Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/fs/open.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
8 | #include <linux/string.h> | |
9 | #include <linux/mm.h> | |
1da177e4 | 10 | #include <linux/file.h> |
9f3acc31 | 11 | #include <linux/fdtable.h> |
0eeca283 | 12 | #include <linux/fsnotify.h> |
1da177e4 | 13 | #include <linux/module.h> |
1da177e4 LT |
14 | #include <linux/tty.h> |
15 | #include <linux/namei.h> | |
16 | #include <linux/backing-dev.h> | |
16f7e0fe | 17 | #include <linux/capability.h> |
086f7316 | 18 | #include <linux/securebits.h> |
1da177e4 LT |
19 | #include <linux/security.h> |
20 | #include <linux/mount.h> | |
5590ff0d | 21 | #include <linux/fcntl.h> |
5a0e3ad6 | 22 | #include <linux/slab.h> |
7c0f6ba6 | 23 | #include <linux/uaccess.h> |
1da177e4 | 24 | #include <linux/fs.h> |
ef3daeda | 25 | #include <linux/personality.h> |
1da177e4 LT |
26 | #include <linux/pagemap.h> |
27 | #include <linux/syscalls.h> | |
ab2af1f5 | 28 | #include <linux/rcupdate.h> |
73241ccc | 29 | #include <linux/audit.h> |
97ac7350 | 30 | #include <linux/falloc.h> |
5ad4e53b | 31 | #include <linux/fs_struct.h> |
b65a9cfc | 32 | #include <linux/ima.h> |
2dfc1cae | 33 | #include <linux/dnotify.h> |
3f6d078d | 34 | #include <linux/compat.h> |
a793d79e | 35 | #include <linux/mnt_idmapping.h> |
1da177e4 | 36 | |
e81e3f4d EP |
37 | #include "internal.h" |
38 | ||
abf08576 | 39 | int do_truncate(struct mnt_idmap *idmap, struct dentry *dentry, |
643fe55a | 40 | loff_t length, unsigned int time_attrs, struct file *filp) |
1da177e4 | 41 | { |
abf08576 | 42 | struct user_namespace *mnt_userns = mnt_idmap_owner(idmap); |
939a9421 | 43 | int ret; |
1da177e4 LT |
44 | struct iattr newattrs; |
45 | ||
46 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
47 | if (length < 0) | |
48 | return -EINVAL; | |
49 | ||
50 | newattrs.ia_size = length; | |
4a30131e | 51 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
52 | if (filp) { |
53 | newattrs.ia_file = filp; | |
54 | newattrs.ia_valid |= ATTR_FILE; | |
55 | } | |
1da177e4 | 56 | |
45f147a1 | 57 | /* Remove suid, sgid, and file capabilities on truncate too */ |
ed5a7047 | 58 | ret = dentry_needs_remove_privs(mnt_userns, dentry); |
45f147a1 JK |
59 | if (ret < 0) |
60 | return ret; | |
939a9421 AW |
61 | if (ret) |
62 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 63 | |
5955102c | 64 | inode_lock(dentry->d_inode); |
27ac0ffe | 65 | /* Note any delegations or leases have already been broken: */ |
abf08576 | 66 | ret = notify_change(idmap, dentry, &newattrs, NULL); |
5955102c | 67 | inode_unlock(dentry->d_inode); |
939a9421 | 68 | return ret; |
1da177e4 LT |
69 | } |
70 | ||
7df818b2 | 71 | long vfs_truncate(const struct path *path, loff_t length) |
1da177e4 | 72 | { |
abf08576 | 73 | struct mnt_idmap *idmap; |
2d8f3038 | 74 | struct inode *inode; |
a02de960 | 75 | long error; |
1da177e4 | 76 | |
a02de960 | 77 | inode = path->dentry->d_inode; |
1da177e4 LT |
78 | |
79 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
1da177e4 | 80 | if (S_ISDIR(inode->i_mode)) |
a02de960 | 81 | return -EISDIR; |
1da177e4 | 82 | if (!S_ISREG(inode->i_mode)) |
a02de960 | 83 | return -EINVAL; |
1da177e4 | 84 | |
a02de960 | 85 | error = mnt_want_write(path->mnt); |
1da177e4 | 86 | if (error) |
a02de960 | 87 | goto out; |
1da177e4 | 88 | |
abf08576 | 89 | idmap = mnt_idmap(path->mnt); |
4609e1f1 | 90 | error = inode_permission(idmap, inode, MAY_WRITE); |
9ac9b847 DH |
91 | if (error) |
92 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
93 | |
94 | error = -EPERM; | |
c82e42da | 95 | if (IS_APPEND(inode)) |
9ac9b847 | 96 | goto mnt_drop_write_and_out; |
1da177e4 | 97 | |
8cf9ee50 | 98 | error = get_write_access(inode); |
1da177e4 | 99 | if (error) |
9ac9b847 | 100 | goto mnt_drop_write_and_out; |
1da177e4 | 101 | |
9700382c | 102 | /* |
103 | * Make sure that there are no leases. get_write_access() protects | |
104 | * against the truncate racing with a lease-granting setlease(). | |
105 | */ | |
8737c930 | 106 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 107 | if (error) |
9700382c | 108 | goto put_write_and_out; |
1da177e4 | 109 | |
f7e33bdb | 110 | error = security_path_truncate(path); |
907f4554 | 111 | if (!error) |
abf08576 | 112 | error = do_truncate(idmap, path->dentry, length, 0, NULL); |
1da177e4 | 113 | |
9700382c | 114 | put_write_and_out: |
8cf9ee50 | 115 | put_write_access(inode); |
9ac9b847 | 116 | mnt_drop_write_and_out: |
a02de960 | 117 | mnt_drop_write(path->mnt); |
1da177e4 LT |
118 | out: |
119 | return error; | |
120 | } | |
a02de960 DH |
121 | EXPORT_SYMBOL_GPL(vfs_truncate); |
122 | ||
df260e21 | 123 | long do_sys_truncate(const char __user *pathname, loff_t length) |
a02de960 | 124 | { |
48f7530d | 125 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
126 | struct path path; |
127 | int error; | |
128 | ||
129 | if (length < 0) /* sorry, but loff_t says... */ | |
130 | return -EINVAL; | |
131 | ||
48f7530d JL |
132 | retry: |
133 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
134 | if (!error) { |
135 | error = vfs_truncate(&path, length); | |
136 | path_put(&path); | |
137 | } | |
48f7530d JL |
138 | if (retry_estale(error, lookup_flags)) { |
139 | lookup_flags |= LOOKUP_REVAL; | |
140 | goto retry; | |
141 | } | |
a02de960 DH |
142 | return error; |
143 | } | |
1da177e4 | 144 | |
4fd8da8d | 145 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 146 | { |
4fd8da8d | 147 | return do_sys_truncate(path, length); |
1da177e4 LT |
148 | } |
149 | ||
3f6d078d AV |
150 | #ifdef CONFIG_COMPAT |
151 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
152 | { | |
153 | return do_sys_truncate(path, length); | |
154 | } | |
155 | #endif | |
156 | ||
411d9475 | 157 | long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 | 158 | { |
bf2965d5 | 159 | struct inode *inode; |
1da177e4 | 160 | struct dentry *dentry; |
2903ff01 | 161 | struct fd f; |
1da177e4 LT |
162 | int error; |
163 | ||
164 | error = -EINVAL; | |
165 | if (length < 0) | |
166 | goto out; | |
167 | error = -EBADF; | |
2903ff01 AV |
168 | f = fdget(fd); |
169 | if (!f.file) | |
1da177e4 LT |
170 | goto out; |
171 | ||
172 | /* explicitly opened as large or we are on 64-bit box */ | |
2903ff01 | 173 | if (f.file->f_flags & O_LARGEFILE) |
1da177e4 LT |
174 | small = 0; |
175 | ||
2903ff01 | 176 | dentry = f.file->f_path.dentry; |
1da177e4 LT |
177 | inode = dentry->d_inode; |
178 | error = -EINVAL; | |
2903ff01 | 179 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
180 | goto out_putf; |
181 | ||
182 | error = -EINVAL; | |
183 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
184 | if (small && length > MAX_NON_LFS) | |
185 | goto out_putf; | |
186 | ||
187 | error = -EPERM; | |
78757af6 AG |
188 | /* Check IS_APPEND on real upper inode */ |
189 | if (IS_APPEND(file_inode(f.file))) | |
1da177e4 | 190 | goto out_putf; |
14da9200 | 191 | sb_start_write(inode->i_sb); |
3350607d | 192 | error = security_file_truncate(f.file); |
1da177e4 | 193 | if (!error) |
abf08576 | 194 | error = do_truncate(file_mnt_idmap(f.file), dentry, length, |
