]>
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 LT |
67 | } |
68 | ||
7df818b2 | 69 | long vfs_truncate(const struct path *path, loff_t length) |
1da177e4 | 70 | { |
2d8f3038 | 71 | struct inode *inode; |
a02de960 | 72 | long error; |
1da177e4 | 73 | |
a02de960 | 74 | inode = path->dentry->d_inode; |
1da177e4 LT |
75 | |
76 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
1da177e4 | 77 | if (S_ISDIR(inode->i_mode)) |
a02de960 | 78 | return -EISDIR; |
1da177e4 | 79 | if (!S_ISREG(inode->i_mode)) |
a02de960 | 80 | return -EINVAL; |
1da177e4 | 81 | |
a02de960 | 82 | error = mnt_want_write(path->mnt); |
1da177e4 | 83 | if (error) |
a02de960 | 84 | goto out; |
1da177e4 | 85 | |
256984a8 | 86 | error = inode_permission(inode, MAY_WRITE); |
9ac9b847 DH |
87 | if (error) |
88 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
89 | |
90 | error = -EPERM; | |
c82e42da | 91 | if (IS_APPEND(inode)) |
9ac9b847 | 92 | goto mnt_drop_write_and_out; |
1da177e4 | 93 | |
8cf9ee50 | 94 | error = get_write_access(inode); |
1da177e4 | 95 | if (error) |
9ac9b847 | 96 | goto mnt_drop_write_and_out; |
1da177e4 | 97 | |
9700382c | 98 | /* |
99 | * Make sure that there are no leases. get_write_access() protects | |
100 | * against the truncate racing with a lease-granting setlease(). | |
101 | */ | |
8737c930 | 102 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 103 | if (error) |
9700382c | 104 | goto put_write_and_out; |
1da177e4 LT |
105 | |
106 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 | 107 | if (!error) |
a02de960 | 108 | error = security_path_truncate(path); |
907f4554 | 109 | if (!error) |
a02de960 | 110 | error = do_truncate(path->dentry, length, 0, NULL); |
1da177e4 | 111 | |
9700382c | 112 | put_write_and_out: |
8cf9ee50 | 113 | put_write_access(inode); |
9ac9b847 | 114 | mnt_drop_write_and_out: |
a02de960 | 115 | mnt_drop_write(path->mnt); |
1da177e4 LT |
116 | out: |
117 | return error; | |
118 | } | |
a02de960 DH |
119 | EXPORT_SYMBOL_GPL(vfs_truncate); |
120 | ||
df260e21 | 121 | long do_sys_truncate(const char __user *pathname, loff_t length) |
a02de960 | 122 | { |
48f7530d | 123 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
124 | struct path path; |
125 | int error; | |
126 | ||
127 | if (length < 0) /* sorry, but loff_t says... */ | |
128 | return -EINVAL; | |
129 | ||
48f7530d JL |
130 | retry: |
131 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
132 | if (!error) { |
133 | error = vfs_truncate(&path, length); | |
134 | path_put(&path); | |
135 | } | |
48f7530d JL |
136 | if (retry_estale(error, lookup_flags)) { |
137 | lookup_flags |= LOOKUP_REVAL; | |
138 | goto retry; | |
139 | } | |
a02de960 DH |
140 | return error; |
141 | } | |
1da177e4 | 142 | |
4fd8da8d | 143 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 144 | { |
4fd8da8d | 145 | return do_sys_truncate(path, length); |
1da177e4 LT |
146 | } |
147 | ||
3f6d078d AV |
148 | #ifdef CONFIG_COMPAT |
149 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
150 | { | |
151 | return do_sys_truncate(path, length); | |
152 | } | |
153 | #endif | |
154 | ||
411d9475 | 155 | long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 | 156 | { |
bf2965d5 | 157 | struct inode *inode; |
1da177e4 | 158 | struct dentry *dentry; |
2903ff01 | 159 | struct fd f; |
1da177e4 LT |
160 | int error; |
161 | ||
162 | error = -EINVAL; | |
163 | if (length < 0) | |
164 | goto out; | |
165 | error = -EBADF; | |
2903ff01 AV |
166 | f = fdget(fd); |
167 | if (!f.file) | |
1da177e4 LT |
168 | goto out; |
169 | ||
170 | /* explicitly opened as large or we are on 64-bit box */ | |
2903ff01 | 171 | if (f.file->f_flags & O_LARGEFILE) |
1da177e4 LT |
172 | small = 0; |
173 | ||
2903ff01 | 174 | dentry = f.file->f_path.dentry; |
1da177e4 LT |
175 | inode = dentry->d_inode; |
176 | error = -EINVAL; | |
2903ff01 | 177 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
178 | goto out_putf; |
179 | ||
180 | error = -EINVAL; | |
181 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
182 | if (small && length > MAX_NON_LFS) | |
183 | goto out_putf; | |
184 | ||
185 | error = -EPERM; | |
78757af6 AG |
186 | /* Check IS_APPEND on real upper inode */ |
187 | if (IS_APPEND(file_inode(f.file))) | |
1da177e4 LT |
188 | goto out_putf; |
189 | ||
14da9200 | 190 | sb_start_write(inode->i_sb); |
2903ff01 | 191 | error = locks_verify_truncate(inode, f.file, length); |
be6d3e56 | 192 | if (!error) |
2903ff01 | 193 | error = security_path_truncate(&f.file->f_path); |
1da177e4 | 194 | if (!error) |
2903ff01 | 195 | error = do_truncate(dentry, length, 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 | |
3e63cbb1 | 228 | |
72c72bdf | 229 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 230 | { |
496ad9aa | 231 | struct inode *inode = file_inode(file); |
3e63cbb1 | 232 | long ret; |
97ac7350 AA |
233 | |
234 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 235 | return -EINVAL; |
97ac7350 AA |
236 | |
237 | /* Return error if mode is not supported */ | |
dd46c787 | 238 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) |
409332b6 LC |
239 | return -EOPNOTSUPP; |
240 | ||
241 | /* Punch hole and zero range are mutually exclusive */ | |
242 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
243 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
244 | return -EOPNOTSUPP; |
245 | ||
246 | /* Punch hole must have keep size set */ | |
247 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
248 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 249 | return -EOPNOTSUPP; |
97ac7350 | 250 | |
00f5e619 NJ |
251 | /* Collapse range should only be used exclusively. */ |
252 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
253 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
254 | return -EINVAL; | |
255 | ||
dd46c787 NJ |
256 | /* Insert range should only be used exclusively. */ |
