Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
7b718769 NS |
2 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. |
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 | */ |
1da177e4 LT |
18 | #include "xfs.h" |
19 | #include "xfs_fs.h" | |
6ca1c906 | 20 | #include "xfs_format.h" |
ef14f0c1 | 21 | #include "xfs_acl.h" |
1da177e4 LT |
22 | #include "xfs_log.h" |
23 | #include "xfs_trans.h" | |
24 | #include "xfs_sb.h" | |
25 | #include "xfs_ag.h" | |
1da177e4 | 26 | #include "xfs_alloc.h" |
1da177e4 LT |
27 | #include "xfs_quota.h" |
28 | #include "xfs_mount.h" | |
1da177e4 | 29 | #include "xfs_bmap_btree.h" |
1da177e4 LT |
30 | #include "xfs_dinode.h" |
31 | #include "xfs_inode.h" | |
32 | #include "xfs_bmap.h" | |
68988114 | 33 | #include "xfs_bmap_util.h" |
1da177e4 LT |
34 | #include "xfs_rtalloc.h" |
35 | #include "xfs_error.h" | |
36 | #include "xfs_itable.h" | |
1da177e4 LT |
37 | #include "xfs_attr.h" |
38 | #include "xfs_buf_item.h" | |
39 | #include "xfs_utils.h" | |
739bfb2a | 40 | #include "xfs_vnodeops.h" |
c4ed4243 | 41 | #include "xfs_inode_item.h" |
0b1b213f | 42 | #include "xfs_trace.h" |
27b52867 | 43 | #include "xfs_icache.h" |
1da177e4 | 44 | |
16f7e0fe | 45 | #include <linux/capability.h> |
1da177e4 LT |
46 | #include <linux/xattr.h> |
47 | #include <linux/namei.h> | |
ef14f0c1 | 48 | #include <linux/posix_acl.h> |
446ada4a | 49 | #include <linux/security.h> |
f35642e2 | 50 | #include <linux/fiemap.h> |
5a0e3ad6 | 51 | #include <linux/slab.h> |
1da177e4 | 52 | |
8d2a5e6e DC |
53 | static int |
54 | xfs_initxattrs( | |
55 | struct inode *inode, | |
56 | const struct xattr *xattr_array, | |
57 | void *fs_info) | |
9d8f13ba | 58 | { |
8d2a5e6e DC |
59 | const struct xattr *xattr; |
60 | struct xfs_inode *ip = XFS_I(inode); | |
61 | int error = 0; | |
9d8f13ba MZ |
62 | |
63 | for (xattr = xattr_array; xattr->name != NULL; xattr++) { | |
64 | error = xfs_attr_set(ip, xattr->name, xattr->value, | |
65 | xattr->value_len, ATTR_SECURE); | |
66 | if (error < 0) | |
67 | break; | |
68 | } | |
69 | return error; | |
70 | } | |
71 | ||
446ada4a NS |
72 | /* |
73 | * Hook in SELinux. This is not quite correct yet, what we really need | |
74 | * here (as we do for default ACLs) is a mechanism by which creation of | |
75 | * these attrs can be journalled at inode creation time (along with the | |
76 | * inode, of course, such that log replay can't cause these to be lost). | |
77 | */ | |
9d8f13ba | 78 | |
446ada4a | 79 | STATIC int |
416c6d5b | 80 | xfs_init_security( |
af048193 | 81 | struct inode *inode, |
2a7dba39 EP |
82 | struct inode *dir, |
83 | const struct qstr *qstr) | |
446ada4a | 84 | { |
9d8f13ba MZ |
85 | return security_inode_init_security(inode, dir, qstr, |
86 | &xfs_initxattrs, NULL); | |
446ada4a NS |
87 | } |
88 | ||
556b8b16 BN |
89 | static void |
90 | xfs_dentry_to_name( | |
91 | struct xfs_name *namep, | |
92 | struct dentry *dentry) | |
93 | { | |
94 | namep->name = dentry->d_name.name; | |
95 | namep->len = dentry->d_name.len; | |
96 | } | |
97 | ||
7989cb8e | 98 | STATIC void |
416c6d5b | 99 | xfs_cleanup_inode( |
739bfb2a | 100 | struct inode *dir, |
af048193 | 101 | struct inode *inode, |
8f112e3b | 102 | struct dentry *dentry) |
3a69c7dc | 103 | { |
556b8b16 | 104 | struct xfs_name teardown; |
3a69c7dc YL |
105 | |
106 | /* Oh, the horror. | |
220b5284 | 107 | * If we can't add the ACL or we fail in |
416c6d5b | 108 | * xfs_init_security we must back out. |
3a69c7dc YL |
109 | * ENOSPC can hit here, among other things. |
110 | */ | |
556b8b16 | 111 | xfs_dentry_to_name(&teardown, dentry); |
3a69c7dc | 112 | |
8f112e3b | 113 | xfs_remove(XFS_I(dir), &teardown, XFS_I(inode)); |
af048193 | 114 | iput(inode); |
3a69c7dc YL |
115 | } |
116 | ||
1da177e4 | 117 | STATIC int |
416c6d5b | 118 | xfs_vn_mknod( |
1da177e4 LT |
119 | struct inode *dir, |
120 | struct dentry *dentry, | |
1a67aafb | 121 | umode_t mode, |
1da177e4 LT |
122 | dev_t rdev) |
123 | { | |
db0bb7ba | 124 | struct inode *inode; |
979ebab1 | 125 | struct xfs_inode *ip = NULL; |
ef14f0c1 | 126 | struct posix_acl *default_acl = NULL; |
556b8b16 | 127 | struct xfs_name name; |
1da177e4 LT |
128 | int error; |
129 | ||
130 | /* | |
131 | * Irix uses Missed'em'V split, but doesn't want to see | |
132 | * the upper 5 bits of (14bit) major. | |
133 | */ | |
517b5e8c CH |
134 | if (S_ISCHR(mode) || S_ISBLK(mode)) { |
135 | if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff)) | |
136 | return -EINVAL; | |
137 | rdev = sysv_encode_dev(rdev); | |
138 | } else { | |
139 | rdev = 0; | |
140 | } | |
1da177e4 | 141 | |
ef14f0c1 CH |
142 | if (IS_POSIXACL(dir)) { |
143 | default_acl = xfs_get_acl(dir, ACL_TYPE_DEFAULT); | |
144 | if (IS_ERR(default_acl)) | |
c46a131c | 145 | return PTR_ERR(default_acl); |
1da177e4 | 146 | |
e83f1eb6 CH |
147 | if (!default_acl) |
148 | mode &= ~current_umask(); | |
ef14f0c1 | 149 | } |
1da177e4 | 150 | |
517b5e8c | 151 | xfs_dentry_to_name(&name, dentry); |
6c77b0ea | 152 | error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip); |
db0bb7ba CH |
153 | if (unlikely(error)) |
154 | goto out_free_acl; | |
446ada4a | 155 | |
01651646 | 156 | inode = VFS_I(ip); |
979ebab1 | 157 | |
2a7dba39 | 158 | error = xfs_init_security(inode, dir, &dentry->d_name); |
db0bb7ba CH |
159 | if (unlikely(error)) |
160 | goto out_cleanup_inode; | |
161 | ||
162 | if (default_acl) { | |
ef14f0c1 | 163 | error = -xfs_inherit_acl(inode, default_acl); |
826cae2f | 164 | default_acl = NULL; |
db0bb7ba CH |
165 | if (unlikely(error)) |
166 | goto out_cleanup_inode; | |
1da177e4 LT |
