Commit | Line | Data |
---|---|---|
0b61f8a4 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 | 2 | /* |
7b718769 NS |
3 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. |
4 | * All Rights Reserved. | |
1da177e4 | 5 | */ |
1da177e4 LT |
6 | #include "xfs.h" |
7 | #include "xfs_fs.h" | |
70a9883c | 8 | #include "xfs_shared.h" |
239880ef DC |
9 | #include "xfs_format.h" |
10 | #include "xfs_log_format.h" | |
11 | #include "xfs_trans_resv.h" | |
1da177e4 | 12 | #include "xfs_mount.h" |
57062787 | 13 | #include "xfs_da_format.h" |
1da177e4 LT |
14 | #include "xfs_inode.h" |
15 | #include "xfs_bmap.h" | |
68988114 | 16 | #include "xfs_bmap_util.h" |
239880ef | 17 | #include "xfs_acl.h" |
239880ef | 18 | #include "xfs_quota.h" |
1da177e4 | 19 | #include "xfs_error.h" |
1da177e4 | 20 | #include "xfs_attr.h" |
239880ef | 21 | #include "xfs_trans.h" |
0b1b213f | 22 | #include "xfs_trace.h" |
27b52867 | 23 | #include "xfs_icache.h" |
c24b5dfa | 24 | #include "xfs_symlink.h" |
0cb97766 | 25 | #include "xfs_da_btree.h" |
1b767ee3 | 26 | #include "xfs_dir2.h" |
99b6436b | 27 | #include "xfs_trans_space.h" |
68a9f5e7 | 28 | #include "xfs_iomap.h" |
02dff7bf | 29 | #include "xfs_defer.h" |
1da177e4 | 30 | |
16f7e0fe | 31 | #include <linux/capability.h> |
1da177e4 | 32 | #include <linux/xattr.h> |
ef14f0c1 | 33 | #include <linux/posix_acl.h> |
446ada4a | 34 | #include <linux/security.h> |
d2bb140e | 35 | #include <linux/iomap.h> |
5a0e3ad6 | 36 | #include <linux/slab.h> |
c3b1b131 | 37 | #include <linux/iversion.h> |
1da177e4 | 38 | |
93a8614e DC |
39 | /* |
40 | * Directories have different lock order w.r.t. mmap_sem compared to regular | |
41 | * files. This is due to readdir potentially triggering page faults on a user | |
42 | * buffer inside filldir(), and this happens with the ilock on the directory | |
43 | * held. For regular files, the lock order is the other way around - the | |
44 | * mmap_sem is taken during the page fault, and then we lock the ilock to do | |
45 | * block mapping. Hence we need a different class for the directory ilock so | |
46 | * that lockdep can tell them apart. | |
47 | */ | |
48 | static struct lock_class_key xfs_nondir_ilock_class; | |
49 | static struct lock_class_key xfs_dir_ilock_class; | |
50 | ||
8d2a5e6e DC |
51 | static int |
52 | xfs_initxattrs( | |
53 | struct inode *inode, | |
54 | const struct xattr *xattr_array, | |
55 | void *fs_info) | |
9d8f13ba | 56 | { |
8d2a5e6e DC |
57 | const struct xattr *xattr; |
58 | struct xfs_inode *ip = XFS_I(inode); | |
59 | int error = 0; | |
9d8f13ba MZ |
60 | |
61 | for (xattr = xattr_array; xattr->name != NULL; xattr++) { | |
2451337d | 62 | error = xfs_attr_set(ip, xattr->name, xattr->value, |
a5a14de2 | 63 | xattr->value_len, ATTR_SECURE); |
9d8f13ba MZ |
64 | if (error < 0) |
65 | break; | |
66 | } | |
67 | return error; | |
68 | } | |
69 | ||
446ada4a NS |
70 | /* |
71 | * Hook in SELinux. This is not quite correct yet, what we really need | |
72 | * here (as we do for default ACLs) is a mechanism by which creation of | |
73 | * these attrs can be journalled at inode creation time (along with the | |
74 | * inode, of course, such that log replay can't cause these to be lost). | |
75 | */ | |
9d8f13ba | 76 | |
446ada4a | 77 | STATIC int |
416c6d5b | 78 | xfs_init_security( |
af048193 | 79 | struct inode *inode, |
2a7dba39 EP |
80 | struct inode *dir, |
81 | const struct qstr *qstr) | |
446ada4a | 82 | { |
2451337d | 83 | return security_inode_init_security(inode, dir, qstr, |
a5a14de2 | 84 | &xfs_initxattrs, NULL); |
446ada4a NS |
85 | } |
86 | ||
556b8b16 BN |
87 | static void |
88 | xfs_dentry_to_name( | |
fab8eef8 AG |
89 | struct xfs_name *namep, |
90 | struct dentry *dentry) | |
91 | { | |
92 | namep->name = dentry->d_name.name; | |
93 | namep->len = dentry->d_name.len; | |
94 | namep->type = XFS_DIR3_FT_UNKNOWN; | |
95 | } | |
96 | ||
97 | static int | |
98 | xfs_dentry_mode_to_name( | |
556b8b16 | 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; | |
1fc4d33f | 105 | namep->type = xfs_mode_to_ftype(mode); |
fab8eef8 AG |
106 | |
107 | if (unlikely(namep->type == XFS_DIR3_FT_UNKNOWN)) | |
108 | return -EFSCORRUPTED; | |
109 | ||
110 | return 0; | |
556b8b16 BN |
111 | } |
112 | ||
7989cb8e | 113 | STATIC void |
416c6d5b | 114 | xfs_cleanup_inode( |
739bfb2a | 115 | struct inode *dir, |
af048193 | 116 | struct inode *inode, |
8f112e3b | 117 | struct dentry *dentry) |
3a69c7dc | 118 | { |
556b8b16 | 119 | struct xfs_name teardown; |
3a69c7dc YL |
120 | |
121 | /* Oh, the horror. | |
220b5284 | 122 | * If we can't add the ACL or we fail in |
416c6d5b | 123 | * xfs_init_security we must back out. |
3a69c7dc YL |
124 | * ENOSPC can hit here, among other things. |
125 | */ | |
fab8eef8 | 126 | xfs_dentry_to_name(&teardown, dentry); |
3a69c7dc | 127 | |
8f112e3b | 128 | xfs_remove(XFS_I(dir), &teardown, XFS_I(inode)); |
3a69c7dc YL |
129 | } |
130 | ||
1da177e4 | 131 | STATIC int |
d540e43b | 132 | xfs_generic_create( |
1da177e4 LT |
133 | struct inode *dir, |
134 | struct dentry *dentry, | |
1a67aafb | 135 | umode_t mode, |
d540e43b BF |
136 | dev_t rdev, |
137 | bool tmpfile) /* unnamed file */ | |
1da177e4 | 138 | { |
db0bb7ba | 139 | struct inode *inode; |
979ebab1 | 140 | struct xfs_inode *ip = NULL; |
2401dc29 | 141 | struct posix_acl *default_acl, *acl; |
556b8b16 | 142 | struct xfs_name name; |
1da177e4 LT |
143 | int error; |
144 | ||
145 | /* | |
146 | * Irix uses Missed'em'V split, but doesn't want to see | |
147 | * the upper 5 bits of (14bit) major. | |
148 | */ | |
517b5e8c CH |
149 | if (S_ISCHR(mode) || S_ISBLK(mode)) { |
150 | if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff)) | |
151 | return -EINVAL; | |
517b5e8c CH |
152 | } else { |
153 | rdev = 0; | |
154 | } | |
1da177e4 | 155 | |
2401dc29 CH |
156 | error = posix_acl_create(dir, &mode, &default_acl, &acl); |
157 | if (error) | |
158 | return error; | |
1da177e4 | 159 | |
fab8eef8 AG |
160 | /* Verify mode is valid also for tmpfile case */ |
161 | error = xfs_dentry_mode_to_name(&name, dentry, mode); | |
162 | if (unlikely(error)) | |
163 | goto out_free_acl; | |
164 | ||
d540e43b | 165 | if (!tmpfile) { |
d540e43b BF |
166 | error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip); |
167 | } else { | |
a1f69417 | 168 | error = xfs_create_tmpfile(XFS_I(dir), mode, &ip); |
d540e43b | 169 | } |
db0bb7ba CH |
170 | if (unlikely(error)) |
171 | goto out_free_acl; | |
446ada4a | 172 | |
01651646 | 173 | inode = VFS_I(ip); |
979ebab1 | 174 | |