257 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
258 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
259 | return -EINVAL; | |
260 | ||
71be6b49 DW |
261 | /* Unshare range should only be used with allocate mode. */ |
262 | if ((mode & FALLOC_FL_UNSHARE_RANGE) && | |
263 | (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE))) | |
264 | return -EINVAL; | |
265 | ||
97ac7350 | 266 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 267 | return -EBADF; |
1ca551c6 | 268 | |
00f5e619 | 269 | /* |
8fc61d92 | 270 | * We can only allow pure fallocate on append only files |
00f5e619 | 271 | */ |
8fc61d92 | 272 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
273 | return -EPERM; |
274 | ||
275 | if (IS_IMMUTABLE(inode)) | |
276 | return -EPERM; | |
277 | ||
0790b31b | 278 | /* |
6d2b6170 | 279 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
280 | */ |
281 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 282 | return -ETXTBSY; |
0790b31b | 283 | |
97ac7350 AA |
284 | /* |
285 | * Revalidate the write permissions, in case security policy has | |
286 | * changed since the files were opened. | |
287 | */ | |
288 | ret = security_file_permission(file, MAY_WRITE); | |
289 | if (ret) | |
3e63cbb1 | 290 | return ret; |
97ac7350 | 291 | |
97ac7350 | 292 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 293 | return -ESPIPE; |
97ac7350 | 294 | |
9e79b132 AG |
295 | if (S_ISDIR(inode->i_mode)) |
296 | return -EISDIR; | |
297 | ||
298 | if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode)) | |
3e63cbb1 | 299 | return -ENODEV; |
97ac7350 | 300 | |
97ac7350 AA |
301 | /* Check for wrap through zero too */ |
302 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 303 | return -EFBIG; |
97ac7350 | 304 | |
2fe17c10 | 305 | if (!file->f_op->fallocate) |
3e63cbb1 | 306 | return -EOPNOTSUPP; |
97ac7350 | 307 | |
bfe219d3 | 308 | file_start_write(file); |
14da9200 | 309 | ret = file->f_op->fallocate(file, mode, offset, len); |
820c12d5 HS |
310 | |
311 | /* | |
312 | * Create inotify and fanotify events. | |
313 | * | |
314 | * To keep the logic simple always create events if fallocate succeeds. | |
315 | * This implies that events are even created if the file size remains | |
316 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
317 | */ | |
318 | if (ret == 0) | |
319 | fsnotify_modify(file); | |
320 | ||
bfe219d3 | 321 | file_end_write(file); |
14da9200 | 322 | return ret; |
3e63cbb1 | 323 | } |
72c72bdf | 324 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 325 | |
edf292c7 | 326 | int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len) |
3e63cbb1 | 327 | { |
2903ff01 | 328 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
329 | int error = -EBADF; |
330 | ||
2903ff01 | 331 | if (f.file) { |
72c72bdf | 332 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 333 | fdput(f); |
3e63cbb1 | 334 | } |
3e63cbb1 | 335 | return error; |
97ac7350 | 336 | } |
3e63cbb1 | 337 | |
edf292c7 DB |
338 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
339 | { | |
340 | return ksys_fallocate(fd, mode, offset, len); | |
341 | } | |
342 | ||
1da177e4 LT |
343 | /* |
344 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
345 | * We do this by temporarily clearing all FS-related capabilities and | |
346 | * switching the fsuid/fsgid around to the real ones. | |
347 | */ | |
94704515 | 348 | static const struct cred *access_override_creds(void) |
1da177e4 | 349 | { |
d84f4f99 DH |
350 | const struct cred *old_cred; |
351 | struct cred *override_cred; | |
1da177e4 | 352 | |
d84f4f99 DH |
353 | override_cred = prepare_creds(); |
354 | if (!override_cred) | |
94704515 | 355 | return NULL; |
1da177e4 | 356 | |
d84f4f99 DH |
357 | override_cred->fsuid = override_cred->uid; |
358 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 359 | |
086f7316 | 360 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 361 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
362 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
363 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 364 | cap_clear(override_cred->cap_effective); |
086f7316 | 365 | else |
d84f4f99 DH |
366 | override_cred->cap_effective = |
367 | override_cred->cap_permitted; | |
086f7316 | 368 | } |
1da177e4 | 369 | |
d7852fbd LT |
370 | /* |
371 | * The new set of credentials can *only* be used in | |
372 | * task-synchronous circumstances, and does not need | |
373 | * RCU freeing, unless somebody then takes a separate | |
374 | * reference to it. | |
375 | * | |
376 | * NOTE! This is _only_ true because this credential | |
377 | * is used purely for override_creds() that installs | |
378 | * it as the subjective cred. Other threads will be | |
379 | * accessing ->real_cred, not the subjective cred. | |
380 | * | |
381 | * If somebody _does_ make a copy of this (using the | |
382 | * 'get_current_cred()' function), that will clear the | |
383 | * non_rcu field, because now that other user may be | |
384 | * expecting RCU freeing. But normal thread-synchronous | |
385 | * cred accesses will keep things non-RCY. | |
386 | */ | |
387 | override_cred->non_rcu = 1; | |
388 | ||
d84f4f99 | 389 | old_cred = override_creds(override_cred); |
94704515 MS |
390 | |
391 | /* override_cred() gets its own ref */ | |
392 | put_cred(override_cred); | |
393 | ||
394 | return old_cred; | |
395 | } | |
396 | ||
c8ffd8bc | 397 | long do_faccessat(int dfd, const char __user *filename, int mode, int flags) |
94704515 MS |
398 | { |
399 | struct path path; | |
400 | struct inode *inode; | |
401 | int res; | |
402 | unsigned int lookup_flags = LOOKUP_FOLLOW; | |
c8ffd8bc | 403 | const struct cred *old_cred = NULL; |
94704515 MS |
404 | |
405 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
406 | return -EINVAL; | |
407 | ||
c8ffd8bc MS |
408 | if (flags & ~(AT_EACCESS | AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) |
