]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
a805bad5 | 2 | * Copyright (c) 2000-2006 Silicon Graphics, Inc. |
7b718769 | 3 | * All Rights Reserved. |
1da177e4 | 4 | * |
7b718769 NS |
5 | * This program is free software; you can redistribute it and/or |
6 | * modify it under the terms of the GNU General Public License as | |
1da177e4 LT |
7 | * published by the Free Software Foundation. |
8 | * | |
7b718769 NS |
9 | * This program is distributed in the hope that it would be useful, |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
1da177e4 | 13 | * |
7b718769 NS |
14 | * You should have received a copy of the GNU General Public License |
15 | * along with this program; if not, write the Free Software Foundation, | |
16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
1da177e4 | 17 | */ |
0b1b213f | 18 | |
1da177e4 | 19 | #include "xfs.h" |
70a9883c | 20 | #include "xfs_shared.h" |
6ca1c906 | 21 | #include "xfs_format.h" |
239880ef DC |
22 | #include "xfs_log_format.h" |
23 | #include "xfs_trans_resv.h" | |
1da177e4 | 24 | #include "xfs_sb.h" |
1da177e4 | 25 | #include "xfs_mount.h" |
57062787 | 26 | #include "xfs_da_format.h" |
1da177e4 | 27 | #include "xfs_inode.h" |
a844f451 | 28 | #include "xfs_btree.h" |
1da177e4 | 29 | #include "xfs_bmap.h" |
a4fbe6ab | 30 | #include "xfs_alloc.h" |
1da177e4 | 31 | #include "xfs_error.h" |
9909c4aa | 32 | #include "xfs_fsops.h" |
239880ef | 33 | #include "xfs_trans.h" |
1da177e4 | 34 | #include "xfs_buf_item.h" |
239880ef | 35 | #include "xfs_log.h" |
a67d7c5f | 36 | #include "xfs_log_priv.h" |
9f8868ff | 37 | #include "xfs_da_btree.h" |
2b9ab5ab | 38 | #include "xfs_dir2.h" |
9f8868ff CH |
39 | #include "xfs_extfree_item.h" |
40 | #include "xfs_mru_cache.h" | |
41 | #include "xfs_inode_item.h" | |
6d8b79cf | 42 | #include "xfs_icache.h" |
0b1b213f | 43 | #include "xfs_trace.h" |
3ebe7d2d | 44 | #include "xfs_icreate_item.h" |
a4fbe6ab DC |
45 | #include "xfs_filestream.h" |
46 | #include "xfs_quota.h" | |
65b65735 | 47 | #include "xfs_sysfs.h" |
30cbc591 | 48 | #include "xfs_ondisk.h" |
5880f2d7 | 49 | #include "xfs_rmap_item.h" |
baf4bcac | 50 | #include "xfs_refcount_item.h" |
6413a014 | 51 | #include "xfs_bmap_item.h" |
5e7e605c | 52 | #include "xfs_reflink.h" |
1da177e4 LT |
53 | |
54 | #include <linux/namei.h> | |
ef510424 | 55 | #include <linux/dax.h> |
1da177e4 | 56 | #include <linux/init.h> |
5a0e3ad6 | 57 | #include <linux/slab.h> |
1da177e4 | 58 | #include <linux/mount.h> |
0829c360 | 59 | #include <linux/mempool.h> |
1da177e4 | 60 | #include <linux/writeback.h> |
4df08c52 | 61 | #include <linux/kthread.h> |
7dfb7103 | 62 | #include <linux/freezer.h> |
62a877e3 | 63 | #include <linux/parser.h> |
1da177e4 | 64 | |
b87221de | 65 | static const struct super_operations xfs_super_operations; |
0e51a8e1 | 66 | struct bio_set *xfs_ioend_bioset; |
65b65735 | 67 | |
e3aed1a0 | 68 | static struct kset *xfs_kset; /* top-level xfs sysfs dir */ |
65b65735 BF |
69 | #ifdef DEBUG |
70 | static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */ | |
71 | #endif | |
1da177e4 | 72 | |
62a877e3 CH |
73 | /* |
74 | * Table driven mount option parser. | |
62a877e3 CH |
75 | */ |
76 | enum { | |
2e74af0e ES |
77 | Opt_logbufs, Opt_logbsize, Opt_logdev, Opt_rtdev, Opt_biosize, |
78 | Opt_wsync, Opt_noalign, Opt_swalloc, Opt_sunit, Opt_swidth, Opt_nouuid, | |
79 | Opt_mtpt, Opt_grpid, Opt_nogrpid, Opt_bsdgroups, Opt_sysvgroups, | |
80 | Opt_allocsize, Opt_norecovery, Opt_barrier, Opt_nobarrier, | |
81 | Opt_inode64, Opt_inode32, Opt_ikeep, Opt_noikeep, | |
82 | Opt_largeio, Opt_nolargeio, Opt_attr2, Opt_noattr2, Opt_filestreams, | |
83 | Opt_quota, Opt_noquota, Opt_usrquota, Opt_grpquota, Opt_prjquota, | |
84 | Opt_uquota, Opt_gquota, Opt_pquota, | |
85 | Opt_uqnoenforce, Opt_gqnoenforce, Opt_pqnoenforce, Opt_qnoenforce, | |
86 | Opt_discard, Opt_nodiscard, Opt_dax, Opt_err, | |
62a877e3 CH |
87 | }; |
88 | ||
a447c093 | 89 | static const match_table_t tokens = { |
2e74af0e ES |
90 | {Opt_logbufs, "logbufs=%u"}, /* number of XFS log buffers */ |
91 | {Opt_logbsize, "logbsize=%s"}, /* size of XFS log buffers */ | |
92 | {Opt_logdev, "logdev=%s"}, /* log device */ | |
93 | {Opt_rtdev, "rtdev=%s"}, /* realtime I/O device */ | |
94 | {Opt_biosize, "biosize=%u"}, /* log2 of preferred buffered io size */ | |
95 | {Opt_wsync, "wsync"}, /* safe-mode nfs compatible mount */ | |
96 | {Opt_noalign, "noalign"}, /* turn off stripe alignment */ | |
97 | {Opt_swalloc, "swalloc"}, /* turn on stripe width allocation */ | |
98 | {Opt_sunit, "sunit=%u"}, /* data volume stripe unit */ | |
99 | {Opt_swidth, "swidth=%u"}, /* data volume stripe width */ | |
100 | {Opt_nouuid, "nouuid"}, /* ignore filesystem UUID */ | |
101 | {Opt_mtpt, "mtpt"}, /* filesystem mount point */ | |
102 | {Opt_grpid, "grpid"}, /* group-ID from parent directory */ | |
103 | {Opt_nogrpid, "nogrpid"}, /* group-ID from current process */ | |
104 | {Opt_bsdgroups, "bsdgroups"}, /* group-ID from parent directory */ | |
105 | {Opt_sysvgroups,"sysvgroups"}, /* group-ID from current process */ | |
106 | {Opt_allocsize, "allocsize=%s"},/* preferred allocation size */ | |
107 | {Opt_norecovery,"norecovery"}, /* don't run XFS recovery */ | |
2e74af0e ES |
108 | {Opt_inode64, "inode64"}, /* inodes can be allocated anywhere */ |
109 | {Opt_inode32, "inode32"}, /* inode allocation limited to | |
110 | * XFS_MAXINUMBER_32 */ | |
111 | {Opt_ikeep, "ikeep"}, /* do not free empty inode clusters */ | |
112 | {Opt_noikeep, "noikeep"}, /* free empty inode clusters */ | |
113 | {Opt_largeio, "largeio"}, /* report large I/O sizes in stat() */ | |
114 | {Opt_nolargeio, "nolargeio"}, /* do not report large I/O sizes | |
115 | * in stat(). */ | |
116 | {Opt_attr2, "attr2"}, /* do use attr2 attribute format */ | |
117 | {Opt_noattr2, "noattr2"}, /* do not use attr2 attribute format */ | |
118 | {Opt_filestreams,"filestreams"},/* use filestreams allocator */ | |
119 | {Opt_quota, "quota"}, /* disk quotas (user) */ | |
120 | {Opt_noquota, "noquota"}, /* no quotas */ | |
121 | {Opt_usrquota, "usrquota"}, /* user quota enabled */ | |
122 | {Opt_grpquota, "grpquota"}, /* group quota enabled */ | |
123 | {Opt_prjquota, "prjquota"}, /* project quota enabled */ | |
124 | {Opt_uquota, "uquota"}, /* user quota (IRIX variant) */ | |
125 | {Opt_gquota, "gquota"}, /* group quota (IRIX variant) */ | |
126 | {Opt_pquota, "pquota"}, /* project quota (IRIX variant) */ | |
127 | {Opt_uqnoenforce,"uqnoenforce"},/* user quota limit enforcement */ | |
128 | {Opt_gqnoenforce,"gqnoenforce"},/* group quota limit enforcement */ | |
129 | {Opt_pqnoenforce,"pqnoenforce"},/* project quota limit enforcement */ | |
130 | {Opt_qnoenforce, "qnoenforce"}, /* same as uqnoenforce */ | |
131 | {Opt_discard, "discard"}, /* Discard unused blocks */ | |
132 | {Opt_nodiscard, "nodiscard"}, /* Do not discard unused blocks */ | |
133 | ||
134 | {Opt_dax, "dax"}, /* Enable direct access to bdev pages */ | |
4cf4573d DC |
135 | |
136 | /* Deprecated mount options scheduled for removal */ | |
137 | {Opt_barrier, "barrier"}, /* use writer barriers for log write and | |
138 | * unwritten extent conversion */ | |
139 | {Opt_nobarrier, "nobarrier"}, /* .. disable */ | |
140 | ||
2e74af0e | 141 | {Opt_err, NULL}, |
62a877e3 CH |
142 | }; |
143 | ||
144 | ||
a841b64d | 145 | STATIC int |
2e74af0e | 146 | suffix_kstrtoint(const substring_t *s, unsigned int base, int *res) |
a67d7c5f | 147 | { |
a17164e5 | 148 | int last, shift_left_factor = 0, _res; |
2e74af0e ES |
149 | char *value; |
150 | int ret = 0; | |
151 | ||
152 | value = match_strdup(s); | |
153 | if (!value) | |
154 | return -ENOMEM; | |
a67d7c5f DC |
155 | |
156 | last = strlen(value) - 1; | |
157 | if (value[last] == 'K' || value[last] == 'k') { | |
158 | shift_left_factor = 10; | |
159 | value[last] = '\0'; | |
160 | } | |
161 | if (value[last] == 'M' || value[last] == 'm') { | |
162 | shift_left_factor = 20; | |
163 | value[last] = '\0'; | |
164 | } | |
165 | if (value[last] == 'G' || value[last] == 'g') { | |
166 | shift_left_factor = 30; | |
167 | value[last] = '\0'; | |
168 | } | |
169 | ||
2e74af0e ES |
170 | if (kstrtoint(value, base, &_res)) |
171 | ret = -EINVAL; | |
172 | kfree(value); | |
a17164e5 | 173 | *res = _res << shift_left_factor; |
2e74af0e | 174 | return ret; |
a67d7c5f DC |
175 | } |
176 | ||
9d565ffa CH |
177 | /* |
178 | * This function fills in xfs_mount_t fields based on mount args. | |
179 | * Note: the superblock has _not_ yet been read in. | |
180 | * | |
181 | * Note that this function leaks the various device name allocations on | |
182 | * failure. The caller takes care of them. | |
a08ee40a ES |
183 | * |
184 | * *sb is const because this is also used to test options on the remount | |
185 | * path, and we don't want this to have any side effects at remount time. | |
186 | * Today this function does not change *sb, but just to future-proof... | |
9d565ffa | 187 | */ |
a67d7c5f DC |
188 | STATIC int |
189 | xfs_parseargs( | |
190 | struct xfs_mount *mp, | |
288699fe | 191 | char *options) |
a67d7c5f | 192 | { |
a08ee40a | 193 | const struct super_block *sb = mp->m_super; |
2e74af0e ES |
194 | char *p; |
195 | substring_t args[MAX_OPT_ARGS]; | |
9d565ffa CH |
196 | int dsunit = 0; |
197 | int dswidth = 0; | |
198 | int iosize = 0; | |
c8ce540d | 199 | uint8_t iosizelog = 0; |
9d565ffa | 200 | |
4f10700a DC |
201 | /* |
202 | * set up the mount name first so all the errors will refer to the | |
203 | * correct device. | |
204 | */ | |
205 | mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL); | |
206 | if (!mp->m_fsname) | |
2451337d | 207 | return -ENOMEM; |
4f10700a DC |
208 | mp->m_fsname_len = strlen(mp->m_fsname) + 1; |
209 | ||
9d565ffa CH |
210 | /* |
211 | * Copy binary VFS mount flags we are interested in. | |
212 | */ | |
bc98a42c | 213 | if (sb_rdonly(sb)) |
9d565ffa CH |
214 | mp->m_flags |= XFS_MOUNT_RDONLY; |
215 | if (sb->s_flags & MS_DIRSYNC) | |
216 | mp->m_flags |= XFS_MOUNT_DIRSYNC; | |
217 | if (sb->s_flags & MS_SYNCHRONOUS) | |
218 | mp->m_flags |= XFS_MOUNT_WSYNC; | |
219 | ||
220 | /* | |
221 | * Set some default flags that could be cleared by the mount option | |
222 | * parsing. | |
223 | */ | |
224 | mp->m_flags |= XFS_MOUNT_BARRIER; | |
225 | mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE; | |
a67d7c5f | 226 | |
9d565ffa CH |
227 | /* |
228 | * These can be overridden by the mount option parsing. | |
229 | */ | |
