]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/open.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | #include <linux/string.h> | |
8 | #include <linux/mm.h> | |
1da177e4 | 9 | #include <linux/file.h> |
9f3acc31 | 10 | #include <linux/fdtable.h> |
0eeca283 | 11 | #include <linux/fsnotify.h> |
1da177e4 | 12 | #include <linux/module.h> |
1da177e4 LT |
13 | #include <linux/tty.h> |
14 | #include <linux/namei.h> | |
15 | #include <linux/backing-dev.h> | |
16f7e0fe | 16 | #include <linux/capability.h> |
086f7316 | 17 | #include <linux/securebits.h> |
1da177e4 LT |
18 | #include <linux/security.h> |
19 | #include <linux/mount.h> | |
5590ff0d | 20 | #include <linux/fcntl.h> |
5a0e3ad6 | 21 | #include <linux/slab.h> |
7c0f6ba6 | 22 | #include <linux/uaccess.h> |
1da177e4 | 23 | #include <linux/fs.h> |
ef3daeda | 24 | #include <linux/personality.h> |
1da177e4 LT |
25 | #include <linux/pagemap.h> |
26 | #include <linux/syscalls.h> | |
ab2af1f5 | 27 | #include <linux/rcupdate.h> |
73241ccc | 28 | #include <linux/audit.h> |
97ac7350 | 29 | #include <linux/falloc.h> |
5ad4e53b | 30 | #include <linux/fs_struct.h> |
b65a9cfc | 31 | #include <linux/ima.h> |
2dfc1cae | 32 | #include <linux/dnotify.h> |
3f6d078d | 33 | #include <linux/compat.h> |
1da177e4 | 34 | |
e81e3f4d EP |
35 | #include "internal.h" |
36 | ||
308a1eef SJR |
37 | #define CREATE_TRACE_POINTS |
38 | #include <trace/events/fs.h> | |
39 | ||
4a30131e N |
40 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, |
41 | struct file *filp) | |
1da177e4 | 42 | { |
939a9421 | 43 | int ret; |
1da177e4 LT |
44 | struct iattr newattrs; |
45 | ||
46 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
47 | if (length < 0) | |
48 | return -EINVAL; | |
49 | ||
50 | newattrs.ia_size = length; | |
4a30131e | 51 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
52 | if (filp) { |
53 | newattrs.ia_file = filp; | |
54 | newattrs.ia_valid |= ATTR_FILE; | |
55 | } | |
1da177e4 | 56 | |
45f147a1 JK |
57 | /* Remove suid, sgid, and file capabilities on truncate too */ |
58 | ret = dentry_needs_remove_privs(dentry); | |
59 | if (ret < 0) | |
60 | return ret; | |
939a9421 AW |
61 | if (ret) |
62 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 63 | |
5955102c | 64 | inode_lock(dentry->d_inode); |
27ac0ffe BF |
65 | /* Note any delegations or leases have already been broken: */ |
66 | ret = notify_change(dentry, &newattrs, NULL); | |
5955102c | 67 | inode_unlock(dentry->d_inode); |
939a9421 | 68 | return ret; |
1da177e4 LT |
69 | } |
70 | ||
7df818b2 | 71 | long vfs_truncate(const struct path *path, loff_t length) |
1da177e4 | 72 | { |
2d8f3038 | 73 | struct inode *inode; |
4d0c5ba2 | 74 | struct dentry *upperdentry; |
a02de960 | 75 | long error; |
1da177e4 | 76 | |
a02de960 | 77 | inode = path->dentry->d_inode; |
1da177e4 LT |
78 | |
79 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
1da177e4 | 80 | if (S_ISDIR(inode->i_mode)) |
a02de960 | 81 | return -EISDIR; |
1da177e4 | 82 | if (!S_ISREG(inode->i_mode)) |
a02de960 | 83 | return -EINVAL; |
1da177e4 | 84 | |
a02de960 | 85 | error = mnt_want_write(path->mnt); |
1da177e4 | 86 | if (error) |
a02de960 | 87 | goto out; |
1da177e4 | 88 | |
256984a8 | 89 | error = inode_permission(inode, MAY_WRITE); |
9ac9b847 DH |
90 | if (error) |
91 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
92 | |
93 | error = -EPERM; | |
c82e42da | 94 | if (IS_APPEND(inode)) |
9ac9b847 | 95 | goto mnt_drop_write_and_out; |
1da177e4 | 96 | |
4d0c5ba2 MS |
97 | /* |
98 | * If this is an overlayfs then do as if opening the file so we get | |
99 | * write access on the upper inode, not on the overlay inode. For | |
100 | * non-overlay filesystems d_real() is an identity function. | |
101 | */ | |
102 | upperdentry = d_real(path->dentry, NULL, O_WRONLY); | |
103 | error = PTR_ERR(upperdentry); | |
104 | if (IS_ERR(upperdentry)) | |
105 | goto mnt_drop_write_and_out; | |
106 | ||
107 | error = get_write_access(upperdentry->d_inode); | |
1da177e4 | 108 | if (error) |
9ac9b847 | 109 | goto mnt_drop_write_and_out; |
1da177e4 | 110 | |
9700382c | 111 | /* |
112 | * Make sure that there are no leases. get_write_access() protects | |
113 | * against the truncate racing with a lease-granting setlease(). | |
114 | */ | |
8737c930 | 115 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 116 | if (error) |