167 | } |
168 | ||
1da177e4 | 169 | |
db0bb7ba | 170 | d_instantiate(dentry, inode); |
db0bb7ba CH |
171 | return -error; |
172 | ||
173 | out_cleanup_inode: | |
8f112e3b | 174 | xfs_cleanup_inode(dir, inode, dentry); |
db0bb7ba | 175 | out_free_acl: |
ef14f0c1 | 176 | posix_acl_release(default_acl); |
1da177e4 LT |
177 | return -error; |
178 | } | |
179 | ||
180 | STATIC int | |
416c6d5b | 181 | xfs_vn_create( |
1da177e4 LT |
182 | struct inode *dir, |
183 | struct dentry *dentry, | |
4acdaf27 | 184 | umode_t mode, |
ebfc3b49 | 185 | bool flags) |
1da177e4 | 186 | { |
416c6d5b | 187 | return xfs_vn_mknod(dir, dentry, mode, 0); |
1da177e4 LT |
188 | } |
189 | ||
190 | STATIC int | |
416c6d5b | 191 | xfs_vn_mkdir( |
1da177e4 LT |
192 | struct inode *dir, |
193 | struct dentry *dentry, | |
18bb1db3 | 194 | umode_t mode) |
1da177e4 | 195 | { |
416c6d5b | 196 | return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0); |
1da177e4 LT |
197 | } |
198 | ||
199 | STATIC struct dentry * | |
416c6d5b | 200 | xfs_vn_lookup( |
1da177e4 LT |
201 | struct inode *dir, |
202 | struct dentry *dentry, | |
00cd8dd3 | 203 | unsigned int flags) |
1da177e4 | 204 | { |
ef1f5e7a | 205 | struct xfs_inode *cip; |
556b8b16 | 206 | struct xfs_name name; |
1da177e4 LT |
207 | int error; |
208 | ||
209 | if (dentry->d_name.len >= MAXNAMELEN) | |
210 | return ERR_PTR(-ENAMETOOLONG); | |
211 | ||
556b8b16 | 212 | xfs_dentry_to_name(&name, dentry); |
384f3ced | 213 | error = xfs_lookup(XFS_I(dir), &name, &cip, NULL); |
67fcaa73 | 214 | if (unlikely(error)) { |
1da177e4 LT |
215 | if (unlikely(error != ENOENT)) |
216 | return ERR_PTR(-error); | |
217 | d_add(dentry, NULL); | |
218 | return NULL; | |
219 | } | |
220 | ||
01651646 | 221 | return d_splice_alias(VFS_I(cip), dentry); |
1da177e4 LT |
222 | } |
223 | ||
384f3ced BN |
224 | STATIC struct dentry * |
225 | xfs_vn_ci_lookup( | |
226 | struct inode *dir, | |
227 | struct dentry *dentry, | |
00cd8dd3 | 228 | unsigned int flags) |
384f3ced BN |
229 | { |
230 | struct xfs_inode *ip; | |
231 | struct xfs_name xname; | |
232 | struct xfs_name ci_name; | |
233 | struct qstr dname; | |
234 | int error; | |
235 | ||
236 | if (dentry->d_name.len >= MAXNAMELEN) | |
237 | return ERR_PTR(-ENAMETOOLONG); | |
238 | ||
239 | xfs_dentry_to_name(&xname, dentry); | |
240 | error = xfs_lookup(XFS_I(dir), &xname, &ip, &ci_name); | |
241 | if (unlikely(error)) { | |
242 | if (unlikely(error != ENOENT)) | |
243 | return ERR_PTR(-error); | |
866d5dc9 BN |
244 | /* |
245 | * call d_add(dentry, NULL) here when d_drop_negative_children | |
246 | * is called in xfs_vn_mknod (ie. allow negative dentries | |
247 | * with CI filesystems). | |
248 | */ | |
384f3ced BN |
249 | return NULL; |
250 | } | |
251 | ||
252 | /* if exact match, just splice and exit */ | |
253 | if (!ci_name.name) | |
01651646 | 254 | return d_splice_alias(VFS_I(ip), dentry); |
384f3ced BN |
255 | |
256 | /* else case-insensitive match... */ | |
257 | dname.name = ci_name.name; | |
258 | dname.len = ci_name.len; | |
e45b590b | 259 | dentry = d_add_ci(dentry, VFS_I(ip), &dname); |
384f3ced BN |
260 | kmem_free(ci_name.name); |
261 | return dentry; | |
262 | } | |
263 | ||
1da177e4 | 264 | STATIC int |
416c6d5b | 265 | xfs_vn_link( |
1da177e4 LT |
266 | struct dentry *old_dentry, |
267 | struct inode *dir, | |
268 | struct dentry *dentry) | |
269 | { | |
d9424b3c | 270 | struct inode *inode = old_dentry->d_inode; |
556b8b16 | 271 | struct xfs_name name; |
1da177e4 LT |
272 | int error; |
273 | ||
556b8b16 | 274 | xfs_dentry_to_name(&name, dentry); |
1da177e4 | 275 | |
556b8b16 | 276 | error = xfs_link(XFS_I(dir), XFS_I(inode), &name); |
d9424b3c | 277 | if (unlikely(error)) |
a3da7896 | 278 | return -error; |
a3da7896 | 279 | |
7de9c6ee | 280 | ihold(inode); |
a3da7896 CH |
281 | d_instantiate(dentry, inode); |
282 | return 0; | |
1da177e4 LT |
283 | } |
284 | ||
285 | STATIC int | |
416c6d5b | 286 | xfs_vn_unlink( |
1da177e4 LT |
287 | struct inode *dir, |
288 | struct dentry *dentry) | |
289 | { | |
556b8b16 | 290 | struct xfs_name name; |
1da177e4 LT |
291 | int error; |
292 | ||
556b8b16 | 293 | xfs_dentry_to_name(&name, dentry); |
1da177e4 | 294 | |
e5700704 CH |
295 | error = -xfs_remove(XFS_I(dir), &name, XFS_I(dentry->d_inode)); |
296 | if (error) | |
297 | return error; | |
298 | ||
299 | /* | |
300 | * With unlink, the VFS makes the dentry "negative": no inode, | |
301 | * but still hashed. This is incompatible with case-insensitive | |
302 | * mode, so invalidate (unhash) the dentry in CI-mode. | |
303 | */ | |
304 | if (xfs_sb_version_hasasciici(&XFS_M(dir->i_sb)->m_sb)) | |
305 | d_invalidate(dentry); | |
306 | return 0; | |
1da177e4 LT |
307 | } |
308 | ||
309 | STATIC int | |
416c6d5b | 310 | xfs_vn_symlink( |
1da177e4 LT |
311 | struct inode *dir, |
312 | struct dentry *dentry, | |
313 | const char *symname) | |
314 | { | |
3937be5b CH |
315 | struct inode *inode; |
316 | struct xfs_inode *cip = NULL; | |
556b8b16 | 317 | struct xfs_name name; |
1da177e4 | 318 | int error; |
576b1d67 | 319 | umode_t mode; |
1da177e4 | 320 | |
3e5daf05 | 321 | mode = S_IFLNK | |
ce3b0f8d | 322 | (irix_symlink_mode ? 0777 & ~current_umask() : S_IRWXUGO); |
556b8b16 | 323 | xfs_dentry_to_name(&name, dentry); |
1da177e4 | 324 | |
6c77b0ea | 325 | error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip); |
3937be5b CH |
326 | if (unlikely(error)) |
327 | goto out; | |
328 | ||
01651646 | 329 | inode = VFS_I(cip); |
3937be5b | 330 | |
2a7dba39 | 331 | error = xfs_init_security(inode, dir, &dentry->d_name); |
3937be5b CH |
