]>
Commit | Line | Data |
---|---|---|
6cbd5570 CM |
1 | /* |
2 | * Copyright (C) 2007 Oracle. All rights reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or | |
5 | * modify it under the terms of the GNU General Public | |
6 | * License v2 as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
11 | * General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public | |
14 | * License along with this program; if not, write to the | |
15 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
16 | * Boston, MA 021110-1307, USA. | |
17 | */ | |
18 | ||
4b82d6e4 | 19 | #include <linux/blkdev.h> |
2e635a27 | 20 | #include <linux/module.h> |
e20d96d6 | 21 | #include <linux/buffer_head.h> |
2e635a27 CM |
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> | |
a9572a15 | 27 | #include <linux/seq_file.h> |
2e635a27 | 28 | #include <linux/string.h> |
2e635a27 | 29 | #include <linux/backing-dev.h> |
4b82d6e4 | 30 | #include <linux/mount.h> |
dee26a9f | 31 | #include <linux/mpage.h> |
75dfe396 CM |
32 | #include <linux/swap.h> |
33 | #include <linux/writeback.h> | |
8fd17795 | 34 | #include <linux/statfs.h> |
08607c1b | 35 | #include <linux/compat.h> |
95e05289 | 36 | #include <linux/parser.h> |
c59f8951 | 37 | #include <linux/ctype.h> |
6da6abae | 38 | #include <linux/namei.h> |
a9218f6b | 39 | #include <linux/miscdevice.h> |
1bcbf313 | 40 | #include <linux/magic.h> |
5a0e3ad6 | 41 | #include <linux/slab.h> |
90a887c9 | 42 | #include <linux/cleancache.h> |
830c4adb | 43 | #include <linux/mnt_namespace.h> |
4b4e25f2 | 44 | #include "compat.h" |
16cdcec7 | 45 | #include "delayed-inode.h" |
2e635a27 | 46 | #include "ctree.h" |
e20d96d6 | 47 | #include "disk-io.h" |
d5719762 | 48 | #include "transaction.h" |
2c90e5d6 | 49 | #include "btrfs_inode.h" |
c5739bba | 50 | #include "ioctl.h" |
3a686375 | 51 | #include "print-tree.h" |
5103e947 | 52 | #include "xattr.h" |
8a4b83cc | 53 | #include "volumes.h" |
b3c3da71 | 54 | #include "version.h" |
be6e8dc0 | 55 | #include "export.h" |
c8b97818 | 56 | #include "compression.h" |
2e635a27 | 57 | |
1abe9b8a | 58 | #define CREATE_TRACE_POINTS |
59 | #include <trace/events/btrfs.h> | |
60 | ||
b87221de | 61 | static const struct super_operations btrfs_super_ops; |
830c4adb | 62 | static struct file_system_type btrfs_fs_type; |
75dfe396 | 63 | |
acce952b | 64 | static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno, |
65 | char nbuf[16]) | |
66 | { | |
67 | char *errstr = NULL; | |
68 | ||
69 | switch (errno) { | |
70 | case -EIO: | |
71 | errstr = "IO failure"; | |
72 | break; | |
73 | case -ENOMEM: | |
74 | errstr = "Out of memory"; | |
75 | break; | |
76 | case -EROFS: | |
77 | errstr = "Readonly filesystem"; | |
78 | break; | |
79 | default: | |
80 | if (nbuf) { | |
81 | if (snprintf(nbuf, 16, "error %d", -errno) >= 0) | |
82 | errstr = nbuf; | |
83 | } | |
84 | break; | |
85 | } | |
86 | ||
87 | return errstr; | |
88 | } | |
89 | ||
90 | static void __save_error_info(struct btrfs_fs_info *fs_info) | |
91 | { | |
92 | /* | |
93 | * today we only save the error info into ram. Long term we'll | |
94 | * also send it down to the disk | |
95 | */ | |
96 | fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR; | |
97 | } | |
98 | ||
99 | /* NOTE: | |
100 | * We move write_super stuff at umount in order to avoid deadlock | |
101 | * for umount hold all lock. | |
102 | */ | |
103 | static void save_error_info(struct btrfs_fs_info *fs_info) | |
104 | { | |
105 | __save_error_info(fs_info); | |
106 | } | |
107 | ||
108 | /* btrfs handle error by forcing the filesystem readonly */ | |
109 | static void btrfs_handle_error(struct btrfs_fs_info *fs_info) | |
110 | { | |
111 | struct super_block *sb = fs_info->sb; | |
112 | ||
113 | if (sb->s_flags & MS_RDONLY) | |
114 | return; | |
115 | ||
116 | if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { | |
117 | sb->s_flags |= MS_RDONLY; | |
118 | printk(KERN_INFO "btrfs is forced readonly\n"); | |
119 | } | |
120 | } | |
121 | ||
122 | /* | |
123 | * __btrfs_std_error decodes expected errors from the caller and | |
124 | * invokes the approciate error response. | |
125 | */ | |
126 | void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function, | |
127 | unsigned int line, int errno) | |
128 | { | |
129 | struct super_block *sb = fs_info->sb; | |
130 | char nbuf[16]; | |
131 | const char *errstr; | |
132 | ||
133 | /* | |
134 | * Special case: if the error is EROFS, and we're already | |
135 | * under MS_RDONLY, then it is safe here. | |
136 | */ | |
137 | if (errno == -EROFS && (sb->s_flags & MS_RDONLY)) | |
138 | return; | |
139 | ||
140 | errstr = btrfs_decode_error(fs_info, errno, nbuf); | |
141 | printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n", | |
142 | sb->s_id, function, line, errstr); | |
143 | save_error_info(fs_info); | |
144 | ||
145 | btrfs_handle_error(fs_info); | |
146 | } | |
147 | ||
d397712b | 148 | static void btrfs_put_super(struct super_block *sb) |
b18c6685 | 149 | { |
39279cc3 | 150 | struct btrfs_root *root = btrfs_sb(sb); |
b18c6685 | 151 | int ret; |
b18c6685 | 152 | |
39279cc3 | 153 | ret = close_ctree(root); |
39279cc3 | 154 | sb->s_fs_info = NULL; |
559af821 AK |
155 | |
156 | (void)ret; /* FIXME: need to fix VFS to return error? */ | |
75dfe396 CM |
157 | } |
158 | ||
95e05289 | 159 | enum { |
73f73415 | 160 | Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum, |
287a0ab9 JB |
161 | Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd, |
162 | Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress, | |
261507a0 LZ |
163 | Opt_compress_type, Opt_compress_force, Opt_compress_force_type, |
164 | Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard, | |
91435650 | 165 | Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed, |
4b9465cb | 166 | Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, |
73bc1876 | 167 | Opt_inode_cache, Opt_no_space_cache, Opt_err, |
95e05289 CM |
168 | }; |
169 | ||
170 | static match_table_t tokens = { | |
dfe25020 | 171 | {Opt_degraded, "degraded"}, |
95e05289 | 172 | {Opt_subvol, "subvol=%s"}, |
73f73415 | 173 | {Opt_subvolid, "subvolid=%d"}, |
43e570b0 | 174 | {Opt_device, "device=%s"}, |
b6cda9bc | 175 | {Opt_nodatasum, "nodatasum"}, |
be20aa9d | 176 | {Opt_nodatacow, "nodatacow"}, |
21ad10cf | 177 | {Opt_nobarrier, "nobarrier"}, |
6f568d35 | 178 | {Opt_max_inline, "max_inline=%s"}, |
