs_info->transaction_wait); | |
271 | wake_up(&root->fs_info->transaction_blocked_wait); | |
272 | __btrfs_std_error(root->fs_info, function, line, errno, NULL); | |
273 | } | |
274 | /* | |
275 | * __btrfs_panic decodes unexpected, fatal errors from the caller, | |
276 | * issues an alert, and either panics or BUGs, depending on mount options. | |
277 | */ | |
278 | void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function, | |
279 | unsigned int line, int errno, const char *fmt, ...) | |
280 | { | |
281 | char *s_id = "<unknown>"; | |
282 | const char *errstr; | |
283 | struct va_format vaf = { .fmt = fmt }; | |
284 | va_list args; | |
285 | ||
286 | if (fs_info) | |
287 | s_id = fs_info->sb->s_id; | |
288 | ||
289 | va_start(args, fmt); | |
290 | vaf.va = &args; | |
291 | ||
292 | errstr = btrfs_decode_error(errno); | |
293 | if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR)) | |
294 | panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n", | |
295 | s_id, function, line, &vaf, errno, errstr); | |
296 | ||
297 | printk(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n", | |
298 | s_id, function, line, &vaf, errno, errstr); | |
299 | va_end(args); | |
300 | /* Caller calls BUG() */ | |
301 | } | |
302 | ||
303 | static void btrfs_put_super(struct super_block *sb) | |
304 | { | |
305 | (void)close_ctree(btrfs_sb(sb)->tree_root); | |
306 | /* FIXME: need to fix VFS to return error? */ | |
307 | /* AV: return it _where_? ->put_super() can be triggered by any number | |
308 | * of async events, up to and including delivery of SIGKILL to the | |
309 | * last process that kept it busy. Or segfault in the aforementioned | |
310 | * process... Whom would you report that to? | |
311 | */ | |
312 | } | |
313 | ||
314 | enum { | |
315 | Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum, | |
316 | Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd, | |
317 | Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress, | |
318 | Opt_compress_type, Opt_compress_force, Opt_compress_force_type, | |
319 | Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard, | |
320 | Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed, | |
321 | Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache, | |
322 | Opt_no_space_cache, Opt_recovery, Opt_skip_balance, | |
323 | Opt_check_integrity, Opt_check_integrity_including_extent_data, | |
324 | Opt_check_integrity_print_mask, Opt_fatal_errors, | |
325 | Opt_commit_interval, | |
326 | Opt_err, | |
327 | }; | |
328 | ||
329 | static match_table_t tokens = { | |
330 | {Opt_degraded, "degraded"}, | |
331 | {Opt_subvol, "subvol=%s"}, | |
332 | {Opt_subvolid, "subvolid=%s"}, | |
333 | {Opt_device, "device=%s"}, | |
334 | {Opt_nodatasum, "nodatasum"}, | |
335 | {Opt_nodatacow, "nodatacow"}, | |
336 | {Opt_nobarrier, "nobarrier"}, | |
337 | {Opt_max_inline, "max_inline=%s"}, | |
338 | {Opt_alloc_start, "alloc_start=%s"}, | |
339 | {Opt_thread_pool, "thread_pool=%d"}, | |
340 | {Opt_compress, "compress"}, | |
341 | {Opt_compress_type, "compress=%s"}, | |
342 | {Opt_compress_force, "compress-force"}, | |
343 | {Opt_compress_force_type, "compress-force=%s"}, | |
344 | {Opt_ssd, "ssd"}, | |
345 | {Opt_ssd_spread, "ssd_spread"}, | |
346 | {Opt_nossd, "nossd"}, | |
347 | {Opt_noacl, "noacl"}, | |
348 | {Opt_notreelog, "notreelog"}, | |
349 | {Opt_flushoncommit, "flushoncommit"}, | |
350 | {Opt_ratio, "metadata_ratio=%d"}, | |
351 | {Opt_discard, "discard"}, | |
352 | {Opt_space_cache, "space_cache"}, | |
353 | {Opt_clear_cache, "clear_cache"}, | |
354 | {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"}, | |
355 | {Opt_enospc_debug, "enospc_debug"}, | |
356 | {Opt_subvolrootid, "subvolrootid=%d"}, | |
357 | {Opt_defrag, "autodefrag"}, | |
358 | {Opt_inode_cache, "inode_cache"}, | |
359 | {Opt_no_space_cache, "nospace_cache"}, | |
360 | {Opt_recovery, "recovery"}, | |
361 | {Opt_skip_balance, "skip_balance"}, | |
362 | {Opt_check_integrity, "check_int"}, | |
363 | {Opt_check_integrity_including_extent_data, "check_int_data"}, | |
364 | {Opt_check_integrity_print_mask, "check_int_print_mask=%d"}, | |
365 | {Opt_fatal_errors, "fatal_errors=%s"}, | |
366 | {Opt_commit_interval, "commit=%d"}, | |
367 | {Opt_err, NULL}, | |
368 | }; | |
369 | ||
370 | /* | |
371 | * Regular mount options parser. Everything that is needed only when | |
372 | * reading in a new superblock is parsed here. | |
373 | * XXX JDM: This needs to be cleaned up for remount. | |
374 | */ | |
375 | int btrfs_parse_options(struct btrfs_root *root, char *options) | |
376 | { | |
377 | struct btrfs_fs_info *info = root->fs_info; | |
378 | substring_t args[MAX_OPT_ARGS]; | |
379 | char *p, *num, *orig = NULL; | |
380 | u64 cache_gen; | |
381 | int intarg; | |
382 | int ret = 0; | |
383 | char *compress_type; | |
384 | bool compress_force = false; | |
385 | ||
386 | cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy); | |
387 | if (cache_gen) | |
388 | btrfs_set_opt(info->mount_opt, SPACE_CACHE); | |
389 | ||
390 | if (!options) | |
391 | goto out; | |
392 | ||
393 | /* | |
394 | * strsep changes the string, duplicate it because parse_options | |
395 | * gets called twice | |
396 | */ | |
397 | options = kstrdup(options, GFP_NOFS); | |
398 | if (!options) | |
399 | return -ENOMEM; | |
400 | ||
401 | orig = options; | |
402 | ||
403 | while ((p = strsep(&options, ",")) != NULL) { | |
404 | int token; | |
405 | if (!*p) | |
406 | continue; | |
407 | ||
408 | token = match_token(p, tokens, args); | |
409 | switch (token) { | |
410 | case Opt_degraded: | |
411 | printk(KERN_INFO "btrfs: allowing degraded mounts\n"); | |
412 | btrfs_set_opt(info->mount_opt, DEGRADED); | |
413 | break; | |
414 | case Opt_subvol: | |
415 | case Opt_subvolid: | |
416 | case Opt_subvolrootid: | |
417 | case Opt_device: | |
418 | /* | |
419 | * These are parsed by btrfs_parse_early_options | |
420 | * and can be happily ignored here. | |
421 | */ | |
422 | break; | |
423 | case Opt_nodatasum: | |
424 | printk(KERN_INFO "btrfs: setting nodatasum\n"); | |
425 | btrfs_set_opt(info->mount_opt, NODATASUM); | |
426 | break; | |
427 | case Opt_nodatacow: | |
428 | if (!btrfs_test_opt(root, COMPRESS) || | |
429 | !btrfs_test_opt(root, FORCE_COMPRESS)) { | |
430 | printk(KERN_INFO "btrfs: setting nodatacow, compression disabled\n"); | |
431 | } else { | |
432 | printk(KERN_INFO "btrfs: setting nodatacow\n"); | |
433 | } | |
434 | info->compress_type = BTRFS_COMPRESS_NONE; | |
435 | btrfs_clear_opt(info->mount_opt, COMPRESS); | |
436 | btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); | |
437 | btrfs_set_opt(info->mount_opt, NODATACOW); | |
438 | btrfs_set_opt(info->mount_opt, NODATASUM); | |
439 | break; | |
440 | case Opt_compress_force: | |
441 | case Opt_compress_force_type: | |
442 | compress_force = true; | |
443 | /* Fallthrough */ | |
444 | case Opt_compress: | |
445 | case Opt_compress_type: | |
446 | if (token == Opt_compress || | |
447 | token == Opt_compress_force || | |
448 | strcmp(args[0].from, "zlib") == 0) { | |
