]>
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> |
1da177e4 | 35 | |
e81e3f4d EP |
36 | #include "internal.h" |
37 | ||
4a30131e N |
38 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, |
39 | struct file *filp) | |
1da177e4 | 40 | { |
939a9421 | 41 | int ret; |
1da177e4 LT |
42 | struct iattr newattrs; |
43 | ||
44 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
45 | if (length < 0) | |
46 | return -EINVAL; | |
47 | ||
48 | newattrs.ia_size = length; | |
4a30131e | 49 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
50 | if (filp) { |
51 | newattrs.ia_file = filp; | |
52 | newattrs.ia_valid |= ATTR_FILE; | |
53 | } | |
1da177e4 | 54 | |
45f147a1 JK |
55 | /* Remove suid, sgid, and file capabilities on truncate too */ |
56 | ret = dentry_needs_remove_privs(dentry); | |
57 | if (ret < 0) | |
58 | return ret; | |
939a9421 AW |
59 | if (ret) |
60 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 61 | |
5955102c | 62 | inode_lock(dentry->d_inode); |
27ac0ffe BF |
63 | /* Note any delegations or leases have already been broken: */ |
64 | ret = notify_change(dentry, &newattrs, NULL); | |
5955102c | 65 | inode_unlock(dentry->d_inode); |
939a9421 | 66 | return ret; |
1da177e4 | 67 | } |
0d331203 | 68 | EXPORT_SYMBOL_GPL(do_truncate); |
1da177e4 | 69 | |
7df818b2 | 70 | long vfs_truncate(const struct path *path, loff_t length) |
1da177e4 | 71 | { |
2d8f3038 | 72 | struct inode *inode; |
a02de960 | 73 | long error; |
1da177e4 | 74 | |
a02de960 | 75 | inode = path->dentry->d_inode; |
1da177e4 LT |
76 | |
77 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
1da177e4 | 78 | if (S_ISDIR(inode->i_mode)) |
a02de960 | 79 | return -EISDIR; |
1da177e4 | 80 | if (!S_ISREG(inode->i_mode)) |
a02de960 | 81 | return -EINVAL; |
1da177e4 | 82 | |
a02de960 | 83 | error = mnt_want_write(path->mnt); |
1da177e4 | 84 | if (error) |
a02de960 | 85 | goto out; |
1da177e4 | 86 | |
256984a8 | 87 | error = inode_permission(inode, MAY_WRITE); |
9ac9b847 DH |
88 | if (error) |
89 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
90 | |
91 | error = -EPERM; | |
c82e42da | 92 | if (IS_APPEND(inode)) |
9ac9b847 | 93 | goto mnt_drop_write_and_out; |
1da177e4 | 94 | |
8cf9ee50 | 95 | error = get_write_access(inode); |
1da177e4 | 96 | if (error) |
9ac9b847 | 97 | goto mnt_drop_write_and_out; |
1da177e4 | 98 | |
9700382c | 99 | /* |
100 | * Make sure that there are no leases. get_write_access() protects | |
101 | * against the truncate racing with a lease-granting setlease(). | |
102 | */ | |
8737c930 | 103 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 104 | if (error) |
9700382c | 105 | goto put_write_and_out; |
1da177e4 LT |
106 | |
107 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 | 108 | if (!error) |
a02de960 | 109 | error = security_path_truncate(path); |
907f4554 | 110 | if (!error) |
a02de960 | 111 | error = do_truncate(path->dentry, length, 0, NULL); |
1da177e4 | 112 | |
9700382c | 113 | put_write_and_out: |
8cf9ee50 | 114 | put_write_access(inode); |
9ac9b847 | 115 | mnt_drop_write_and_out: |
a02de960 | 116 | mnt_drop_write(path->mnt); |
1da177e4 LT |
117 | out: |
118 | return error; | |
119 | } | |
a02de960 DH |
120 | EXPORT_SYMBOL_GPL(vfs_truncate); |
121 | ||
df260e21 | 122 | long do_sys_truncate(const char __user *pathname, loff_t length) |
a02de960 | 123 | { |
48f7530d | 124 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
125 | struct path path; |
126 | int error; | |
127 | ||
128 | if (length < 0) /* sorry, but loff_t says... */ | |
129 | return -EINVAL; | |
130 | ||
48f7530d JL |
131 | retry: |
132 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
133 | if (!error) { |
134 | error = vfs_truncate(&path, length); | |
135 | path_put(&path); | |
136 | } | |
48f7530d JL |
137 | if (retry_estale(error, lookup_flags)) { |
138 | lookup_flags |= LOOKUP_REVAL; | |
139 | goto retry; | |
140 | } | |
a02de960 DH |
141 | return error; |
142 | } | |
1da177e4 | 143 | |
4fd8da8d | 144 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 145 | { |
4fd8da8d | 146 | return do_sys_truncate(path, length); |
1da177e4 LT |
147 | } |
148 | ||
3f6d078d AV |
149 | #ifdef CONFIG_COMPAT |
150 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
151 | { | |
152 | return do_sys_truncate(path, length); | |
153 | } | |
154 | #endif | |
155 | ||
411d9475 | 156 | long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 | 157 | { |
bf2965d5 | 158 | struct inode *inode; |
1da177e4 | 159 | struct dentry *dentry; |
2903ff01 | 160 | struct fd f; |
1da177e4 LT |
161 | int error; |
162 | ||
163 | error = -EINVAL; | |
164 | if (length < 0) | |
165 | goto out; | |
166 | error = -EBADF; | |
2903ff01 AV |
167 | f = fdget(fd); |
168 | if (!f.file) | |
1da177e4 LT |
169 | goto out; |
170 | ||
171 | /* explicitly opened as large or we are on 64-bit box */ | |
2903ff01 | 172 | if (f.file->f_flags & O_LARGEFILE) |
1da177e4 LT |
173 | small = 0; |
174 | ||
2903ff01 | 175 | dentry = f.file->f_path.dentry; |
1da177e4 LT |
176 | inode = dentry->d_inode; |
177 | error = -EINVAL; | |
2903ff01 | 178 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
179 | goto out_putf; |
180 | ||
181 | error = -EINVAL; | |
182 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
183 | if (small && length > MAX_NON_LFS) | |
184 | goto out_putf; | |
185 | ||
186 | error = -EPERM; | |
78757af6 AG |
187 | /* Check IS_APPEND on real upper inode */ |
188 | if (IS_APPEND(file_inode(f.file))) | |
1da177e4 LT |
189 | goto out_putf; |
190 | ||
14da9200 | 191 | sb_start_write(inode->i_sb); |
2903ff01 | 192 | error = locks_verify_truncate(inode, f.file, length); |
be6d3e56 | 193 | if (!error) |
2903ff01 | 194 | error = security_path_truncate(&f.file->f_path); |
1da177e4 | 195 | if (!error) |
2903ff01 | 196 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file); |
14da9200 | 197 | sb_end_write(inode->i_sb); |
1da177e4 | 198 | out_putf: |
2903ff01 | 199 | fdput(f); |
1da177e4 LT |
200 | out: |
201 | return error; | |
202 | } | |
203 | ||
bdc480e3 | 204 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 205 | { |
2cf09666 | 206 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
207 | } |
208 | ||
3f6d078d AV |
209 | #ifdef CONFIG_COMPAT |
210 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