409 | return -EINVAL; | |
410 | ||
411 | if (flags & AT_SYMLINK_NOFOLLOW) | |
412 | lookup_flags &= ~LOOKUP_FOLLOW; | |
413 | if (flags & AT_EMPTY_PATH) | |
414 | lookup_flags |= LOOKUP_EMPTY; | |
415 | ||
416 | if (!(flags & AT_EACCESS)) { | |
417 | old_cred = access_override_creds(); | |
418 | if (!old_cred) | |
419 | return -ENOMEM; | |
420 | } | |
94704515 | 421 | |
87fa5595 JL |
422 | retry: |
423 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
424 | if (res) |
425 | goto out; | |
426 | ||
63afdfc7 | 427 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
428 | |
429 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
430 | /* |
431 | * MAY_EXEC on regular files is denied if the fs is mounted | |
432 | * with the "noexec" flag. | |
433 | */ | |
434 | res = -EACCES; | |
90f8572b | 435 | if (path_noexec(&path)) |
30524472 AV |
436 | goto out_path_release; |
437 | } | |
438 | ||
256984a8 | 439 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 440 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 441 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 442 | goto out_path_release; |
2f676cbc DH |
443 | /* |
444 | * This is a rare case where using __mnt_is_readonly() | |
445 | * is OK without a mnt_want/drop_write() pair. Since | |
446 | * no actual write to the fs is performed here, we do | |
447 | * not need to telegraph to that to anyone. | |
448 | * | |
449 | * By doing this, we accept that this access is | |
450 | * inherently racy and know that the fs may change | |
451 | * state before we even see this result. | |
452 | */ | |
2d8f3038 | 453 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 454 | res = -EROFS; |
1da177e4 | 455 | |
6902d925 | 456 | out_path_release: |
2d8f3038 | 457 | path_put(&path); |
87fa5595 JL |
458 | if (retry_estale(res, lookup_flags)) { |
459 | lookup_flags |= LOOKUP_REVAL; | |
460 | goto retry; | |
461 | } | |
6902d925 | 462 | out: |
c8ffd8bc MS |
463 | if (old_cred) |
464 | revert_creds(old_cred); | |
465 | ||
1da177e4 LT |
466 | return res; |
467 | } | |
468 | ||
cbfe20f5 DB |
469 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
470 | { | |
c8ffd8bc MS |
471 | return do_faccessat(dfd, filename, mode, 0); |
472 | } | |
473 | ||
474 | SYSCALL_DEFINE4(faccessat2, int, dfd, const char __user *, filename, int, mode, | |
475 | int, flags) | |
476 | { | |
477 | return do_faccessat(dfd, filename, mode, flags); | |
cbfe20f5 DB |
478 | } |
479 | ||
ca013e94 | 480 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d | 481 | { |
c8ffd8bc | 482 | return do_faccessat(AT_FDCWD, filename, mode, 0); |
5590ff0d UD |
483 | } |
484 | ||
447016e9 | 485 | int ksys_chdir(const char __user *filename) |
1da177e4 | 486 | { |
2d8f3038 | 487 | struct path path; |
1da177e4 | 488 | int error; |
0291c0a5 JL |
489 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
490 | retry: | |
491 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
492 | if (error) |
493 | goto out; | |
494 | ||
9cfcac81 | 495 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
496 | if (error) |
497 | goto dput_and_out; | |
498 | ||
2d8f3038 | 499 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
500 | |
501 | dput_and_out: | |
2d8f3038 | 502 | path_put(&path); |
0291c0a5 JL |
503 | if (retry_estale(error, lookup_flags)) { |
504 | lookup_flags |= LOOKUP_REVAL; | |
505 | goto retry; | |
506 | } | |
1da177e4 LT |
507 | out: |
508 | return error; | |
509 | } | |
510 | ||
447016e9 DB |
511 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
512 | { | |
513 | return ksys_chdir(filename); | |
514 | } | |
515 | ||
3cdad428 | 516 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 517 | { |
2903ff01 | 518 | struct fd f = fdget_raw(fd); |
159b0956 | 519 | int error; |
1da177e4 LT |
520 | |
521 | error = -EBADF; | |
2903ff01 | 522 | if (!f.file) |
1da177e4 LT |
523 | goto out; |
524 | ||
1da177e4 | 525 | error = -ENOTDIR; |
159b0956 | 526 | if (!d_can_lookup(f.file->f_path.dentry)) |
1da177e4 LT |
527 | goto out_putf; |
528 | ||
159b0956 | 529 | error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR); |
1da177e4 | 530 | if (!error) |
2903ff01 | 531 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 532 | out_putf: |
2903ff01 | 533 | fdput(f); |
1da177e4 LT |
534 | out: |
535 | return error; | |
536 | } | |
537 | ||
a16fe33a | 538 | int ksys_chroot(const char __user *filename) |
1da177e4 | 539 | { |
2d8f3038 | 540 | struct path path; |
1da177e4 | 541 | int error; |
2771261e JL |
542 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
543 | retry: | |
544 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
545 | if (error) |
546 | goto out; | |
547 | ||
9cfcac81 | 548 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
549 | if (error) |
550 | goto dput_and_out; | |
551 | ||
552 | error = -EPERM; | |
c7b96acf | 553 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 554 | goto dput_and_out; |
8b8efb44 TH |
555 | error = security_path_chroot(&path); |
556 | if (error) | |
557 | goto dput_and_out; | |
1da177e4 | 558 | |
2d8f3038 | 559 | set_fs_root(current->fs, &path); |
1da177e4 LT |
560 | error = 0; |
561 | dput_and_out: | |
2d8f3038 | 562 | path_put(&path); |
2771261e JL |
563 | if (retry_estale(error, lookup_flags)) { |
564 | lookup_flags |= LOOKUP_REVAL; | |
565 | goto retry; | |
566 | } | |
1da177e4 LT |
567 | out: |
568 | return error; | |
569 | } | |
570 | ||
a16fe33a DB |
571 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
572 | { | |
573 | return ksys_chroot(filename); | |
574 | } | |
575 | ||
be01f9f2 | 576 | static int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 577 | { |
e57712eb | 578 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 579 | struct inode *delegated_inode = NULL; |
1da177e4 | 580 | struct iattr newattrs; |
e57712eb | 581 | int error; |
1da177e4 | 582 | |