8f662a76 | 179 | {Opt_alloc_start, "alloc_start=%s"}, |
4543df7e | 180 | {Opt_thread_pool, "thread_pool=%d"}, |
c8b97818 | 181 | {Opt_compress, "compress"}, |
261507a0 | 182 | {Opt_compress_type, "compress=%s"}, |
a555f810 | 183 | {Opt_compress_force, "compress-force"}, |
261507a0 | 184 | {Opt_compress_force_type, "compress-force=%s"}, |
e18e4809 | 185 | {Opt_ssd, "ssd"}, |
451d7585 | 186 | {Opt_ssd_spread, "ssd_spread"}, |
3b30c22f | 187 | {Opt_nossd, "nossd"}, |
33268eaf | 188 | {Opt_noacl, "noacl"}, |
3a5e1404 | 189 | {Opt_notreelog, "notreelog"}, |
dccae999 | 190 | {Opt_flushoncommit, "flushoncommit"}, |
97e728d4 | 191 | {Opt_ratio, "metadata_ratio=%d"}, |
e244a0ae | 192 | {Opt_discard, "discard"}, |
0af3d00b | 193 | {Opt_space_cache, "space_cache"}, |
88c2ba3b | 194 | {Opt_clear_cache, "clear_cache"}, |
4260f7c7 | 195 | {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"}, |
91435650 | 196 | {Opt_enospc_debug, "enospc_debug"}, |
e15d0542 | 197 | {Opt_subvolrootid, "subvolrootid=%d"}, |
4cb5300b | 198 | {Opt_defrag, "autodefrag"}, |
4b9465cb | 199 | {Opt_inode_cache, "inode_cache"}, |
73bc1876 | 200 | {Opt_no_space_cache, "no_space_cache"}, |
33268eaf | 201 | {Opt_err, NULL}, |
95e05289 CM |
202 | }; |
203 | ||
edf24abe CH |
204 | /* |
205 | * Regular mount options parser. Everything that is needed only when | |
206 | * reading in a new superblock is parsed here. | |
207 | */ | |
208 | int btrfs_parse_options(struct btrfs_root *root, char *options) | |
95e05289 | 209 | { |
edf24abe | 210 | struct btrfs_fs_info *info = root->fs_info; |
95e05289 | 211 | substring_t args[MAX_OPT_ARGS]; |
73bc1876 JB |
212 | char *p, *num, *orig = NULL; |
213 | u64 cache_gen; | |
4543df7e | 214 | int intarg; |
a7a3f7ca | 215 | int ret = 0; |
261507a0 LZ |
216 | char *compress_type; |
217 | bool compress_force = false; | |
b6cda9bc | 218 | |
73bc1876 JB |
219 | cache_gen = btrfs_super_cache_generation(&root->fs_info->super_copy); |
220 | if (cache_gen) | |
221 | btrfs_set_opt(info->mount_opt, SPACE_CACHE); | |
222 | ||
95e05289 | 223 | if (!options) |
73bc1876 | 224 | goto out; |
95e05289 | 225 | |
be20aa9d CM |
226 | /* |
227 | * strsep changes the string, duplicate it because parse_options | |
228 | * gets called twice | |
229 | */ | |
230 | options = kstrdup(options, GFP_NOFS); | |
231 | if (!options) | |
232 | return -ENOMEM; | |
233 | ||
da495ecc | 234 | orig = options; |
be20aa9d | 235 | |
edf24abe | 236 | while ((p = strsep(&options, ",")) != NULL) { |
95e05289 CM |
237 | int token; |
238 | if (!*p) | |
239 | continue; | |
240 | ||
241 | token = match_token(p, tokens, args); | |
242 | switch (token) { | |
dfe25020 | 243 | case Opt_degraded: |
edf24abe CH |
244 | printk(KERN_INFO "btrfs: allowing degraded mounts\n"); |
245 | btrfs_set_opt(info->mount_opt, DEGRADED); | |
dfe25020 | 246 | break; |
95e05289 | 247 | case Opt_subvol: |
73f73415 | 248 | case Opt_subvolid: |
e15d0542 | 249 | case Opt_subvolrootid: |
43e570b0 | 250 | case Opt_device: |
edf24abe | 251 | /* |
43e570b0 | 252 | * These are parsed by btrfs_parse_early_options |
edf24abe CH |
253 | * and can be happily ignored here. |
254 | */ | |
b6cda9bc CM |
255 | break; |
256 | case Opt_nodatasum: | |
067c28ad | 257 | printk(KERN_INFO "btrfs: setting nodatasum\n"); |
edf24abe | 258 | btrfs_set_opt(info->mount_opt, NODATASUM); |
be20aa9d CM |
259 | break; |
260 | case Opt_nodatacow: | |
edf24abe CH |
261 | printk(KERN_INFO "btrfs: setting nodatacow\n"); |
262 | btrfs_set_opt(info->mount_opt, NODATACOW); | |
263 | btrfs_set_opt(info->mount_opt, NODATASUM); | |
95e05289 | 264 | break; |
a555f810 | 265 | case Opt_compress_force: |
261507a0 LZ |
266 | case Opt_compress_force_type: |
267 | compress_force = true; | |
268 | case Opt_compress: | |
269 | case Opt_compress_type: | |
270 | if (token == Opt_compress || | |
271 | token == Opt_compress_force || | |
272 | strcmp(args[0].from, "zlib") == 0) { | |
273 | compress_type = "zlib"; | |
274 | info->compress_type = BTRFS_COMPRESS_ZLIB; | |
a6fa6fae LZ |
275 | } else if (strcmp(args[0].from, "lzo") == 0) { |
276 | compress_type = "lzo"; | |
277 | info->compress_type = BTRFS_COMPRESS_LZO; | |
261507a0 LZ |
278 | } else { |
279 | ret = -EINVAL; | |
280 | goto out; | |
281 | } | |
282 | ||
a555f810 | 283 | btrfs_set_opt(info->mount_opt, COMPRESS); |
261507a0 LZ |
284 | if (compress_force) { |
285 | btrfs_set_opt(info->mount_opt, FORCE_COMPRESS); | |
286 | pr_info("btrfs: force %s compression\n", | |
287 | compress_type); | |
288 | } else | |
289 | pr_info("btrfs: use %s compression\n", | |
290 | compress_type); | |
a555f810 | 291 | break; |
e18e4809 | 292 | case Opt_ssd: |
edf24abe CH |
293 | printk(KERN_INFO "btrfs: use ssd allocation scheme\n"); |
294 | btrfs_set_opt(info->mount_opt, SSD); | |
e18e4809 | 295 | break; |
451d7585 CM |
296 | case Opt_ssd_spread: |
297 | printk(KERN_INFO "btrfs: use spread ssd " | |
298 | "allocation scheme\n"); | |
299 | btrfs_set_opt(info->mount_opt, SSD); | |
300 | btrfs_set_opt(info->mount_opt, SSD_SPREAD); | |
301 | break; | |
3b30c22f | 302 | case Opt_nossd: |
451d7585 CM |
303 | printk(KERN_INFO "btrfs: not using ssd allocation " |
304 | "scheme\n"); | |
c289811c | 305 | btrfs_set_opt(info->mount_opt, NOSSD); |
3b30c22f | 306 | btrfs_clear_opt(info->mount_opt, SSD); |
451d7585 | 307 | btrfs_clear_opt(info->mount_opt, SSD_SPREAD); |
3b30c22f | 308 | break; |
21ad10cf | 309 | case Opt_nobarrier: |
edf24abe CH |
310 | printk(KERN_INFO "btrfs: turning off barriers\n"); |
311 | btrfs_set_opt(info->mount_opt, NOBARRIER); | |
21ad10cf | 312 | break; |
4543df7e CM |
313 | case Opt_thread_pool: |
314 | intarg = 0; | |
315 | match_int(&args[0], &intarg); | |
316 | if (intarg) { | |
317 | info->thread_pool_size = intarg; | |
318 | printk(KERN_INFO "btrfs: thread pool %d\n", | |
319 | info->thread_pool_size); | |
320 | } | |
321 | break; | |
6f568d35 | 322 | case Opt_max_inline: |
edf24abe CH |
323 | num = match_strdup(&args[0]); |
324 | if (num) { | |
91748467 | 325 | info->max_inline = memparse(num, NULL); |
edf24abe CH |
326 | kfree(num); |
327 | ||
15ada040 CM |
328 | if (info->max_inline) { |
329 | info->max_inline = max_t(u64, | |
330 | info->max_inline, | |
331 | root->sectorsize); | |
332 | } | |
edf24abe | 333 | printk(KERN_INFO "btrfs: max_inline at %llu\n", |
21380931 | 334 | (unsigned long long)info->max_inline); |