211 | { | |
212 | return do_sys_ftruncate(fd, length, 1); | |
213 | } | |
214 | #endif | |
215 | ||
1da177e4 LT |
216 | /* LFS versions of truncate are only needed on 32 bit machines */ |
217 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 218 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
219 | { |
220 | return do_sys_truncate(path, length); | |
221 | } | |
222 | ||
4a0fd5bf | 223 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 224 | { |
2cf09666 | 225 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 226 | } |
6673e0c3 | 227 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 228 | |
3e63cbb1 | 229 | |
72c72bdf | 230 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 231 | { |
496ad9aa | 232 | struct inode *inode = file_inode(file); |
3e63cbb1 | 233 | long ret; |
97ac7350 AA |
234 | |
235 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 236 | return -EINVAL; |
97ac7350 AA |
237 | |
238 | /* Return error if mode is not supported */ | |
dd46c787 | 239 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) |
409332b6 LC |
240 | return -EOPNOTSUPP; |
241 | ||
242 | /* Punch hole and zero range are mutually exclusive */ | |
243 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
244 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
245 | return -EOPNOTSUPP; |
246 | ||
247 | /* Punch hole must have keep size set */ | |
248 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
249 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 250 | return -EOPNOTSUPP; |
97ac7350 | 251 | |
00f5e619 NJ |
252 | /* Collapse range should only be used exclusively. */ |
253 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
254 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
255 | return -EINVAL; | |
256 | ||
dd46c787 NJ |
257 | /* Insert range should only be used exclusively. */ |
258 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
259 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
260 | return -EINVAL; | |
261 | ||
71be6b49 DW |
262 | /* Unshare range should only be used with allocate mode. */ |
263 | if ((mode & FALLOC_FL_UNSHARE_RANGE) && | |
264 | (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE))) | |
265 | return -EINVAL; | |
266 | ||
97ac7350 | 267 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 268 | return -EBADF; |
1ca551c6 | 269 | |
00f5e619 | 270 | /* |
8fc61d92 | 271 | * We can only allow pure fallocate on append only files |
00f5e619 | 272 | */ |
8fc61d92 | 273 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
274 | return -EPERM; |
275 | ||
276 | if (IS_IMMUTABLE(inode)) | |
277 | return -EPERM; | |
278 | ||
0790b31b | 279 | /* |
6d2b6170 | 280 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
281 | */ |
282 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 283 | return -ETXTBSY; |
0790b31b | 284 | |
97ac7350 AA |
285 | /* |
286 | * Revalidate the write permissions, in case security policy has | |
287 | * changed since the files were opened. | |
288 | */ | |
289 | ret = security_file_permission(file, MAY_WRITE); | |
290 | if (ret) | |
3e63cbb1 | 291 | return ret; |
97ac7350 | 292 | |
97ac7350 | 293 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 294 | return -ESPIPE; |
97ac7350 | 295 | |
9e79b132 AG |
296 | if (S_ISDIR(inode->i_mode)) |
297 | return -EISDIR; | |
298 | ||
299 | if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode)) | |
3e63cbb1 | 300 | return -ENODEV; |
97ac7350 | 301 | |
97ac7350 AA |
302 | /* Check for wrap through zero too */ |
303 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 304 | return -EFBIG; |
97ac7350 | 305 | |
2fe17c10 | 306 | if (!file->f_op->fallocate) |
3e63cbb1 | 307 | return -EOPNOTSUPP; |
97ac7350 | 308 | |
bfe219d3 | 309 | file_start_write(file); |
14da9200 | 310 | ret = file->f_op->fallocate(file, mode, offset, len); |
820c12d5 HS |
311 | |
312 | /* | |
313 | * Create inotify and fanotify events. | |
314 | * | |
315 | * To keep the logic simple always create events if fallocate succeeds. | |
316 | * This implies that events are even created if the file size remains | |
317 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
318 | */ | |
319 | if (ret == 0) | |
320 | fsnotify_modify(file); | |
321 | ||
bfe219d3 | 322 | file_end_write(file); |
14da9200 | 323 | return ret; |
3e63cbb1 | 324 | } |
72c72bdf | 325 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 326 | |
edf292c7 | 327 | int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len) |
3e63cbb1 | 328 | { |
2903ff01 | 329 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
330 | int error = -EBADF; |
331 | ||
2903ff01 | 332 | if (f.file) { |
72c72bdf | 333 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 334 | fdput(f); |
3e63cbb1 | 335 | } |
3e63cbb1 | 336 | return error; |
97ac7350 | 337 | } |
3e63cbb1 | 338 | |
edf292c7 DB |
339 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
340 | { | |
341 | return ksys_fallocate(fd, mode, offset, len); | |
342 | } | |
343 | ||
1da177e4 LT |
344 | /* |
345 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
346 | * We do this by temporarily clearing all FS-related capabilities and | |
347 | * switching the fsuid/fsgid around to the real ones. | |
348 | */ | |
94704515 | 349 | static const struct cred *access_override_creds(void) |
1da177e4 | 350 | { |
d84f4f99 DH |
351 | const struct cred *old_cred; |
352 | struct cred *override_cred; | |
1da177e4 | 353 | |
d84f4f99 DH |
354 | override_cred = prepare_creds(); |
355 | if (!override_cred) | |
94704515 | 356 | return NULL; |
1da177e4 | 357 | |
d84f4f99 DH |
358 | override_cred->fsuid = override_cred->uid; |
359 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 360 | |
086f7316 | 361 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 362 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
363 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
364 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 365 | cap_clear(override_cred->cap_effective); |
086f7316 | 366 | else |
d84f4f99 DH |
367 | override_cred->cap_effective = |
368 | override_cred->cap_permitted; | |
086f7316 | 369 | } |
1da177e4 | 370 | |
d7852fbd LT |
371 | /* |
372 | * The new set of credentials can *only* be used in | |
373 | * task-synchronous circumstances, and does not need | |