332 | if (unlikely(error)) |
333 | goto out_cleanup_inode; | |
334 | ||
335 | d_instantiate(dentry, inode); | |
3937be5b CH |
336 | return 0; |
337 | ||
338 | out_cleanup_inode: | |
8f112e3b | 339 | xfs_cleanup_inode(dir, inode, dentry); |
3937be5b | 340 | out: |
1da177e4 LT |
341 | return -error; |
342 | } | |
343 | ||
1da177e4 | 344 | STATIC int |
416c6d5b | 345 | xfs_vn_rename( |
1da177e4 LT |
346 | struct inode *odir, |
347 | struct dentry *odentry, | |
348 | struct inode *ndir, | |
349 | struct dentry *ndentry) | |
350 | { | |
351 | struct inode *new_inode = ndentry->d_inode; | |
556b8b16 BN |
352 | struct xfs_name oname; |
353 | struct xfs_name nname; | |
1da177e4 | 354 | |
556b8b16 BN |
355 | xfs_dentry_to_name(&oname, odentry); |
356 | xfs_dentry_to_name(&nname, ndentry); | |
357 | ||
e5700704 | 358 | return -xfs_rename(XFS_I(odir), &oname, XFS_I(odentry->d_inode), |
cfa853e4 CH |
359 | XFS_I(ndir), &nname, new_inode ? |
360 | XFS_I(new_inode) : NULL); | |
1da177e4 LT |
361 | } |
362 | ||
363 | /* | |
364 | * careful here - this function can get called recursively, so | |
365 | * we need to be very careful about how much stack we use. | |
366 | * uio is kmalloced for this reason... | |
367 | */ | |
008b150a | 368 | STATIC void * |
416c6d5b | 369 | xfs_vn_follow_link( |
1da177e4 LT |
370 | struct dentry *dentry, |
371 | struct nameidata *nd) | |
372 | { | |
1da177e4 | 373 | char *link; |
804c83c3 | 374 | int error = -ENOMEM; |
1da177e4 | 375 | |
f52720ca | 376 | link = kmalloc(MAXPATHLEN+1, GFP_KERNEL); |
804c83c3 CH |
377 | if (!link) |
378 | goto out_err; | |
1da177e4 | 379 | |
739bfb2a | 380 | error = -xfs_readlink(XFS_I(dentry->d_inode), link); |
804c83c3 CH |
381 | if (unlikely(error)) |
382 | goto out_kfree; | |
1da177e4 LT |
383 | |
384 | nd_set_link(nd, link); | |
008b150a | 385 | return NULL; |
804c83c3 CH |
386 | |
387 | out_kfree: | |
388 | kfree(link); | |
389 | out_err: | |
390 | nd_set_link(nd, ERR_PTR(error)); | |
391 | return NULL; | |
1da177e4 LT |
392 | } |
393 | ||
cde410a9 | 394 | STATIC void |
416c6d5b | 395 | xfs_vn_put_link( |
cde410a9 NS |
396 | struct dentry *dentry, |
397 | struct nameidata *nd, | |
398 | void *p) | |
1da177e4 | 399 | { |
cde410a9 NS |
400 | char *s = nd_get_link(nd); |
401 | ||
1da177e4 LT |
402 | if (!IS_ERR(s)) |
403 | kfree(s); | |
404 | } | |
405 | ||
1da177e4 | 406 | STATIC int |
416c6d5b | 407 | xfs_vn_getattr( |
c43f4087 CH |
408 | struct vfsmount *mnt, |
409 | struct dentry *dentry, | |
410 | struct kstat *stat) | |
1da177e4 | 411 | { |
c43f4087 CH |
412 | struct inode *inode = dentry->d_inode; |
413 | struct xfs_inode *ip = XFS_I(inode); | |
414 | struct xfs_mount *mp = ip->i_mount; | |
415 | ||
cca28fb8 | 416 | trace_xfs_getattr(ip); |
c43f4087 CH |
417 | |
418 | if (XFS_FORCED_SHUTDOWN(mp)) | |
ed32201e | 419 | return -XFS_ERROR(EIO); |
c43f4087 CH |
420 | |
421 | stat->size = XFS_ISIZE(ip); | |
422 | stat->dev = inode->i_sb->s_dev; | |
423 | stat->mode = ip->i_d.di_mode; | |
424 | stat->nlink = ip->i_d.di_nlink; | |
425 | stat->uid = ip->i_d.di_uid; | |
426 | stat->gid = ip->i_d.di_gid; | |
427 | stat->ino = ip->i_ino; | |
c43f4087 | 428 | stat->atime = inode->i_atime; |
f9581b14 CH |
429 | stat->mtime = inode->i_mtime; |
430 | stat->ctime = inode->i_ctime; | |
c43f4087 CH |
431 | stat->blocks = |
432 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
433 | ||
434 | ||
435 | switch (inode->i_mode & S_IFMT) { | |
436 | case S_IFBLK: | |
437 | case S_IFCHR: | |
438 | stat->blksize = BLKDEV_IOSIZE; | |
439 | stat->rdev = MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff, | |
440 | sysv_minor(ip->i_df.if_u2.if_rdev)); | |
441 | break; | |
442 | default: | |
71ddabb9 | 443 | if (XFS_IS_REALTIME_INODE(ip)) { |
c43f4087 CH |
444 | /* |
445 | * If the file blocks are being allocated from a | |
446 | * realtime volume, then return the inode's realtime | |
447 | * extent size or the realtime volume's extent size. | |
448 | */ | |
449 | stat->blksize = | |
450 | xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog; | |
451 | } else | |
452 | stat->blksize = xfs_preferred_iosize(mp); | |
453 | stat->rdev = 0; | |
454 | break; | |
69e23b9a | 455 | } |
c43f4087 CH |
456 | |
457 | return 0; | |
1da177e4 LT |
458 | } |
459 | ||
56c19e89 DC |
460 | static void |
461 | xfs_setattr_mode( | |
462 | struct xfs_trans *tp, | |
463 | struct xfs_inode *ip, | |
464 | struct iattr *iattr) | |
465 | { | |
466 | struct inode *inode = VFS_I(ip); | |
467 | umode_t mode = iattr->ia_mode; | |
468 | ||
469 | ASSERT(tp); | |
470 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); | |
471 | ||
56c19e89 DC |
472 | ip->i_d.di_mode &= S_IFMT; |
473 | ip->i_d.di_mode |= mode & ~S_IFMT; | |
474 | ||
475 | inode->i_mode &= S_IFMT; | |
476 | inode->i_mode |= mode & ~S_IFMT; | |
477 | } | |
478 | ||
c4ed4243 CH |
479 | int |
480 | xfs_setattr_nonsize( | |
481 | struct xfs_inode *ip, | |
482 | struct iattr *iattr, | |
483 | int flags) | |
484 | { | |
485 | xfs_mount_t *mp = ip->i_mount; | |
486 | struct inode *inode = VFS_I(ip); | |
487 | int mask = iattr->ia_valid; | |
488 | xfs_trans_t *tp; | |
489 | int error; | |
490 | uid_t uid = 0, iuid = 0; | |
491 | gid_t gid = 0, igid = 0; | |
492 | struct xfs_dquot *udqp = NULL, *gdqp = NULL; | |
493 | struct xfs_dquot *olddquot1 = NULL, *olddquot2 = NULL; | |
494 | ||
495 | trace_xfs_setattr(ip); | |
496 | ||
42c49d7f CM |
497 | /* If acls are being inherited, we already have this checked */ |
498 | if (!(flags & XFS_ATTR_NOACL)) { | |
499 | if (mp->m_flags & XFS_MOUNT_RDONLY) | |
500 | return XFS_ERROR(EROFS); | |