2a7dba39 | 175 | error = xfs_init_security(inode, dir, &dentry->d_name); |
db0bb7ba CH |
176 | if (unlikely(error)) |
177 | goto out_cleanup_inode; | |
178 | ||
2401dc29 | 179 | #ifdef CONFIG_XFS_POSIX_ACL |
db0bb7ba | 180 | if (default_acl) { |
8ba35875 | 181 | error = __xfs_set_acl(inode, default_acl, ACL_TYPE_DEFAULT); |
2401dc29 | 182 | if (error) |
db0bb7ba | 183 | goto out_cleanup_inode; |
1da177e4 | 184 | } |
2401dc29 | 185 | if (acl) { |
8ba35875 | 186 | error = __xfs_set_acl(inode, acl, ACL_TYPE_ACCESS); |
2401dc29 CH |
187 | if (error) |
188 | goto out_cleanup_inode; | |
189 | } | |
190 | #endif | |
1da177e4 | 191 | |
2b3d1d41 CH |
192 | xfs_setup_iops(ip); |
193 | ||
c4a6bf7f DW |
194 | if (tmpfile) { |
195 | /* | |
196 | * The VFS requires that any inode fed to d_tmpfile must have | |
197 | * nlink == 1 so that it can decrement the nlink in d_tmpfile. | |
198 | * However, we created the temp file with nlink == 0 because | |
199 | * we're not allowed to put an inode with nlink > 0 on the | |
200 | * unlinked list. Therefore we have to set nlink to 1 so that | |
201 | * d_tmpfile can immediately set it back to zero. | |
202 | */ | |
203 | set_nlink(inode, 1); | |
d540e43b | 204 | d_tmpfile(dentry, inode); |
c4a6bf7f | 205 | } else |
d540e43b BF |
206 | d_instantiate(dentry, inode); |
207 | ||
58c90473 DC |
208 | xfs_finish_inode_setup(ip); |
209 | ||
2401dc29 CH |
210 | out_free_acl: |
211 | if (default_acl) | |
212 | posix_acl_release(default_acl); | |
213 | if (acl) | |
214 | posix_acl_release(acl); | |
2451337d | 215 | return error; |
db0bb7ba CH |
216 | |
217 | out_cleanup_inode: | |
58c90473 | 218 | xfs_finish_inode_setup(ip); |
d540e43b BF |
219 | if (!tmpfile) |
220 | xfs_cleanup_inode(dir, inode, dentry); | |
44a8736b | 221 | xfs_irele(ip); |
2401dc29 | 222 | goto out_free_acl; |
1da177e4 LT |
223 | } |
224 | ||
d540e43b BF |
225 | STATIC int |
226 | xfs_vn_mknod( | |
227 | struct inode *dir, | |
228 | struct dentry *dentry, | |
229 | umode_t mode, | |
230 | dev_t rdev) | |
231 | { | |
232 | return xfs_generic_create(dir, dentry, mode, rdev, false); | |
233 | } | |
234 | ||
1da177e4 | 235 | STATIC int |
416c6d5b | 236 | xfs_vn_create( |
1da177e4 LT |
237 | struct inode *dir, |
238 | struct dentry *dentry, | |
4acdaf27 | 239 | umode_t mode, |
ebfc3b49 | 240 | bool flags) |
1da177e4 | 241 | { |
416c6d5b | 242 | return xfs_vn_mknod(dir, dentry, mode, 0); |
1da177e4 LT |
243 | } |
244 | ||
245 | STATIC int | |
416c6d5b | 246 | xfs_vn_mkdir( |
1da177e4 LT |
247 | struct inode *dir, |
248 | struct dentry *dentry, | |
18bb1db3 | 249 | umode_t mode) |
1da177e4 | 250 | { |
416c6d5b | 251 | return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0); |
1da177e4 LT |
252 | } |
253 | ||
254 | STATIC struct dentry * | |
416c6d5b | 255 | xfs_vn_lookup( |
1da177e4 LT |
256 | struct inode *dir, |
257 | struct dentry *dentry, | |
00cd8dd3 | 258 | unsigned int flags) |
1da177e4 | 259 | { |
b113a6d3 | 260 | struct inode *inode; |
ef1f5e7a | 261 | struct xfs_inode *cip; |
556b8b16 | 262 | struct xfs_name name; |
1da177e4 LT |
263 | int error; |
264 | ||
265 | if (dentry->d_name.len >= MAXNAMELEN) | |
266 | return ERR_PTR(-ENAMETOOLONG); | |
267 | ||
fab8eef8 | 268 | xfs_dentry_to_name(&name, dentry); |
384f3ced | 269 | error = xfs_lookup(XFS_I(dir), &name, &cip, NULL); |
b113a6d3 AV |
270 | if (likely(!error)) |
271 | inode = VFS_I(cip); | |
272 | else if (likely(error == -ENOENT)) | |
273 | inode = NULL; | |
274 | else | |
275 | inode = ERR_PTR(error); | |
276 | return d_splice_alias(inode, dentry); | |
1da177e4 LT |
277 | } |
278 | ||
384f3ced BN |
279 | STATIC struct dentry * |
280 | xfs_vn_ci_lookup( | |
281 | struct inode *dir, | |
282 | struct dentry *dentry, | |
00cd8dd3 | 283 | unsigned int flags) |
384f3ced BN |
284 | { |
285 | struct xfs_inode *ip; | |
286 | struct xfs_name xname; | |
287 | struct xfs_name ci_name; | |
288 | struct qstr dname; | |
289 | int error; | |
290 | ||
291 | if (dentry->d_name.len >= MAXNAMELEN) | |
292 | return ERR_PTR(-ENAMETOOLONG); | |
293 | ||
fab8eef8 | 294 | xfs_dentry_to_name(&xname, dentry); |
384f3ced BN |
295 | error = xfs_lookup(XFS_I(dir), &xname, &ip, &ci_name); |
296 | if (unlikely(error)) { | |
2451337d DC |
297 | if (unlikely(error != -ENOENT)) |
298 | return ERR_PTR(error); | |
866d5dc9 BN |
299 | /* |
300 | * call d_add(dentry, NULL) here when d_drop_negative_children | |
301 | * is called in xfs_vn_mknod (ie. allow negative dentries | |
302 | * with CI filesystems). | |
303 | */ | |
384f3ced BN |
304 | return NULL; |
305 | } | |
306 | ||
307 | /* if exact match, just splice and exit */ | |
308 | if (!ci_name.name) | |
01651646 | 309 | return d_splice_alias(VFS_I(ip), dentry); |
384f3ced BN |
310 | |
311 | /* else case-insensitive match... */ | |
312 | dname.name = ci_name.name; | |
313 | dname.len = ci_name.len; | |
e45b590b | 314 | dentry = d_add_ci(dentry, VFS_I(ip), &dname); |
384f3ced BN |
315 | kmem_free(ci_name.name); |
316 | return dentry; | |
317 | } | |
318 | ||
1da177e4 | 319 | STATIC int |
416c6d5b | 320 | xfs_vn_link( |
1da177e4 LT |
321 | struct dentry *old_dentry, |
322 | struct inode *dir, | |
323 | struct dentry *dentry) | |
324 | { | |
2b0143b5 | 325 | struct inode *inode = d_inode(old_dentry); |
556b8b16 | 326 | struct xfs_name name; |
1da177e4 LT |
327 | int error; |
328 | ||
fab8eef8 AG |
329 | error = xfs_dentry_mode_to_name(&name, dentry, inode->i_mode); |
330 | if (unlikely(error)) | |
331 | return error; | |
1da177e4 | 332 | |
556b8b16 | 333 | error = xfs_link(XFS_I(dir), XFS_I(inode), &name); |
d9424b3c | 334 | if (unlikely(error)) |
2451337d | 335 | return error; |
a3da7896 | 336 | |
7de9c6ee | 337 | ihold(inode); |
a3da7896 CH |
338 | d_instantiate(dentry, inode); |
339 | return 0; | |
1da177e4 LT |
340 | } |
341 | ||
342 | STATIC int | |
416c6d5b | 343 | xfs_vn_unlink( |
1da177e4 LT |
344 | struct inode *dir, |
345 | struct dentry *dentry) | |
346 | { | |
556b8b16 | 347 | struct xfs_name name; |
1da177e4 LT |
348 | int error; |
349 | ||
fab8eef8 | 350 | xfs_dentry_to_name(&name, dentry); |
1da177e4 | 351 | |
2b0143b5 | 352 | error = xfs_remove(XFS_I(dir), &name, XFS_I(d_inode(dentry))); |
e5700704 CH |
353 | if (error) |
354 | return error; | |
355 | ||
356 | /* | |
357 | * With unlink, the VFS makes the dentry "negative": no inode, | |
358 | * but still hashed. This is incompatible with case-insensitive | |
359 | * mode, so invalidate (unhash) the dentry in CI-mode. | |
360 | */ | |