230 | mp->m_logbufs = -1; | |
231 | mp->m_logbsize = -1; | |
a67d7c5f DC |
232 | |
233 | if (!options) | |
234 | goto done; | |
235 | ||
2e74af0e ES |
236 | while ((p = strsep(&options, ",")) != NULL) { |
237 | int token; | |
238 | ||
239 | if (!*p) | |
a67d7c5f | 240 | continue; |
a67d7c5f | 241 | |
2e74af0e ES |
242 | token = match_token(p, tokens, args); |
243 | switch (token) { | |
244 | case Opt_logbufs: | |
245 | if (match_int(args, &mp->m_logbufs)) | |
2451337d | 246 | return -EINVAL; |
2e74af0e ES |
247 | break; |
248 | case Opt_logbsize: | |
249 | if (suffix_kstrtoint(args, 10, &mp->m_logbsize)) | |
2451337d | 250 | return -EINVAL; |
2e74af0e ES |
251 | break; |
252 | case Opt_logdev: | |
253 | mp->m_logname = match_strdup(args); | |
9d565ffa | 254 | if (!mp->m_logname) |
2451337d | 255 | return -ENOMEM; |
2e74af0e ES |
256 | break; |
257 | case Opt_mtpt: | |
258 | xfs_warn(mp, "%s option not allowed on this system", p); | |
2451337d | 259 | return -EINVAL; |
2e74af0e ES |
260 | case Opt_rtdev: |
261 | mp->m_rtname = match_strdup(args); | |
9d565ffa | 262 | if (!mp->m_rtname) |
2451337d | 263 | return -ENOMEM; |
2e74af0e ES |
264 | break; |
265 | case Opt_allocsize: | |
266 | case Opt_biosize: | |
267 | if (suffix_kstrtoint(args, 10, &iosize)) | |
2451337d | 268 | return -EINVAL; |
9d565ffa | 269 | iosizelog = ffs(iosize) - 1; |
2e74af0e ES |
270 | break; |
271 | case Opt_grpid: | |
272 | case Opt_bsdgroups: | |
a67d7c5f | 273 | mp->m_flags |= XFS_MOUNT_GRPID; |
2e74af0e ES |
274 | break; |
275 | case Opt_nogrpid: | |
276 | case Opt_sysvgroups: | |
a67d7c5f | 277 | mp->m_flags &= ~XFS_MOUNT_GRPID; |
2e74af0e ES |
278 | break; |
279 | case Opt_wsync: | |
9d565ffa | 280 | mp->m_flags |= XFS_MOUNT_WSYNC; |
2e74af0e ES |
281 | break; |
282 | case Opt_norecovery: | |
9d565ffa | 283 | mp->m_flags |= XFS_MOUNT_NORECOVERY; |
2e74af0e ES |
284 | break; |
285 | case Opt_noalign: | |
9d565ffa | 286 | mp->m_flags |= XFS_MOUNT_NOALIGN; |
2e74af0e ES |
287 | break; |
288 | case Opt_swalloc: | |
9d565ffa | 289 | mp->m_flags |= XFS_MOUNT_SWALLOC; |
2e74af0e ES |
290 | break; |
291 | case Opt_sunit: | |
292 | if (match_int(args, &dsunit)) | |
2451337d | 293 | return -EINVAL; |
2e74af0e ES |
294 | break; |
295 | case Opt_swidth: | |
296 | if (match_int(args, &dswidth)) | |
2451337d | 297 | return -EINVAL; |
2e74af0e ES |
298 | break; |
299 | case Opt_inode32: | |
08bf5404 | 300 | mp->m_flags |= XFS_MOUNT_SMALL_INUMS; |
2e74af0e ES |
301 | break; |
302 | case Opt_inode64: | |
9d565ffa | 303 | mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS; |
2e74af0e ES |
304 | break; |
305 | case Opt_nouuid: | |
9d565ffa | 306 | mp->m_flags |= XFS_MOUNT_NOUUID; |
2e74af0e | 307 | break; |
2e74af0e | 308 | case Opt_ikeep: |
9d565ffa | 309 | mp->m_flags |= XFS_MOUNT_IKEEP; |
2e74af0e ES |
310 | break; |
311 | case Opt_noikeep: | |
9d565ffa | 312 | mp->m_flags &= ~XFS_MOUNT_IKEEP; |
2e74af0e ES |
313 | break; |
314 | case Opt_largeio: | |
9d565ffa | 315 | mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE; |
2e74af0e ES |
316 | break; |
317 | case Opt_nolargeio: | |
9d565ffa | 318 | mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE; |
2e74af0e ES |
319 | break; |
320 | case Opt_attr2: | |
9d565ffa | 321 | mp->m_flags |= XFS_MOUNT_ATTR2; |
2e74af0e ES |
322 | break; |
323 | case Opt_noattr2: | |
9d565ffa CH |
324 | mp->m_flags &= ~XFS_MOUNT_ATTR2; |
325 | mp->m_flags |= XFS_MOUNT_NOATTR2; | |
2e74af0e ES |
326 | break; |
327 | case Opt_filestreams: | |
9d565ffa | 328 | mp->m_flags |= XFS_MOUNT_FILESTREAMS; |
2e74af0e ES |
329 | break; |
330 | case Opt_noquota: | |
4177af3a CS |
331 | mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT; |
332 | mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD; | |
333 | mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE; | |
2e74af0e ES |
334 | break; |
335 | case Opt_quota: | |
336 | case Opt_uquota: | |
337 | case Opt_usrquota: | |
9d565ffa CH |
338 | mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE | |
339 | XFS_UQUOTA_ENFD); | |
2e74af0e ES |
340 | break; |
341 | case Opt_qnoenforce: | |
342 | case Opt_uqnoenforce: | |
9d565ffa CH |
343 | mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE); |
344 | mp->m_qflags &= ~XFS_UQUOTA_ENFD; | |
2e74af0e ES |
345 | break; |
346 | case Opt_pquota: | |
347 | case Opt_prjquota: | |
9d565ffa | 348 | mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE | |
83e782e1 | 349 | XFS_PQUOTA_ENFD); |
2e74af0e ES |
350 | break; |
351 | case Opt_pqnoenforce: | |
9d565ffa | 352 | mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE); |
83e782e1 | 353 | mp->m_qflags &= ~XFS_PQUOTA_ENFD; |
ce5c767d | 354 | break; |
2e74af0e ES |
355 | case Opt_gquota: |
356 | case Opt_grpquota: | |
9d565ffa | 357 | mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE | |
83e782e1 | 358 | XFS_GQUOTA_ENFD); |
2e74af0e ES |
359 | break; |
360 | case Opt_gqnoenforce: | |
9d565ffa | 361 | mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE); |
83e782e1 | 362 | mp->m_qflags &= ~XFS_GQUOTA_ENFD; |
2e74af0e ES |
363 | break; |
364 | case Opt_discard: | |
e84661aa | 365 | mp->m_flags |= XFS_MOUNT_DISCARD; |
2e74af0e ES |
366 | break; |
367 | case Opt_nodiscard: | |
e84661aa | 368 | mp->m_flags &= ~XFS_MOUNT_DISCARD; |
2e74af0e | 369 | break; |
cbe4dab1 | 370 | #ifdef CONFIG_FS_DAX |
2e74af0e | 371 | case Opt_dax: |
cbe4dab1 | 372 | mp->m_flags |= XFS_MOUNT_DAX; |
2e74af0e | 373 | break; |
cbe4dab1 | 374 | #endif |
4cf4573d DC |
375 | case Opt_barrier: |
376 | xfs_warn(mp, "%s option is deprecated, ignoring.", p); | |
377 | mp->m_flags |= XFS_MOUNT_BARRIER; | |
378 | break; | |
379 | case Opt_nobarrier: | |
380 | xfs_warn(mp, "%s option is deprecated, ignoring.", p); | |
381 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
382 | break; | |
2e74af0e ES |
383 | default: |
384 | xfs_warn(mp, "unknown mount option [%s].", p); | |
2451337d | 385 | return -EINVAL; |
a67d7c5f DC |
386 | } |
387 | } | |
388 | ||
9d565ffa CH |
389 | /* |
390 | * no recovery flag requires a read-only mount | |
391 | */ | |
392 | if ((mp->m_flags & XFS_MOUNT_NORECOVERY) && | |
393 | !(mp->m_flags & XFS_MOUNT_RDONLY)) { | |
4f10700a | 394 | xfs_warn(mp, "no-recovery mounts must be read-only."); |
2451337d | 395 | return -EINVAL; |
a67d7c5f DC |
396 | } |
397 | ||
9d565ffa | 398 | if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) { |
4f10700a DC |
399 | xfs_warn(mp, |
400 | "sunit and swidth options incompatible with the noalign option"); | |
2451337d | 401 | return -EINVAL; |
a67d7c5f DC |
402 | } |
403 | ||
7d095257 CH |
404 | #ifndef CONFIG_XFS_QUOTA |
405 | if (XFS_IS_QUOTA_RUNNING(mp)) { | |
4f10700a | 406 | xfs_warn(mp, "quota support not available in this kernel."); |
2451337d | 407 | return -EINVAL; |
7d095257 CH |
408 | } |
409 | #endif | |
410 | ||
a67d7c5f | 411 | if ((dsunit && !dswidth) || (!dsunit && dswidth)) { |
4f10700a | 412 | xfs_warn(mp, "sunit and swidth must be specified together"); |
2451337d | 413 | return -EINVAL; |
a67d7c5f DC |
414 | } |
415 | ||
416 | if (dsunit && (dswidth % dsunit != 0)) { | |
4f10700a DC |
417 | xfs_warn(mp, |
418 | "stripe width (%d) must be a multiple of the stripe unit (%d)", | |
a67d7c5f | 419 | dswidth, dsunit); |
2451337d | 420 | return -EINVAL; |
a67d7c5f DC |
421 | } |
422 | ||
9d565ffa | 423 | done: |
39a45d84 | 424 | if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) { |
9d565ffa CH |
425 | /* |
426 | * At this point the superblock has not been read | |
427 | * in, therefore we do not know the block size. | |
428 | * Before the mount call ends we will convert | |
429 | * these to FSBs. | |
430 | */ | |
39a45d84 JL |
431 | mp->m_dalign = dsunit; |
432 | mp->m_swidth = dswidth; | |
9d565ffa CH |
433 | } |
434 | ||
435 | if (mp->m_logbufs != -1 && | |
436 | mp->m_logbufs != 0 && | |
437 | (mp->m_logbufs < XLOG_MIN_ICLOGS || | |
438 | mp->m_logbufs > XLOG_MAX_ICLOGS)) { | |
4f10700a | 439 | xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]", |
9d565ffa | 440 | mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS); |
2451337d | 441 | return -EINVAL; |
9d565ffa CH |
442 | } |
443 | if (mp->m_logbsize != -1 && | |
444 | mp->m_logbsize != 0 && | |
445 | (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE || | |
446 | mp->m_logbsize > XLOG_MAX_RECORD_BSIZE || | |
447 | !is_power_of_2(mp->m_logbsize))) { | |
4f10700a DC |
448 | xfs_warn(mp, |
449 | "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]", | |
9d565ffa | 450 | mp->m_logbsize); |
2451337d | 451 | return -EINVAL; |
9d565ffa CH |
452 | } |
453 | ||
9d565ffa CH |
454 | if (iosizelog) { |
455 | if (iosizelog > XFS_MAX_IO_LOG || | |
456 | iosizelog < XFS_MIN_IO_LOG) { | |
4f10700a | 457 | xfs_warn(mp, "invalid log iosize: %d [not %d-%d]", |
9d565ffa CH |
458 | iosizelog, XFS_MIN_IO_LOG, |
459 | XFS_MAX_IO_LOG); | |
2451337d | 460 | return -EINVAL; |
9d565ffa CH |
461 | } |
462 | ||
463 | mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE; | |
464 | mp->m_readio_log = iosizelog; | |
465 | mp->m_writeio_log = iosizelog; | |
a67d7c5f DC |
466 | } |
467 | ||
a67d7c5f DC |
468 | return 0; |
469 | } | |
470 | ||
471 | struct proc_xfs_info { | |
cbe4dab1 DC |
472 | uint64_t flag; |
473 | char *str; | |
a67d7c5f DC |
474 | }; |
475 | ||
476 | STATIC int | |
477 | xfs_showargs( | |
478 | struct xfs_mount *mp, | |
479 | struct seq_file *m) | |
480 | { | |
481 | static struct proc_xfs_info xfs_info_set[] = { | |
482 | /* the few simple ones we can get from the mount struct */ | |
2e74af0e ES |
483 | { XFS_MOUNT_IKEEP, ",ikeep" }, |
484 | { XFS_MOUNT_WSYNC, ",wsync" }, | |
485 | { XFS_MOUNT_NOALIGN, ",noalign" }, | |
486 | { XFS_MOUNT_SWALLOC, ",swalloc" }, | |
487 | { XFS_MOUNT_NOUUID, ",nouuid" }, | |
488 | { XFS_MOUNT_NORECOVERY, ",norecovery" }, | |
489 | { XFS_MOUNT_ATTR2, ",attr2" }, | |
490 | { XFS_MOUNT_FILESTREAMS, ",filestreams" }, | |
491 | { XFS_MOUNT_GRPID, ",grpid" }, | |
492 | { XFS_MOUNT_DISCARD, ",discard" }, | |
493 | { XFS_MOUNT_SMALL_INUMS, ",inode32" }, | |
494 | { XFS_MOUNT_DAX, ",dax" }, | |
a67d7c5f DC |
495 | { 0, NULL } |
496 | }; | |
497 | static struct proc_xfs_info xfs_info_unset[] = { | |
498 | /* the few simple ones we can get from the mount struct */ | |
2e74af0e ES |
499 | { XFS_MOUNT_COMPAT_IOSIZE, ",largeio" }, |
500 | { XFS_MOUNT_BARRIER, ",nobarrier" }, | |
501 | { XFS_MOUNT_SMALL_INUMS, ",inode64" }, | |
a67d7c5f DC |
502 | { 0, NULL } |
503 | }; | |