6f568d35 CM |
335 | } |
336 | break; | |
8f662a76 | 337 | case Opt_alloc_start: |
edf24abe CH |
338 | num = match_strdup(&args[0]); |
339 | if (num) { | |
91748467 | 340 | info->alloc_start = memparse(num, NULL); |
edf24abe CH |
341 | kfree(num); |
342 | printk(KERN_INFO | |
343 | "btrfs: allocations start at %llu\n", | |
21380931 | 344 | (unsigned long long)info->alloc_start); |
8f662a76 CM |
345 | } |
346 | break; | |
33268eaf JB |
347 | case Opt_noacl: |
348 | root->fs_info->sb->s_flags &= ~MS_POSIXACL; | |
349 | break; | |
3a5e1404 SW |
350 | case Opt_notreelog: |
351 | printk(KERN_INFO "btrfs: disabling tree log\n"); | |
352 | btrfs_set_opt(info->mount_opt, NOTREELOG); | |
353 | break; | |
dccae999 SW |
354 | case Opt_flushoncommit: |
355 | printk(KERN_INFO "btrfs: turning on flush-on-commit\n"); | |
356 | btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT); | |
357 | break; | |
97e728d4 JB |
358 | case Opt_ratio: |
359 | intarg = 0; | |
360 | match_int(&args[0], &intarg); | |
361 | if (intarg) { | |
362 | info->metadata_ratio = intarg; | |
363 | printk(KERN_INFO "btrfs: metadata ratio %d\n", | |
364 | info->metadata_ratio); | |
365 | } | |
366 | break; | |
e244a0ae CH |
367 | case Opt_discard: |
368 | btrfs_set_opt(info->mount_opt, DISCARD); | |
369 | break; | |
0af3d00b | 370 | case Opt_space_cache: |
0af3d00b | 371 | btrfs_set_opt(info->mount_opt, SPACE_CACHE); |
0de90876 | 372 | break; |
73bc1876 JB |
373 | case Opt_no_space_cache: |
374 | printk(KERN_INFO "btrfs: disabling disk space caching\n"); | |
375 | btrfs_clear_opt(info->mount_opt, SPACE_CACHE); | |
376 | break; | |
4b9465cb CM |
377 | case Opt_inode_cache: |
378 | printk(KERN_INFO "btrfs: enabling inode map caching\n"); | |
379 | btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE); | |
380 | break; | |
88c2ba3b JB |
381 | case Opt_clear_cache: |
382 | printk(KERN_INFO "btrfs: force clearing of disk cache\n"); | |
383 | btrfs_set_opt(info->mount_opt, CLEAR_CACHE); | |
0af3d00b | 384 | break; |
4260f7c7 SW |
385 | case Opt_user_subvol_rm_allowed: |
386 | btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED); | |
387 | break; | |
91435650 CM |
388 | case Opt_enospc_debug: |
389 | btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG); | |
390 | break; | |
4cb5300b CM |
391 | case Opt_defrag: |
392 | printk(KERN_INFO "btrfs: enabling auto defrag"); | |
393 | btrfs_set_opt(info->mount_opt, AUTO_DEFRAG); | |
394 | break; | |
a7a3f7ca SW |
395 | case Opt_err: |
396 | printk(KERN_INFO "btrfs: unrecognized mount option " | |
397 | "'%s'\n", p); | |
398 | ret = -EINVAL; | |
399 | goto out; | |
95e05289 | 400 | default: |
be20aa9d | 401 | break; |
95e05289 CM |
402 | } |
403 | } | |
a7a3f7ca | 404 | out: |
73bc1876 JB |
405 | if (!ret && btrfs_test_opt(root, SPACE_CACHE)) |
406 | printk(KERN_INFO "btrfs: disk space caching is enabled\n"); | |
da495ecc | 407 | kfree(orig); |
a7a3f7ca | 408 | return ret; |
edf24abe CH |
409 | } |
410 | ||
411 | /* | |
412 | * Parse mount options that are required early in the mount process. | |
413 | * | |
414 | * All other options will be parsed on much later in the mount process and | |
415 | * only when we need to allocate a new super block. | |
416 | */ | |
97288f2c | 417 | static int btrfs_parse_early_options(const char *options, fmode_t flags, |
73f73415 | 418 | void *holder, char **subvol_name, u64 *subvol_objectid, |
e15d0542 | 419 | u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices) |
edf24abe CH |
420 | { |
421 | substring_t args[MAX_OPT_ARGS]; | |
83c8c9bd | 422 | char *device_name, *opts, *orig, *p; |
edf24abe | 423 | int error = 0; |
73f73415 | 424 | int intarg; |
edf24abe CH |
425 | |
426 | if (!options) | |
830c4adb | 427 | return 0; |
edf24abe CH |
428 | |
429 | /* | |
430 | * strsep changes the string, duplicate it because parse_options | |
431 | * gets called twice | |
432 | */ | |
433 | opts = kstrdup(options, GFP_KERNEL); | |
434 | if (!opts) | |
435 | return -ENOMEM; | |
3f3d0bc0 | 436 | orig = opts; |
edf24abe CH |
437 | |
438 | while ((p = strsep(&opts, ",")) != NULL) { | |
439 | int token; | |
440 | if (!*p) | |
441 | continue; | |
442 | ||
443 | token = match_token(p, tokens, args); | |
444 | switch (token) { | |
445 | case Opt_subvol: | |
446 | *subvol_name = match_strdup(&args[0]); | |
447 | break; | |
73f73415 JB |
448 | case Opt_subvolid: |
449 | intarg = 0; | |
4849f01d JB |
450 | error = match_int(&args[0], &intarg); |
451 | if (!error) { | |
452 | /* we want the original fs_tree */ | |
453 | if (!intarg) | |
454 | *subvol_objectid = | |
455 | BTRFS_FS_TREE_OBJECTID; | |
456 | else | |
457 | *subvol_objectid = intarg; | |
458 | } | |
73f73415 | 459 | break; |
e15d0542 XZ |
460 | case Opt_subvolrootid: |
461 | intarg = 0; | |
462 | error = match_int(&args[0], &intarg); | |
463 | if (!error) { | |
464 | /* we want the original fs_tree */ | |
465 | if (!intarg) | |
466 | *subvol_rootid = | |
467 | BTRFS_FS_TREE_OBJECTID; | |
468 | else | |
469 | *subvol_rootid = intarg; | |
470 | } | |
471 | break; | |
43e570b0 | 472 | case Opt_device: |
83c8c9bd JL |
473 | device_name = match_strdup(&args[0]); |
474 | if (!device_name) { | |
475 | error = -ENOMEM; | |
476 | goto out; | |
477 | } | |
478 | error = btrfs_scan_one_device(device_name, | |
43e570b0 | 479 | flags, holder, fs_devices); |
83c8c9bd | 480 | kfree(device_name); |
43e570b0 | 481 | if (error) |
830c4adb | 482 | goto out; |
43e570b0 | 483 | break; |
edf24abe CH |
484 | default: |
485 | break; | |
486 | } | |
487 | } | |
488 | ||
830c4adb | 489 | out: |
3f3d0bc0 | 490 | kfree(orig); |
edf24abe | 491 | return error; |
95e05289 CM |
492 | } |
493 | ||
73f73415 JB |
494 | static struct dentry *get_default_root(struct super_block *sb, |
495 | u64 subvol_objectid) | |
496 | { | |
497 | struct btrfs_root *root = sb->s_fs_info; | |
498 | struct btrfs_root *new_root; | |
499 | struct btrfs_dir_item *di; | |
500 | struct btrfs_path *path; | |
501 | struct btrfs_key location; | |
502 | struct inode *inode; | |
73f73415 JB |
503 | u64 dir_id; |
504 | int new = 0; | |
505 | ||
506 | /* | |
507 | * We have a specific subvol we want to mount, just setup location and | |
508 | * go look up the root. | |
509 | */ | |
510 | if (subvol_objectid) { | |
511 | location.objectid = subvol_objectid; | |
512 | location.type = BTRFS_ROOT_ITEM_KEY; | |