361 | if (xfs_sb_version_hasasciici(&XFS_M(dir->i_sb)->m_sb)) | |
362 | d_invalidate(dentry); | |
363 | return 0; | |
1da177e4 LT |
364 | } |
365 | ||
366 | STATIC int | |
416c6d5b | 367 | xfs_vn_symlink( |
1da177e4 LT |
368 | struct inode *dir, |
369 | struct dentry *dentry, | |
370 | const char *symname) | |
371 | { | |
3937be5b CH |
372 | struct inode *inode; |
373 | struct xfs_inode *cip = NULL; | |
556b8b16 | 374 | struct xfs_name name; |
1da177e4 | 375 | int error; |
576b1d67 | 376 | umode_t mode; |
1da177e4 | 377 | |
3e5daf05 | 378 | mode = S_IFLNK | |
ce3b0f8d | 379 | (irix_symlink_mode ? 0777 & ~current_umask() : S_IRWXUGO); |
fab8eef8 AG |
380 | error = xfs_dentry_mode_to_name(&name, dentry, mode); |
381 | if (unlikely(error)) | |
382 | goto out; | |
1da177e4 | 383 | |
6c77b0ea | 384 | error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip); |
3937be5b CH |
385 | if (unlikely(error)) |
386 | goto out; | |
387 | ||
01651646 | 388 | inode = VFS_I(cip); |
3937be5b | 389 | |
2a7dba39 | 390 | error = xfs_init_security(inode, dir, &dentry->d_name); |
3937be5b CH |
391 | if (unlikely(error)) |
392 | goto out_cleanup_inode; | |
393 | ||
2b3d1d41 CH |
394 | xfs_setup_iops(cip); |
395 | ||
3937be5b | 396 | d_instantiate(dentry, inode); |
58c90473 | 397 | xfs_finish_inode_setup(cip); |
3937be5b CH |
398 | return 0; |
399 | ||
400 | out_cleanup_inode: | |
58c90473 | 401 | xfs_finish_inode_setup(cip); |
8f112e3b | 402 | xfs_cleanup_inode(dir, inode, dentry); |
44a8736b | 403 | xfs_irele(cip); |
3937be5b | 404 | out: |
2451337d | 405 | return error; |
1da177e4 LT |
406 | } |
407 | ||
1da177e4 | 408 | STATIC int |
416c6d5b | 409 | xfs_vn_rename( |
1da177e4 LT |
410 | struct inode *odir, |
411 | struct dentry *odentry, | |
412 | struct inode *ndir, | |
dbe1b5ca CM |
413 | struct dentry *ndentry, |
414 | unsigned int flags) | |
1da177e4 | 415 | { |
2b0143b5 | 416 | struct inode *new_inode = d_inode(ndentry); |
d31a1825 | 417 | int omode = 0; |
fab8eef8 | 418 | int error; |
556b8b16 BN |
419 | struct xfs_name oname; |
420 | struct xfs_name nname; | |
1da177e4 | 421 | |
7dcf5c3e | 422 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
dbe1b5ca CM |
423 | return -EINVAL; |
424 | ||
d31a1825 CM |
425 | /* if we are exchanging files, we need to set i_mode of both files */ |
426 | if (flags & RENAME_EXCHANGE) | |
2b0143b5 | 427 | omode = d_inode(ndentry)->i_mode; |
d31a1825 | 428 | |
fab8eef8 AG |
429 | error = xfs_dentry_mode_to_name(&oname, odentry, omode); |
430 | if (omode && unlikely(error)) | |
431 | return error; | |
432 | ||
433 | error = xfs_dentry_mode_to_name(&nname, ndentry, | |
434 | d_inode(odentry)->i_mode); | |
435 | if (unlikely(error)) | |
436 | return error; | |
556b8b16 | 437 | |
2b0143b5 | 438 | return xfs_rename(XFS_I(odir), &oname, XFS_I(d_inode(odentry)), |
dbe1b5ca | 439 | XFS_I(ndir), &nname, |
d31a1825 | 440 | new_inode ? XFS_I(new_inode) : NULL, flags); |
1da177e4 LT |
441 | } |
442 | ||
443 | /* | |
444 | * careful here - this function can get called recursively, so | |
445 | * we need to be very careful about how much stack we use. | |
446 | * uio is kmalloced for this reason... | |
447 | */ | |
680baacb | 448 | STATIC const char * |
6b255391 | 449 | xfs_vn_get_link( |
1da177e4 | 450 | struct dentry *dentry, |
6b255391 | 451 | struct inode *inode, |
fceef393 | 452 | struct delayed_call *done) |
1da177e4 | 453 | { |
1da177e4 | 454 | char *link; |
804c83c3 | 455 | int error = -ENOMEM; |
1da177e4 | 456 | |
6b255391 AV |
457 | if (!dentry) |
458 | return ERR_PTR(-ECHILD); | |
459 | ||
6eb0b8df | 460 | link = kmalloc(XFS_SYMLINK_MAXLEN+1, GFP_KERNEL); |
804c83c3 CH |
461 | if (!link) |
462 | goto out_err; | |
1da177e4 | 463 | |
2b0143b5 | 464 | error = xfs_readlink(XFS_I(d_inode(dentry)), link); |
804c83c3 CH |
465 | if (unlikely(error)) |
466 | goto out_kfree; | |
1da177e4 | 467 | |
fceef393 AV |
468 | set_delayed_call(done, kfree_link, link); |
469 | return link; | |
804c83c3 CH |
470 | |
471 | out_kfree: | |
472 | kfree(link); | |
473 | out_err: | |
680baacb | 474 | return ERR_PTR(error); |
1da177e4 LT |
475 | } |
476 | ||
30ee052e CH |
477 | STATIC const char * |
478 | xfs_vn_get_link_inline( | |
479 | struct dentry *dentry, | |
480 | struct inode *inode, | |
481 | struct delayed_call *done) | |
482 | { | |
ae294787 DW |
483 | char *link; |
484 | ||
30ee052e | 485 | ASSERT(XFS_I(inode)->i_df.if_flags & XFS_IFINLINE); |
ae294787 DW |
486 | |
487 | /* | |
488 | * The VFS crashes on a NULL pointer, so return -EFSCORRUPTED if | |
489 | * if_data is junk. | |
490 | */ | |
491 | link = XFS_I(inode)->i_df.if_u1.if_data; | |
492 | if (!link) | |
493 | return ERR_PTR(-EFSCORRUPTED); | |
494 | return link; | |
30ee052e CH |
495 | } |
496 | ||
1da177e4 | 497 | STATIC int |
416c6d5b | 498 | xfs_vn_getattr( |
a528d35e DH |
499 | const struct path *path, |
500 | struct kstat *stat, | |
501 | u32 request_mask, | |
502 | unsigned int query_flags) | |
1da177e4 | 503 | { |
a528d35e | 504 | struct inode *inode = d_inode(path->dentry); |
c43f4087 CH |
505 | struct xfs_inode *ip = XFS_I(inode); |
506 | struct xfs_mount *mp = ip->i_mount; | |
507 | ||
cca28fb8 | 508 | trace_xfs_getattr(ip); |
c43f4087 CH |
509 | |
510 | if (XFS_FORCED_SHUTDOWN(mp)) | |
b474c7ae | 511 | return -EIO; |
c43f4087 CH |
512 | |
513 | stat->size = XFS_ISIZE(ip); | |
514 | stat->dev = inode->i_sb->s_dev; | |
c19b3b05 | 515 | stat->mode = inode->i_mode; |
54d7b5c1 | 516 | stat->nlink = inode->i_nlink; |
7aab1b28 DE |
517 | stat->uid = inode->i_uid; |
518 | stat->gid = inode->i_gid; | |
c43f4087 | 519 | stat->ino = ip->i_ino; |
c43f4087 | 520 | stat->atime = inode->i_atime; |
f9581b14 CH |
521 | stat->mtime = inode->i_mtime; |
522 | stat->ctime = inode->i_ctime; | |
c43f4087 CH |
523 | stat->blocks = |
524 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
525 | ||
5f955f26 DW |
526 | if (ip->i_d.di_version == 3) { |
527 | if (request_mask & STATX_BTIME) { | |
528 | stat->result_mask |= STATX_BTIME; | |
529 | stat->btime.tv_sec = ip->i_d.di_crtime.t_sec; | |
530 | stat->btime.tv_nsec = ip->i_d.di_crtime.t_nsec; | |
531 | } | |
532 | } | |
533 | ||
1b9598c8 LR |
534 | /* |
535 | * Note: If you add another clause to set an attribute flag, please | |
536 | * update attributes_mask below. | |
537 | */ | |
5f955f26 DW |
538 | if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE) |
539 | stat->attributes |= STATX_ATTR_IMMUTABLE; | |
540 | if (ip->i_d.di_flags & XFS_DIFLAG_APPEND) | |
541 | stat->attributes |= STATX_ATTR_APPEND; | |