9700382c | 117 | goto put_write_and_out; |
1da177e4 LT |
118 | |
119 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 | 120 | if (!error) |
a02de960 | 121 | error = security_path_truncate(path); |
907f4554 | 122 | if (!error) |
a02de960 | 123 | error = do_truncate(path->dentry, length, 0, NULL); |
1da177e4 | 124 | |
9700382c | 125 | put_write_and_out: |
4d0c5ba2 | 126 | put_write_access(upperdentry->d_inode); |
9ac9b847 | 127 | mnt_drop_write_and_out: |
a02de960 | 128 | mnt_drop_write(path->mnt); |
1da177e4 LT |
129 | out: |
130 | return error; | |
131 | } | |
a02de960 DH |
132 | EXPORT_SYMBOL_GPL(vfs_truncate); |
133 | ||
134 | static long do_sys_truncate(const char __user *pathname, loff_t length) | |
135 | { | |
48f7530d | 136 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
137 | struct path path; |
138 | int error; | |
139 | ||
140 | if (length < 0) /* sorry, but loff_t says... */ | |
141 | return -EINVAL; | |
142 | ||
48f7530d JL |
143 | retry: |
144 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
145 | if (!error) { |
146 | error = vfs_truncate(&path, length); | |
147 | path_put(&path); | |
148 | } | |
48f7530d JL |
149 | if (retry_estale(error, lookup_flags)) { |
150 | lookup_flags |= LOOKUP_REVAL; | |
151 | goto retry; | |
152 | } | |
a02de960 DH |
153 | return error; |
154 | } | |
1da177e4 | 155 | |
4fd8da8d | 156 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 157 | { |
4fd8da8d | 158 | return do_sys_truncate(path, length); |
1da177e4 LT |
159 | } |
160 | ||
3f6d078d AV |
161 | #ifdef CONFIG_COMPAT |
162 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
163 | { | |
164 | return do_sys_truncate(path, length); | |
165 | } | |
166 | #endif | |
167 | ||
b01ec0ef | 168 | static long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 | 169 | { |
bf2965d5 | 170 | struct inode *inode; |
1da177e4 | 171 | struct dentry *dentry; |
2903ff01 | 172 | struct fd f; |
1da177e4 LT |
173 | int error; |
174 | ||
175 | error = -EINVAL; | |
176 | if (length < 0) | |
177 | goto out; | |
178 | error = -EBADF; | |
2903ff01 AV |
179 | f = fdget(fd); |
180 | if (!f.file) | |
1da177e4 LT |
181 | goto out; |
182 | ||
183 | /* explicitly opened as large or we are on 64-bit box */ | |
2903ff01 | 184 | if (f.file->f_flags & O_LARGEFILE) |
1da177e4 LT |
185 | small = 0; |
186 | ||
2903ff01 | 187 | dentry = f.file->f_path.dentry; |
1da177e4 LT |
188 | inode = dentry->d_inode; |
189 | error = -EINVAL; | |
2903ff01 | 190 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
191 | goto out_putf; |
192 | ||
193 | error = -EINVAL; | |
194 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
195 | if (small && length > MAX_NON_LFS) | |
196 | goto out_putf; | |
197 | ||
198 | error = -EPERM; | |
78757af6 AG |
199 | /* Check IS_APPEND on real upper inode */ |
200 | if (IS_APPEND(file_inode(f.file))) | |
1da177e4 LT |
201 | goto out_putf; |
202 | ||
14da9200 | 203 | sb_start_write(inode->i_sb); |
2903ff01 | 204 | error = locks_verify_truncate(inode, f.file, length); |
be6d3e56 | 205 | if (!error) |
2903ff01 | 206 | error = security_path_truncate(&f.file->f_path); |
1da177e4 | 207 | if (!error) |
2903ff01 | 208 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file); |
14da9200 | 209 | sb_end_write(inode->i_sb); |
1da177e4 | 210 | out_putf: |
2903ff01 | 211 | fdput(f); |
1da177e4 LT |
212 | out: |
213 | return error; | |
214 | } | |
215 | ||
bdc480e3 | 216 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 217 | { |
2cf09666 | 218 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
219 | } |
220 | ||
3f6d078d AV |
221 | #ifdef CONFIG_COMPAT |
222 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
223 | { | |
224 | return do_sys_ftruncate(fd, length, 1); | |
225 | } | |
226 | #endif | |
227 | ||
1da177e4 LT |
228 | /* LFS versions of truncate are only needed on 32 bit machines */ |
229 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 230 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
231 | { |
232 | return do_sys_truncate(path, length); | |
233 | } | |
234 | ||
4a0fd5bf | 235 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 236 | { |
2cf09666 | 237 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 238 | } |
6673e0c3 | 239 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 240 | |
3e63cbb1 | 241 | |