513 | location.offset = (u64)-1; | |
514 | goto find_root; | |
515 | } | |
516 | ||
517 | path = btrfs_alloc_path(); | |
518 | if (!path) | |
519 | return ERR_PTR(-ENOMEM); | |
520 | path->leave_spinning = 1; | |
521 | ||
522 | /* | |
523 | * Find the "default" dir item which points to the root item that we | |
524 | * will mount by default if we haven't been given a specific subvolume | |
525 | * to mount. | |
526 | */ | |
527 | dir_id = btrfs_super_root_dir(&root->fs_info->super_copy); | |
528 | di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0); | |
b0839166 JL |
529 | if (IS_ERR(di)) { |
530 | btrfs_free_path(path); | |
fb4f6f91 | 531 | return ERR_CAST(di); |
b0839166 | 532 | } |
73f73415 JB |
533 | if (!di) { |
534 | /* | |
535 | * Ok the default dir item isn't there. This is weird since | |
536 | * it's always been there, but don't freak out, just try and | |
537 | * mount to root most subvolume. | |
538 | */ | |
539 | btrfs_free_path(path); | |
540 | dir_id = BTRFS_FIRST_FREE_OBJECTID; | |
541 | new_root = root->fs_info->fs_root; | |
542 | goto setup_root; | |
543 | } | |
544 | ||
545 | btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); | |
546 | btrfs_free_path(path); | |
547 | ||
548 | find_root: | |
549 | new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); | |
550 | if (IS_ERR(new_root)) | |
d0b678cb | 551 | return ERR_CAST(new_root); |
73f73415 JB |
552 | |
553 | if (btrfs_root_refs(&new_root->root_item) == 0) | |
554 | return ERR_PTR(-ENOENT); | |
555 | ||
556 | dir_id = btrfs_root_dirid(&new_root->root_item); | |
557 | setup_root: | |
558 | location.objectid = dir_id; | |
559 | location.type = BTRFS_INODE_ITEM_KEY; | |
560 | location.offset = 0; | |
561 | ||
562 | inode = btrfs_iget(sb, &location, new_root, &new); | |
4cbd1149 DC |
563 | if (IS_ERR(inode)) |
564 | return ERR_CAST(inode); | |
73f73415 JB |
565 | |
566 | /* | |
567 | * If we're just mounting the root most subvol put the inode and return | |
568 | * a reference to the dentry. We will have already gotten a reference | |
569 | * to the inode in btrfs_fill_super so we're good to go. | |
570 | */ | |
571 | if (!new && sb->s_root->d_inode == inode) { | |
572 | iput(inode); | |
573 | return dget(sb->s_root); | |
574 | } | |
575 | ||
ba5b8958 | 576 | return d_obtain_alias(inode); |
73f73415 JB |
577 | } |
578 | ||
d397712b | 579 | static int btrfs_fill_super(struct super_block *sb, |
8a4b83cc | 580 | struct btrfs_fs_devices *fs_devices, |
d397712b | 581 | void *data, int silent) |
75dfe396 | 582 | { |
d397712b CM |
583 | struct inode *inode; |
584 | struct dentry *root_dentry; | |
39279cc3 | 585 | struct btrfs_root *tree_root; |
5d4f98a2 | 586 | struct btrfs_key key; |
39279cc3 | 587 | int err; |
a429e513 | 588 | |
39279cc3 CM |
589 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
590 | sb->s_magic = BTRFS_SUPER_MAGIC; | |
591 | sb->s_op = &btrfs_super_ops; | |
af53d29a | 592 | sb->s_d_op = &btrfs_dentry_operations; |
be6e8dc0 | 593 | sb->s_export_op = &btrfs_export_ops; |
5103e947 | 594 | sb->s_xattr = btrfs_xattr_handlers; |
39279cc3 | 595 | sb->s_time_gran = 1; |
0eda294d | 596 | #ifdef CONFIG_BTRFS_FS_POSIX_ACL |
33268eaf | 597 | sb->s_flags |= MS_POSIXACL; |
49cf6f45 | 598 | #endif |
a429e513 | 599 | |
dfe25020 | 600 | tree_root = open_ctree(sb, fs_devices, (char *)data); |
6567e837 | 601 | |
e58ca020 | 602 | if (IS_ERR(tree_root)) { |
39279cc3 | 603 | printk("btrfs: open_ctree failed\n"); |
e58ca020 | 604 | return PTR_ERR(tree_root); |
a429e513 | 605 | } |
39279cc3 | 606 | sb->s_fs_info = tree_root; |
a429e513 | 607 | |
5d4f98a2 YZ |
608 | key.objectid = BTRFS_FIRST_FREE_OBJECTID; |
609 | key.type = BTRFS_INODE_ITEM_KEY; | |
610 | key.offset = 0; | |
73f73415 | 611 | inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL); |
5d4f98a2 YZ |
612 | if (IS_ERR(inode)) { |
613 | err = PTR_ERR(inode); | |
39279cc3 | 614 | goto fail_close; |
f254e52c | 615 | } |
f254e52c | 616 | |
39279cc3 CM |
617 | root_dentry = d_alloc_root(inode); |
618 | if (!root_dentry) { | |
619 | iput(inode); | |
620 | err = -ENOMEM; | |
621 | goto fail_close; | |
f254e52c | 622 | } |
58176a96 | 623 | |
39279cc3 | 624 | sb->s_root = root_dentry; |
6885f308 | 625 | |
6885f308 | 626 | save_mount_options(sb, data); |
90a887c9 | 627 | cleancache_init_fs(sb); |
2619ba1f | 628 | return 0; |
39279cc3 CM |
629 | |
630 | fail_close: | |
631 | close_ctree(tree_root); | |
632 | return err; | |
2619ba1f CM |
633 | } |
634 | ||
6bf13c0c | 635 | int btrfs_sync_fs(struct super_block *sb, int wait) |
c5739bba CM |
636 | { |
637 | struct btrfs_trans_handle *trans; | |
dccae999 | 638 | struct btrfs_root *root = btrfs_sb(sb); |
c5739bba | 639 | int ret; |
2619ba1f | 640 | |
1abe9b8a | 641 | trace_btrfs_sync_fs(wait); |
642 | ||
39279cc3 CM |
643 | if (!wait) { |
644 | filemap_flush(root->fs_info->btree_inode->i_mapping); | |
645 | return 0; | |
646 | } | |
771ed689 | 647 | |
24bbcf04 YZ |
648 | btrfs_start_delalloc_inodes(root, 0); |
649 | btrfs_wait_ordered_extents(root, 0, 0); | |
771ed689 | 650 | |
a22285a6 | 651 | trans = btrfs_start_transaction(root, 0); |
98d5dc13 TI |
652 | if (IS_ERR(trans)) |
653 | return PTR_ERR(trans); | |
c5739bba | 654 | ret = btrfs_commit_transaction(trans, root); |
54aa1f4d | 655 | return ret; |
2c90e5d6 CM |
656 | } |
657 | ||
a9572a15 EP |
658 | static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs) |
659 | { | |
660 | struct btrfs_root *root = btrfs_sb(vfs->mnt_sb); | |
661 | struct btrfs_fs_info *info = root->fs_info; | |
200da64e | 662 | char *compress_type; |
a9572a15 EP |
663 | |
664 | if (btrfs_test_opt(root, DEGRADED)) | |
665 | seq_puts(seq, ",degraded"); | |
666 | if (btrfs_test_opt(root, NODATASUM)) | |
667 | seq_puts(seq, ",nodatasum"); | |
668 | if (btrfs_test_opt(root, NODATACOW)) | |
669 | seq_puts(seq, ",nodatacow"); | |
670 | if (btrfs_test_opt(root, NOBARRIER)) | |
671 | seq_puts(seq, ",nobarrier"); | |
a9572a15 | 672 | if (info->max_inline != 8192 * 1024) |
21380931 JB |
673 | seq_printf(seq, ",max_inline=%llu", |
674 | (unsigned long long)info->max_inline); | |
a9572a15 | 675 | if (info->alloc_start != 0) |
21380931 JB |
676 | seq_printf(seq, ",alloc_start=%llu", |
677 | (unsigned long long)info->alloc_start); | |
a9572a15 EP |
678 | if (info->thread_pool_size != min_t(unsigned long, |
679 | num_online_cpus() + 2, 8)) | |
