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 | ||
643fe55a CB |
38 | int do_truncate(struct user_namespace *mnt_userns, struct dentry *dentry, |
39 | loff_t length, unsigned int time_attrs, 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 | 63 | /* Note any delegations or leases have already been broken: */ |
643fe55a | 64 | ret = notify_change(mnt_userns, 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 | { |
643fe55a | 71 | struct user_namespace *mnt_userns; |
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 | |
643fe55a CB |
87 | mnt_userns = mnt_user_ns(path->mnt); |
88 | error = inode_permission(mnt_userns, inode, MAY_WRITE); | |
9ac9b847 DH |
89 | if (error) |
90 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
91 | |
92 | error = -EPERM; | |
c82e42da | 93 | if (IS_APPEND(inode)) |
9ac9b847 | 94 | goto mnt_drop_write_and_out; |
1da177e4 | 95 | |
8cf9ee50 | 96 | error = get_write_access(inode); |
1da177e4 | 97 | if (error) |
9ac9b847 | 98 | goto mnt_drop_write_and_out; |
1da177e4 | 99 | |
9700382c | 100 | /* |
101 | * Make sure that there are no leases. get_write_access() protects | |
102 | * against the truncate racing with a lease-granting setlease(). | |
103 | */ | |
8737c930 | 104 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 105 | if (error) |
9700382c | 106 | goto put_write_and_out; |
1da177e4 LT |
107 | |
108 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 | 109 | if (!error) |
a02de960 | 110 | error = security_path_truncate(path); |
907f4554 | 111 | if (!error) |
643fe55a | 112 | error = do_truncate(mnt_userns, path->dentry, length, 0, NULL); |
1da177e4 | 113 | |
9700382c | 114 | put_write_and_out: |
8cf9ee50 | 115 | put_write_access(inode); |
9ac9b847 | 116 | mnt_drop_write_and_out: |
a02de960 | 117 | mnt_drop_write(path->mnt); |
1da177e4 LT |
118 | out: |
119 | return error; | |
120 | } | |
a02de960 DH |
121 | EXPORT_SYMBOL_GPL(vfs_truncate); |
122 | ||
df260e21 | 123 | long do_sys_truncate(const char __user *pathname, loff_t length) |
a02de960 | 124 | { |
48f7530d | 125 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
126 | struct path path; |
127 | int error; | |
128 | ||
129 | if (length < 0) /* sorry, but loff_t says... */ | |
130 | return -EINVAL; | |
131 | ||
48f7530d JL |
132 | retry: |
133 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
134 | if (!error) { |
135 | error = vfs_truncate(&path, length); | |
136 | path_put(&path); | |
137 | } | |
48f7530d JL |
138 | if (retry_estale(error, lookup_flags)) { |
139 | lookup_flags |= LOOKUP_REVAL; | |
140 | goto retry; | |
141 | } | |
a02de960 DH |
142 | return error; |
143 | } | |
1da177e4 | 144 | |
4fd8da8d | 145 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 146 | { |
4fd8da8d | 147 | return do_sys_truncate(path, length); |
1da177e4 LT |
148 | } |
149 | ||
3f6d078d AV |
150 | #ifdef CONFIG_COMPAT |
151 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
152 | { | |
153 | return do_sys_truncate(path, length); | |
154 | } | |
155 | #endif | |
156 | ||
411d9475 | 157 | long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 | 158 | { |
bf2965d5 | 159 | struct inode *inode; |
1da177e4 | 160 | struct dentry *dentry; |
2903ff01 | 161 | struct fd f; |
1da177e4 LT |
162 | int error; |
163 | ||
164 | error = -EINVAL; | |
165 | if (length < 0) | |
166 | goto out; | |
167 | error = -EBADF; | |
2903ff01 AV |
168 | f = fdget(fd); |
169 | if (!f.file) | |
1da177e4 LT |
170 | goto out; |
171 | ||
172 | /* explicitly opened as large or we are on 64-bit box */ | |
2903ff01 | 173 | if (f.file->f_flags & O_LARGEFILE) |
1da177e4 LT |
174 | small = 0; |
175 | ||
2903ff01 | 176 | dentry = f.file->f_path.dentry; |
1da177e4 LT |
177 | inode = dentry->d_inode; |
178 | error = -EINVAL; | |
2903ff01 | 179 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
180 | goto out_putf; |
181 | ||
182 | error = -EINVAL; | |
183 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
184 | if (small && length > MAX_NON_LFS) | |
185 | goto out_putf; | |
186 | ||
187 | error = -EPERM; | |
78757af6 AG |
188 | /* Check IS_APPEND on real upper inode */ |
189 | if (IS_APPEND(file_inode(f.file))) | |
1da177e4 | 190 | goto out_putf; |
14da9200 | 191 | sb_start_write(inode->i_sb); |
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) |
643fe55a CB |
196 | error = do_truncate(file_mnt_user_ns(f.file), dentry, length, |
197 | ATTR_MTIME | ATTR_CTIME, f.file); | |
14da9200 | 198 | sb_end_write(inode->i_sb); |
1da177e4 | 199 | out_putf: |
2903ff01 | 200 | fdput(f); |
1da177e4 LT |
201 | out: |
202 | return error; | |
203 | } | |
204 | ||
bdc480e3 | 205 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 206 | { |
2cf09666 | 207 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
208 | } |
209 | ||
3f6d078d AV |
210 | #ifdef CONFIG_COMPAT |
211 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
212 | { | |
213 | return do_sys_ftruncate(fd, length, 1); | |
214 | } | |
215 | #endif | |
216 | ||
1da177e4 LT |
217 | /* LFS versions of truncate are only needed on 32 bit machines */ |
218 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 219 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
220 | { |
221 | return do_sys_truncate(path, length); | |
222 | } | |
223 | ||
4a0fd5bf | 224 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 225 | { |
2cf09666 | 226 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 227 | } |
6673e0c3 | 228 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 229 | |
3e63cbb1 | 230 | |
72c72bdf | 231 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 232 | { |
496ad9aa | 233 | struct inode *inode = file_inode(file); |
3e63cbb1 | 234 | long ret; |
97ac7350 AA |
235 | |
236 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 237 | return -EINVAL; |
97ac7350 AA |
238 | |
239 | /* Return error if mode is not supported */ | |
dd46c787 | 240 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) |
409332b6 LC |
241 | return -EOPNOTSUPP; |
242 | ||
243 | /* Punch hole and zero range are mutually exclusive */ | |
244 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