374 | * RCU freeing, unless somebody then takes a separate | |
375 | * reference to it. | |
376 | * | |
377 | * NOTE! This is _only_ true because this credential | |
378 | * is used purely for override_creds() that installs | |
379 | * it as the subjective cred. Other threads will be | |
380 | * accessing ->real_cred, not the subjective cred. | |
381 | * | |
382 | * If somebody _does_ make a copy of this (using the | |
383 | * 'get_current_cred()' function), that will clear the | |
384 | * non_rcu field, because now that other user may be | |
385 | * expecting RCU freeing. But normal thread-synchronous | |
386 | * cred accesses will keep things non-RCY. | |
387 | */ | |
388 | override_cred->non_rcu = 1; | |
389 | ||
d84f4f99 | 390 | old_cred = override_creds(override_cred); |
94704515 MS |
391 | |
392 | /* override_cred() gets its own ref */ | |
393 | put_cred(override_cred); | |
394 | ||
395 | return old_cred; | |
396 | } | |
397 | ||
eb9d7d39 | 398 | static long do_faccessat(int dfd, const char __user *filename, int mode, int flags) |
94704515 MS |
399 | { |
400 | struct path path; | |
401 | struct inode *inode; | |
402 | int res; | |
403 | unsigned int lookup_flags = LOOKUP_FOLLOW; | |
c8ffd8bc | 404 | const struct cred *old_cred = NULL; |
94704515 MS |
405 | |
406 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
407 | return -EINVAL; | |
408 | ||
c8ffd8bc MS |
409 | if (flags & ~(AT_EACCESS | AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) |
410 | return -EINVAL; | |
411 | ||
412 | if (flags & AT_SYMLINK_NOFOLLOW) | |
413 | lookup_flags &= ~LOOKUP_FOLLOW; | |
414 | if (flags & AT_EMPTY_PATH) | |
415 | lookup_flags |= LOOKUP_EMPTY; | |
416 | ||
417 | if (!(flags & AT_EACCESS)) { | |
418 | old_cred = access_override_creds(); | |
419 | if (!old_cred) | |
420 | return -ENOMEM; | |
421 | } | |
94704515 | 422 | |
87fa5595 JL |
423 | retry: |
424 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
425 | if (res) |
426 | goto out; | |
427 | ||
63afdfc7 | 428 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
429 | |
430 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
431 | /* |
432 | * MAY_EXEC on regular files is denied if the fs is mounted | |
433 | * with the "noexec" flag. | |
434 | */ | |
435 | res = -EACCES; | |
90f8572b | 436 | if (path_noexec(&path)) |
30524472 AV |
437 | goto out_path_release; |
438 | } | |
439 | ||
256984a8 | 440 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 441 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 442 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 443 | goto out_path_release; |
2f676cbc DH |
444 | /* |
445 | * This is a rare case where using __mnt_is_readonly() | |
446 | * is OK without a mnt_want/drop_write() pair. Since | |
447 | * no actual write to the fs is performed here, we do | |
448 | * not need to telegraph to that to anyone. | |
449 | * | |
450 | * By doing this, we accept that this access is | |
451 | * inherently racy and know that the fs may change | |
452 | * state before we even see this result. | |
453 | */ | |
2d8f3038 | 454 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 455 | res = -EROFS; |
1da177e4 | 456 | |
6902d925 | 457 | out_path_release: |
2d8f3038 | 458 | path_put(&path); |
87fa5595 JL |
459 | if (retry_estale(res, lookup_flags)) { |
460 | lookup_flags |= LOOKUP_REVAL; | |
461 | goto retry; | |
462 | } | |
6902d925 | 463 | out: |
c8ffd8bc MS |
464 | if (old_cred) |
465 | revert_creds(old_cred); | |
466 | ||
1da177e4 LT |
467 | return res; |
468 | } | |
469 | ||
cbfe20f5 DB |
470 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
471 | { | |
c8ffd8bc MS |
472 | return do_faccessat(dfd, filename, mode, 0); |
473 | } | |
474 | ||
475 | SYSCALL_DEFINE4(faccessat2, int, dfd, const char __user *, filename, int, mode, | |
476 | int, flags) | |
477 | { | |
478 | return do_faccessat(dfd, filename, mode, flags); | |
cbfe20f5 DB |
479 | } |
480 | ||
ca013e94 | 481 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d | 482 | { |
c8ffd8bc | 483 | return do_faccessat(AT_FDCWD, filename, mode, 0); |
5590ff0d UD |
484 | } |
485 | ||
db63f1e3 | 486 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 487 | { |
2d8f3038 | 488 | struct path path; |
1da177e4 | 489 | int error; |
0291c0a5 JL |
490 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
491 | retry: | |
492 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
493 | if (error) |
494 | goto out; | |
495 | ||
9cfcac81 | 496 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
497 | if (error) |
498 | goto dput_and_out; | |
499 | ||
2d8f3038 | 500 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
501 | |
502 | dput_and_out: | |
2d8f3038 | 503 | path_put(&path); |
0291c0a5 JL |
504 | if (retry_estale(error, lookup_flags)) { |
505 | lookup_flags |= LOOKUP_REVAL; | |
506 | goto retry; | |
507 | } | |
1da177e4 LT |
508 | out: |
509 | return error; | |
510 | } | |
511 | ||
3cdad428 | 512 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 513 | { |
2903ff01 | 514 | struct fd f = fdget_raw(fd); |
159b0956 | 515 | int error; |
1da177e4 LT |
516 | |
517 | error = -EBADF; | |
2903ff01 | 518 | if (!f.file) |
1da177e4 LT |
519 | goto out; |
520 | ||
1da177e4 | 521 | error = -ENOTDIR; |
159b0956 | 522 | if (!d_can_lookup(f.file->f_path.dentry)) |
1da177e4 LT |
523 | goto out_putf; |
524 | ||
159b0956 | 525 | error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR); |
1da177e4 | 526 | if (!error) |
2903ff01 | 527 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 528 | out_putf: |
2903ff01 | 529 | fdput(f); |
1da177e4 LT |
530 | out: |
531 | return error; | |
532 | } | |
533 | ||
4b7ca501 | 534 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 535 | { |
2d8f3038 | 536 | struct path path; |
1da177e4 | 537 | int error; |
2771261e JL |
538 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
539 | retry: | |
540 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
541 | if (error) |
542 | goto out; | |
543 | ||
9cfcac81 | 544 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
545 | if (error) |
546 | goto dput_and_out; | |
547 | ||
548 | error = -EPERM; | |
c7b96acf | 549 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 550 | goto dput_and_out; |
8b8efb44 TH |
551 | error = security_path_chroot(&path); |