680 | seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); | |
200da64e TI |
681 | if (btrfs_test_opt(root, COMPRESS)) { |
682 | if (info->compress_type == BTRFS_COMPRESS_ZLIB) | |
683 | compress_type = "zlib"; | |
684 | else | |
685 | compress_type = "lzo"; | |
686 | if (btrfs_test_opt(root, FORCE_COMPRESS)) | |
687 | seq_printf(seq, ",compress-force=%s", compress_type); | |
688 | else | |
689 | seq_printf(seq, ",compress=%s", compress_type); | |
690 | } | |
c289811c CM |
691 | if (btrfs_test_opt(root, NOSSD)) |
692 | seq_puts(seq, ",nossd"); | |
451d7585 CM |
693 | if (btrfs_test_opt(root, SSD_SPREAD)) |
694 | seq_puts(seq, ",ssd_spread"); | |
695 | else if (btrfs_test_opt(root, SSD)) | |
a9572a15 | 696 | seq_puts(seq, ",ssd"); |
3a5e1404 | 697 | if (btrfs_test_opt(root, NOTREELOG)) |
6b65c5c6 | 698 | seq_puts(seq, ",notreelog"); |
dccae999 | 699 | if (btrfs_test_opt(root, FLUSHONCOMMIT)) |
6b65c5c6 | 700 | seq_puts(seq, ",flushoncommit"); |
20a5239a MW |
701 | if (btrfs_test_opt(root, DISCARD)) |
702 | seq_puts(seq, ",discard"); | |
a9572a15 EP |
703 | if (!(root->fs_info->sb->s_flags & MS_POSIXACL)) |
704 | seq_puts(seq, ",noacl"); | |
200da64e TI |
705 | if (btrfs_test_opt(root, SPACE_CACHE)) |
706 | seq_puts(seq, ",space_cache"); | |
73bc1876 JB |
707 | else |
708 | seq_puts(seq, ",no_space_cache"); | |
200da64e TI |
709 | if (btrfs_test_opt(root, CLEAR_CACHE)) |
710 | seq_puts(seq, ",clear_cache"); | |
711 | if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) | |
712 | seq_puts(seq, ",user_subvol_rm_allowed"); | |
0942caa3 DS |
713 | if (btrfs_test_opt(root, ENOSPC_DEBUG)) |
714 | seq_puts(seq, ",enospc_debug"); | |
715 | if (btrfs_test_opt(root, AUTO_DEFRAG)) | |
716 | seq_puts(seq, ",autodefrag"); | |
717 | if (btrfs_test_opt(root, INODE_MAP_CACHE)) | |
718 | seq_puts(seq, ",inode_cache"); | |
a9572a15 EP |
719 | return 0; |
720 | } | |
721 | ||
a061fc8d | 722 | static int btrfs_test_super(struct super_block *s, void *data) |
4b82d6e4 | 723 | { |
450ba0ea | 724 | struct btrfs_root *test_root = data; |
a061fc8d | 725 | struct btrfs_root *root = btrfs_sb(s); |
4b82d6e4 | 726 | |
619c8c76 IK |
727 | /* |
728 | * If this super block is going away, return false as it | |
729 | * can't match as an existing super block. | |
730 | */ | |
731 | if (!atomic_read(&s->s_active)) | |
732 | return 0; | |
450ba0ea | 733 | return root->fs_info->fs_devices == test_root->fs_info->fs_devices; |
4b82d6e4 Y |
734 | } |
735 | ||
450ba0ea JB |
736 | static int btrfs_set_super(struct super_block *s, void *data) |
737 | { | |
738 | s->s_fs_info = data; | |
739 | ||
740 | return set_anon_super(s, data); | |
4b82d6e4 Y |
741 | } |
742 | ||
830c4adb JB |
743 | /* |
744 | * This will strip out the subvol=%s argument for an argument string and add | |
745 | * subvolid=0 to make sure we get the actual tree root for path walking to the | |
746 | * subvol we want. | |
747 | */ | |
748 | static char *setup_root_args(char *args) | |
749 | { | |
750 | unsigned copied = 0; | |
751 | unsigned len = strlen(args) + 2; | |
752 | char *pos; | |
753 | char *ret; | |
754 | ||
755 | /* | |
756 | * We need the same args as before, but minus | |
757 | * | |
758 | * subvol=a | |
759 | * | |
760 | * and add | |
761 | * | |
762 | * subvolid=0 | |
763 | * | |
764 | * which is a difference of 2 characters, so we allocate strlen(args) + | |
765 | * 2 characters. | |
766 | */ | |
767 | ret = kzalloc(len * sizeof(char), GFP_NOFS); | |
768 | if (!ret) | |
769 | return NULL; | |
770 | pos = strstr(args, "subvol="); | |
771 | ||
772 | /* This shouldn't happen, but just in case.. */ | |
773 | if (!pos) { | |
774 | kfree(ret); | |
775 | return NULL; | |
776 | } | |
777 | ||
778 | /* | |
779 | * The subvol=<> arg is not at the front of the string, copy everybody | |
780 | * up to that into ret. | |
781 | */ | |
782 | if (pos != args) { | |
783 | *pos = '\0'; | |
784 | strcpy(ret, args); | |
785 | copied += strlen(args); | |
786 | pos++; | |
787 | } | |
788 | ||
789 | strncpy(ret + copied, "subvolid=0", len - copied); | |
790 | ||
791 | /* Length of subvolid=0 */ | |
792 | copied += 10; | |
793 | ||
794 | /* | |
795 | * If there is no , after the subvol= option then we know there's no | |
796 | * other options and we can just return. | |
797 | */ | |
798 | pos = strchr(pos, ','); | |
799 | if (!pos) | |
800 | return ret; | |
801 | ||
802 | /* Copy the rest of the arguments into our buffer */ | |
803 | strncpy(ret + copied, pos, len - copied); | |
804 | copied += strlen(pos); | |
805 | ||
806 | return ret; | |
807 | } | |
808 | ||
809 | static struct dentry *mount_subvol(const char *subvol_name, int flags, | |
810 | const char *device_name, char *data) | |
811 | { | |
812 | struct super_block *s; | |
813 | struct dentry *root; | |
814 | struct vfsmount *mnt; | |
815 | struct mnt_namespace *ns_private; | |
816 | char *newargs; | |
817 | struct path path; | |
818 | int error; | |
819 | ||
820 | newargs = setup_root_args(data); | |
821 | if (!newargs) | |
822 | return ERR_PTR(-ENOMEM); | |
823 | mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, | |
824 | newargs); | |
825 | kfree(newargs); | |
826 | if (IS_ERR(mnt)) | |
827 | return ERR_CAST(mnt); | |
828 | ||
829 | ns_private = create_mnt_ns(mnt); | |
830 | if (IS_ERR(ns_private)) { | |
831 | mntput(mnt); | |
832 | return ERR_CAST(ns_private); | |
833 | } | |
834 | ||
835 | /* | |
836 | * This will trigger the automount of the subvol so we can just | |
837 | * drop the mnt we have here and return the dentry that we | |
838 | * found. | |
839 | */ | |
840 | error = vfs_path_lookup(mnt->mnt_root, mnt, subvol_name, | |
841 | LOOKUP_FOLLOW, &path); | |
842 | put_mnt_ns(ns_private); | |
843 | if (error) | |
844 | return ERR_PTR(error); | |
845 | ||
846 | /* Get a ref to the sb and the dentry we found and return it */ | |
847 | s = path.mnt->mnt_sb; | |
848 | atomic_inc(&s->s_active); | |
849 | root = dget(path.dentry); | |
850 | path_put(&path); | |
851 | down_write(&s->s_umount); | |
852 | ||
853 | return root; | |
854 | } | |
450ba0ea | 855 | |
edf24abe CH |
856 | /* |
857 | * Find a superblock for the given device / mount point. | |
858 | * | |
859 | * Note: This is based on get_sb_bdev from fs/super.c with a few additions | |
860 | * for multiple device setup. Make sure to keep it in sync. | |
861 | */ | |