504 | struct proc_xfs_info *xfs_infop; | |
505 | ||
506 | for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) { | |
507 | if (mp->m_flags & xfs_infop->flag) | |
508 | seq_puts(m, xfs_infop->str); | |
509 | } | |
510 | for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) { | |
511 | if (!(mp->m_flags & xfs_infop->flag)) | |
512 | seq_puts(m, xfs_infop->str); | |
513 | } | |
514 | ||
515 | if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) | |
2e74af0e | 516 | seq_printf(m, ",allocsize=%dk", |
a67d7c5f DC |
517 | (int)(1 << mp->m_writeio_log) >> 10); |
518 | ||
519 | if (mp->m_logbufs > 0) | |
2e74af0e | 520 | seq_printf(m, ",logbufs=%d", mp->m_logbufs); |
a67d7c5f | 521 | if (mp->m_logbsize > 0) |
2e74af0e | 522 | seq_printf(m, ",logbsize=%dk", mp->m_logbsize >> 10); |
a67d7c5f DC |
523 | |
524 | if (mp->m_logname) | |
2e74af0e | 525 | seq_show_option(m, "logdev", mp->m_logname); |
a67d7c5f | 526 | if (mp->m_rtname) |
2e74af0e | 527 | seq_show_option(m, "rtdev", mp->m_rtname); |
a67d7c5f DC |
528 | |
529 | if (mp->m_dalign > 0) | |
2e74af0e | 530 | seq_printf(m, ",sunit=%d", |
a67d7c5f DC |
531 | (int)XFS_FSB_TO_BB(mp, mp->m_dalign)); |
532 | if (mp->m_swidth > 0) | |
2e74af0e | 533 | seq_printf(m, ",swidth=%d", |
a67d7c5f DC |
534 | (int)XFS_FSB_TO_BB(mp, mp->m_swidth)); |
535 | ||
536 | if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD)) | |
2e74af0e | 537 | seq_puts(m, ",usrquota"); |
a67d7c5f | 538 | else if (mp->m_qflags & XFS_UQUOTA_ACCT) |
2e74af0e | 539 | seq_puts(m, ",uqnoenforce"); |
a67d7c5f | 540 | |
988abe40 | 541 | if (mp->m_qflags & XFS_PQUOTA_ACCT) { |
83e782e1 | 542 | if (mp->m_qflags & XFS_PQUOTA_ENFD) |
2e74af0e | 543 | seq_puts(m, ",prjquota"); |
988abe40 | 544 | else |
2e74af0e | 545 | seq_puts(m, ",pqnoenforce"); |
d892d586 CS |
546 | } |
547 | if (mp->m_qflags & XFS_GQUOTA_ACCT) { | |
83e782e1 | 548 | if (mp->m_qflags & XFS_GQUOTA_ENFD) |
2e74af0e | 549 | seq_puts(m, ",grpquota"); |
988abe40 | 550 | else |
2e74af0e | 551 | seq_puts(m, ",gqnoenforce"); |
988abe40 | 552 | } |
a67d7c5f DC |
553 | |
554 | if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT)) | |
2e74af0e | 555 | seq_puts(m, ",noquota"); |
a67d7c5f DC |
556 | |
557 | return 0; | |
558 | } | |
c8ce540d | 559 | static uint64_t |
1da177e4 LT |
560 | xfs_max_file_offset( |
561 | unsigned int blockshift) | |
562 | { | |
563 | unsigned int pagefactor = 1; | |
564 | unsigned int bitshift = BITS_PER_LONG - 1; | |
565 | ||
566 | /* Figure out maximum filesize, on Linux this can depend on | |
567 | * the filesystem blocksize (on 32 bit platforms). | |
ebdec241 | 568 | * __block_write_begin does this in an [unsigned] long... |
ea1754a0 | 569 | * page->index << (PAGE_SHIFT - bbits) |
1da177e4 LT |
570 | * So, for page sized blocks (4K on 32 bit platforms), |
571 | * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is | |
ea1754a0 | 572 | * (((u64)PAGE_SIZE << (BITS_PER_LONG-1))-1) |
1da177e4 LT |
573 | * but for smaller blocksizes it is less (bbits = log2 bsize). |
574 | * Note1: get_block_t takes a long (implicit cast from above) | |
575 | * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch | |
576 | * can optionally convert the [unsigned] long from above into | |
577 | * an [unsigned] long long. | |
578 | */ | |
579 | ||
580 | #if BITS_PER_LONG == 32 | |
90c699a9 | 581 | # if defined(CONFIG_LBDAF) |
1da177e4 | 582 | ASSERT(sizeof(sector_t) == 8); |
09cbfeaf | 583 | pagefactor = PAGE_SIZE; |
1da177e4 LT |
584 | bitshift = BITS_PER_LONG; |
585 | # else | |
09cbfeaf | 586 | pagefactor = PAGE_SIZE >> (PAGE_SHIFT - blockshift); |
1da177e4 LT |
587 | # endif |
588 | #endif | |
589 | ||
c8ce540d | 590 | return (((uint64_t)pagefactor) << bitshift) - 1; |
1da177e4 LT |
591 | } |
592 | ||
9de67c3b | 593 | /* |
12c3f05c ES |
594 | * Set parameters for inode allocation heuristics, taking into account |
595 | * filesystem size and inode32/inode64 mount options; i.e. specifically | |
596 | * whether or not XFS_MOUNT_SMALL_INUMS is set. | |
597 | * | |
598 | * Inode allocation patterns are altered only if inode32 is requested | |
599 | * (XFS_MOUNT_SMALL_INUMS), and the filesystem is sufficiently large. | |
600 | * If altered, XFS_MOUNT_32BITINODES is set as well. | |
601 | * | |
602 | * An agcount independent of that in the mount structure is provided | |
603 | * because in the growfs case, mp->m_sb.sb_agcount is not yet updated | |
604 | * to the potentially higher ag count. | |
605 | * | |
606 | * Returns the maximum AG index which may contain inodes. | |
9de67c3b | 607 | */ |
2d2194f6 | 608 | xfs_agnumber_t |
12c3f05c ES |
609 | xfs_set_inode_alloc( |
610 | struct xfs_mount *mp, | |
611 | xfs_agnumber_t agcount) | |
2d2194f6 | 612 | { |
12c3f05c | 613 | xfs_agnumber_t index; |
4056c1d0 | 614 | xfs_agnumber_t maxagi = 0; |
2d2194f6 CM |
615 | xfs_sb_t *sbp = &mp->m_sb; |
616 | xfs_agnumber_t max_metadata; | |
54aa61f8 ES |
617 | xfs_agino_t agino; |
618 | xfs_ino_t ino; | |
2d2194f6 | 619 | |
12c3f05c ES |
620 | /* |
621 | * Calculate how much should be reserved for inodes to meet | |
622 | * the max inode percentage. Used only for inode32. | |
2d2194f6 CM |
623 | */ |
624 | if (mp->m_maxicount) { | |
c8ce540d | 625 | uint64_t icount; |
2d2194f6 CM |
626 | |
627 | icount = sbp->sb_dblocks * sbp->sb_imax_pct; | |
628 | do_div(icount, 100); | |
629 | icount += sbp->sb_agblocks - 1; | |
630 | do_div(icount, sbp->sb_agblocks); | |
631 | max_metadata = icount; | |
632 | } else { | |
9de67c3b | 633 | max_metadata = agcount; |
2d2194f6 CM |
634 | } |
635 | ||
12c3f05c | 636 | /* Get the last possible inode in the filesystem */ |
54aa61f8 | 637 | agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks - 1, 0); |
12c3f05c ES |
638 | ino = XFS_AGINO_TO_INO(mp, agcount - 1, agino); |
639 | ||
640 | /* | |
641 | * If user asked for no more than 32-bit inodes, and the fs is | |
642 | * sufficiently large, set XFS_MOUNT_32BITINODES if we must alter | |
643 | * the allocator to accommodate the request. | |
644 | */ | |
645 | if ((mp->m_flags & XFS_MOUNT_SMALL_INUMS) && ino > XFS_MAXINUMBER_32) | |
646 | mp->m_flags |= XFS_MOUNT_32BITINODES; | |
647 | else | |
648 | mp->m_flags &= ~XFS_MOUNT_32BITINODES; | |
54aa61f8 | 649 | |
9de67c3b | 650 | for (index = 0; index < agcount; index++) { |
12c3f05c | 651 | struct xfs_perag *pag; |
4056c1d0 | 652 | |
12c3f05c | 653 | ino = XFS_AGINO_TO_INO(mp, index, agino); |
2d2194f6 CM |
654 | |
655 | pag = xfs_perag_get(mp, index); | |
2d2194f6 | 656 | |
12c3f05c ES |
657 | if (mp->m_flags & XFS_MOUNT_32BITINODES) { |
658 | if (ino > XFS_MAXINUMBER_32) { | |
659 | pag->pagi_inodeok = 0; | |
660 | pag->pagf_metadata = 0; | |
661 | } else { | |
662 | pag->pagi_inodeok = 1; | |
663 | maxagi++; | |
664 | if (index < max_metadata) | |
665 | pag->pagf_metadata = 1; | |
666 | else | |
667 | pag->pagf_metadata = 0; | |
668 | } | |
669 | } else { | |
670 | pag->pagi_inodeok = 1; | |
671 | pag->pagf_metadata = 0; | |
672 | } | |
2d2194f6 | 673 | |
2d2194f6 CM |
674 | xfs_perag_put(pag); |
675 | } | |
676 | ||
12c3f05c | 677 | return (mp->m_flags & XFS_MOUNT_32BITINODES) ? maxagi : agcount; |
2d2194f6 CM |
678 | } |
679 | ||
3180e66d | 680 | STATIC int |
1da177e4 LT |
681 | xfs_blkdev_get( |
682 | xfs_mount_t *mp, | |
683 | const char *name, | |
684 | struct block_device **bdevp) | |
685 | { | |
686 | int error = 0; | |
687 | ||
d4d77629 TH |
688 | *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL, |
689 | mp); | |
1da177e4 LT |
690 | if (IS_ERR(*bdevp)) { |
691 | error = PTR_ERR(*bdevp); | |
77af574e | 692 | xfs_warn(mp, "Invalid device [%s], error=%d", name, error); |
1da177e4 LT |
693 | } |
694 | ||
2451337d | 695 | return error; |
1da177e4 LT |
696 | } |
697 | ||
3180e66d | 698 | STATIC void |
1da177e4 LT |
699 | xfs_blkdev_put( |
700 | struct block_device *bdev) | |
701 | { | |
702 | if (bdev) | |
e525fd89 | 703 | blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); |
1da177e4 LT |
704 | } |
705 | ||
f538d4da CH |
706 | void |
707 | xfs_blkdev_issue_flush( | |
708 | xfs_buftarg_t *buftarg) | |
709 | { | |
7582df51 | 710 | blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL); |
f538d4da | 711 | } |
1da177e4 | 712 | |
19f354d4 CH |
713 | STATIC void |
714 | xfs_close_devices( | |
715 | struct xfs_mount *mp) | |
716 | { | |
717 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { | |
c032bfcf | 718 | struct block_device *logdev = mp->m_logdev_targp->bt_bdev; |
b7963133 | 719 | xfs_free_buftarg(mp, mp->m_logdev_targp); |
c032bfcf | 720 | xfs_blkdev_put(logdev); |
19f354d4 CH |
721 | } |
722 | if (mp->m_rtdev_targp) { | |
c032bfcf | 723 | struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev; |
b7963133 | 724 | xfs_free_buftarg(mp, mp->m_rtdev_targp); |
c032bfcf | 725 | xfs_blkdev_put(rtdev); |
19f354d4 | 726 | } |
b7963133 | 727 | xfs_free_buftarg(mp, mp->m_ddev_targp); |
19f354d4 CH |
728 | } |
729 | ||
730 | /* | |
731 | * The file system configurations are: | |
732 | * (1) device (partition) with data and internal log | |
733 | * (2) logical volume with data and log subvolumes. | |
734 | * (3) logical volume with data, log, and realtime subvolumes. | |
735 | * | |
736 | * We only have to handle opening the log and realtime volumes here if | |
737 | * they are present. The data subvolume has already been opened by | |
738 | * get_sb_bdev() and is stored in sb->s_bdev. | |
739 | */ | |
740 | STATIC int | |
741 | xfs_open_devices( | |
9d565ffa | 742 | struct xfs_mount *mp) |
19f354d4 CH |
743 | { |
744 | struct block_device *ddev = mp->m_super->s_bdev; | |
745 | struct block_device *logdev = NULL, *rtdev = NULL; | |
746 | int error; | |
747 | ||
748 | /* | |
749 | * Open real time and log devices - order is important. | |
750 | */ | |
9d565ffa CH |
751 | if (mp->m_logname) { |
752 | error = xfs_blkdev_get(mp, mp->m_logname, &logdev); | |
19f354d4 CH |
753 | if (error) |
754 | goto out; | |
755 | } | |
756 | ||
9d565ffa CH |
757 | if (mp->m_rtname) { |
758 | error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev); | |
19f354d4 CH |
759 | if (error) |
760 | goto out_close_logdev; | |
761 | ||
762 | if (rtdev == ddev || rtdev == logdev) { | |
4f10700a DC |
763 | xfs_warn(mp, |
764 | "Cannot mount filesystem with identical rtdev and ddev/logdev."); | |
2451337d | 765 | error = -EINVAL; |
19f354d4 CH |
766 | goto out_close_rtdev; |
767 | } | |
768 | } | |
769 | ||
770 | /* | |
771 | * Setup xfs_mount buffer target pointers | |
772 | */ | |