449 | compress_type = "zlib"; | |
450 | info->compress_type = BTRFS_COMPRESS_ZLIB; | |
451 | btrfs_set_opt(info->mount_opt, COMPRESS); | |
452 | btrfs_clear_opt(info->mount_opt, NODATACOW); | |
453 | btrfs_clear_opt(info->mount_opt, NODATASUM); | |
454 | } else if (strcmp(args[0].from, "lzo") == 0) { | |
455 | compress_type = "lzo"; | |
456 | info->compress_type = BTRFS_COMPRESS_LZO; | |
457 | btrfs_set_opt(info->mount_opt, COMPRESS); | |
458 | btrfs_clear_opt(info->mount_opt, NODATACOW); | |
459 | btrfs_clear_opt(info->mount_opt, NODATASUM); | |
460 | btrfs_set_fs_incompat(info, COMPRESS_LZO); | |
461 | } else if (strncmp(args[0].from, "no", 2) == 0) { | |
462 | compress_type = "no"; | |
463 | info->compress_type = BTRFS_COMPRESS_NONE; | |
464 | btrfs_clear_opt(info->mount_opt, COMPRESS); | |
465 | btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); | |
466 | compress_force = false; | |
467 | } else { | |
468 | ret = -EINVAL; | |
469 | goto out; | |
470 | } | |
471 | ||
472 | if (compress_force) { | |
473 | btrfs_set_opt(info->mount_opt, FORCE_COMPRESS); | |
474 | pr_info("btrfs: force %s compression\n", | |
475 | compress_type); | |
476 | } else | |
477 | pr_info("btrfs: use %s compression\n", | |
478 | compress_type); | |
479 | break; | |
480 | case Opt_ssd: | |
481 | printk(KERN_INFO "btrfs: use ssd allocation scheme\n"); | |
482 | btrfs_set_opt(info->mount_opt, SSD); | |
483 | break; | |
484 | case Opt_ssd_spread: | |
485 | printk(KERN_INFO "btrfs: use spread ssd " | |
486 | "allocation scheme\n"); | |
487 | btrfs_set_opt(info->mount_opt, SSD); | |
488 | btrfs_set_opt(info->mount_opt, SSD_SPREAD); | |
489 | break; | |
490 | case Opt_nossd: | |
491 | printk(KERN_INFO "btrfs: not using ssd allocation " | |
492 | "scheme\n"); | |
493 | btrfs_set_opt(info->mount_opt, NOSSD); | |
494 | btrfs_clear_opt(info->mount_opt, SSD); | |
495 | btrfs_clear_opt(info->mount_opt, SSD_SPREAD); | |
496 | break; | |
497 | case Opt_nobarrier: | |
498 | printk(KERN_INFO "btrfs: turning off barriers\n"); | |
499 | btrfs_set_opt(info->mount_opt, NOBARRIER); | |
500 | break; | |
501 | case Opt_thread_pool: | |
502 | ret = match_int(&args[0], &intarg); | |
503 | if (ret) { | |
504 | goto out; | |
505 | } else if (intarg > 0) { | |
506 | info->thread_pool_size = intarg; | |
507 | } else { | |
508 | ret = -EINVAL; | |
509 | goto out; | |
510 | } | |
511 | break; | |
512 | case Opt_max_inline: | |
513 | num = match_strdup(&args[0]); | |
514 | if (num) { | |
515 | info->max_inline = memparse(num, NULL); | |
516 | kfree(num); | |
517 | ||
518 | if (info->max_inline) { | |
519 | info->max_inline = max_t(u64, | |
520 | info->max_inline, | |
521 | root->sectorsize); | |
522 | } | |
523 | printk(KERN_INFO "btrfs: max_inline at %llu\n", | |
524 | (unsigned long long)info->max_inline); | |
525 | } else { | |
526 | ret = -ENOMEM; | |
527 | goto out; | |
528 | } | |
529 | break; | |
530 | case Opt_alloc_start: | |
531 | num = match_strdup(&args[0]); | |
532 | if (num) { | |
533 | mutex_lock(&info->chunk_mutex); | |
534 | info->alloc_start = memparse(num, NULL); | |
535 | mutex_unlock(&info->chunk_mutex); | |
536 | kfree(num); | |
537 | printk(KERN_INFO | |
538 | "btrfs: allocations start at %llu\n", | |
539 | (unsigned long long)info->alloc_start); | |
540 | } else { | |
541 | ret = -ENOMEM; | |
542 | goto out; | |
543 | } | |
544 | break; | |
545 | case Opt_noacl: | |
546 | root->fs_info->sb->s_flags &= ~MS_POSIXACL; | |
547 | break; | |
548 | case Opt_notreelog: | |
549 | printk(KERN_INFO "btrfs: disabling tree log\n"); | |
550 | btrfs_set_opt(info->mount_opt, NOTREELOG); | |
551 | break; | |
552 | case Opt_flushoncommit: | |
553 | printk(KERN_INFO "btrfs: turning on flush-on-commit\n"); | |
554 | btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT); | |
555 | break; | |
556 | case Opt_ratio: | |
557 | ret = match_int(&args[0], &intarg); | |
558 | if (ret) { | |
559 | goto out; | |
560 | } else if (intarg >= 0) { | |
561 | info->metadata_ratio = intarg; | |
562 | printk(KERN_INFO "btrfs: metadata ratio %d\n", | |
563 | info->metadata_ratio); | |
564 | } else { | |
565 | ret = -EINVAL; | |
566 | goto out; | |
567 | } | |
568 | break; | |
569 | case Opt_discard: | |
570 | btrfs_set_opt(info->mount_opt, DISCARD); | |
571 | break; | |
572 | case Opt_space_cache: | |
573 | btrfs_set_opt(info->mount_opt, SPACE_CACHE); | |
574 | break; | |
575 | case Opt_no_space_cache: | |
576 | printk(KERN_INFO "btrfs: disabling disk space caching\n"); | |
577 | btrfs_clear_opt(info->mount_opt, SPACE_CACHE); | |
578 | break; | |
579 | case Opt_inode_cache: | |
580 | printk(KERN_INFO "btrfs: enabling inode map caching\n"); | |
581 | btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE); | |
582 | break; | |
583 | case Opt_clear_cache: | |
584 | printk(KERN_INFO "btrfs: force clearing of disk cache\n"); | |
585 | btrfs_set_opt(info->mount_opt, CLEAR_CACHE); | |
586 | break; | |
587 | case Opt_user_subvol_rm_allowed: | |
588 | btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED); | |
589 | break; | |
590 | case Opt_enospc_debug: | |
591 | btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG); | |
592 | break; | |
593 | case Opt_defrag: | |
594 | printk(KERN_INFO "btrfs: enabling auto defrag\n"); | |
595 | btrfs_set_opt(info->mount_opt, AUTO_DEFRAG); | |
596 | break; | |
597 | case Opt_recovery: | |
598 | printk(KERN_INFO "btrfs: enabling auto recovery\n"); | |
599 | btrfs_set_opt(info->mount_opt, RECOVERY); | |
600 | break; | |
601 | case Opt_skip_balance: | |
602 | btrfs_set_opt(info->mount_opt, SKIP_BALANCE); | |
603 | break; | |
604 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY | |
605 | case Opt_check_integrity_including_extent_data: | |
606 | printk(KERN_INFO "btrfs: enabling check integrity" | |
607 | " including extent data\n"); | |
608 | btrfs_set_opt(info->mount_opt, | |
609 | CHECK_INTEGRITY_INCLUDING_EXTENT_DATA); | |
610 | btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY); | |
611 | break; | |
612 | case Opt_check_integrity: | |
613 | printk(KERN_INFO "btrfs: enabling check integrity\n"); | |
614 | btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY); | |
615 | break; | |
616 | case Opt_check_integrity_print_mask: | |
617 | ret = match_int(&args[0], &intarg); | |
618 | if (ret) { | |
619 | goto out; | |
620 | } else if (intarg >= 0) { | |
621 | info->check_integrity_print_mask = intarg; | |
622 | printk(KERN_INFO "btrfs:" | |
623 | " check_integrity_print_mask 0x%x\n", | |
624 | info->check_integrity_print_mask); | |
625 | } else { | |
626 | ret = -EINVAL; | |
627 | goto out; | |
628 | } | |
629 | break; | |
630 | #else | |
631 | case Opt_check_integrity_including_extent_data: | |
632 | case Opt_check_integrity: | |
633 | case Opt_check_integrity_print_mask: | |
634 | printk(KERN_ERR "btrfs: support for check_integrity*" | |
635 | " not compiled in!\n"); | |
636 | ret = -EINVAL; | |
637 | goto out; | |
638 | #endif | |
639 | case Opt_fatal_errors: | |
640 | if (strcmp(args[0].from, "panic") == 0) | |
