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" | |
70a9883c | 20 | #include "xfs_shared.h" |
239880ef DC |
21 | #include "xfs_format.h" |
22 | #include "xfs_log_format.h" | |
23 | #include "xfs_trans_resv.h" | |
1da177e4 | 24 | #include "xfs_mount.h" |
57062787 | 25 | #include "xfs_da_format.h" |
1da177e4 LT |
26 | #include "xfs_inode.h" |
27 | #include "xfs_bmap.h" | |
68988114 | 28 | #include "xfs_bmap_util.h" |
239880ef | 29 | #include "xfs_acl.h" |
239880ef | 30 | #include "xfs_quota.h" |
1da177e4 | 31 | #include "xfs_error.h" |
1da177e4 | 32 | #include "xfs_attr.h" |
239880ef | 33 | #include "xfs_trans.h" |
0b1b213f | 34 | #include "xfs_trace.h" |
27b52867 | 35 | #include "xfs_icache.h" |
c24b5dfa | 36 | #include "xfs_symlink.h" |
0cb97766 | 37 | #include "xfs_da_btree.h" |
1b767ee3 | 38 | #include "xfs_dir2.h" |
99b6436b | 39 | #include "xfs_trans_space.h" |
781355c6 | 40 | #include "xfs_pnfs.h" |
1da177e4 | 41 | |
16f7e0fe | 42 | #include <linux/capability.h> |
1da177e4 | 43 | #include <linux/xattr.h> |
ef14f0c1 | 44 | #include <linux/posix_acl.h> |
446ada4a | 45 | #include <linux/security.h> |
f35642e2 | 46 | #include <linux/fiemap.h> |
5a0e3ad6 | 47 | #include <linux/slab.h> |
1da177e4 | 48 | |
93a8614e DC |
49 | /* |
50 | * Directories have different lock order w.r.t. mmap_sem compared to regular | |
51 | * files. This is due to readdir potentially triggering page faults on a user | |
52 | * buffer inside filldir(), and this happens with the ilock on the directory | |
53 | * held. For regular files, the lock order is the other way around - the | |
54 | * mmap_sem is taken during the page fault, and then we lock the ilock to do | |
55 | * block mapping. Hence we need a different class for the directory ilock so | |
56 | * that lockdep can tell them apart. | |
57 | */ | |
58 | static struct lock_class_key xfs_nondir_ilock_class; | |
59 | static struct lock_class_key xfs_dir_ilock_class; | |
60 | ||
8d2a5e6e DC |
61 | static int |
62 | xfs_initxattrs( | |
63 | struct inode *inode, | |
64 | const struct xattr *xattr_array, | |
65 | void *fs_info) | |
9d8f13ba | 66 | { |
8d2a5e6e DC |
67 | const struct xattr *xattr; |
68 | struct xfs_inode *ip = XFS_I(inode); | |
69 | int error = 0; | |
9d8f13ba MZ |
70 | |
71 | for (xattr = xattr_array; xattr->name != NULL; xattr++) { | |
2451337d | 72 | error = xfs_attr_set(ip, xattr->name, xattr->value, |
a5a14de2 | 73 | xattr->value_len, ATTR_SECURE); |
9d8f13ba MZ |
74 | if (error < 0) |
75 | break; | |
76 | } | |
77 | return error; | |
78 | } | |
79 | ||
446ada4a NS |
80 | /* |
81 | * Hook in SELinux. This is not quite correct yet, what we really need | |
82 | * here (as we do for default ACLs) is a mechanism by which creation of | |
83 | * these attrs can be journalled at inode creation time (along with the | |
84 | * inode, of course, such that log replay can't cause these to be lost). | |
85 | */ | |
9d8f13ba | 86 | |
446ada4a | 87 | STATIC int |
416c6d5b | 88 | xfs_init_security( |
af048193 | 89 | struct inode *inode, |
2a7dba39 EP |
90 | struct inode *dir, |
91 | const struct qstr *qstr) | |
446ada4a | 92 | { |
2451337d | 93 | return security_inode_init_security(inode, dir, qstr, |
a5a14de2 | 94 | &xfs_initxattrs, NULL); |
446ada4a NS |
95 | } |
96 | ||
556b8b16 BN |
97 | static void |
98 | xfs_dentry_to_name( | |
99 | struct xfs_name *namep, | |
0cb97766 DC |
100 | struct dentry *dentry, |
101 | int mode) | |
556b8b16 BN |
102 | { |
103 | namep->name = dentry->d_name.name; | |
104 | namep->len = dentry->d_name.len; | |
0cb97766 | 105 | namep->type = xfs_mode_to_ftype[(mode & S_IFMT) >> S_SHIFT]; |
556b8b16 BN |
106 | } |
107 | ||
7989cb8e | 108 | STATIC void |
416c6d5b | 109 | xfs_cleanup_inode( |
739bfb2a | 110 | struct inode *dir, |
af048193 | 111 | struct inode *inode, |
8f112e3b | 112 | struct dentry *dentry) |
3a69c7dc | 113 | { |
556b8b16 | 114 | struct xfs_name teardown; |
3a69c7dc YL |
115 | |
116 | /* Oh, the horror. | |
220b5284 | 117 | * If we can't add the ACL or we fail in |
416c6d5b | 118 | * xfs_init_security we must back out. |
3a69c7dc YL |
119 | * ENOSPC can hit here, among other things. |
120 | */ | |
0cb97766 | 121 | xfs_dentry_to_name(&teardown, dentry, 0); |
3a69c7dc | 122 | |
8f112e3b | 123 | xfs_remove(XFS_I(dir), &teardown, XFS_I(inode)); |
3a69c7dc YL |
124 | } |
125 | ||
1da177e4 | 126 | STATIC int |
d540e43b | 127 | xfs_generic_create( |
1da177e4 LT |
128 | struct inode *dir, |
129 | struct dentry *dentry, | |
1a67aafb | 130 | umode_t mode, |
d540e43b BF |
131 | dev_t rdev, |
132 | bool tmpfile) /* unnamed file */ | |
1da177e4 | 133 | { |
db0bb7ba | 134 | struct inode *inode; |
979ebab1 | 135 | struct xfs_inode *ip = NULL; |
2401dc29 | 136 | struct posix_acl *default_acl, *acl; |
556b8b16 | 137 | struct xfs_name name; |
1da177e4 LT |
138 | int error; |
139 | ||
140 | /* | |
141 | * Irix uses Missed'em'V split, but doesn't want to see | |
142 | * the upper 5 bits of (14bit) major. | |
143 | */ | |
517b5e8c CH |
144 | if (S_ISCHR(mode) || S_ISBLK(mode)) { |
145 | if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff)) | |
146 | return -EINVAL; | |
147 | rdev = sysv_encode_dev(rdev); | |
148 | } else { | |
149 | rdev = 0; | |
150 | } | |
1da177e4 | 151 | |
2401dc29 CH |
152 | error = posix_acl_create(dir, &mode, &default_acl, &acl); |
153 | if (error) | |
154 | return error; | |
1da177e4 | 155 | |
d540e43b BF |
156 | if (!tmpfile) { |
157 | xfs_dentry_to_name(&name, dentry, mode); | |
158 | error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip); | |
159 | } else { | |
160 | error = xfs_create_tmpfile(XFS_I(dir), dentry, mode, &ip); | |
161 | } | |
db0bb7ba CH |
162 | if (unlikely(error)) |
163 | goto out_free_acl; | |
446ada4a | 164 | |
01651646 | 165 | inode = VFS_I(ip); |
979ebab1 | 166 | |
2a7dba39 | 167 | error = xfs_init_security(inode, dir, &dentry->d_name); |
db0bb7ba CH |
168 | if (unlikely(error)) |
169 | goto out_cleanup_inode; | |
170 | ||
2401dc29 | 171 | #ifdef CONFIG_XFS_POSIX_ACL |
db0bb7ba | 172 | if (default_acl) { |
2451337d | 173 | error = xfs_set_acl(inode, default_acl, ACL_TYPE_DEFAULT); |
2401dc29 | 174 | if (error) |
db0bb7ba | 175 | goto out_cleanup_inode; |
1da177e4 | 176 | } |