552 | if (error) | |
553 | goto dput_and_out; | |
1da177e4 | 554 | |
2d8f3038 | 555 | set_fs_root(current->fs, &path); |
1da177e4 LT |
556 | error = 0; |
557 | dput_and_out: | |
2d8f3038 | 558 | path_put(&path); |
2771261e JL |
559 | if (retry_estale(error, lookup_flags)) { |
560 | lookup_flags |= LOOKUP_REVAL; | |
561 | goto retry; | |
562 | } | |
1da177e4 LT |
563 | out: |
564 | return error; | |
565 | } | |
566 | ||
1097742e | 567 | int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 568 | { |
e57712eb | 569 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 570 | struct inode *delegated_inode = NULL; |
1da177e4 | 571 | struct iattr newattrs; |
e57712eb | 572 | int error; |
1da177e4 | 573 | |
e57712eb AV |
574 | error = mnt_want_write(path->mnt); |
575 | if (error) | |
576 | return error; | |
27ac0ffe | 577 | retry_deleg: |
5955102c | 578 | inode_lock(inode); |
cdcf116d | 579 | error = security_path_chmod(path, mode); |
e57712eb | 580 | if (error) |
fe542cf5 | 581 | goto out_unlock; |
1da177e4 LT |
582 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
583 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 584 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 585 | out_unlock: |
5955102c | 586 | inode_unlock(inode); |
27ac0ffe BF |
587 | if (delegated_inode) { |
588 | error = break_deleg_wait(&delegated_inode); | |
589 | if (!error) | |
590 | goto retry_deleg; | |
591 | } | |
e57712eb AV |
592 | mnt_drop_write(path->mnt); |
593 | return error; | |
594 | } | |
595 | ||
9e96c8c0 CH |
596 | int vfs_fchmod(struct file *file, umode_t mode) |
597 | { | |
598 | audit_file(file); | |
599 | return chmod_common(&file->f_path, mode); | |
600 | } | |
601 | ||
b25ba7c3 | 602 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 603 | { |
173c8401 | 604 | struct fd f = fdget(fd); |
e57712eb AV |
605 | int err = -EBADF; |
606 | ||
173c8401 | 607 | if (f.file) { |
9e96c8c0 | 608 | err = vfs_fchmod(f.file, mode); |
173c8401 | 609 | fdput(f); |
e57712eb | 610 | } |
1da177e4 LT |
611 | return err; |
612 | } | |
613 | ||
1097742e | 614 | static int do_fchmodat(int dfd, const char __user *filename, umode_t mode) |
1da177e4 | 615 | { |
2d8f3038 | 616 | struct path path; |
1da177e4 | 617 | int error; |
14ff690c JL |
618 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
619 | retry: | |
620 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
621 | if (!error) { |
622 | error = chmod_common(&path, mode); | |
623 | path_put(&path); | |
14ff690c JL |
624 | if (retry_estale(error, lookup_flags)) { |
625 | lookup_flags |= LOOKUP_REVAL; | |
626 | goto retry; | |
627 | } | |
e57712eb | 628 | } |
1da177e4 LT |
629 | return error; |
630 | } | |
631 | ||
03450e27 DB |
632 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, |
633 | umode_t, mode) | |
634 | { | |
635 | return do_fchmodat(dfd, filename, mode); | |
636 | } | |
637 | ||
49f0a076 | 638 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d | 639 | { |
03450e27 | 640 | return do_fchmodat(AT_FDCWD, filename, mode); |
5590ff0d UD |
641 | } |
642 | ||
b873498f | 643 | int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 644 | { |
fe542cf5 | 645 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 646 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
647 | int error; |
648 | struct iattr newattrs; | |
52137abe EB |
649 | kuid_t uid; |
650 | kgid_t gid; | |
651 | ||
652 | uid = make_kuid(current_user_ns(), user); | |
653 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 654 | |
c1b8940b | 655 | retry_deleg: |
1da177e4 LT |
656 | newattrs.ia_valid = ATTR_CTIME; |
657 | if (user != (uid_t) -1) { | |
52137abe EB |
658 | if (!uid_valid(uid)) |
659 | return -EINVAL; | |
1da177e4 | 660 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 661 | newattrs.ia_uid = uid; |
1da177e4 LT |
662 | } |
663 | if (group != (gid_t) -1) { | |
52137abe EB |
664 | if (!gid_valid(gid)) |
665 | return -EINVAL; | |
1da177e4 | 666 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 667 | newattrs.ia_gid = gid; |
1da177e4 LT |
668 | } |
669 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
670 | newattrs.ia_valid |= |
671 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
5955102c | 672 | inode_lock(inode); |
d2b31ca6 | 673 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 674 | if (!error) |
27ac0ffe | 675 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
5955102c | 676 | inode_unlock(inode); |
27ac0ffe BF |
677 | if (delegated_inode) { |
678 | error = break_deleg_wait(&delegated_inode); | |
679 | if (!error) | |
680 | goto retry_deleg; | |
681 | } | |
1da177e4 LT |
682 | return error; |
683 | } | |
684 | ||
55731b3c DB |
685 | int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group, |
686 | int flag) | |
5590ff0d | 687 | { |
2d8f3038 | 688 | struct path path; |
5590ff0d | 689 | int error = -EINVAL; |
65cfc672 | 690 | int lookup_flags; |
5590ff0d | 691 | |
65cfc672 | 692 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
693 | goto out; |
694 | ||
65cfc672 AV |
695 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
696 | if (flag & AT_EMPTY_PATH) | |
697 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 698 | retry: |
65cfc672 | 699 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
700 | if (error) |
701 | goto out; | |
2d8f3038 | 702 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
703 | if (error) |
704 | goto out_release; | |
fe542cf5 | 705 | error = chown_common(&path, user, group); |
2d8f3038 | 706 | mnt_drop_write(path.mnt); |
2af482a7 | 707 | out_release: |
2d8f3038 | 708 | path_put(&path); |
99a5df37 JL |
709 | if (retry_estale(error, lookup_flags)) { |
710 | lookup_flags |= LOOKUP_REVAL; | |
711 | goto retry; | |
712 | } | |
5590ff0d UD |
713 | out: |
714 | return error; | |
715 | } | |
716 | ||
55731b3c DB |
717 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
718 | gid_t, group, int, flag) | |
719 | { | |
720 | return do_fchownat(dfd, filename, user, group, flag); | |
721 | } | |
722 | ||
55e4def0 | 723 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 724 | { |
55731b3c | 725 | return do_fchownat(AT_FDCWD, filename, user, group, 0); |
55e4def0 | 726 | } |
1da177e4 | 727 | |
55e4def0 DH |
728 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