641 | btrfs_set_opt(info->mount_opt, | |
642 | PANIC_ON_FATAL_ERROR); | |
643 | else if (strcmp(args[0].from, "bug") == 0) | |
644 | btrfs_clear_opt(info->mount_opt, | |
645 | PANIC_ON_FATAL_ERROR); | |
646 | else { | |
647 | ret = -EINVAL; | |
648 | goto out; | |
649 | } | |
650 | break; | |
651 | case Opt_commit_interval: | |
652 | intarg = 0; | |
653 | ret = match_int(&args[0], &intarg); | |
654 | if (ret < 0) { | |
655 | printk(KERN_ERR | |
656 | "btrfs: invalid commit interval\n"); | |
657 | ret = -EINVAL; | |
658 | goto out; | |
659 | } | |
660 | if (intarg > 0) { | |
661 | if (intarg > 300) { | |
662 | printk(KERN_WARNING | |
663 | "btrfs: excessive commit interval %d\n", | |
664 | intarg); | |
665 | } | |
666 | info->commit_interval = intarg; | |
667 | } else { | |
668 | printk(KERN_INFO | |
669 | "btrfs: using default commit interval %ds\n", | |
670 | BTRFS_DEFAULT_COMMIT_INTERVAL); | |
671 | info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; | |
672 | } | |
673 | break; | |
674 | case Opt_err: | |
675 | printk(KERN_INFO "btrfs: unrecognized mount option " | |
676 | "'%s'\n", p); | |
677 | ret = -EINVAL; | |
678 | goto out; | |
679 | default: | |
680 | break; | |
681 | } | |
682 | } | |
683 | out: | |
684 | if (!ret && btrfs_test_opt(root, SPACE_CACHE)) | |
685 | printk(KERN_INFO "btrfs: disk space caching is enabled\n"); | |
686 | kfree(orig); | |
687 | return ret; | |
688 | } | |
689 | ||
690 | /* | |
691 | * Parse mount options that are required early in the mount process. | |
692 | * | |
693 | * All other options will be parsed on much later in the mount process and | |
694 | * only when we need to allocate a new super block. | |
695 | */ | |
696 | static int btrfs_parse_early_options(const char *options, fmode_t flags, | |
697 | void *holder, char **subvol_name, u64 *subvol_objectid, | |
698 | struct btrfs_fs_devices **fs_devices) | |
699 | { | |
700 | substring_t args[MAX_OPT_ARGS]; | |
701 | char *device_name, *opts, *orig, *p; | |
702 | char *num = NULL; | |
703 | int error = 0; | |
704 | ||
705 | if (!options) | |
706 | return 0; | |
707 | ||
708 | /* | |
709 | * strsep changes the string, duplicate it because parse_options | |
710 | * gets called twice | |
711 | */ | |
712 | opts = kstrdup(options, GFP_KERNEL); | |
713 | if (!opts) | |
714 | return -ENOMEM; | |
715 | orig = opts; | |
716 | ||
717 | while ((p = strsep(&opts, ",")) != NULL) { | |
718 | int token; | |
719 | if (!*p) | |
720 | continue; | |
721 | ||
722 | token = match_token(p, tokens, args); | |
723 | switch (token) { | |
724 | case Opt_subvol: | |
725 | kfree(*subvol_name); | |
726 | *subvol_name = match_strdup(&args[0]); | |
727 | if (!*subvol_name) { | |
728 | error = -ENOMEM; | |
729 | goto out; | |
730 | } | |
731 | break; | |
732 | case Opt_subvolid: | |
733 | num = match_strdup(&args[0]); | |
734 | if (num) { | |
735 | *subvol_objectid = memparse(num, NULL); | |
736 | kfree(num); | |
737 | /* we want the original fs_tree */ | |
738 | if (!*subvol_objectid) | |
739 | *subvol_objectid = | |
740 | BTRFS_FS_TREE_OBJECTID; | |
741 | } else { | |
742 | error = -EINVAL; | |
743 | goto out; | |
744 | } | |
745 | break; | |
746 | case Opt_subvolrootid: | |
747 | printk(KERN_WARNING | |
748 | "btrfs: 'subvolrootid' mount option is deprecated and has no effect\n"); | |
749 | break; | |
750 | case Opt_device: | |
751 | device_name = match_strdup(&args[0]); | |
752 | if (!device_name) { | |
753 | error = -ENOMEM; | |
754 | goto out; | |
755 | } | |
756 | error = btrfs_scan_one_device(device_name, | |
757 | flags, holder, fs_devices); | |
758 | kfree(device_name); | |
759 | if (error) | |
760 | goto out; | |
761 | break; | |
762 | default: | |
763 | break; | |
764 | } | |
765 | } | |
766 | ||
767 | out: | |
768 | kfree(orig); | |
769 | return error; | |
770 | } | |
771 | ||
772 | static struct dentry *get_default_root(struct super_block *sb, | |
773 | u64 subvol_objectid) | |
774 | { | |
775 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
776 | struct btrfs_root *root = fs_info->tree_root; | |
777 | struct btrfs_root *new_root; | |
778 | struct btrfs_dir_item *di; | |
779 | struct btrfs_path *path; | |
780 | struct btrfs_key location; | |
781 | struct inode *inode; | |
782 | u64 dir_id; | |
783 | int new = 0; | |
784 | ||
785 | /* | |
786 | * We have a specific subvol we want to mount, just setup location and | |
787 | * go look up the root. | |
788 | */ | |
789 | if (subvol_objectid) { | |
790 | location.objectid = subvol_objectid; | |
791 | location.type = BTRFS_ROOT_ITEM_KEY; | |
792 | location.offset = (u64)-1; | |
793 | goto find_root; | |
794 | } | |
795 | ||
796 | path = btrfs_alloc_path(); | |
797 | if (!path) | |
798 | return ERR_PTR(-ENOMEM); | |
799 | path->leave_spinning = 1; | |
800 | ||
801 | /* | |
802 | * Find the "default" dir item which points to the root item that we | |
803 | * will mount by default if we haven't been given a specific subvolume | |
804 | * to mount. | |
805 | */ | |
806 | dir_id = btrfs_super_root_dir(fs_info->super_copy); | |
807 | di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0); | |
808 | if (IS_ERR(di)) { | |
809 | btrfs_free_path(path); | |
810 | return ERR_CAST(di); | |
811 | } | |
812 | if (!di) { | |
813 | /* | |
814 | * Ok the default dir item isn't there. This is weird since | |
815 | * it's always been there, but don't freak out, just try and | |
816 | * mount to root most subvolume. | |
817 | */ | |
818 | btrfs_free_path(path); | |
819 | dir_id = BTRFS_FIRST_FREE_OBJECTID; | |
820 | new_root = fs_info->fs_root; | |
821 | goto setup_root; | |
822 | } | |
823 | ||
824 | btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); | |
825 | btrfs_free_path(path); | |
826 | ||
827 | find_root: | |
828 | new_root = btrfs_read_fs_root_no_name(fs_info, &location); | |
829 | if (IS_ERR(new_root)) | |
830 | return ERR_CAST(new_root); | |
831 | ||
832 | dir_id = btrfs_root_dirid(&new_root->root_item); | |
833 | setup_root: | |
834 | location.objectid = dir_id; | |
835 | location.type = BTRFS_INODE_ITEM_KEY; | |
836 | location.offset = 0; | |
837 | ||
838 | inode = btrfs_iget(sb, &location, new_root, &new); | |
839 | if (IS_ERR(inode)) | |
840 | return ERR_CAST(inode); | |
841 | ||
842 | /* | |
843 | * If we're just mounting the root most subvol put the inode and return | |
844 | * a reference to the dentry. We will have already gotten a reference | |
845 | * to the inode in btrfs_fill_super so we're good to go. | |
846 | */ | |
847 | if (!new && sb->s_root->d_inode == inode) { | |
848 | iput(inode); | |
849 | return dget(sb->s_root); | |
850 | } | |
851 | ||
852 | return d_obtain_alias(inode); | |
853 | } | |
854 | ||
855 | static int btrfs_fill_super(struct super_block *sb, | |
856 | struct btrfs_fs_devices *fs_devices, | |
857 | void *data, int silent) | |
858 | { | |
859 | struct inode *inode; | |
860 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
861 | struct btrfs_key key; | |
862 | int err; | |
863 | ||
864 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
865 | sb->s_magic = BTRFS_SUPER_MAGIC; | |