72c72bdf | 242 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 243 | { |
496ad9aa | 244 | struct inode *inode = file_inode(file); |
3e63cbb1 | 245 | long ret; |
97ac7350 AA |
246 | |
247 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 248 | return -EINVAL; |
97ac7350 AA |
249 | |
250 | /* Return error if mode is not supported */ | |
dd46c787 | 251 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) |
409332b6 LC |
252 | return -EOPNOTSUPP; |
253 | ||
254 | /* Punch hole and zero range are mutually exclusive */ | |
255 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
256 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
257 | return -EOPNOTSUPP; |
258 | ||
259 | /* Punch hole must have keep size set */ | |
260 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
261 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 262 | return -EOPNOTSUPP; |
97ac7350 | 263 | |
00f5e619 NJ |
264 | /* Collapse range should only be used exclusively. */ |
265 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
266 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
267 | return -EINVAL; | |
268 | ||
dd46c787 NJ |
269 | /* Insert range should only be used exclusively. */ |
270 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
271 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
272 | return -EINVAL; | |
273 | ||
71be6b49 DW |
274 | /* Unshare range should only be used with allocate mode. */ |
275 | if ((mode & FALLOC_FL_UNSHARE_RANGE) && | |
276 | (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE))) | |
277 | return -EINVAL; | |
278 | ||
97ac7350 | 279 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 280 | return -EBADF; |
1ca551c6 | 281 | |
00f5e619 | 282 | /* |
8fc61d92 | 283 | * We can only allow pure fallocate on append only files |
00f5e619 | 284 | */ |
8fc61d92 | 285 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
286 | return -EPERM; |
287 | ||
288 | if (IS_IMMUTABLE(inode)) | |
289 | return -EPERM; | |
290 | ||
0790b31b | 291 | /* |
6d2b6170 | 292 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
293 | */ |
294 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 295 | return -ETXTBSY; |
0790b31b | 296 | |
97ac7350 AA |
297 | /* |
298 | * Revalidate the write permissions, in case security policy has | |
299 | * changed since the files were opened. | |
300 | */ | |
301 | ret = security_file_permission(file, MAY_WRITE); | |
302 | if (ret) | |
3e63cbb1 | 303 | return ret; |
97ac7350 | 304 | |
97ac7350 | 305 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 306 | return -ESPIPE; |
97ac7350 | 307 | |
9e79b132 AG |
308 | if (S_ISDIR(inode->i_mode)) |
309 | return -EISDIR; | |
310 | ||
311 | if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode)) | |
3e63cbb1 | 312 | return -ENODEV; |
97ac7350 | 313 | |
97ac7350 AA |
314 | /* Check for wrap through zero too */ |
315 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 316 | return -EFBIG; |
97ac7350 | 317 | |
2fe17c10 | 318 | if (!file->f_op->fallocate) |
3e63cbb1 | 319 | return -EOPNOTSUPP; |
97ac7350 | 320 | |
bfe219d3 | 321 | file_start_write(file); |
14da9200 | 322 | ret = file->f_op->fallocate(file, mode, offset, len); |
820c12d5 HS |
323 | |
324 | /* | |
325 | * Create inotify and fanotify events. | |
326 | * | |
327 | * To keep the logic simple always create events if fallocate succeeds. | |
328 | * This implies that events are even created if the file size remains | |
329 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
330 | */ | |
331 | if (ret == 0) | |
332 | fsnotify_modify(file); | |
333 | ||
bfe219d3 | 334 | file_end_write(file); |
14da9200 | 335 | return ret; |
3e63cbb1 | 336 | } |
72c72bdf | 337 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 338 | |
4a0fd5bf | 339 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
3e63cbb1 | 340 | { |
2903ff01 | 341 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
342 | int error = -EBADF; |
343 | ||
2903ff01 | 344 | if (f.file) { |
72c72bdf | 345 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 346 | fdput(f); |
3e63cbb1 | 347 | } |
3e63cbb1 | 348 | return error; |
97ac7350 | 349 | } |
3e63cbb1 | 350 | |
1da177e4 LT |
351 | /* |
352 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
353 | * We do this by temporarily clearing all FS-related capabilities and | |