643fe55a | 195 | ATTR_MTIME | ATTR_CTIME, f.file); |
14da9200 | 196 | sb_end_write(inode->i_sb); |
1da177e4 | 197 | out_putf: |
2903ff01 | 198 | fdput(f); |
1da177e4 LT |
199 | out: |
200 | return error; | |
201 | } | |
202 | ||
bdc480e3 | 203 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 204 | { |
2cf09666 | 205 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
206 | } |
207 | ||
3f6d078d AV |
208 | #ifdef CONFIG_COMPAT |
209 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
210 | { | |
211 | return do_sys_ftruncate(fd, length, 1); | |
212 | } | |
213 | #endif | |
214 | ||
1da177e4 LT |
215 | /* LFS versions of truncate are only needed on 32 bit machines */ |
216 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 217 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
218 | { |
219 | return do_sys_truncate(path, length); | |
220 | } | |
221 | ||
4a0fd5bf | 222 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 223 | { |
2cf09666 | 224 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 225 | } |
6673e0c3 | 226 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 227 | |
59c10c52 GR |
228 | #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_TRUNCATE64) |
229 | COMPAT_SYSCALL_DEFINE3(truncate64, const char __user *, pathname, | |
230 | compat_arg_u64_dual(length)) | |
231 | { | |
232 | return ksys_truncate(pathname, compat_arg_u64_glue(length)); | |
233 | } | |
234 | #endif | |
235 | ||
236 | #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_FTRUNCATE64) | |
237 | COMPAT_SYSCALL_DEFINE3(ftruncate64, unsigned int, fd, | |
238 | compat_arg_u64_dual(length)) | |
239 | { | |
240 | return ksys_ftruncate(fd, compat_arg_u64_glue(length)); | |
241 | } | |
242 | #endif | |
3e63cbb1 | 243 | |
72c72bdf | 244 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 245 | { |
496ad9aa | 246 | struct inode *inode = file_inode(file); |
3e63cbb1 | 247 | long ret; |
97ac7350 AA |
248 | |
249 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 250 | return -EINVAL; |
97ac7350 AA |
251 | |
252 | /* Return error if mode is not supported */ | |
dd46c787 | 253 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) |
409332b6 LC |
254 | return -EOPNOTSUPP; |
255 | ||
256 | /* Punch hole and zero range are mutually exclusive */ | |
257 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
258 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
259 | return -EOPNOTSUPP; |
260 | ||
261 | /* Punch hole must have keep size set */ | |
262 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
263 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 264 | return -EOPNOTSUPP; |
97ac7350 | 265 | |
00f5e619 NJ |
266 | /* Collapse range should only be used exclusively. */ |
267 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
268 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
269 | return -EINVAL; | |
270 | ||
dd46c787 NJ |
271 | /* Insert range should only be used exclusively. */ |
272 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
273 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
274 | return -EINVAL; | |
275 | ||
71be6b49 DW |
276 | /* Unshare range should only be used with allocate mode. */ |
277 | if ((mode & FALLOC_FL_UNSHARE_RANGE) && | |
278 | (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE))) | |
279 | return -EINVAL; | |
280 | ||
97ac7350 | 281 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 282 | return -EBADF; |
1ca551c6 | 283 | |
00f5e619 | 284 | /* |
8fc61d92 | 285 | * We can only allow pure fallocate on append only files |
00f5e619 | 286 | */ |
8fc61d92 | 287 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
288 | return -EPERM; |
289 | ||
290 | if (IS_IMMUTABLE(inode)) | |
291 | return -EPERM; | |
292 | ||
0790b31b | 293 | /* |
6d2b6170 | 294 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
295 | */ |
296 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 297 | return -ETXTBSY; |
0790b31b | 298 | |
97ac7350 AA |
299 | /* |
300 | * Revalidate the write permissions, in case security policy has | |
301 | * changed since the files were opened. | |
302 | */ | |
303 | ret = security_file_permission(file, MAY_WRITE); | |
304 | if (ret) | |
3e63cbb1 | 305 | return ret; |
97ac7350 | 306 | |
97ac7350 | 307 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 308 | return -ESPIPE; |
97ac7350 | 309 | |
9e79b132 AG |
310 | if (S_ISDIR(inode->i_mode)) |
311 | return -EISDIR; | |
312 | ||
313 | if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode)) | |
3e63cbb1 | 314 | return -ENODEV; |
97ac7350 | 315 | |
97ac7350 AA |
316 | /* Check for wrap through zero too */ |
317 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 318 | return -EFBIG; |
97ac7350 | 319 | |
2fe17c10 | 320 | if (!file->f_op->fallocate) |
3e63cbb1 | 321 | return -EOPNOTSUPP; |
97ac7350 | 322 | |
bfe219d3 | 323 | file_start_write(file); |
14da9200 | 324 | ret = file->f_op->fallocate(file, mode, offset, len); |
820c12d5 HS |
325 | |
326 | /* | |
327 | * Create inotify and fanotify events. | |
328 | * | |
329 | * To keep the logic simple always create events if fallocate succeeds. | |
330 | * This implies that events are even created if the file size remains | |
331 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
332 | */ | |
333 | if (ret == 0) | |
334 | fsnotify_modify(file); | |
335 | ||
bfe219d3 | 336 | file_end_write(file); |
14da9200 | 337 | return ret; |
3e63cbb1 | 338 | } |
72c72bdf | 339 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 340 | |
edf292c7 | 341 | int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len) |
3e63cbb1 | 342 | { |
2903ff01 | 343 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
344 | int error = -EBADF; |
345 | ||
2903ff01 | 346 | if (f.file) { |
72c72bdf | 347 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 348 | fdput(f); |
3e63cbb1 | 349 | } |
3e63cbb1 | 350 | return error; |
97ac7350 | 351 | } |
3e63cbb1 | 352 | |
edf292c7 DB |
353 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
354 | { | |
355 | return ksys_fallocate(fd, mode, offset, len); | |
356 | } | |
357 | ||
59c10c52 GR |
358 | #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_FALLOCATE) |
359 | COMPAT_SYSCALL_DEFINE6(fallocate, int, fd, int, mode, compat_arg_u64_dual(offset), | |
360 | compat_arg_u64_dual(len)) | |
361 | { | |
362 | return ksys_fallocate(fd, mode, compat_arg_u64_glue(offset), | |
363 | compat_arg_u64_glue(len)); | |
364 | } | |
365 | #endif | |
366 | ||
1da177e4 LT |
367 | /* |
368 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
369 | * We do this by temporarily clearing all FS-related capabilities and | |
370 | * switching the fsuid/fsgid around to the real ones. | |
371 | */ | |
94704515 | 372 | static const struct cred *access_override_creds(void) |
1da177e4 | 373 | { |
d84f4f99 DH |
374 | const struct cred *old_cred; |
375 | struct cred *override_cred; | |
1da177e4 | 376 | |
d84f4f99 DH |
377 | override_cred = prepare_creds(); |
378 | if (!override_cred) | |
94704515 | 379 | return NULL; |
1da177e4 | 380 | |
d84f4f99 DH |
381 | override_cred->fsuid = override_cred->uid; |
382 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 383 | |