c4ed4243 | 501 | |
42c49d7f CM |
502 | if (XFS_FORCED_SHUTDOWN(mp)) |
503 | return XFS_ERROR(EIO); | |
c4ed4243 | 504 | |
42c49d7f CM |
505 | error = -inode_change_ok(inode, iattr); |
506 | if (error) | |
507 | return XFS_ERROR(error); | |
508 | } | |
c4ed4243 CH |
509 | |
510 | ASSERT((mask & ATTR_SIZE) == 0); | |
511 | ||
512 | /* | |
513 | * If disk quotas is on, we make sure that the dquots do exist on disk, | |
514 | * before we start any other transactions. Trying to do this later | |
515 | * is messy. We don't care to take a readlock to look at the ids | |
516 | * in inode here, because we can't hold it across the trans_reserve. | |
517 | * If the IDs do change before we take the ilock, we're covered | |
518 | * because the i_*dquot fields will get updated anyway. | |
519 | */ | |
520 | if (XFS_IS_QUOTA_ON(mp) && (mask & (ATTR_UID|ATTR_GID))) { | |
521 | uint qflags = 0; | |
522 | ||
523 | if ((mask & ATTR_UID) && XFS_IS_UQUOTA_ON(mp)) { | |
524 | uid = iattr->ia_uid; | |
525 | qflags |= XFS_QMOPT_UQUOTA; | |
526 | } else { | |
527 | uid = ip->i_d.di_uid; | |
528 | } | |
529 | if ((mask & ATTR_GID) && XFS_IS_GQUOTA_ON(mp)) { | |
530 | gid = iattr->ia_gid; | |
531 | qflags |= XFS_QMOPT_GQUOTA; | |
532 | } else { | |
533 | gid = ip->i_d.di_gid; | |
534 | } | |
535 | ||
536 | /* | |
537 | * We take a reference when we initialize udqp and gdqp, | |
538 | * so it is important that we never blindly double trip on | |
539 | * the same variable. See xfs_create() for an example. | |
540 | */ | |
541 | ASSERT(udqp == NULL); | |
542 | ASSERT(gdqp == NULL); | |
543 | error = xfs_qm_vop_dqalloc(ip, uid, gid, xfs_get_projid(ip), | |
92f8ff73 | 544 | qflags, &udqp, &gdqp, NULL); |
c4ed4243 CH |
545 | if (error) |
546 | return error; | |
547 | } | |
548 | ||
549 | tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE); | |
550 | error = xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0); | |
551 | if (error) | |
552 | goto out_dqrele; | |
553 | ||
554 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
555 | ||
556 | /* | |
557 | * Change file ownership. Must be the owner or privileged. | |
558 | */ | |
559 | if (mask & (ATTR_UID|ATTR_GID)) { | |
560 | /* | |
561 | * These IDs could have changed since we last looked at them. | |
562 | * But, we're assured that if the ownership did change | |
563 | * while we didn't have the inode locked, inode's dquot(s) | |
564 | * would have changed also. | |
565 | */ | |
566 | iuid = ip->i_d.di_uid; | |
567 | igid = ip->i_d.di_gid; | |
568 | gid = (mask & ATTR_GID) ? iattr->ia_gid : igid; | |
569 | uid = (mask & ATTR_UID) ? iattr->ia_uid : iuid; | |
570 | ||
571 | /* | |
572 | * Do a quota reservation only if uid/gid is actually | |
573 | * going to change. | |
574 | */ | |
575 | if (XFS_IS_QUOTA_RUNNING(mp) && | |
576 | ((XFS_IS_UQUOTA_ON(mp) && iuid != uid) || | |
577 | (XFS_IS_GQUOTA_ON(mp) && igid != gid))) { | |
578 | ASSERT(tp); | |
579 | error = xfs_qm_vop_chown_reserve(tp, ip, udqp, gdqp, | |
92f8ff73 | 580 | NULL, capable(CAP_FOWNER) ? |
c4ed4243 CH |
581 | XFS_QMOPT_FORCE_RES : 0); |
582 | if (error) /* out of quota */ | |
583 | goto out_trans_cancel; | |
584 | } | |
585 | } | |
586 | ||
ddc3415a | 587 | xfs_trans_ijoin(tp, ip, 0); |
c4ed4243 CH |
588 | |
589 | /* | |
590 | * Change file ownership. Must be the owner or privileged. | |
591 | */ | |
592 | if (mask & (ATTR_UID|ATTR_GID)) { | |
593 | /* | |
594 | * CAP_FSETID overrides the following restrictions: | |
595 | * | |
596 | * The set-user-ID and set-group-ID bits of a file will be | |
597 | * cleared upon successful return from chown() | |
598 | */ | |
599 | if ((ip->i_d.di_mode & (S_ISUID|S_ISGID)) && | |
600 | !capable(CAP_FSETID)) | |
601 | ip->i_d.di_mode &= ~(S_ISUID|S_ISGID); | |
602 | ||
603 | /* | |
604 | * Change the ownerships and register quota modifications | |
605 | * in the transaction. | |
606 | */ | |
607 | if (iuid != uid) { | |
608 | if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_UQUOTA_ON(mp)) { | |
609 | ASSERT(mask & ATTR_UID); | |
610 | ASSERT(udqp); | |
611 | olddquot1 = xfs_qm_vop_chown(tp, ip, | |
612 | &ip->i_udquot, udqp); | |
613 | } | |
614 | ip->i_d.di_uid = uid; | |
615 | inode->i_uid = uid; | |
616 | } | |
617 | if (igid != gid) { | |
618 | if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_GQUOTA_ON(mp)) { | |
619 | ASSERT(!XFS_IS_PQUOTA_ON(mp)); | |
620 | ASSERT(mask & ATTR_GID); | |
621 | ASSERT(gdqp); | |
622 | olddquot2 = xfs_qm_vop_chown(tp, ip, | |
623 | &ip->i_gdquot, gdqp); | |
624 | } | |
625 | ip->i_d.di_gid = gid; | |
626 | inode->i_gid = gid; | |
627 | } | |
628 | } | |
629 | ||
630 | /* | |
631 | * Change file access modes. | |
632 | */ | |
56c19e89 DC |
633 | if (mask & ATTR_MODE) |
634 | xfs_setattr_mode(tp, ip, iattr); | |
c4ed4243 CH |
635 | |
636 | /* | |
637 | * Change file access or modified times. | |
638 | */ | |
639 | if (mask & ATTR_ATIME) { | |
640 | inode->i_atime = iattr->ia_atime; | |
641 | ip->i_d.di_atime.t_sec = iattr->ia_atime.tv_sec; | |
642 | ip->i_d.di_atime.t_nsec = iattr->ia_atime.tv_nsec; | |
c4ed4243 CH |
643 | } |
644 | if (mask & ATTR_CTIME) { | |
645 | inode->i_ctime = iattr->ia_ctime; | |
646 | ip->i_d.di_ctime.t_sec = iattr->ia_ctime.tv_sec; | |
647 | ip->i_d.di_ctime.t_nsec = iattr->ia_ctime.tv_nsec; | |
c4ed4243 CH |
648 | } |
649 | if (mask & ATTR_MTIME) { | |
650 | inode->i_mtime = iattr->ia_mtime; | |
651 | ip->i_d.di_mtime.t_sec = iattr->ia_mtime.tv_sec; | |
652 | ip->i_d.di_mtime.t_nsec = iattr->ia_mtime.tv_nsec; | |
c4ed4243 CH |
653 | } |
654 | ||
655 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); | |