354 | * switching the fsuid/fsgid around to the real ones. | |
355 | */ | |
6559eed8 | 356 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
1da177e4 | 357 | { |
d84f4f99 DH |
358 | const struct cred *old_cred; |
359 | struct cred *override_cred; | |
2d8f3038 | 360 | struct path path; |
256984a8 | 361 | struct inode *inode; |
1da177e4 | 362 | int res; |
87fa5595 | 363 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
1da177e4 LT |
364 | |
365 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
366 | return -EINVAL; | |
367 | ||
d84f4f99 DH |
368 | override_cred = prepare_creds(); |
369 | if (!override_cred) | |
370 | return -ENOMEM; | |
1da177e4 | 371 | |
d84f4f99 DH |
372 | override_cred->fsuid = override_cred->uid; |
373 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 374 | |
086f7316 | 375 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 376 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
377 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
378 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 379 | cap_clear(override_cred->cap_effective); |
086f7316 | 380 | else |
d84f4f99 DH |
381 | override_cred->cap_effective = |
382 | override_cred->cap_permitted; | |
086f7316 | 383 | } |
1da177e4 | 384 | |
d84f4f99 | 385 | old_cred = override_creds(override_cred); |
87fa5595 JL |
386 | retry: |
387 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
388 | if (res) |
389 | goto out; | |
390 | ||
63afdfc7 | 391 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
392 | |
393 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
394 | /* |
395 | * MAY_EXEC on regular files is denied if the fs is mounted | |
396 | * with the "noexec" flag. | |
397 | */ | |
398 | res = -EACCES; | |
90f8572b | 399 | if (path_noexec(&path)) |
30524472 AV |
400 | goto out_path_release; |
401 | } | |
402 | ||
256984a8 | 403 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 404 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 405 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 406 | goto out_path_release; |
2f676cbc DH |
407 | /* |
408 | * This is a rare case where using __mnt_is_readonly() | |
409 | * is OK without a mnt_want/drop_write() pair. Since | |
410 | * no actual write to the fs is performed here, we do | |
411 | * not need to telegraph to that to anyone. | |
412 | * | |
413 | * By doing this, we accept that this access is | |
414 | * inherently racy and know that the fs may change | |
415 | * state before we even see this result. | |
416 | */ | |
2d8f3038 | 417 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 418 | res = -EROFS; |
1da177e4 | 419 | |
6902d925 | 420 | out_path_release: |
2d8f3038 | 421 | path_put(&path); |
87fa5595 JL |
422 | if (retry_estale(res, lookup_flags)) { |
423 | lookup_flags |= LOOKUP_REVAL; | |
424 | goto retry; | |
425 | } | |
6902d925 | 426 | out: |
d84f4f99 DH |
427 | revert_creds(old_cred); |
428 | put_cred(override_cred); | |
1da177e4 LT |
429 | return res; |
430 | } | |
431 | ||
ca013e94 | 432 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d UD |
433 | { |
434 | return sys_faccessat(AT_FDCWD, filename, mode); | |
435 | } | |
436 | ||
3cdad428 | 437 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 438 | { |
2d8f3038 | 439 | struct path path; |
1da177e4 | 440 | int error; |
0291c0a5 JL |
441 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
442 | retry: | |
443 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
444 | if (error) |
445 | goto out; | |
446 | ||
9cfcac81 | 447 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
448 | if (error) |
449 | goto dput_and_out; | |
450 | ||
2d8f3038 | 451 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
452 | |
453 | dput_and_out: | |
2d8f3038 | 454 | path_put(&path); |
0291c0a5 JL |
455 | if (retry_estale(error, lookup_flags)) { |
456 | lookup_flags |= LOOKUP_REVAL; | |
457 | goto retry; | |
458 | } | |
1da177e4 LT |
459 | out: |
460 | return error; | |
461 | } | |
462 | ||
3cdad428 | 463 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 464 | { |
2903ff01 | 465 | struct fd f = fdget_raw(fd); |
159b0956 | 466 | int error; |
1da177e4 LT |
467 | |
468 | error = -EBADF; | |
2903ff01 | 469 | if (!f.file) |
1da177e4 LT |
470 | goto out; |
471 | ||
1da177e4 | 472 | error = -ENOTDIR; |
159b0956 | 473 | if (!d_can_lookup(f.file->f_path.dentry)) |
1da177e4 LT |