086f7316 | 384 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 385 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
386 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
387 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 388 | cap_clear(override_cred->cap_effective); |
086f7316 | 389 | else |
d84f4f99 DH |
390 | override_cred->cap_effective = |
391 | override_cred->cap_permitted; | |
086f7316 | 392 | } |
1da177e4 | 393 | |
d7852fbd LT |
394 | /* |
395 | * The new set of credentials can *only* be used in | |
396 | * task-synchronous circumstances, and does not need | |
397 | * RCU freeing, unless somebody then takes a separate | |
398 | * reference to it. | |
399 | * | |
400 | * NOTE! This is _only_ true because this credential | |
401 | * is used purely for override_creds() that installs | |
402 | * it as the subjective cred. Other threads will be | |
403 | * accessing ->real_cred, not the subjective cred. | |
404 | * | |
405 | * If somebody _does_ make a copy of this (using the | |
406 | * 'get_current_cred()' function), that will clear the | |
407 | * non_rcu field, because now that other user may be | |
408 | * expecting RCU freeing. But normal thread-synchronous | |
409 | * cred accesses will keep things non-RCY. | |
410 | */ | |
411 | override_cred->non_rcu = 1; | |
412 | ||
d84f4f99 | 413 | old_cred = override_creds(override_cred); |
94704515 MS |
414 | |
415 | /* override_cred() gets its own ref */ | |
416 | put_cred(override_cred); | |
417 | ||
418 | return old_cred; | |
419 | } | |
420 | ||
eb9d7d39 | 421 | static long do_faccessat(int dfd, const char __user *filename, int mode, int flags) |
94704515 MS |
422 | { |
423 | struct path path; | |
424 | struct inode *inode; | |
425 | int res; | |
426 | unsigned int lookup_flags = LOOKUP_FOLLOW; | |
c8ffd8bc | 427 | const struct cred *old_cred = NULL; |
94704515 MS |
428 | |
429 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
430 | return -EINVAL; | |
431 | ||
c8ffd8bc MS |
432 | if (flags & ~(AT_EACCESS | AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) |
433 | return -EINVAL; | |
434 | ||
435 | if (flags & AT_SYMLINK_NOFOLLOW) | |
436 | lookup_flags &= ~LOOKUP_FOLLOW; | |
437 | if (flags & AT_EMPTY_PATH) | |
438 | lookup_flags |= LOOKUP_EMPTY; | |
439 | ||
440 | if (!(flags & AT_EACCESS)) { | |
441 | old_cred = access_override_creds(); | |
442 | if (!old_cred) | |
443 | return -ENOMEM; | |
444 | } | |
94704515 | 445 | |
87fa5595 JL |
446 | retry: |
447 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
448 | if (res) |
449 | goto out; | |
450 | ||
63afdfc7 | 451 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
452 | |
453 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
454 | /* |
455 | * MAY_EXEC on regular files is denied if the fs is mounted | |
456 | * with the "noexec" flag. | |
457 | */ | |
458 | res = -EACCES; | |
90f8572b | 459 | if (path_noexec(&path)) |
30524472 AV |
460 | goto out_path_release; |
461 | } | |
462 | ||
4609e1f1 | 463 | res = inode_permission(mnt_idmap(path.mnt), inode, mode | MAY_ACCESS); |
6902d925 | 464 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 465 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 466 | goto out_path_release; |
2f676cbc DH |
467 | /* |
468 | * This is a rare case where using __mnt_is_readonly() | |
469 | * is OK without a mnt_want/drop_write() pair. Since | |
470 | * no actual write to the fs is performed here, we do | |
471 | * not need to telegraph to that to anyone. | |
472 | * | |
473 | * By doing this, we accept that this access is | |
474 | * inherently racy and know that the fs may change | |
475 | * state before we even see this result. | |
476 | */ | |
2d8f3038 | 477 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 478 | res = -EROFS; |
1da177e4 | 479 | |
6902d925 | 480 | out_path_release: |
2d8f3038 | 481 | path_put(&path); |
87fa5595 JL |
482 | if (retry_estale(res, lookup_flags)) { |
483 | lookup_flags |= LOOKUP_REVAL; | |
484 | goto retry; | |
485 | } | |
6902d925 | 486 | out: |
c8ffd8bc MS |
487 | if (old_cred) |
488 | revert_creds(old_cred); | |
489 | ||
1da177e4 LT |
490 | return res; |
491 | } | |
492 | ||
cbfe20f5 DB |
493 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
494 | { | |
c8ffd8bc MS |
495 | return do_faccessat(dfd, filename, mode, 0); |
496 | } | |
497 | ||
498 | SYSCALL_DEFINE4(faccessat2, int, dfd, const char __user *, filename, int, mode, | |
499 | int, flags) | |
500 | { | |
501 | return do_faccessat(dfd, filename, mode, flags); | |
cbfe20f5 DB |
502 | } |
503 | ||
ca013e94 | 504 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d | 505 | { |
c8ffd8bc | 506 | return do_faccessat(AT_FDCWD, filename, mode, 0); |
5590ff0d UD |
507 | } |
508 | ||
db63f1e3 | 509 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 510 | { |
2d8f3038 | 511 | struct path path; |
1da177e4 | 512 | int error; |
0291c0a5 JL |
513 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
514 | retry: | |
515 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
516 | if (error) |
517 | goto out; | |
518 | ||
02f92b38 | 519 | error = path_permission(&path, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
520 | if (error) |
521 | goto dput_and_out; | |
522 | ||
2d8f3038 | 523 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
524 | |
525 | dput_and_out: | |
2d8f3038 | 526 | path_put(&path); |
0291c0a5 JL |
527 | if (retry_estale(error, lookup_flags)) { |
528 | lookup_flags |= LOOKUP_REVAL; | |
529 | goto retry; | |
530 | } | |
1da177e4 LT |
531 | out: |
532 | return error; | |
533 | } | |
534 | ||
3cdad428 | 535 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 536 | { |
2903ff01 | 537 | struct fd f = fdget_raw(fd); |
159b0956 | 538 | int error; |
1da177e4 LT |
539 | |
540 | error = -EBADF; | |
2903ff01 | 541 | if (!f.file) |
1da177e4 LT |
542 | goto out; |
543 | ||
1da177e4 | 544 | error = -ENOTDIR; |
159b0956 | 545 | if (!d_can_lookup(f.file->f_path.dentry)) |
1da177e4 LT |
546 | goto out_putf; |
547 | ||
02f92b38 | 548 | error = file_permission(f.file, MAY_EXEC | MAY_CHDIR); |
1da177e4 | 549 | if (!error) |
2903ff01 | 550 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 551 | out_putf: |
2903ff01 | 552 | fdput(f); |
1da177e4 LT |
553 | out: |
554 | return error; | |
555 | } | |
556 | ||
4b7ca501 | 557 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 558 | { |
2d8f3038 | 559 | struct path path; |
1da177e4 | 560 | int error; |
2771261e JL |
561 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
562 | retry: | |
563 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
564 | if (error) |
565 | goto out; | |
566 | ||
02f92b38 | 567 | error = path_permission(&path, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
568 | if (error) |
569 | goto dput_and_out; | |
570 | ||
571 | error = -EPERM; | |
c7b96acf | 572 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 573 | goto dput_and_out; |
8b8efb44 TH |
574 | error = security_path_chroot(&path); |
575 | if (error) | |
576 | goto dput_and_out; | |
1da177e4 | 577 | |
2d8f3038 | 578 | set_fs_root(current->fs, &path); |
1da177e4 LT |
579 | error = 0; |
580 | dput_and_out: | |