656 | ||
657 | XFS_STATS_INC(xs_ig_attrchg); | |
658 | ||
659 | if (mp->m_flags & XFS_MOUNT_WSYNC) | |
660 | xfs_trans_set_sync(tp); | |
661 | error = xfs_trans_commit(tp, 0); | |
662 | ||
663 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
664 | ||
665 | /* | |
666 | * Release any dquot(s) the inode had kept before chown. | |
667 | */ | |
668 | xfs_qm_dqrele(olddquot1); | |
669 | xfs_qm_dqrele(olddquot2); | |
670 | xfs_qm_dqrele(udqp); | |
671 | xfs_qm_dqrele(gdqp); | |
672 | ||
673 | if (error) | |
674 | return XFS_ERROR(error); | |
675 | ||
676 | /* | |
677 | * XXX(hch): Updating the ACL entries is not atomic vs the i_mode | |
678 | * update. We could avoid this with linked transactions | |
679 | * and passing down the transaction pointer all the way | |
680 | * to attr_set. No previous user of the generic | |
681 | * Posix ACL code seems to care about this issue either. | |
682 | */ | |
683 | if ((mask & ATTR_MODE) && !(flags & XFS_ATTR_NOACL)) { | |
684 | error = -xfs_acl_chmod(inode); | |
685 | if (error) | |
686 | return XFS_ERROR(error); | |
687 | } | |
688 | ||
689 | return 0; | |
690 | ||
691 | out_trans_cancel: | |
692 | xfs_trans_cancel(tp, 0); | |
693 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
694 | out_dqrele: | |
695 | xfs_qm_dqrele(udqp); | |
696 | xfs_qm_dqrele(gdqp); | |
697 | return error; | |
698 | } | |
699 | ||
700 | /* | |
701 | * Truncate file. Must have write permission and not be a directory. | |
702 | */ | |
703 | int | |
704 | xfs_setattr_size( | |
705 | struct xfs_inode *ip, | |
706 | struct iattr *iattr, | |
707 | int flags) | |
708 | { | |
709 | struct xfs_mount *mp = ip->i_mount; | |
710 | struct inode *inode = VFS_I(ip); | |
711 | int mask = iattr->ia_valid; | |
673e8e59 | 712 | xfs_off_t oldsize, newsize; |
c4ed4243 CH |
713 | struct xfs_trans *tp; |
714 | int error; | |
f38996f5 | 715 | uint lock_flags = 0; |
c4ed4243 CH |
716 | uint commit_flags = 0; |
717 | ||
718 | trace_xfs_setattr(ip); | |
719 | ||
720 | if (mp->m_flags & XFS_MOUNT_RDONLY) | |
721 | return XFS_ERROR(EROFS); | |
722 | ||
723 | if (XFS_FORCED_SHUTDOWN(mp)) | |
724 | return XFS_ERROR(EIO); | |
725 | ||
726 | error = -inode_change_ok(inode, iattr); | |
727 | if (error) | |
728 | return XFS_ERROR(error); | |
729 | ||
730 | ASSERT(S_ISREG(ip->i_d.di_mode)); | |
56c19e89 DC |
731 | ASSERT((mask & (ATTR_UID|ATTR_GID|ATTR_ATIME|ATTR_ATIME_SET| |
732 | ATTR_MTIME_SET|ATTR_KILL_PRIV|ATTR_TIMES_SET)) == 0); | |
c4ed4243 | 733 | |
f38996f5 | 734 | if (!(flags & XFS_ATTR_NOLOCK)) { |
c4ed4243 | 735 | lock_flags |= XFS_IOLOCK_EXCL; |
f38996f5 CH |
736 | xfs_ilock(ip, lock_flags); |
737 | } | |
c4ed4243 | 738 | |
ce7ae151 | 739 | oldsize = inode->i_size; |
673e8e59 CH |
740 | newsize = iattr->ia_size; |
741 | ||
c4ed4243 CH |
742 | /* |
743 | * Short circuit the truncate case for zero length files. | |
744 | */ | |
673e8e59 | 745 | if (newsize == 0 && oldsize == 0 && ip->i_d.di_nextents == 0) { |
681b1200 CH |
746 | if (!(mask & (ATTR_CTIME|ATTR_MTIME))) |
747 | goto out_unlock; | |
748 | ||
749 | /* | |
750 | * Use the regular setattr path to update the timestamps. | |
751 | */ | |
752 | xfs_iunlock(ip, lock_flags); | |
753 | iattr->ia_valid &= ~ATTR_SIZE; | |
754 | return xfs_setattr_nonsize(ip, iattr, 0); | |
c4ed4243 CH |
755 | } |
756 | ||
757 | /* | |
758 | * Make sure that the dquots are attached to the inode. | |
759 | */ | |
f38996f5 | 760 | error = xfs_qm_dqattach(ip, 0); |
c4ed4243 CH |
761 | if (error) |
762 | goto out_unlock; | |
763 | ||
764 | /* | |
765 | * Now we can make the changes. Before we join the inode to the | |
766 | * transaction, take care of the part of the truncation that must be | |
767 | * done without the inode lock. This needs to be done before joining | |
768 | * the inode to the transaction, because the inode cannot be unlocked | |
769 | * once it is a part of the transaction. | |
770 | */ | |
673e8e59 | 771 | if (newsize > oldsize) { |
c4ed4243 CH |
772 | /* |
773 | * Do the first part of growing a file: zero any data in the | |
774 | * last block that is beyond the old EOF. We need to do this | |
775 | * before the inode is joined to the transaction to modify | |
776 | * i_size. | |
777 | */ | |
673e8e59 | 778 | error = xfs_zero_eof(ip, newsize, oldsize); |
c4ed4243 CH |
779 | if (error) |
780 | goto out_unlock; | |
781 | } | |
c4ed4243 CH |
782 | |
783 | /* | |
784 | * We are going to log the inode size change in this transaction so | |
785 | * any previous writes that are beyond the on disk EOF and the new | |
786 | * EOF that have not been written out need to be written here. If we | |
787 | * do not write the data out, we expose ourselves to the null files | |
788 | * problem. | |
789 | * | |
790 | * Only flush from the on disk size to the smaller of the in memory | |
791 | * file size or the new size as that's the range we really care about | |
792 | * here and prevents waiting for other data not within the range we | |
793 | * care about here. | |
794 | */ | |
673e8e59 | 795 | if (oldsize != ip->i_d.di_size && newsize > ip->i_d.di_size) { |
4bc1ea6b DC |
796 | error = -filemap_write_and_wait_range(VFS_I(ip)->i_mapping, |
797 | ip->i_d.di_size, newsize); | |
c4ed4243 CH |
798 | if (error) |
799 | goto out_unlock; | |
800 | } | |
801 | ||
802 | /* | |
4a06fd26 | 803 | * Wait for all direct I/O to complete. |
c4ed4243 | 804 | */ |
4a06fd26 | 805 | inode_dio_wait(inode); |
c4ed4243 | 806 | |
673e8e59 | 807 | error = -block_truncate_page(inode->i_mapping, newsize, xfs_get_blocks); |
c4ed4243 CH |
808 | if (error) |