2451337d | 773 | error = -ENOMEM; |
34dcefd7 | 774 | mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev); |
19f354d4 CH |
775 | if (!mp->m_ddev_targp) |
776 | goto out_close_rtdev; | |
777 | ||
778 | if (rtdev) { | |
34dcefd7 | 779 | mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev); |
19f354d4 CH |
780 | if (!mp->m_rtdev_targp) |
781 | goto out_free_ddev_targ; | |
782 | } | |
783 | ||
784 | if (logdev && logdev != ddev) { | |
34dcefd7 | 785 | mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev); |
19f354d4 CH |
786 | if (!mp->m_logdev_targp) |
787 | goto out_free_rtdev_targ; | |
788 | } else { | |
789 | mp->m_logdev_targp = mp->m_ddev_targp; | |
790 | } | |
791 | ||
792 | return 0; | |
793 | ||
794 | out_free_rtdev_targ: | |
795 | if (mp->m_rtdev_targp) | |
b7963133 | 796 | xfs_free_buftarg(mp, mp->m_rtdev_targp); |
19f354d4 | 797 | out_free_ddev_targ: |
b7963133 | 798 | xfs_free_buftarg(mp, mp->m_ddev_targp); |
19f354d4 | 799 | out_close_rtdev: |
d2a5e3c6 | 800 | xfs_blkdev_put(rtdev); |
19f354d4 CH |
801 | out_close_logdev: |
802 | if (logdev && logdev != ddev) | |
803 | xfs_blkdev_put(logdev); | |
804 | out: | |
805 | return error; | |
806 | } | |
807 | ||
e34b562c CH |
808 | /* |
809 | * Setup xfs_mount buffer target pointers based on superblock | |
810 | */ | |
811 | STATIC int | |
812 | xfs_setup_devices( | |
813 | struct xfs_mount *mp) | |
814 | { | |
815 | int error; | |
19f354d4 | 816 | |
a96c4151 | 817 | error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize); |
e34b562c CH |
818 | if (error) |
819 | return error; | |
820 | ||
821 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { | |
822 | unsigned int log_sector_size = BBSIZE; | |
823 | ||
824 | if (xfs_sb_version_hassector(&mp->m_sb)) | |
825 | log_sector_size = mp->m_sb.sb_logsectsize; | |
826 | error = xfs_setsize_buftarg(mp->m_logdev_targp, | |
e34b562c CH |
827 | log_sector_size); |
828 | if (error) | |
829 | return error; | |
830 | } | |
831 | if (mp->m_rtdev_targp) { | |
832 | error = xfs_setsize_buftarg(mp->m_rtdev_targp, | |
e34b562c CH |
833 | mp->m_sb.sb_sectsize); |
834 | if (error) | |
835 | return error; | |
836 | } | |
837 | ||
838 | return 0; | |
839 | } | |
19f354d4 | 840 | |
aa6bf01d CH |
841 | STATIC int |
842 | xfs_init_mount_workqueues( | |
843 | struct xfs_mount *mp) | |
844 | { | |
78c931b8 | 845 | mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s", |
b29c70f5 | 846 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 1, mp->m_fsname); |
78c931b8 BF |
847 | if (!mp->m_buf_workqueue) |
848 | goto out; | |
849 | ||
aa6bf01d | 850 | mp->m_data_workqueue = alloc_workqueue("xfs-data/%s", |
8018ec08 | 851 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname); |
aa6bf01d | 852 | if (!mp->m_data_workqueue) |
78c931b8 | 853 | goto out_destroy_buf; |
aa6bf01d CH |
854 | |
855 | mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s", | |
8018ec08 | 856 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname); |
aa6bf01d CH |
857 | if (!mp->m_unwritten_workqueue) |
858 | goto out_destroy_data_iodone_queue; | |
859 | ||
4c2d542f | 860 | mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s", |
8018ec08 | 861 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname); |
4c2d542f DC |
862 | if (!mp->m_cil_workqueue) |
863 | goto out_destroy_unwritten; | |
5889608d DC |
864 | |
865 | mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s", | |
7a29ac47 | 866 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname); |
5889608d DC |
867 | if (!mp->m_reclaim_workqueue) |
868 | goto out_destroy_cil; | |
869 | ||
870 | mp->m_log_workqueue = alloc_workqueue("xfs-log/%s", | |
7a29ac47 CM |
871 | WQ_MEM_RECLAIM|WQ_FREEZABLE|WQ_HIGHPRI, 0, |
872 | mp->m_fsname); | |
5889608d DC |
873 | if (!mp->m_log_workqueue) |
874 | goto out_destroy_reclaim; | |
875 | ||
579b62fa | 876 | mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s", |
7a29ac47 | 877 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname); |
579b62fa BF |
878 | if (!mp->m_eofblocks_workqueue) |
879 | goto out_destroy_log; | |
880 | ||
696a5620 BF |
881 | mp->m_sync_workqueue = alloc_workqueue("xfs-sync/%s", WQ_FREEZABLE, 0, |
882 | mp->m_fsname); | |
883 | if (!mp->m_sync_workqueue) | |
884 | goto out_destroy_eofb; | |
885 | ||
aa6bf01d CH |
886 | return 0; |
887 | ||
696a5620 BF |
888 | out_destroy_eofb: |
889 | destroy_workqueue(mp->m_eofblocks_workqueue); | |
579b62fa BF |
890 | out_destroy_log: |
891 | destroy_workqueue(mp->m_log_workqueue); | |
5889608d DC |
892 | out_destroy_reclaim: |
893 | destroy_workqueue(mp->m_reclaim_workqueue); | |
894 | out_destroy_cil: | |
895 | destroy_workqueue(mp->m_cil_workqueue); | |
4c2d542f DC |
896 | out_destroy_unwritten: |
897 | destroy_workqueue(mp->m_unwritten_workqueue); | |
aa6bf01d CH |
898 | out_destroy_data_iodone_queue: |
899 | destroy_workqueue(mp->m_data_workqueue); | |
78c931b8 BF |
900 | out_destroy_buf: |
901 | destroy_workqueue(mp->m_buf_workqueue); | |
aa6bf01d CH |
902 | out: |
903 | return -ENOMEM; | |
904 | } | |
905 | ||
906 | STATIC void | |
907 | xfs_destroy_mount_workqueues( | |
908 | struct xfs_mount *mp) | |
909 | { | |
696a5620 | 910 | destroy_workqueue(mp->m_sync_workqueue); |
579b62fa | 911 | destroy_workqueue(mp->m_eofblocks_workqueue); |
5889608d DC |
912 | destroy_workqueue(mp->m_log_workqueue); |
913 | destroy_workqueue(mp->m_reclaim_workqueue); | |
4c2d542f | 914 | destroy_workqueue(mp->m_cil_workqueue); |
aa6bf01d CH |
915 | destroy_workqueue(mp->m_data_workqueue); |
916 | destroy_workqueue(mp->m_unwritten_workqueue); | |
78c931b8 | 917 | destroy_workqueue(mp->m_buf_workqueue); |
aa6bf01d CH |
918 | } |
919 | ||
9aa05000 DC |
920 | /* |
921 | * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK | |
922 | * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting | |
923 | * for IO to complete so that we effectively throttle multiple callers to the | |
924 | * rate at which IO is completing. | |
925 | */ | |
926 | void | |
927 | xfs_flush_inodes( | |
928 | struct xfs_mount *mp) | |
929 | { | |
930 | struct super_block *sb = mp->m_super; | |
931 | ||
932 | if (down_read_trylock(&sb->s_umount)) { | |
0dc83bd3 | 933 | sync_inodes_sb(sb); |
9aa05000 DC |
934 | up_read(&sb->s_umount); |
935 | } | |
936 | } | |
937 | ||
bf904248 | 938 | /* Catch misguided souls that try to use this interface on XFS */ |
1da177e4 | 939 | STATIC struct inode * |
a50cd269 | 940 | xfs_fs_alloc_inode( |
1da177e4 LT |
941 | struct super_block *sb) |
942 | { | |
bf904248 | 943 | BUG(); |
493dca61 | 944 | return NULL; |
1da177e4 LT |
945 | } |
946 | ||
bf904248 | 947 | /* |
99fa8cb3 | 948 | * Now that the generic code is guaranteed not to be accessing |
8179c036 | 949 | * the linux inode, we can inactivate and reclaim the inode. |
bf904248 | 950 | */ |
1da177e4 | 951 | STATIC void |
a50cd269 | 952 | xfs_fs_destroy_inode( |
848ce8f7 | 953 | struct inode *inode) |
1da177e4 | 954 | { |
848ce8f7 | 955 | struct xfs_inode *ip = XFS_I(inode); |
5e7e605c | 956 | int error; |
848ce8f7 | 957 | |
cca28fb8 | 958 | trace_xfs_destroy_inode(ip); |
99fa8cb3 | 959 | |
65523218 | 960 | ASSERT(!rwsem_is_locked(&inode->i_rwsem)); |
8179c036 DC |
961 | XFS_STATS_INC(ip->i_mount, vn_rele); |
962 | XFS_STATS_INC(ip->i_mount, vn_remove); | |
963 | ||
63646fc5 | 964 | if (xfs_is_reflink_inode(ip)) { |
3802a345 | 965 | error = xfs_reflink_cancel_cow_range(ip, 0, NULLFILEOFF, true); |
63646fc5 DW |
966 | if (error && !XFS_FORCED_SHUTDOWN(ip->i_mount)) |
967 | xfs_warn(ip->i_mount, | |
968 | "Error %d while evicting CoW blocks for inode %llu.", | |
969 | error, ip->i_ino); | |
970 | } | |
5e7e605c | 971 | |
8179c036 | 972 | xfs_inactive(ip); |
848ce8f7 | 973 | |
848ce8f7 | 974 | ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0); |
8179c036 | 975 | XFS_STATS_INC(ip->i_mount, vn_reclaim); |
848ce8f7 CH |
976 | |
977 | /* | |
978 | * We should never get here with one of the reclaim flags already set. | |
979 | */ | |
980 | ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE)); | |
981 | ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM)); | |
982 | ||
983 | /* | |
57817c68 DC |
984 | * We always use background reclaim here because even if the |
985 | * inode is clean, it still may be under IO and hence we have | |
986 | * to take the flush lock. The background reclaim path handles | |
987 | * this more efficiently than we can here, so simply let background | |
988 | * reclaim tear down all inodes. | |
848ce8f7 | 989 | */ |
57817c68 | 990 | xfs_inode_set_reclaim_tag(ip); |
1da177e4 LT |
991 | } |
992 | ||
07c8f675 DC |
993 | /* |
994 | * Slab object creation initialisation for the XFS inode. | |
995 | * This covers only the idempotent fields in the XFS inode; | |
996 | * all other fields need to be initialised on allocation | |
b595076a | 997 | * from the slab. This avoids the need to repeatedly initialise |
07c8f675 DC |
998 | * fields in the xfs inode that left in the initialise state |
999 | * when freeing the inode. | |
1000 | */ | |
bf904248 DC |
1001 | STATIC void |
1002 | xfs_fs_inode_init_once( | |
07c8f675 DC |
1003 | void *inode) |
1004 | { | |
1005 | struct xfs_inode *ip = inode; | |
1006 | ||
1007 | memset(ip, 0, sizeof(struct xfs_inode)); | |
bf904248 DC |
1008 | |
1009 | /* vfs inode */ | |
1010 | inode_init_once(VFS_I(ip)); | |
1011 | ||
1012 | /* xfs inode */ | |
07c8f675 DC |
1013 | atomic_set(&ip->i_pincount, 0); |
1014 | spin_lock_init(&ip->i_flags_lock); | |
07c8f675 | 1015 | |
653c60b6 DC |
1016 | mrlock_init(&ip->i_mmaplock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, |
1017 | "xfsino", ip->i_ino); | |
07c8f675 DC |
1018 | mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, |
1019 | "xfsino", ip->i_ino); | |
07c8f675 DC |
1020 | } |
1021 | ||
5132ba8f DC |
1022 | /* |
1023 | * We do an unlocked check for XFS_IDONTCACHE here because we are already | |
1024 | * serialised against cache hits here via the inode->i_lock and igrab() in | |
1025 | * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be | |
1026 | * racing with us, and it avoids needing to grab a spinlock here for every inode | |
1027 | * we drop the final reference on. | |
1028 | */ | |
1029 | STATIC int | |
1030 | xfs_fs_drop_inode( | |
1031 | struct inode *inode) | |
1032 | { | |
1033 | struct xfs_inode *ip = XFS_I(inode); | |
1034 | ||
17c12bcd DW |
1035 | /* |
1036 | * If this unlinked inode is in the middle of recovery, don't | |
1037 | * drop the inode just yet; log recovery will take care of | |