474 | goto out_putf; |
475 | ||
159b0956 | 476 | error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR); |
1da177e4 | 477 | if (!error) |
2903ff01 | 478 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 479 | out_putf: |
2903ff01 | 480 | fdput(f); |
1da177e4 LT |
481 | out: |
482 | return error; | |
483 | } | |
484 | ||
3480b257 | 485 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 486 | { |
2d8f3038 | 487 | struct path path; |
1da177e4 | 488 | int error; |
2771261e 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 | ||
499 | error = -EPERM; | |
c7b96acf | 500 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 501 | goto dput_and_out; |
8b8efb44 TH |
502 | error = security_path_chroot(&path); |
503 | if (error) | |
504 | goto dput_and_out; | |
1da177e4 | 505 | |
2d8f3038 | 506 | set_fs_root(current->fs, &path); |
1da177e4 LT |
507 | error = 0; |
508 | dput_and_out: | |
2d8f3038 | 509 | path_put(&path); |
2771261e JL |
510 | if (retry_estale(error, lookup_flags)) { |
511 | lookup_flags |= LOOKUP_REVAL; | |
512 | goto retry; | |
513 | } | |
1da177e4 LT |
514 | out: |
515 | return error; | |
516 | } | |
517 | ||
be01f9f2 | 518 | static int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 519 | { |
e57712eb | 520 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 521 | struct inode *delegated_inode = NULL; |
1da177e4 | 522 | struct iattr newattrs; |
e57712eb | 523 | int error; |
1da177e4 | 524 | |
e57712eb AV |
525 | error = mnt_want_write(path->mnt); |
526 | if (error) | |
527 | return error; | |
27ac0ffe | 528 | retry_deleg: |
5955102c | 529 | inode_lock(inode); |
cdcf116d | 530 | error = security_path_chmod(path, mode); |
e57712eb | 531 | if (error) |
fe542cf5 | 532 | goto out_unlock; |
1da177e4 LT |
533 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
534 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 535 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 536 | out_unlock: |
5955102c | 537 | inode_unlock(inode); |
27ac0ffe BF |
538 | if (delegated_inode) { |
539 | error = break_deleg_wait(&delegated_inode); | |
540 | if (!error) | |
541 | goto retry_deleg; | |
542 | } | |
e57712eb AV |
543 | mnt_drop_write(path->mnt); |
544 | return error; | |
545 | } | |
546 | ||
49f0a076 | 547 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 548 | { |
173c8401 | 549 | struct fd f = fdget(fd); |
e57712eb AV |
550 | int err = -EBADF; |
551 | ||
173c8401 | 552 | if (f.file) { |
9f45f5bf | 553 | audit_file(f.file); |
173c8401 AV |
554 | err = chmod_common(&f.file->f_path, mode); |
555 | fdput(f); | |
e57712eb | 556 | } |
1da177e4 LT |
557 | return err; |
558 | } | |
559 | ||
49f0a076 | 560 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode) |
1da177e4 | 561 | { |
2d8f3038 | 562 | struct path path; |
1da177e4 | 563 | int error; |
14ff690c JL |
564 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
565 | retry: | |
566 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
567 | if (!error) { |
568 | error = chmod_common(&path, mode); | |
569 | path_put(&path); | |
14ff690c JL |
570 | if (retry_estale(error, lookup_flags)) { |
571 | lookup_flags |= LOOKUP_REVAL; | |
572 | goto retry; | |
573 | } | |
e57712eb | 574 | } |
1da177e4 LT |
575 | return error; |
576 | } | |
577 | ||
49f0a076 | 578 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d UD |
579 | { |
580 | return sys_fchmodat(AT_FDCWD, filename, mode); | |
581 | } | |
582 | ||
7fd25dac | 583 | static int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 584 | { |
fe542cf5 | 585 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 586 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
587 | int error; |
588 | struct iattr newattrs; | |
52137abe EB |
589 | kuid_t uid; |
590 | kgid_t gid; | |
591 | ||
592 | uid = make_kuid(current_user_ns(), user); | |
593 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 594 | |
c1b8940b | 595 | retry_deleg: |
1da177e4 LT |
596 | newattrs.ia_valid = ATTR_CTIME; |
597 | if (user != (uid_t) -1) { | |
52137abe EB |
598 | if (!uid_valid(uid)) |
599 | return -EINVAL; | |
1da177e4 | 600 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 601 | newattrs.ia_uid = uid; |
1da177e4 LT |
602 | } |
603 | if (group != (gid_t) -1) { | |
52137abe EB |
604 | if (!gid_valid(gid)) |
605 | return -EINVAL; | |