2401dc29 | 177 | if (acl) { |
2451337d | 178 | error = xfs_set_acl(inode, acl, ACL_TYPE_ACCESS); |
2401dc29 CH |
179 | if (error) |
180 | goto out_cleanup_inode; | |
181 | } | |
182 | #endif | |
1da177e4 | 183 | |
2b3d1d41 CH |
184 | xfs_setup_iops(ip); |
185 | ||
d540e43b BF |
186 | if (tmpfile) |
187 | d_tmpfile(dentry, inode); | |
188 | else | |
189 | d_instantiate(dentry, inode); | |
190 | ||
58c90473 DC |
191 | xfs_finish_inode_setup(ip); |
192 | ||
2401dc29 CH |
193 | out_free_acl: |
194 | if (default_acl) | |
195 | posix_acl_release(default_acl); | |
196 | if (acl) | |
197 | posix_acl_release(acl); | |
2451337d | 198 | return error; |
db0bb7ba CH |
199 | |
200 | out_cleanup_inode: | |
58c90473 | 201 | xfs_finish_inode_setup(ip); |
d540e43b BF |
202 | if (!tmpfile) |
203 | xfs_cleanup_inode(dir, inode, dentry); | |
204 | iput(inode); | |
2401dc29 | 205 | goto out_free_acl; |
1da177e4 LT |
206 | } |
207 | ||
d540e43b BF |
208 | STATIC int |
209 | xfs_vn_mknod( | |
210 | struct inode *dir, | |
211 | struct dentry *dentry, | |
212 | umode_t mode, | |
213 | dev_t rdev) | |
214 | { | |
215 | return xfs_generic_create(dir, dentry, mode, rdev, false); | |
216 | } | |
217 | ||
1da177e4 | 218 | STATIC int |
416c6d5b | 219 | xfs_vn_create( |
1da177e4 LT |
220 | struct inode *dir, |
221 | struct dentry *dentry, | |
4acdaf27 | 222 | umode_t mode, |
ebfc3b49 | 223 | bool flags) |
1da177e4 | 224 | { |
416c6d5b | 225 | return xfs_vn_mknod(dir, dentry, mode, 0); |
1da177e4 LT |
226 | } |
227 | ||
228 | STATIC int | |
416c6d5b | 229 | xfs_vn_mkdir( |
1da177e4 LT |
230 | struct inode *dir, |
231 | struct dentry *dentry, | |
18bb1db3 | 232 | umode_t mode) |
1da177e4 | 233 | { |
416c6d5b | 234 | return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0); |
1da177e4 LT |
235 | } |
236 | ||
237 | STATIC struct dentry * | |
416c6d5b | 238 | xfs_vn_lookup( |
1da177e4 LT |
239 | struct inode *dir, |
240 | struct dentry *dentry, | |
00cd8dd3 | 241 | unsigned int flags) |
1da177e4 | 242 | { |
ef1f5e7a | 243 | struct xfs_inode *cip; |
556b8b16 | 244 | struct xfs_name name; |
1da177e4 LT |
245 | int error; |
246 | ||
247 | if (dentry->d_name.len >= MAXNAMELEN) | |
248 | return ERR_PTR(-ENAMETOOLONG); | |
249 | ||
0cb97766 | 250 | xfs_dentry_to_name(&name, dentry, 0); |
384f3ced | 251 | error = xfs_lookup(XFS_I(dir), &name, &cip, NULL); |
67fcaa73 | 252 | if (unlikely(error)) { |
2451337d DC |
253 | if (unlikely(error != -ENOENT)) |
254 | return ERR_PTR(error); | |
1da177e4 LT |
255 | d_add(dentry, NULL); |
256 | return NULL; | |
257 | } | |
258 | ||
01651646 | 259 | return d_splice_alias(VFS_I(cip), dentry); |
1da177e4 LT |
260 | } |
261 | ||
384f3ced BN |
262 | STATIC struct dentry * |
263 | xfs_vn_ci_lookup( | |
264 | struct inode *dir, | |
265 | struct dentry *dentry, | |
00cd8dd3 | 266 | unsigned int flags) |
384f3ced BN |
267 | { |
268 | struct xfs_inode *ip; | |
269 | struct xfs_name xname; | |
270 | struct xfs_name ci_name; | |
271 | struct qstr dname; | |
272 | int error; | |
273 | ||
274 | if (dentry->d_name.len >= MAXNAMELEN) | |
275 | return ERR_PTR(-ENAMETOOLONG); | |
276 | ||
0cb97766 | 277 | xfs_dentry_to_name(&xname, dentry, 0); |
384f3ced BN |
278 | error = xfs_lookup(XFS_I(dir), &xname, &ip, &ci_name); |
279 | if (unlikely(error)) { | |
2451337d DC |
280 | if (unlikely(error != -ENOENT)) |
281 | return ERR_PTR(error); | |
866d5dc9 BN |
282 | /* |
283 | * call d_add(dentry, NULL) here when d_drop_negative_children | |
284 | * is called in xfs_vn_mknod (ie. allow negative dentries | |
285 | * with CI filesystems). | |
286 | */ | |
384f3ced BN |
287 | return NULL; |
288 | } | |
289 | ||
290 | /* if exact match, just splice and exit */ | |
291 | if (!ci_name.name) | |
01651646 | 292 | return d_splice_alias(VFS_I(ip), dentry); |
384f3ced BN |
293 | |
294 | /* else case-insensitive match... */ | |
295 | dname.name = ci_name.name; | |
296 | dname.len = ci_name.len; | |
e45b590b | 297 | dentry = d_add_ci(dentry, VFS_I(ip), &dname); |
384f3ced BN |
298 | kmem_free(ci_name.name); |
299 | return dentry; | |
300 | } | |
301 | ||
1da177e4 | 302 | STATIC int |
416c6d5b | 303 | xfs_vn_link( |
1da177e4 LT |
304 | struct dentry *old_dentry, |
305 | struct inode *dir, | |
306 | struct dentry *dentry) | |
307 | { | |
2b0143b5 | 308 | struct inode *inode = d_inode(old_dentry); |
556b8b16 | 309 | struct xfs_name name; |
1da177e4 LT |
310 | int error; |
311 | ||
0cb97766 | 312 | xfs_dentry_to_name(&name, dentry, inode->i_mode); |
1da177e4 | 313 | |
556b8b16 | 314 | error = xfs_link(XFS_I(dir), XFS_I(inode), &name); |
d9424b3c | 315 | if (unlikely(error)) |
2451337d | 316 | return error; |
a3da7896 | 317 | |
7de9c6ee | 318 | ihold(inode); |
a3da7896 CH |
319 | d_instantiate(dentry, inode); |
320 | return 0; | |
1da177e4 LT |
321 | } |
322 | ||
323 | STATIC int | |
416c6d5b | 324 | xfs_vn_unlink( |
1da177e4 LT |
325 | struct inode *dir, |
326 | struct dentry *dentry) | |
327 | { | |
556b8b16 | 328 | struct xfs_name name; |
1da177e4 LT |
329 | int error; |
330 | ||
0cb97766 | 331 | xfs_dentry_to_name(&name, dentry, 0); |
1da177e4 | 332 | |
2b0143b5 | 333 | error = xfs_remove(XFS_I(dir), &name, XFS_I(d_inode(dentry))); |
e5700704 CH |
334 | if (error) |
335 | return error; | |
336 | ||
337 | /* | |
338 | * With unlink, the VFS makes the dentry "negative": no inode, | |
339 | * but still hashed. This is incompatible with case-insensitive | |
340 | * mode, so invalidate (unhash) the dentry in CI-mode. | |
341 | */ | |
342 | if (xfs_sb_version_hasasciici(&XFS_M(dir->i_sb)->m_sb)) | |
343 | d_invalidate(dentry); | |
344 | return 0; | |
1da177e4 LT |
345 | } |
346 | ||
347 | STATIC int | |
416c6d5b | 348 | xfs_vn_symlink( |
1da177e4 LT |
349 | struct inode *dir, |
350 | struct dentry *dentry, | |
351 | const char *symname) | |
352 | { | |
3937be5b CH |
353 | struct inode *inode; |
354 | struct xfs_inode *cip = NULL; | |
556b8b16 | 355 | struct xfs_name name; |
1da177e4 | 356 | int error; |
576b1d67 | 357 | umode_t mode; |
1da177e4 | 358 | |
3e5daf05 | 359 | mode = S_IFLNK | |
ce3b0f8d | 360 | (irix_symlink_mode ? 0777 & ~current_umask() : S_IRWXUGO); |
0cb97766 | 361 | xfs_dentry_to_name(&name, dentry, mode); |
1da177e4 | 362 | |