866 | sb->s_op = &btrfs_super_ops; | |
867 | sb->s_d_op = &btrfs_dentry_operations; | |
868 | sb->s_export_op = &btrfs_export_ops; | |
869 | sb->s_xattr = btrfs_xattr_handlers; | |
870 | sb->s_time_gran = 1; | |
871 | #ifdef CONFIG_BTRFS_FS_POSIX_ACL | |
872 | sb->s_flags |= MS_POSIXACL; | |
873 | #endif | |
874 | sb->s_flags |= MS_I_VERSION; | |
875 | err = open_ctree(sb, fs_devices, (char *)data); | |
876 | if (err) { | |
877 | printk("btrfs: open_ctree failed\n"); | |
878 | return err; | |
879 | } | |
880 | ||
881 | key.objectid = BTRFS_FIRST_FREE_OBJECTID; | |
882 | key.type = BTRFS_INODE_ITEM_KEY; | |
883 | key.offset = 0; | |
884 | inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL); | |
885 | if (IS_ERR(inode)) { | |
886 | err = PTR_ERR(inode); | |
887 | goto fail_close; | |
888 | } | |
889 | ||
890 | sb->s_root = d_make_root(inode); | |
891 | if (!sb->s_root) { | |
892 | err = -ENOMEM; | |
893 | goto fail_close; | |
894 | } | |
895 | ||
896 | save_mount_options(sb, data); | |
897 | cleancache_init_fs(sb); | |
898 | sb->s_flags |= MS_ACTIVE; | |
899 | return 0; | |
900 | ||
901 | fail_close: | |
902 | close_ctree(fs_info->tree_root); | |
903 | return err; | |
904 | } | |
905 | ||
906 | int btrfs_sync_fs(struct super_block *sb, int wait) | |
907 | { | |
908 | struct btrfs_trans_handle *trans; | |
909 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
910 | struct btrfs_root *root = fs_info->tree_root; | |
911 | ||
912 | trace_btrfs_sync_fs(wait); | |
913 | ||
914 | if (!wait) { | |
915 | filemap_flush(fs_info->btree_inode->i_mapping); | |
916 | return 0; | |
917 | } | |
918 | ||
919 | btrfs_wait_all_ordered_extents(fs_info, 1); | |
920 | ||
921 | trans = btrfs_attach_transaction_barrier(root); | |
922 | if (IS_ERR(trans)) { | |
923 | /* no transaction, don't bother */ | |
924 | if (PTR_ERR(trans) == -ENOENT) | |
925 | return 0; | |
926 | return PTR_ERR(trans); | |
927 | } | |
928 | return btrfs_commit_transaction(trans, root); | |
929 | } | |
930 | ||
931 | static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) | |
932 | { | |
933 | struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb); | |
934 | struct btrfs_root *root = info->tree_root; | |
935 | char *compress_type; | |
936 | ||
937 | if (btrfs_test_opt(root, DEGRADED)) | |
938 | seq_puts(seq, ",degraded"); | |
939 | if (btrfs_test_opt(root, NODATASUM)) | |
940 | seq_puts(seq, ",nodatasum"); | |
941 | if (btrfs_test_opt(root, NODATACOW)) | |
942 | seq_puts(seq, ",nodatacow"); | |
943 | if (btrfs_test_opt(root, NOBARRIER)) | |
944 | seq_puts(seq, ",nobarrier"); | |
945 | if (info->max_inline != 8192 * 1024) | |
946 | seq_printf(seq, ",max_inline=%llu", | |
947 | (unsigned long long)info->max_inline); | |
948 | if (info->alloc_start != 0) | |
949 | seq_printf(seq, ",alloc_start=%llu", | |
950 | (unsigned long long)info->alloc_start); | |
951 | if (info->thread_pool_size != min_t(unsigned long, | |
952 | num_online_cpus() + 2, 8)) | |
953 | seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); | |
954 | if (btrfs_test_opt(root, COMPRESS)) { | |
955 | if (info->compress_type == BTRFS_COMPRESS_ZLIB) | |
956 | compress_type = "zlib"; | |
957 | else | |
958 | compress_type = "lzo"; | |
959 | if (btrfs_test_opt(root, FORCE_COMPRESS)) | |
960 | seq_printf(seq, ",compress-force=%s", compress_type); | |
961 | else | |
962 | seq_printf(seq, ",compress=%s", compress_type); | |
963 | } | |
964 | if (btrfs_test_opt(root, NOSSD)) | |
965 | seq_puts(seq, ",nossd"); | |
966 | if (btrfs_test_opt(root, SSD_SPREAD)) | |
967 | seq_puts(seq, ",ssd_spread"); | |
968 | else if (btrfs_test_opt(root, SSD)) | |
969 | seq_puts(seq, ",ssd"); | |
970 | if (btrfs_test_opt(root, NOTREELOG)) | |
971 | seq_puts(seq, ",notreelog"); | |
972 | if (btrfs_test_opt(root, FLUSHONCOMMIT)) | |
973 | seq_puts(seq, ",flushoncommit"); | |
974 | if (btrfs_test_opt(root, DISCARD)) | |
975 | seq_puts(seq, ",discard"); | |
976 | if (!(root->fs_info->sb->s_flags & MS_POSIXACL)) | |
977 | seq_puts(seq, ",noacl"); | |
978 | if (btrfs_test_opt(root, SPACE_CACHE)) | |
979 | seq_puts(seq, ",space_cache"); | |
980 | else | |
981 | seq_puts(seq, ",nospace_cache"); | |
982 | if (btrfs_test_opt(root, CLEAR_CACHE)) | |
983 | seq_puts(seq, ",clear_cache"); | |
984 | if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) | |
985 | seq_puts(seq, ",user_subvol_rm_allowed"); | |
986 | if (btrfs_test_opt(root, ENOSPC_DEBUG)) | |
987 | seq_puts(seq, ",enospc_debug"); | |
988 | if (btrfs_test_opt(root, AUTO_DEFRAG)) | |
989 | seq_puts(seq, ",autodefrag"); | |
990 | if (btrfs_test_opt(root, INODE_MAP_CACHE)) | |
991 | seq_puts(seq, ",inode_cache"); | |
992 | if (btrfs_test_opt(root, SKIP_BALANCE)) | |
993 | seq_puts(seq, ",skip_balance"); | |
994 | if (btrfs_test_opt(root, RECOVERY)) | |
995 | seq_puts(seq, ",recovery"); | |
996 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY | |
997 | if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA)) | |
998 | seq_puts(seq, ",check_int_data"); | |
999 | else if (btrfs_test_opt(root, CHECK_INTEGRITY)) | |
1000 | seq_puts(seq, ",check_int"); | |
1001 | if (info->check_integrity_print_mask) | |
1002 | seq_printf(seq, ",check_int_print_mask=%d", | |
1003 | info->check_integrity_print_mask); | |
1004 | #endif | |
1005 | if (info->metadata_ratio) | |
1006 | seq_printf(seq, ",metadata_ratio=%d", | |
1007 | info->metadata_ratio); | |
1008 | if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR)) | |
1009 | seq_puts(seq, ",fatal_errors=panic"); | |
1010 | if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL) | |
1011 | seq_printf(seq, ",commit=%d", info->commit_interval); | |
1012 | return 0; | |
1013 | } | |
1014 | ||
1015 | static int btrfs_test_super(struct super_block *s, void *data) | |
1016 | { | |
1017 | struct btrfs_fs_info *p = data; | |
1018 | struct btrfs_fs_info *fs_info = btrfs_sb(s); | |
1019 | ||
1020 | return fs_info->fs_devices == p->fs_devices; | |
1021 | } | |
1022 | ||
1023 | static int btrfs_set_super(struct super_block *s, void *data) | |
1024 | { | |
1025 | int err = set_anon_super(s, data); | |
1026 | if (!err) | |
1027 | s->s_fs_info = data; | |
1028 | return err; | |
1029 | } | |
1030 | ||
1031 | /* | |
1032 | * subvolumes are identified by ino 256 | |
1033 | */ | |
1034 | static inline int is_subvolume_inode(struct inode *inode) | |
1035 | { | |
1036 | if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) | |
1037 | return 1; | |
1038 | return 0; | |
1039 | } | |
1040 | ||
1041 | /* | |
1042 | * This will strip out the subvol=%s argument for an argument string and add | |
1043 | * subvolid=0 to make sure we get the actual tree root for path walking to the | |
1044 | * subvol we want. | |
1045 | */ | |
1046 | static char *setup_root_args(char *args) | |
1047 | { | |
1048 | unsigned len = strlen(args) + 2 + 1; | |
1049 | char *src, *dst, *buf; | |
1050 | ||
1051 | /* | |
1052 | * We need the same args as before, but with this substitution: | |
1053 | * s!subvol=[^,]+!subvolid=0! | |
1054 | * | |