1da177e4 | 606 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 607 | newattrs.ia_gid = gid; |
1da177e4 LT |
608 | } |
609 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
610 | newattrs.ia_valid |= |
611 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
5955102c | 612 | inode_lock(inode); |
d2b31ca6 | 613 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 614 | if (!error) |
27ac0ffe | 615 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
5955102c | 616 | inode_unlock(inode); |
27ac0ffe BF |
617 | if (delegated_inode) { |
618 | error = break_deleg_wait(&delegated_inode); | |
619 | if (!error) | |
620 | goto retry_deleg; | |
621 | } | |
1da177e4 LT |
622 | return error; |
623 | } | |
624 | ||
6559eed8 HC |
625 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
626 | gid_t, group, int, flag) | |
5590ff0d | 627 | { |
2d8f3038 | 628 | struct path path; |
5590ff0d | 629 | int error = -EINVAL; |
65cfc672 | 630 | int lookup_flags; |
5590ff0d | 631 | |
65cfc672 | 632 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
633 | goto out; |
634 | ||
65cfc672 AV |
635 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
636 | if (flag & AT_EMPTY_PATH) | |
637 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 638 | retry: |
65cfc672 | 639 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
640 | if (error) |
641 | goto out; | |
2d8f3038 | 642 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
643 | if (error) |
644 | goto out_release; | |
fe542cf5 | 645 | error = chown_common(&path, user, group); |
2d8f3038 | 646 | mnt_drop_write(path.mnt); |
2af482a7 | 647 | out_release: |
2d8f3038 | 648 | path_put(&path); |
99a5df37 JL |
649 | if (retry_estale(error, lookup_flags)) { |
650 | lookup_flags |= LOOKUP_REVAL; | |
651 | goto retry; | |
652 | } | |
5590ff0d UD |
653 | out: |
654 | return error; | |
655 | } | |
656 | ||
55e4def0 | 657 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 658 | { |
55e4def0 DH |
659 | return sys_fchownat(AT_FDCWD, filename, user, group, 0); |
660 | } | |
1da177e4 | 661 | |
55e4def0 DH |
662 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
663 | { | |
664 | return sys_fchownat(AT_FDCWD, filename, user, group, | |
665 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
666 | } |
667 | ||
ca013e94 | 668 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
1da177e4 | 669 | { |
2903ff01 | 670 | struct fd f = fdget(fd); |
1da177e4 LT |
671 | int error = -EBADF; |
672 | ||
2903ff01 | 673 | if (!f.file) |
6902d925 DH |
674 | goto out; |
675 | ||
2903ff01 | 676 | error = mnt_want_write_file(f.file); |
2af482a7 DH |
677 | if (error) |
678 | goto out_fput; | |
9f45f5bf | 679 | audit_file(f.file); |
2903ff01 AV |
680 | error = chown_common(&f.file->f_path, user, group); |
681 | mnt_drop_write_file(f.file); | |
2af482a7 | 682 | out_fput: |
2903ff01 | 683 | fdput(f); |
6902d925 | 684 | out: |
1da177e4 LT |
685 | return error; |
686 | } | |
687 | ||
90ad1a8e MS |
688 | int open_check_o_direct(struct file *f) |
689 | { | |
690 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
691 | if (f->f_flags & O_DIRECT) { | |
e748dcd0 | 692 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) |
90ad1a8e | 693 | return -EINVAL; |
90ad1a8e MS |
694 | } |
695 | return 0; | |
696 | } | |
697 | ||
02e5180d | 698 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 699 | struct inode *inode, |
96b7e579 AV |
700 | int (*open)(struct inode *, struct file *), |
701 | const struct cred *cred) | |
1da177e4 | 702 | { |
1abf0c71 | 703 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
704 | int error; |
705 | ||
5300990c | 706 | f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK | |
a1a5b3d9 | 707 | FMODE_PREAD | FMODE_PWRITE; |
1abf0c71 | 708 | |
b5bcdda3 | 709 | path_get(&f->f_path); |
4bacc9c9 | 710 | f->f_inode = inode; |
1da177e4 | 711 | f->f_mapping = inode->i_mapping; |
1da177e4 | 712 | |
5660e13d JL |
713 | /* Ensure that we skip any errors that predate opening of the file */ |
714 | f->f_wb_err = filemap_sample_wb_err(f->f_mapping); | |
715 | ||
3f4d5a00 AV |
716 | if (unlikely(f->f_flags & O_PATH)) { |
717 | f->f_mode = FMODE_PATH; | |
1abf0c71 | 718 | f->f_op = &empty_fops; |
96b7e579 | 719 | return 0; |
1abf0c71 AV |
720 | } |
721 | ||
dd20908a | 722 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 723 | error = get_write_access(inode); |