729 | { | |
55731b3c DB |
730 | return do_fchownat(AT_FDCWD, filename, user, group, |
731 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
732 | } |
733 | ||
c04011fe CH |
734 | int vfs_fchown(struct file *file, uid_t user, gid_t group) |
735 | { | |
736 | int error; | |
737 | ||
738 | error = mnt_want_write_file(file); | |
739 | if (error) | |
740 | return error; | |
741 | audit_file(file); | |
742 | error = chown_common(&file->f_path, user, group); | |
743 | mnt_drop_write_file(file); | |
744 | return error; | |
745 | } | |
746 | ||
55731b3c | 747 | int ksys_fchown(unsigned int fd, uid_t user, gid_t group) |
1da177e4 | 748 | { |
2903ff01 | 749 | struct fd f = fdget(fd); |
1da177e4 LT |
750 | int error = -EBADF; |
751 | ||
c04011fe CH |
752 | if (f.file) { |
753 | error = vfs_fchown(f.file, user, group); | |
754 | fdput(f); | |
755 | } | |
1da177e4 LT |
756 | return error; |
757 | } | |
758 | ||
55731b3c DB |
759 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
760 | { | |
761 | return ksys_fchown(fd, user, group); | |
762 | } | |
763 | ||
02e5180d | 764 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 765 | struct inode *inode, |
ae2bb293 | 766 | int (*open)(struct inode *, struct file *)) |
1da177e4 | 767 | { |
1abf0c71 | 768 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
769 | int error; |
770 | ||
b5bcdda3 | 771 | path_get(&f->f_path); |
4bacc9c9 | 772 | f->f_inode = inode; |
1da177e4 | 773 | f->f_mapping = inode->i_mapping; |
5660e13d | 774 | f->f_wb_err = filemap_sample_wb_err(f->f_mapping); |
735e4ae5 | 775 | f->f_sb_err = file_sample_sb_err(f); |
5660e13d | 776 | |
3f4d5a00 | 777 | if (unlikely(f->f_flags & O_PATH)) { |
f5d11409 | 778 | f->f_mode = FMODE_PATH | FMODE_OPENED; |
1abf0c71 | 779 | f->f_op = &empty_fops; |
af04fadc | 780 | return 0; |
1abf0c71 AV |
781 | } |
782 | ||
dd20908a | 783 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 784 | error = get_write_access(inode); |
3f4d5a00 | 785 | if (unlikely(error)) |
1da177e4 | 786 | goto cleanup_file; |
0ccb2863 | 787 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 788 | if (unlikely(error)) { |
0ccb2863 AV |
789 | put_write_access(inode); |
790 | goto cleanup_file; | |
791 | } | |
83f936c7 | 792 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
793 | } |
794 | ||
2be7d348 LT |
795 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
796 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) | |
797 | f->f_mode |= FMODE_ATOMIC_POS; | |
798 | ||
1abf0c71 | 799 | f->f_op = fops_get(inode->i_fop); |
7159d544 | 800 | if (WARN_ON(!f->f_op)) { |
72c2d531 AV |
801 | error = -ENODEV; |
802 | goto cleanup_all; | |
803 | } | |
1abf0c71 | 804 | |
e3f20ae2 | 805 | error = security_file_open(f); |
788e7dd4 YN |
806 | if (error) |
807 | goto cleanup_all; | |
808 | ||
c568d683 | 809 | error = break_lease(locks_inode(f), f->f_flags); |
f3c7691e BF |
810 | if (error) |
811 | goto cleanup_all; | |
812 | ||
ea73ea72 AV |
813 | /* normally all 3 are set; ->open() can clear them if needed */ |
814 | f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE; | |
72c2d531 | 815 | if (!open) |
834f2a4a TM |
816 | open = f->f_op->open; |
817 | if (open) { | |
818 | error = open(inode, f); | |
1da177e4 LT |
819 | if (error) |
820 | goto cleanup_all; | |
821 | } | |
f5d11409 | 822 | f->f_mode |= FMODE_OPENED; |
890275b5 MZ |
823 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
824 | i_readcount_inc(inode); | |
293bc982 | 825 | if ((f->f_mode & FMODE_READ) && |
84363182 | 826 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 827 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 828 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 829 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 830 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 831 | |
c75b1d94 | 832 | f->f_write_hint = WRITE_LIFE_NOT_SET; |
1da177e4 LT |
833 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
834 | ||
835 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
af04fadc | 836 | |
69527c55 AV |
837 | /* NB: we're sure to have correct a_ops only after f_op->open */ |
838 | if (f->f_flags & O_DIRECT) { | |
839 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) | |
840 | return -EINVAL; | |
841 | } | |
09d91cda SL |
842 | |
843 | /* | |
844 | * XXX: Huge page cache doesn't support writing yet. Drop all page | |
845 | * cache for this file before processing writes. | |
846 | */ | |
847 | if ((f->f_mode & FMODE_WRITE) && filemap_nr_thps(inode->i_mapping)) | |
848 | truncate_pagecache(inode, 0); | |
849 | ||
96b7e579 | 850 | return 0; |
1da177e4 LT |
851 | |
852 | cleanup_all: | |
6b4e8085 AV |
853 | if (WARN_ON_ONCE(error > 0)) |
854 | error = -EINVAL; | |
1da177e4 | 855 | fops_put(f->f_op); |
83f936c7 | 856 | if (f->f_mode & FMODE_WRITER) { |
1da177e4 | 857 | put_write_access(inode); |
83f936c7 | 858 | __mnt_drop_write(f->f_path.mnt); |
4a3fd211 | 859 | } |
1da177e4 | 860 | cleanup_file: |
02e5180d AV |
861 | path_put(&f->f_path); |
862 | f->f_path.mnt = NULL; | |
863 | f->f_path.dentry = NULL; | |
dd37978c | 864 | f->f_inode = NULL; |
96b7e579 | 865 | return error; |
1da177e4 LT |
866 | } |
867 | ||
d18e9008 MS |
868 | /** |
869 | * finish_open - finish opening a file | |
0854d450 | 870 | * @file: file pointer |
d18e9008 MS |
871 | * @dentry: pointer to dentry |
872 | * @open: open callback | |
0854d450 | 873 | * @opened: state of open |
d18e9008 MS |
874 | * |
875 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
876 | * | |
877 | * If the open callback is set to NULL, then the standard f_op->open() | |
878 | * filesystem callback is substituted. | |
0854d450 MS |
879 | * |
880 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
881 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
882 | * on it after finish_open(). | |
883 | * | |
0854d450 | 884 | * Returns zero on success or -errno if the open failed. |
d18e9008 | 885 | */ |
30d90494 | 886 | int finish_open(struct file *file, struct dentry *dentry, |
be12af3e | 887 | int (*open)(struct inode *, struct file *)) |
d18e9008 | 888 | { |
aad888f8 | 889 | BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 890 | |
b5bcdda3 | 891 | file->f_path.dentry = dentry; |
aad888f8 | 892 | return do_dentry_open(file, d_backing_inode(dentry), open); |