e57712eb AV |
583 | error = mnt_want_write(path->mnt); |
584 | if (error) | |
585 | return error; | |
27ac0ffe | 586 | retry_deleg: |
5955102c | 587 | inode_lock(inode); |
cdcf116d | 588 | error = security_path_chmod(path, mode); |
e57712eb | 589 | if (error) |
fe542cf5 | 590 | goto out_unlock; |
1da177e4 LT |
591 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
592 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 593 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 594 | out_unlock: |
5955102c | 595 | inode_unlock(inode); |
27ac0ffe BF |
596 | if (delegated_inode) { |
597 | error = break_deleg_wait(&delegated_inode); | |
598 | if (!error) | |
599 | goto retry_deleg; | |
600 | } | |
e57712eb AV |
601 | mnt_drop_write(path->mnt); |
602 | return error; | |
603 | } | |
604 | ||
03450e27 | 605 | int ksys_fchmod(unsigned int fd, umode_t mode) |
e57712eb | 606 | { |
173c8401 | 607 | struct fd f = fdget(fd); |
e57712eb AV |
608 | int err = -EBADF; |
609 | ||
173c8401 | 610 | if (f.file) { |
9f45f5bf | 611 | audit_file(f.file); |
173c8401 AV |
612 | err = chmod_common(&f.file->f_path, mode); |
613 | fdput(f); | |
e57712eb | 614 | } |
1da177e4 LT |
615 | return err; |
616 | } | |
617 | ||
03450e27 DB |
618 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
619 | { | |
620 | return ksys_fchmod(fd, mode); | |
621 | } | |
622 | ||
623 | int do_fchmodat(int dfd, const char __user *filename, umode_t mode) | |
1da177e4 | 624 | { |
2d8f3038 | 625 | struct path path; |
1da177e4 | 626 | int error; |
14ff690c JL |
627 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
628 | retry: | |
629 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
630 | if (!error) { |
631 | error = chmod_common(&path, mode); | |
632 | path_put(&path); | |
14ff690c JL |
633 | if (retry_estale(error, lookup_flags)) { |
634 | lookup_flags |= LOOKUP_REVAL; | |
635 | goto retry; | |
636 | } | |
e57712eb | 637 | } |
1da177e4 LT |
638 | return error; |
639 | } | |
640 | ||
03450e27 DB |
641 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, |
642 | umode_t, mode) | |
643 | { | |
644 | return do_fchmodat(dfd, filename, mode); | |
645 | } | |
646 | ||
49f0a076 | 647 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d | 648 | { |
03450e27 | 649 | return do_fchmodat(AT_FDCWD, filename, mode); |
5590ff0d UD |
650 | } |
651 | ||
7fd25dac | 652 | static int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 653 | { |
fe542cf5 | 654 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 655 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
656 | int error; |
657 | struct iattr newattrs; | |
52137abe EB |
658 | kuid_t uid; |
659 | kgid_t gid; | |
660 | ||
661 | uid = make_kuid(current_user_ns(), user); | |
662 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 663 | |
c1b8940b | 664 | retry_deleg: |
1da177e4 LT |
665 | newattrs.ia_valid = ATTR_CTIME; |
666 | if (user != (uid_t) -1) { | |
52137abe EB |
667 | if (!uid_valid(uid)) |
668 | return -EINVAL; | |
1da177e4 | 669 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 670 | newattrs.ia_uid = uid; |
1da177e4 LT |
671 | } |
672 | if (group != (gid_t) -1) { | |
52137abe EB |
673 | if (!gid_valid(gid)) |
674 | return -EINVAL; | |
1da177e4 | 675 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 676 | newattrs.ia_gid = gid; |
1da177e4 LT |
677 | } |
678 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
679 | newattrs.ia_valid |= |
680 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
5955102c | 681 | inode_lock(inode); |
d2b31ca6 | 682 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 683 | if (!error) |
27ac0ffe | 684 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
5955102c | 685 | inode_unlock(inode); |
27ac0ffe BF |
686 | if (delegated_inode) { |
687 | error = break_deleg_wait(&delegated_inode); | |
688 | if (!error) | |
689 | goto retry_deleg; | |
690 | } | |
1da177e4 LT |
691 | return error; |
692 | } | |
693 | ||
55731b3c DB |
694 | int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group, |
695 | int flag) | |
5590ff0d | 696 | { |
2d8f3038 | 697 | struct path path; |
5590ff0d | 698 | int error = -EINVAL; |
65cfc672 | 699 | int lookup_flags; |
5590ff0d | 700 | |
65cfc672 | 701 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
702 | goto out; |
703 | ||
65cfc672 AV |
704 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
705 | if (flag & AT_EMPTY_PATH) | |
706 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 707 | retry: |
65cfc672 | 708 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
709 | if (error) |
710 | goto out; | |
2d8f3038 | 711 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
712 | if (error) |
713 | goto out_release; | |
fe542cf5 | 714 | error = chown_common(&path, user, group); |
2d8f3038 | 715 | mnt_drop_write(path.mnt); |
2af482a7 | 716 | out_release: |
2d8f3038 | 717 | path_put(&path); |
99a5df37 JL |
718 | if (retry_estale(error, lookup_flags)) { |
719 | lookup_flags |= LOOKUP_REVAL; | |
720 | goto retry; | |
721 | } | |
5590ff0d UD |
722 | out: |
723 | return error; | |
724 | } | |
725 | ||
55731b3c DB |
726 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
727 | gid_t, group, int, flag) | |
728 | { | |
729 | return do_fchownat(dfd, filename, user, group, flag); | |
730 | } | |
731 | ||
55e4def0 | 732 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 733 | { |
55731b3c | 734 | return do_fchownat(AT_FDCWD, filename, user, group, 0); |
55e4def0 | 735 | } |
1da177e4 | 736 | |
55e4def0 DH |
737 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
738 | { | |
55731b3c DB |
739 | return do_fchownat(AT_FDCWD, filename, user, group, |
740 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
741 | } |
742 | ||
55731b3c | 743 | int ksys_fchown(unsigned int fd, uid_t user, gid_t group) |
1da177e4 | 744 | { |
2903ff01 | 745 | struct fd f = fdget(fd); |
1da177e4 LT |