3f4d5a00 | 724 | if (unlikely(error)) |
1da177e4 | 725 | goto cleanup_file; |
0ccb2863 | 726 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 727 | if (unlikely(error)) { |
0ccb2863 AV |
728 | put_write_access(inode); |
729 | goto cleanup_file; | |
730 | } | |
83f936c7 | 731 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
732 | } |
733 | ||
9c225f26 | 734 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
63b6df14 | 735 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) |
9c225f26 LT |
736 | f->f_mode |= FMODE_ATOMIC_POS; |
737 | ||
1abf0c71 | 738 | f->f_op = fops_get(inode->i_fop); |
72c2d531 AV |
739 | if (unlikely(WARN_ON(!f->f_op))) { |
740 | error = -ENODEV; | |
741 | goto cleanup_all; | |
742 | } | |
1abf0c71 | 743 | |
83d49856 | 744 | error = security_file_open(f, cred); |
788e7dd4 YN |
745 | if (error) |
746 | goto cleanup_all; | |
747 | ||
c568d683 | 748 | error = break_lease(locks_inode(f), f->f_flags); |
f3c7691e BF |
749 | if (error) |
750 | goto cleanup_all; | |
751 | ||
72c2d531 | 752 | if (!open) |
834f2a4a TM |
753 | open = f->f_op->open; |
754 | if (open) { | |
755 | error = open(inode, f); | |
1da177e4 LT |
756 | if (error) |
757 | goto cleanup_all; | |
758 | } | |
890275b5 MZ |
759 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
760 | i_readcount_inc(inode); | |
293bc982 | 761 | if ((f->f_mode & FMODE_READ) && |
84363182 | 762 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 763 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 764 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 765 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 766 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 767 | |
c75b1d94 | 768 | f->f_write_hint = WRITE_LIFE_NOT_SET; |
1da177e4 LT |
769 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
770 | ||
771 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
772 | ||
96b7e579 | 773 | return 0; |
1da177e4 LT |
774 | |
775 | cleanup_all: | |
776 | fops_put(f->f_op); | |
83f936c7 | 777 | if (f->f_mode & FMODE_WRITER) { |
1da177e4 | 778 | put_write_access(inode); |
83f936c7 | 779 | __mnt_drop_write(f->f_path.mnt); |
4a3fd211 | 780 | } |
1da177e4 | 781 | cleanup_file: |
02e5180d AV |
782 | path_put(&f->f_path); |
783 | f->f_path.mnt = NULL; | |
784 | f->f_path.dentry = NULL; | |
dd37978c | 785 | f->f_inode = NULL; |
96b7e579 | 786 | return error; |
1da177e4 LT |
787 | } |
788 | ||
d18e9008 MS |
789 | /** |
790 | * finish_open - finish opening a file | |
0854d450 | 791 | * @file: file pointer |
d18e9008 MS |
792 | * @dentry: pointer to dentry |
793 | * @open: open callback | |
0854d450 | 794 | * @opened: state of open |
d18e9008 MS |
795 | * |
796 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
797 | * | |
798 | * If the open callback is set to NULL, then the standard f_op->open() | |
799 | * filesystem callback is substituted. | |
0854d450 MS |
800 | * |
801 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
802 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
803 | * on it after finish_open(). | |
804 | * | |
805 | * On successful return @file is a fully instantiated open file. After this, if | |
806 | * an error occurs in ->atomic_open(), it needs to clean up with fput(). | |
807 | * | |
808 | * Returns zero on success or -errno if the open failed. | |
d18e9008 | 809 | */ |
30d90494 AV |
810 | int finish_open(struct file *file, struct dentry *dentry, |
811 | int (*open)(struct inode *, struct file *), | |
812 | int *opened) | |
d18e9008 | 813 | { |
96b7e579 | 814 | int error; |
3d8a00d2 | 815 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 816 | |
b5bcdda3 | 817 | file->f_path.dentry = dentry; |
4bacc9c9 DH |
818 | error = do_dentry_open(file, d_backing_inode(dentry), open, |
819 | current_cred()); | |
96b7e579 | 820 | if (!error) |
47237687 | 821 | *opened |= FILE_OPENED; |
d18e9008 | 822 | |
96b7e579 | 823 | return error; |
d18e9008 MS |
824 | } |
825 | EXPORT_SYMBOL(finish_open); | |
826 | ||
827 | /** | |
828 | * finish_no_open - finish ->atomic_open() without opening the file | |
829 | * | |
0854d450 | 830 | * @file: file pointer |
d18e9008 MS |
831 | * @dentry: dentry or NULL (as returned from ->lookup()) |
832 | * | |
833 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
834 | * |
835 | * NB: unlike finish_open() this function does consume the dentry reference and | |