6c77b0ea | 363 | error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip); |
3937be5b CH |
364 | if (unlikely(error)) |
365 | goto out; | |
366 | ||
01651646 | 367 | inode = VFS_I(cip); |
3937be5b | 368 | |
2a7dba39 | 369 | error = xfs_init_security(inode, dir, &dentry->d_name); |
3937be5b CH |
370 | if (unlikely(error)) |
371 | goto out_cleanup_inode; | |
372 | ||
2b3d1d41 CH |
373 | xfs_setup_iops(cip); |
374 | ||
3937be5b | 375 | d_instantiate(dentry, inode); |
58c90473 | 376 | xfs_finish_inode_setup(cip); |
3937be5b CH |
377 | return 0; |
378 | ||
379 | out_cleanup_inode: | |
58c90473 | 380 | xfs_finish_inode_setup(cip); |
8f112e3b | 381 | xfs_cleanup_inode(dir, inode, dentry); |
d540e43b | 382 | iput(inode); |
3937be5b | 383 | out: |
2451337d | 384 | return error; |
1da177e4 LT |
385 | } |
386 | ||
1da177e4 | 387 | STATIC int |
416c6d5b | 388 | xfs_vn_rename( |
1da177e4 LT |
389 | struct inode *odir, |
390 | struct dentry *odentry, | |
391 | struct inode *ndir, | |
dbe1b5ca CM |
392 | struct dentry *ndentry, |
393 | unsigned int flags) | |
1da177e4 | 394 | { |
2b0143b5 | 395 | struct inode *new_inode = d_inode(ndentry); |
d31a1825 | 396 | int omode = 0; |
556b8b16 BN |
397 | struct xfs_name oname; |
398 | struct xfs_name nname; | |
1da177e4 | 399 | |
7dcf5c3e | 400 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
dbe1b5ca CM |
401 | return -EINVAL; |
402 | ||
d31a1825 CM |
403 | /* if we are exchanging files, we need to set i_mode of both files */ |
404 | if (flags & RENAME_EXCHANGE) | |
2b0143b5 | 405 | omode = d_inode(ndentry)->i_mode; |
d31a1825 CM |
406 | |
407 | xfs_dentry_to_name(&oname, odentry, omode); | |
2b0143b5 | 408 | xfs_dentry_to_name(&nname, ndentry, d_inode(odentry)->i_mode); |
556b8b16 | 409 | |
2b0143b5 | 410 | return xfs_rename(XFS_I(odir), &oname, XFS_I(d_inode(odentry)), |
dbe1b5ca | 411 | XFS_I(ndir), &nname, |
d31a1825 | 412 | new_inode ? XFS_I(new_inode) : NULL, flags); |
1da177e4 LT |
413 | } |
414 | ||
415 | /* | |
416 | * careful here - this function can get called recursively, so | |
417 | * we need to be very careful about how much stack we use. | |
418 | * uio is kmalloced for this reason... | |
419 | */ | |
680baacb | 420 | STATIC const char * |
6b255391 | 421 | xfs_vn_get_link( |
1da177e4 | 422 | struct dentry *dentry, |
6b255391 | 423 | struct inode *inode, |
fceef393 | 424 | struct delayed_call *done) |
1da177e4 | 425 | { |
1da177e4 | 426 | char *link; |
804c83c3 | 427 | int error = -ENOMEM; |
1da177e4 | 428 | |
6b255391 AV |
429 | if (!dentry) |
430 | return ERR_PTR(-ECHILD); | |
431 | ||
f52720ca | 432 | link = kmalloc(MAXPATHLEN+1, GFP_KERNEL); |
804c83c3 CH |
433 | if (!link) |
434 | goto out_err; | |
1da177e4 | 435 | |
2b0143b5 | 436 | error = xfs_readlink(XFS_I(d_inode(dentry)), link); |
804c83c3 CH |
437 | if (unlikely(error)) |
438 | goto out_kfree; | |
1da177e4 | 439 | |
fceef393 AV |
440 | set_delayed_call(done, kfree_link, link); |
441 | return link; | |
804c83c3 CH |
442 | |
443 | out_kfree: | |
444 | kfree(link); | |
445 | out_err: | |
680baacb | 446 | return ERR_PTR(error); |
1da177e4 LT |
447 | } |
448 | ||
30ee052e CH |
449 | STATIC const char * |
450 | xfs_vn_get_link_inline( | |
451 | struct dentry *dentry, | |
452 | struct inode *inode, | |
453 | struct delayed_call *done) | |
454 | { | |
455 | ASSERT(XFS_I(inode)->i_df.if_flags & XFS_IFINLINE); | |
456 | return XFS_I(inode)->i_df.if_u1.if_data; | |
457 | } | |
458 | ||
1da177e4 | 459 | STATIC int |
416c6d5b | 460 | xfs_vn_getattr( |
c43f4087 CH |
461 | struct vfsmount *mnt, |
462 | struct dentry *dentry, | |
463 | struct kstat *stat) | |
1da177e4 | 464 | { |
2b0143b5 | 465 | struct inode *inode = d_inode(dentry); |
c43f4087 CH |
466 | struct xfs_inode *ip = XFS_I(inode); |
467 | struct xfs_mount *mp = ip->i_mount; | |
468 | ||
cca28fb8 | 469 | trace_xfs_getattr(ip); |
c43f4087 CH |
470 | |
471 | if (XFS_FORCED_SHUTDOWN(mp)) | |
b474c7ae | 472 | return -EIO; |
c43f4087 CH |
473 | |
474 | stat->size = XFS_ISIZE(ip); | |
475 | stat->dev = inode->i_sb->s_dev; | |
c19b3b05 | 476 | stat->mode = inode->i_mode; |
54d7b5c1 | 477 | stat->nlink = inode->i_nlink; |
7aab1b28 DE |
478 | stat->uid = inode->i_uid; |
479 | stat->gid = inode->i_gid; | |
c43f4087 | 480 | stat->ino = ip->i_ino; |
c43f4087 | 481 | stat->atime = inode->i_atime; |
f9581b14 CH |
482 | stat->mtime = inode->i_mtime; |
483 | stat->ctime = inode->i_ctime; | |
c43f4087 CH |
484 | stat->blocks = |
485 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
486 | ||
487 | ||
488 | switch (inode->i_mode & S_IFMT) { | |
489 | case S_IFBLK: | |
490 | case S_IFCHR: | |
491 | stat->blksize = BLKDEV_IOSIZE; | |
492 | stat->rdev = MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff, | |
493 | sysv_minor(ip->i_df.if_u2.if_rdev)); | |
494 | break; | |
495 | default: | |
71ddabb9 | 496 | if (XFS_IS_REALTIME_INODE(ip)) { |
c43f4087 CH |
497 | /* |
498 | * If the file blocks are being allocated from a | |
499 | * realtime volume, then return the inode's realtime | |
500 | * extent size or the realtime volume's extent size. | |
501 | */ | |
502 | stat->blksize = | |
503 | xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog; | |
504 | } else | |
505 | stat->blksize = xfs_preferred_iosize(mp); | |
506 | stat->rdev = 0; | |
507 | break; | |
69e23b9a | 508 | } |
c43f4087 CH |
509 | |
510 | return 0; | |
1da177e4 LT |
511 | } |
512 | ||
56c19e89 DC |
513 | static void |
514 | xfs_setattr_mode( | |
56c19e89 DC |
515 | struct xfs_inode *ip, |
516 | struct iattr *iattr) | |
517 | { | |
0c3d88df CH |
518 | struct inode *inode = VFS_I(ip); |
519 | umode_t mode = iattr->ia_mode; | |
56c19e89 | 520 | |
56c19e89 DC |
521 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); |
522 | ||
56c19e89 DC |
523 | inode->i_mode &= S_IFMT; |
524 | inode->i_mode |= mode & ~S_IFMT; | |
525 | } | |
526 | ||
52785112 | 527 | void |
c91c46c1 CH |
528 | xfs_setattr_time( |
529 | struct xfs_inode *ip, | |
530 | struct iattr *iattr) | |
531 | { | |
532 | struct inode *inode = VFS_I(ip); | |
533 | ||
534 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); | |
535 | ||
3987848c | 536 | if (iattr->ia_valid & ATTR_ATIME) |
c91c46c1 | 537 | inode->i_atime = iattr->ia_atime; |