746 | int error = -EBADF; |
747 | ||
2903ff01 | 748 | if (!f.file) |
6902d925 DH |
749 | goto out; |
750 | ||
6742cee0 | 751 | error = mnt_want_write_file(f.file); |
2af482a7 DH |
752 | if (error) |
753 | goto out_fput; | |
9f45f5bf | 754 | audit_file(f.file); |
2903ff01 | 755 | error = chown_common(&f.file->f_path, user, group); |
6742cee0 | 756 | mnt_drop_write_file(f.file); |
2af482a7 | 757 | out_fput: |
2903ff01 | 758 | fdput(f); |
6902d925 | 759 | out: |
1da177e4 LT |
760 | return error; |
761 | } | |
762 | ||
55731b3c DB |
763 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
764 | { | |
765 | return ksys_fchown(fd, user, group); | |
766 | } | |
767 | ||
02e5180d | 768 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 769 | struct inode *inode, |
ae2bb293 | 770 | int (*open)(struct inode *, struct file *)) |
1da177e4 | 771 | { |
1abf0c71 | 772 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
773 | int error; |
774 | ||
b5bcdda3 | 775 | path_get(&f->f_path); |
4bacc9c9 | 776 | f->f_inode = inode; |
1da177e4 | 777 | f->f_mapping = inode->i_mapping; |
5660e13d | 778 | f->f_wb_err = filemap_sample_wb_err(f->f_mapping); |
735e4ae5 | 779 | f->f_sb_err = file_sample_sb_err(f); |
5660e13d | 780 | |
3f4d5a00 | 781 | if (unlikely(f->f_flags & O_PATH)) { |
f5d11409 | 782 | f->f_mode = FMODE_PATH | FMODE_OPENED; |
1abf0c71 | 783 | f->f_op = &empty_fops; |
af04fadc | 784 | return 0; |
1abf0c71 AV |
785 | } |
786 | ||
73601ea5 TH |
787 | /* Any file opened for execve()/uselib() has to be a regular file. */ |
788 | if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) { | |
789 | error = -EACCES; | |
790 | goto cleanup_file; | |
791 | } | |
792 | ||
dd20908a | 793 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 794 | error = get_write_access(inode); |
3f4d5a00 | 795 | if (unlikely(error)) |
1da177e4 | 796 | goto cleanup_file; |
0ccb2863 | 797 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 798 | if (unlikely(error)) { |
0ccb2863 AV |
799 | put_write_access(inode); |
800 | goto cleanup_file; | |
801 | } | |
83f936c7 | 802 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
803 | } |
804 | ||
2be7d348 LT |
805 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
806 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) | |
807 | f->f_mode |= FMODE_ATOMIC_POS; | |
808 | ||
1abf0c71 | 809 | f->f_op = fops_get(inode->i_fop); |
7159d544 | 810 | if (WARN_ON(!f->f_op)) { |
72c2d531 AV |
811 | error = -ENODEV; |
812 | goto cleanup_all; | |
813 | } | |
1abf0c71 | 814 | |
e3f20ae2 | 815 | error = security_file_open(f); |
788e7dd4 YN |
816 | if (error) |
817 | goto cleanup_all; | |
818 | ||
c568d683 | 819 | error = break_lease(locks_inode(f), f->f_flags); |
f3c7691e BF |
820 | if (error) |
821 | goto cleanup_all; | |
822 | ||
ea73ea72 AV |
823 | /* normally all 3 are set; ->open() can clear them if needed */ |
824 | f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE; | |
72c2d531 | 825 | if (!open) |
834f2a4a TM |
826 | open = f->f_op->open; |
827 | if (open) { | |
828 | error = open(inode, f); | |
1da177e4 LT |
829 | if (error) |
830 | goto cleanup_all; | |
831 | } | |
f5d11409 | 832 | f->f_mode |= FMODE_OPENED; |
890275b5 MZ |
833 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
834 | i_readcount_inc(inode); | |
293bc982 | 835 | if ((f->f_mode & FMODE_READ) && |
84363182 | 836 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 837 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 838 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 839 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 840 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 841 | |
c75b1d94 | 842 | f->f_write_hint = WRITE_LIFE_NOT_SET; |
1da177e4 LT |
843 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
844 | ||
845 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
af04fadc | 846 | |
69527c55 AV |
847 | /* NB: we're sure to have correct a_ops only after f_op->open */ |
848 | if (f->f_flags & O_DIRECT) { | |
849 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) | |
850 | return -EINVAL; | |
851 | } | |
09d91cda SL |
852 | |
853 | /* | |
854 | * XXX: Huge page cache doesn't support writing yet. Drop all page | |
855 | * cache for this file before processing writes. | |
856 | */ | |
857 | if ((f->f_mode & FMODE_WRITE) && filemap_nr_thps(inode->i_mapping)) | |
858 | truncate_pagecache(inode, 0); | |
859 | ||
96b7e579 | 860 | return 0; |
1da177e4 LT |
861 | |
862 | cleanup_all: | |
6b4e8085 AV |
863 | if (WARN_ON_ONCE(error > 0)) |
864 | error = -EINVAL; | |
1da177e4 | 865 | fops_put(f->f_op); |
83f936c7 | 866 | if (f->f_mode & FMODE_WRITER) { |
1da177e4 | 867 | put_write_access(inode); |
83f936c7 | 868 | __mnt_drop_write(f->f_path.mnt); |
4a3fd211 | 869 | } |
1da177e4 | 870 | cleanup_file: |
02e5180d AV |
871 | path_put(&f->f_path); |
872 | f->f_path.mnt = NULL; | |
873 | f->f_path.dentry = NULL; | |
dd37978c | 874 | f->f_inode = NULL; |
96b7e579 | 875 | return error; |
1da177e4 LT |
876 | } |
877 | ||
d18e9008 MS |
878 | /** |
879 | * finish_open - finish opening a file | |
0854d450 | 880 | * @file: file pointer |
d18e9008 MS |
881 | * @dentry: pointer to dentry |
882 | * @open: open callback | |
0854d450 | 883 | * @opened: state of open |
d18e9008 MS |
884 | * |
885 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
886 | * | |
887 | * If the open callback is set to NULL, then the standard f_op->open() | |
888 | * filesystem callback is substituted. | |
0854d450 MS |
889 | * |
890 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
891 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
892 | * on it after finish_open(). | |
893 | * | |
0854d450 | 894 | * Returns zero on success or -errno if the open failed. |
d18e9008 | 895 | */ |
30d90494 | 896 | int finish_open(struct file *file, struct dentry *dentry, |