245 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
246 | return -EOPNOTSUPP; |
247 | ||
248 | /* Punch hole must have keep size set */ | |
249 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
250 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 251 | return -EOPNOTSUPP; |
97ac7350 | 252 | |
00f5e619 NJ |
253 | /* Collapse range should only be used exclusively. */ |
254 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
255 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
256 | return -EINVAL; | |
257 | ||
dd46c787 NJ |
258 | /* Insert range should only be used exclusively. */ |
259 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
260 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
261 | return -EINVAL; | |
262 | ||
71be6b49 DW |
263 | /* Unshare range should only be used with allocate mode. */ |
264 | if ((mode & FALLOC_FL_UNSHARE_RANGE) && | |
265 | (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE))) | |
266 | return -EINVAL; | |
267 | ||
97ac7350 | 268 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 269 | return -EBADF; |
1ca551c6 | 270 | |
00f5e619 | 271 | /* |
8fc61d92 | 272 | * We can only allow pure fallocate on append only files |
00f5e619 | 273 | */ |
8fc61d92 | 274 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
275 | return -EPERM; |
276 | ||
277 | if (IS_IMMUTABLE(inode)) | |
278 | return -EPERM; | |
279 | ||
0790b31b | 280 | /* |
6d2b6170 | 281 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
282 | */ |
283 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 284 | return -ETXTBSY; |
0790b31b | 285 | |
97ac7350 AA |
286 | /* |
287 | * Revalidate the write permissions, in case security policy has | |
288 | * changed since the files were opened. | |
289 | */ | |
290 | ret = security_file_permission(file, MAY_WRITE); | |
291 | if (ret) | |
3e63cbb1 | 292 | return ret; |
97ac7350 | 293 | |
97ac7350 | 294 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 295 | return -ESPIPE; |
97ac7350 | 296 | |
9e79b132 AG |
297 | if (S_ISDIR(inode->i_mode)) |
298 | return -EISDIR; | |
299 | ||
300 | if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode)) | |
3e63cbb1 | 301 | return -ENODEV; |
97ac7350 | 302 | |
97ac7350 AA |
303 | /* Check for wrap through zero too */ |
304 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 305 | return -EFBIG; |
97ac7350 | 306 | |
2fe17c10 | 307 | if (!file->f_op->fallocate) |
3e63cbb1 | 308 | return -EOPNOTSUPP; |
97ac7350 | 309 | |
bfe219d3 | 310 | file_start_write(file); |
14da9200 | 311 | ret = file->f_op->fallocate(file, mode, offset, len); |
820c12d5 HS |
312 | |
313 | /* | |
314 | * Create inotify and fanotify events. | |
315 | * | |
316 | * To keep the logic simple always create events if fallocate succeeds. | |
317 | * This implies that events are even created if the file size remains | |
318 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
319 | */ | |
320 | if (ret == 0) | |
321 | fsnotify_modify(file); | |
322 | ||
bfe219d3 | 323 | file_end_write(file); |
14da9200 | 324 | return ret; |
3e63cbb1 | 325 | } |
72c72bdf | 326 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 327 | |
edf292c7 | 328 | int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len) |
3e63cbb1 | 329 | { |
2903ff01 | 330 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
331 | int error = -EBADF; |
332 | ||
2903ff01 | 333 | if (f.file) { |
72c72bdf | 334 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 335 | fdput(f); |
3e63cbb1 | 336 | } |
3e63cbb1 | 337 | return error; |
97ac7350 | 338 | } |
3e63cbb1 | 339 | |
edf292c7 DB |
340 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
341 | { | |
342 | return ksys_fallocate(fd, mode, offset, len); | |
343 | } | |
344 | ||
1da177e4 LT |
345 | /* |
346 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
347 | * We do this by temporarily clearing all FS-related capabilities and | |
348 | * switching the fsuid/fsgid around to the real ones. | |
349 | */ | |
94704515 | 350 | static const struct cred *access_override_creds(void) |
1da177e4 | 351 | { |
d84f4f99 DH |
352 | const struct cred *old_cred; |
353 | struct cred *override_cred; | |
1da177e4 | 354 | |
d84f4f99 DH |
355 | override_cred = prepare_creds(); |
356 | if (!override_cred) | |
94704515 | 357 | return NULL; |
1da177e4 | 358 | |
d84f4f99 DH |
359 | override_cred->fsuid = override_cred->uid; |
360 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 361 | |
086f7316 | 362 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 363 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
364 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
365 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 366 | cap_clear(override_cred->cap_effective); |
086f7316 | 367 | else |
d84f4f99 DH |
368 | override_cred->cap_effective = |
369 | override_cred->cap_permitted; | |
086f7316 | 370 | } |
1da177e4 | 371 | |
d7852fbd LT |
372 | /* |
373 | * The new set of credentials can *only* be used in | |
374 | * task-synchronous circumstances, and does not need | |
375 | * RCU freeing, unless somebody then takes a separate | |
376 | * reference to it. | |
377 | * | |
378 | * NOTE! This is _only_ true because this credential | |
379 | * is used purely for override_creds() that installs | |
380 | * it as the subjective cred. Other threads will be | |
381 | * accessing ->real_cred, not the subjective cred. | |
382 | * | |
383 | * If somebody _does_ make a copy of this (using the | |
384 | * 'get_current_cred()' function), that will clear the | |
385 | * non_rcu field, because now that other user may be | |
386 | * expecting RCU freeing. But normal thread-synchronous | |
387 | * cred accesses will keep things non-RCY. | |
388 | */ | |
389 | override_cred->non_rcu = 1; | |
390 | ||
d84f4f99 | 391 | old_cred = override_creds(override_cred); |
94704515 MS |
392 | |
393 | /* override_cred() gets its own ref */ | |
394 | put_cred(override_cred); | |
395 | ||
396 | return old_cred; | |
397 | } | |
398 | ||
eb9d7d39 | 399 | static long do_faccessat(int dfd, const char __user *filename, int mode, int flags) |