809 | goto out_unlock; | |
810 | ||
811 | tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_SIZE); | |
812 | error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0, | |
813 | XFS_TRANS_PERM_LOG_RES, | |
814 | XFS_ITRUNCATE_LOG_COUNT); | |
815 | if (error) | |
816 | goto out_trans_cancel; | |
817 | ||
673e8e59 | 818 | truncate_setsize(inode, newsize); |
c4ed4243 CH |
819 | |
820 | commit_flags = XFS_TRANS_RELEASE_LOG_RES; | |
821 | lock_flags |= XFS_ILOCK_EXCL; | |
822 | ||
823 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
824 | ||
ddc3415a | 825 | xfs_trans_ijoin(tp, ip, 0); |
c4ed4243 CH |
826 | |
827 | /* | |
828 | * Only change the c/mtime if we are changing the size or we are | |
829 | * explicitly asked to change it. This handles the semantic difference | |
830 | * between truncate() and ftruncate() as implemented in the VFS. | |
831 | * | |
832 | * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a | |
833 | * special case where we need to update the times despite not having | |
834 | * these flags set. For all other operations the VFS set these flags | |
835 | * explicitly if it wants a timestamp update. | |
836 | */ | |
673e8e59 | 837 | if (newsize != oldsize && (!(mask & (ATTR_CTIME | ATTR_MTIME)))) { |
c4ed4243 CH |
838 | iattr->ia_ctime = iattr->ia_mtime = |
839 | current_fs_time(inode->i_sb); | |
840 | mask |= ATTR_CTIME | ATTR_MTIME; | |
841 | } | |
842 | ||
673e8e59 CH |
843 | /* |
844 | * The first thing we do is set the size to new_size permanently on | |
845 | * disk. This way we don't have to worry about anyone ever being able | |
846 | * to look at the data being freed even in the face of a crash. | |
847 | * What we're getting around here is the case where we free a block, it | |
848 | * is allocated to another file, it is written to, and then we crash. | |
849 | * If the new data gets written to the file but the log buffers | |
850 | * containing the free and reallocation don't, then we'd end up with | |
851 | * garbage in the blocks being freed. As long as we make the new size | |
852 | * permanent before actually freeing any blocks it doesn't matter if | |
853 | * they get written to. | |
854 | */ | |
855 | ip->i_d.di_size = newsize; | |
673e8e59 CH |
856 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); |
857 | ||
858 | if (newsize <= oldsize) { | |
859 | error = xfs_itruncate_extents(&tp, ip, XFS_DATA_FORK, newsize); | |
c4ed4243 CH |
860 | if (error) |
861 | goto out_trans_abort; | |
862 | ||
863 | /* | |
864 | * Truncated "down", so we're removing references to old data | |
865 | * here - if we delay flushing for a long time, we expose | |
866 | * ourselves unduly to the notorious NULL files problem. So, | |
867 | * we mark this inode and flush it when the file is closed, | |
868 | * and do not wait the usual (long) time for writeout. | |
869 | */ | |
870 | xfs_iflags_set(ip, XFS_ITRUNCATED); | |
27b52867 BF |
871 | |
872 | /* A truncate down always removes post-EOF blocks. */ | |
873 | xfs_inode_clear_eofblocks_tag(ip); | |
c4ed4243 CH |
874 | } |
875 | ||
56c19e89 DC |
876 | /* |
877 | * Change file access modes. | |
878 | */ | |
879 | if (mask & ATTR_MODE) | |
880 | xfs_setattr_mode(tp, ip, iattr); | |
881 | ||
c4ed4243 CH |
882 | if (mask & ATTR_CTIME) { |
883 | inode->i_ctime = iattr->ia_ctime; | |
884 | ip->i_d.di_ctime.t_sec = iattr->ia_ctime.tv_sec; | |
885 | ip->i_d.di_ctime.t_nsec = iattr->ia_ctime.tv_nsec; | |
c4ed4243 CH |
886 | } |
887 | if (mask & ATTR_MTIME) { | |
888 | inode->i_mtime = iattr->ia_mtime; | |
889 | ip->i_d.di_mtime.t_sec = iattr->ia_mtime.tv_sec; | |
890 | ip->i_d.di_mtime.t_nsec = iattr->ia_mtime.tv_nsec; | |
c4ed4243 CH |
891 | } |
892 | ||
893 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); | |
894 | ||
895 | XFS_STATS_INC(xs_ig_attrchg); | |
896 | ||
897 | if (mp->m_flags & XFS_MOUNT_WSYNC) | |
898 | xfs_trans_set_sync(tp); | |
899 | ||
900 | error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES); | |
901 | out_unlock: | |
902 | if (lock_flags) | |
903 | xfs_iunlock(ip, lock_flags); | |
904 | return error; | |
905 | ||
906 | out_trans_abort: | |
907 | commit_flags |= XFS_TRANS_ABORT; | |
908 | out_trans_cancel: | |
909 | xfs_trans_cancel(tp, commit_flags); | |
910 | goto out_unlock; | |
911 | } | |
912 | ||
1da177e4 | 913 | STATIC int |
416c6d5b | 914 | xfs_vn_setattr( |
1da177e4 | 915 | struct dentry *dentry, |
0f285c8a | 916 | struct iattr *iattr) |
1da177e4 | 917 | { |
c4ed4243 CH |
918 | if (iattr->ia_valid & ATTR_SIZE) |
919 | return -xfs_setattr_size(XFS_I(dentry->d_inode), iattr, 0); | |
920 | return -xfs_setattr_nonsize(XFS_I(dentry->d_inode), iattr, 0); | |
1da177e4 LT |
921 | } |
922 | ||
69ff2826 CH |
923 | STATIC int |
924 | xfs_vn_update_time( | |
925 | struct inode *inode, | |
926 | struct timespec *now, | |
927 | int flags) | |
928 | { | |
929 | struct xfs_inode *ip = XFS_I(inode); | |
930 | struct xfs_mount *mp = ip->i_mount; | |
931 | struct xfs_trans *tp; | |
932 | int error; | |
933 | ||
934 | trace_xfs_update_time(ip); | |
935 | ||
936 | tp = xfs_trans_alloc(mp, XFS_TRANS_FSYNC_TS); | |
937 | error = xfs_trans_reserve(tp, 0, XFS_FSYNC_TS_LOG_RES(mp), 0, 0, 0); | |
938 | if (error) { | |
939 | xfs_trans_cancel(tp, 0); | |
940 | return -error; | |
941 | } | |
942 | ||
943 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
944 | if (flags & S_CTIME) { | |
945 | inode->i_ctime = *now; | |
946 | ip->i_d.di_ctime.t_sec = (__int32_t)now->tv_sec; | |
947 | ip->i_d.di_ctime.t_nsec = (__int32_t)now->tv_nsec; | |
948 | } | |
949 | if (flags & S_MTIME) { | |
950 | inode->i_mtime = *now; | |