2d8f3038 | 581 | path_put(&path); |
2771261e JL |
582 | if (retry_estale(error, lookup_flags)) { |
583 | lookup_flags |= LOOKUP_REVAL; | |
584 | goto retry; | |
585 | } | |
1da177e4 LT |
586 | out: |
587 | return error; | |
588 | } | |
589 | ||
1097742e | 590 | int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 591 | { |
e57712eb | 592 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 593 | struct inode *delegated_inode = NULL; |
1da177e4 | 594 | struct iattr newattrs; |
e57712eb | 595 | int error; |
1da177e4 | 596 | |
e57712eb AV |
597 | error = mnt_want_write(path->mnt); |
598 | if (error) | |
599 | return error; | |
27ac0ffe | 600 | retry_deleg: |
5955102c | 601 | inode_lock(inode); |
cdcf116d | 602 | error = security_path_chmod(path, mode); |
e57712eb | 603 | if (error) |
fe542cf5 | 604 | goto out_unlock; |
1da177e4 LT |
605 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
606 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
abf08576 | 607 | error = notify_change(mnt_idmap(path->mnt), path->dentry, |
b8b546a0 | 608 | &newattrs, &delegated_inode); |
fe542cf5 | 609 | out_unlock: |
5955102c | 610 | inode_unlock(inode); |
27ac0ffe BF |
611 | if (delegated_inode) { |
612 | error = break_deleg_wait(&delegated_inode); | |
613 | if (!error) | |
614 | goto retry_deleg; | |
615 | } | |
e57712eb AV |
616 | mnt_drop_write(path->mnt); |
617 | return error; | |
618 | } | |
619 | ||
9e96c8c0 CH |
620 | int vfs_fchmod(struct file *file, umode_t mode) |
621 | { | |
622 | audit_file(file); | |
623 | return chmod_common(&file->f_path, mode); | |
624 | } | |
625 | ||
b25ba7c3 | 626 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 627 | { |
173c8401 | 628 | struct fd f = fdget(fd); |
e57712eb AV |
629 | int err = -EBADF; |
630 | ||
173c8401 | 631 | if (f.file) { |
9e96c8c0 | 632 | err = vfs_fchmod(f.file, mode); |
173c8401 | 633 | fdput(f); |
e57712eb | 634 | } |
1da177e4 LT |
635 | return err; |
636 | } | |
637 | ||
1097742e | 638 | static int do_fchmodat(int dfd, const char __user *filename, umode_t mode) |
1da177e4 | 639 | { |
2d8f3038 | 640 | struct path path; |
1da177e4 | 641 | int error; |
14ff690c JL |
642 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
643 | retry: | |
644 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
645 | if (!error) { |
646 | error = chmod_common(&path, mode); | |
647 | path_put(&path); | |
14ff690c JL |
648 | if (retry_estale(error, lookup_flags)) { |
649 | lookup_flags |= LOOKUP_REVAL; | |
650 | goto retry; | |
651 | } | |
e57712eb | 652 | } |
1da177e4 LT |
653 | return error; |
654 | } | |
655 | ||
03450e27 DB |
656 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, |
657 | umode_t, mode) | |
658 | { | |
659 | return do_fchmodat(dfd, filename, mode); | |
660 | } | |
661 | ||
49f0a076 | 662 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d | 663 | { |
03450e27 | 664 | return do_fchmodat(AT_FDCWD, filename, mode); |
5590ff0d UD |
665 | } |
666 | ||
b27c82e1 CB |
667 | /** |
668 | * setattr_vfsuid - check and set ia_fsuid attribute | |
669 | * @kuid: new inode owner | |
670 | * | |
671 | * Check whether @kuid is valid and if so generate and set vfsuid_t in | |
672 | * ia_vfsuid. | |
673 | * | |
674 | * Return: true if @kuid is valid, false if not. | |
675 | */ | |
676 | static inline bool setattr_vfsuid(struct iattr *attr, kuid_t kuid) | |
677 | { | |
678 | if (!uid_valid(kuid)) | |
679 | return false; | |
680 | attr->ia_valid |= ATTR_UID; | |
681 | attr->ia_vfsuid = VFSUIDT_INIT(kuid); | |
682 | return true; | |
683 | } | |
684 | ||
685 | /** | |
686 | * setattr_vfsgid - check and set ia_fsgid attribute | |
687 | * @kgid: new inode owner | |
688 | * | |
689 | * Check whether @kgid is valid and if so generate and set vfsgid_t in | |
690 | * ia_vfsgid. | |
691 | * | |
692 | * Return: true if @kgid is valid, false if not. | |
693 | */ | |
694 | static inline bool setattr_vfsgid(struct iattr *attr, kgid_t kgid) | |
695 | { | |
696 | if (!gid_valid(kgid)) | |
697 | return false; | |
698 | attr->ia_valid |= ATTR_GID; | |
699 | attr->ia_vfsgid = VFSGIDT_INIT(kgid); | |
700 | return true; | |
701 | } | |
702 | ||
b873498f | 703 | int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 704 | { |
abf08576 | 705 | struct mnt_idmap *idmap; |
bd303368 | 706 | struct user_namespace *mnt_userns, *fs_userns; |
fe542cf5 | 707 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 708 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
709 | int error; |
710 | struct iattr newattrs; | |
52137abe EB |
711 | kuid_t uid; |
712 | kgid_t gid; | |
713 | ||
714 | uid = make_kuid(current_user_ns(), user); | |
715 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 716 | |
abf08576 CB |
717 | idmap = mnt_idmap(path->mnt); |
718 | mnt_userns = mnt_idmap_owner(idmap); | |
bd303368 | 719 | fs_userns = i_user_ns(inode); |
b8b546a0 | 720 | |
c1b8940b | 721 | retry_deleg: |
f52d74b1 TH |
722 | newattrs.ia_vfsuid = INVALID_VFSUID; |
723 | newattrs.ia_vfsgid = INVALID_VFSGID; | |
1da177e4 | 724 | newattrs.ia_valid = ATTR_CTIME; |
b27c82e1 CB |
725 | if ((user != (uid_t)-1) && !setattr_vfsuid(&newattrs, uid)) |
726 | return -EINVAL; | |
727 | if ((group != (gid_t)-1) && !setattr_vfsgid(&newattrs, gid)) | |
728 | return -EINVAL; | |
5955102c | 729 | inode_lock(inode); |
ed5a7047 CB |
730 | if (!S_ISDIR(inode->i_mode)) |
731 | newattrs.ia_valid |= ATTR_KILL_SUID | ATTR_KILL_PRIV | | |
732 | setattr_should_drop_sgid(mnt_userns, inode); | |
b27c82e1 CB |
733 | /* Continue to send actual fs values, not the mount values. */ |
734 | error = security_path_chown( | |
735 | path, | |
736 | from_vfsuid(mnt_userns, fs_userns, newattrs.ia_vfsuid), | |
737 | from_vfsgid(mnt_userns, fs_userns, newattrs.ia_vfsgid)); | |
fe542cf5 | 738 | if (!error) |
abf08576 | 739 | error = notify_change(idmap, path->dentry, &newattrs, |
2f221d6f | 740 | &delegated_inode); |
5955102c | 741 | inode_unlock(inode); |
27ac0ffe BF |
742 | if (delegated_inode) { |
743 | error = break_deleg_wait(&delegated_inode); | |
744 | if (!error) | |
745 | goto retry_deleg; | |
746 | } | |
1da177e4 LT |
747 | return error; |
748 | } | |
749 | ||
55731b3c DB |
750 | int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group, |
751 | int flag) | |
5590ff0d | 752 | { |
2d8f3038 | 753 | struct path path; |
5590ff0d | 754 | int error = -EINVAL; |
65cfc672 | 755 | int lookup_flags; |
5590ff0d | 756 | |
65cfc672 | 757 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
758 | goto out; |
759 | ||
65cfc672 AV |
760 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
761 | if (flag & AT_EMPTY_PATH) | |
762 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 763 | retry: |
65cfc672 | 764 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
765 | if (error) |
766 | goto out; | |
2d8f3038 | 767 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
768 | if (error) |
769 | goto out_release; | |
fe542cf5 | 770 | error = chown_common(&path, user, group); |
2d8f3038 | 771 | mnt_drop_write(path.mnt); |
2af482a7 | 772 | out_release: |
2d8f3038 | 773 | path_put(&path); |
99a5df37 JL |
774 | if (retry_estale(error, lookup_flags)) { |