1055 | * Since the replacement string is up to 2 bytes longer than the | |
1056 | * original, allocate strlen(args) + 2 + 1 bytes. | |
1057 | */ | |
1058 | ||
1059 | src = strstr(args, "subvol="); | |
1060 | /* This shouldn't happen, but just in case.. */ | |
1061 | if (!src) | |
1062 | return NULL; | |
1063 | ||
1064 | buf = dst = kmalloc(len, GFP_NOFS); | |
1065 | if (!buf) | |
1066 | return NULL; | |
1067 | ||
1068 | /* | |
1069 | * If the subvol= arg is not at the start of the string, | |
1070 | * copy whatever precedes it into buf. | |
1071 | */ | |
1072 | if (src != args) { | |
1073 | *src++ = '\0'; | |
1074 | strcpy(buf, args); | |
1075 | dst += strlen(args); | |
1076 | } | |
1077 | ||
1078 | strcpy(dst, "subvolid=0"); | |
1079 | dst += strlen("subvolid=0"); | |
1080 | ||
1081 | /* | |
1082 | * If there is a "," after the original subvol=... string, | |
1083 | * copy that suffix into our buffer. Otherwise, we're done. | |
1084 | */ | |
1085 | src = strchr(src, ','); | |
1086 | if (src) | |
1087 | strcpy(dst, src); | |
1088 | ||
1089 | return buf; | |
1090 | } | |
1091 | ||
1092 | static struct dentry *mount_subvol(const char *subvol_name, int flags, | |
1093 | const char *device_name, char *data) | |
1094 | { | |
1095 | struct dentry *root; | |
1096 | struct vfsmount *mnt; | |
1097 | char *newargs; | |
1098 | ||
1099 | newargs = setup_root_args(data); | |
1100 | if (!newargs) | |
1101 | return ERR_PTR(-ENOMEM); | |
1102 | mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, | |
1103 | newargs); | |
1104 | kfree(newargs); | |
1105 | if (IS_ERR(mnt)) | |
1106 | return ERR_CAST(mnt); | |
1107 | ||
1108 | root = mount_subtree(mnt, subvol_name); | |
1109 | ||
1110 | if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) { | |
1111 | struct super_block *s = root->d_sb; | |
1112 | dput(root); | |
1113 | root = ERR_PTR(-EINVAL); | |
1114 | deactivate_locked_super(s); | |
1115 | printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n", | |
1116 | subvol_name); | |
1117 | } | |
1118 | ||
1119 | return root; | |
1120 | } | |
1121 | ||
1122 | /* | |
1123 | * Find a superblock for the given device / mount point. | |
1124 | * | |
1125 | * Note: This is based on get_sb_bdev from fs/super.c with a few additions | |
1126 | * for multiple device setup. Make sure to keep it in sync. | |
1127 | */ | |
1128 | static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags, | |
1129 | const char *device_name, void *data) | |
1130 | { | |
1131 | struct block_device *bdev = NULL; | |
1132 | struct super_block *s; | |
1133 | struct dentry *root; | |
1134 | struct btrfs_fs_devices *fs_devices = NULL; | |
1135 | struct btrfs_fs_info *fs_info = NULL; | |
1136 | fmode_t mode = FMODE_READ; | |
1137 | char *subvol_name = NULL; | |
1138 | u64 subvol_objectid = 0; | |
1139 | int error = 0; | |
1140 | ||
1141 | if (!(flags & MS_RDONLY)) | |
1142 | mode |= FMODE_WRITE; | |
1143 | ||
1144 | error = btrfs_parse_early_options(data, mode, fs_type, | |
1145 | &subvol_name, &subvol_objectid, | |
1146 | &fs_devices); | |
1147 | if (error) { | |
1148 | kfree(subvol_name); | |
1149 | return ERR_PTR(error); | |
1150 | } | |
1151 | ||
1152 | if (subvol_name) { | |
1153 | root = mount_subvol(subvol_name, flags, device_name, data); | |
1154 | kfree(subvol_name); | |
1155 | return root; | |
1156 | } | |
1157 | ||
1158 | error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices); | |
1159 | if (error) | |
1160 | return ERR_PTR(error); | |
1161 | ||
1162 | /* | |
1163 | * Setup a dummy root and fs_info for test/set super. This is because | |
1164 | * we don't actually fill this stuff out until open_ctree, but we need | |
1165 | * it for searching for existing supers, so this lets us do that and | |
1166 | * then open_ctree will properly initialize everything later. | |
1167 | */ | |
1168 | fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS); | |
1169 | if (!fs_info) | |
1170 | return ERR_PTR(-ENOMEM); | |
1171 | ||
1172 | fs_info->fs_devices = fs_devices; | |
1173 | ||
1174 | fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS); | |
1175 | fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS); | |
1176 | if (!fs_info->super_copy || !fs_info->super_for_commit) { | |
1177 | error = -ENOMEM; | |
1178 | goto error_fs_info; | |
1179 | } | |
1180 | ||
1181 | error = btrfs_open_devices(fs_devices, mode, fs_type); | |
1182 | if (error) | |
1183 | goto error_fs_info; | |
1184 | ||
1185 | if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) { | |
1186 | error = -EACCES; | |
1187 | goto error_close_devices; | |
1188 | } | |
1189 | ||
1190 | bdev = fs_devices->latest_bdev; | |
1191 | s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC, | |
1192 | fs_info); | |
1193 | if (IS_ERR(s)) { | |
1194 | error = PTR_ERR(s); | |
1195 | goto error_close_devices; | |
1196 | } | |
1197 | ||
1198 | if (s->s_root) { | |
1199 | btrfs_close_devices(fs_devices); | |
1200 | free_fs_info(fs_info); | |
1201 | if ((flags ^ s->s_flags) & MS_RDONLY) | |
1202 | error = -EBUSY; | |
1203 | } else { | |
1204 | char b[BDEVNAME_SIZE]; | |
1205 | ||
1206 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); | |
1207 | btrfs_sb(s)->bdev_holder = fs_type; | |
1208 | error = btrfs_fill_super(s, fs_devices, data, | |
1209 | flags & MS_SILENT ? 1 : 0); | |
1210 | } | |
1211 | ||
1212 | root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error); | |
1213 | if (IS_ERR(root)) | |
1214 | deactivate_locked_super(s); | |
1215 | ||
1216 | return root; | |
1217 | ||
1218 | error_close_devices: | |
1219 | btrfs_close_devices(fs_devices); | |
1220 | error_fs_info: | |
1221 | free_fs_info(fs_info); | |
1222 | return ERR_PTR(error); | |
1223 | } | |
1224 | ||
1225 | static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit) | |
1226 | { | |
1227 | spin_lock_irq(&workers->lock); | |
1228 | workers->max_workers = new_limit; | |
1229 | spin_unlock_irq(&workers->lock); | |
1230 | } | |
1231 | ||
1232 | static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info, | |
1233 | int new_pool_size, int old_pool_size) | |
1234 | { | |
1235 | if (new_pool_size == old_pool_size) | |
1236 | return; | |
1237 | ||
1238 | fs_info->thread_pool_size = new_pool_size; | |
1239 | ||
1240 | printk(KERN_INFO "btrfs: resize thread pool %d -> %d\n", | |
1241 | old_pool_size, new_pool_size); | |
1242 | ||
1243 | btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size); | |
1244 | btrfs_set_max_workers(&fs_info->workers, new_pool_size); | |
1245 | btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size); | |
1246 | btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size); | |
1247 | btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size); | |
1248 | btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size); | |
1249 | btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size); | |
1250 | btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size); | |
1251 | btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size); | |
1252 | btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size); | |