1038 | * that. See the comment for this inode flag. | |
1039 | */ | |
1040 | if (ip->i_flags & XFS_IRECOVERY) { | |
1041 | ASSERT(ip->i_mount->m_log->l_flags & XLOG_RECOVERY_NEEDED); | |
1042 | return 0; | |
1043 | } | |
1044 | ||
5132ba8f DC |
1045 | return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE); |
1046 | } | |
1047 | ||
a738159d CH |
1048 | STATIC void |
1049 | xfs_free_fsname( | |
1050 | struct xfs_mount *mp) | |
1051 | { | |
1052 | kfree(mp->m_fsname); | |
1053 | kfree(mp->m_rtname); | |
1054 | kfree(mp->m_logname); | |
1055 | } | |
1056 | ||
1da177e4 | 1057 | STATIC int |
69961a26 | 1058 | xfs_fs_sync_fs( |
1da177e4 LT |
1059 | struct super_block *sb, |
1060 | int wait) | |
1061 | { | |
745f6919 | 1062 | struct xfs_mount *mp = XFS_M(sb); |
1da177e4 | 1063 | |
e893bffd | 1064 | /* |
34625c66 | 1065 | * Doing anything during the async pass would be counterproductive. |
e893bffd | 1066 | */ |
34625c66 | 1067 | if (!wait) |
69961a26 | 1068 | return 0; |
69961a26 | 1069 | |
34061f5c | 1070 | xfs_log_force(mp, XFS_LOG_SYNC); |
69961a26 | 1071 | if (laptop_mode) { |
1da177e4 LT |
1072 | /* |
1073 | * The disk must be active because we're syncing. | |
f661f1e0 | 1074 | * We schedule log work now (now that the disk is |
1da177e4 LT |
1075 | * active) instead of later (when it might not be). |
1076 | */ | |
f661f1e0 | 1077 | flush_delayed_work(&mp->m_log->l_work); |
1da177e4 LT |
1078 | } |
1079 | ||
69961a26 | 1080 | return 0; |
1da177e4 LT |
1081 | } |
1082 | ||
1083 | STATIC int | |
a50cd269 | 1084 | xfs_fs_statfs( |
726c3342 | 1085 | struct dentry *dentry, |
1da177e4 LT |
1086 | struct kstatfs *statp) |
1087 | { | |
4ca488eb CH |
1088 | struct xfs_mount *mp = XFS_M(dentry->d_sb); |
1089 | xfs_sb_t *sbp = &mp->m_sb; | |
2b0143b5 | 1090 | struct xfs_inode *ip = XFS_I(d_inode(dentry)); |
c8ce540d DW |
1091 | uint64_t fakeinos, id; |
1092 | uint64_t icount; | |
1093 | uint64_t ifree; | |
1094 | uint64_t fdblocks; | |
4ca488eb | 1095 | xfs_extlen_t lsize; |
c8ce540d | 1096 | int64_t ffree; |
4ca488eb CH |
1097 | |
1098 | statp->f_type = XFS_SB_MAGIC; | |
1099 | statp->f_namelen = MAXNAMELEN - 1; | |
1100 | ||
1101 | id = huge_encode_dev(mp->m_ddev_targp->bt_dev); | |
1102 | statp->f_fsid.val[0] = (u32)id; | |
1103 | statp->f_fsid.val[1] = (u32)(id >> 32); | |
1104 | ||
501ab323 | 1105 | icount = percpu_counter_sum(&mp->m_icount); |
e88b64ea | 1106 | ifree = percpu_counter_sum(&mp->m_ifree); |
0d485ada | 1107 | fdblocks = percpu_counter_sum(&mp->m_fdblocks); |
4ca488eb CH |
1108 | |
1109 | spin_lock(&mp->m_sb_lock); | |
1110 | statp->f_bsize = sbp->sb_blocksize; | |
1111 | lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0; | |
1112 | statp->f_blocks = sbp->sb_dblocks - lsize; | |
0d485ada DC |
1113 | spin_unlock(&mp->m_sb_lock); |
1114 | ||
52548852 | 1115 | statp->f_bfree = fdblocks - mp->m_alloc_set_aside; |
0d485ada DC |
1116 | statp->f_bavail = statp->f_bfree; |
1117 | ||
4ca488eb | 1118 | fakeinos = statp->f_bfree << sbp->sb_inopblog; |
c8ce540d | 1119 | statp->f_files = MIN(icount + fakeinos, (uint64_t)XFS_MAXINUMBER); |
4ca488eb | 1120 | if (mp->m_maxicount) |
a19d9f88 CH |
1121 | statp->f_files = min_t(typeof(statp->f_files), |
1122 | statp->f_files, | |
1123 | mp->m_maxicount); | |
2fe33661 | 1124 | |
01f9882e ES |
1125 | /* If sb_icount overshot maxicount, report actual allocation */ |
1126 | statp->f_files = max_t(typeof(statp->f_files), | |
1127 | statp->f_files, | |
1128 | sbp->sb_icount); | |
1129 | ||
2fe33661 | 1130 | /* make sure statp->f_ffree does not underflow */ |
e88b64ea | 1131 | ffree = statp->f_files - (icount - ifree); |
c8ce540d | 1132 | statp->f_ffree = max_t(int64_t, ffree, 0); |
2fe33661 | 1133 | |
4ca488eb | 1134 | |
da5bf95e | 1135 | if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && |
83e782e1 CS |
1136 | ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) == |
1137 | (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD)) | |
7d095257 | 1138 | xfs_qm_statvfs(ip, statp); |
4ca488eb | 1139 | return 0; |
1da177e4 LT |
1140 | } |
1141 | ||
d5db0f97 ES |
1142 | STATIC void |
1143 | xfs_save_resvblks(struct xfs_mount *mp) | |
1144 | { | |
c8ce540d | 1145 | uint64_t resblks = 0; |
d5db0f97 ES |
1146 | |
1147 | mp->m_resblks_save = mp->m_resblks; | |
1148 | xfs_reserve_blocks(mp, &resblks, NULL); | |
1149 | } | |
1150 | ||
1151 | STATIC void | |
1152 | xfs_restore_resvblks(struct xfs_mount *mp) | |
1153 | { | |
c8ce540d | 1154 | uint64_t resblks; |
d5db0f97 ES |
1155 | |
1156 | if (mp->m_resblks_save) { | |
1157 | resblks = mp->m_resblks_save; | |
1158 | mp->m_resblks_save = 0; | |
1159 | } else | |
1160 | resblks = xfs_default_resblks(mp); | |
1161 | ||
1162 | xfs_reserve_blocks(mp, &resblks, NULL); | |
1163 | } | |
1164 | ||
c7eea6f7 DC |
1165 | /* |
1166 | * Trigger writeback of all the dirty metadata in the file system. | |
1167 | * | |
1168 | * This ensures that the metadata is written to their location on disk rather | |
1169 | * than just existing in transactions in the log. This means after a quiesce | |
c75921a7 DC |
1170 | * there is no log replay required to write the inodes to disk - this is the |
1171 | * primary difference between a sync and a quiesce. | |
c7eea6f7 | 1172 | * |
c75921a7 DC |
1173 | * Note: xfs_log_quiesce() stops background log work - the callers must ensure |
1174 | * it is started again when appropriate. | |
c7eea6f7 | 1175 | */ |
ddeb14f4 | 1176 | void |
c7eea6f7 DC |
1177 | xfs_quiesce_attr( |
1178 | struct xfs_mount *mp) | |
1179 | { | |
1180 | int error = 0; | |
1181 | ||
1182 | /* wait for all modifications to complete */ | |
1183 | while (atomic_read(&mp->m_active_trans) > 0) | |
1184 | delay(100); | |
1185 | ||
1186 | /* force the log to unpin objects from the now complete transactions */ | |
1187 | xfs_log_force(mp, XFS_LOG_SYNC); | |
1188 | ||
1189 | /* reclaim inodes to do any IO before the freeze completes */ | |
1190 | xfs_reclaim_inodes(mp, 0); | |
1191 | xfs_reclaim_inodes(mp, SYNC_WAIT); | |
1192 | ||
c7eea6f7 DC |
1193 | /* Push the superblock and write an unmount record */ |
1194 | error = xfs_log_sbcount(mp); | |
1195 | if (error) | |
1196 | xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. " | |
1197 | "Frozen image may not be consistent."); | |
c7eea6f7 | 1198 | /* |
c75921a7 DC |
1199 | * Just warn here till VFS can correctly support |
1200 | * read-only remount without racing. | |
c7eea6f7 | 1201 | */ |
c75921a7 | 1202 | WARN_ON(atomic_read(&mp->m_active_trans) != 0); |
c7eea6f7 | 1203 | |
c75921a7 | 1204 | xfs_log_quiesce(mp); |
c7eea6f7 DC |
1205 | } |
1206 | ||
a08ee40a ES |
1207 | STATIC int |
1208 | xfs_test_remount_options( | |
1209 | struct super_block *sb, | |
1210 | struct xfs_mount *mp, | |
1211 | char *options) | |
1212 | { | |
1213 | int error = 0; | |
1214 | struct xfs_mount *tmp_mp; | |
1215 | ||
1216 | tmp_mp = kmem_zalloc(sizeof(*tmp_mp), KM_MAYFAIL); | |
1217 | if (!tmp_mp) | |
1218 | return -ENOMEM; | |
1219 | ||
1220 | tmp_mp->m_super = sb; | |
1221 | error = xfs_parseargs(tmp_mp, options); | |
1222 | xfs_free_fsname(tmp_mp); | |
1223 | kfree(tmp_mp); | |
1224 | ||
1225 | return error; | |
1226 | } | |
1227 | ||
1da177e4 | 1228 | STATIC int |
a50cd269 | 1229 | xfs_fs_remount( |
1da177e4 LT |
1230 | struct super_block *sb, |
1231 | int *flags, | |
1232 | char *options) | |
1233 | { | |
745f6919 | 1234 | struct xfs_mount *mp = XFS_M(sb); |
9de67c3b | 1235 | xfs_sb_t *sbp = &mp->m_sb; |
62a877e3 CH |
1236 | substring_t args[MAX_OPT_ARGS]; |
1237 | char *p; | |
7884bc86 | 1238 | int error; |
1da177e4 | 1239 | |
a08ee40a ES |
1240 | /* First, check for complete junk; i.e. invalid options */ |
1241 | error = xfs_test_remount_options(sb, mp, options); | |
1242 | if (error) | |
1243 | return error; | |
1244 | ||
02b9984d | 1245 | sync_filesystem(sb); |
62a877e3 CH |
1246 | while ((p = strsep(&options, ",")) != NULL) { |
1247 | int token; | |
bdd907ba | 1248 | |
62a877e3 CH |
1249 | if (!*p) |
1250 | continue; | |
48b62a1a | 1251 | |
62a877e3 CH |
1252 | token = match_token(p, tokens, args); |
1253 | switch (token) { | |
1254 | case Opt_barrier: | |
4cf4573d | 1255 | xfs_warn(mp, "%s option is deprecated, ignoring.", p); |
48b62a1a | 1256 | mp->m_flags |= XFS_MOUNT_BARRIER; |
62a877e3 CH |
1257 | break; |
1258 | case Opt_nobarrier: | |
4cf4573d | 1259 | xfs_warn(mp, "%s option is deprecated, ignoring.", p); |
48b62a1a | 1260 | mp->m_flags &= ~XFS_MOUNT_BARRIER; |
62a877e3 | 1261 | break; |
c3a58fec | 1262 | case Opt_inode64: |
12c3f05c ES |
1263 | mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS; |
1264 | mp->m_maxagi = xfs_set_inode_alloc(mp, sbp->sb_agcount); | |
c3a58fec | 1265 | break; |
2ea03929 | 1266 | case Opt_inode32: |
12c3f05c ES |
1267 | mp->m_flags |= XFS_MOUNT_SMALL_INUMS; |
1268 | mp->m_maxagi = xfs_set_inode_alloc(mp, sbp->sb_agcount); | |
2ea03929 | 1269 | break; |
62a877e3 | 1270 | default: |
6efdf281 CH |
1271 | /* |
1272 | * Logically we would return an error here to prevent | |
1273 | * users from believing they might have changed | |
1274 | * mount options using remount which can't be changed. | |
1275 | * | |
1276 | * But unfortunately mount(8) adds all options from | |
1277 | * mtab and fstab to the mount arguments in some cases | |
1278 | * so we can't blindly reject options, but have to | |
1279 | * check for each specified option if it actually | |
1280 | * differs from the currently set option and only | |
1281 | * reject it if that's the case. | |
1282 | * | |
1283 | * Until that is implemented we return success for | |
1284 | * every remount request, and silently ignore all | |
1285 | * options that we can't actually change. | |
1286 | */ | |
1287 | #if 0 | |
4f10700a | 1288 | xfs_info(mp, |
08e96e1a | 1289 | "mount option \"%s\" not supported for remount", p); |
62a877e3 | 1290 | return -EINVAL; |
6efdf281 | 1291 | #else |
6c5e51da | 1292 | break; |
6efdf281 | 1293 | #endif |
48b62a1a | 1294 | } |
62a877e3 CH |
1295 | } |
1296 | ||
7884bc86 | 1297 | /* ro -> rw */ |
62a877e3 | 1298 | if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) { |
bbe051c8 ES |
1299 | if (mp->m_flags & XFS_MOUNT_NORECOVERY) { |
1300 | xfs_warn(mp, | |
1301 | "ro->rw transition prohibited on norecovery mount"); | |
1302 | return -EINVAL; | |
1303 | } | |
1304 | ||
d0a58e83 ES |
1305 | if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 && |
1306 | xfs_sb_has_ro_compat_feature(sbp, | |