3987848c | 538 | if (iattr->ia_valid & ATTR_CTIME) |
c91c46c1 | 539 | inode->i_ctime = iattr->ia_ctime; |
3987848c | 540 | if (iattr->ia_valid & ATTR_MTIME) |
c91c46c1 | 541 | inode->i_mtime = iattr->ia_mtime; |
c91c46c1 CH |
542 | } |
543 | ||
c4ed4243 CH |
544 | int |
545 | xfs_setattr_nonsize( | |
546 | struct xfs_inode *ip, | |
547 | struct iattr *iattr, | |
548 | int flags) | |
549 | { | |
550 | xfs_mount_t *mp = ip->i_mount; | |
551 | struct inode *inode = VFS_I(ip); | |
552 | int mask = iattr->ia_valid; | |
553 | xfs_trans_t *tp; | |
554 | int error; | |
7aab1b28 DE |
555 | kuid_t uid = GLOBAL_ROOT_UID, iuid = GLOBAL_ROOT_UID; |
556 | kgid_t gid = GLOBAL_ROOT_GID, igid = GLOBAL_ROOT_GID; | |
c4ed4243 CH |
557 | struct xfs_dquot *udqp = NULL, *gdqp = NULL; |
558 | struct xfs_dquot *olddquot1 = NULL, *olddquot2 = NULL; | |
559 | ||
560 | trace_xfs_setattr(ip); | |
561 | ||
42c49d7f CM |
562 | /* If acls are being inherited, we already have this checked */ |
563 | if (!(flags & XFS_ATTR_NOACL)) { | |
564 | if (mp->m_flags & XFS_MOUNT_RDONLY) | |
2451337d | 565 | return -EROFS; |
c4ed4243 | 566 | |
42c49d7f | 567 | if (XFS_FORCED_SHUTDOWN(mp)) |
2451337d | 568 | return -EIO; |
c4ed4243 | 569 | |
2451337d | 570 | error = inode_change_ok(inode, iattr); |
42c49d7f | 571 | if (error) |
b474c7ae | 572 | return error; |
42c49d7f | 573 | } |
c4ed4243 CH |
574 | |
575 | ASSERT((mask & ATTR_SIZE) == 0); | |
576 | ||
577 | /* | |
578 | * If disk quotas is on, we make sure that the dquots do exist on disk, | |
579 | * before we start any other transactions. Trying to do this later | |
580 | * is messy. We don't care to take a readlock to look at the ids | |
581 | * in inode here, because we can't hold it across the trans_reserve. | |
582 | * If the IDs do change before we take the ilock, we're covered | |
583 | * because the i_*dquot fields will get updated anyway. | |
584 | */ | |
585 | if (XFS_IS_QUOTA_ON(mp) && (mask & (ATTR_UID|ATTR_GID))) { | |
586 | uint qflags = 0; | |
587 | ||
588 | if ((mask & ATTR_UID) && XFS_IS_UQUOTA_ON(mp)) { | |
589 | uid = iattr->ia_uid; | |
590 | qflags |= XFS_QMOPT_UQUOTA; | |
591 | } else { | |
7aab1b28 | 592 | uid = inode->i_uid; |
c4ed4243 CH |
593 | } |
594 | if ((mask & ATTR_GID) && XFS_IS_GQUOTA_ON(mp)) { | |
595 | gid = iattr->ia_gid; | |
596 | qflags |= XFS_QMOPT_GQUOTA; | |
597 | } else { | |
7aab1b28 | 598 | gid = inode->i_gid; |
c4ed4243 CH |
599 | } |
600 | ||
601 | /* | |
602 | * We take a reference when we initialize udqp and gdqp, | |
603 | * so it is important that we never blindly double trip on | |
604 | * the same variable. See xfs_create() for an example. | |
605 | */ | |
606 | ASSERT(udqp == NULL); | |
607 | ASSERT(gdqp == NULL); | |
7aab1b28 DE |
608 | error = xfs_qm_vop_dqalloc(ip, xfs_kuid_to_uid(uid), |
609 | xfs_kgid_to_gid(gid), | |
610 | xfs_get_projid(ip), | |
611 | qflags, &udqp, &gdqp, NULL); | |
c4ed4243 CH |
612 | if (error) |
613 | return error; | |
614 | } | |
615 | ||
253f4911 | 616 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp); |
c4ed4243 | 617 | if (error) |
253f4911 | 618 | goto out_dqrele; |
c4ed4243 CH |
619 | |
620 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
253f4911 | 621 | xfs_trans_ijoin(tp, ip, 0); |
c4ed4243 CH |
622 | |
623 | /* | |
624 | * Change file ownership. Must be the owner or privileged. | |
625 | */ | |
626 | if (mask & (ATTR_UID|ATTR_GID)) { | |
627 | /* | |
628 | * These IDs could have changed since we last looked at them. | |
629 | * But, we're assured that if the ownership did change | |
630 | * while we didn't have the inode locked, inode's dquot(s) | |
631 | * would have changed also. | |
632 | */ | |
7aab1b28 DE |
633 | iuid = inode->i_uid; |
634 | igid = inode->i_gid; | |
c4ed4243 CH |
635 | gid = (mask & ATTR_GID) ? iattr->ia_gid : igid; |
636 | uid = (mask & ATTR_UID) ? iattr->ia_uid : iuid; | |
637 | ||
638 | /* | |
639 | * Do a quota reservation only if uid/gid is actually | |
640 | * going to change. | |
641 | */ | |
642 | if (XFS_IS_QUOTA_RUNNING(mp) && | |
7aab1b28 DE |
643 | ((XFS_IS_UQUOTA_ON(mp) && !uid_eq(iuid, uid)) || |
644 | (XFS_IS_GQUOTA_ON(mp) && !gid_eq(igid, gid)))) { | |
c4ed4243 CH |
645 | ASSERT(tp); |
646 | error = xfs_qm_vop_chown_reserve(tp, ip, udqp, gdqp, | |
92f8ff73 | 647 | NULL, capable(CAP_FOWNER) ? |
c4ed4243 CH |
648 | XFS_QMOPT_FORCE_RES : 0); |
649 | if (error) /* out of quota */ | |
253f4911 | 650 | goto out_cancel; |
c4ed4243 CH |
651 | } |
652 | } | |
653 | ||
c4ed4243 CH |
654 | /* |
655 | * Change file ownership. Must be the owner or privileged. | |
656 | */ | |
657 | if (mask & (ATTR_UID|ATTR_GID)) { | |
658 | /* | |
659 | * CAP_FSETID overrides the following restrictions: | |
660 | * | |
661 | * The set-user-ID and set-group-ID bits of a file will be | |
662 | * cleared upon successful return from chown() | |
663 | */ | |
c19b3b05 | 664 | if ((inode->i_mode & (S_ISUID|S_ISGID)) && |
c4ed4243 | 665 | !capable(CAP_FSETID)) |
c19b3b05 | 666 | inode->i_mode &= ~(S_ISUID|S_ISGID); |
c4ed4243 CH |
667 | |
668 | /* | |
669 | * Change the ownerships and register quota modifications | |
670 | * in the transaction. | |
671 | */ | |
7aab1b28 | 672 | if (!uid_eq(iuid, uid)) { |
c4ed4243 CH |
673 | if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_UQUOTA_ON(mp)) { |
674 | ASSERT(mask & ATTR_UID); | |
675 | ASSERT(udqp); | |
676 | olddquot1 = xfs_qm_vop_chown(tp, ip, | |
677 | &ip->i_udquot, udqp); | |
678 | } | |
7aab1b28 | 679 | ip->i_d.di_uid = xfs_kuid_to_uid(uid); |
c4ed4243 CH |
680 | inode->i_uid = uid; |
681 | } | |
7aab1b28 | 682 | if (!gid_eq(igid, gid)) { |
c4ed4243 | 683 | if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_GQUOTA_ON(mp)) { |
5a01dd54 JL |
684 | ASSERT(xfs_sb_version_has_pquotino(&mp->m_sb) || |
685 | !XFS_IS_PQUOTA_ON(mp)); | |
c4ed4243 CH |
686 | ASSERT(mask & ATTR_GID); |
687 | ASSERT(gdqp); | |
688 | olddquot2 = xfs_qm_vop_chown(tp, ip, | |
689 | &ip->i_gdquot, gdqp); | |
690 | } | |
7aab1b28 | 691 | ip->i_d.di_gid = xfs_kgid_to_gid(gid); |
c4ed4243 CH |
692 | inode->i_gid = gid; |
693 | } | |
694 | } | |
695 | ||
56c19e89 | 696 | if (mask & ATTR_MODE) |
0c3d88df | 697 | xfs_setattr_mode(ip, iattr); |