be12af3e | 897 | int (*open)(struct inode *, struct file *)) |
d18e9008 | 898 | { |
aad888f8 | 899 | BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 900 | |
b5bcdda3 | 901 | file->f_path.dentry = dentry; |
aad888f8 | 902 | return do_dentry_open(file, d_backing_inode(dentry), open); |
d18e9008 MS |
903 | } |
904 | EXPORT_SYMBOL(finish_open); | |
905 | ||
906 | /** | |
907 | * finish_no_open - finish ->atomic_open() without opening the file | |
908 | * | |
0854d450 | 909 | * @file: file pointer |
d18e9008 MS |
910 | * @dentry: dentry or NULL (as returned from ->lookup()) |
911 | * | |
912 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
913 | * |
914 | * NB: unlike finish_open() this function does consume the dentry reference and | |
915 | * the caller need not dput() it. | |
916 | * | |
64e1ac4d | 917 | * Returns "0" which must be the return value of ->atomic_open() after having |
0854d450 | 918 | * called this function. |
d18e9008 | 919 | */ |
e45198a6 | 920 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 921 | { |
30d90494 | 922 | file->f_path.dentry = dentry; |
64e1ac4d | 923 | return 0; |
d18e9008 MS |
924 | } |
925 | EXPORT_SYMBOL(finish_no_open); | |
926 | ||
9bf39ab2 MS |
927 | char *file_path(struct file *filp, char *buf, int buflen) |
928 | { | |
929 | return d_path(&filp->f_path, buf, buflen); | |
930 | } | |
931 | EXPORT_SYMBOL(file_path); | |
932 | ||
4bacc9c9 DH |
933 | /** |
934 | * vfs_open - open the file at the given path | |
935 | * @path: path to open | |
936 | * @file: newly allocated file with f_flag initialized | |
937 | * @cred: credentials to use | |
938 | */ | |
ae2bb293 | 939 | int vfs_open(const struct path *path, struct file *file) |
4bacc9c9 | 940 | { |
54d5ca87 | 941 | file->f_path = *path; |
a6518f73 | 942 | return do_dentry_open(file, d_backing_inode(path->dentry), NULL); |
4bacc9c9 DH |
943 | } |
944 | ||
765927b2 | 945 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 946 | const struct cred *cred) |
a1a5b3d9 PS |
947 | { |
948 | int error; | |
949 | struct file *f; | |
950 | ||
e0e81739 DH |
951 | validate_creds(cred); |
952 | ||
c212f9aa | 953 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 954 | BUG_ON(!path->mnt); |
322ee5b3 | 955 | |
ea73ea72 | 956 | f = alloc_empty_file(flags, cred); |
af04fadc | 957 | if (!IS_ERR(f)) { |
ae2bb293 | 958 | error = vfs_open(path, f); |
4d27f326 AV |
959 | if (error) { |
960 | fput(f); | |
af04fadc AV |
961 | f = ERR_PTR(error); |
962 | } | |
2a027e7a AV |
963 | } |
964 | return f; | |
a1a5b3d9 | 965 | } |
1da177e4 LT |
966 | EXPORT_SYMBOL(dentry_open); |
967 | ||
2abc77af AV |
968 | struct file *open_with_fake_path(const struct path *path, int flags, |
969 | struct inode *inode, const struct cred *cred) | |
970 | { | |
d3b1084d | 971 | struct file *f = alloc_empty_file_noaccount(flags, cred); |
2abc77af AV |
972 | if (!IS_ERR(f)) { |
973 | int error; | |
974 | ||
975 | f->f_path = *path; | |
976 | error = do_dentry_open(f, inode, NULL); | |
977 | if (error) { | |
978 | fput(f); | |
979 | f = ERR_PTR(error); | |
980 | } | |
981 | } | |
982 | return f; | |
983 | } | |
984 | EXPORT_SYMBOL(open_with_fake_path); | |
985 | ||
fddb5d43 AS |
986 | #define WILL_CREATE(flags) (flags & (O_CREAT | __O_TMPFILE)) |
987 | #define O_PATH_FLAGS (O_DIRECTORY | O_NOFOLLOW | O_PATH | O_CLOEXEC) | |
988 | ||
35cb6d54 | 989 | inline struct open_how build_open_how(int flags, umode_t mode) |
fddb5d43 AS |
990 | { |
991 | struct open_how how = { | |
992 | .flags = flags & VALID_OPEN_FLAGS, | |
993 | .mode = mode & S_IALLUGO, | |
994 | }; | |
995 | ||
996 | /* O_PATH beats everything else. */ | |
997 | if (how.flags & O_PATH) | |
998 | how.flags &= O_PATH_FLAGS; | |
999 | /* Modes should only be set for create-like flags. */ | |
1000 | if (!WILL_CREATE(how.flags)) | |
1001 | how.mode = 0; | |
1002 | return how; | |
1003 | } | |
1004 | ||
35cb6d54 | 1005 | inline int build_open_flags(const struct open_how *how, struct open_flags *op) |
47c805dc | 1006 | { |
fddb5d43 | 1007 | int flags = how->flags; |
47c805dc | 1008 | int lookup_flags = 0; |
62fb4a15 | 1009 | int acc_mode = ACC_MODE(flags); |
47c805dc | 1010 | |
fddb5d43 AS |
1011 | /* Must never be set by userspace */ |
1012 | flags &= ~(FMODE_NONOTIFY | O_CLOEXEC); | |
1013 | ||
629e014b | 1014 | /* |
fddb5d43 AS |
1015 | * Older syscalls implicitly clear all of the invalid flags or argument |
1016 | * values before calling build_open_flags(), but openat2(2) checks all | |
1017 | * of its arguments. | |
629e014b | 1018 | */ |
fddb5d43 AS |
1019 | if (flags & ~VALID_OPEN_FLAGS) |
1020 | return -EINVAL; | |
1021 | if (how->resolve & ~VALID_RESOLVE_FLAGS) | |
1022 | return -EINVAL; | |
629e014b | 1023 | |
fddb5d43 AS |
1024 | /* Deal with the mode. */ |
1025 | if (WILL_CREATE(flags)) { | |
1026 | if (how->mode & ~S_IALLUGO) | |
1027 | return -EINVAL; | |
1028 | op->mode = how->mode | S_IFREG; | |
1029 | } else { | |
1030 | if (how->mode != 0) | |
1031 | return -EINVAL; | |
e68726ff | 1032 | op->mode = 0; |
fddb5d43 | 1033 | } |
47c805dc AV |
1034 | |
1035 | /* | |
fddb5d43 AS |
1036 | * In order to ensure programs get explicit errors when trying to use |
1037 | * O_TMPFILE on old kernels, O_TMPFILE is implemented such that it | |
1038 | * looks like (O_DIRECTORY|O_RDWR & ~O_CREAT) to old kernels. But we | |
1039 | * have to require userspace to explicitly set it. | |
47c805dc | 1040 | */ |
bb458c64 AV |
1041 | if (flags & __O_TMPFILE) { |
1042 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d | 1043 | return -EINVAL; |
ba57ea64 AV |
1044 | if (!(acc_mode & MAY_WRITE)) |
1045 | return -EINVAL; | |
fddb5d43 AS |
1046 | } |
1047 | if (flags & O_PATH) { | |
1048 | /* O_PATH only permits certain other flags to be set. */ | |
1049 | if (flags & ~O_PATH_FLAGS) | |
1050 | return -EINVAL; | |
1abf0c71 | 1051 | acc_mode = 0; |
1abf0c71 | 1052 | } |
47c805dc | 1053 | |