061dbc6b | 862 | static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags, |
306e16ce | 863 | const char *device_name, void *data) |
4b82d6e4 Y |
864 | { |
865 | struct block_device *bdev = NULL; | |
866 | struct super_block *s; | |
867 | struct dentry *root; | |
8a4b83cc | 868 | struct btrfs_fs_devices *fs_devices = NULL; |
450ba0ea JB |
869 | struct btrfs_root *tree_root = NULL; |
870 | struct btrfs_fs_info *fs_info = NULL; | |
97288f2c | 871 | fmode_t mode = FMODE_READ; |
73f73415 JB |
872 | char *subvol_name = NULL; |
873 | u64 subvol_objectid = 0; | |
e15d0542 | 874 | u64 subvol_rootid = 0; |
4b82d6e4 Y |
875 | int error = 0; |
876 | ||
97288f2c CH |
877 | if (!(flags & MS_RDONLY)) |
878 | mode |= FMODE_WRITE; | |
879 | ||
880 | error = btrfs_parse_early_options(data, mode, fs_type, | |
73f73415 | 881 | &subvol_name, &subvol_objectid, |
e15d0542 | 882 | &subvol_rootid, &fs_devices); |
edf24abe | 883 | if (error) |
061dbc6b | 884 | return ERR_PTR(error); |
edf24abe | 885 | |
830c4adb JB |
886 | if (subvol_name) { |
887 | root = mount_subvol(subvol_name, flags, device_name, data); | |
888 | kfree(subvol_name); | |
889 | return root; | |
890 | } | |
891 | ||
306e16ce | 892 | error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices); |
8a4b83cc | 893 | if (error) |
830c4adb | 894 | return ERR_PTR(error); |
4b82d6e4 | 895 | |
97288f2c | 896 | error = btrfs_open_devices(fs_devices, mode, fs_type); |
8a4b83cc | 897 | if (error) |
830c4adb | 898 | return ERR_PTR(error); |
8a4b83cc | 899 | |
2b82032c YZ |
900 | if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) { |
901 | error = -EACCES; | |
902 | goto error_close_devices; | |
903 | } | |
904 | ||
450ba0ea JB |
905 | /* |
906 | * Setup a dummy root and fs_info for test/set super. This is because | |
907 | * we don't actually fill this stuff out until open_ctree, but we need | |
908 | * it for searching for existing supers, so this lets us do that and | |
909 | * then open_ctree will properly initialize everything later. | |
910 | */ | |
911 | fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS); | |
912 | tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); | |
913 | if (!fs_info || !tree_root) { | |
914 | error = -ENOMEM; | |
915 | goto error_close_devices; | |
916 | } | |
917 | fs_info->tree_root = tree_root; | |
918 | fs_info->fs_devices = fs_devices; | |
919 | tree_root->fs_info = fs_info; | |
920 | ||
dfe25020 | 921 | bdev = fs_devices->latest_bdev; |
450ba0ea | 922 | s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root); |
830c4adb JB |
923 | if (IS_ERR(s)) { |
924 | error = PTR_ERR(s); | |
925 | goto error_close_devices; | |
926 | } | |
4b82d6e4 Y |
927 | |
928 | if (s->s_root) { | |
929 | if ((flags ^ s->s_flags) & MS_RDONLY) { | |
6f5bbff9 | 930 | deactivate_locked_super(s); |
830c4adb | 931 | return ERR_PTR(-EBUSY); |
4b82d6e4 Y |
932 | } |
933 | ||
2b82032c | 934 | btrfs_close_devices(fs_devices); |
bdc924bb IK |
935 | kfree(fs_info); |
936 | kfree(tree_root); | |
4b82d6e4 Y |
937 | } else { |
938 | char b[BDEVNAME_SIZE]; | |
939 | ||
9e1f1de0 | 940 | s->s_flags = flags | MS_NOSEC; |
4b82d6e4 | 941 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); |
5f524444 | 942 | btrfs_sb(s)->fs_info->bdev_holder = fs_type; |
8a4b83cc CM |
943 | error = btrfs_fill_super(s, fs_devices, data, |
944 | flags & MS_SILENT ? 1 : 0); | |
4b82d6e4 | 945 | if (error) { |
6f5bbff9 | 946 | deactivate_locked_super(s); |
830c4adb | 947 | return ERR_PTR(error); |
4b82d6e4 Y |
948 | } |
949 | ||
950 | s->s_flags |= MS_ACTIVE; | |
951 | } | |
952 | ||
830c4adb JB |
953 | root = get_default_root(s, subvol_objectid); |
954 | if (IS_ERR(root)) { | |
955 | deactivate_locked_super(s); | |
956 | return root; | |
4b82d6e4 Y |
957 | } |
958 | ||
061dbc6b | 959 | return root; |
4b82d6e4 | 960 | |
c146afad | 961 | error_close_devices: |
8a4b83cc | 962 | btrfs_close_devices(fs_devices); |
450ba0ea JB |
963 | kfree(fs_info); |
964 | kfree(tree_root); | |
061dbc6b | 965 | return ERR_PTR(error); |
4b82d6e4 | 966 | } |
2e635a27 | 967 | |
c146afad YZ |
968 | static int btrfs_remount(struct super_block *sb, int *flags, char *data) |
969 | { | |
970 | struct btrfs_root *root = btrfs_sb(sb); | |
971 | int ret; | |
972 | ||
b288052e CM |
973 | ret = btrfs_parse_options(root, data); |
974 | if (ret) | |
975 | return -EINVAL; | |
976 | ||
c146afad YZ |
977 | if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) |
978 | return 0; | |
979 | ||
980 | if (*flags & MS_RDONLY) { | |
981 | sb->s_flags |= MS_RDONLY; | |
982 | ||
983 | ret = btrfs_commit_super(root); | |
984 | WARN_ON(ret); | |
985 | } else { | |
2b82032c YZ |
986 | if (root->fs_info->fs_devices->rw_devices == 0) |
987 | return -EACCES; | |
988 | ||
c146afad YZ |
989 | if (btrfs_super_log_root(&root->fs_info->super_copy) != 0) |
990 | return -EINVAL; | |
991 | ||
d68fc57b | 992 | ret = btrfs_cleanup_fs_roots(root->fs_info); |
c146afad YZ |
993 | WARN_ON(ret); |
994 | ||
d68fc57b YZ |
995 | /* recover relocation */ |
996 | ret = btrfs_recover_relocation(root); | |
c146afad YZ |
997 | WARN_ON(ret); |
998 | ||
999 | sb->s_flags &= ~MS_RDONLY; | |
1000 | } | |
1001 | ||
1002 | return 0; | |
1003 | } | |
1004 | ||
bcd53741 AJ |
1005 | /* Used to sort the devices by max_avail(descending sort) */ |
1006 | static int btrfs_cmp_device_free_bytes(const void *dev_info1, | |
1007 | const void *dev_info2) | |
1008 | { | |
1009 | if (((struct btrfs_device_info *)dev_info1)->max_avail > | |
1010 | ((struct btrfs_device_info *)dev_info2)->max_avail) | |
1011 | return -1; | |
1012 | else if (((struct btrfs_device_info *)dev_info1)->max_avail < | |
1013 | ((struct btrfs_device_info *)dev_info2)->max_avail) | |
1014 | return 1; | |
1015 | else | |
1016 | return 0; | |
1017 | } | |
1018 | ||
1019 | /* | |
1020 | * sort the devices by max_avail, in which max free extent size of each device | |
1021 | * is stored.(Descending Sort) | |
1022 | */ | |
1023 | static inline void btrfs_descending_sort_devices( | |
1024 | struct btrfs_device_info *devices, | |
1025 | size_t nr_devices) | |
1026 | { | |
1027 | sort(devices, nr_devices, sizeof(struct btrfs_device_info), | |
1028 | btrfs_cmp_device_free_bytes, NULL); | |
1029 | } | |
1030 | ||
6d07bcec MX |
1031 | /* |
1032 | * The helper to calc the free space on the devices that can be used to store | |
1033 | * file data. | |
1034 | */ | |