951 | ip->i_d.di_mtime.t_sec = (__int32_t)now->tv_sec; | |
952 | ip->i_d.di_mtime.t_nsec = (__int32_t)now->tv_nsec; | |
953 | } | |
954 | if (flags & S_ATIME) { | |
955 | inode->i_atime = *now; | |
956 | ip->i_d.di_atime.t_sec = (__int32_t)now->tv_sec; | |
957 | ip->i_d.di_atime.t_nsec = (__int32_t)now->tv_nsec; | |
958 | } | |
959 | xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); | |
960 | xfs_trans_log_inode(tp, ip, XFS_ILOG_TIMESTAMP); | |
961 | return -xfs_trans_commit(tp, 0); | |
962 | } | |
963 | ||
f35642e2 ES |
964 | #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR) |
965 | ||
966 | /* | |
967 | * Call fiemap helper to fill in user data. | |
968 | * Returns positive errors to xfs_getbmap. | |
969 | */ | |
970 | STATIC int | |
971 | xfs_fiemap_format( | |
972 | void **arg, | |
973 | struct getbmapx *bmv, | |
974 | int *full) | |
975 | { | |
976 | int error; | |
977 | struct fiemap_extent_info *fieinfo = *arg; | |
978 | u32 fiemap_flags = 0; | |
979 | u64 logical, physical, length; | |
980 | ||
981 | /* Do nothing for a hole */ | |
982 | if (bmv->bmv_block == -1LL) | |
983 | return 0; | |
984 | ||
985 | logical = BBTOB(bmv->bmv_offset); | |
986 | physical = BBTOB(bmv->bmv_block); | |
987 | length = BBTOB(bmv->bmv_length); | |
988 | ||
989 | if (bmv->bmv_oflags & BMV_OF_PREALLOC) | |
990 | fiemap_flags |= FIEMAP_EXTENT_UNWRITTEN; | |
991 | else if (bmv->bmv_oflags & BMV_OF_DELALLOC) { | |
635c4d0b JL |
992 | fiemap_flags |= (FIEMAP_EXTENT_DELALLOC | |
993 | FIEMAP_EXTENT_UNKNOWN); | |
f35642e2 ES |
994 | physical = 0; /* no block yet */ |
995 | } | |
996 | if (bmv->bmv_oflags & BMV_OF_LAST) | |
997 | fiemap_flags |= FIEMAP_EXTENT_LAST; | |
998 | ||
999 | error = fiemap_fill_next_extent(fieinfo, logical, physical, | |
1000 | length, fiemap_flags); | |
1001 | if (error > 0) { | |
1002 | error = 0; | |
1003 | *full = 1; /* user array now full */ | |
1004 | } | |
1005 | ||
1006 | return -error; | |
1007 | } | |
1008 | ||
1009 | STATIC int | |
1010 | xfs_vn_fiemap( | |
1011 | struct inode *inode, | |
1012 | struct fiemap_extent_info *fieinfo, | |
1013 | u64 start, | |
1014 | u64 length) | |
1015 | { | |
1016 | xfs_inode_t *ip = XFS_I(inode); | |
1017 | struct getbmapx bm; | |
1018 | int error; | |
1019 | ||
1020 | error = fiemap_check_flags(fieinfo, XFS_FIEMAP_FLAGS); | |
1021 | if (error) | |
1022 | return error; | |
1023 | ||
1024 | /* Set up bmap header for xfs internal routine */ | |
1025 | bm.bmv_offset = BTOBB(start); | |
1026 | /* Special case for whole file */ | |
1027 | if (length == FIEMAP_MAX_OFFSET) | |
1028 | bm.bmv_length = -1LL; | |
1029 | else | |
1030 | bm.bmv_length = BTOBB(length); | |
1031 | ||
97db39a1 | 1032 | /* We add one because in getbmap world count includes the header */ |
2d1ff3c7 TM |
1033 | bm.bmv_count = !fieinfo->fi_extents_max ? MAXEXTNUM : |
1034 | fieinfo->fi_extents_max + 1; | |
1035 | bm.bmv_count = min_t(__s32, bm.bmv_count, | |
1036 | (PAGE_SIZE * 16 / sizeof(struct getbmapx))); | |
9af25465 | 1037 | bm.bmv_iflags = BMV_IF_PREALLOC | BMV_IF_NO_HOLES; |
f35642e2 ES |
1038 | if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) |
1039 | bm.bmv_iflags |= BMV_IF_ATTRFORK; | |
1040 | if (!(fieinfo->fi_flags & FIEMAP_FLAG_SYNC)) | |
1041 | bm.bmv_iflags |= BMV_IF_DELALLOC; | |
1042 | ||
1043 | error = xfs_getbmap(ip, &bm, xfs_fiemap_format, fieinfo); | |
1044 | if (error) | |
1045 | return -error; | |
1046 | ||
1047 | return 0; | |
1048 | } | |
1049 | ||
41be8bed | 1050 | static const struct inode_operations xfs_inode_operations = { |
4e34e719 | 1051 | .get_acl = xfs_get_acl, |
416c6d5b NS |
1052 | .getattr = xfs_vn_getattr, |
1053 | .setattr = xfs_vn_setattr, | |
0ec58516 LM |
1054 | .setxattr = generic_setxattr, |
1055 | .getxattr = generic_getxattr, | |
1056 | .removexattr = generic_removexattr, | |
416c6d5b | 1057 | .listxattr = xfs_vn_listxattr, |
f35642e2 | 1058 | .fiemap = xfs_vn_fiemap, |
69ff2826 | 1059 | .update_time = xfs_vn_update_time, |
1da177e4 LT |
1060 | }; |
1061 | ||
41be8bed | 1062 | static const struct inode_operations xfs_dir_inode_operations = { |
416c6d5b NS |
1063 | .create = xfs_vn_create, |
1064 | .lookup = xfs_vn_lookup, | |
1065 | .link = xfs_vn_link, | |
1066 | .unlink = xfs_vn_unlink, | |
1067 | .symlink = xfs_vn_symlink, | |
1068 | .mkdir = xfs_vn_mkdir, | |
8f112e3b CH |
1069 | /* |
1070 | * Yes, XFS uses the same method for rmdir and unlink. | |
1071 | * | |
1072 | * There are some subtile differences deeper in the code, | |
1073 | * but we use S_ISDIR to check for those. | |
1074 | */ | |
1075 | .rmdir = xfs_vn_unlink, | |
416c6d5b NS |
1076 | .mknod = xfs_vn_mknod, |
1077 | .rename = xfs_vn_rename, | |
4e34e719 | 1078 | .get_acl = xfs_get_acl, |
416c6d5b NS |
1079 | .getattr = xfs_vn_getattr, |
1080 | .setattr = xfs_vn_setattr, | |
0ec58516 LM |
1081 | .setxattr = generic_setxattr, |
1082 | .getxattr = generic_getxattr, | |
1083 | .removexattr = generic_removexattr, | |
416c6d5b | 1084 | .listxattr = xfs_vn_listxattr, |
69ff2826 | 1085 | .update_time = xfs_vn_update_time, |
1da177e4 LT |
1086 | }; |
1087 | ||
41be8bed | 1088 | static const struct inode_operations xfs_dir_ci_inode_operations = { |
384f3ced BN |
1089 | .create = xfs_vn_create, |
1090 | .lookup = xfs_vn_ci_lookup, | |
1091 | .link = xfs_vn_link, | |
1092 | .unlink = xfs_vn_unlink, | |
1093 | .symlink = xfs_vn_symlink, | |
1094 | .mkdir = xfs_vn_mkdir, | |
8f112e3b CH |
1095 | /* |
1096 | * Yes, XFS uses the same method for rmdir and unlink. | |
1097 | * | |
1098 | * There are some subtile differences deeper in the code, | |