775 | lookup_flags |= LOOKUP_REVAL; | |
776 | goto retry; | |
777 | } | |
5590ff0d UD |
778 | out: |
779 | return error; | |
780 | } | |
781 | ||
55731b3c DB |
782 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
783 | gid_t, group, int, flag) | |
784 | { | |
785 | return do_fchownat(dfd, filename, user, group, flag); | |
786 | } | |
787 | ||
55e4def0 | 788 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 789 | { |
55731b3c | 790 | return do_fchownat(AT_FDCWD, filename, user, group, 0); |
55e4def0 | 791 | } |
1da177e4 | 792 | |
55e4def0 DH |
793 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
794 | { | |
55731b3c DB |
795 | return do_fchownat(AT_FDCWD, filename, user, group, |
796 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
797 | } |
798 | ||
c04011fe CH |
799 | int vfs_fchown(struct file *file, uid_t user, gid_t group) |
800 | { | |
801 | int error; | |
802 | ||
803 | error = mnt_want_write_file(file); | |
804 | if (error) | |
805 | return error; | |
806 | audit_file(file); | |
807 | error = chown_common(&file->f_path, user, group); | |
808 | mnt_drop_write_file(file); | |
809 | return error; | |
810 | } | |
811 | ||
55731b3c | 812 | int ksys_fchown(unsigned int fd, uid_t user, gid_t group) |
1da177e4 | 813 | { |
2903ff01 | 814 | struct fd f = fdget(fd); |
1da177e4 LT |
815 | int error = -EBADF; |
816 | ||
c04011fe CH |
817 | if (f.file) { |
818 | error = vfs_fchown(f.file, user, group); | |
819 | fdput(f); | |
820 | } | |
1da177e4 LT |
821 | return error; |
822 | } | |
823 | ||
55731b3c DB |
824 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
825 | { | |
826 | return ksys_fchown(fd, user, group); | |
827 | } | |
828 | ||
02e5180d | 829 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 830 | struct inode *inode, |
ae2bb293 | 831 | int (*open)(struct inode *, struct file *)) |
1da177e4 | 832 | { |
1abf0c71 | 833 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
834 | int error; |
835 | ||
b5bcdda3 | 836 | path_get(&f->f_path); |
4bacc9c9 | 837 | f->f_inode = inode; |
1da177e4 | 838 | f->f_mapping = inode->i_mapping; |
5660e13d | 839 | f->f_wb_err = filemap_sample_wb_err(f->f_mapping); |
735e4ae5 | 840 | f->f_sb_err = file_sample_sb_err(f); |
5660e13d | 841 | |
3f4d5a00 | 842 | if (unlikely(f->f_flags & O_PATH)) { |
f5d11409 | 843 | f->f_mode = FMODE_PATH | FMODE_OPENED; |
1abf0c71 | 844 | f->f_op = &empty_fops; |
af04fadc | 845 | return 0; |
1abf0c71 AV |
846 | } |
847 | ||
d6da19c9 AG |
848 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) { |
849 | i_readcount_inc(inode); | |
850 | } else if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { | |
0ccb2863 | 851 | error = get_write_access(inode); |
3f4d5a00 | 852 | if (unlikely(error)) |
1da177e4 | 853 | goto cleanup_file; |
0ccb2863 | 854 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 855 | if (unlikely(error)) { |
0ccb2863 AV |
856 | put_write_access(inode); |
857 | goto cleanup_file; | |
858 | } | |
83f936c7 | 859 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
860 | } |
861 | ||
2be7d348 LT |
862 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
863 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) | |
864 | f->f_mode |= FMODE_ATOMIC_POS; | |
865 | ||
1abf0c71 | 866 | f->f_op = fops_get(inode->i_fop); |
7159d544 | 867 | if (WARN_ON(!f->f_op)) { |
72c2d531 AV |
868 | error = -ENODEV; |
869 | goto cleanup_all; | |
870 | } | |
1abf0c71 | 871 | |
e3f20ae2 | 872 | error = security_file_open(f); |
788e7dd4 YN |
873 | if (error) |
874 | goto cleanup_all; | |
875 | ||
c568d683 | 876 | error = break_lease(locks_inode(f), f->f_flags); |
f3c7691e BF |
877 | if (error) |
878 | goto cleanup_all; | |
879 | ||
ea73ea72 AV |
880 | /* normally all 3 are set; ->open() can clear them if needed */ |
881 | f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE; | |
72c2d531 | 882 | if (!open) |
834f2a4a TM |
883 | open = f->f_op->open; |
884 | if (open) { | |
885 | error = open(inode, f); | |
1da177e4 LT |
886 | if (error) |
887 | goto cleanup_all; | |
888 | } | |
f5d11409 | 889 | f->f_mode |= FMODE_OPENED; |
293bc982 | 890 | if ((f->f_mode & FMODE_READ) && |
84363182 | 891 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 892 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 893 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 894 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 895 | f->f_mode |= FMODE_CAN_WRITE; |
e7478158 JD |
896 | if ((f->f_mode & FMODE_LSEEK) && !f->f_op->llseek) |
897 | f->f_mode &= ~FMODE_LSEEK; | |
a2ad63da N |
898 | if (f->f_mapping->a_ops && f->f_mapping->a_ops->direct_IO) |
899 | f->f_mode |= FMODE_CAN_ODIRECT; | |
834f2a4a | 900 | |
1da177e4 | 901 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
164f4064 | 902 | f->f_iocb_flags = iocb_flags(f); |
1da177e4 LT |
903 | |
904 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
af04fadc | 905 | |
a2ad63da N |
906 | if ((f->f_flags & O_DIRECT) && !(f->f_mode & FMODE_CAN_ODIRECT)) |
907 | return -EINVAL; | |
09d91cda SL |
908 | |
909 | /* | |
910 | * XXX: Huge page cache doesn't support writing yet. Drop all page | |
911 | * cache for this file before processing writes. | |
912 | */ | |
eb6ecbed CF |
913 | if (f->f_mode & FMODE_WRITE) { |
914 | /* | |
915 | * Paired with smp_mb() in collapse_file() to ensure nr_thps | |
916 | * is up to date and the update to i_writecount by | |
917 | * get_write_access() is visible. Ensures subsequent insertion | |
918 | * of THPs into the page cache will fail. | |
919 | */ | |
920 | smp_mb(); | |
55fc0d91 | 921 | if (filemap_nr_thps(inode->i_mapping)) { |
8468e937 RW |
922 | struct address_space *mapping = inode->i_mapping; |
923 | ||
55fc0d91 | 924 | filemap_invalidate_lock(inode->i_mapping); |
8468e937 RW |
925 | /* |
926 | * unmap_mapping_range just need to be called once | |
927 | * here, because the private pages is not need to be | |
928 | * unmapped mapping (e.g. data segment of dynamic | |
929 | * shared libraries here). | |
930 | */ | |
931 | unmap_mapping_range(mapping, 0, 0, 0); | |
932 | truncate_inode_pages(mapping, 0); | |
55fc0d91 RW |
933 | filemap_invalidate_unlock(inode->i_mapping); |
934 | } | |
eb6ecbed | 935 | } |
09d91cda | 936 | |
96b7e579 | 937 | return 0; |
1da177e4 LT |
938 | |
939 | cleanup_all: | |
6b4e8085 AV |
940 | if (WARN_ON_ONCE(error > 0)) |
941 | error = -EINVAL; | |
1da177e4 | 942 | fops_put(f->f_op); |
d6da19c9 | 943 | put_file_access(f); |
1da177e4 | 944 | cleanup_file: |
02e5180d AV |
945 | path_put(&f->f_path); |
946 | f->f_path.mnt = NULL; | |
947 | f->f_path.dentry = NULL; | |
dd37978c | 948 | f->f_inode = NULL; |
96b7e579 | 949 | return error; |
1da177e4 LT |
950 | } |
951 | ||
d18e9008 MS |
952 | /** |
953 | * finish_open - finish opening a file | |
0854d450 | 954 | * @file: file pointer |
d18e9008 MS |
955 | * @dentry: pointer to dentry |
956 | * @open: open callback | |
0854d450 | 957 | * @opened: state of open |
d18e9008 MS |
958 | * |
959 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
960 | * | |
961 | * If the open callback is set to NULL, then the standard f_op->open() | |
962 | * filesystem callback is substituted. | |