1035 | static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes) | |
1036 | { | |
1037 | struct btrfs_fs_info *fs_info = root->fs_info; | |
1038 | struct btrfs_device_info *devices_info; | |
1039 | struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; | |
1040 | struct btrfs_device *device; | |
1041 | u64 skip_space; | |
1042 | u64 type; | |
1043 | u64 avail_space; | |
1044 | u64 used_space; | |
1045 | u64 min_stripe_size; | |
1046 | int min_stripes = 1; | |
1047 | int i = 0, nr_devices; | |
1048 | int ret; | |
1049 | ||
1050 | nr_devices = fs_info->fs_devices->rw_devices; | |
1051 | BUG_ON(!nr_devices); | |
1052 | ||
1053 | devices_info = kmalloc(sizeof(*devices_info) * nr_devices, | |
1054 | GFP_NOFS); | |
1055 | if (!devices_info) | |
1056 | return -ENOMEM; | |
1057 | ||
1058 | /* calc min stripe number for data space alloction */ | |
1059 | type = btrfs_get_alloc_profile(root, 1); | |
1060 | if (type & BTRFS_BLOCK_GROUP_RAID0) | |
1061 | min_stripes = 2; | |
1062 | else if (type & BTRFS_BLOCK_GROUP_RAID1) | |
1063 | min_stripes = 2; | |
1064 | else if (type & BTRFS_BLOCK_GROUP_RAID10) | |
1065 | min_stripes = 4; | |
1066 | ||
1067 | if (type & BTRFS_BLOCK_GROUP_DUP) | |
1068 | min_stripe_size = 2 * BTRFS_STRIPE_LEN; | |
1069 | else | |
1070 | min_stripe_size = BTRFS_STRIPE_LEN; | |
1071 | ||
1072 | list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) { | |
1073 | if (!device->in_fs_metadata) | |
1074 | continue; | |
1075 | ||
1076 | avail_space = device->total_bytes - device->bytes_used; | |
1077 | ||
1078 | /* align with stripe_len */ | |
1079 | do_div(avail_space, BTRFS_STRIPE_LEN); | |
1080 | avail_space *= BTRFS_STRIPE_LEN; | |
1081 | ||
1082 | /* | |
1083 | * In order to avoid overwritting the superblock on the drive, | |
1084 | * btrfs starts at an offset of at least 1MB when doing chunk | |
1085 | * allocation. | |
1086 | */ | |
1087 | skip_space = 1024 * 1024; | |
1088 | ||
1089 | /* user can set the offset in fs_info->alloc_start. */ | |
1090 | if (fs_info->alloc_start + BTRFS_STRIPE_LEN <= | |
1091 | device->total_bytes) | |
1092 | skip_space = max(fs_info->alloc_start, skip_space); | |
1093 | ||
1094 | /* | |
1095 | * btrfs can not use the free space in [0, skip_space - 1], | |
1096 | * we must subtract it from the total. In order to implement | |
1097 | * it, we account the used space in this range first. | |
1098 | */ | |
1099 | ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1, | |
1100 | &used_space); | |
1101 | if (ret) { | |
1102 | kfree(devices_info); | |
1103 | return ret; | |
1104 | } | |
1105 | ||
1106 | /* calc the free space in [0, skip_space - 1] */ | |
1107 | skip_space -= used_space; | |
1108 | ||
1109 | /* | |
1110 | * we can use the free space in [0, skip_space - 1], subtract | |
1111 | * it from the total. | |
1112 | */ | |
1113 | if (avail_space && avail_space >= skip_space) | |
1114 | avail_space -= skip_space; | |
1115 | else | |
1116 | avail_space = 0; | |
1117 | ||
1118 | if (avail_space < min_stripe_size) | |
1119 | continue; | |
1120 | ||
1121 | devices_info[i].dev = device; | |
1122 | devices_info[i].max_avail = avail_space; | |
1123 | ||
1124 | i++; | |
1125 | } | |
1126 | ||
1127 | nr_devices = i; | |
1128 | ||
1129 | btrfs_descending_sort_devices(devices_info, nr_devices); | |
1130 | ||
1131 | i = nr_devices - 1; | |
1132 | avail_space = 0; | |
1133 | while (nr_devices >= min_stripes) { | |
1134 | if (devices_info[i].max_avail >= min_stripe_size) { | |
1135 | int j; | |
1136 | u64 alloc_size; | |
1137 | ||
1138 | avail_space += devices_info[i].max_avail * min_stripes; | |
1139 | alloc_size = devices_info[i].max_avail; | |
1140 | for (j = i + 1 - min_stripes; j <= i; j++) | |
1141 | devices_info[j].max_avail -= alloc_size; | |
1142 | } | |
1143 | i--; | |
1144 | nr_devices--; | |
1145 | } | |
1146 | ||
1147 | kfree(devices_info); | |
1148 | *free_bytes = avail_space; | |
1149 | return 0; | |
1150 | } | |
1151 | ||
8fd17795 CM |
1152 | static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) |
1153 | { | |
1154 | struct btrfs_root *root = btrfs_sb(dentry->d_sb); | |
4b52dff6 | 1155 | struct btrfs_super_block *disk_super = &root->fs_info->super_copy; |
bd4d1088 JB |
1156 | struct list_head *head = &root->fs_info->space_info; |
1157 | struct btrfs_space_info *found; | |
1158 | u64 total_used = 0; | |
6d07bcec | 1159 | u64 total_free_data = 0; |
db94535d | 1160 | int bits = dentry->d_sb->s_blocksize_bits; |
9d03632e | 1161 | __be32 *fsid = (__be32 *)root->fs_info->fsid; |
6d07bcec | 1162 | int ret; |
8fd17795 | 1163 | |
6d07bcec MX |
1164 | /* holding chunk_muext to avoid allocating new chunks */ |
1165 | mutex_lock(&root->fs_info->chunk_mutex); | |
bd4d1088 | 1166 | rcu_read_lock(); |
89a55897 | 1167 | list_for_each_entry_rcu(found, head, list) { |
6d07bcec MX |
1168 | if (found->flags & BTRFS_BLOCK_GROUP_DATA) { |
1169 | total_free_data += found->disk_total - found->disk_used; | |
1170 | total_free_data -= | |
1171 | btrfs_account_ro_block_groups_free_space(found); | |
1172 | } | |
1173 | ||
b742bb82 | 1174 | total_used += found->disk_used; |
89a55897 | 1175 | } |
bd4d1088 JB |
1176 | rcu_read_unlock(); |
1177 | ||
8fd17795 | 1178 | buf->f_namelen = BTRFS_NAME_LEN; |
db94535d | 1179 | buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits; |
bd4d1088 | 1180 | buf->f_bfree = buf->f_blocks - (total_used >> bits); |
8fd17795 CM |
1181 | buf->f_bsize = dentry->d_sb->s_blocksize; |
1182 | buf->f_type = BTRFS_SUPER_MAGIC; | |
6d07bcec MX |
1183 | buf->f_bavail = total_free_data; |
1184 | ret = btrfs_calc_avail_data_space(root, &total_free_data); | |
1185 | if (ret) { | |
1186 | mutex_unlock(&root->fs_info->chunk_mutex); | |
1187 | return ret; | |
1188 | } | |
1189 | buf->f_bavail += total_free_data; | |
1190 | buf->f_bavail = buf->f_bavail >> bits; | |
1191 | mutex_unlock(&root->fs_info->chunk_mutex); | |
d397712b | 1192 | |
9d03632e | 1193 | /* We treat it as constant endianness (it doesn't matter _which_) |
d397712b | 1194 | because we want the fsid to come out the same whether mounted |
9d03632e DW |
1195 | on a big-endian or little-endian host */ |
1196 | buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); | |
1197 | buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); | |
32d48fa1 DW |
1198 | /* Mask in the root object ID too, to disambiguate subvols */ |