c91c46c1 CH |
698 | if (mask & (ATTR_ATIME|ATTR_CTIME|ATTR_MTIME)) |
699 | xfs_setattr_time(ip, iattr); | |
c4ed4243 CH |
700 | |
701 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); | |
702 | ||
ff6d6af2 | 703 | XFS_STATS_INC(mp, xs_ig_attrchg); |
c4ed4243 CH |
704 | |
705 | if (mp->m_flags & XFS_MOUNT_WSYNC) | |
706 | xfs_trans_set_sync(tp); | |
70393313 | 707 | error = xfs_trans_commit(tp); |
c4ed4243 CH |
708 | |
709 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
710 | ||
711 | /* | |
712 | * Release any dquot(s) the inode had kept before chown. | |
713 | */ | |
714 | xfs_qm_dqrele(olddquot1); | |
715 | xfs_qm_dqrele(olddquot2); | |
716 | xfs_qm_dqrele(udqp); | |
717 | xfs_qm_dqrele(gdqp); | |
718 | ||
719 | if (error) | |
b474c7ae | 720 | return error; |
c4ed4243 CH |
721 | |
722 | /* | |
723 | * XXX(hch): Updating the ACL entries is not atomic vs the i_mode | |
724 | * update. We could avoid this with linked transactions | |
725 | * and passing down the transaction pointer all the way | |
726 | * to attr_set. No previous user of the generic | |
727 | * Posix ACL code seems to care about this issue either. | |
728 | */ | |
729 | if ((mask & ATTR_MODE) && !(flags & XFS_ATTR_NOACL)) { | |
2451337d | 730 | error = posix_acl_chmod(inode, inode->i_mode); |
c4ed4243 | 731 | if (error) |
b474c7ae | 732 | return error; |
c4ed4243 CH |
733 | } |
734 | ||
735 | return 0; | |
736 | ||
253f4911 | 737 | out_cancel: |
4906e215 | 738 | xfs_trans_cancel(tp); |
253f4911 | 739 | out_dqrele: |
c4ed4243 CH |
740 | xfs_qm_dqrele(udqp); |
741 | xfs_qm_dqrele(gdqp); | |
742 | return error; | |
743 | } | |
744 | ||
745 | /* | |
746 | * Truncate file. Must have write permission and not be a directory. | |
747 | */ | |
748 | int | |
749 | xfs_setattr_size( | |
750 | struct xfs_inode *ip, | |
76ca4c23 | 751 | struct iattr *iattr) |
c4ed4243 CH |
752 | { |
753 | struct xfs_mount *mp = ip->i_mount; | |
754 | struct inode *inode = VFS_I(ip); | |
673e8e59 | 755 | xfs_off_t oldsize, newsize; |
c4ed4243 CH |
756 | struct xfs_trans *tp; |
757 | int error; | |
f38996f5 | 758 | uint lock_flags = 0; |
5885ebda | 759 | bool did_zeroing = false; |
c4ed4243 CH |
760 | |
761 | trace_xfs_setattr(ip); | |
762 | ||
763 | if (mp->m_flags & XFS_MOUNT_RDONLY) | |
2451337d | 764 | return -EROFS; |
c4ed4243 CH |
765 | |
766 | if (XFS_FORCED_SHUTDOWN(mp)) | |
2451337d | 767 | return -EIO; |
c4ed4243 | 768 | |
2451337d | 769 | error = inode_change_ok(inode, iattr); |
c4ed4243 | 770 | if (error) |
b474c7ae | 771 | return error; |
c4ed4243 | 772 | |
76ca4c23 | 773 | ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL)); |
e8e9ad42 | 774 | ASSERT(xfs_isilocked(ip, XFS_MMAPLOCK_EXCL)); |
c19b3b05 | 775 | ASSERT(S_ISREG(inode->i_mode)); |
fe60a8a0 CH |
776 | ASSERT((iattr->ia_valid & (ATTR_UID|ATTR_GID|ATTR_ATIME|ATTR_ATIME_SET| |
777 | ATTR_MTIME_SET|ATTR_KILL_PRIV|ATTR_TIMES_SET)) == 0); | |
c4ed4243 | 778 | |
ce7ae151 | 779 | oldsize = inode->i_size; |
673e8e59 CH |
780 | newsize = iattr->ia_size; |
781 | ||
c4ed4243 CH |
782 | /* |
783 | * Short circuit the truncate case for zero length files. | |
784 | */ | |
673e8e59 | 785 | if (newsize == 0 && oldsize == 0 && ip->i_d.di_nextents == 0) { |
fe60a8a0 | 786 | if (!(iattr->ia_valid & (ATTR_CTIME|ATTR_MTIME))) |
76ca4c23 | 787 | return 0; |
681b1200 CH |
788 | |
789 | /* | |
790 | * Use the regular setattr path to update the timestamps. | |
791 | */ | |
681b1200 CH |
792 | iattr->ia_valid &= ~ATTR_SIZE; |
793 | return xfs_setattr_nonsize(ip, iattr, 0); | |
c4ed4243 CH |
794 | } |
795 | ||
796 | /* | |
797 | * Make sure that the dquots are attached to the inode. | |
798 | */ | |
f38996f5 | 799 | error = xfs_qm_dqattach(ip, 0); |
c4ed4243 | 800 | if (error) |
76ca4c23 | 801 | return error; |
c4ed4243 CH |
802 | |
803 | /* | |
5885ebda DC |
804 | * File data changes must be complete before we start the transaction to |
805 | * modify the inode. This needs to be done before joining the inode to | |
806 | * the transaction because the inode cannot be unlocked once it is a | |
807 | * part of the transaction. | |
808 | * | |
809 | * Start with zeroing any data block beyond EOF that we may expose on | |
810 | * file extension. | |
c4ed4243 | 811 | */ |
673e8e59 | 812 | if (newsize > oldsize) { |
5885ebda | 813 | error = xfs_zero_eof(ip, newsize, oldsize, &did_zeroing); |
c4ed4243 | 814 | if (error) |
76ca4c23 | 815 | return error; |
c4ed4243 | 816 | } |
c4ed4243 CH |
817 | |
818 | /* | |
819 | * We are going to log the inode size change in this transaction so | |
820 | * any previous writes that are beyond the on disk EOF and the new | |
821 | * EOF that have not been written out need to be written here. If we | |
822 | * do not write the data out, we expose ourselves to the null files | |
5885ebda DC |
823 | * problem. Note that this includes any block zeroing we did above; |
824 | * otherwise those blocks may not be zeroed after a crash. | |
c4ed4243 | 825 | */ |
5885ebda DC |
826 | if (newsize > ip->i_d.di_size && |
827 | (oldsize != ip->i_d.di_size || did_zeroing)) { | |
2451337d | 828 | error = filemap_write_and_wait_range(VFS_I(ip)->i_mapping, |
4bc1ea6b | 829 | ip->i_d.di_size, newsize); |
c4ed4243 | 830 | if (error) |
76ca4c23 | 831 | return error; |
c4ed4243 CH |
832 | } |
833 | ||
5885ebda | 834 | /* Now wait for all direct I/O to complete. */ |
4a06fd26 | 835 | inode_dio_wait(inode); |
c4ed4243 | 836 | |
49abc3a8 | 837 | /* |
0f9160b4 DC |
838 | * We've already locked out new page faults, so now we can safely remove |
839 | * pages from the page cache knowing they won't get refaulted until we | |
840 | * drop the XFS_MMAP_EXCL lock after the extent manipulations are | |
841 | * complete. The truncate_setsize() call also cleans partial EOF page | |
842 | * PTEs on extending truncates and hence ensures sub-page block size | |
843 | * filesystems are correctly handled, too. | |
49abc3a8 | 844 | * |
0f9160b4 DC |
845 | * We have to do all the page cache truncate work outside the |
846 | * transaction context as the "lock" order is page lock->log space | |
847 | * reservation as defined by extent allocation in the writeback path. | |