94704515 MS |
400 | { |
401 | struct path path; | |
402 | struct inode *inode; | |
403 | int res; | |
404 | unsigned int lookup_flags = LOOKUP_FOLLOW; | |
c8ffd8bc | 405 | const struct cred *old_cred = NULL; |
94704515 MS |
406 | |
407 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
408 | return -EINVAL; | |
409 | ||
c8ffd8bc MS |
410 | if (flags & ~(AT_EACCESS | AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) |
411 | return -EINVAL; | |
412 | ||
413 | if (flags & AT_SYMLINK_NOFOLLOW) | |
414 | lookup_flags &= ~LOOKUP_FOLLOW; | |
415 | if (flags & AT_EMPTY_PATH) | |
416 | lookup_flags |= LOOKUP_EMPTY; | |
417 | ||
418 | if (!(flags & AT_EACCESS)) { | |
419 | old_cred = access_override_creds(); | |
420 | if (!old_cred) | |
421 | return -ENOMEM; | |
422 | } | |
94704515 | 423 | |
87fa5595 JL |
424 | retry: |
425 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
426 | if (res) |
427 | goto out; | |
428 | ||
63afdfc7 | 429 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
430 | |
431 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
432 | /* |
433 | * MAY_EXEC on regular files is denied if the fs is mounted | |
434 | * with the "noexec" flag. | |
435 | */ | |
436 | res = -EACCES; | |
90f8572b | 437 | if (path_noexec(&path)) |
30524472 AV |
438 | goto out_path_release; |
439 | } | |
440 | ||
47291baa | 441 | res = inode_permission(&init_user_ns, inode, mode | MAY_ACCESS); |
6902d925 | 442 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 443 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 444 | goto out_path_release; |
2f676cbc DH |
445 | /* |
446 | * This is a rare case where using __mnt_is_readonly() | |
447 | * is OK without a mnt_want/drop_write() pair. Since | |
448 | * no actual write to the fs is performed here, we do | |
449 | * not need to telegraph to that to anyone. | |
450 | * | |
451 | * By doing this, we accept that this access is | |
452 | * inherently racy and know that the fs may change | |
453 | * state before we even see this result. | |
454 | */ | |
2d8f3038 | 455 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 456 | res = -EROFS; |
1da177e4 | 457 | |
6902d925 | 458 | out_path_release: |
2d8f3038 | 459 | path_put(&path); |
87fa5595 JL |
460 | if (retry_estale(res, lookup_flags)) { |
461 | lookup_flags |= LOOKUP_REVAL; | |
462 | goto retry; | |
463 | } | |
6902d925 | 464 | out: |
c8ffd8bc MS |
465 | if (old_cred) |
466 | revert_creds(old_cred); | |
467 | ||
1da177e4 LT |
468 | return res; |
469 | } | |
470 | ||
cbfe20f5 DB |
471 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
472 | { | |
c8ffd8bc MS |
473 | return do_faccessat(dfd, filename, mode, 0); |
474 | } | |
475 | ||
476 | SYSCALL_DEFINE4(faccessat2, int, dfd, const char __user *, filename, int, mode, | |
477 | int, flags) | |
478 | { | |
479 | return do_faccessat(dfd, filename, mode, flags); | |
cbfe20f5 DB |
480 | } |
481 | ||
ca013e94 | 482 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d | 483 | { |
c8ffd8bc | 484 | return do_faccessat(AT_FDCWD, filename, mode, 0); |
5590ff0d UD |
485 | } |
486 | ||
db63f1e3 | 487 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 488 | { |
2d8f3038 | 489 | struct path path; |
1da177e4 | 490 | int error; |
0291c0a5 JL |
491 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
492 | retry: | |
493 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
494 | if (error) |
495 | goto out; | |
496 | ||
02f92b38 | 497 | error = path_permission(&path, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
498 | if (error) |
499 | goto dput_and_out; | |
500 | ||
2d8f3038 | 501 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
502 | |
503 | dput_and_out: | |
2d8f3038 | 504 | path_put(&path); |
0291c0a5 JL |
505 | if (retry_estale(error, lookup_flags)) { |
506 | lookup_flags |= LOOKUP_REVAL; | |
507 | goto retry; | |
508 | } | |
1da177e4 LT |
509 | out: |
510 | return error; | |
511 | } | |
512 | ||
3cdad428 | 513 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 514 | { |
2903ff01 | 515 | struct fd f = fdget_raw(fd); |
159b0956 | 516 | int error; |
1da177e4 LT |
517 | |
518 | error = -EBADF; | |
2903ff01 | 519 | if (!f.file) |
1da177e4 LT |
520 | goto out; |
521 | ||
1da177e4 | 522 | error = -ENOTDIR; |
159b0956 | 523 | if (!d_can_lookup(f.file->f_path.dentry)) |
1da177e4 LT |
524 | goto out_putf; |
525 | ||
02f92b38 | 526 | error = file_permission(f.file, MAY_EXEC | MAY_CHDIR); |
1da177e4 | 527 | if (!error) |
2903ff01 | 528 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 529 | out_putf: |
2903ff01 | 530 | fdput(f); |
1da177e4 LT |
531 | out: |
532 | return error; | |
533 | } | |
534 | ||
4b7ca501 | 535 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 536 | { |
2d8f3038 | 537 | struct path path; |
1da177e4 | 538 | int error; |
2771261e JL |
539 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
540 | retry: | |
541 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
542 | if (error) |
543 | goto out; | |
544 | ||
02f92b38 | 545 | error = path_permission(&path, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
546 | if (error) |
547 | goto dput_and_out; | |
548 | ||
549 | error = -EPERM; | |
c7b96acf | 550 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 551 | goto dput_and_out; |
8b8efb44 TH |
552 | error = security_path_chroot(&path); |
553 | if (error) | |
554 | goto dput_and_out; | |
1da177e4 | 555 | |
2d8f3038 | 556 | set_fs_root(current->fs, &path); |
1da177e4 LT |
557 | error = 0; |
558 | dput_and_out: | |
2d8f3038 | 559 | path_put(&path); |
2771261e JL |
560 | if (retry_estale(error, lookup_flags)) { |
561 | lookup_flags |= LOOKUP_REVAL; | |
562 | goto retry; | |
563 | } | |
1da177e4 LT |
564 | out: |
565 | return error; | |
566 | } | |
567 | ||
1097742e | 568 | int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 569 | { |
e57712eb | 570 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 571 | struct inode *delegated_inode = NULL; |
1da177e4 | 572 | struct iattr newattrs; |
e57712eb | 573 | int error; |
1da177e4 | 574 | |
e57712eb AV |