0854d450 MS |
963 | * |
964 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
965 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
966 | * on it after finish_open(). | |
967 | * | |
0854d450 | 968 | * Returns zero on success or -errno if the open failed. |
d18e9008 | 969 | */ |
30d90494 | 970 | int finish_open(struct file *file, struct dentry *dentry, |
be12af3e | 971 | int (*open)(struct inode *, struct file *)) |
d18e9008 | 972 | { |
aad888f8 | 973 | BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 974 | |
b5bcdda3 | 975 | file->f_path.dentry = dentry; |
aad888f8 | 976 | return do_dentry_open(file, d_backing_inode(dentry), open); |
d18e9008 MS |
977 | } |
978 | EXPORT_SYMBOL(finish_open); | |
979 | ||
980 | /** | |
981 | * finish_no_open - finish ->atomic_open() without opening the file | |
982 | * | |
0854d450 | 983 | * @file: file pointer |
d18e9008 MS |
984 | * @dentry: dentry or NULL (as returned from ->lookup()) |
985 | * | |
986 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
987 | * |
988 | * NB: unlike finish_open() this function does consume the dentry reference and | |
989 | * the caller need not dput() it. | |
990 | * | |
64e1ac4d | 991 | * Returns "0" which must be the return value of ->atomic_open() after having |
0854d450 | 992 | * called this function. |
d18e9008 | 993 | */ |
e45198a6 | 994 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 995 | { |
30d90494 | 996 | file->f_path.dentry = dentry; |
64e1ac4d | 997 | return 0; |
d18e9008 MS |
998 | } |
999 | EXPORT_SYMBOL(finish_no_open); | |
1000 | ||
9bf39ab2 MS |
1001 | char *file_path(struct file *filp, char *buf, int buflen) |
1002 | { | |
1003 | return d_path(&filp->f_path, buf, buflen); | |
1004 | } | |
1005 | EXPORT_SYMBOL(file_path); | |
1006 | ||
4bacc9c9 DH |
1007 | /** |
1008 | * vfs_open - open the file at the given path | |
1009 | * @path: path to open | |
1010 | * @file: newly allocated file with f_flag initialized | |
1011 | * @cred: credentials to use | |
1012 | */ | |
ae2bb293 | 1013 | int vfs_open(const struct path *path, struct file *file) |
4bacc9c9 | 1014 | { |
54d5ca87 | 1015 | file->f_path = *path; |
a6518f73 | 1016 | return do_dentry_open(file, d_backing_inode(path->dentry), NULL); |
4bacc9c9 DH |
1017 | } |
1018 | ||
765927b2 | 1019 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 1020 | const struct cred *cred) |
a1a5b3d9 PS |
1021 | { |
1022 | int error; | |
1023 | struct file *f; | |
1024 | ||
e0e81739 DH |
1025 | validate_creds(cred); |
1026 | ||
c212f9aa | 1027 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 1028 | BUG_ON(!path->mnt); |
322ee5b3 | 1029 | |
ea73ea72 | 1030 | f = alloc_empty_file(flags, cred); |
af04fadc | 1031 | if (!IS_ERR(f)) { |
ae2bb293 | 1032 | error = vfs_open(path, f); |
4d27f326 AV |
1033 | if (error) { |
1034 | fput(f); | |
af04fadc AV |
1035 | f = ERR_PTR(error); |
1036 | } | |
2a027e7a AV |
1037 | } |
1038 | return f; | |
a1a5b3d9 | 1039 | } |
1da177e4 LT |
1040 | EXPORT_SYMBOL(dentry_open); |
1041 | ||
fb70bf12 CL |
1042 | /** |
1043 | * dentry_create - Create and open a file | |
1044 | * @path: path to create | |
1045 | * @flags: O_ flags | |
1046 | * @mode: mode bits for new file | |
1047 | * @cred: credentials to use | |
1048 | * | |
1049 | * Caller must hold the parent directory's lock, and have prepared | |
1050 | * a negative dentry, placed in @path->dentry, for the new file. | |
1051 | * | |
1052 | * Caller sets @path->mnt to the vfsmount of the filesystem where | |
1053 | * the new file is to be created. The parent directory and the | |
1054 | * negative dentry must reside on the same filesystem instance. | |
1055 | * | |
1056 | * On success, returns a "struct file *". Otherwise a ERR_PTR | |
1057 | * is returned. | |
1058 | */ | |
1059 | struct file *dentry_create(const struct path *path, int flags, umode_t mode, | |
1060 | const struct cred *cred) | |
1061 | { | |
1062 | struct file *f; | |
1063 | int error; | |
1064 | ||
1065 | validate_creds(cred); | |
1066 | f = alloc_empty_file(flags, cred); | |
1067 | if (IS_ERR(f)) | |
1068 | return f; | |
1069 | ||
abf08576 | 1070 | error = vfs_create(mnt_idmap(path->mnt), |
fb70bf12 CL |
1071 | d_inode(path->dentry->d_parent), |
1072 | path->dentry, mode, true); | |
1073 | if (!error) | |
1074 | error = vfs_open(path, f); | |
1075 | ||
1076 | if (unlikely(error)) { | |
1077 | fput(f); | |
1078 | return ERR_PTR(error); | |
1079 | } | |
1080 | return f; | |
1081 | } | |
1082 | EXPORT_SYMBOL(dentry_create); | |
1083 | ||
2abc77af AV |
1084 | struct file *open_with_fake_path(const struct path *path, int flags, |
1085 | struct inode *inode, const struct cred *cred) | |
1086 | { | |
d3b1084d | 1087 | struct file *f = alloc_empty_file_noaccount(flags, cred); |
2abc77af AV |
1088 | if (!IS_ERR(f)) { |
1089 | int error; | |
1090 | ||
1091 | f->f_path = *path; | |
1092 | error = do_dentry_open(f, inode, NULL); | |
1093 | if (error) { | |
1094 | fput(f); | |
1095 | f = ERR_PTR(error); | |
1096 | } | |
1097 | } | |
1098 | return f; | |
1099 | } | |
1100 | EXPORT_SYMBOL(open_with_fake_path); | |
1101 | ||
fddb5d43 AS |
1102 | #define WILL_CREATE(flags) (flags & (O_CREAT | __O_TMPFILE)) |
1103 | #define O_PATH_FLAGS (O_DIRECTORY | O_NOFOLLOW | O_PATH | O_CLOEXEC) | |
1104 | ||
35cb6d54 | 1105 | inline struct open_how build_open_how(int flags, umode_t mode) |
fddb5d43 AS |
1106 | { |
1107 | struct open_how how = { | |
1108 | .flags = flags & VALID_OPEN_FLAGS, | |
1109 | .mode = mode & S_IALLUGO, | |
1110 | }; | |
1111 | ||
1112 | /* O_PATH beats everything else. */ | |
1113 | if (how.flags & O_PATH) | |
1114 | how.flags &= O_PATH_FLAGS; | |
1115 | /* Modes should only be set for create-like flags. */ | |
1116 | if (!WILL_CREATE(how.flags)) | |
1117 | how.mode = 0; | |
1118 | return how; | |
1119 | } | |
1120 | ||
35cb6d54 | 1121 | inline int build_open_flags(const struct open_how *how, struct open_flags *op) |
47c805dc | 1122 | { |
cfe80306 CB |
1123 | u64 flags = how->flags; |
1124 | u64 strip = FMODE_NONOTIFY | O_CLOEXEC; | |
47c805dc | 1125 | int lookup_flags = 0; |
62fb4a15 | 1126 | int acc_mode = ACC_MODE(flags); |
47c805dc | 1127 | |
cfe80306 CB |
1128 | BUILD_BUG_ON_MSG(upper_32_bits(VALID_OPEN_FLAGS), |
1129 | "struct open_flags doesn't yet handle flags > 32 bits"); | |
1130 | ||
1131 | /* | |
1132 | * Strip flags that either shouldn't be set by userspace like | |
1133 | * FMODE_NONOTIFY or that aren't relevant in determining struct | |
1134 | * open_flags like O_CLOEXEC. | |
1135 | */ | |
1136 | flags &= ~strip; | |
fddb5d43 | 1137 | |
629e014b | 1138 | /* |
fddb5d43 AS |
1139 | * Older syscalls implicitly clear all of the invalid flags or argument |
1140 | * values before calling build_open_flags(), but openat2(2) checks all | |
1141 | * of its arguments. | |
629e014b | 1142 | */ |
fddb5d43 AS |
1143 | if (flags & ~VALID_OPEN_FLAGS) |
1144 | return -EINVAL; | |
1145 | if (how->resolve & ~VALID_RESOLVE_FLAGS) | |
1146 | return -EINVAL; | |
629e014b | 1147 | |
398840f8 AS |
1148 | /* Scoping flags are mutually exclusive. */ |
1149 | if ((how->resolve & RESOLVE_BENEATH) && (how->resolve & RESOLVE_IN_ROOT)) | |