1253 | btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size); | |
1254 | btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size); | |
1255 | btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size); | |
1256 | btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers, | |
1257 | new_pool_size); | |
1258 | } | |
1259 | ||
1260 | static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info) | |
1261 | { | |
1262 | set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); | |
1263 | } | |
1264 | ||
1265 | static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info, | |
1266 | unsigned long old_opts, int flags) | |
1267 | { | |
1268 | if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && | |
1269 | (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || | |
1270 | (flags & MS_RDONLY))) { | |
1271 | /* wait for any defraggers to finish */ | |
1272 | wait_event(fs_info->transaction_wait, | |
1273 | (atomic_read(&fs_info->defrag_running) == 0)); | |
1274 | if (flags & MS_RDONLY) | |
1275 | sync_filesystem(fs_info->sb); | |
1276 | } | |
1277 | } | |
1278 | ||
1279 | static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info, | |
1280 | unsigned long old_opts) | |
1281 | { | |
1282 | /* | |
1283 | * We need cleanup all defragable inodes if the autodefragment is | |
1284 | * close or the fs is R/O. | |
1285 | */ | |
1286 | if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && | |
1287 | (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || | |
1288 | (fs_info->sb->s_flags & MS_RDONLY))) { | |
1289 | btrfs_cleanup_defrag_inodes(fs_info); | |
1290 | } | |
1291 | ||
1292 | clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); | |
1293 | } | |
1294 | ||
1295 | static int btrfs_remount(struct super_block *sb, int *flags, char *data) | |
1296 | { | |
1297 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
1298 | struct btrfs_root *root = fs_info->tree_root; | |
1299 | unsigned old_flags = sb->s_flags; | |
1300 | unsigned long old_opts = fs_info->mount_opt; | |
1301 | unsigned long old_compress_type = fs_info->compress_type; | |
1302 | u64 old_max_inline = fs_info->max_inline; | |
1303 | u64 old_alloc_start = fs_info->alloc_start; | |
1304 | int old_thread_pool_size = fs_info->thread_pool_size; | |
1305 | unsigned int old_metadata_ratio = fs_info->metadata_ratio; | |
1306 | int ret; | |
1307 | ||
1308 | btrfs_remount_prepare(fs_info); | |
1309 | ||
1310 | ret = btrfs_parse_options(root, data); | |
1311 | if (ret) { | |
1312 | ret = -EINVAL; | |
1313 | goto restore; | |
1314 | } | |
1315 | ||
1316 | btrfs_remount_begin(fs_info, old_opts, *flags); | |
1317 | btrfs_resize_thread_pool(fs_info, | |
1318 | fs_info->thread_pool_size, old_thread_pool_size); | |
1319 | ||
1320 | if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) | |
1321 | goto out; | |
1322 | ||
1323 | if (*flags & MS_RDONLY) { | |
1324 | /* | |
1325 | * this also happens on 'umount -rf' or on shutdown, when | |
1326 | * the filesystem is busy. | |
1327 | */ | |
1328 | sb->s_flags |= MS_RDONLY; | |
1329 | ||
1330 | btrfs_dev_replace_suspend_for_unmount(fs_info); | |
1331 | btrfs_scrub_cancel(fs_info); | |
1332 | btrfs_pause_balance(fs_info); | |
1333 | ||
1334 | ret = btrfs_commit_super(root); | |
1335 | if (ret) | |
1336 | goto restore; | |
1337 | } else { | |
1338 | if (fs_info->fs_devices->rw_devices == 0) { | |
1339 | ret = -EACCES; | |
1340 | goto restore; | |
1341 | } | |
1342 | ||
1343 | if (fs_info->fs_devices->missing_devices > | |
1344 | fs_info->num_tolerated_disk_barrier_failures && | |
1345 | !(*flags & MS_RDONLY)) { | |
1346 | printk(KERN_WARNING | |
1347 | "Btrfs: too many missing devices, writeable remount is not allowed\n"); | |
1348 | ret = -EACCES; | |
1349 | goto restore; | |
1350 | } | |
1351 | ||
1352 | if (btrfs_super_log_root(fs_info->super_copy) != 0) { | |
1353 | ret = -EINVAL; | |
1354 | goto restore; | |
1355 | } | |
1356 | ||
1357 | ret = btrfs_cleanup_fs_roots(fs_info); | |
1358 | if (ret) | |
1359 | goto restore; | |
1360 | ||
1361 | /* recover relocation */ | |
1362 | ret = btrfs_recover_relocation(root); | |
1363 | if (ret) | |
1364 | goto restore; | |
1365 | ||
1366 | ret = btrfs_resume_balance_async(fs_info); | |
1367 | if (ret) | |
1368 | goto restore; | |
1369 | ||
1370 | ret = btrfs_resume_dev_replace_async(fs_info); | |
1371 | if (ret) { | |
1372 | pr_warn("btrfs: failed to resume dev_replace\n"); | |
1373 | goto restore; | |
1374 | } | |
1375 | sb->s_flags &= ~MS_RDONLY; | |
1376 | } | |
1377 | out: | |
1378 | btrfs_remount_cleanup(fs_info, old_opts); | |
1379 | return 0; | |
1380 | ||
1381 | restore: | |
1382 | /* We've hit an error - don't reset MS_RDONLY */ | |
1383 | if (sb->s_flags & MS_RDONLY) | |
1384 | old_flags |= MS_RDONLY; | |
1385 | sb->s_flags = old_flags; | |
1386 | fs_info->mount_opt = old_opts; | |
1387 | fs_info->compress_type = old_compress_type; | |
1388 | fs_info->max_inline = old_max_inline; | |
1389 | mutex_lock(&fs_info->chunk_mutex); | |
1390 | fs_info->alloc_start = old_alloc_start; | |
1391 | mutex_unlock(&fs_info->chunk_mutex); | |
1392 | btrfs_resize_thread_pool(fs_info, | |
1393 | old_thread_pool_size, fs_info->thread_pool_size); | |
1394 | fs_info->metadata_ratio = old_metadata_ratio; | |
1395 | btrfs_remount_cleanup(fs_info, old_opts); | |
1396 | return ret; | |
1397 | } | |
1398 | ||
1399 | /* Used to sort the devices by max_avail(descending sort) */ | |
1400 | static int btrfs_cmp_device_free_bytes(const void *dev_info1, | |
1401 | const void *dev_info2) | |
1402 | { | |
1403 | if (((struct btrfs_device_info *)dev_info1)->max_avail > | |
1404 | ((struct btrfs_device_info *)dev_info2)->max_avail) | |
1405 | return -1; | |
1406 | else if (((struct btrfs_device_info *)dev_info1)->max_avail < | |
1407 | ((struct btrfs_device_info *)dev_info2)->max_avail) | |
1408 | return 1; | |
1409 | else | |
1410 | return 0; | |
1411 | } | |
1412 | ||
1413 | /* | |
1414 | * sort the devices by max_avail, in which max free extent size of each device | |
1415 | * is stored.(Descending Sort) | |
1416 | */ | |
1417 | static inline void btrfs_descending_sort_devices( | |
1418 | struct btrfs_device_info *devices, | |
1419 | size_t nr_devices) | |
1420 | { | |
1421 | sort(devices, nr_devices, sizeof(struct btrfs_device_info), | |
1422 | btrfs_cmp_device_free_bytes, NULL); | |
1423 | } | |
1424 | ||
1425 | /* | |
1426 | * The helper to calc the free space on the devices that can be used to store | |
1427 | * file data. | |
1428 | */ | |
1429 | static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes) | |
1430 | { | |
1431 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1432 | struct btrfs_device_info *devices_info; | |
1433 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; | |
1434 | struct btrfs_device *device; | |
1435 | u64 skip_space; | |
1436 | u64 type; | |
1437 | u64 avail_space; | |
1438 | u64 used_space; | |
1439 | u64 min_stripe_size; | |
1440 | int min_stripes = 1, num_stripes = 1; | |
1441 | int i = 0, nr_devices; | |
1442 | int ret; | |
1443 | ||