575 | error = mnt_want_write(path->mnt); |
576 | if (error) | |
577 | return error; | |
27ac0ffe | 578 | retry_deleg: |
5955102c | 579 | inode_lock(inode); |
cdcf116d | 580 | error = security_path_chmod(path, mode); |
e57712eb | 581 | if (error) |
fe542cf5 | 582 | goto out_unlock; |
1da177e4 LT |
583 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
584 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
2f221d6f CB |
585 | error = notify_change(&init_user_ns, path->dentry, &newattrs, |
586 | &delegated_inode); | |
fe542cf5 | 587 | out_unlock: |
5955102c | 588 | inode_unlock(inode); |
27ac0ffe BF |
589 | if (delegated_inode) { |
590 | error = break_deleg_wait(&delegated_inode); | |
591 | if (!error) | |
592 | goto retry_deleg; | |
593 | } | |
e57712eb AV |
594 | mnt_drop_write(path->mnt); |
595 | return error; | |
596 | } | |
597 | ||
9e96c8c0 CH |
598 | int vfs_fchmod(struct file *file, umode_t mode) |
599 | { | |
600 | audit_file(file); | |
601 | return chmod_common(&file->f_path, mode); | |
602 | } | |
603 | ||
b25ba7c3 | 604 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 605 | { |
173c8401 | 606 | struct fd f = fdget(fd); |
e57712eb AV |
607 | int err = -EBADF; |
608 | ||
173c8401 | 609 | if (f.file) { |
9e96c8c0 | 610 | err = vfs_fchmod(f.file, mode); |
173c8401 | 611 | fdput(f); |
e57712eb | 612 | } |
1da177e4 LT |
613 | return err; |
614 | } | |
615 | ||
1097742e | 616 | static int do_fchmodat(int dfd, const char __user *filename, umode_t mode) |
1da177e4 | 617 | { |
2d8f3038 | 618 | struct path path; |
1da177e4 | 619 | int error; |
14ff690c JL |
620 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
621 | retry: | |
622 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
623 | if (!error) { |
624 | error = chmod_common(&path, mode); | |
625 | path_put(&path); | |
14ff690c JL |
626 | if (retry_estale(error, lookup_flags)) { |
627 | lookup_flags |= LOOKUP_REVAL; | |
628 | goto retry; | |
629 | } | |
e57712eb | 630 | } |
1da177e4 LT |
631 | return error; |
632 | } | |
633 | ||
03450e27 DB |
634 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, |
635 | umode_t, mode) | |
636 | { | |
637 | return do_fchmodat(dfd, filename, mode); | |
638 | } | |
639 | ||
49f0a076 | 640 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d | 641 | { |
03450e27 | 642 | return do_fchmodat(AT_FDCWD, filename, mode); |
5590ff0d UD |
643 | } |
644 | ||
b873498f | 645 | int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 646 | { |
fe542cf5 | 647 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 648 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
649 | int error; |
650 | struct iattr newattrs; | |
52137abe EB |
651 | kuid_t uid; |
652 | kgid_t gid; | |
653 | ||
654 | uid = make_kuid(current_user_ns(), user); | |
655 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 656 | |
c1b8940b | 657 | retry_deleg: |
1da177e4 LT |
658 | newattrs.ia_valid = ATTR_CTIME; |
659 | if (user != (uid_t) -1) { | |
52137abe EB |
660 | if (!uid_valid(uid)) |
661 | return -EINVAL; | |
1da177e4 | 662 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 663 | newattrs.ia_uid = uid; |
1da177e4 LT |
664 | } |
665 | if (group != (gid_t) -1) { | |
52137abe EB |
666 | if (!gid_valid(gid)) |
667 | return -EINVAL; | |
1da177e4 | 668 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 669 | newattrs.ia_gid = gid; |
1da177e4 LT |
670 | } |
671 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
672 | newattrs.ia_valid |= |
673 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
5955102c | 674 | inode_lock(inode); |
d2b31ca6 | 675 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 676 | if (!error) |
2f221d6f CB |
677 | error = notify_change(&init_user_ns, path->dentry, &newattrs, |
678 | &delegated_inode); | |
5955102c | 679 | inode_unlock(inode); |
27ac0ffe BF |
680 | if (delegated_inode) { |
681 | error = break_deleg_wait(&delegated_inode); | |
682 | if (!error) | |
683 | goto retry_deleg; | |
684 | } | |
1da177e4 LT |
685 | return error; |
686 | } | |
687 | ||
55731b3c DB |
688 | int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group, |
689 | int flag) | |
5590ff0d | 690 | { |
2d8f3038 | 691 | struct path path; |
5590ff0d | 692 | int error = -EINVAL; |
65cfc672 | 693 | int lookup_flags; |
5590ff0d | 694 | |
65cfc672 | 695 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
696 | goto out; |
697 | ||
65cfc672 AV |
698 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
699 | if (flag & AT_EMPTY_PATH) | |
700 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 701 | retry: |
65cfc672 | 702 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
703 | if (error) |
704 | goto out; | |
2d8f3038 | 705 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
706 | if (error) |
707 | goto out_release; | |
fe542cf5 | 708 | error = chown_common(&path, user, group); |
2d8f3038 | 709 | mnt_drop_write(path.mnt); |
2af482a7 | 710 | out_release: |
2d8f3038 | 711 | path_put(&path); |
99a5df37 JL |
712 | if (retry_estale(error, lookup_flags)) { |
713 | lookup_flags |= LOOKUP_REVAL; | |
714 | goto retry; | |
715 | } | |
5590ff0d UD |
716 | out: |
717 | return error; | |
718 | } | |
719 | ||
55731b3c DB |
720 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
721 | gid_t, group, int, flag) | |
722 | { | |
723 | return do_fchownat(dfd, filename, user, group, flag); | |
724 | } | |
725 | ||
55e4def0 | 726 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 727 | { |
55731b3c | 728 | return do_fchownat(AT_FDCWD, filename, user, group, 0); |
55e4def0 | 729 | } |
1da177e4 | 730 | |
55e4def0 DH |
731 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
732 | { | |
55731b3c DB |
733 | return do_fchownat(AT_FDCWD, filename, user, group, |
734 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
735 | } |
736 | ||
c04011fe CH |
737 | int vfs_fchown(struct file *file, uid_t user, gid_t group) |
738 | { | |
739 | int error; | |
740 | ||
741 | error = mnt_want_write_file(file); | |
742 | if (error) | |
743 | return error; | |
744 | audit_file(file); | |