1150 | return -EINVAL; | |
1151 | ||
fddb5d43 AS |
1152 | /* Deal with the mode. */ |
1153 | if (WILL_CREATE(flags)) { | |
1154 | if (how->mode & ~S_IALLUGO) | |
1155 | return -EINVAL; | |
1156 | op->mode = how->mode | S_IFREG; | |
1157 | } else { | |
1158 | if (how->mode != 0) | |
1159 | return -EINVAL; | |
e68726ff | 1160 | op->mode = 0; |
fddb5d43 | 1161 | } |
47c805dc AV |
1162 | |
1163 | /* | |
fddb5d43 AS |
1164 | * In order to ensure programs get explicit errors when trying to use |
1165 | * O_TMPFILE on old kernels, O_TMPFILE is implemented such that it | |
1166 | * looks like (O_DIRECTORY|O_RDWR & ~O_CREAT) to old kernels. But we | |
1167 | * have to require userspace to explicitly set it. | |
47c805dc | 1168 | */ |
bb458c64 AV |
1169 | if (flags & __O_TMPFILE) { |
1170 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d | 1171 | return -EINVAL; |
ba57ea64 AV |
1172 | if (!(acc_mode & MAY_WRITE)) |
1173 | return -EINVAL; | |
fddb5d43 AS |
1174 | } |
1175 | if (flags & O_PATH) { | |
1176 | /* O_PATH only permits certain other flags to be set. */ | |
1177 | if (flags & ~O_PATH_FLAGS) | |
1178 | return -EINVAL; | |
1abf0c71 | 1179 | acc_mode = 0; |
1abf0c71 | 1180 | } |
47c805dc | 1181 | |
fddb5d43 AS |
1182 | /* |
1183 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1184 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1185 | * always set instead of having to deal with possibly weird behaviour | |
1186 | * for malicious applications setting only __O_SYNC. | |
1187 | */ | |
1188 | if (flags & __O_SYNC) | |
1189 | flags |= O_DSYNC; | |
1190 | ||
1abf0c71 | 1191 | op->open_flag = flags; |
47c805dc AV |
1192 | |
1193 | /* O_TRUNC implies we need access checks for write permissions */ | |
1194 | if (flags & O_TRUNC) | |
1195 | acc_mode |= MAY_WRITE; | |
1196 | ||
1197 | /* Allow the LSM permission hook to distinguish append | |
1198 | access from general write access. */ | |
1199 | if (flags & O_APPEND) | |
1200 | acc_mode |= MAY_APPEND; | |
1201 | ||
1202 | op->acc_mode = acc_mode; | |
1203 | ||
1abf0c71 AV |
1204 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
1205 | ||
47c805dc AV |
1206 | if (flags & O_CREAT) { |
1207 | op->intent |= LOOKUP_CREATE; | |
31d1726d | 1208 | if (flags & O_EXCL) { |
47c805dc | 1209 | op->intent |= LOOKUP_EXCL; |
31d1726d AV |
1210 | flags |= O_NOFOLLOW; |
1211 | } | |
47c805dc AV |
1212 | } |
1213 | ||
1214 | if (flags & O_DIRECTORY) | |
1215 | lookup_flags |= LOOKUP_DIRECTORY; | |
1216 | if (!(flags & O_NOFOLLOW)) | |
1217 | lookup_flags |= LOOKUP_FOLLOW; | |
fddb5d43 AS |
1218 | |
1219 | if (how->resolve & RESOLVE_NO_XDEV) | |
1220 | lookup_flags |= LOOKUP_NO_XDEV; | |
1221 | if (how->resolve & RESOLVE_NO_MAGICLINKS) | |
1222 | lookup_flags |= LOOKUP_NO_MAGICLINKS; | |
1223 | if (how->resolve & RESOLVE_NO_SYMLINKS) | |
1224 | lookup_flags |= LOOKUP_NO_SYMLINKS; | |
1225 | if (how->resolve & RESOLVE_BENEATH) | |
1226 | lookup_flags |= LOOKUP_BENEATH; | |
1227 | if (how->resolve & RESOLVE_IN_ROOT) | |
1228 | lookup_flags |= LOOKUP_IN_ROOT; | |
99668f61 JA |
1229 | if (how->resolve & RESOLVE_CACHED) { |
1230 | /* Don't bother even trying for create/truncate/tmpfile open */ | |
1231 | if (flags & (O_TRUNC | O_CREAT | O_TMPFILE)) | |
1232 | return -EAGAIN; | |
1233 | lookup_flags |= LOOKUP_CACHED; | |
1234 | } | |
fddb5d43 | 1235 | |
f9652e10 AV |
1236 | op->lookup_flags = lookup_flags; |
1237 | return 0; | |
47c805dc AV |
1238 | } |
1239 | ||
669abf4e JL |
1240 | /** |
1241 | * file_open_name - open file and return file pointer | |
1242 | * | |
1243 | * @name: struct filename containing path to open | |
1244 | * @flags: open flags as per the open(2) second argument | |
1245 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1246 | * | |
1247 | * This is the helper to open a file from kernelspace if you really | |
1248 | * have to. But in generally you should not do this, so please move | |
1249 | * along, nothing to see here.. | |
1250 | */ | |
1251 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
1252 | { | |
1253 | struct open_flags op; | |
fddb5d43 AS |
1254 | struct open_how how = build_open_how(flags, mode); |
1255 | int err = build_open_flags(&how, &op); | |
1256 | if (err) | |
1257 | return ERR_PTR(err); | |
1258 | return do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
1259 | } |
1260 | ||
47c805dc AV |
1261 | /** |
1262 | * filp_open - open file and return file pointer | |
1263 | * | |
1264 | * @filename: path to open | |
1265 | * @flags: open flags as per the open(2) second argument | |
1266 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1267 | * | |
1268 | * This is the helper to open a file from kernelspace if you really | |
1269 | * have to. But in generally you should not do this, so please move | |
1270 | * along, nothing to see here.. | |
1271 | */ | |
a218d0fd | 1272 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 1273 | { |
51689104 PM |
1274 | struct filename *name = getname_kernel(filename); |
1275 | struct file *file = ERR_CAST(name); | |
1276 | ||
1277 | if (!IS_ERR(name)) { | |
1278 | file = file_open_name(name, flags, mode); | |
1279 | putname(name); | |
1280 | } | |
1281 | return file; | |
47c805dc AV |
1282 | } |
1283 | EXPORT_SYMBOL(filp_open); | |
1284 | ||
ffb37ca3 | 1285 | struct file *file_open_root(const struct path *root, |
378c6520 | 1286 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1287 | { |
1288 | struct open_flags op; | |
fddb5d43 AS |
1289 | struct open_how how = build_open_how(flags, mode); |
1290 | int err = build_open_flags(&how, &op); | |
f9652e10 AV |
1291 | if (err) |
1292 | return ERR_PTR(err); | |
ffb37ca3 | 1293 | return do_file_open_root(root, filename, &op); |
73d049a4 AV |
1294 | } |
1295 | EXPORT_SYMBOL(file_open_root); | |
1296 | ||
fddb5d43 AS |
1297 | static long do_sys_openat2(int dfd, const char __user *filename, |
1298 | struct open_how *how) | |
1da177e4 | 1299 | { |
47c805dc | 1300 | struct open_flags op; |
fddb5d43 | 1301 | int fd = build_open_flags(how, &op); |
f9652e10 AV |
1302 | struct filename *tmp; |
1303 | ||
1304 | if (fd) | |
1305 | return fd; | |
1306 | ||
1307 | tmp = getname(filename); | |
1308 | if (IS_ERR(tmp)) | |
1309 | return PTR_ERR(tmp); | |
1310 | ||
fddb5d43 | 1311 | fd = get_unused_fd_flags(how->flags); |
f9652e10 AV |
1312 | if (fd >= 0) { |
1313 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1314 | if (IS_ERR(f)) { | |
1315 | put_unused_fd(fd); | |
1316 | fd = PTR_ERR(f); | |
1317 | } else { | |
1318 | fsnotify_open(f); | |
1319 | fd_install(fd, f); | |
1da177e4 | 1320 | } |
1da177e4 | 1321 | } |
f9652e10 | 1322 | putname(tmp); |
1da177e4 | 1323 | return fd; |
1da177e4 | 1324 | } |
e922efc3 | 1325 | |
fddb5d43 | 1326 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
e922efc3 | 1327 | { |
fddb5d43 AS |
1328 | struct open_how how = build_open_how(flags, mode); |
1329 | return do_sys_openat2(dfd, filename, &how); | |
1330 | } | |
e922efc3 | 1331 | |
fddb5d43 AS |
1332 | |
1333 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1334 | { | |
166e07c3 CH |
1335 | if (force_o_largefile()) |
1336 | flags |= O_LARGEFILE; | |