253f4911 | 848 | * Hence a truncate can fail with ENOMEM from xfs_trans_alloc(), but |
0f9160b4 DC |
849 | * having already truncated the in-memory version of the file (i.e. made |
850 | * user visible changes). There's not much we can do about this, except | |
851 | * to hope that the caller sees ENOMEM and retries the truncate | |
852 | * operation. | |
49abc3a8 | 853 | */ |
9969441f DC |
854 | if (IS_DAX(inode)) |
855 | error = dax_truncate_page(inode, newsize, xfs_get_blocks_direct); | |
856 | else | |
857 | error = block_truncate_page(inode->i_mapping, newsize, | |
858 | xfs_get_blocks); | |
c4ed4243 | 859 | if (error) |
76ca4c23 | 860 | return error; |
49abc3a8 | 861 | truncate_setsize(inode, newsize); |
c4ed4243 | 862 | |
253f4911 | 863 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, &tp); |
c4ed4243 | 864 | if (error) |
253f4911 | 865 | return error; |
c4ed4243 | 866 | |
c4ed4243 | 867 | lock_flags |= XFS_ILOCK_EXCL; |
c4ed4243 | 868 | xfs_ilock(ip, XFS_ILOCK_EXCL); |
ddc3415a | 869 | xfs_trans_ijoin(tp, ip, 0); |
c4ed4243 CH |
870 | |
871 | /* | |
872 | * Only change the c/mtime if we are changing the size or we are | |
873 | * explicitly asked to change it. This handles the semantic difference | |
874 | * between truncate() and ftruncate() as implemented in the VFS. | |
875 | * | |
876 | * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a | |
877 | * special case where we need to update the times despite not having | |
878 | * these flags set. For all other operations the VFS set these flags | |
879 | * explicitly if it wants a timestamp update. | |
880 | */ | |
fe60a8a0 CH |
881 | if (newsize != oldsize && |
882 | !(iattr->ia_valid & (ATTR_CTIME | ATTR_MTIME))) { | |
c4ed4243 CH |
883 | iattr->ia_ctime = iattr->ia_mtime = |
884 | current_fs_time(inode->i_sb); | |
fe60a8a0 | 885 | iattr->ia_valid |= ATTR_CTIME | ATTR_MTIME; |
c4ed4243 CH |
886 | } |
887 | ||
673e8e59 CH |
888 | /* |
889 | * The first thing we do is set the size to new_size permanently on | |
890 | * disk. This way we don't have to worry about anyone ever being able | |
891 | * to look at the data being freed even in the face of a crash. | |
892 | * What we're getting around here is the case where we free a block, it | |
893 | * is allocated to another file, it is written to, and then we crash. | |
894 | * If the new data gets written to the file but the log buffers | |
895 | * containing the free and reallocation don't, then we'd end up with | |
896 | * garbage in the blocks being freed. As long as we make the new size | |
897 | * permanent before actually freeing any blocks it doesn't matter if | |
898 | * they get written to. | |
899 | */ | |
900 | ip->i_d.di_size = newsize; | |
673e8e59 CH |
901 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); |
902 | ||
903 | if (newsize <= oldsize) { | |
904 | error = xfs_itruncate_extents(&tp, ip, XFS_DATA_FORK, newsize); | |
c4ed4243 | 905 | if (error) |
4906e215 | 906 | goto out_trans_cancel; |
c4ed4243 CH |
907 | |
908 | /* | |
909 | * Truncated "down", so we're removing references to old data | |
910 | * here - if we delay flushing for a long time, we expose | |
911 | * ourselves unduly to the notorious NULL files problem. So, | |
912 | * we mark this inode and flush it when the file is closed, | |
913 | * and do not wait the usual (long) time for writeout. | |
914 | */ | |
915 | xfs_iflags_set(ip, XFS_ITRUNCATED); | |
27b52867 BF |
916 | |
917 | /* A truncate down always removes post-EOF blocks. */ | |
918 | xfs_inode_clear_eofblocks_tag(ip); | |
c4ed4243 CH |
919 | } |
920 | ||
fe60a8a0 | 921 | if (iattr->ia_valid & ATTR_MODE) |
0c3d88df | 922 | xfs_setattr_mode(ip, iattr); |
fe60a8a0 | 923 | if (iattr->ia_valid & (ATTR_ATIME|ATTR_CTIME|ATTR_MTIME)) |
c91c46c1 | 924 | xfs_setattr_time(ip, iattr); |
c4ed4243 CH |
925 | |
926 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); | |
927 | ||
ff6d6af2 | 928 | XFS_STATS_INC(mp, xs_ig_attrchg); |
c4ed4243 CH |
929 | |
930 | if (mp->m_flags & XFS_MOUNT_WSYNC) | |
931 | xfs_trans_set_sync(tp); | |
932 | ||
70393313 | 933 | error = xfs_trans_commit(tp); |
c4ed4243 CH |
934 | out_unlock: |
935 | if (lock_flags) | |
936 | xfs_iunlock(ip, lock_flags); | |
937 | return error; | |
938 | ||
c4ed4243 | 939 | out_trans_cancel: |
4906e215 | 940 | xfs_trans_cancel(tp); |
c4ed4243 CH |
941 | goto out_unlock; |
942 | } | |
943 | ||
1da177e4 | 944 | STATIC int |
416c6d5b | 945 | xfs_vn_setattr( |
76ca4c23 CH |
946 | struct dentry *dentry, |
947 | struct iattr *iattr) | |
1da177e4 | 948 | { |
2b0143b5 | 949 | struct xfs_inode *ip = XFS_I(d_inode(dentry)); |
76ca4c23 CH |
950 | int error; |
951 | ||
952 | if (iattr->ia_valid & ATTR_SIZE) { | |
781355c6 CH |
953 | uint iolock = XFS_IOLOCK_EXCL; |
954 | ||
955 | xfs_ilock(ip, iolock); | |
9ec3a646 | 956 | error = xfs_break_layouts(d_inode(dentry), &iolock, true); |
e8e9ad42 DC |
957 | if (!error) { |
958 | xfs_ilock(ip, XFS_MMAPLOCK_EXCL); | |
959 | iolock |= XFS_MMAPLOCK_EXCL; | |
960 | ||
781355c6 | 961 | error = xfs_setattr_size(ip, iattr); |
e8e9ad42 | 962 | } |
781355c6 | 963 | xfs_iunlock(ip, iolock); |
76ca4c23 CH |
964 | } else { |
965 | error = xfs_setattr_nonsize(ip, iattr, 0); | |
966 | } | |
967 | ||
2451337d | 968 | return error; |
1da177e4 LT |
969 | } |
970 | ||
69ff2826 CH |
971 | STATIC int |
972 | xfs_vn_update_time( | |
973 | struct inode *inode, | |
974 | struct timespec *now, | |
975 | int flags) | |
976 | { | |
977 | struct xfs_inode *ip = XFS_I(inode); | |
978 | struct xfs_mount *mp = ip->i_mount; | |
979 | struct xfs_trans *tp; | |
980 | int error; | |
981 | ||
982 | trace_xfs_update_time(ip); | |
983 | ||
253f4911 CH |
984 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_fsyncts, 0, 0, 0, &tp); |
985 | if (error) | |
2451337d | 986 | return error; |
69ff2826 CH |
987 | |
988 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
3987848c | 989 | if (flags & S_CTIME) |
69ff2826 | 990 | inode->i_ctime = *now; |
3987848c | 991 | if (flags & S_MTIME) |
69ff2826 | 992 | inode->i_mtime = *now; |
3987848c | 993 | if (flags & S_ATIME) |
69ff2826 | 994 | inode->i_atime = *now; |
3987848c | 995 | |
69ff2826 CH |
996 | xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); |