1307 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) { | |
1308 | xfs_warn(mp, | |
1309 | "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem", | |
1310 | (sbp->sb_features_ro_compat & | |
1311 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)); | |
1312 | return -EINVAL; | |
1313 | } | |
1314 | ||
62a877e3 | 1315 | mp->m_flags &= ~XFS_MOUNT_RDONLY; |
7884bc86 CH |
1316 | |
1317 | /* | |
1318 | * If this is the first remount to writeable state we | |
1319 | * might have some superblock changes to update. | |
1320 | */ | |
61e63ecb DC |
1321 | if (mp->m_update_sb) { |
1322 | error = xfs_sync_sb(mp, false); | |
7884bc86 | 1323 | if (error) { |
4f10700a | 1324 | xfs_warn(mp, "failed to write sb changes"); |
7884bc86 CH |
1325 | return error; |
1326 | } | |
61e63ecb | 1327 | mp->m_update_sb = false; |
7884bc86 | 1328 | } |
cbe132a8 DC |
1329 | |
1330 | /* | |
1331 | * Fill out the reserve pool if it is empty. Use the stashed | |
1332 | * value if it is non-zero, otherwise go with the default. | |
1333 | */ | |
d5db0f97 | 1334 | xfs_restore_resvblks(mp); |
f661f1e0 | 1335 | xfs_log_work_queue(mp); |
fa5a4f57 | 1336 | xfs_queue_eofblocks(mp); |
174edb0e DW |
1337 | |
1338 | /* Recover any CoW blocks that never got remapped. */ | |
1339 | error = xfs_reflink_recover_cow(mp); | |
1340 | if (error) { | |
1341 | xfs_err(mp, | |
1342 | "Error %d recovering leftover CoW allocations.", error); | |
1343 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); | |
1344 | return error; | |
1345 | } | |
84d69619 DW |
1346 | |
1347 | /* Create the per-AG metadata reservation pool .*/ | |
1348 | error = xfs_fs_reserve_ag_blocks(mp); | |
1349 | if (error && error != -ENOSPC) | |
1350 | return error; | |
62a877e3 CH |
1351 | } |
1352 | ||
1353 | /* rw -> ro */ | |
1354 | if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) { | |
84d69619 DW |
1355 | /* Free the per-AG metadata reservation pool. */ |
1356 | error = xfs_fs_unreserve_ag_blocks(mp); | |
1357 | if (error) { | |
1358 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); | |
1359 | return error; | |
1360 | } | |
1361 | ||
cbe132a8 | 1362 | /* |
34061f5c DC |
1363 | * Before we sync the metadata, we need to free up the reserve |
1364 | * block pool so that the used block count in the superblock on | |
1365 | * disk is correct at the end of the remount. Stash the current | |
1366 | * reserve pool size so that if we get remounted rw, we can | |
1367 | * return it to the same size. | |
cbe132a8 | 1368 | */ |
d5db0f97 | 1369 | xfs_save_resvblks(mp); |
fa5a4f57 BF |
1370 | |
1371 | /* | |
1372 | * Cancel background eofb scanning so it cannot race with the | |
1373 | * final log force+buftarg wait and deadlock the remount. | |
1374 | */ | |
1375 | cancel_delayed_work_sync(&mp->m_eofblocks_work); | |
1376 | ||
76bf105c | 1377 | xfs_quiesce_attr(mp); |
48b62a1a CH |
1378 | mp->m_flags |= XFS_MOUNT_RDONLY; |
1379 | } | |
1380 | ||
62a877e3 | 1381 | return 0; |
1da177e4 LT |
1382 | } |
1383 | ||
9909c4aa CH |
1384 | /* |
1385 | * Second stage of a freeze. The data is already frozen so we only | |
61e63ecb DC |
1386 | * need to take care of the metadata. Once that's done sync the superblock |
1387 | * to the log to dirty it in case of a crash while frozen. This ensures that we | |
1388 | * will recover the unlinked inode lists on the next mount. | |
9909c4aa | 1389 | */ |
c4be0c1d TS |
1390 | STATIC int |
1391 | xfs_fs_freeze( | |
1da177e4 LT |
1392 | struct super_block *sb) |
1393 | { | |
9909c4aa CH |
1394 | struct xfs_mount *mp = XFS_M(sb); |
1395 | ||
d5db0f97 | 1396 | xfs_save_resvblks(mp); |
76bf105c | 1397 | xfs_quiesce_attr(mp); |
61e63ecb | 1398 | return xfs_sync_sb(mp, true); |
1da177e4 LT |
1399 | } |
1400 | ||
d5db0f97 ES |
1401 | STATIC int |
1402 | xfs_fs_unfreeze( | |
1403 | struct super_block *sb) | |
1404 | { | |
1405 | struct xfs_mount *mp = XFS_M(sb); | |
1406 | ||
1407 | xfs_restore_resvblks(mp); | |
f661f1e0 | 1408 | xfs_log_work_queue(mp); |
d5db0f97 ES |
1409 | return 0; |
1410 | } | |
1411 | ||
1da177e4 | 1412 | STATIC int |
a50cd269 | 1413 | xfs_fs_show_options( |
1da177e4 | 1414 | struct seq_file *m, |
34c80b1d | 1415 | struct dentry *root) |
1da177e4 | 1416 | { |
2451337d | 1417 | return xfs_showargs(XFS_M(root->d_sb), m); |
1da177e4 LT |
1418 | } |
1419 | ||
f8f15e42 CH |
1420 | /* |
1421 | * This function fills in xfs_mount_t fields based on mount args. | |
1422 | * Note: the superblock _has_ now been read in. | |
1423 | */ | |
1424 | STATIC int | |
1425 | xfs_finish_flags( | |
f8f15e42 CH |
1426 | struct xfs_mount *mp) |
1427 | { | |
1428 | int ronly = (mp->m_flags & XFS_MOUNT_RDONLY); | |
1429 | ||
025dfdaf | 1430 | /* Fail a mount where the logbuf is smaller than the log stripe */ |
f8f15e42 | 1431 | if (xfs_sb_version_haslogv2(&mp->m_sb)) { |
9d565ffa CH |
1432 | if (mp->m_logbsize <= 0 && |
1433 | mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) { | |
f8f15e42 | 1434 | mp->m_logbsize = mp->m_sb.sb_logsunit; |
9d565ffa CH |
1435 | } else if (mp->m_logbsize > 0 && |
1436 | mp->m_logbsize < mp->m_sb.sb_logsunit) { | |
4f10700a DC |
1437 | xfs_warn(mp, |
1438 | "logbuf size must be greater than or equal to log stripe size"); | |
2451337d | 1439 | return -EINVAL; |
f8f15e42 CH |
1440 | } |
1441 | } else { | |
1442 | /* Fail a mount if the logbuf is larger than 32K */ | |
9d565ffa | 1443 | if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) { |
4f10700a DC |
1444 | xfs_warn(mp, |
1445 | "logbuf size for version 1 logs must be 16K or 32K"); | |
2451337d | 1446 | return -EINVAL; |
f8f15e42 CH |
1447 | } |
1448 | } | |
1449 | ||
d3eaace8 DC |
1450 | /* |
1451 | * V5 filesystems always use attr2 format for attributes. | |
1452 | */ | |
1453 | if (xfs_sb_version_hascrc(&mp->m_sb) && | |
1454 | (mp->m_flags & XFS_MOUNT_NOATTR2)) { | |
2e74af0e ES |
1455 | xfs_warn(mp, "Cannot mount a V5 filesystem as noattr2. " |
1456 | "attr2 is always enabled for V5 filesystems."); | |
2451337d | 1457 | return -EINVAL; |
d3eaace8 DC |
1458 | } |
1459 | ||
f8f15e42 CH |
1460 | /* |
1461 | * mkfs'ed attr2 will turn on attr2 mount unless explicitly | |
1462 | * told by noattr2 to turn it off | |
1463 | */ | |
1464 | if (xfs_sb_version_hasattr2(&mp->m_sb) && | |
9d565ffa | 1465 | !(mp->m_flags & XFS_MOUNT_NOATTR2)) |
f8f15e42 CH |
1466 | mp->m_flags |= XFS_MOUNT_ATTR2; |
1467 | ||
1468 | /* | |
1469 | * prohibit r/w mounts of read-only filesystems | |
1470 | */ | |
1471 | if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) { | |
4f10700a DC |
1472 | xfs_warn(mp, |
1473 | "cannot mount a read-only filesystem as read-write"); | |
2451337d | 1474 | return -EROFS; |
f8f15e42 CH |
1475 | } |
1476 | ||
d892d586 CS |
1477 | if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) && |
1478 | (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) && | |
1479 | !xfs_sb_version_has_pquotino(&mp->m_sb)) { | |
1480 | xfs_warn(mp, | |
1481 | "Super block does not support project and group quota together"); | |
2451337d | 1482 | return -EINVAL; |
d892d586 CS |
1483 | } |
1484 | ||
f8f15e42 CH |
1485 | return 0; |
1486 | } | |
1487 | ||
5681ca40 DC |
1488 | static int |
1489 | xfs_init_percpu_counters( | |
1490 | struct xfs_mount *mp) | |
1491 | { | |
1492 | int error; | |
1493 | ||
1494 | error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL); | |
1495 | if (error) | |
5e9383f9 | 1496 | return -ENOMEM; |
5681ca40 DC |
1497 | |
1498 | error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL); | |
1499 | if (error) | |
1500 | goto free_icount; | |
1501 | ||
1502 | error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL); | |
1503 | if (error) | |
1504 | goto free_ifree; | |
1505 | ||
1506 | return 0; | |
1507 | ||
1508 | free_ifree: | |
1509 | percpu_counter_destroy(&mp->m_ifree); | |
1510 | free_icount: | |
1511 | percpu_counter_destroy(&mp->m_icount); | |
1512 | return -ENOMEM; | |
1513 | } | |
1514 | ||
1515 | void | |
1516 | xfs_reinit_percpu_counters( | |
1517 | struct xfs_mount *mp) | |
1518 | { | |
1519 | percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount); | |
1520 | percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree); | |
1521 | percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks); | |
1522 | } | |
1523 | ||
1524 | static void | |
1525 | xfs_destroy_percpu_counters( | |
1526 | struct xfs_mount *mp) | |
1527 | { | |
1528 | percpu_counter_destroy(&mp->m_icount); | |
1529 | percpu_counter_destroy(&mp->m_ifree); | |
1530 | percpu_counter_destroy(&mp->m_fdblocks); | |
1531 | } | |
1532 | ||
1da177e4 | 1533 | STATIC int |
a50cd269 | 1534 | xfs_fs_fill_super( |
1da177e4 LT |
1535 | struct super_block *sb, |
1536 | void *data, | |
1537 | int silent) | |
1538 | { | |
f3dcc13f | 1539 | struct inode *root; |
745f6919 | 1540 | struct xfs_mount *mp = NULL; |
2451337d | 1541 | int flags = 0, error = -ENOMEM; |
bdd907ba | 1542 | |
c962fb79 CH |
1543 | mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL); |
1544 | if (!mp) | |
9d565ffa | 1545 | goto out; |
1da177e4 | 1546 | |
c962fb79 | 1547 | spin_lock_init(&mp->m_sb_lock); |
c962fb79 CH |
1548 | mutex_init(&mp->m_growlock); |
1549 | atomic_set(&mp->m_active_trans, 0); | |
7e18530b | 1550 | INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker); |
579b62fa | 1551 | INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker); |
83104d44 | 1552 | INIT_DELAYED_WORK(&mp->m_cowblocks_work, xfs_cowblocks_worker); |
e3aed1a0 | 1553 | mp->m_kobj.kobject.kset = xfs_kset; |
74394496 | 1554 | |
b267ce99 CH |
1555 | mp->m_super = sb; |
1556 | sb->s_fs_info = mp; | |
1da177e4 | 1557 | |
288699fe | 1558 | error = xfs_parseargs(mp, (char *)data); |
745f6919 | 1559 | if (error) |
9d565ffa | 1560 | goto out_free_fsname; |
1da177e4 LT |
1561 | |
1562 | sb_min_blocksize(sb, BBSIZE); | |
0ec58516 | 1563 | sb->s_xattr = xfs_xattr_handlers; |
a50cd269 | 1564 | sb->s_export_op = &xfs_export_operations; |
fcafb71b | 1565 | #ifdef CONFIG_XFS_QUOTA |
a50cd269 | 1566 | sb->s_qcop = &xfs_quotactl_operations; |
17ef4fdd | 1567 | sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; |
fcafb71b | 1568 | #endif |
a50cd269 | 1569 | sb->s_op = &xfs_super_operations; |
1da177e4 | 1570 | |
9d565ffa | 1571 | if (silent) |
f8f15e42 CH |
1572 | flags |= XFS_MFSI_QUIET; |
1573 | ||
9d565ffa | 1574 | error = xfs_open_devices(mp); |
19f354d4 | 1575 | if (error) |
288699fe | 1576 | goto out_free_fsname; |
f8f15e42 | 1577 | |
2451337d | 1578 | error = xfs_init_mount_workqueues(mp); |