542 | if (ip->i_d.di_flags & XFS_DIFLAG_NODUMP) | |
543 | stat->attributes |= STATX_ATTR_NODUMP; | |
c43f4087 | 544 | |
1b9598c8 LR |
545 | stat->attributes_mask |= (STATX_ATTR_IMMUTABLE | |
546 | STATX_ATTR_APPEND | | |
547 | STATX_ATTR_NODUMP); | |
548 | ||
c43f4087 CH |
549 | switch (inode->i_mode & S_IFMT) { |
550 | case S_IFBLK: | |
551 | case S_IFCHR: | |
552 | stat->blksize = BLKDEV_IOSIZE; | |
66f36464 | 553 | stat->rdev = inode->i_rdev; |
c43f4087 CH |
554 | break; |
555 | default: | |
71ddabb9 | 556 | if (XFS_IS_REALTIME_INODE(ip)) { |
c43f4087 CH |
557 | /* |
558 | * If the file blocks are being allocated from a | |
559 | * realtime volume, then return the inode's realtime | |
560 | * extent size or the realtime volume's extent size. | |
561 | */ | |
562 | stat->blksize = | |
563 | xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog; | |
564 | } else | |
565 | stat->blksize = xfs_preferred_iosize(mp); | |
566 | stat->rdev = 0; | |
567 | break; | |
69e23b9a | 568 | } |
c43f4087 CH |
569 | |
570 | return 0; | |
1da177e4 LT |
571 | } |
572 | ||
56c19e89 DC |
573 | static void |
574 | xfs_setattr_mode( | |
56c19e89 DC |
575 | struct xfs_inode *ip, |
576 | struct iattr *iattr) | |
577 | { | |
0c3d88df CH |
578 | struct inode *inode = VFS_I(ip); |
579 | umode_t mode = iattr->ia_mode; | |
56c19e89 | 580 | |
56c19e89 DC |
581 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); |
582 | ||
56c19e89 DC |
583 | inode->i_mode &= S_IFMT; |
584 | inode->i_mode |= mode & ~S_IFMT; | |
585 | } | |
586 | ||
52785112 | 587 | void |
c91c46c1 CH |
588 | xfs_setattr_time( |
589 | struct xfs_inode *ip, | |
590 | struct iattr *iattr) | |
591 | { | |
592 | struct inode *inode = VFS_I(ip); | |
593 | ||
594 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); | |
595 | ||
3987848c | 596 | if (iattr->ia_valid & ATTR_ATIME) |
c91c46c1 | 597 | inode->i_atime = iattr->ia_atime; |
3987848c | 598 | if (iattr->ia_valid & ATTR_CTIME) |
c91c46c1 | 599 | inode->i_ctime = iattr->ia_ctime; |
3987848c | 600 | if (iattr->ia_valid & ATTR_MTIME) |
c91c46c1 | 601 | inode->i_mtime = iattr->ia_mtime; |
c91c46c1 CH |
602 | } |
603 | ||
69bca807 JK |
604 | static int |
605 | xfs_vn_change_ok( | |
606 | struct dentry *dentry, | |
607 | struct iattr *iattr) | |
608 | { | |
31051c85 | 609 | struct xfs_mount *mp = XFS_I(d_inode(dentry))->i_mount; |
69bca807 JK |
610 | |
611 | if (mp->m_flags & XFS_MOUNT_RDONLY) | |
612 | return -EROFS; | |
613 | ||
614 | if (XFS_FORCED_SHUTDOWN(mp)) | |
615 | return -EIO; | |
616 | ||
31051c85 | 617 | return setattr_prepare(dentry, iattr); |
69bca807 JK |
618 | } |
619 | ||
620 | /* | |
621 | * Set non-size attributes of an inode. | |
622 | * | |
623 | * Caution: The caller of this function is responsible for calling | |
31051c85 | 624 | * setattr_prepare() or otherwise verifying the change is fine. |
69bca807 | 625 | */ |
c4ed4243 CH |
626 | int |
627 | xfs_setattr_nonsize( | |
628 | struct xfs_inode *ip, | |
629 | struct iattr *iattr, | |
630 | int flags) | |
631 | { | |
632 | xfs_mount_t *mp = ip->i_mount; | |
633 | struct inode *inode = VFS_I(ip); | |
634 | int mask = iattr->ia_valid; | |
635 | xfs_trans_t *tp; | |
636 | int error; | |
7aab1b28 DE |
637 | kuid_t uid = GLOBAL_ROOT_UID, iuid = GLOBAL_ROOT_UID; |
638 | kgid_t gid = GLOBAL_ROOT_GID, igid = GLOBAL_ROOT_GID; | |
c4ed4243 CH |
639 | struct xfs_dquot *udqp = NULL, *gdqp = NULL; |
640 | struct xfs_dquot *olddquot1 = NULL, *olddquot2 = NULL; | |
641 | ||
c4ed4243 CH |
642 | ASSERT((mask & ATTR_SIZE) == 0); |
643 | ||
644 | /* | |
645 | * If disk quotas is on, we make sure that the dquots do exist on disk, | |
646 | * before we start any other transactions. Trying to do this later | |
647 | * is messy. We don't care to take a readlock to look at the ids | |
648 | * in inode here, because we can't hold it across the trans_reserve. | |
649 | * If the IDs do change before we take the ilock, we're covered | |
650 | * because the i_*dquot fields will get updated anyway. | |
651 | */ | |
652 | if (XFS_IS_QUOTA_ON(mp) && (mask & (ATTR_UID|ATTR_GID))) { | |
653 | uint qflags = 0; | |
654 | ||
655 | if ((mask & ATTR_UID) && XFS_IS_UQUOTA_ON(mp)) { | |
656 | uid = iattr->ia_uid; | |
657 | qflags |= XFS_QMOPT_UQUOTA; | |
658 | } else { | |
7aab1b28 | 659 | uid = inode->i_uid; |
c4ed4243 CH |
660 | } |
661 | if ((mask & ATTR_GID) && XFS_IS_GQUOTA_ON(mp)) { | |
662 | gid = iattr->ia_gid; | |
663 | qflags |= XFS_QMOPT_GQUOTA; | |
664 | } else { | |
7aab1b28 | 665 | gid = inode->i_gid; |
c4ed4243 CH |
666 | } |
667 | ||
668 | /* | |
669 | * We take a reference when we initialize udqp and gdqp, | |
670 | * so it is important that we never blindly double trip on | |
671 | * the same variable. See xfs_create() for an example. | |
672 | */ | |
673 | ASSERT(udqp == NULL); | |
674 | ASSERT(gdqp == NULL); | |
7aab1b28 DE |
675 | error = xfs_qm_vop_dqalloc(ip, xfs_kuid_to_uid(uid), |
676 | xfs_kgid_to_gid(gid), | |
677 | xfs_get_projid(ip), | |
678 | qflags, &udqp, &gdqp, NULL); | |
c4ed4243 CH |
679 | if (error) |
680 | return error; | |
681 | } | |
682 | ||
253f4911 | 683 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp); |
c4ed4243 | 684 | if (error) |
253f4911 | 685 | goto out_dqrele; |
c4ed4243 CH |
686 | |
687 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
253f4911 | 688 | xfs_trans_ijoin(tp, ip, 0); |
c4ed4243 CH |
689 | |
690 | /* | |
691 | * Change file ownership. Must be the owner or privileged. | |
692 | */ | |
693 | if (mask & (ATTR_UID|ATTR_GID)) { | |
694 | /* | |
695 | * These IDs could have changed since we last looked at them. | |
696 | * But, we're assured that if the ownership did change | |
697 | * while we didn't have the inode locked, inode's dquot(s) | |
698 | * would have changed also. | |
699 | */ | |
7aab1b28 DE |
700 | iuid = inode->i_uid; |
701 | igid = inode->i_gid; | |
c4ed4243 CH |
702 | gid = (mask & ATTR_GID) ? iattr->ia_gid : igid; |
703 | uid = (mask & ATTR_UID) ? iattr->ia_uid : iuid; | |
704 | ||
705 | /* | |
706 | * Do a quota reservation only if uid/gid is actually | |
707 | * going to change. | |
708 | */ | |
709 | if (XFS_IS_QUOTA_RUNNING(mp) && | |
7aab1b28 DE |
710 | ((XFS_IS_UQUOTA_ON(mp) && !uid_eq(iuid, uid)) || |
711 | (XFS_IS_GQUOTA_ON(mp) && !gid_eq(igid, gid)))) { | |
c4ed4243 CH |
712 | ASSERT(tp); |
713 | error = xfs_qm_vop_chown_reserve(tp, ip, udqp, gdqp, | |
92f8ff73 | 714 | NULL, capable(CAP_FOWNER) ? |
c4ed4243 CH |
715 | XFS_QMOPT_FORCE_RES : 0); |
716 | if (error) /* out of quota */ | |
253f4911 | 717 | goto out_cancel; |