1199 | buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32; | |
1200 | buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid; | |
1201 | ||
8fd17795 CM |
1202 | return 0; |
1203 | } | |
b5133862 | 1204 | |
2e635a27 CM |
1205 | static struct file_system_type btrfs_fs_type = { |
1206 | .owner = THIS_MODULE, | |
1207 | .name = "btrfs", | |
061dbc6b | 1208 | .mount = btrfs_mount, |
a061fc8d | 1209 | .kill_sb = kill_anon_super, |
2e635a27 CM |
1210 | .fs_flags = FS_REQUIRES_DEV, |
1211 | }; | |
a9218f6b | 1212 | |
d352ac68 CM |
1213 | /* |
1214 | * used by btrfsctl to scan devices when no FS is mounted | |
1215 | */ | |
8a4b83cc CM |
1216 | static long btrfs_control_ioctl(struct file *file, unsigned int cmd, |
1217 | unsigned long arg) | |
1218 | { | |
1219 | struct btrfs_ioctl_vol_args *vol; | |
1220 | struct btrfs_fs_devices *fs_devices; | |
c071fcfd | 1221 | int ret = -ENOTTY; |
8a4b83cc | 1222 | |
e441d54d CM |
1223 | if (!capable(CAP_SYS_ADMIN)) |
1224 | return -EPERM; | |
1225 | ||
dae7b665 LZ |
1226 | vol = memdup_user((void __user *)arg, sizeof(*vol)); |
1227 | if (IS_ERR(vol)) | |
1228 | return PTR_ERR(vol); | |
c071fcfd | 1229 | |
8a4b83cc CM |
1230 | switch (cmd) { |
1231 | case BTRFS_IOC_SCAN_DEV: | |
97288f2c | 1232 | ret = btrfs_scan_one_device(vol->name, FMODE_READ, |
8a4b83cc CM |
1233 | &btrfs_fs_type, &fs_devices); |
1234 | break; | |
1235 | } | |
dae7b665 | 1236 | |
8a4b83cc | 1237 | kfree(vol); |
f819d837 | 1238 | return ret; |
8a4b83cc CM |
1239 | } |
1240 | ||
0176260f | 1241 | static int btrfs_freeze(struct super_block *sb) |
ed0dab6b Y |
1242 | { |
1243 | struct btrfs_root *root = btrfs_sb(sb); | |
a74a4b97 CM |
1244 | mutex_lock(&root->fs_info->transaction_kthread_mutex); |
1245 | mutex_lock(&root->fs_info->cleaner_mutex); | |
0176260f | 1246 | return 0; |
ed0dab6b Y |
1247 | } |
1248 | ||
0176260f | 1249 | static int btrfs_unfreeze(struct super_block *sb) |
ed0dab6b Y |
1250 | { |
1251 | struct btrfs_root *root = btrfs_sb(sb); | |
a74a4b97 CM |
1252 | mutex_unlock(&root->fs_info->cleaner_mutex); |
1253 | mutex_unlock(&root->fs_info->transaction_kthread_mutex); | |
0176260f | 1254 | return 0; |
ed0dab6b | 1255 | } |
2e635a27 | 1256 | |
b87221de | 1257 | static const struct super_operations btrfs_super_ops = { |
76dda93c | 1258 | .drop_inode = btrfs_drop_inode, |
bd555975 | 1259 | .evict_inode = btrfs_evict_inode, |
e20d96d6 | 1260 | .put_super = btrfs_put_super, |
d5719762 | 1261 | .sync_fs = btrfs_sync_fs, |
a9572a15 | 1262 | .show_options = btrfs_show_options, |
4730a4bc | 1263 | .write_inode = btrfs_write_inode, |
b5133862 | 1264 | .dirty_inode = btrfs_dirty_inode, |
2c90e5d6 CM |
1265 | .alloc_inode = btrfs_alloc_inode, |
1266 | .destroy_inode = btrfs_destroy_inode, | |
8fd17795 | 1267 | .statfs = btrfs_statfs, |
c146afad | 1268 | .remount_fs = btrfs_remount, |
0176260f LT |
1269 | .freeze_fs = btrfs_freeze, |
1270 | .unfreeze_fs = btrfs_unfreeze, | |
e20d96d6 | 1271 | }; |
a9218f6b CM |
1272 | |
1273 | static const struct file_operations btrfs_ctl_fops = { | |
1274 | .unlocked_ioctl = btrfs_control_ioctl, | |
1275 | .compat_ioctl = btrfs_control_ioctl, | |
1276 | .owner = THIS_MODULE, | |
6038f373 | 1277 | .llseek = noop_llseek, |
a9218f6b CM |
1278 | }; |
1279 | ||
1280 | static struct miscdevice btrfs_misc = { | |
578454ff | 1281 | .minor = BTRFS_MINOR, |
a9218f6b CM |
1282 | .name = "btrfs-control", |
1283 | .fops = &btrfs_ctl_fops | |
1284 | }; | |
1285 | ||
578454ff KS |
1286 | MODULE_ALIAS_MISCDEV(BTRFS_MINOR); |
1287 | MODULE_ALIAS("devname:btrfs-control"); | |
1288 | ||
a9218f6b CM |
1289 | static int btrfs_interface_init(void) |
1290 | { | |
1291 | return misc_register(&btrfs_misc); | |
1292 | } | |
1293 | ||
b2950863 | 1294 | static void btrfs_interface_exit(void) |
a9218f6b CM |
1295 | { |
1296 | if (misc_deregister(&btrfs_misc) < 0) | |
d397712b | 1297 | printk(KERN_INFO "misc_deregister failed for control device"); |
a9218f6b CM |
1298 | } |
1299 | ||
2e635a27 CM |
1300 | static int __init init_btrfs_fs(void) |
1301 | { | |
2c90e5d6 | 1302 | int err; |
58176a96 JB |
1303 | |
1304 | err = btrfs_init_sysfs(); | |
1305 | if (err) | |
1306 | return err; | |
1307 | ||
261507a0 | 1308 | err = btrfs_init_compress(); |
2c90e5d6 | 1309 | if (err) |
a74a4b97 | 1310 | goto free_sysfs; |
d1310b2e | 1311 | |
261507a0 LZ |
1312 | err = btrfs_init_cachep(); |
1313 | if (err) | |
1314 | goto free_compress; | |
1315 | ||
d1310b2e | 1316 | err = extent_io_init(); |
2f4cbe64 WB |
1317 | if (err) |
1318 | goto free_cachep; | |
1319 | ||
d1310b2e CM |
1320 | err = extent_map_init(); |
1321 | if (err) | |
1322 | goto free_extent_io; | |
1323 | ||
16cdcec7 | 1324 | err = btrfs_delayed_inode_init(); |
2f4cbe64 WB |
1325 | if (err) |
1326 | goto free_extent_map; | |
c8b97818 | 1327 | |
16cdcec7 MX |
1328 | err = btrfs_interface_init(); |
1329 | if (err) | |
1330 | goto free_delayed_inode; | |
1331 | ||
a9218f6b CM |
1332 | err = register_filesystem(&btrfs_fs_type); |
1333 | if (err) | |
1334 | goto unregister_ioctl; | |
b3c3da71 CM |
1335 | |
1336 | printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION); | |
2f4cbe64 WB |
1337 | return 0; |
1338 | ||
a9218f6b CM |
1339 | unregister_ioctl: |
1340 | btrfs_interface_exit(); | |
16cdcec7 MX |
1341 | free_delayed_inode: |
1342 | btrfs_delayed_inode_exit(); | |
2f4cbe64 WB |
1343 | free_extent_map: |
1344 | extent_map_exit(); | |
d1310b2e CM |
1345 | free_extent_io: |
1346 | extent_io_exit(); | |
2f4cbe64 WB |
1347 | free_cachep: |
1348 | btrfs_destroy_cachep(); | |
261507a0 LZ |
1349 | free_compress: |
1350 | btrfs_exit_compress(); | |
a74a4b97 | 1351 | free_sysfs: |
2f4cbe64 WB |
1352 | btrfs_exit_sysfs(); |
1353 | return err; | |
2e635a27 CM |
1354 | } |
1355 | ||
1356 | static void __exit exit_btrfs_fs(void) | |
1357 | { | |
39279cc3 | 1358 | btrfs_destroy_cachep(); |
16cdcec7 | 1359 | btrfs_delayed_inode_exit(); |
a52d9a80 | 1360 | extent_map_exit(); |
d1310b2e | 1361 | extent_io_exit(); |
a9218f6b | 1362 | btrfs_interface_exit(); |
2e635a27 | 1363 | unregister_filesystem(&btrfs_fs_type); |
58176a96 | 1364 | btrfs_exit_sysfs(); |
8a4b83cc | 1365 | btrfs_cleanup_fs_uuids(); |
261507a0 | 1366 | btrfs_exit_compress(); |
2e635a27 CM |
1367 | } |
1368 | ||
1369 | module_init(init_btrfs_fs) | |
1370 | module_exit(exit_btrfs_fs) | |
1371 | ||
1372 | MODULE_LICENSE("GPL"); |