61ba35de CH |
1579 | if (error) |
1580 | goto out_close_devices; | |
c962fb79 | 1581 | |
5681ca40 | 1582 | error = xfs_init_percpu_counters(mp); |
aa6bf01d CH |
1583 | if (error) |
1584 | goto out_destroy_workqueues; | |
1585 | ||
225e4635 BD |
1586 | /* Allocate stats memory before we do operations that might use it */ |
1587 | mp->m_stats.xs_stats = alloc_percpu(struct xfsstats); | |
1588 | if (!mp->m_stats.xs_stats) { | |
f9d460b3 | 1589 | error = -ENOMEM; |
225e4635 BD |
1590 | goto out_destroy_counters; |
1591 | } | |
1592 | ||
f8f15e42 CH |
1593 | error = xfs_readsb(mp, flags); |
1594 | if (error) | |
225e4635 | 1595 | goto out_free_stats; |
9d565ffa CH |
1596 | |
1597 | error = xfs_finish_flags(mp); | |
f8f15e42 | 1598 | if (error) |
effa2eda | 1599 | goto out_free_sb; |
f8f15e42 | 1600 | |
e34b562c | 1601 | error = xfs_setup_devices(mp); |
19f354d4 | 1602 | if (error) |
effa2eda | 1603 | goto out_free_sb; |
f8f15e42 | 1604 | |
f8f15e42 CH |
1605 | error = xfs_filestream_mount(mp); |
1606 | if (error) | |
effa2eda | 1607 | goto out_free_sb; |
f8f15e42 | 1608 | |
704b2907 DC |
1609 | /* |
1610 | * we must configure the block size in the superblock before we run the | |
1611 | * full mount process as the mount process can lookup and cache inodes. | |
704b2907 | 1612 | */ |
4ca488eb CH |
1613 | sb->s_magic = XFS_SB_MAGIC; |
1614 | sb->s_blocksize = mp->m_sb.sb_blocksize; | |
1615 | sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1; | |
1da177e4 | 1616 | sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); |
8de52778 | 1617 | sb->s_max_links = XFS_MAXLINK; |
1da177e4 LT |
1618 | sb->s_time_gran = 1; |
1619 | set_posix_acl_flag(sb); | |
1620 | ||
dc037ad7 DC |
1621 | /* version 5 superblocks support inode version counters. */ |
1622 | if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5) | |
1623 | sb->s_flags |= MS_I_VERSION; | |
1624 | ||
cbe4dab1 DC |
1625 | if (mp->m_flags & XFS_MOUNT_DAX) { |
1626 | xfs_warn(mp, | |
1e937cdd TK |
1627 | "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); |
1628 | ||
1629 | error = bdev_dax_supported(sb, sb->s_blocksize); | |
1630 | if (error) { | |
cbe4dab1 | 1631 | xfs_alert(mp, |
1e937cdd | 1632 | "DAX unsupported by block device. Turning off DAX."); |
cbe4dab1 DC |
1633 | mp->m_flags &= ~XFS_MOUNT_DAX; |
1634 | } | |
e54b5bf9 DW |
1635 | if (xfs_sb_version_hasreflink(&mp->m_sb)) |
1636 | xfs_alert(mp, | |
1637 | "DAX and reflink have not been tested together!"); | |
cbe4dab1 DC |
1638 | } |
1639 | ||
738f57c1 DW |
1640 | if (xfs_sb_version_hasrmapbt(&mp->m_sb)) { |
1641 | if (mp->m_sb.sb_rblocks) { | |
1642 | xfs_alert(mp, | |
1643 | "EXPERIMENTAL reverse mapping btree not compatible with realtime device!"); | |
1644 | error = -EINVAL; | |
1645 | goto out_filestream_unmount; | |
1646 | } | |
1c0607ac DW |
1647 | xfs_alert(mp, |
1648 | "EXPERIMENTAL reverse mapping btree feature enabled. Use at your own risk!"); | |
738f57c1 | 1649 | } |
1c0607ac | 1650 | |
e54b5bf9 DW |
1651 | if (xfs_sb_version_hasreflink(&mp->m_sb)) |
1652 | xfs_alert(mp, | |
1653 | "EXPERIMENTAL reflink feature enabled. Use at your own risk!"); | |
1654 | ||
8a00ebe4 | 1655 | error = xfs_mountfs(mp); |
2bcf6e97 | 1656 | if (error) |
7e18530b | 1657 | goto out_filestream_unmount; |
704b2907 | 1658 | |
01651646 | 1659 | root = igrab(VFS_I(mp->m_rootip)); |
f3dcc13f | 1660 | if (!root) { |
2451337d | 1661 | error = -ENOENT; |
8a00ebe4 | 1662 | goto out_unmount; |
cbc89dcf | 1663 | } |
48fde701 | 1664 | sb->s_root = d_make_root(root); |
f3dcc13f | 1665 | if (!sb->s_root) { |
2451337d | 1666 | error = -ENOMEM; |
8a00ebe4 | 1667 | goto out_unmount; |
1da177e4 | 1668 | } |
74394496 | 1669 | |
1da177e4 | 1670 | return 0; |
33c7a2bc | 1671 | |
7e18530b | 1672 | out_filestream_unmount: |
120226c1 | 1673 | xfs_filestream_unmount(mp); |
effa2eda CH |
1674 | out_free_sb: |
1675 | xfs_freesb(mp); | |
225e4635 BD |
1676 | out_free_stats: |
1677 | free_percpu(mp->m_stats.xs_stats); | |
9d565ffa | 1678 | out_destroy_counters: |
5681ca40 | 1679 | xfs_destroy_percpu_counters(mp); |
225e4635 | 1680 | out_destroy_workqueues: |
aa6bf01d | 1681 | xfs_destroy_mount_workqueues(mp); |
61ba35de | 1682 | out_close_devices: |
19f354d4 | 1683 | xfs_close_devices(mp); |
9d565ffa CH |
1684 | out_free_fsname: |
1685 | xfs_free_fsname(mp); | |
c962fb79 | 1686 | kfree(mp); |
9d565ffa | 1687 | out: |
2451337d | 1688 | return error; |
f8f15e42 | 1689 | |
2bcf6e97 | 1690 | out_unmount: |
e48ad316 | 1691 | xfs_filestream_unmount(mp); |
19f354d4 | 1692 | xfs_unmountfs(mp); |
6203300e | 1693 | goto out_free_sb; |
1da177e4 LT |
1694 | } |
1695 | ||
5681ca40 DC |
1696 | STATIC void |
1697 | xfs_fs_put_super( | |
1698 | struct super_block *sb) | |
1699 | { | |
1700 | struct xfs_mount *mp = XFS_M(sb); | |
1701 | ||
4225441a | 1702 | xfs_notice(mp, "Unmounting Filesystem"); |
5681ca40 DC |
1703 | xfs_filestream_unmount(mp); |
1704 | xfs_unmountfs(mp); | |
1705 | ||
1706 | xfs_freesb(mp); | |
225e4635 | 1707 | free_percpu(mp->m_stats.xs_stats); |
5681ca40 DC |
1708 | xfs_destroy_percpu_counters(mp); |
1709 | xfs_destroy_mount_workqueues(mp); | |
1710 | xfs_close_devices(mp); | |
1711 | xfs_free_fsname(mp); | |
1712 | kfree(mp); | |
1713 | } | |
1714 | ||
152a0836 AV |
1715 | STATIC struct dentry * |
1716 | xfs_fs_mount( | |
1da177e4 LT |
1717 | struct file_system_type *fs_type, |
1718 | int flags, | |
1719 | const char *dev_name, | |
152a0836 | 1720 | void *data) |
1da177e4 | 1721 | { |
152a0836 | 1722 | return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super); |
a50cd269 NS |
1723 | } |
1724 | ||
0a234c6d | 1725 | static long |
8daaa831 | 1726 | xfs_fs_nr_cached_objects( |
9b17c623 | 1727 | struct super_block *sb, |
4101b624 | 1728 | struct shrink_control *sc) |
8daaa831 DC |
1729 | { |
1730 | return xfs_reclaim_inodes_count(XFS_M(sb)); | |
1731 | } | |
1732 | ||
0a234c6d | 1733 | static long |
8daaa831 DC |
1734 | xfs_fs_free_cached_objects( |
1735 | struct super_block *sb, | |
4101b624 | 1736 | struct shrink_control *sc) |
8daaa831 | 1737 | { |
4101b624 | 1738 | return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan); |
8daaa831 DC |
1739 | } |
1740 | ||
b87221de | 1741 | static const struct super_operations xfs_super_operations = { |
a50cd269 NS |
1742 | .alloc_inode = xfs_fs_alloc_inode, |
1743 | .destroy_inode = xfs_fs_destroy_inode, | |
5132ba8f | 1744 | .drop_inode = xfs_fs_drop_inode, |
a50cd269 | 1745 | .put_super = xfs_fs_put_super, |
69961a26 | 1746 | .sync_fs = xfs_fs_sync_fs, |
c4be0c1d | 1747 | .freeze_fs = xfs_fs_freeze, |
d5db0f97 | 1748 | .unfreeze_fs = xfs_fs_unfreeze, |
a50cd269 NS |
1749 | .statfs = xfs_fs_statfs, |
1750 | .remount_fs = xfs_fs_remount, | |
1751 | .show_options = xfs_fs_show_options, | |
8daaa831 DC |
1752 | .nr_cached_objects = xfs_fs_nr_cached_objects, |
1753 | .free_cached_objects = xfs_fs_free_cached_objects, | |
1da177e4 LT |
1754 | }; |
1755 | ||
5085b607 | 1756 | static struct file_system_type xfs_fs_type = { |
1da177e4 LT |
1757 | .owner = THIS_MODULE, |
1758 | .name = "xfs", | |
152a0836 | 1759 | .mount = xfs_fs_mount, |
1da177e4 LT |
1760 | .kill_sb = kill_block_super, |
1761 | .fs_flags = FS_REQUIRES_DEV, | |
1762 | }; | |
7f78e035 | 1763 | MODULE_ALIAS_FS("xfs"); |
1da177e4 | 1764 | |
9f8868ff CH |
1765 | STATIC int __init |
1766 | xfs_init_zones(void) | |
1767 | { | |
0e51a8e1 | 1768 | xfs_ioend_bioset = bioset_create(4 * MAX_BUF_PER_PAGE, |
011067b0 N |
1769 | offsetof(struct xfs_ioend, io_inline_bio), |
1770 | BIOSET_NEED_BVECS); | |
0e51a8e1 | 1771 | if (!xfs_ioend_bioset) |
bf904248 | 1772 | goto out; |
9f8868ff | 1773 | |
9f8868ff CH |
1774 | xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t), |
1775 | "xfs_log_ticket"); | |
1776 | if (!xfs_log_ticket_zone) | |
0e51a8e1 | 1777 | goto out_free_ioend_bioset; |
9f8868ff | 1778 | |
e66a4c67 | 1779 | xfs_bmap_free_item_zone = kmem_zone_init( |
310a75a3 | 1780 | sizeof(struct xfs_extent_free_item), |
e66a4c67 | 1781 | "xfs_bmap_free_item"); |
9f8868ff CH |
1782 | if (!xfs_bmap_free_item_zone) |
1783 | goto out_destroy_log_ticket_zone; | |
bf904248 | 1784 | |
9f8868ff CH |
1785 | xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t), |
1786 | "xfs_btree_cur"); | |
1787 | if (!xfs_btree_cur_zone) | |
1788 | goto out_destroy_bmap_free_item_zone; | |
1789 | ||
1790 | xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t), | |
1791 | "xfs_da_state"); | |
1792 | if (!xfs_da_state_zone) | |
1793 | goto out_destroy_btree_cur_zone; | |
1794 | ||
9f8868ff CH |
1795 | xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork"); |
1796 | if (!xfs_ifork_zone) | |
1d9025e5 | 1797 | goto out_destroy_da_state_zone; |
9f8868ff CH |
1798 | |
1799 | xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans"); | |
1800 | if (!xfs_trans_zone) | |
1801 | goto out_destroy_ifork_zone; | |
1802 | ||
e98c414f CH |
1803 | xfs_log_item_desc_zone = |
1804 | kmem_zone_init(sizeof(struct xfs_log_item_desc), | |
1805 | "xfs_log_item_desc"); | |
1806 | if (!xfs_log_item_desc_zone) | |
1807 | goto out_destroy_trans_zone; | |
1808 | ||
9f8868ff CH |
1809 | /* |
1810 | * The size of the zone allocated buf log item is the maximum | |
1811 | * size possible under XFS. This wastes a little bit of memory, | |
1812 | * but it is much faster. | |
1813 | */ | |
77c1a08f DC |
1814 | xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item), |
1815 | "xfs_buf_item"); | |
9f8868ff | 1816 | if (!xfs_buf_item_zone) |
e98c414f | 1817 | goto out_destroy_log_item_desc_zone; |
9f8868ff CH |
1818 | |
1819 | xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) + | |
1820 | ((XFS_EFD_MAX_FAST_EXTENTS - 1) * | |
1821 | sizeof(xfs_extent_t))), "xfs_efd_item"); | |
1822 | if (!xfs_efd_zone) | |
1823 | goto out_destroy_buf_item_zone; | |
1824 | ||
1825 | xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) + | |
1826 | ((XFS_EFI_MAX_FAST_EXTENTS - 1) * | |
1827 | sizeof(xfs_extent_t))), "xfs_efi_item"); | |
1828 | if (!xfs_efi_zone) | |
1829 | goto out_destroy_efd_zone; | |
1830 | ||
1831 | xfs_inode_zone = | |
1832 | kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode", | |
5d097056 VD |
1833 | KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD | |
1834 | KM_ZONE_ACCOUNT, xfs_fs_inode_init_once); | |
9f8868ff CH |
1835 | if (!xfs_inode_zone) |
1836 | goto out_destroy_efi_zone; | |
1837 | ||
1838 | xfs_ili_zone = | |
1839 | kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili", | |