1099 | * but we use S_ISDIR to check for those. | |
1100 | */ | |
1101 | .rmdir = xfs_vn_unlink, | |
384f3ced BN |
1102 | .mknod = xfs_vn_mknod, |
1103 | .rename = xfs_vn_rename, | |
4e34e719 | 1104 | .get_acl = xfs_get_acl, |
384f3ced BN |
1105 | .getattr = xfs_vn_getattr, |
1106 | .setattr = xfs_vn_setattr, | |
0ec58516 LM |
1107 | .setxattr = generic_setxattr, |
1108 | .getxattr = generic_getxattr, | |
1109 | .removexattr = generic_removexattr, | |
384f3ced | 1110 | .listxattr = xfs_vn_listxattr, |
69ff2826 | 1111 | .update_time = xfs_vn_update_time, |
384f3ced BN |
1112 | }; |
1113 | ||
41be8bed | 1114 | static const struct inode_operations xfs_symlink_inode_operations = { |
1da177e4 | 1115 | .readlink = generic_readlink, |
416c6d5b NS |
1116 | .follow_link = xfs_vn_follow_link, |
1117 | .put_link = xfs_vn_put_link, | |
4e34e719 | 1118 | .get_acl = xfs_get_acl, |
416c6d5b NS |
1119 | .getattr = xfs_vn_getattr, |
1120 | .setattr = xfs_vn_setattr, | |
0ec58516 LM |
1121 | .setxattr = generic_setxattr, |
1122 | .getxattr = generic_getxattr, | |
1123 | .removexattr = generic_removexattr, | |
416c6d5b | 1124 | .listxattr = xfs_vn_listxattr, |
69ff2826 | 1125 | .update_time = xfs_vn_update_time, |
1da177e4 | 1126 | }; |
41be8bed CH |
1127 | |
1128 | STATIC void | |
1129 | xfs_diflags_to_iflags( | |
1130 | struct inode *inode, | |
1131 | struct xfs_inode *ip) | |
1132 | { | |
1133 | if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE) | |
1134 | inode->i_flags |= S_IMMUTABLE; | |
1135 | else | |
1136 | inode->i_flags &= ~S_IMMUTABLE; | |
1137 | if (ip->i_d.di_flags & XFS_DIFLAG_APPEND) | |
1138 | inode->i_flags |= S_APPEND; | |
1139 | else | |
1140 | inode->i_flags &= ~S_APPEND; | |
1141 | if (ip->i_d.di_flags & XFS_DIFLAG_SYNC) | |
1142 | inode->i_flags |= S_SYNC; | |
1143 | else | |
1144 | inode->i_flags &= ~S_SYNC; | |
1145 | if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME) | |
1146 | inode->i_flags |= S_NOATIME; | |
1147 | else | |
1148 | inode->i_flags &= ~S_NOATIME; | |
1149 | } | |
1150 | ||
1151 | /* | |
1152 | * Initialize the Linux inode, set up the operation vectors and | |
1153 | * unlock the inode. | |
1154 | * | |
1155 | * When reading existing inodes from disk this is called directly | |
1156 | * from xfs_iget, when creating a new inode it is called from | |
1157 | * xfs_ialloc after setting up the inode. | |
bf904248 DC |
1158 | * |
1159 | * We are always called with an uninitialised linux inode here. | |
1160 | * We need to initialise the necessary fields and take a reference | |
1161 | * on it. | |
41be8bed CH |
1162 | */ |
1163 | void | |
1164 | xfs_setup_inode( | |
1165 | struct xfs_inode *ip) | |
1166 | { | |
bf904248 DC |
1167 | struct inode *inode = &ip->i_vnode; |
1168 | ||
1169 | inode->i_ino = ip->i_ino; | |
eaff8079 | 1170 | inode->i_state = I_NEW; |
646ec461 CH |
1171 | |
1172 | inode_sb_list_add(inode); | |
c6f6cd06 CH |
1173 | /* make the inode look hashed for the writeback code */ |
1174 | hlist_add_fake(&inode->i_hash); | |
41be8bed CH |
1175 | |
1176 | inode->i_mode = ip->i_d.di_mode; | |
bfe86848 | 1177 | set_nlink(inode, ip->i_d.di_nlink); |
41be8bed CH |
1178 | inode->i_uid = ip->i_d.di_uid; |
1179 | inode->i_gid = ip->i_d.di_gid; | |
1180 | ||
1181 | switch (inode->i_mode & S_IFMT) { | |
1182 | case S_IFBLK: | |
1183 | case S_IFCHR: | |
1184 | inode->i_rdev = | |
1185 | MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff, | |
1186 | sysv_minor(ip->i_df.if_u2.if_rdev)); | |
1187 | break; | |
1188 | default: | |
1189 | inode->i_rdev = 0; | |
1190 | break; | |
1191 | } | |
1192 | ||
1193 | inode->i_generation = ip->i_d.di_gen; | |
1194 | i_size_write(inode, ip->i_d.di_size); | |
1195 | inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec; | |
1196 | inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec; | |
1197 | inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec; | |
1198 | inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec; | |
1199 | inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec; | |
1200 | inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec; | |
1201 | xfs_diflags_to_iflags(inode, ip); | |
41be8bed CH |
1202 | |
1203 | switch (inode->i_mode & S_IFMT) { | |
1204 | case S_IFREG: | |
1205 | inode->i_op = &xfs_inode_operations; | |
1206 | inode->i_fop = &xfs_file_operations; | |
1207 | inode->i_mapping->a_ops = &xfs_address_space_operations; | |
1208 | break; | |
1209 | case S_IFDIR: | |
1210 | if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb)) | |
1211 | inode->i_op = &xfs_dir_ci_inode_operations; | |
1212 | else | |
1213 | inode->i_op = &xfs_dir_inode_operations; | |
1214 | inode->i_fop = &xfs_dir_file_operations; | |
1215 | break; | |
1216 | case S_IFLNK: | |
1217 | inode->i_op = &xfs_symlink_inode_operations; | |
1218 | if (!(ip->i_df.if_flags & XFS_IFINLINE)) | |
1219 | inode->i_mapping->a_ops = &xfs_address_space_operations; | |
1220 | break; | |
1221 | default: | |
1222 | inode->i_op = &xfs_inode_operations; | |
1223 | init_special_inode(inode, inode->i_mode, inode->i_rdev); | |
1224 | break; | |
1225 | } | |
1226 | ||
510792ee CH |
1227 | /* |
1228 | * If there is no attribute fork no ACL can exist on this inode, | |
1229 | * and it can't have any file capabilities attached to it either. | |
1230 | */ | |
1231 | if (!XFS_IFORK_Q(ip)) { | |
1232 | inode_has_no_xattr(inode); | |
6311b108 | 1233 | cache_no_acl(inode); |
510792ee | 1234 | } |
6311b108 | 1235 | |
41be8bed CH |
1236 | xfs_iflags_clear(ip, XFS_INEW); |
1237 | barrier(); | |
1238 | ||
1239 | unlock_new_inode(inode); | |
1240 | } |