1444 | nr_devices = fs_info->fs_devices->open_devices; | |
1445 | BUG_ON(!nr_devices); | |
1446 | ||
1447 | devices_info = kmalloc(sizeof(*devices_info) * nr_devices, | |
1448 | GFP_NOFS); | |
1449 | if (!devices_info) | |
1450 | return -ENOMEM; | |
1451 | ||
1452 | /* calc min stripe number for data space alloction */ | |
1453 | type = btrfs_get_alloc_profile(root, 1); | |
1454 | if (type & BTRFS_BLOCK_GROUP_RAID0) { | |
1455 | min_stripes = 2; | |
1456 | num_stripes = nr_devices; | |
1457 | } else if (type & BTRFS_BLOCK_GROUP_RAID1) { | |
1458 | min_stripes = 2; | |
1459 | num_stripes = 2; | |
1460 | } else if (type & BTRFS_BLOCK_GROUP_RAID10) { | |
1461 | min_stripes = 4; | |
1462 | num_stripes = 4; | |
1463 | } | |
1464 | ||
1465 | if (type & BTRFS_BLOCK_GROUP_DUP) | |
1466 | min_stripe_size = 2 * BTRFS_STRIPE_LEN; | |
1467 | else | |
1468 | min_stripe_size = BTRFS_STRIPE_LEN; | |
1469 | ||
1470 | list_for_each_entry(device, &fs_devices->devices, dev_list) { | |
1471 | if (!device->in_fs_metadata || !device->bdev || | |
1472 | device->is_tgtdev_for_dev_replace) | |
1473 | continue; | |
1474 | ||
1475 | avail_space = device->total_bytes - device->bytes_used; | |
1476 | ||
1477 | /* align with stripe_len */ | |
1478 | do_div(avail_space, BTRFS_STRIPE_LEN); | |
1479 | avail_space *= BTRFS_STRIPE_LEN; | |
1480 | ||
1481 | /* | |
1482 | * In order to avoid overwritting the superblock on the drive, | |
1483 | * btrfs starts at an offset of at least 1MB when doing chunk | |
1484 | * allocation. | |
1485 | */ | |
1486 | skip_space = 1024 * 1024; | |
1487 | ||
1488 | /* user can set the offset in fs_info->alloc_start. */ | |
1489 | if (fs_info->alloc_start + BTRFS_STRIPE_LEN <= | |
1490 | device->total_bytes) | |
1491 | skip_space = max(fs_info->alloc_start, skip_space); | |
1492 | ||
1493 | /* | |
1494 | * btrfs can not use the free space in [0, skip_space - 1], | |
1495 | * we must subtract it from the total. In order to implement | |
1496 | * it, we account the used space in this range first. | |
1497 | */ | |
1498 | ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1, | |
1499 | &used_space); | |
1500 | if (ret) { | |
1501 | kfree(devices_info); | |
1502 | return ret; | |
1503 | } | |
1504 | ||
1505 | /* calc the free space in [0, skip_space - 1] */ | |
1506 | skip_space -= used_space; | |
1507 | ||
1508 | /* | |
1509 | * we can use the free space in [0, skip_space - 1], subtract | |
1510 | * it from the total. | |
1511 | */ | |
1512 | if (avail_space && avail_space >= skip_space) | |
1513 | avail_space -= skip_space; | |
1514 | else | |
1515 | avail_space = 0; | |
1516 | ||
1517 | if (avail_space < min_stripe_size) | |
1518 | continue; | |
1519 | ||
1520 | devices_info[i].dev = device; | |
1521 | devices_info[i].max_avail = avail_space; | |
1522 | ||
1523 | i++; | |
1524 | } | |
1525 | ||
1526 | nr_devices = i; | |
1527 | ||
1528 | btrfs_descending_sort_devices(devices_info, nr_devices); | |
1529 | ||
1530 | i = nr_devices - 1; | |
1531 | avail_space = 0; | |
1532 | while (nr_devices >= min_stripes) { | |
1533 | if (num_stripes > nr_devices) | |
1534 | num_stripes = nr_devices; | |
1535 | ||
1536 | if (devices_info[i].max_avail >= min_stripe_size) { | |
1537 | int j; | |
1538 | u64 alloc_size; | |
1539 | ||
1540 | avail_space += devices_info[i].max_avail * num_stripes; | |
1541 | alloc_size = devices_info[i].max_avail; | |
1542 | for (j = i + 1 - num_stripes; j <= i; j++) | |
1543 | devices_info[j].max_avail -= alloc_size; | |
1544 | } | |
1545 | i--; | |
1546 | nr_devices--; | |
1547 | } | |
1548 | ||
1549 | kfree(devices_info); | |
1550 | *free_bytes = avail_space; | |
1551 | return 0; | |
1552 | } | |
1553 | ||
1554 | static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) | |
1555 | { | |
1556 | struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb); | |
1557 | struct btrfs_super_block *disk_super = fs_info->super_copy; | |
1558 | struct list_head *head = &fs_info->space_info; | |
1559 | struct btrfs_space_info *found; | |
1560 | u64 total_used = 0; | |
1561 | u64 total_free_data = 0; | |
1562 | int bits = dentry->d_sb->s_blocksize_bits; | |
1563 | __be32 *fsid = (__be32 *)fs_info->fsid; | |
1564 | int ret; | |
1565 | ||
1566 | /* holding chunk_muext to avoid allocating new chunks */ | |
1567 | mutex_lock(&fs_info->chunk_mutex); | |
1568 | rcu_read_lock(); | |
1569 | list_for_each_entry_rcu(found, head, list) { | |
1570 | if (found->flags & BTRFS_BLOCK_GROUP_DATA) { | |
1571 | total_free_data += found->disk_total - found->disk_used; | |
1572 | total_free_data -= | |
1573 | btrfs_account_ro_block_groups_free_space(found); | |
1574 | } | |
1575 | ||
1576 | total_used += found->disk_used; | |
1577 | } | |
1578 | rcu_read_unlock(); | |
1579 | ||
1580 | buf->f_namelen = BTRFS_NAME_LEN; | |
1581 | buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits; | |
1582 | buf->f_bfree = buf->f_blocks - (total_used >> bits); | |
1583 | buf->f_bsize = dentry->d_sb->s_blocksize; | |
1584 | buf->f_type = BTRFS_SUPER_MAGIC; | |
1585 | buf->f_bavail = total_free_data; | |
1586 | ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data); | |
1587 | if (ret) { | |
1588 | mutex_unlock(&fs_info->chunk_mutex); | |
1589 | return ret; | |
1590 | } | |
1591 | buf->f_bavail += total_free_data; | |
1592 | buf->f_bavail = buf->f_bavail >> bits; | |
1593 | mutex_unlock(&fs_info->chunk_mutex); | |
1594 | ||
1595 | /* We treat it as constant endianness (it doesn't matter _which_) | |
1596 | because we want the fsid to come out the same whether mounted | |
1597 | on a big-endian or little-endian host */ | |
1598 | buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); | |
1599 | buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); | |
1600 | /* Mask in the root object ID too, to disambiguate subvols */ | |
1601 | buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32; | |
1602 | buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid; | |
1603 | ||
1604 | return 0; | |
1605 | } | |
1606 | ||
1607 | static void btrfs_kill_super(struct super_block *sb) | |
1608 | { | |
1609 | struct btrfs_fs_info *fs_info = btrfs_sb(sb); | |
1610 | kill_anon_super(sb); | |
1611 | free_fs_info(fs_info); | |
1612 | } | |
1613 | ||
1614 | static struct file_system_type btrfs_fs_type = { | |
1615 | .owner = THIS_MODULE, | |
1616 | .name = "btrfs", | |
1617 | .mount = btrfs_mount, | |
1618 | .kill_sb = btrfs_kill_super, | |
1619 | .fs_flags = FS_REQUIRES_DEV, | |
1620 | }; | |
1621 | MODULE_ALIAS_FS("btrfs"); | |
1622 | ||
1623 | /* | |
1624 | * used by btrfsctl to scan devices when no FS is mounted | |
1625 | */ | |
1626 | static long btrfs_control_ioctl(struct file *file, unsigned int cmd, | |
1627 | unsigned long arg) | |
1628 | { | |
1629 | struct btrfs_ioctl_vol_args *vol; | |
1630 | struct btrfs_fs_devices *fs_devices; | |
1631 | int ret = -ENOTTY; | |
1632 | ||
1633 | if (!capable(CAP_SYS_ADMIN)) | |
1634 | return -EPERM; | |
1635 | ||
1636 | vol = memdup_user((void __user *)arg, sizeof(*vol)); | |
1637 | if (IS_ERR(vol)) | |