745 | error = chown_common(&file->f_path, user, group); | |
746 | mnt_drop_write_file(file); | |
747 | return error; | |
748 | } | |
749 | ||
55731b3c | 750 | int ksys_fchown(unsigned int fd, uid_t user, gid_t group) |
1da177e4 | 751 | { |
2903ff01 | 752 | struct fd f = fdget(fd); |
1da177e4 LT |
753 | int error = -EBADF; |
754 | ||
c04011fe CH |
755 | if (f.file) { |
756 | error = vfs_fchown(f.file, user, group); | |
757 | fdput(f); | |
758 | } | |
1da177e4 LT |
759 | return error; |
760 | } | |
761 | ||
55731b3c DB |
762 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
763 | { | |
764 | return ksys_fchown(fd, user, group); | |
765 | } | |
766 | ||
02e5180d | 767 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 768 | struct inode *inode, |
ae2bb293 | 769 | int (*open)(struct inode *, struct file *)) |
1da177e4 | 770 | { |
1abf0c71 | 771 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
772 | int error; |
773 | ||
b5bcdda3 | 774 | path_get(&f->f_path); |
4bacc9c9 | 775 | f->f_inode = inode; |
1da177e4 | 776 | f->f_mapping = inode->i_mapping; |
5660e13d | 777 | f->f_wb_err = filemap_sample_wb_err(f->f_mapping); |
735e4ae5 | 778 | f->f_sb_err = file_sample_sb_err(f); |
5660e13d | 779 | |
3f4d5a00 | 780 | if (unlikely(f->f_flags & O_PATH)) { |
f5d11409 | 781 | f->f_mode = FMODE_PATH | FMODE_OPENED; |
1abf0c71 | 782 | f->f_op = &empty_fops; |
af04fadc | 783 | return 0; |
1abf0c71 AV |
784 | } |
785 | ||
dd20908a | 786 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 787 | error = get_write_access(inode); |
3f4d5a00 | 788 | if (unlikely(error)) |
1da177e4 | 789 | goto cleanup_file; |
0ccb2863 | 790 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 791 | if (unlikely(error)) { |
0ccb2863 AV |
792 | put_write_access(inode); |
793 | goto cleanup_file; | |
794 | } | |
83f936c7 | 795 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
796 | } |
797 | ||
2be7d348 LT |
798 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
799 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) | |
800 | f->f_mode |= FMODE_ATOMIC_POS; | |
801 | ||
1abf0c71 | 802 | f->f_op = fops_get(inode->i_fop); |
7159d544 | 803 | if (WARN_ON(!f->f_op)) { |
72c2d531 AV |
804 | error = -ENODEV; |
805 | goto cleanup_all; | |
806 | } | |
1abf0c71 | 807 | |
e3f20ae2 | 808 | error = security_file_open(f); |
788e7dd4 YN |
809 | if (error) |
810 | goto cleanup_all; | |
811 | ||
c568d683 | 812 | error = break_lease(locks_inode(f), f->f_flags); |
f3c7691e BF |
813 | if (error) |
814 | goto cleanup_all; | |
815 | ||
ea73ea72 AV |
816 | /* normally all 3 are set; ->open() can clear them if needed */ |
817 | f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE; | |
72c2d531 | 818 | if (!open) |
834f2a4a TM |
819 | open = f->f_op->open; |
820 | if (open) { | |
821 | error = open(inode, f); | |
1da177e4 LT |
822 | if (error) |
823 | goto cleanup_all; | |
824 | } | |
f5d11409 | 825 | f->f_mode |= FMODE_OPENED; |
890275b5 MZ |
826 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
827 | i_readcount_inc(inode); | |
293bc982 | 828 | if ((f->f_mode & FMODE_READ) && |
84363182 | 829 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 830 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 831 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 832 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 833 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 834 | |
c75b1d94 | 835 | f->f_write_hint = WRITE_LIFE_NOT_SET; |
1da177e4 LT |
836 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
837 | ||
838 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
af04fadc | 839 | |
69527c55 AV |
840 | /* NB: we're sure to have correct a_ops only after f_op->open */ |
841 | if (f->f_flags & O_DIRECT) { | |
842 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) | |
843 | return -EINVAL; | |
844 | } | |
09d91cda SL |
845 | |
846 | /* | |
847 | * XXX: Huge page cache doesn't support writing yet. Drop all page | |
848 | * cache for this file before processing writes. | |
849 | */ | |
850 | if ((f->f_mode & FMODE_WRITE) && filemap_nr_thps(inode->i_mapping)) | |
851 | truncate_pagecache(inode, 0); | |
852 | ||
96b7e579 | 853 | return 0; |
1da177e4 LT |
854 | |
855 | cleanup_all: | |
6b4e8085 AV |
856 | if (WARN_ON_ONCE(error > 0)) |
857 | error = -EINVAL; | |
1da177e4 | 858 | fops_put(f->f_op); |
83f936c7 | 859 | if (f->f_mode & FMODE_WRITER) { |
1da177e4 | 860 | put_write_access(inode); |
83f936c7 | 861 | __mnt_drop_write(f->f_path.mnt); |
4a3fd211 | 862 | } |
1da177e4 | 863 | cleanup_file: |
02e5180d AV |
864 | path_put(&f->f_path); |
865 | f->f_path.mnt = NULL; | |
866 | f->f_path.dentry = NULL; | |
dd37978c | 867 | f->f_inode = NULL; |
96b7e579 | 868 | return error; |
1da177e4 LT |
869 | } |
870 | ||
d18e9008 MS |
871 | /** |
872 | * finish_open - finish opening a file | |
0854d450 | 873 | * @file: file pointer |
d18e9008 MS |
874 | * @dentry: pointer to dentry |
875 | * @open: open callback | |
0854d450 | 876 | * @opened: state of open |
d18e9008 MS |
877 | * |
878 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
879 | * | |
880 | * If the open callback is set to NULL, then the standard f_op->open() | |
881 | * filesystem callback is substituted. | |
0854d450 MS |
882 | * |
883 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
884 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
885 | * on it after finish_open(). | |
886 | * | |
0854d450 | 887 | * Returns zero on success or -errno if the open failed. |
d18e9008 | 888 | */ |
30d90494 | 889 | int finish_open(struct file *file, struct dentry *dentry, |
be12af3e | 890 | int (*open)(struct inode *, struct file *)) |
d18e9008 | 891 | { |
aad888f8 | 892 | BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 893 | |
b5bcdda3 | 894 | file->f_path.dentry = dentry; |
aad888f8 | 895 | return do_dentry_open(file, d_backing_inode(dentry), open); |
d18e9008 MS |
896 | } |
897 | EXPORT_SYMBOL(finish_open); | |
898 | ||
899 | /** | |
900 | * finish_no_open - finish ->atomic_open() without opening the file | |
901 | * | |
0854d450 | 902 | * @file: file pointer |
d18e9008 MS |