c4ed4243 CH |
718 | } |
719 | } | |
720 | ||
c4ed4243 CH |
721 | /* |
722 | * Change file ownership. Must be the owner or privileged. | |
723 | */ | |
724 | if (mask & (ATTR_UID|ATTR_GID)) { | |
725 | /* | |
726 | * CAP_FSETID overrides the following restrictions: | |
727 | * | |
728 | * The set-user-ID and set-group-ID bits of a file will be | |
729 | * cleared upon successful return from chown() | |
730 | */ | |
c19b3b05 | 731 | if ((inode->i_mode & (S_ISUID|S_ISGID)) && |
c4ed4243 | 732 | !capable(CAP_FSETID)) |
c19b3b05 | 733 | inode->i_mode &= ~(S_ISUID|S_ISGID); |
c4ed4243 CH |
734 | |
735 | /* | |
736 | * Change the ownerships and register quota modifications | |
737 | * in the transaction. | |
738 | */ | |
7aab1b28 | 739 | if (!uid_eq(iuid, uid)) { |
c4ed4243 CH |
740 | if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_UQUOTA_ON(mp)) { |
741 | ASSERT(mask & ATTR_UID); | |
742 | ASSERT(udqp); | |
743 | olddquot1 = xfs_qm_vop_chown(tp, ip, | |
744 | &ip->i_udquot, udqp); | |
745 | } | |
7aab1b28 | 746 | ip->i_d.di_uid = xfs_kuid_to_uid(uid); |
c4ed4243 CH |
747 | inode->i_uid = uid; |
748 | } | |
7aab1b28 | 749 | if (!gid_eq(igid, gid)) { |
c4ed4243 | 750 | if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_GQUOTA_ON(mp)) { |
5a01dd54 JL |
751 | ASSERT(xfs_sb_version_has_pquotino(&mp->m_sb) || |
752 | !XFS_IS_PQUOTA_ON(mp)); | |
c4ed4243 CH |
753 | ASSERT(mask & ATTR_GID); |
754 | ASSERT(gdqp); | |
755 | olddquot2 = xfs_qm_vop_chown(tp, ip, | |
756 | &ip->i_gdquot, gdqp); | |
757 | } | |
7aab1b28 | 758 | ip->i_d.di_gid = xfs_kgid_to_gid(gid); |
c4ed4243 CH |
759 | inode->i_gid = gid; |
760 | } | |
761 | } | |
762 | ||
56c19e89 | 763 | if (mask & ATTR_MODE) |
0c3d88df | 764 | xfs_setattr_mode(ip, iattr); |
c91c46c1 CH |
765 | if (mask & (ATTR_ATIME|ATTR_CTIME|ATTR_MTIME)) |
766 | xfs_setattr_time(ip, iattr); | |
c4ed4243 CH |
767 | |
768 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); | |
769 | ||
ff6d6af2 | 770 | XFS_STATS_INC(mp, xs_ig_attrchg); |
c4ed4243 CH |
771 | |
772 | if (mp->m_flags & XFS_MOUNT_WSYNC) | |
773 | xfs_trans_set_sync(tp); | |
70393313 | 774 | error = xfs_trans_commit(tp); |
c4ed4243 CH |
775 | |
776 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
777 | ||
778 | /* | |
779 | * Release any dquot(s) the inode had kept before chown. | |
780 | */ | |
781 | xfs_qm_dqrele(olddquot1); | |
782 | xfs_qm_dqrele(olddquot2); | |
783 | xfs_qm_dqrele(udqp); | |
784 | xfs_qm_dqrele(gdqp); | |
785 | ||
786 | if (error) | |
b474c7ae | 787 | return error; |
c4ed4243 CH |
788 | |
789 | /* | |
790 | * XXX(hch): Updating the ACL entries is not atomic vs the i_mode | |
791 | * update. We could avoid this with linked transactions | |
792 | * and passing down the transaction pointer all the way | |
793 | * to attr_set. No previous user of the generic | |
794 | * Posix ACL code seems to care about this issue either. | |
795 | */ | |
796 | if ((mask & ATTR_MODE) && !(flags & XFS_ATTR_NOACL)) { | |
2451337d | 797 | error = posix_acl_chmod(inode, inode->i_mode); |
c4ed4243 | 798 | if (error) |
b474c7ae | 799 | return error; |
c4ed4243 CH |
800 | } |
801 | ||
802 | return 0; | |
803 | ||
253f4911 | 804 | out_cancel: |
4906e215 | 805 | xfs_trans_cancel(tp); |
253f4911 | 806 | out_dqrele: |
c4ed4243 CH |
807 | xfs_qm_dqrele(udqp); |
808 | xfs_qm_dqrele(gdqp); | |
809 | return error; | |
810 | } | |
811 | ||
69bca807 JK |
812 | int |
813 | xfs_vn_setattr_nonsize( | |
814 | struct dentry *dentry, | |
815 | struct iattr *iattr) | |
816 | { | |
817 | struct xfs_inode *ip = XFS_I(d_inode(dentry)); | |
818 | int error; | |
819 | ||
820 | trace_xfs_setattr(ip); | |
821 | ||
822 | error = xfs_vn_change_ok(dentry, iattr); | |
823 | if (error) | |
824 | return error; | |
825 | return xfs_setattr_nonsize(ip, iattr, 0); | |
826 | } | |
827 | ||
c4ed4243 CH |
828 | /* |
829 | * Truncate file. Must have write permission and not be a directory. | |
69bca807 JK |
830 | * |
831 | * Caution: The caller of this function is responsible for calling | |
31051c85 | 832 | * setattr_prepare() or otherwise verifying the change is fine. |
c4ed4243 | 833 | */ |
7bf7a193 | 834 | STATIC int |
c4ed4243 CH |
835 | xfs_setattr_size( |
836 | struct xfs_inode *ip, | |
76ca4c23 | 837 | struct iattr *iattr) |
c4ed4243 CH |
838 | { |
839 | struct xfs_mount *mp = ip->i_mount; | |
840 | struct inode *inode = VFS_I(ip); | |
673e8e59 | 841 | xfs_off_t oldsize, newsize; |
c4ed4243 CH |
842 | struct xfs_trans *tp; |
843 | int error; | |
f38996f5 | 844 | uint lock_flags = 0; |
5885ebda | 845 | bool did_zeroing = false; |
c4ed4243 | 846 | |
76ca4c23 | 847 | ASSERT(xfs_isilocked(ip, XFS_IOLOCK_EXCL)); |
e8e9ad42 | 848 | ASSERT(xfs_isilocked(ip, XFS_MMAPLOCK_EXCL)); |
c19b3b05 | 849 | ASSERT(S_ISREG(inode->i_mode)); |
fe60a8a0 CH |
850 | ASSERT((iattr->ia_valid & (ATTR_UID|ATTR_GID|ATTR_ATIME|ATTR_ATIME_SET| |
851 | ATTR_MTIME_SET|ATTR_KILL_PRIV|ATTR_TIMES_SET)) == 0); | |
c4ed4243 | 852 | |
ce7ae151 | 853 | oldsize = inode->i_size; |
673e8e59 CH |
854 | newsize = iattr->ia_size; |
855 | ||
c4ed4243 CH |
856 | /* |
857 | * Short circuit the truncate case for zero length files. | |
858 | */ | |
673e8e59 | 859 | if (newsize == 0 && oldsize == 0 && ip->i_d.di_nextents == 0) { |
fe60a8a0 | 860 | if (!(iattr->ia_valid & (ATTR_CTIME|ATTR_MTIME))) |
76ca4c23 | 861 | return 0; |
681b1200 CH |
862 | |
863 | /* | |
864 | * Use the regular setattr path to update the timestamps. | |
865 | */ | |
681b1200 CH |
866 | iattr->ia_valid &= ~ATTR_SIZE; |
867 | return xfs_setattr_nonsize(ip, iattr, 0); | |
c4ed4243 CH |
868 | } |
869 | ||
870 | /* | |
871 | * Make sure that the dquots are attached to the inode. | |
872 | */ | |
c14cfcca | 873 | error = xfs_qm_dqattach(ip); |
c4ed4243 | 874 | if (error) |
76ca4c23 | 875 | return error; |
c4ed4243 | 876 | |
f0c6bcba CH |
877 | /* |
878 | * Wait for all direct I/O to complete. | |
879 | */ | |
880 | inode_dio_wait(inode); | |
881 | ||
c4ed4243 | 882 | /* |
5885ebda DC |
883 | * File data changes must be complete before we start the transaction to |
884 | * modify the inode. This needs to be done before joining the inode to | |
885 | * the transaction because the inode cannot be unlocked once it is a | |
886 | * part of the transaction. | |
887 | * | |
f0c6bcba CH |
888 | * Start with zeroing any data beyond EOF that we may expose on file |
889 | * extension, or zeroing out the rest of the block on a downward | |
890 | * truncate. | |
c4ed4243 | 891 | */ |
673e8e59 | 892 | if (newsize > oldsize) { |