1638 | return PTR_ERR(vol); | |
1639 | ||
1640 | switch (cmd) { | |
1641 | case BTRFS_IOC_SCAN_DEV: | |
1642 | ret = btrfs_scan_one_device(vol->name, FMODE_READ, | |
1643 | &btrfs_fs_type, &fs_devices); | |
1644 | break; | |
1645 | case BTRFS_IOC_DEVICES_READY: | |
1646 | ret = btrfs_scan_one_device(vol->name, FMODE_READ, | |
1647 | &btrfs_fs_type, &fs_devices); | |
1648 | if (ret) | |
1649 | break; | |
1650 | ret = !(fs_devices->num_devices == fs_devices->total_devices); | |
1651 | break; | |
1652 | } | |
1653 | ||
1654 | kfree(vol); | |
1655 | return ret; | |
1656 | } | |
1657 | ||
1658 | static int btrfs_freeze(struct super_block *sb) | |
1659 | { | |
1660 | struct btrfs_trans_handle *trans; | |
1661 | struct btrfs_root *root = btrfs_sb(sb)->tree_root; | |
1662 | ||
1663 | trans = btrfs_attach_transaction_barrier(root); | |
1664 | if (IS_ERR(trans)) { | |
1665 | /* no transaction, don't bother */ | |
1666 | if (PTR_ERR(trans) == -ENOENT) | |
1667 | return 0; | |
1668 | return PTR_ERR(trans); | |
1669 | } | |
1670 | return btrfs_commit_transaction(trans, root); | |
1671 | } | |
1672 | ||
1673 | static int btrfs_unfreeze(struct super_block *sb) | |
1674 | { | |
1675 | return 0; | |
1676 | } | |
1677 | ||
1678 | static int btrfs_show_devname(struct seq_file *m, struct dentry *root) | |
1679 | { | |
1680 | struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb); | |
1681 | struct btrfs_fs_devices *cur_devices; | |
1682 | struct btrfs_device *dev, *first_dev = NULL; | |
1683 | struct list_head *head; | |
1684 | struct rcu_string *name; | |
1685 | ||
1686 | mutex_lock(&fs_info->fs_devices->device_list_mutex); | |
1687 | cur_devices = fs_info->fs_devices; | |
1688 | while (cur_devices) { | |
1689 | head = &cur_devices->devices; | |
1690 | list_for_each_entry(dev, head, dev_list) { | |
1691 | if (dev->missing) | |
1692 | continue; | |
1693 | if (!first_dev || dev->devid < first_dev->devid) | |
1694 | first_dev = dev; | |
1695 | } | |
1696 | cur_devices = cur_devices->seed; | |
1697 | } | |
1698 | ||
1699 | if (first_dev) { | |
1700 | rcu_read_lock(); | |
1701 | name = rcu_dereference(first_dev->name); | |
1702 | seq_escape(m, name->str, " \t\n\\"); | |
1703 | rcu_read_unlock(); | |
1704 | } else { | |
1705 | WARN_ON(1); | |
1706 | } | |
1707 | mutex_unlock(&fs_info->fs_devices->device_list_mutex); | |
1708 | return 0; | |
1709 | } | |
1710 | ||
1711 | static const struct super_operations btrfs_super_ops = { | |
1712 | .drop_inode = btrfs_drop_inode, | |
1713 | .evict_inode = btrfs_evict_inode, | |
1714 | .put_super = btrfs_put_super, | |
1715 | .sync_fs = btrfs_sync_fs, | |
1716 | .show_options = btrfs_show_options, | |
1717 | .show_devname = btrfs_show_devname, | |
1718 | .write_inode = btrfs_write_inode, | |
1719 | .alloc_inode = btrfs_alloc_inode, | |
1720 | .destroy_inode = btrfs_destroy_inode, | |
1721 | .statfs = btrfs_statfs, | |
1722 | .remount_fs = btrfs_remount, | |
1723 | .freeze_fs = btrfs_freeze, | |
1724 | .unfreeze_fs = btrfs_unfreeze, | |
1725 | }; | |
1726 | ||
1727 | static const struct file_operations btrfs_ctl_fops = { | |
1728 | .unlocked_ioctl = btrfs_control_ioctl, | |
1729 | .compat_ioctl = btrfs_control_ioctl, | |
1730 | .owner = THIS_MODULE, | |
1731 | .llseek = noop_llseek, | |
1732 | }; | |
1733 | ||
1734 | static struct miscdevice btrfs_misc = { | |
1735 | .minor = BTRFS_MINOR, | |
1736 | .name = "btrfs-control", | |
1737 | .fops = &btrfs_ctl_fops | |
1738 | }; | |
1739 | ||
1740 | MODULE_ALIAS_MISCDEV(BTRFS_MINOR); | |
1741 | MODULE_ALIAS("devname:btrfs-control"); | |
1742 | ||
1743 | static int btrfs_interface_init(void) | |
1744 | { | |
1745 | return misc_register(&btrfs_misc); | |
1746 | } | |
1747 | ||
1748 | static void btrfs_interface_exit(void) | |
1749 | { | |
1750 | if (misc_deregister(&btrfs_misc) < 0) | |
1751 | printk(KERN_INFO "btrfs: misc_deregister failed for control device\n"); | |
1752 | } | |
1753 | ||
1754 | static void btrfs_print_info(void) | |
1755 | { | |
1756 | printk(KERN_INFO "Btrfs loaded" | |
1757 | #ifdef CONFIG_BTRFS_DEBUG | |
1758 | ", debug=on" | |
1759 | #endif | |
1760 | #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY | |
1761 | ", integrity-checker=on" | |
1762 | #endif | |
1763 | "\n"); | |
1764 | } | |
1765 | ||
1766 | static int btrfs_run_sanity_tests(void) | |
1767 | { | |
1768 | return btrfs_test_free_space_cache(); | |
1769 | } | |
1770 | ||
1771 | static int __init init_btrfs_fs(void) | |
1772 | { | |
1773 | int err; | |
1774 | ||
1775 | err = btrfs_init_sysfs(); | |
1776 | if (err) | |
1777 | return err; | |
1778 | ||
1779 | btrfs_init_compress(); | |
1780 | ||
1781 | err = btrfs_init_cachep(); | |
1782 | if (err) | |
1783 | goto free_compress; | |
1784 | ||
1785 | err = extent_io_init(); | |
1786 | if (err) | |
1787 | goto free_cachep; | |
1788 | ||
1789 | err = extent_map_init(); | |
1790 | if (err) | |
1791 | goto free_extent_io; | |
1792 | ||
1793 | err = ordered_data_init(); | |
1794 | if (err) | |
1795 | goto free_extent_map; | |
1796 | ||
1797 | err = btrfs_delayed_inode_init(); | |
1798 | if (err) | |
1799 | goto free_ordered_data; | |
1800 | ||
1801 | err = btrfs_auto_defrag_init(); | |
1802 | if (err) | |
1803 | goto free_delayed_inode; | |
1804 | ||
1805 | err = btrfs_delayed_ref_init(); | |
1806 | if (err) | |
1807 | goto free_auto_defrag; | |
1808 | ||
1809 | err = btrfs_interface_init(); | |
1810 | if (err) | |
1811 | goto free_delayed_ref; | |
1812 | ||
1813 | btrfs_init_lockdep(); | |
1814 | ||
1815 | btrfs_print_info(); | |
1816 | ||
1817 | err = btrfs_run_sanity_tests(); | |
1818 | if (err) | |
1819 | goto unregister_ioctl; | |
1820 | ||
1821 | err = register_filesystem(&btrfs_fs_type); | |
1822 | if (err) | |
1823 | goto unregister_ioctl; | |
1824 | ||
1825 | return 0; | |
1826 | ||
1827 | unregister_ioctl: | |
1828 | btrfs_interface_exit(); | |
1829 | free_delayed_ref: | |
1830 | btrfs_delayed_ref_exit(); | |
1831 | free_auto_defrag: | |
1832 | btrfs_auto_defrag_exit(); | |
1833 | free_delayed_inode: | |
1834 | btrfs_delayed_inode_exit(); | |
1835 | free_ordered_data: | |
1836 | ordered_data_exit(); | |
1837 | free_extent_map: | |
1838 | extent_map_exit(); | |
1839 | free_extent_io: | |
1840 | extent_io_exit(); | |
1841 | free_cachep: | |
1842 | btrfs_destroy_cachep(); | |
1843 | free_compress: | |
1844 | btrfs_exit_compress(); | |
1845 | btrfs_exit_sysfs(); | |
1846 | return err; | |
1847 | } | |
1848 | ||
1849 | static void __exit exit_btrfs_fs(void) | |
1850 | { | |
1851 | btrfs_destroy_cachep(); | |
1852 | btrfs_delayed_ref_exit(); | |
1853 | btrfs_auto_defrag_exit(); | |
1854 | btrfs_delayed_inode_exit(); | |
1855 | ordered_data_exit(); | |
1856 | extent_map_exit(); | |
1857 | extent_io_exit(); | |
1858 | btrfs_interface_exit(); | |
1859 | unregister_filesystem(&btrfs_fs_type); | |
1860 | btrfs_exit_sysfs(); | |
1861 | btrfs_cleanup_fs_uuids(); | |
1862 | btrfs_exit_compress(); | |
1863 | } | |
1864 | ||
1865 | module_init(init_btrfs_fs) | |
1866 | module_exit(exit_btrfs_fs) | |
1867 | ||
1868 | MODULE_LICENSE("GPL"); |