1337 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
e922efc3 | 1338 | } |
1da177e4 | 1339 | |
6559eed8 | 1340 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1341 | umode_t, mode) |
5590ff0d UD |
1342 | { |
1343 | if (force_o_largefile()) | |
1344 | flags |= O_LARGEFILE; | |
2cf09666 | 1345 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1346 | } |
5590ff0d | 1347 | |
fddb5d43 AS |
1348 | SYSCALL_DEFINE4(openat2, int, dfd, const char __user *, filename, |
1349 | struct open_how __user *, how, size_t, usize) | |
1350 | { | |
1351 | int err; | |
1352 | struct open_how tmp; | |
1353 | ||
1354 | BUILD_BUG_ON(sizeof(struct open_how) < OPEN_HOW_SIZE_VER0); | |
1355 | BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_LATEST); | |
1356 | ||
1357 | if (unlikely(usize < OPEN_HOW_SIZE_VER0)) | |
1358 | return -EINVAL; | |
1359 | ||
1360 | err = copy_struct_from_user(&tmp, sizeof(tmp), how, usize); | |
1361 | if (err) | |
1362 | return err; | |
1363 | ||
571e5c0e RGB |
1364 | audit_openat2_how(&tmp); |
1365 | ||
fddb5d43 AS |
1366 | /* O_LARGEFILE is only allowed for non-O_PATH. */ |
1367 | if (!(tmp.flags & O_PATH) && force_o_largefile()) | |
1368 | tmp.flags |= O_LARGEFILE; | |
1369 | ||
1370 | return do_sys_openat2(dfd, filename, &tmp); | |
1371 | } | |
1372 | ||
e35d49f6 AV |
1373 | #ifdef CONFIG_COMPAT |
1374 | /* | |
1375 | * Exactly like sys_open(), except that it doesn't set the | |
1376 | * O_LARGEFILE flag. | |
1377 | */ | |
1378 | COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1379 | { | |
1380 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
1381 | } | |
1382 | ||
1383 | /* | |
1384 | * Exactly like sys_openat(), except that it doesn't set the | |
1385 | * O_LARGEFILE flag. | |
1386 | */ | |
1387 | COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode) | |
1388 | { | |
1389 | return do_sys_open(dfd, filename, flags, mode); | |
1390 | } | |
1391 | #endif | |
1392 | ||
1da177e4 LT |
1393 | #ifndef __alpha__ |
1394 | ||
1395 | /* | |
1396 | * For backward compatibility? Maybe this should be moved | |
1397 | * into arch/i386 instead? | |
1398 | */ | |
a218d0fd | 1399 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 | 1400 | { |
166e07c3 | 1401 | int flags = O_CREAT | O_WRONLY | O_TRUNC; |
1da177e4 | 1402 | |
166e07c3 CH |
1403 | if (force_o_largefile()) |
1404 | flags |= O_LARGEFILE; | |
1405 | return do_sys_open(AT_FDCWD, pathname, flags, mode); | |
1406 | } | |
1da177e4 LT |
1407 | #endif |
1408 | ||
1409 | /* | |
1410 | * "id" is the POSIX thread ID. We use the | |
1411 | * files pointer for this.. | |
1412 | */ | |
1413 | int filp_close(struct file *filp, fl_owner_t id) | |
1414 | { | |
45778ca8 | 1415 | int retval = 0; |
1da177e4 LT |
1416 | |
1417 | if (!file_count(filp)) { | |
1418 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1419 | return 0; |
1da177e4 LT |
1420 | } |
1421 | ||
72c2d531 | 1422 | if (filp->f_op->flush) |
75e1fcc0 | 1423 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1424 | |
1abf0c71 AV |
1425 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1426 | dnotify_flush(filp, id); | |
1427 | locks_remove_posix(filp, id); | |
1428 | } | |
1da177e4 LT |
1429 | fput(filp); |
1430 | return retval; | |
1431 | } | |
1432 | ||
1433 | EXPORT_SYMBOL(filp_close); | |
1434 | ||
1435 | /* | |
1436 | * Careful here! We test whether the file pointer is NULL before | |
1437 | * releasing the fd. This ensures that one clone task can't release | |
1438 | * an fd while another clone is opening it. | |
1439 | */ | |
ca013e94 | 1440 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1441 | { |
8760c909 | 1442 | int retval = close_fd(fd); |
ee731f4f EP |
1443 | |
1444 | /* can't restart close syscall because file table entry was cleared */ | |
1445 | if (unlikely(retval == -ERESTARTSYS || | |
1446 | retval == -ERESTARTNOINTR || | |
1447 | retval == -ERESTARTNOHAND || | |
1448 | retval == -ERESTART_RESTARTBLOCK)) | |
1449 | retval = -EINTR; | |
1450 | ||
1451 | return retval; | |
1da177e4 | 1452 | } |
1da177e4 | 1453 | |
278a5fba CB |
1454 | /** |
1455 | * close_range() - Close all file descriptors in a given range. | |
1456 | * | |
1457 | * @fd: starting file descriptor to close | |
1458 | * @max_fd: last file descriptor to close | |
1459 | * @flags: reserved for future extensions | |
1460 | * | |
1461 | * This closes a range of file descriptors. All file descriptors | |
1462 | * from @fd up to and including @max_fd are closed. | |
1463 | * Currently, errors to close a given file descriptor are ignored. | |
1464 | */ | |
1465 | SYSCALL_DEFINE3(close_range, unsigned int, fd, unsigned int, max_fd, | |
1466 | unsigned int, flags) | |
1467 | { | |
60997c3d | 1468 | return __close_range(fd, max_fd, flags); |
278a5fba CB |
1469 | } |
1470 | ||
1da177e4 LT |
1471 | /* |
1472 | * This routine simulates a hangup on the tty, to arrange that users | |
1473 | * are given clean terminals at login time. | |
1474 | */ | |
ca013e94 | 1475 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1476 | { |
1477 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1478 | tty_vhangup_self(); |
1da177e4 LT |
1479 | return 0; |
1480 | } | |
1481 | return -EPERM; | |
1482 | } | |
1483 | ||
1484 | /* | |
1485 | * Called when an inode is about to be open. | |
1486 | * We use this to disallow opening large files on 32bit systems if | |
1487 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1488 | * on this flag in sys_open. | |
1489 | */ | |
1490 | int generic_file_open(struct inode * inode, struct file * filp) | |
1491 | { | |
1492 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1493 | return -EOVERFLOW; |
1da177e4 LT |
1494 | return 0; |
1495 | } | |
1496 | ||
1497 | EXPORT_SYMBOL(generic_file_open); | |
1498 | ||
1499 | /* | |
1500 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1501 | * file descriptors. The function is not supposed to ever fail, the only |
1502 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1503 | * directly into file_operations structure. | |
1da177e4 LT |
1504 | */ |
1505 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1506 | { | |
1507 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1508 | return 0; | |
1509 | } | |
1510 | ||
1511 | EXPORT_SYMBOL(nonseekable_open); | |
10dce8af KS |
1512 | |
1513 | /* | |
1514 | * stream_open is used by subsystems that want stream-like file descriptors. | |
1515 | * Such file descriptors are not seekable and don't have notion of position | |
438ab720 KS |
1516 | * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL). |
1517 | * Contrary to file descriptors of other regular files, .read() and .write() | |
1518 | * can run simultaneously. | |
10dce8af KS |
1519 | * |
1520 | * stream_open never fails and is marked to return int so that it could be | |
1521 | * directly used as file_operations.open . | |
1522 | */ | |
1523 | int stream_open(struct inode *inode, struct file *filp) | |
1524 | { | |
2be7d348 | 1525 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS); |
10dce8af KS |
1526 | filp->f_mode |= FMODE_STREAM; |
1527 | return 0; | |
1528 | } | |
1529 | ||
1530 | EXPORT_SYMBOL(stream_open); |