f5c54717 CH |
893 | trace_xfs_zero_eof(ip, oldsize, newsize - oldsize); |
894 | error = iomap_zero_range(inode, oldsize, newsize - oldsize, | |
895 | &did_zeroing, &xfs_iomap_ops); | |
f0c6bcba | 896 | } else { |
459f0fbc CH |
897 | error = iomap_truncate_page(inode, newsize, &did_zeroing, |
898 | &xfs_iomap_ops); | |
c4ed4243 | 899 | } |
c4ed4243 | 900 | |
f0c6bcba CH |
901 | if (error) |
902 | return error; | |
903 | ||
49abc3a8 | 904 | /* |
0f9160b4 DC |
905 | * We've already locked out new page faults, so now we can safely remove |
906 | * pages from the page cache knowing they won't get refaulted until we | |
907 | * drop the XFS_MMAP_EXCL lock after the extent manipulations are | |
908 | * complete. The truncate_setsize() call also cleans partial EOF page | |
909 | * PTEs on extending truncates and hence ensures sub-page block size | |
910 | * filesystems are correctly handled, too. | |
49abc3a8 | 911 | * |
0f9160b4 DC |
912 | * We have to do all the page cache truncate work outside the |
913 | * transaction context as the "lock" order is page lock->log space | |
914 | * reservation as defined by extent allocation in the writeback path. | |
253f4911 | 915 | * Hence a truncate can fail with ENOMEM from xfs_trans_alloc(), but |
0f9160b4 DC |
916 | * having already truncated the in-memory version of the file (i.e. made |
917 | * user visible changes). There's not much we can do about this, except | |
918 | * to hope that the caller sees ENOMEM and retries the truncate | |
919 | * operation. | |
350976ae EG |
920 | * |
921 | * And we update in-core i_size and truncate page cache beyond newsize | |
922 | * before writeback the [di_size, newsize] range, so we're guaranteed | |
923 | * not to write stale data past the new EOF on truncate down. | |
49abc3a8 | 924 | */ |
49abc3a8 | 925 | truncate_setsize(inode, newsize); |
c4ed4243 | 926 | |
350976ae EG |
927 | /* |
928 | * We are going to log the inode size change in this transaction so | |
929 | * any previous writes that are beyond the on disk EOF and the new | |
930 | * EOF that have not been written out need to be written here. If we | |
931 | * do not write the data out, we expose ourselves to the null files | |
932 | * problem. Note that this includes any block zeroing we did above; | |
933 | * otherwise those blocks may not be zeroed after a crash. | |
934 | */ | |
935 | if (did_zeroing || | |
936 | (newsize > ip->i_d.di_size && oldsize != ip->i_d.di_size)) { | |
937 | error = filemap_write_and_wait_range(VFS_I(ip)->i_mapping, | |
938 | ip->i_d.di_size, newsize - 1); | |
939 | if (error) | |
940 | return error; | |
941 | } | |
942 | ||
253f4911 | 943 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, &tp); |
c4ed4243 | 944 | if (error) |
253f4911 | 945 | return error; |
c4ed4243 | 946 | |
c4ed4243 | 947 | lock_flags |= XFS_ILOCK_EXCL; |
c4ed4243 | 948 | xfs_ilock(ip, XFS_ILOCK_EXCL); |
ddc3415a | 949 | xfs_trans_ijoin(tp, ip, 0); |
c4ed4243 CH |
950 | |
951 | /* | |
952 | * Only change the c/mtime if we are changing the size or we are | |
953 | * explicitly asked to change it. This handles the semantic difference | |
954 | * between truncate() and ftruncate() as implemented in the VFS. | |
955 | * | |
956 | * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a | |
957 | * special case where we need to update the times despite not having | |
958 | * these flags set. For all other operations the VFS set these flags | |
959 | * explicitly if it wants a timestamp update. | |
960 | */ | |
fe60a8a0 CH |
961 | if (newsize != oldsize && |
962 | !(iattr->ia_valid & (ATTR_CTIME | ATTR_MTIME))) { | |
c4ed4243 | 963 | iattr->ia_ctime = iattr->ia_mtime = |
c2050a45 | 964 | current_time(inode); |
fe60a8a0 | 965 | iattr->ia_valid |= ATTR_CTIME | ATTR_MTIME; |
c4ed4243 CH |
966 | } |
967 | ||
673e8e59 CH |
968 | /* |
969 | * The first thing we do is set the size to new_size permanently on | |
970 | * disk. This way we don't have to worry about anyone ever being able | |
971 | * to look at the data being freed even in the face of a crash. | |
972 | * What we're getting around here is the case where we free a block, it | |
973 | * is allocated to another file, it is written to, and then we crash. | |
974 | * If the new data gets written to the file but the log buffers | |
975 | * containing the free and reallocation don't, then we'd end up with | |
976 | * garbage in the blocks being freed. As long as we make the new size | |
977 | * permanent before actually freeing any blocks it doesn't matter if | |
978 | * they get written to. | |
979 | */ | |
980 | ip->i_d.di_size = newsize; | |
673e8e59 CH |
981 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); |
982 | ||
983 | if (newsize <= oldsize) { | |
984 | error = xfs_itruncate_extents(&tp, ip, XFS_DATA_FORK, newsize); | |
c4ed4243 | 985 | if (error) |
4906e215 | 986 | goto out_trans_cancel; |
c4ed4243 CH |
987 | |
988 | /* | |
989 | * Truncated "down", so we're removing references to old data | |
990 | * here - if we delay flushing for a long time, we expose | |
991 | * ourselves unduly to the notorious NULL files problem. So, | |
992 | * we mark this inode and flush it when the file is closed, | |
993 | * and do not wait the usual (long) time for writeout. | |
994 | */ | |
995 | xfs_iflags_set(ip, XFS_ITRUNCATED); | |
27b52867 BF |
996 | |
997 | /* A truncate down always removes post-EOF blocks. */ | |
998 | xfs_inode_clear_eofblocks_tag(ip); | |
c4ed4243 CH |
999 | } |
1000 | ||
fe60a8a0 | 1001 | if (iattr->ia_valid & ATTR_MODE) |
0c3d88df | 1002 | xfs_setattr_mode(ip, iattr); |
fe60a8a0 | 1003 | if (iattr->ia_valid & (ATTR_ATIME|ATTR_CTIME|ATTR_MTIME)) |
c91c46c1 | 1004 | xfs_setattr_time(ip, iattr); |
c4ed4243 CH |
1005 | |
1006 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); | |
1007 | ||
ff6d6af2 | 1008 | XFS_STATS_INC(mp, xs_ig_attrchg); |
c4ed4243 CH |
1009 | |
1010 | if (mp->m_flags & XFS_MOUNT_WSYNC) | |
1011 | xfs_trans_set_sync(tp); | |
1012 | ||
70393313 | 1013 | error = xfs_trans_commit(tp); |
c4ed4243 CH |
1014 | out_unlock: |
1015 | if (lock_flags) | |
1016 | xfs_iunlock(ip, lock_flags); | |
1017 | return error; | |
1018 | ||
c4ed4243 | 1019 | out_trans_cancel: |
4906e215 | 1020 | xfs_trans_cancel(tp); |
c4ed4243 CH |
1021 | goto out_unlock; |
1022 | } | |
1023 | ||
69bca807 JK |
1024 | int |
1025 | xfs_vn_setattr_size( | |
1026 | struct dentry *dentry, | |
1027 | struct iattr *iattr) | |
1028 | { | |
1029 | struct xfs_inode *ip = XFS_I(d_inode(dentry)); | |
1030 | int error; | |