903 | * @dentry: dentry or NULL (as returned from ->lookup()) |
904 | * | |
905 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
906 | * |
907 | * NB: unlike finish_open() this function does consume the dentry reference and | |
908 | * the caller need not dput() it. | |
909 | * | |
64e1ac4d | 910 | * Returns "0" which must be the return value of ->atomic_open() after having |
0854d450 | 911 | * called this function. |
d18e9008 | 912 | */ |
e45198a6 | 913 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 914 | { |
30d90494 | 915 | file->f_path.dentry = dentry; |
64e1ac4d | 916 | return 0; |
d18e9008 MS |
917 | } |
918 | EXPORT_SYMBOL(finish_no_open); | |
919 | ||
9bf39ab2 MS |
920 | char *file_path(struct file *filp, char *buf, int buflen) |
921 | { | |
922 | return d_path(&filp->f_path, buf, buflen); | |
923 | } | |
924 | EXPORT_SYMBOL(file_path); | |
925 | ||
4bacc9c9 DH |
926 | /** |
927 | * vfs_open - open the file at the given path | |
928 | * @path: path to open | |
929 | * @file: newly allocated file with f_flag initialized | |
930 | * @cred: credentials to use | |
931 | */ | |
ae2bb293 | 932 | int vfs_open(const struct path *path, struct file *file) |
4bacc9c9 | 933 | { |
54d5ca87 | 934 | file->f_path = *path; |
a6518f73 | 935 | return do_dentry_open(file, d_backing_inode(path->dentry), NULL); |
4bacc9c9 DH |
936 | } |
937 | ||
765927b2 | 938 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 939 | const struct cred *cred) |
a1a5b3d9 PS |
940 | { |
941 | int error; | |
942 | struct file *f; | |
943 | ||
e0e81739 DH |
944 | validate_creds(cred); |
945 | ||
c212f9aa | 946 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 947 | BUG_ON(!path->mnt); |
322ee5b3 | 948 | |
ea73ea72 | 949 | f = alloc_empty_file(flags, cred); |
af04fadc | 950 | if (!IS_ERR(f)) { |
ae2bb293 | 951 | error = vfs_open(path, f); |
4d27f326 AV |
952 | if (error) { |
953 | fput(f); | |
af04fadc AV |
954 | f = ERR_PTR(error); |
955 | } | |
2a027e7a AV |
956 | } |
957 | return f; | |
a1a5b3d9 | 958 | } |
1da177e4 LT |
959 | EXPORT_SYMBOL(dentry_open); |
960 | ||
2abc77af AV |
961 | struct file *open_with_fake_path(const struct path *path, int flags, |
962 | struct inode *inode, const struct cred *cred) | |
963 | { | |
d3b1084d | 964 | struct file *f = alloc_empty_file_noaccount(flags, cred); |
2abc77af AV |
965 | if (!IS_ERR(f)) { |
966 | int error; | |
967 | ||
968 | f->f_path = *path; | |
969 | error = do_dentry_open(f, inode, NULL); | |
970 | if (error) { | |
971 | fput(f); | |
972 | f = ERR_PTR(error); | |
973 | } | |
974 | } | |
975 | return f; | |
976 | } | |
977 | EXPORT_SYMBOL(open_with_fake_path); | |
978 | ||
fddb5d43 AS |
979 | #define WILL_CREATE(flags) (flags & (O_CREAT | __O_TMPFILE)) |
980 | #define O_PATH_FLAGS (O_DIRECTORY | O_NOFOLLOW | O_PATH | O_CLOEXEC) | |
981 | ||
35cb6d54 | 982 | inline struct open_how build_open_how(int flags, umode_t mode) |
fddb5d43 AS |
983 | { |
984 | struct open_how how = { | |
985 | .flags = flags & VALID_OPEN_FLAGS, | |
986 | .mode = mode & S_IALLUGO, | |
987 | }; | |
988 | ||
989 | /* O_PATH beats everything else. */ | |
990 | if (how.flags & O_PATH) | |
991 | how.flags &= O_PATH_FLAGS; | |
992 | /* Modes should only be set for create-like flags. */ | |
993 | if (!WILL_CREATE(how.flags)) | |
994 | how.mode = 0; | |
995 | return how; | |
996 | } | |
997 | ||
35cb6d54 | 998 | inline int build_open_flags(const struct open_how *how, struct open_flags *op) |
47c805dc | 999 | { |
fddb5d43 | 1000 | int flags = how->flags; |
47c805dc | 1001 | int lookup_flags = 0; |
62fb4a15 | 1002 | int acc_mode = ACC_MODE(flags); |
47c805dc | 1003 | |
fddb5d43 AS |
1004 | /* Must never be set by userspace */ |
1005 | flags &= ~(FMODE_NONOTIFY | O_CLOEXEC); | |
1006 | ||
629e014b | 1007 | /* |
fddb5d43 AS |
1008 | * Older syscalls implicitly clear all of the invalid flags or argument |
1009 | * values before calling build_open_flags(), but openat2(2) checks all | |
1010 | * of its arguments. | |
629e014b | 1011 | */ |
fddb5d43 AS |
1012 | if (flags & ~VALID_OPEN_FLAGS) |
1013 | return -EINVAL; | |
1014 | if (how->resolve & ~VALID_RESOLVE_FLAGS) | |
1015 | return -EINVAL; | |
629e014b | 1016 | |
398840f8 AS |
1017 | /* Scoping flags are mutually exclusive. */ |
1018 | if ((how->resolve & RESOLVE_BENEATH) && (how->resolve & RESOLVE_IN_ROOT)) | |
1019 | return -EINVAL; | |
1020 | ||
fddb5d43 AS |
1021 | /* Deal with the mode. */ |
1022 | if (WILL_CREATE(flags)) { | |
1023 | if (how->mode & ~S_IALLUGO) | |
1024 | return -EINVAL; | |
1025 | op->mode = how->mode | S_IFREG; | |
1026 | } else { | |
1027 | if (how->mode != 0) | |
1028 | return -EINVAL; | |
e68726ff | 1029 | op->mode = 0; |
fddb5d43 | 1030 | } |
47c805dc AV |
1031 | |
1032 | /* | |
fddb5d43 AS |
1033 | * In order to ensure programs get explicit errors when trying to use |
1034 | * O_TMPFILE on old kernels, O_TMPFILE is implemented such that it | |
1035 | * looks like (O_DIRECTORY|O_RDWR & ~O_CREAT) to old kernels. But we | |
1036 | * have to require userspace to explicitly set it. | |
47c805dc | 1037 | */ |
bb458c64 AV |
1038 | if (flags & __O_TMPFILE) { |
1039 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d | 1040 | return -EINVAL; |
ba57ea64 AV |
1041 | if (!(acc_mode & MAY_WRITE)) |
1042 | return -EINVAL; | |
fddb5d43 AS |
1043 | } |
1044 | if (flags & O_PATH) { | |
1045 | /* O_PATH only permits certain other flags to be set. */ | |
1046 | if (flags & ~O_PATH_FLAGS) | |
1047 | return -EINVAL; | |
1abf0c71 | 1048 | acc_mode = 0; |
1abf0c71 | 1049 | } |
47c805dc | 1050 | |
fddb5d43 AS |
1051 | /* |
1052 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1053 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1054 | * always set instead of having to deal with possibly weird behaviour | |
1055 | * for malicious applications setting only __O_SYNC. | |
1056 | */ | |
1057 | if (flags & __O_SYNC) | |
1058 | flags |= O_DSYNC; | |
1059 | ||
1abf0c71 | 1060 | op->open_flag = flags; |
47c805dc AV |
1061 | |
1062 | /* O_TRUNC implies we need access checks for write permissions */ | |