1840 | KM_ZONE_SPREAD, NULL); | |
1841 | if (!xfs_ili_zone) | |
1842 | goto out_destroy_inode_zone; | |
3ebe7d2d DC |
1843 | xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item), |
1844 | "xfs_icr"); | |
1845 | if (!xfs_icreate_zone) | |
1846 | goto out_destroy_ili_zone; | |
9f8868ff | 1847 | |
722e2517 | 1848 | xfs_rud_zone = kmem_zone_init(sizeof(struct xfs_rud_log_item), |
5880f2d7 DW |
1849 | "xfs_rud_item"); |
1850 | if (!xfs_rud_zone) | |
1851 | goto out_destroy_icreate_zone; | |
1852 | ||
cd00158c DW |
1853 | xfs_rui_zone = kmem_zone_init( |
1854 | xfs_rui_log_item_sizeof(XFS_RUI_MAX_FAST_EXTENTS), | |
5880f2d7 DW |
1855 | "xfs_rui_item"); |
1856 | if (!xfs_rui_zone) | |
1857 | goto out_destroy_rud_zone; | |
1858 | ||
baf4bcac DW |
1859 | xfs_cud_zone = kmem_zone_init(sizeof(struct xfs_cud_log_item), |
1860 | "xfs_cud_item"); | |
1861 | if (!xfs_cud_zone) | |
1862 | goto out_destroy_rui_zone; | |
1863 | ||
1864 | xfs_cui_zone = kmem_zone_init( | |
1865 | xfs_cui_log_item_sizeof(XFS_CUI_MAX_FAST_EXTENTS), | |
1866 | "xfs_cui_item"); | |
1867 | if (!xfs_cui_zone) | |
1868 | goto out_destroy_cud_zone; | |
1869 | ||
6413a014 DW |
1870 | xfs_bud_zone = kmem_zone_init(sizeof(struct xfs_bud_log_item), |
1871 | "xfs_bud_item"); | |
1872 | if (!xfs_bud_zone) | |
1873 | goto out_destroy_cui_zone; | |
1874 | ||
1875 | xfs_bui_zone = kmem_zone_init( | |
1876 | xfs_bui_log_item_sizeof(XFS_BUI_MAX_FAST_EXTENTS), | |
1877 | "xfs_bui_item"); | |
1878 | if (!xfs_bui_zone) | |
1879 | goto out_destroy_bud_zone; | |
1880 | ||
9f8868ff CH |
1881 | return 0; |
1882 | ||
6413a014 DW |
1883 | out_destroy_bud_zone: |
1884 | kmem_zone_destroy(xfs_bud_zone); | |
1885 | out_destroy_cui_zone: | |
1886 | kmem_zone_destroy(xfs_cui_zone); | |
baf4bcac DW |
1887 | out_destroy_cud_zone: |
1888 | kmem_zone_destroy(xfs_cud_zone); | |
1889 | out_destroy_rui_zone: | |
1890 | kmem_zone_destroy(xfs_rui_zone); | |
5880f2d7 DW |
1891 | out_destroy_rud_zone: |
1892 | kmem_zone_destroy(xfs_rud_zone); | |
1893 | out_destroy_icreate_zone: | |
1894 | kmem_zone_destroy(xfs_icreate_zone); | |
3ebe7d2d DC |
1895 | out_destroy_ili_zone: |
1896 | kmem_zone_destroy(xfs_ili_zone); | |
9f8868ff CH |
1897 | out_destroy_inode_zone: |
1898 | kmem_zone_destroy(xfs_inode_zone); | |
1899 | out_destroy_efi_zone: | |
1900 | kmem_zone_destroy(xfs_efi_zone); | |
1901 | out_destroy_efd_zone: | |
1902 | kmem_zone_destroy(xfs_efd_zone); | |
1903 | out_destroy_buf_item_zone: | |
1904 | kmem_zone_destroy(xfs_buf_item_zone); | |
e98c414f CH |
1905 | out_destroy_log_item_desc_zone: |
1906 | kmem_zone_destroy(xfs_log_item_desc_zone); | |
9f8868ff CH |
1907 | out_destroy_trans_zone: |
1908 | kmem_zone_destroy(xfs_trans_zone); | |
1909 | out_destroy_ifork_zone: | |
1910 | kmem_zone_destroy(xfs_ifork_zone); | |
9f8868ff CH |
1911 | out_destroy_da_state_zone: |
1912 | kmem_zone_destroy(xfs_da_state_zone); | |
1913 | out_destroy_btree_cur_zone: | |
1914 | kmem_zone_destroy(xfs_btree_cur_zone); | |
1915 | out_destroy_bmap_free_item_zone: | |
1916 | kmem_zone_destroy(xfs_bmap_free_item_zone); | |
1917 | out_destroy_log_ticket_zone: | |
1918 | kmem_zone_destroy(xfs_log_ticket_zone); | |
0e51a8e1 CH |
1919 | out_free_ioend_bioset: |
1920 | bioset_free(xfs_ioend_bioset); | |
9f8868ff CH |
1921 | out: |
1922 | return -ENOMEM; | |
1923 | } | |
1924 | ||
1925 | STATIC void | |
1926 | xfs_destroy_zones(void) | |
1927 | { | |
8c0a8537 KS |
1928 | /* |
1929 | * Make sure all delayed rcu free are flushed before we | |
1930 | * destroy caches. | |
1931 | */ | |
1932 | rcu_barrier(); | |
6413a014 DW |
1933 | kmem_zone_destroy(xfs_bui_zone); |
1934 | kmem_zone_destroy(xfs_bud_zone); | |
baf4bcac DW |
1935 | kmem_zone_destroy(xfs_cui_zone); |
1936 | kmem_zone_destroy(xfs_cud_zone); | |
5880f2d7 DW |
1937 | kmem_zone_destroy(xfs_rui_zone); |
1938 | kmem_zone_destroy(xfs_rud_zone); | |
3ebe7d2d | 1939 | kmem_zone_destroy(xfs_icreate_zone); |
9f8868ff CH |
1940 | kmem_zone_destroy(xfs_ili_zone); |
1941 | kmem_zone_destroy(xfs_inode_zone); | |
1942 | kmem_zone_destroy(xfs_efi_zone); | |
1943 | kmem_zone_destroy(xfs_efd_zone); | |
1944 | kmem_zone_destroy(xfs_buf_item_zone); | |
e98c414f | 1945 | kmem_zone_destroy(xfs_log_item_desc_zone); |
9f8868ff CH |
1946 | kmem_zone_destroy(xfs_trans_zone); |
1947 | kmem_zone_destroy(xfs_ifork_zone); | |
9f8868ff CH |
1948 | kmem_zone_destroy(xfs_da_state_zone); |
1949 | kmem_zone_destroy(xfs_btree_cur_zone); | |
1950 | kmem_zone_destroy(xfs_bmap_free_item_zone); | |
1951 | kmem_zone_destroy(xfs_log_ticket_zone); | |
0e51a8e1 | 1952 | bioset_free(xfs_ioend_bioset); |
9f8868ff | 1953 | } |
1da177e4 | 1954 | |
0bf6a5bd DC |
1955 | STATIC int __init |
1956 | xfs_init_workqueues(void) | |
1957 | { | |
c999a223 DC |
1958 | /* |
1959 | * The allocation workqueue can be used in memory reclaim situations | |
1960 | * (writepage path), and parallelism is only limited by the number of | |
1961 | * AGs in all the filesystems mounted. Hence use the default large | |
1962 | * max_active value for this workqueue. | |
1963 | */ | |
8018ec08 BF |
1964 | xfs_alloc_wq = alloc_workqueue("xfsalloc", |
1965 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 0); | |
c999a223 | 1966 | if (!xfs_alloc_wq) |
5889608d | 1967 | return -ENOMEM; |
c999a223 | 1968 | |
4560e78f CH |
1969 | xfs_discard_wq = alloc_workqueue("xfsdiscard", WQ_UNBOUND, 0); |
1970 | if (!xfs_discard_wq) | |
1971 | goto out_free_alloc_wq; | |
1972 | ||
0bf6a5bd | 1973 | return 0; |
4560e78f CH |
1974 | out_free_alloc_wq: |
1975 | destroy_workqueue(xfs_alloc_wq); | |
1976 | return -ENOMEM; | |
0bf6a5bd DC |
1977 | } |
1978 | ||
39411f81 | 1979 | STATIC void |
0bf6a5bd DC |
1980 | xfs_destroy_workqueues(void) |
1981 | { | |
4560e78f | 1982 | destroy_workqueue(xfs_discard_wq); |
c999a223 | 1983 | destroy_workqueue(xfs_alloc_wq); |
0bf6a5bd DC |
1984 | } |
1985 | ||
1da177e4 | 1986 | STATIC int __init |
9f8868ff | 1987 | init_xfs_fs(void) |
1da177e4 LT |
1988 | { |
1989 | int error; | |
1da177e4 | 1990 | |
30cbc591 DW |
1991 | xfs_check_ondisk_structs(); |
1992 | ||
65795910 CH |
1993 | printk(KERN_INFO XFS_VERSION_STRING " with " |
1994 | XFS_BUILD_OPTIONS " enabled\n"); | |
1da177e4 | 1995 | |
9749fee8 | 1996 | xfs_extent_free_init_defer_op(); |
f8dbebef | 1997 | xfs_rmap_update_init_defer_op(); |
33ba6129 | 1998 | xfs_refcount_update_init_defer_op(); |
9f3afb57 | 1999 | xfs_bmap_update_init_defer_op(); |
9749fee8 | 2000 | |
9f8868ff | 2001 | xfs_dir_startup(); |
1da177e4 | 2002 | |
8758280f | 2003 | error = xfs_init_zones(); |
9f8868ff CH |
2004 | if (error) |
2005 | goto out; | |
2006 | ||
0bf6a5bd | 2007 | error = xfs_init_workqueues(); |
9f8868ff | 2008 | if (error) |
0b1b213f | 2009 | goto out_destroy_zones; |
9f8868ff | 2010 | |
0bf6a5bd DC |
2011 | error = xfs_mru_cache_init(); |
2012 | if (error) | |
2013 | goto out_destroy_wq; | |
2014 | ||
ce8e922c | 2015 | error = xfs_buf_init(); |
9f8868ff | 2016 | if (error) |
1919adda | 2017 | goto out_mru_cache_uninit; |
9f8868ff CH |
2018 | |
2019 | error = xfs_init_procfs(); | |
2020 | if (error) | |
2021 | goto out_buf_terminate; | |
2022 | ||
2023 | error = xfs_sysctl_register(); | |
2024 | if (error) | |
2025 | goto out_cleanup_procfs; | |
1da177e4 | 2026 | |
3d871226 BF |
2027 | xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj); |
2028 | if (!xfs_kset) { | |
2029 | error = -ENOMEM; | |
bb230c12 | 2030 | goto out_sysctl_unregister; |
3d871226 BF |
2031 | } |
2032 | ||
80529c45 BD |
2033 | xfsstats.xs_kobj.kobject.kset = xfs_kset; |
2034 | ||
2035 | xfsstats.xs_stats = alloc_percpu(struct xfsstats); | |
2036 | if (!xfsstats.xs_stats) { | |
2037 | error = -ENOMEM; | |
2038 | goto out_kset_unregister; | |
2039 | } | |
2040 | ||
2041 | error = xfs_sysfs_init(&xfsstats.xs_kobj, &xfs_stats_ktype, NULL, | |
bb230c12 BD |
2042 | "stats"); |
2043 | if (error) | |
80529c45 | 2044 | goto out_free_stats; |
bb230c12 | 2045 | |
65b65735 BF |
2046 | #ifdef DEBUG |
2047 | xfs_dbg_kobj.kobject.kset = xfs_kset; | |
2048 | error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug"); | |
a05931ce | 2049 | if (error) |
bb230c12 | 2050 | goto out_remove_stats_kobj; |
65b65735 BF |
2051 | #endif |
2052 | ||
2053 | error = xfs_qm_init(); | |
2054 | if (error) | |
bb230c12 | 2055 | goto out_remove_dbg_kobj; |
1da177e4 LT |
2056 | |
2057 | error = register_filesystem(&xfs_fs_type); | |
2058 | if (error) | |
a05931ce | 2059 | goto out_qm_exit; |
1da177e4 LT |
2060 | return 0; |
2061 | ||
a05931ce CH |
2062 | out_qm_exit: |
2063 | xfs_qm_exit(); | |
bb230c12 | 2064 | out_remove_dbg_kobj: |
65b65735 BF |
2065 | #ifdef DEBUG |
2066 | xfs_sysfs_del(&xfs_dbg_kobj); | |
bb230c12 | 2067 | out_remove_stats_kobj: |
65b65735 | 2068 | #endif |
80529c45 BD |
2069 | xfs_sysfs_del(&xfsstats.xs_kobj); |
2070 | out_free_stats: | |
2071 | free_percpu(xfsstats.xs_stats); | |
bb230c12 | 2072 | out_kset_unregister: |
3d871226 | 2073 | kset_unregister(xfs_kset); |
9f8868ff CH |
2074 | out_sysctl_unregister: |
2075 | xfs_sysctl_unregister(); | |
2076 | out_cleanup_procfs: | |
2077 | xfs_cleanup_procfs(); | |
2078 | out_buf_terminate: | |
ce8e922c | 2079 | xfs_buf_terminate(); |
9f8868ff CH |
2080 | out_mru_cache_uninit: |
2081 | xfs_mru_cache_uninit(); | |
0bf6a5bd DC |
2082 | out_destroy_wq: |
2083 | xfs_destroy_workqueues(); | |
9f8868ff | 2084 | out_destroy_zones: |
8758280f | 2085 | xfs_destroy_zones(); |
9f8868ff | 2086 | out: |
1da177e4 LT |
2087 | return error; |
2088 | } | |
2089 | ||
2090 | STATIC void __exit | |
9f8868ff | 2091 | exit_xfs_fs(void) |
1da177e4 | 2092 | { |
a05931ce | 2093 | xfs_qm_exit(); |
1da177e4 | 2094 | unregister_filesystem(&xfs_fs_type); |
65b65735 BF |
2095 | #ifdef DEBUG |
2096 | xfs_sysfs_del(&xfs_dbg_kobj); | |
2097 | #endif | |
80529c45 BD |
2098 | xfs_sysfs_del(&xfsstats.xs_kobj); |
2099 | free_percpu(xfsstats.xs_stats); | |
3d871226 | 2100 | kset_unregister(xfs_kset); |
9f8868ff CH |
2101 | xfs_sysctl_unregister(); |
2102 | xfs_cleanup_procfs(); | |
ce8e922c | 2103 | xfs_buf_terminate(); |
9f8868ff | 2104 | xfs_mru_cache_uninit(); |
0bf6a5bd | 2105 | xfs_destroy_workqueues(); |
8758280f | 2106 | xfs_destroy_zones(); |
af3b6382 | 2107 | xfs_uuid_table_free(); |
1da177e4 LT |
2108 | } |
2109 | ||
2110 | module_init(init_xfs_fs); | |
2111 | module_exit(exit_xfs_fs); | |
2112 | ||
2113 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
2114 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
2115 | MODULE_LICENSE("GPL"); |