fddb5d43 AS |
1054 | /* |
1055 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1056 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1057 | * always set instead of having to deal with possibly weird behaviour | |
1058 | * for malicious applications setting only __O_SYNC. | |
1059 | */ | |
1060 | if (flags & __O_SYNC) | |
1061 | flags |= O_DSYNC; | |
1062 | ||
1abf0c71 | 1063 | op->open_flag = flags; |
47c805dc AV |
1064 | |
1065 | /* O_TRUNC implies we need access checks for write permissions */ | |
1066 | if (flags & O_TRUNC) | |
1067 | acc_mode |= MAY_WRITE; | |
1068 | ||
1069 | /* Allow the LSM permission hook to distinguish append | |
1070 | access from general write access. */ | |
1071 | if (flags & O_APPEND) | |
1072 | acc_mode |= MAY_APPEND; | |
1073 | ||
1074 | op->acc_mode = acc_mode; | |
1075 | ||
1abf0c71 AV |
1076 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
1077 | ||
47c805dc AV |
1078 | if (flags & O_CREAT) { |
1079 | op->intent |= LOOKUP_CREATE; | |
31d1726d | 1080 | if (flags & O_EXCL) { |
47c805dc | 1081 | op->intent |= LOOKUP_EXCL; |
31d1726d AV |
1082 | flags |= O_NOFOLLOW; |
1083 | } | |
47c805dc AV |
1084 | } |
1085 | ||
1086 | if (flags & O_DIRECTORY) | |
1087 | lookup_flags |= LOOKUP_DIRECTORY; | |
1088 | if (!(flags & O_NOFOLLOW)) | |
1089 | lookup_flags |= LOOKUP_FOLLOW; | |
fddb5d43 AS |
1090 | |
1091 | if (how->resolve & RESOLVE_NO_XDEV) | |
1092 | lookup_flags |= LOOKUP_NO_XDEV; | |
1093 | if (how->resolve & RESOLVE_NO_MAGICLINKS) | |
1094 | lookup_flags |= LOOKUP_NO_MAGICLINKS; | |
1095 | if (how->resolve & RESOLVE_NO_SYMLINKS) | |
1096 | lookup_flags |= LOOKUP_NO_SYMLINKS; | |
1097 | if (how->resolve & RESOLVE_BENEATH) | |
1098 | lookup_flags |= LOOKUP_BENEATH; | |
1099 | if (how->resolve & RESOLVE_IN_ROOT) | |
1100 | lookup_flags |= LOOKUP_IN_ROOT; | |
1101 | ||
f9652e10 AV |
1102 | op->lookup_flags = lookup_flags; |
1103 | return 0; | |
47c805dc AV |
1104 | } |
1105 | ||
669abf4e JL |
1106 | /** |
1107 | * file_open_name - open file and return file pointer | |
1108 | * | |
1109 | * @name: struct filename containing path to open | |
1110 | * @flags: open flags as per the open(2) second argument | |
1111 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1112 | * | |
1113 | * This is the helper to open a file from kernelspace if you really | |
1114 | * have to. But in generally you should not do this, so please move | |
1115 | * along, nothing to see here.. | |
1116 | */ | |
1117 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
1118 | { | |
1119 | struct open_flags op; | |
fddb5d43 AS |
1120 | struct open_how how = build_open_how(flags, mode); |
1121 | int err = build_open_flags(&how, &op); | |
1122 | if (err) | |
1123 | return ERR_PTR(err); | |
1124 | return do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
1125 | } |
1126 | ||
47c805dc AV |
1127 | /** |
1128 | * filp_open - open file and return file pointer | |
1129 | * | |
1130 | * @filename: path to open | |
1131 | * @flags: open flags as per the open(2) second argument | |
1132 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1133 | * | |
1134 | * This is the helper to open a file from kernelspace if you really | |
1135 | * have to. But in generally you should not do this, so please move | |
1136 | * along, nothing to see here.. | |
1137 | */ | |
a218d0fd | 1138 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 1139 | { |
51689104 PM |
1140 | struct filename *name = getname_kernel(filename); |
1141 | struct file *file = ERR_CAST(name); | |
1142 | ||
1143 | if (!IS_ERR(name)) { | |
1144 | file = file_open_name(name, flags, mode); | |
1145 | putname(name); | |
1146 | } | |
1147 | return file; | |
47c805dc AV |
1148 | } |
1149 | EXPORT_SYMBOL(filp_open); | |
1150 | ||
73d049a4 | 1151 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
378c6520 | 1152 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1153 | { |
1154 | struct open_flags op; | |
fddb5d43 AS |
1155 | struct open_how how = build_open_how(flags, mode); |
1156 | int err = build_open_flags(&how, &op); | |
f9652e10 AV |
1157 | if (err) |
1158 | return ERR_PTR(err); | |
f9652e10 | 1159 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
1160 | } |
1161 | EXPORT_SYMBOL(file_open_root); | |
1162 | ||
fddb5d43 AS |
1163 | static long do_sys_openat2(int dfd, const char __user *filename, |
1164 | struct open_how *how) | |
1da177e4 | 1165 | { |
47c805dc | 1166 | struct open_flags op; |
fddb5d43 | 1167 | int fd = build_open_flags(how, &op); |
f9652e10 AV |
1168 | struct filename *tmp; |
1169 | ||
1170 | if (fd) | |
1171 | return fd; | |
1172 | ||
1173 | tmp = getname(filename); | |
1174 | if (IS_ERR(tmp)) | |
1175 | return PTR_ERR(tmp); | |
1176 | ||
fddb5d43 | 1177 | fd = get_unused_fd_flags(how->flags); |
f9652e10 AV |
1178 | if (fd >= 0) { |
1179 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1180 | if (IS_ERR(f)) { | |
1181 | put_unused_fd(fd); | |
1182 | fd = PTR_ERR(f); | |
1183 | } else { | |
1184 | fsnotify_open(f); | |
1185 | fd_install(fd, f); | |
1da177e4 | 1186 | } |
1da177e4 | 1187 | } |
f9652e10 | 1188 | putname(tmp); |
1da177e4 | 1189 | return fd; |
1da177e4 | 1190 | } |
e922efc3 | 1191 | |
fddb5d43 | 1192 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
e922efc3 | 1193 | { |
fddb5d43 AS |
1194 | struct open_how how = build_open_how(flags, mode); |
1195 | return do_sys_openat2(dfd, filename, &how); | |
1196 | } | |
e922efc3 | 1197 | |
fddb5d43 AS |
1198 | |
1199 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1200 | { | |
1201 | return ksys_open(filename, flags, mode); | |
e922efc3 | 1202 | } |
1da177e4 | 1203 | |
6559eed8 | 1204 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1205 | umode_t, mode) |
5590ff0d UD |
1206 | { |
1207 | if (force_o_largefile()) | |
1208 | flags |= O_LARGEFILE; | |
2cf09666 | 1209 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1210 | } |