1063 | if (flags & O_TRUNC) | |
1064 | acc_mode |= MAY_WRITE; | |
1065 | ||
1066 | /* Allow the LSM permission hook to distinguish append | |
1067 | access from general write access. */ | |
1068 | if (flags & O_APPEND) | |
1069 | acc_mode |= MAY_APPEND; | |
1070 | ||
1071 | op->acc_mode = acc_mode; | |
1072 | ||
1abf0c71 AV |
1073 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
1074 | ||
47c805dc AV |
1075 | if (flags & O_CREAT) { |
1076 | op->intent |= LOOKUP_CREATE; | |
31d1726d | 1077 | if (flags & O_EXCL) { |
47c805dc | 1078 | op->intent |= LOOKUP_EXCL; |
31d1726d AV |
1079 | flags |= O_NOFOLLOW; |
1080 | } | |
47c805dc AV |
1081 | } |
1082 | ||
1083 | if (flags & O_DIRECTORY) | |
1084 | lookup_flags |= LOOKUP_DIRECTORY; | |
1085 | if (!(flags & O_NOFOLLOW)) | |
1086 | lookup_flags |= LOOKUP_FOLLOW; | |
fddb5d43 AS |
1087 | |
1088 | if (how->resolve & RESOLVE_NO_XDEV) | |
1089 | lookup_flags |= LOOKUP_NO_XDEV; | |
1090 | if (how->resolve & RESOLVE_NO_MAGICLINKS) | |
1091 | lookup_flags |= LOOKUP_NO_MAGICLINKS; | |
1092 | if (how->resolve & RESOLVE_NO_SYMLINKS) | |
1093 | lookup_flags |= LOOKUP_NO_SYMLINKS; | |
1094 | if (how->resolve & RESOLVE_BENEATH) | |
1095 | lookup_flags |= LOOKUP_BENEATH; | |
1096 | if (how->resolve & RESOLVE_IN_ROOT) | |
1097 | lookup_flags |= LOOKUP_IN_ROOT; | |
1098 | ||
f9652e10 AV |
1099 | op->lookup_flags = lookup_flags; |
1100 | return 0; | |
47c805dc AV |
1101 | } |
1102 | ||
669abf4e JL |
1103 | /** |
1104 | * file_open_name - open file and return file pointer | |
1105 | * | |
1106 | * @name: struct filename containing path to open | |
1107 | * @flags: open flags as per the open(2) second argument | |
1108 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1109 | * | |
1110 | * This is the helper to open a file from kernelspace if you really | |
1111 | * have to. But in generally you should not do this, so please move | |
1112 | * along, nothing to see here.. | |
1113 | */ | |
1114 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
1115 | { | |
1116 | struct open_flags op; | |
fddb5d43 AS |
1117 | struct open_how how = build_open_how(flags, mode); |
1118 | int err = build_open_flags(&how, &op); | |
1119 | if (err) | |
1120 | return ERR_PTR(err); | |
1121 | return do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
1122 | } |
1123 | ||
47c805dc AV |
1124 | /** |
1125 | * filp_open - open file and return file pointer | |
1126 | * | |
1127 | * @filename: path to open | |
1128 | * @flags: open flags as per the open(2) second argument | |
1129 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1130 | * | |
1131 | * This is the helper to open a file from kernelspace if you really | |
1132 | * have to. But in generally you should not do this, so please move | |
1133 | * along, nothing to see here.. | |
1134 | */ | |
a218d0fd | 1135 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 1136 | { |
51689104 PM |
1137 | struct filename *name = getname_kernel(filename); |
1138 | struct file *file = ERR_CAST(name); | |
1139 | ||
1140 | if (!IS_ERR(name)) { | |
1141 | file = file_open_name(name, flags, mode); | |
1142 | putname(name); | |
1143 | } | |
1144 | return file; | |
47c805dc AV |
1145 | } |
1146 | EXPORT_SYMBOL(filp_open); | |
1147 | ||
73d049a4 | 1148 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
378c6520 | 1149 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1150 | { |
1151 | struct open_flags op; | |
fddb5d43 AS |
1152 | struct open_how how = build_open_how(flags, mode); |
1153 | int err = build_open_flags(&how, &op); | |
f9652e10 AV |
1154 | if (err) |
1155 | return ERR_PTR(err); | |
f9652e10 | 1156 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
1157 | } |
1158 | EXPORT_SYMBOL(file_open_root); | |
1159 | ||
fddb5d43 AS |
1160 | static long do_sys_openat2(int dfd, const char __user *filename, |
1161 | struct open_how *how) | |
1da177e4 | 1162 | { |
47c805dc | 1163 | struct open_flags op; |
fddb5d43 | 1164 | int fd = build_open_flags(how, &op); |
f9652e10 AV |
1165 | struct filename *tmp; |
1166 | ||
1167 | if (fd) | |
1168 | return fd; | |
1169 | ||
1170 | tmp = getname(filename); | |
1171 | if (IS_ERR(tmp)) | |
1172 | return PTR_ERR(tmp); | |
1173 | ||
fddb5d43 | 1174 | fd = get_unused_fd_flags(how->flags); |
f9652e10 AV |
1175 | if (fd >= 0) { |
1176 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1177 | if (IS_ERR(f)) { | |
1178 | put_unused_fd(fd); | |
1179 | fd = PTR_ERR(f); | |
1180 | } else { | |
1181 | fsnotify_open(f); | |
1182 | fd_install(fd, f); | |
1da177e4 | 1183 | } |
1da177e4 | 1184 | } |
f9652e10 | 1185 | putname(tmp); |
1da177e4 | 1186 | return fd; |
1da177e4 | 1187 | } |
e922efc3 | 1188 | |
fddb5d43 | 1189 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
e922efc3 | 1190 | { |
fddb5d43 AS |
1191 | struct open_how how = build_open_how(flags, mode); |
1192 | return do_sys_openat2(dfd, filename, &how); | |
1193 | } | |
e922efc3 | 1194 | |
fddb5d43 AS |
1195 | |
1196 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1197 | { | |
166e07c3 CH |
1198 | if (force_o_largefile()) |
1199 | flags |= O_LARGEFILE; | |
1200 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
e922efc3 | 1201 | } |
1da177e4 | 1202 | |
6559eed8 | 1203 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1204 | umode_t, mode) |
5590ff0d UD |
1205 | { |
1206 | if (force_o_largefile()) | |
1207 | flags |= O_LARGEFILE; | |
2cf09666 | 1208 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1209 | } |
5590ff0d | 1210 | |
fddb5d43 AS |
1211 | SYSCALL_DEFINE4(openat2, int, dfd, const char __user *, filename, |
1212 | struct open_how __user *, how, size_t, usize) | |
1213 | { | |
1214 | int err; | |
1215 | struct open_how tmp; | |
1216 | ||
1217 | BUILD_BUG_ON(sizeof(struct open_how) < OPEN_HOW_SIZE_VER0); | |
1218 | BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_LATEST); | |
1219 | ||
1220 | if (unlikely(usize < OPEN_HOW_SIZE_VER0)) | |
1221 | return -EINVAL; | |
1222 | ||
1223 | err = copy_struct_from_user(&tmp, sizeof(tmp), how, usize); | |
1224 | if (err) | |
1225 | return err; | |
1226 | ||