d18e9008 MS |
893 | } |
894 | EXPORT_SYMBOL(finish_open); | |
895 | ||
896 | /** | |
897 | * finish_no_open - finish ->atomic_open() without opening the file | |
898 | * | |
0854d450 | 899 | * @file: file pointer |
d18e9008 MS |
900 | * @dentry: dentry or NULL (as returned from ->lookup()) |
901 | * | |
902 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
903 | * |
904 | * NB: unlike finish_open() this function does consume the dentry reference and | |
905 | * the caller need not dput() it. | |
906 | * | |
64e1ac4d | 907 | * Returns "0" which must be the return value of ->atomic_open() after having |
0854d450 | 908 | * called this function. |
d18e9008 | 909 | */ |
e45198a6 | 910 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 911 | { |
30d90494 | 912 | file->f_path.dentry = dentry; |
64e1ac4d | 913 | return 0; |
d18e9008 MS |
914 | } |
915 | EXPORT_SYMBOL(finish_no_open); | |
916 | ||
9bf39ab2 MS |
917 | char *file_path(struct file *filp, char *buf, int buflen) |
918 | { | |
919 | return d_path(&filp->f_path, buf, buflen); | |
920 | } | |
921 | EXPORT_SYMBOL(file_path); | |
922 | ||
4bacc9c9 DH |
923 | /** |
924 | * vfs_open - open the file at the given path | |
925 | * @path: path to open | |
926 | * @file: newly allocated file with f_flag initialized | |
927 | * @cred: credentials to use | |
928 | */ | |
ae2bb293 | 929 | int vfs_open(const struct path *path, struct file *file) |
4bacc9c9 | 930 | { |
54d5ca87 | 931 | file->f_path = *path; |
a6518f73 | 932 | return do_dentry_open(file, d_backing_inode(path->dentry), NULL); |
4bacc9c9 DH |
933 | } |
934 | ||
765927b2 | 935 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 936 | const struct cred *cred) |
a1a5b3d9 PS |
937 | { |
938 | int error; | |
939 | struct file *f; | |
940 | ||
e0e81739 DH |
941 | validate_creds(cred); |
942 | ||
c212f9aa | 943 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 944 | BUG_ON(!path->mnt); |
322ee5b3 | 945 | |
ea73ea72 | 946 | f = alloc_empty_file(flags, cred); |
af04fadc | 947 | if (!IS_ERR(f)) { |
ae2bb293 | 948 | error = vfs_open(path, f); |
4d27f326 AV |
949 | if (error) { |
950 | fput(f); | |
af04fadc AV |
951 | f = ERR_PTR(error); |
952 | } | |
2a027e7a AV |
953 | } |
954 | return f; | |
a1a5b3d9 | 955 | } |
1da177e4 LT |
956 | EXPORT_SYMBOL(dentry_open); |
957 | ||
2abc77af AV |
958 | struct file *open_with_fake_path(const struct path *path, int flags, |
959 | struct inode *inode, const struct cred *cred) | |
960 | { | |
d3b1084d | 961 | struct file *f = alloc_empty_file_noaccount(flags, cred); |
2abc77af AV |
962 | if (!IS_ERR(f)) { |
963 | int error; | |
964 | ||
965 | f->f_path = *path; | |
966 | error = do_dentry_open(f, inode, NULL); | |
967 | if (error) { | |
968 | fput(f); | |
969 | f = ERR_PTR(error); | |
970 | } | |
971 | } | |
972 | return f; | |
973 | } | |
974 | EXPORT_SYMBOL(open_with_fake_path); | |
975 | ||
fddb5d43 AS |
976 | #define WILL_CREATE(flags) (flags & (O_CREAT | __O_TMPFILE)) |
977 | #define O_PATH_FLAGS (O_DIRECTORY | O_NOFOLLOW | O_PATH | O_CLOEXEC) | |
978 | ||
35cb6d54 | 979 | inline struct open_how build_open_how(int flags, umode_t mode) |
fddb5d43 AS |
980 | { |
981 | struct open_how how = { | |
982 | .flags = flags & VALID_OPEN_FLAGS, | |
983 | .mode = mode & S_IALLUGO, | |
984 | }; | |
985 | ||
986 | /* O_PATH beats everything else. */ | |
987 | if (how.flags & O_PATH) | |
988 | how.flags &= O_PATH_FLAGS; | |
989 | /* Modes should only be set for create-like flags. */ | |
990 | if (!WILL_CREATE(how.flags)) | |
991 | how.mode = 0; | |
992 | return how; | |
993 | } | |
994 | ||
35cb6d54 | 995 | inline int build_open_flags(const struct open_how *how, struct open_flags *op) |
47c805dc | 996 | { |
e7a2cf4a CB |
997 | u64 flags = how->flags; |
998 | u64 strip = FMODE_NONOTIFY | O_CLOEXEC; | |
47c805dc | 999 | int lookup_flags = 0; |
62fb4a15 | 1000 | int acc_mode = ACC_MODE(flags); |
47c805dc | 1001 | |
e7a2cf4a CB |
1002 | BUILD_BUG_ON_MSG(upper_32_bits(VALID_OPEN_FLAGS), |
1003 | "struct open_flags doesn't yet handle flags > 32 bits"); | |
1004 | ||
1005 | /* | |
1006 | * Strip flags that either shouldn't be set by userspace like | |
1007 | * FMODE_NONOTIFY or that aren't relevant in determining struct | |
1008 | * open_flags like O_CLOEXEC. | |
1009 | */ | |
1010 | flags &= ~strip; | |
fddb5d43 | 1011 | |
629e014b | 1012 | /* |
fddb5d43 AS |
1013 | * Older syscalls implicitly clear all of the invalid flags or argument |
1014 | * values before calling build_open_flags(), but openat2(2) checks all | |
1015 | * of its arguments. | |
629e014b | 1016 | */ |
fddb5d43 AS |
1017 | if (flags & ~VALID_OPEN_FLAGS) |
1018 | return -EINVAL; | |
1019 | if (how->resolve & ~VALID_RESOLVE_FLAGS) | |
1020 | return -EINVAL; | |
629e014b | 1021 | |
398840f8 AS |
1022 | /* Scoping flags are mutually exclusive. */ |
1023 | if ((how->resolve & RESOLVE_BENEATH) && (how->resolve & RESOLVE_IN_ROOT)) | |
1024 | return -EINVAL; | |
1025 | ||
fddb5d43 AS |
1026 | /* Deal with the mode. */ |
1027 | if (WILL_CREATE(flags)) { | |
1028 | if (how->mode & ~S_IALLUGO) | |
1029 | return -EINVAL; | |
1030 | op->mode = how->mode | S_IFREG; | |
1031 | } else { | |
1032 | if (how->mode != 0) | |
1033 | return -EINVAL; | |
e68726ff | 1034 | op->mode = 0; |
fddb5d43 | 1035 | } |
47c805dc AV |
1036 | |
1037 | /* | |
fddb5d43 AS |
1038 | * In order to ensure programs get explicit errors when trying to use |
1039 | * O_TMPFILE on old kernels, O_TMPFILE is implemented such that it | |
1040 | * looks like (O_DIRECTORY|O_RDWR & ~O_CREAT) to old kernels. But we | |
1041 | * have to require userspace to explicitly set it. | |
47c805dc | 1042 | */ |
bb458c64 AV |
1043 | if (flags & __O_TMPFILE) { |
1044 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d | 1045 | return -EINVAL; |
ba57ea64 AV |
1046 | if (!(acc_mode & MAY_WRITE)) |
1047 | return -EINVAL; | |
fddb5d43 AS |
1048 | } |
1049 | if (flags & O_PATH) { | |
1050 | /* O_PATH only permits certain other flags to be set. */ | |
1051 | if (flags & ~O_PATH_FLAGS) | |
1052 | return -EINVAL; | |
1abf0c71 | 1053 | acc_mode = 0; |
1abf0c71 | 1054 | } |
47c805dc | 1055 | |
fddb5d43 AS |
1056 | /* |
1057 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1058 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1059 | * always set instead of having to deal with possibly weird behaviour | |