997 | xfs_trans_log_inode(tp, ip, XFS_ILOG_TIMESTAMP); | |
70393313 | 998 | return xfs_trans_commit(tp); |
69ff2826 CH |
999 | } |
1000 | ||
f35642e2 ES |
1001 | #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR) |
1002 | ||
1003 | /* | |
1004 | * Call fiemap helper to fill in user data. | |
1005 | * Returns positive errors to xfs_getbmap. | |
1006 | */ | |
1007 | STATIC int | |
1008 | xfs_fiemap_format( | |
1009 | void **arg, | |
1010 | struct getbmapx *bmv, | |
1011 | int *full) | |
1012 | { | |
1013 | int error; | |
1014 | struct fiemap_extent_info *fieinfo = *arg; | |
1015 | u32 fiemap_flags = 0; | |
1016 | u64 logical, physical, length; | |
1017 | ||
1018 | /* Do nothing for a hole */ | |
1019 | if (bmv->bmv_block == -1LL) | |
1020 | return 0; | |
1021 | ||
1022 | logical = BBTOB(bmv->bmv_offset); | |
1023 | physical = BBTOB(bmv->bmv_block); | |
1024 | length = BBTOB(bmv->bmv_length); | |
1025 | ||
1026 | if (bmv->bmv_oflags & BMV_OF_PREALLOC) | |
1027 | fiemap_flags |= FIEMAP_EXTENT_UNWRITTEN; | |
1028 | else if (bmv->bmv_oflags & BMV_OF_DELALLOC) { | |
635c4d0b JL |
1029 | fiemap_flags |= (FIEMAP_EXTENT_DELALLOC | |
1030 | FIEMAP_EXTENT_UNKNOWN); | |
f35642e2 ES |
1031 | physical = 0; /* no block yet */ |
1032 | } | |
1033 | if (bmv->bmv_oflags & BMV_OF_LAST) | |
1034 | fiemap_flags |= FIEMAP_EXTENT_LAST; | |
1035 | ||
1036 | error = fiemap_fill_next_extent(fieinfo, logical, physical, | |
1037 | length, fiemap_flags); | |
1038 | if (error > 0) { | |
1039 | error = 0; | |
1040 | *full = 1; /* user array now full */ | |
1041 | } | |
1042 | ||
2451337d | 1043 | return error; |
f35642e2 ES |
1044 | } |
1045 | ||
1046 | STATIC int | |
1047 | xfs_vn_fiemap( | |
1048 | struct inode *inode, | |
1049 | struct fiemap_extent_info *fieinfo, | |
1050 | u64 start, | |
1051 | u64 length) | |
1052 | { | |
1053 | xfs_inode_t *ip = XFS_I(inode); | |
1054 | struct getbmapx bm; | |
1055 | int error; | |
1056 | ||
1057 | error = fiemap_check_flags(fieinfo, XFS_FIEMAP_FLAGS); | |
1058 | if (error) | |
1059 | return error; | |
1060 | ||
1061 | /* Set up bmap header for xfs internal routine */ | |
eedf32bf | 1062 | bm.bmv_offset = BTOBBT(start); |
f35642e2 ES |
1063 | /* Special case for whole file */ |
1064 | if (length == FIEMAP_MAX_OFFSET) | |
1065 | bm.bmv_length = -1LL; | |
1066 | else | |
eedf32bf | 1067 | bm.bmv_length = BTOBB(start + length) - bm.bmv_offset; |
f35642e2 | 1068 | |
97db39a1 | 1069 | /* We add one because in getbmap world count includes the header */ |
2d1ff3c7 TM |
1070 | bm.bmv_count = !fieinfo->fi_extents_max ? MAXEXTNUM : |
1071 | fieinfo->fi_extents_max + 1; | |
1072 | bm.bmv_count = min_t(__s32, bm.bmv_count, | |
1073 | (PAGE_SIZE * 16 / sizeof(struct getbmapx))); | |
9af25465 | 1074 | bm.bmv_iflags = BMV_IF_PREALLOC | BMV_IF_NO_HOLES; |
f35642e2 ES |
1075 | if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) |
1076 | bm.bmv_iflags |= BMV_IF_ATTRFORK; | |
1077 | if (!(fieinfo->fi_flags & FIEMAP_FLAG_SYNC)) | |
1078 | bm.bmv_iflags |= BMV_IF_DELALLOC; | |
1079 | ||
1080 | error = xfs_getbmap(ip, &bm, xfs_fiemap_format, fieinfo); | |
1081 | if (error) | |
2451337d | 1082 | return error; |
f35642e2 ES |
1083 | |
1084 | return 0; | |
1085 | } | |
1086 | ||
99b6436b ZYW |
1087 | STATIC int |
1088 | xfs_vn_tmpfile( | |
1089 | struct inode *dir, | |
1090 | struct dentry *dentry, | |
1091 | umode_t mode) | |
1092 | { | |
d540e43b | 1093 | return xfs_generic_create(dir, dentry, mode, 0, true); |
99b6436b ZYW |
1094 | } |
1095 | ||
41be8bed | 1096 | static const struct inode_operations xfs_inode_operations = { |
4e34e719 | 1097 | .get_acl = xfs_get_acl, |
2401dc29 | 1098 | .set_acl = xfs_set_acl, |
416c6d5b NS |
1099 | .getattr = xfs_vn_getattr, |
1100 | .setattr = xfs_vn_setattr, | |
0ec58516 LM |
1101 | .setxattr = generic_setxattr, |
1102 | .getxattr = generic_getxattr, | |
1103 | .removexattr = generic_removexattr, | |
416c6d5b | 1104 | .listxattr = xfs_vn_listxattr, |
f35642e2 | 1105 | .fiemap = xfs_vn_fiemap, |
69ff2826 | 1106 | .update_time = xfs_vn_update_time, |
1da177e4 LT |
1107 | }; |
1108 | ||
41be8bed | 1109 | static const struct inode_operations xfs_dir_inode_operations = { |
416c6d5b NS |
1110 | .create = xfs_vn_create, |
1111 | .lookup = xfs_vn_lookup, | |
1112 | .link = xfs_vn_link, | |
1113 | .unlink = xfs_vn_unlink, | |
1114 | .symlink = xfs_vn_symlink, | |
1115 | .mkdir = xfs_vn_mkdir, | |
8f112e3b CH |
1116 | /* |
1117 | * Yes, XFS uses the same method for rmdir and unlink. | |
1118 | * | |
1119 | * There are some subtile differences deeper in the code, | |
1120 | * but we use S_ISDIR to check for those. | |
1121 | */ | |
1122 | .rmdir = xfs_vn_unlink, | |
416c6d5b | 1123 | .mknod = xfs_vn_mknod, |
dbe1b5ca | 1124 | .rename2 = xfs_vn_rename, |
4e34e719 | 1125 | .get_acl = xfs_get_acl, |
2401dc29 | 1126 | .set_acl = xfs_set_acl, |
416c6d5b NS |
1127 | .getattr = xfs_vn_getattr, |
1128 | .setattr = xfs_vn_setattr, | |
0ec58516 LM |
1129 | .setxattr = generic_setxattr, |
1130 | .getxattr = generic_getxattr, | |
1131 | .removexattr = generic_removexattr, | |
416c6d5b | 1132 | .listxattr = xfs_vn_listxattr, |
69ff2826 | 1133 | .update_time = xfs_vn_update_time, |
99b6436b | 1134 | .tmpfile = xfs_vn_tmpfile, |
1da177e4 LT |
1135 | }; |
1136 | ||
41be8bed | 1137 | static const struct inode_operations xfs_dir_ci_inode_operations = { |
384f3ced BN |
1138 | .create = xfs_vn_create, |
1139 | .lookup = xfs_vn_ci_lookup, | |
1140 | .link = xfs_vn_link, | |
1141 | .unlink = xfs_vn_unlink, | |
1142 | .symlink = xfs_vn_symlink, | |
1143 | .mkdir = xfs_vn_mkdir, | |
8f112e3b CH |
1144 | /* |
1145 | * Yes, XFS uses the same method for rmdir and unlink. | |
1146 | * | |
1147 | * There are some subtile differences deeper in the code, | |
1148 | * but we use S_ISDIR to check for those. | |
1149 | */ | |
1150 | .rmdir = xfs_vn_unlink, | |
384f3ced | 1151 | .mknod = xfs_vn_mknod, |
dbe1b5ca | 1152 | .rename2 = xfs_vn_rename, |
4e34e719 | 1153 | .get_acl = xfs_get_acl, |
2401dc29 | 1154 | .set_acl = xfs_set_acl, |
384f3ced BN |
1155 | .getattr = xfs_vn_getattr, |
1156 | .setattr = xfs_vn_setattr, | |
0ec58516 LM |
1157 | .setxattr = generic_setxattr, |
1158 | .getxattr = generic_getxattr, | |