1227 | /* O_LARGEFILE is only allowed for non-O_PATH. */ | |
1228 | if (!(tmp.flags & O_PATH) && force_o_largefile()) | |
1229 | tmp.flags |= O_LARGEFILE; | |
1230 | ||
1231 | return do_sys_openat2(dfd, filename, &tmp); | |
1232 | } | |
1233 | ||
e35d49f6 AV |
1234 | #ifdef CONFIG_COMPAT |
1235 | /* | |
1236 | * Exactly like sys_open(), except that it doesn't set the | |
1237 | * O_LARGEFILE flag. | |
1238 | */ | |
1239 | COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1240 | { | |
1241 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
1242 | } | |
1243 | ||
1244 | /* | |
1245 | * Exactly like sys_openat(), except that it doesn't set the | |
1246 | * O_LARGEFILE flag. | |
1247 | */ | |
1248 | COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode) | |
1249 | { | |
1250 | return do_sys_open(dfd, filename, flags, mode); | |
1251 | } | |
1252 | #endif | |
1253 | ||
1da177e4 LT |
1254 | #ifndef __alpha__ |
1255 | ||
1256 | /* | |
1257 | * For backward compatibility? Maybe this should be moved | |
1258 | * into arch/i386 instead? | |
1259 | */ | |
a218d0fd | 1260 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 | 1261 | { |
166e07c3 | 1262 | int flags = O_CREAT | O_WRONLY | O_TRUNC; |
1da177e4 | 1263 | |
166e07c3 CH |
1264 | if (force_o_largefile()) |
1265 | flags |= O_LARGEFILE; | |
1266 | return do_sys_open(AT_FDCWD, pathname, flags, mode); | |
1267 | } | |
1da177e4 LT |
1268 | #endif |
1269 | ||
1270 | /* | |
1271 | * "id" is the POSIX thread ID. We use the | |
1272 | * files pointer for this.. | |
1273 | */ | |
1274 | int filp_close(struct file *filp, fl_owner_t id) | |
1275 | { | |
45778ca8 | 1276 | int retval = 0; |
1da177e4 LT |
1277 | |
1278 | if (!file_count(filp)) { | |
1279 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1280 | return 0; |
1da177e4 LT |
1281 | } |
1282 | ||
72c2d531 | 1283 | if (filp->f_op->flush) |
75e1fcc0 | 1284 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1285 | |
1abf0c71 AV |
1286 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1287 | dnotify_flush(filp, id); | |
1288 | locks_remove_posix(filp, id); | |
1289 | } | |
1da177e4 LT |
1290 | fput(filp); |
1291 | return retval; | |
1292 | } | |
1293 | ||
1294 | EXPORT_SYMBOL(filp_close); | |
1295 | ||
1296 | /* | |
1297 | * Careful here! We test whether the file pointer is NULL before | |
1298 | * releasing the fd. This ensures that one clone task can't release | |
1299 | * an fd while another clone is opening it. | |
1300 | */ | |
ca013e94 | 1301 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1302 | { |
8760c909 | 1303 | int retval = close_fd(fd); |
ee731f4f EP |
1304 | |
1305 | /* can't restart close syscall because file table entry was cleared */ | |
1306 | if (unlikely(retval == -ERESTARTSYS || | |
1307 | retval == -ERESTARTNOINTR || | |
1308 | retval == -ERESTARTNOHAND || | |
1309 | retval == -ERESTART_RESTARTBLOCK)) | |
1310 | retval = -EINTR; | |
1311 | ||
1312 | return retval; | |
1da177e4 | 1313 | } |
1da177e4 | 1314 | |
278a5fba CB |
1315 | /** |
1316 | * close_range() - Close all file descriptors in a given range. | |
1317 | * | |
1318 | * @fd: starting file descriptor to close | |
1319 | * @max_fd: last file descriptor to close | |
1320 | * @flags: reserved for future extensions | |
1321 | * | |
1322 | * This closes a range of file descriptors. All file descriptors | |
1323 | * from @fd up to and including @max_fd are closed. | |
1324 | * Currently, errors to close a given file descriptor are ignored. | |
1325 | */ | |
1326 | SYSCALL_DEFINE3(close_range, unsigned int, fd, unsigned int, max_fd, | |
1327 | unsigned int, flags) | |
1328 | { | |
60997c3d | 1329 | return __close_range(fd, max_fd, flags); |
278a5fba CB |
1330 | } |
1331 | ||
1da177e4 LT |
1332 | /* |
1333 | * This routine simulates a hangup on the tty, to arrange that users | |
1334 | * are given clean terminals at login time. | |
1335 | */ | |
ca013e94 | 1336 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1337 | { |
1338 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1339 | tty_vhangup_self(); |
1da177e4 LT |
1340 | return 0; |
1341 | } | |
1342 | return -EPERM; | |
1343 | } | |
1344 | ||
1345 | /* | |
1346 | * Called when an inode is about to be open. | |
1347 | * We use this to disallow opening large files on 32bit systems if | |
1348 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1349 | * on this flag in sys_open. | |
1350 | */ | |
1351 | int generic_file_open(struct inode * inode, struct file * filp) | |
1352 | { | |
1353 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1354 | return -EOVERFLOW; |
1da177e4 LT |
1355 | return 0; |
1356 | } | |
1357 | ||
1358 | EXPORT_SYMBOL(generic_file_open); | |
1359 | ||
1360 | /* | |
1361 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1362 | * file descriptors. The function is not supposed to ever fail, the only |
1363 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1364 | * directly into file_operations structure. | |
1da177e4 LT |
1365 | */ |
1366 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1367 | { | |
1368 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1369 | return 0; | |
1370 | } | |
1371 | ||
1372 | EXPORT_SYMBOL(nonseekable_open); | |
10dce8af KS |
1373 | |
1374 | /* | |
1375 | * stream_open is used by subsystems that want stream-like file descriptors. | |
1376 | * Such file descriptors are not seekable and don't have notion of position | |
438ab720 KS |
1377 | * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL). |
1378 | * Contrary to file descriptors of other regular files, .read() and .write() | |
1379 | * can run simultaneously. | |
10dce8af KS |
1380 | * |
1381 | * stream_open never fails and is marked to return int so that it could be | |
1382 | * directly used as file_operations.open . | |
1383 | */ | |
1384 | int stream_open(struct inode *inode, struct file *filp) | |
1385 | { | |
2be7d348 | 1386 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS); |
10dce8af KS |
1387 | filp->f_mode |= FMODE_STREAM; |
1388 | return 0; | |
1389 | } | |
1390 | ||
1391 | EXPORT_SYMBOL(stream_open); |