1060 | * for malicious applications setting only __O_SYNC. | |
1061 | */ | |
1062 | if (flags & __O_SYNC) | |
1063 | flags |= O_DSYNC; | |
1064 | ||
1abf0c71 | 1065 | op->open_flag = flags; |
47c805dc AV |
1066 | |
1067 | /* O_TRUNC implies we need access checks for write permissions */ | |
1068 | if (flags & O_TRUNC) | |
1069 | acc_mode |= MAY_WRITE; | |
1070 | ||
1071 | /* Allow the LSM permission hook to distinguish append | |
1072 | access from general write access. */ | |
1073 | if (flags & O_APPEND) | |
1074 | acc_mode |= MAY_APPEND; | |
1075 | ||
1076 | op->acc_mode = acc_mode; | |
1077 | ||
1abf0c71 AV |
1078 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
1079 | ||
47c805dc AV |
1080 | if (flags & O_CREAT) { |
1081 | op->intent |= LOOKUP_CREATE; | |
31d1726d | 1082 | if (flags & O_EXCL) { |
47c805dc | 1083 | op->intent |= LOOKUP_EXCL; |
31d1726d AV |
1084 | flags |= O_NOFOLLOW; |
1085 | } | |
47c805dc AV |
1086 | } |
1087 | ||
1088 | if (flags & O_DIRECTORY) | |
1089 | lookup_flags |= LOOKUP_DIRECTORY; | |
1090 | if (!(flags & O_NOFOLLOW)) | |
1091 | lookup_flags |= LOOKUP_FOLLOW; | |
fddb5d43 AS |
1092 | |
1093 | if (how->resolve & RESOLVE_NO_XDEV) | |
1094 | lookup_flags |= LOOKUP_NO_XDEV; | |
1095 | if (how->resolve & RESOLVE_NO_MAGICLINKS) | |
1096 | lookup_flags |= LOOKUP_NO_MAGICLINKS; | |
1097 | if (how->resolve & RESOLVE_NO_SYMLINKS) | |
1098 | lookup_flags |= LOOKUP_NO_SYMLINKS; | |
1099 | if (how->resolve & RESOLVE_BENEATH) | |
1100 | lookup_flags |= LOOKUP_BENEATH; | |
1101 | if (how->resolve & RESOLVE_IN_ROOT) | |
1102 | lookup_flags |= LOOKUP_IN_ROOT; | |
1103 | ||
f9652e10 AV |
1104 | op->lookup_flags = lookup_flags; |
1105 | return 0; | |
47c805dc AV |
1106 | } |
1107 | ||
669abf4e JL |
1108 | /** |
1109 | * file_open_name - open file and return file pointer | |
1110 | * | |
1111 | * @name: struct filename containing path to open | |
1112 | * @flags: open flags as per the open(2) second argument | |
1113 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1114 | * | |
1115 | * This is the helper to open a file from kernelspace if you really | |
1116 | * have to. But in generally you should not do this, so please move | |
1117 | * along, nothing to see here.. | |
1118 | */ | |
1119 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
1120 | { | |
1121 | struct open_flags op; | |
fddb5d43 AS |
1122 | struct open_how how = build_open_how(flags, mode); |
1123 | int err = build_open_flags(&how, &op); | |
1124 | if (err) | |
1125 | return ERR_PTR(err); | |
1126 | return do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
1127 | } |
1128 | ||
47c805dc AV |
1129 | /** |
1130 | * filp_open - open file and return file pointer | |
1131 | * | |
1132 | * @filename: path to open | |
1133 | * @flags: open flags as per the open(2) second argument | |
1134 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1135 | * | |
1136 | * This is the helper to open a file from kernelspace if you really | |
1137 | * have to. But in generally you should not do this, so please move | |
1138 | * along, nothing to see here.. | |
1139 | */ | |
a218d0fd | 1140 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 1141 | { |
51689104 PM |
1142 | struct filename *name = getname_kernel(filename); |
1143 | struct file *file = ERR_CAST(name); | |
1144 | ||
1145 | if (!IS_ERR(name)) { | |
1146 | file = file_open_name(name, flags, mode); | |
1147 | putname(name); | |
1148 | } | |
1149 | return file; | |
47c805dc AV |
1150 | } |
1151 | EXPORT_SYMBOL(filp_open); | |
1152 | ||
73d049a4 | 1153 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
378c6520 | 1154 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1155 | { |
1156 | struct open_flags op; | |
fddb5d43 AS |
1157 | struct open_how how = build_open_how(flags, mode); |
1158 | int err = build_open_flags(&how, &op); | |
f9652e10 AV |
1159 | if (err) |
1160 | return ERR_PTR(err); | |
f9652e10 | 1161 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
1162 | } |
1163 | EXPORT_SYMBOL(file_open_root); | |
1164 | ||
fddb5d43 AS |
1165 | static long do_sys_openat2(int dfd, const char __user *filename, |
1166 | struct open_how *how) | |
1da177e4 | 1167 | { |
47c805dc | 1168 | struct open_flags op; |
fddb5d43 | 1169 | int fd = build_open_flags(how, &op); |
f9652e10 AV |
1170 | struct filename *tmp; |
1171 | ||
1172 | if (fd) | |
1173 | return fd; | |
1174 | ||
1175 | tmp = getname(filename); | |
1176 | if (IS_ERR(tmp)) | |
1177 | return PTR_ERR(tmp); | |
1178 | ||
fddb5d43 | 1179 | fd = get_unused_fd_flags(how->flags); |
f9652e10 AV |
1180 | if (fd >= 0) { |
1181 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1182 | if (IS_ERR(f)) { | |
1183 | put_unused_fd(fd); | |
1184 | fd = PTR_ERR(f); | |
1185 | } else { | |
1186 | fsnotify_open(f); | |
1187 | fd_install(fd, f); | |
1da177e4 | 1188 | } |
1da177e4 | 1189 | } |
f9652e10 | 1190 | putname(tmp); |
1da177e4 | 1191 | return fd; |
1da177e4 | 1192 | } |
e922efc3 | 1193 | |
fddb5d43 | 1194 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
e922efc3 | 1195 | { |
fddb5d43 AS |
1196 | struct open_how how = build_open_how(flags, mode); |
1197 | return do_sys_openat2(dfd, filename, &how); | |
1198 | } | |
e922efc3 | 1199 | |
fddb5d43 AS |
1200 | |
1201 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1202 | { | |
166e07c3 CH |
1203 | if (force_o_largefile()) |
1204 | flags |= O_LARGEFILE; | |
1205 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
e922efc3 | 1206 | } |
1da177e4 | 1207 | |
6559eed8 | 1208 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1209 | umode_t, mode) |
5590ff0d UD |
1210 | { |
1211 | if (force_o_largefile()) | |
1212 | flags |= O_LARGEFILE; | |
2cf09666 | 1213 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1214 | } |
5590ff0d | 1215 | |
fddb5d43 AS |
1216 | SYSCALL_DEFINE4(openat2, int, dfd, const char __user *, filename, |
1217 | struct open_how __user *, how, size_t, usize) | |
1218 | { | |
1219 | int err; | |
1220 | struct open_how tmp; | |
1221 | ||
1222 | BUILD_BUG_ON(sizeof(struct open_how) < OPEN_HOW_SIZE_VER0); | |
1223 | BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_LATEST); | |
1224 | ||
1225 | if (unlikely(usize < OPEN_HOW_SIZE_VER0)) | |