1159 | .removexattr = generic_removexattr, | |
384f3ced | 1160 | .listxattr = xfs_vn_listxattr, |
69ff2826 | 1161 | .update_time = xfs_vn_update_time, |
99b6436b | 1162 | .tmpfile = xfs_vn_tmpfile, |
384f3ced BN |
1163 | }; |
1164 | ||
41be8bed | 1165 | static const struct inode_operations xfs_symlink_inode_operations = { |
1da177e4 | 1166 | .readlink = generic_readlink, |
6b255391 | 1167 | .get_link = xfs_vn_get_link, |
416c6d5b NS |
1168 | .getattr = xfs_vn_getattr, |
1169 | .setattr = xfs_vn_setattr, | |
0ec58516 LM |
1170 | .setxattr = generic_setxattr, |
1171 | .getxattr = generic_getxattr, | |
1172 | .removexattr = generic_removexattr, | |
416c6d5b | 1173 | .listxattr = xfs_vn_listxattr, |
69ff2826 | 1174 | .update_time = xfs_vn_update_time, |
1da177e4 | 1175 | }; |
41be8bed | 1176 | |
30ee052e CH |
1177 | static const struct inode_operations xfs_inline_symlink_inode_operations = { |
1178 | .readlink = generic_readlink, | |
1179 | .get_link = xfs_vn_get_link_inline, | |
1180 | .getattr = xfs_vn_getattr, | |
1181 | .setattr = xfs_vn_setattr, | |
1182 | .setxattr = generic_setxattr, | |
1183 | .getxattr = generic_getxattr, | |
1184 | .removexattr = generic_removexattr, | |
1185 | .listxattr = xfs_vn_listxattr, | |
1186 | .update_time = xfs_vn_update_time, | |
1187 | }; | |
1188 | ||
41be8bed CH |
1189 | STATIC void |
1190 | xfs_diflags_to_iflags( | |
1191 | struct inode *inode, | |
1192 | struct xfs_inode *ip) | |
1193 | { | |
cbe4dab1 DC |
1194 | uint16_t flags = ip->i_d.di_flags; |
1195 | ||
1196 | inode->i_flags &= ~(S_IMMUTABLE | S_APPEND | S_SYNC | | |
1197 | S_NOATIME | S_DAX); | |
1198 | ||
1199 | if (flags & XFS_DIFLAG_IMMUTABLE) | |
41be8bed | 1200 | inode->i_flags |= S_IMMUTABLE; |
cbe4dab1 | 1201 | if (flags & XFS_DIFLAG_APPEND) |
41be8bed | 1202 | inode->i_flags |= S_APPEND; |
cbe4dab1 | 1203 | if (flags & XFS_DIFLAG_SYNC) |
41be8bed | 1204 | inode->i_flags |= S_SYNC; |
cbe4dab1 | 1205 | if (flags & XFS_DIFLAG_NOATIME) |
41be8bed | 1206 | inode->i_flags |= S_NOATIME; |
db10c697 | 1207 | if (S_ISREG(inode->i_mode) && |
64485437 | 1208 | ip->i_mount->m_sb.sb_blocksize == PAGE_SIZE && |
db10c697 DC |
1209 | (ip->i_mount->m_flags & XFS_MOUNT_DAX || |
1210 | ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) | |
cbe4dab1 | 1211 | inode->i_flags |= S_DAX; |
41be8bed CH |
1212 | } |
1213 | ||
1214 | /* | |
2b3d1d41 | 1215 | * Initialize the Linux inode. |
bf904248 | 1216 | * |
58c90473 DC |
1217 | * When reading existing inodes from disk this is called directly from xfs_iget, |
1218 | * when creating a new inode it is called from xfs_ialloc after setting up the | |
1219 | * inode. These callers have different criteria for clearing XFS_INEW, so leave | |
1220 | * it up to the caller to deal with unlocking the inode appropriately. | |
41be8bed CH |
1221 | */ |
1222 | void | |
1223 | xfs_setup_inode( | |
1224 | struct xfs_inode *ip) | |
1225 | { | |
bf904248 | 1226 | struct inode *inode = &ip->i_vnode; |
ad22c7a0 | 1227 | gfp_t gfp_mask; |
bf904248 DC |
1228 | |
1229 | inode->i_ino = ip->i_ino; | |
eaff8079 | 1230 | inode->i_state = I_NEW; |
646ec461 CH |
1231 | |
1232 | inode_sb_list_add(inode); | |
c6f6cd06 CH |
1233 | /* make the inode look hashed for the writeback code */ |
1234 | hlist_add_fake(&inode->i_hash); | |
41be8bed | 1235 | |
7aab1b28 DE |
1236 | inode->i_uid = xfs_uid_to_kuid(ip->i_d.di_uid); |
1237 | inode->i_gid = xfs_gid_to_kgid(ip->i_d.di_gid); | |
41be8bed CH |
1238 | |
1239 | switch (inode->i_mode & S_IFMT) { | |
1240 | case S_IFBLK: | |
1241 | case S_IFCHR: | |
1242 | inode->i_rdev = | |
1243 | MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff, | |
1244 | sysv_minor(ip->i_df.if_u2.if_rdev)); | |
1245 | break; | |
1246 | default: | |
1247 | inode->i_rdev = 0; | |
1248 | break; | |
1249 | } | |
1250 | ||
41be8bed | 1251 | i_size_write(inode, ip->i_d.di_size); |
41be8bed | 1252 | xfs_diflags_to_iflags(inode, ip); |
41be8bed | 1253 | |
2b3d1d41 | 1254 | if (S_ISDIR(inode->i_mode)) { |
93a8614e | 1255 | lockdep_set_class(&ip->i_lock.mr_lock, &xfs_dir_ilock_class); |
32c5483a | 1256 | ip->d_ops = ip->i_mount->m_dir_inode_ops; |
2b3d1d41 CH |
1257 | } else { |
1258 | ip->d_ops = ip->i_mount->m_nondir_inode_ops; | |
1259 | lockdep_set_class(&ip->i_lock.mr_lock, &xfs_nondir_ilock_class); | |
41be8bed CH |
1260 | } |
1261 | ||
ad22c7a0 DC |
1262 | /* |
1263 | * Ensure all page cache allocations are done from GFP_NOFS context to | |
1264 | * prevent direct reclaim recursion back into the filesystem and blowing | |
1265 | * stacks or deadlocking. | |
1266 | */ | |
1267 | gfp_mask = mapping_gfp_mask(inode->i_mapping); | |
1268 | mapping_set_gfp_mask(inode->i_mapping, (gfp_mask & ~(__GFP_FS))); | |
1269 | ||
510792ee CH |
1270 | /* |
1271 | * If there is no attribute fork no ACL can exist on this inode, | |
1272 | * and it can't have any file capabilities attached to it either. | |
1273 | */ | |
1274 | if (!XFS_IFORK_Q(ip)) { | |
1275 | inode_has_no_xattr(inode); | |
6311b108 | 1276 | cache_no_acl(inode); |
510792ee | 1277 | } |
41be8bed | 1278 | } |
2b3d1d41 CH |
1279 | |
1280 | void | |
1281 | xfs_setup_iops( | |
1282 | struct xfs_inode *ip) | |
1283 | { | |
1284 | struct inode *inode = &ip->i_vnode; | |
1285 | ||
41be8bed CH |
1286 | switch (inode->i_mode & S_IFMT) { |
1287 | case S_IFREG: | |
1288 | inode->i_op = &xfs_inode_operations; | |
1289 | inode->i_fop = &xfs_file_operations; | |
1290 | inode->i_mapping->a_ops = &xfs_address_space_operations; | |
1291 | break; | |
1292 | case S_IFDIR: | |
1293 | if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb)) | |
1294 | inode->i_op = &xfs_dir_ci_inode_operations; | |
1295 | else | |
1296 | inode->i_op = &xfs_dir_inode_operations; | |
1297 | inode->i_fop = &xfs_dir_file_operations; | |
1298 | break; | |
1299 | case S_IFLNK: | |
30ee052e CH |
1300 | if (ip->i_df.if_flags & XFS_IFINLINE) |
1301 | inode->i_op = &xfs_inline_symlink_inode_operations; | |
1302 | else | |
1303 | inode->i_op = &xfs_symlink_inode_operations; | |
41be8bed CH |
1304 | break; |
1305 | default: | |
1306 | inode->i_op = &xfs_inode_operations; | |
1307 | init_special_inode(inode, inode->i_mode, inode->i_rdev); | |
1308 | break; | |
1309 | } | |
41be8bed | 1310 | } |