5590ff0d | 1211 | |
fddb5d43 AS |
1212 | SYSCALL_DEFINE4(openat2, int, dfd, const char __user *, filename, |
1213 | struct open_how __user *, how, size_t, usize) | |
1214 | { | |
1215 | int err; | |
1216 | struct open_how tmp; | |
1217 | ||
1218 | BUILD_BUG_ON(sizeof(struct open_how) < OPEN_HOW_SIZE_VER0); | |
1219 | BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_LATEST); | |
1220 | ||
1221 | if (unlikely(usize < OPEN_HOW_SIZE_VER0)) | |
1222 | return -EINVAL; | |
1223 | ||
1224 | err = copy_struct_from_user(&tmp, sizeof(tmp), how, usize); | |
1225 | if (err) | |
1226 | return err; | |
1227 | ||
1228 | /* O_LARGEFILE is only allowed for non-O_PATH. */ | |
1229 | if (!(tmp.flags & O_PATH) && force_o_largefile()) | |
1230 | tmp.flags |= O_LARGEFILE; | |
1231 | ||
1232 | return do_sys_openat2(dfd, filename, &tmp); | |
1233 | } | |
1234 | ||
e35d49f6 AV |
1235 | #ifdef CONFIG_COMPAT |
1236 | /* | |
1237 | * Exactly like sys_open(), except that it doesn't set the | |
1238 | * O_LARGEFILE flag. | |
1239 | */ | |
1240 | COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1241 | { | |
1242 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
1243 | } | |
1244 | ||
1245 | /* | |
1246 | * Exactly like sys_openat(), except that it doesn't set the | |
1247 | * O_LARGEFILE flag. | |
1248 | */ | |
1249 | COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode) | |
1250 | { | |
1251 | return do_sys_open(dfd, filename, flags, mode); | |
1252 | } | |
1253 | #endif | |
1254 | ||
1da177e4 LT |
1255 | #ifndef __alpha__ |
1256 | ||
1257 | /* | |
1258 | * For backward compatibility? Maybe this should be moved | |
1259 | * into arch/i386 instead? | |
1260 | */ | |
a218d0fd | 1261 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 | 1262 | { |
bae217ea | 1263 | return ksys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); |
1da177e4 LT |
1264 | } |
1265 | ||
1266 | #endif | |
1267 | ||
1268 | /* | |
1269 | * "id" is the POSIX thread ID. We use the | |
1270 | * files pointer for this.. | |
1271 | */ | |
1272 | int filp_close(struct file *filp, fl_owner_t id) | |
1273 | { | |
45778ca8 | 1274 | int retval = 0; |
1da177e4 LT |
1275 | |
1276 | if (!file_count(filp)) { | |
1277 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1278 | return 0; |
1da177e4 LT |
1279 | } |
1280 | ||
72c2d531 | 1281 | if (filp->f_op->flush) |
75e1fcc0 | 1282 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1283 | |
1abf0c71 AV |
1284 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1285 | dnotify_flush(filp, id); | |
1286 | locks_remove_posix(filp, id); | |
1287 | } | |
1da177e4 LT |
1288 | fput(filp); |
1289 | return retval; | |
1290 | } | |
1291 | ||
1292 | EXPORT_SYMBOL(filp_close); | |
1293 | ||
1294 | /* | |
1295 | * Careful here! We test whether the file pointer is NULL before | |
1296 | * releasing the fd. This ensures that one clone task can't release | |
1297 | * an fd while another clone is opening it. | |
1298 | */ | |
ca013e94 | 1299 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1300 | { |
483ce1d4 | 1301 | int retval = __close_fd(current->files, fd); |
ee731f4f EP |
1302 | |
1303 | /* can't restart close syscall because file table entry was cleared */ | |
1304 | if (unlikely(retval == -ERESTARTSYS || | |
1305 | retval == -ERESTARTNOINTR || | |
1306 | retval == -ERESTARTNOHAND || | |
1307 | retval == -ERESTART_RESTARTBLOCK)) | |
1308 | retval = -EINTR; | |
1309 | ||
1310 | return retval; | |
1da177e4 | 1311 | } |
1da177e4 LT |
1312 | |
1313 | /* | |
1314 | * This routine simulates a hangup on the tty, to arrange that users | |
1315 | * are given clean terminals at login time. | |
1316 | */ | |
ca013e94 | 1317 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1318 | { |
1319 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1320 | tty_vhangup_self(); |
1da177e4 LT |
1321 | return 0; |
1322 | } | |
1323 | return -EPERM; | |
1324 | } | |
1325 | ||
1326 | /* | |
1327 | * Called when an inode is about to be open. | |
1328 | * We use this to disallow opening large files on 32bit systems if | |
1329 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1330 | * on this flag in sys_open. | |
1331 | */ | |
1332 | int generic_file_open(struct inode * inode, struct file * filp) | |
1333 | { | |
1334 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1335 | return -EOVERFLOW; |
1da177e4 LT |
1336 | return 0; |
1337 | } | |
1338 | ||
1339 | EXPORT_SYMBOL(generic_file_open); | |
1340 | ||
1341 | /* | |
1342 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1343 | * file descriptors. The function is not supposed to ever fail, the only |
1344 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1345 | * directly into file_operations structure. | |
1da177e4 LT |
1346 | */ |
1347 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1348 | { | |
1349 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1350 | return 0; | |
1351 | } | |
1352 | ||
1353 | EXPORT_SYMBOL(nonseekable_open); | |
10dce8af KS |
1354 | |
1355 | /* | |
1356 | * stream_open is used by subsystems that want stream-like file descriptors. | |
1357 | * Such file descriptors are not seekable and don't have notion of position | |
438ab720 KS |
1358 | * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL). |
1359 | * Contrary to file descriptors of other regular files, .read() and .write() | |
1360 | * can run simultaneously. | |
10dce8af KS |
1361 | * |
1362 | * stream_open never fails and is marked to return int so that it could be | |
1363 | * directly used as file_operations.open . | |
1364 | */ | |
1365 | int stream_open(struct inode *inode, struct file *filp) | |
1366 | { | |
2be7d348 | 1367 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS); |
10dce8af KS |
1368 | filp->f_mode |= FMODE_STREAM; |
1369 | return 0; | |
1370 | } | |
1371 | ||
1372 | EXPORT_SYMBOL(stream_open); |