1226 | return -EINVAL; | |
1227 | ||
1228 | err = copy_struct_from_user(&tmp, sizeof(tmp), how, usize); | |
1229 | if (err) | |
1230 | return err; | |
1231 | ||
1232 | /* O_LARGEFILE is only allowed for non-O_PATH. */ | |
1233 | if (!(tmp.flags & O_PATH) && force_o_largefile()) | |
1234 | tmp.flags |= O_LARGEFILE; | |
1235 | ||
1236 | return do_sys_openat2(dfd, filename, &tmp); | |
1237 | } | |
1238 | ||
e35d49f6 AV |
1239 | #ifdef CONFIG_COMPAT |
1240 | /* | |
1241 | * Exactly like sys_open(), except that it doesn't set the | |
1242 | * O_LARGEFILE flag. | |
1243 | */ | |
1244 | COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1245 | { | |
1246 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
1247 | } | |
1248 | ||
1249 | /* | |
1250 | * Exactly like sys_openat(), except that it doesn't set the | |
1251 | * O_LARGEFILE flag. | |
1252 | */ | |
1253 | COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode) | |
1254 | { | |
1255 | return do_sys_open(dfd, filename, flags, mode); | |
1256 | } | |
1257 | #endif | |
1258 | ||
1da177e4 LT |
1259 | #ifndef __alpha__ |
1260 | ||
1261 | /* | |
1262 | * For backward compatibility? Maybe this should be moved | |
1263 | * into arch/i386 instead? | |
1264 | */ | |
a218d0fd | 1265 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 | 1266 | { |
166e07c3 | 1267 | int flags = O_CREAT | O_WRONLY | O_TRUNC; |
1da177e4 | 1268 | |
166e07c3 CH |
1269 | if (force_o_largefile()) |
1270 | flags |= O_LARGEFILE; | |
1271 | return do_sys_open(AT_FDCWD, pathname, flags, mode); | |
1272 | } | |
1da177e4 LT |
1273 | #endif |
1274 | ||
1275 | /* | |
1276 | * "id" is the POSIX thread ID. We use the | |
1277 | * files pointer for this.. | |
1278 | */ | |
1279 | int filp_close(struct file *filp, fl_owner_t id) | |
1280 | { | |
45778ca8 | 1281 | int retval = 0; |
1da177e4 LT |
1282 | |
1283 | if (!file_count(filp)) { | |
1284 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1285 | return 0; |
1da177e4 LT |
1286 | } |
1287 | ||
72c2d531 | 1288 | if (filp->f_op->flush) |
75e1fcc0 | 1289 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1290 | |
1abf0c71 AV |
1291 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1292 | dnotify_flush(filp, id); | |
1293 | locks_remove_posix(filp, id); | |
1294 | } | |
1da177e4 LT |
1295 | fput(filp); |
1296 | return retval; | |
1297 | } | |
1298 | ||
1299 | EXPORT_SYMBOL(filp_close); | |
1300 | ||
1301 | /* | |
1302 | * Careful here! We test whether the file pointer is NULL before | |
1303 | * releasing the fd. This ensures that one clone task can't release | |
1304 | * an fd while another clone is opening it. | |
1305 | */ | |
ca013e94 | 1306 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1307 | { |
8760c909 | 1308 | int retval = close_fd(fd); |
ee731f4f EP |
1309 | |
1310 | /* can't restart close syscall because file table entry was cleared */ | |
1311 | if (unlikely(retval == -ERESTARTSYS || | |
1312 | retval == -ERESTARTNOINTR || | |
1313 | retval == -ERESTARTNOHAND || | |
1314 | retval == -ERESTART_RESTARTBLOCK)) | |
1315 | retval = -EINTR; | |
1316 | ||
1317 | return retval; | |
1da177e4 | 1318 | } |
1da177e4 | 1319 | |
278a5fba CB |
1320 | /** |
1321 | * close_range() - Close all file descriptors in a given range. | |
1322 | * | |
1323 | * @fd: starting file descriptor to close | |
1324 | * @max_fd: last file descriptor to close | |
1325 | * @flags: reserved for future extensions | |
1326 | * | |
1327 | * This closes a range of file descriptors. All file descriptors | |
1328 | * from @fd up to and including @max_fd are closed. | |
1329 | * Currently, errors to close a given file descriptor are ignored. | |
1330 | */ | |
1331 | SYSCALL_DEFINE3(close_range, unsigned int, fd, unsigned int, max_fd, | |
1332 | unsigned int, flags) | |
1333 | { | |
60997c3d | 1334 | return __close_range(fd, max_fd, flags); |
278a5fba CB |
1335 | } |
1336 | ||
1da177e4 LT |
1337 | /* |
1338 | * This routine simulates a hangup on the tty, to arrange that users | |
1339 | * are given clean terminals at login time. | |
1340 | */ | |
ca013e94 | 1341 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1342 | { |
1343 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1344 | tty_vhangup_self(); |
1da177e4 LT |
1345 | return 0; |
1346 | } | |
1347 | return -EPERM; | |
1348 | } | |
1349 | ||
1350 | /* | |
1351 | * Called when an inode is about to be open. | |
1352 | * We use this to disallow opening large files on 32bit systems if | |
1353 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1354 | * on this flag in sys_open. | |
1355 | */ | |
1356 | int generic_file_open(struct inode * inode, struct file * filp) | |
1357 | { | |
1358 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1359 | return -EOVERFLOW; |
1da177e4 LT |
1360 | return 0; |
1361 | } | |
1362 | ||
1363 | EXPORT_SYMBOL(generic_file_open); | |
1364 | ||
1365 | /* | |
1366 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1367 | * file descriptors. The function is not supposed to ever fail, the only |
1368 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1369 | * directly into file_operations structure. | |
1da177e4 LT |
1370 | */ |
1371 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1372 | { | |
1373 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1374 | return 0; | |
1375 | } | |
1376 | ||
1377 | EXPORT_SYMBOL(nonseekable_open); | |
10dce8af KS |
1378 | |
1379 | /* | |
1380 | * stream_open is used by subsystems that want stream-like file descriptors. | |
1381 | * Such file descriptors are not seekable and don't have notion of position | |
438ab720 KS |
1382 | * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL). |
1383 | * Contrary to file descriptors of other regular files, .read() and .write() | |
1384 | * can run simultaneously. | |
10dce8af KS |
1385 | * |
1386 | * stream_open never fails and is marked to return int so that it could be | |
1387 | * directly used as file_operations.open . | |
1388 | */ | |
1389 | int stream_open(struct inode *inode, struct file *filp) | |
1390 | { | |
2be7d348 | 1391 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS); |
10dce8af KS |
1392 | filp->f_mode |= FMODE_STREAM; |
1393 | return 0; | |
1394 | } | |
1395 | ||
1396 | EXPORT_SYMBOL(stream_open); |