1031 | ||
1032 | trace_xfs_setattr(ip); | |
1033 | ||
1034 | error = xfs_vn_change_ok(dentry, iattr); | |
1035 | if (error) | |
1036 | return error; | |
1037 | return xfs_setattr_size(ip, iattr); | |
1038 | } | |
1039 | ||
1da177e4 | 1040 | STATIC int |
416c6d5b | 1041 | xfs_vn_setattr( |
76ca4c23 CH |
1042 | struct dentry *dentry, |
1043 | struct iattr *iattr) | |
1da177e4 | 1044 | { |
76ca4c23 CH |
1045 | int error; |
1046 | ||
1047 | if (iattr->ia_valid & ATTR_SIZE) { | |
69eb5fa1 DW |
1048 | struct inode *inode = d_inode(dentry); |
1049 | struct xfs_inode *ip = XFS_I(inode); | |
c63a8eae | 1050 | uint iolock; |
781355c6 | 1051 | |
c63a8eae DW |
1052 | xfs_ilock(ip, XFS_MMAPLOCK_EXCL); |
1053 | iolock = XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL; | |
781355c6 | 1054 | |
69eb5fa1 | 1055 | error = xfs_break_layouts(inode, &iolock, BREAK_UNMAP); |
c63a8eae DW |
1056 | if (error) { |
1057 | xfs_iunlock(ip, XFS_MMAPLOCK_EXCL); | |
65523218 | 1058 | return error; |
c63a8eae | 1059 | } |
e8e9ad42 | 1060 | |
0c187dc5 | 1061 | error = xfs_vn_setattr_size(dentry, iattr); |
65523218 | 1062 | xfs_iunlock(ip, XFS_MMAPLOCK_EXCL); |
76ca4c23 | 1063 | } else { |
69bca807 | 1064 | error = xfs_vn_setattr_nonsize(dentry, iattr); |
76ca4c23 CH |
1065 | } |
1066 | ||
2451337d | 1067 | return error; |
1da177e4 LT |
1068 | } |
1069 | ||
69ff2826 CH |
1070 | STATIC int |
1071 | xfs_vn_update_time( | |
1072 | struct inode *inode, | |
95582b00 | 1073 | struct timespec64 *now, |
69ff2826 CH |
1074 | int flags) |
1075 | { | |
1076 | struct xfs_inode *ip = XFS_I(inode); | |
1077 | struct xfs_mount *mp = ip->i_mount; | |
c3b1b131 | 1078 | int log_flags = XFS_ILOG_TIMESTAMP; |
69ff2826 CH |
1079 | struct xfs_trans *tp; |
1080 | int error; | |
1081 | ||
1082 | trace_xfs_update_time(ip); | |
1083 | ||
c3b1b131 CH |
1084 | if (inode->i_sb->s_flags & SB_LAZYTIME) { |
1085 | if (!((flags & S_VERSION) && | |
1086 | inode_maybe_inc_iversion(inode, false))) | |
1087 | return generic_update_time(inode, now, flags); | |
1088 | ||
1089 | /* Capture the iversion update that just occurred */ | |
1090 | log_flags |= XFS_ILOG_CORE; | |
1091 | } | |
1092 | ||
253f4911 CH |
1093 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_fsyncts, 0, 0, 0, &tp); |
1094 | if (error) | |
2451337d | 1095 | return error; |
69ff2826 CH |
1096 | |
1097 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
3987848c | 1098 | if (flags & S_CTIME) |
69ff2826 | 1099 | inode->i_ctime = *now; |
3987848c | 1100 | if (flags & S_MTIME) |
69ff2826 | 1101 | inode->i_mtime = *now; |
3987848c | 1102 | if (flags & S_ATIME) |
69ff2826 | 1103 | inode->i_atime = *now; |
3987848c | 1104 | |
69ff2826 | 1105 | xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); |
c3b1b131 | 1106 | xfs_trans_log_inode(tp, ip, log_flags); |
70393313 | 1107 | return xfs_trans_commit(tp); |
69ff2826 CH |
1108 | } |
1109 | ||
f35642e2 ES |
1110 | STATIC int |
1111 | xfs_vn_fiemap( | |
1112 | struct inode *inode, | |
1113 | struct fiemap_extent_info *fieinfo, | |
1114 | u64 start, | |
1115 | u64 length) | |
1116 | { | |
f35642e2 ES |
1117 | int error; |
1118 | ||
d2bb140e | 1119 | xfs_ilock(XFS_I(inode), XFS_IOLOCK_SHARED); |
1d4795e7 CH |
1120 | if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) { |
1121 | fieinfo->fi_flags &= ~FIEMAP_FLAG_XATTR; | |
1122 | error = iomap_fiemap(inode, fieinfo, start, length, | |
1123 | &xfs_xattr_iomap_ops); | |
1124 | } else { | |
1125 | error = iomap_fiemap(inode, fieinfo, start, length, | |
1126 | &xfs_iomap_ops); | |
1127 | } | |
d2bb140e | 1128 | xfs_iunlock(XFS_I(inode), XFS_IOLOCK_SHARED); |
f35642e2 | 1129 | |
d2bb140e | 1130 | return error; |
f35642e2 ES |
1131 | } |
1132 | ||
99b6436b ZYW |
1133 | STATIC int |
1134 | xfs_vn_tmpfile( | |
1135 | struct inode *dir, | |
1136 | struct dentry *dentry, | |
1137 | umode_t mode) | |
1138 | { | |
d540e43b | 1139 | return xfs_generic_create(dir, dentry, mode, 0, true); |
99b6436b ZYW |
1140 | } |
1141 | ||
41be8bed | 1142 | static const struct inode_operations xfs_inode_operations = { |
4e34e719 | 1143 | .get_acl = xfs_get_acl, |
2401dc29 | 1144 | .set_acl = xfs_set_acl, |
416c6d5b NS |
1145 | .getattr = xfs_vn_getattr, |
1146 | .setattr = xfs_vn_setattr, | |
416c6d5b | 1147 | .listxattr = xfs_vn_listxattr, |
f35642e2 | 1148 | .fiemap = xfs_vn_fiemap, |
69ff2826 | 1149 | .update_time = xfs_vn_update_time, |
1da177e4 LT |
1150 | }; |
1151 | ||
41be8bed | 1152 | static const struct inode_operations xfs_dir_inode_operations = { |
416c6d5b NS |
1153 | .create = xfs_vn_create, |
1154 | .lookup = xfs_vn_lookup, | |
1155 | .link = xfs_vn_link, | |
1156 | .unlink = xfs_vn_unlink, | |
1157 | .symlink = xfs_vn_symlink, | |
1158 | .mkdir = xfs_vn_mkdir, | |
8f112e3b CH |
1159 | /* |
1160 | * Yes, XFS uses the same method for rmdir and unlink. | |
1161 | * | |
1162 | * There are some subtile differences deeper in the code, | |
1163 | * but we use S_ISDIR to check for those. | |
1164 | */ | |
1165 | .rmdir = xfs_vn_unlink, | |
416c6d5b | 1166 | .mknod = xfs_vn_mknod, |
2773bf00 | 1167 | .rename = xfs_vn_rename, |
4e34e719 | 1168 | .get_acl = xfs_get_acl, |
2401dc29 | 1169 | .set_acl = xfs_set_acl, |
416c6d5b NS |
1170 | .getattr = xfs_vn_getattr, |
1171 | .setattr = xfs_vn_setattr, | |
416c6d5b | 1172 | .listxattr = xfs_vn_listxattr, |
69ff2826 | 1173 | .update_time = xfs_vn_update_time, |
99b6436b | 1174 | .tmpfile = xfs_vn_tmpfile, |
1da177e4 LT |
1175 | }; |
1176 | ||
41be8bed | 1177 | static const struct inode_operations xfs_dir_ci_inode_operations = { |
384f3ced BN |
1178 | .create = xfs_vn_create, |
1179 | .lookup = xfs_vn_ci_lookup, | |
1180 | .link = xfs_vn_link, | |
1181 | .unlink = xfs_vn_unlink, | |
1182 | .symlink = xfs_vn_symlink, | |
1183 | .mkdir = xfs_vn_mkdir, | |
8f112e3b CH |
1184 | /* |
1185 | * Yes, XFS uses the same method for rmdir and unlink. | |
1186 | * | |
1187 | * There are some subtile differences deeper in the code, | |
1188 | * but we use S_ISDIR to check for those. | |
1189 | */ | |
1190 | .rmdir = xfs_vn_unlink, | |
384f3ced | 1191 | .mknod = xfs_vn_mknod, |
2773bf00 | 1192 | .rename = xfs_vn_rename, |
4e34e719 | 1193 | .get_acl = xfs_get_acl, |
2401dc29 | 1194 | .set_acl = xfs_set_acl, |
384f3ced BN |
1195 | .getattr = xfs_vn_getattr, |
1196 | .setattr = xfs_vn_setattr, | |
384f3ced | 1197 | .listxattr = xfs_vn_listxattr, |
69ff2826 | 1198 | .update_time = xfs_vn_update_time, |
99b6436b | 1199 | .tmpfile = xfs_vn_tmpfile, |
384f3ced BN |
1200 | }; |
1201 | ||
41be8bed | 1202 | static const struct inode_operations xfs_symlink_inode_operations = { |
6b255391 | 1203 | .get_link = xfs_vn_get_link, |
416c6d5b NS |
1204 | .getattr = xfs_vn_getattr, |
1205 | .setattr = xfs_vn_setattr, | |
416c6d5b | 1206 | .listxattr = xfs_vn_listxattr, |
69ff2826 | 1207 | .update_time = xfs_vn_update_time, |
1da177e4 | 1208 | }; |
41be8bed | 1209 | |
30ee052e | 1210 | static const struct inode_operations xfs_inline_symlink_inode_operations = { |
30ee052e CH |
1211 | .get_link = xfs_vn_get_link_inline, |
1212 | .getattr = xfs_vn_getattr, | |
1213 | .setattr = xfs_vn_setattr, | |
30ee052e CH |
1214 | .listxattr = xfs_vn_listxattr, |
1215 | .update_time = xfs_vn_update_time, | |
1216 | }; | |
1217 | ||
ba23cba9 DW |
1218 | /* Figure out if this file actually supports DAX. */ |
1219 | static bool | |
1220 | xfs_inode_supports_dax( | |
1221 | struct xfs_inode *ip) | |
1222 | { | |
1223 | struct xfs_mount *mp = ip->i_mount; | |
1224 | ||
1225 | /* Only supported on non-reflinked files. */ | |
1226 | if (!S_ISREG(VFS_I(ip)->i_mode) || xfs_is_reflink_inode(ip)) | |
1227 | return false; | |
1228 | ||
1229 | /* DAX mount option or DAX iflag must be set. */ | |
1230 | if (!(mp->m_flags & XFS_MOUNT_DAX) && | |
1231 | !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) | |
1232 | return false; | |
1233 | ||
1234 | /* Block size must match page size */ | |
1235 | if (mp->m_sb.sb_blocksize != PAGE_SIZE) | |
1236 | return false; | |
1237 | ||
1238 | /* Device has to support DAX too. */ | |
1239 | return xfs_find_daxdev_for_inode(VFS_I(ip)) != NULL; | |
1240 | } | |
1241 | ||
41be8bed CH |
1242 | STATIC void |
1243 | xfs_diflags_to_iflags( | |
1244 | struct inode *inode, | |
1245 | struct xfs_inode *ip) | |
1246 | { | |
cbe4dab1 DC |
1247 | uint16_t flags = ip->i_d.di_flags; |
1248 | ||
1249 | inode->i_flags &= ~(S_IMMUTABLE | S_APPEND | S_SYNC | | |
1250 | S_NOATIME | S_DAX); | |
1251 | ||
1252 | if (flags & XFS_DIFLAG_IMMUTABLE) | |
41be8bed | 1253 | inode->i_flags |= S_IMMUTABLE; |
cbe4dab1 | 1254 | if (flags & XFS_DIFLAG_APPEND) |
41be8bed | 1255 | inode->i_flags |= S_APPEND; |
cbe4dab1 | 1256 | if (flags & XFS_DIFLAG_SYNC) |
41be8bed | 1257 | inode->i_flags |= S_SYNC; |
cbe4dab1 | 1258 | if (flags & XFS_DIFLAG_NOATIME) |
41be8bed | 1259 | inode->i_flags |= S_NOATIME; |
ba23cba9 | 1260 | if (xfs_inode_supports_dax(ip)) |
cbe4dab1 | 1261 | inode->i_flags |= S_DAX; |
41be8bed CH |
1262 | } |
1263 | ||
1264 | /* | |
2b3d1d41 | 1265 | * Initialize the Linux inode. |
bf904248 | 1266 | * |
58c90473 DC |
1267 | * When reading existing inodes from disk this is called directly from xfs_iget, |
1268 | * when creating a new inode it is called from xfs_ialloc after setting up the | |
1269 | * inode. These callers have different criteria for clearing XFS_INEW, so leave | |
1270 | * it up to the caller to deal with unlocking the inode appropriately. | |
41be8bed CH |
1271 | */ |
1272 | void | |
1273 | xfs_setup_inode( | |
1274 | struct xfs_inode *ip) | |
1275 | { | |
bf904248 | 1276 | struct inode *inode = &ip->i_vnode; |
ad22c7a0 | 1277 | gfp_t gfp_mask; |
bf904248 DC |
1278 | |
1279 | inode->i_ino = ip->i_ino; | |
eaff8079 | 1280 | inode->i_state = I_NEW; |
646ec461 CH |
1281 | |
1282 | inode_sb_list_add(inode); | |
c6f6cd06 | 1283 | /* make the inode look hashed for the writeback code */ |
5bef9151 | 1284 | inode_fake_hash(inode); |
41be8bed | 1285 | |
7aab1b28 DE |
1286 | inode->i_uid = xfs_uid_to_kuid(ip->i_d.di_uid); |
1287 | inode->i_gid = xfs_gid_to_kgid(ip->i_d.di_gid); | |
41be8bed | 1288 | |
41be8bed | 1289 | i_size_write(inode, ip->i_d.di_size); |
41be8bed | 1290 | xfs_diflags_to_iflags(inode, ip); |
41be8bed | 1291 | |
2b3d1d41 | 1292 | if (S_ISDIR(inode->i_mode)) { |
ef215e39 DC |
1293 | /* |
1294 | * We set the i_rwsem class here to avoid potential races with | |
1295 | * lockdep_annotate_inode_mutex_key() reinitialising the lock | |
1296 | * after a filehandle lookup has already found the inode in | |
1297 | * cache before it has been unlocked via unlock_new_inode(). | |
1298 | */ | |
1299 | lockdep_set_class(&inode->i_rwsem, | |
1300 | &inode->i_sb->s_type->i_mutex_dir_key); | |
93a8614e | 1301 | lockdep_set_class(&ip->i_lock.mr_lock, &xfs_dir_ilock_class); |
32c5483a | 1302 | ip->d_ops = ip->i_mount->m_dir_inode_ops; |
2b3d1d41 CH |
1303 | } else { |
1304 | ip->d_ops = ip->i_mount->m_nondir_inode_ops; | |
1305 | lockdep_set_class(&ip->i_lock.mr_lock, &xfs_nondir_ilock_class); | |
41be8bed CH |
1306 | } |
1307 | ||
ad22c7a0 DC |
1308 | /* |
1309 | * Ensure all page cache allocations are done from GFP_NOFS context to | |
1310 | * prevent direct reclaim recursion back into the filesystem and blowing | |
1311 | * stacks or deadlocking. | |
1312 | */ | |
1313 | gfp_mask = mapping_gfp_mask(inode->i_mapping); | |
1314 | mapping_set_gfp_mask(inode->i_mapping, (gfp_mask & ~(__GFP_FS))); | |
1315 | ||
510792ee CH |
1316 | /* |
1317 | * If there is no attribute fork no ACL can exist on this inode, | |
1318 | * and it can't have any file capabilities attached to it either. | |
1319 | */ | |
1320 | if (!XFS_IFORK_Q(ip)) { | |
1321 | inode_has_no_xattr(inode); | |
6311b108 | 1322 | cache_no_acl(inode); |
510792ee | 1323 | } |
41be8bed | 1324 | } |
2b3d1d41 CH |
1325 | |
1326 | void | |
1327 | xfs_setup_iops( | |
1328 | struct xfs_inode *ip) | |
1329 | { | |
1330 | struct inode *inode = &ip->i_vnode; | |
1331 | ||
41be8bed CH |
1332 | switch (inode->i_mode & S_IFMT) { |
1333 | case S_IFREG: | |
1334 | inode->i_op = &xfs_inode_operations; | |
1335 | inode->i_fop = &xfs_file_operations; | |
6e2608df DW |
1336 | if (IS_DAX(inode)) |
1337 | inode->i_mapping->a_ops = &xfs_dax_aops; | |
1338 | else | |
1339 | inode->i_mapping->a_ops = &xfs_address_space_operations; | |
41be8bed CH |
1340 | break; |
1341 | case S_IFDIR: | |
1342 | if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb)) | |
1343 | inode->i_op = &xfs_dir_ci_inode_operations; | |
1344 | else | |
1345 | inode->i_op = &xfs_dir_inode_operations; | |
1346 | inode->i_fop = &xfs_dir_file_operations; | |
1347 | break; | |
1348 | case S_IFLNK: | |
30ee052e CH |
1349 | if (ip->i_df.if_flags & XFS_IFINLINE) |
1350 | inode->i_op = &xfs_inline_symlink_inode_operations; | |
1351 | else | |
1352 | inode->i_op = &xfs_symlink_inode_operations; | |
41be8bed CH |
1353 | break; |
1354 | default: | |
1355 | inode->i_op = &xfs_inode_operations; | |
1356 | init_special_inode(inode, inode->i_mode, inode->i_rdev); | |
1357 | break; | |
1358 | } | |
41be8bed | 1359 | } |