836 | * the caller need not dput() it. | |
837 | * | |
838 | * Returns "1" which must be the return value of ->atomic_open() after having | |
839 | * called this function. | |
d18e9008 | 840 | */ |
e45198a6 | 841 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 842 | { |
30d90494 | 843 | file->f_path.dentry = dentry; |
e45198a6 | 844 | return 1; |
d18e9008 MS |
845 | } |
846 | EXPORT_SYMBOL(finish_no_open); | |
847 | ||
9bf39ab2 MS |
848 | char *file_path(struct file *filp, char *buf, int buflen) |
849 | { | |
850 | return d_path(&filp->f_path, buf, buflen); | |
851 | } | |
852 | EXPORT_SYMBOL(file_path); | |
853 | ||
4bacc9c9 DH |
854 | /** |
855 | * vfs_open - open the file at the given path | |
856 | * @path: path to open | |
857 | * @file: newly allocated file with f_flag initialized | |
858 | * @cred: credentials to use | |
859 | */ | |
860 | int vfs_open(const struct path *path, struct file *file, | |
861 | const struct cred *cred) | |
862 | { | |
2d902671 | 863 | struct dentry *dentry = d_real(path->dentry, NULL, file->f_flags); |
4bacc9c9 | 864 | |
2d902671 MS |
865 | if (IS_ERR(dentry)) |
866 | return PTR_ERR(dentry); | |
4bacc9c9 | 867 | |
54d5ca87 | 868 | file->f_path = *path; |
2d902671 | 869 | return do_dentry_open(file, d_backing_inode(dentry), NULL, cred); |
4bacc9c9 DH |
870 | } |
871 | ||
765927b2 | 872 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 873 | const struct cred *cred) |
a1a5b3d9 PS |
874 | { |
875 | int error; | |
876 | struct file *f; | |
877 | ||
e0e81739 DH |
878 | validate_creds(cred); |
879 | ||
c212f9aa | 880 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 881 | BUG_ON(!path->mnt); |
322ee5b3 | 882 | |
a1a5b3d9 | 883 | f = get_empty_filp(); |
1afc99be AV |
884 | if (!IS_ERR(f)) { |
885 | f->f_flags = flags; | |
4aa7c634 | 886 | error = vfs_open(path, f, cred); |
1afc99be AV |
887 | if (!error) { |
888 | /* from now on we need fput() to dispose of f */ | |
889 | error = open_check_o_direct(f); | |
890 | if (error) { | |
891 | fput(f); | |
892 | f = ERR_PTR(error); | |
893 | } | |
894 | } else { | |
895 | put_filp(f); | |
2a027e7a AV |
896 | f = ERR_PTR(error); |
897 | } | |
2a027e7a AV |
898 | } |
899 | return f; | |
a1a5b3d9 | 900 | } |
1da177e4 LT |
901 | EXPORT_SYMBOL(dentry_open); |
902 | ||
a218d0fd | 903 | static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op) |
47c805dc AV |
904 | { |
905 | int lookup_flags = 0; | |
62fb4a15 | 906 | int acc_mode = ACC_MODE(flags); |
47c805dc | 907 | |
629e014b CH |
908 | /* |
909 | * Clear out all open flags we don't know about so that we don't report | |
910 | * them in fcntl(F_GETFD) or similar interfaces. | |
911 | */ | |
912 | flags &= VALID_OPEN_FLAGS; | |
913 | ||
e305f48b | 914 | if (flags & (O_CREAT | __O_TMPFILE)) |
e68726ff MS |
915 | op->mode = (mode & S_IALLUGO) | S_IFREG; |
916 | else | |
917 | op->mode = 0; | |
47c805dc AV |
918 | |
919 | /* Must never be set by userspace */ | |
c6f3d811 | 920 | flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC; |
47c805dc AV |
921 | |
922 | /* | |
923 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
924 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
925 | * always set instead of having to deal with possibly weird behaviour | |
926 | * for malicious applications setting only __O_SYNC. | |
927 | */ | |
928 | if (flags & __O_SYNC) | |
929 | flags |= O_DSYNC; | |
930 | ||
bb458c64 AV |
931 | if (flags & __O_TMPFILE) { |
932 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d | 933 | return -EINVAL; |
ba57ea64 AV |
934 | if (!(acc_mode & MAY_WRITE)) |
935 | return -EINVAL; | |
60545d0d AV |
936 | } else if (flags & O_PATH) { |
937 | /* | |
938 | * If we have O_PATH in the open flag. Then we | |
939 | * cannot have anything other than the below set of flags | |
940 | */ | |
1abf0c71 AV |
941 | flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH; |
942 | acc_mode = 0; | |
1abf0c71 | 943 | } |
47c805dc | 944 | |
1abf0c71 | 945 | op->open_flag = flags; |
47c805dc AV |
946 | |
947 | /* O_TRUNC implies we need access checks for write permissions */ | |
948 | if (flags & O_TRUNC) | |
949 | acc_mode |= MAY_WRITE; | |
950 | ||
951 | /* Allow the LSM permission hook to distinguish append | |
952 | access from general write access. */ | |
953 | if (flags & O_APPEND) | |
954 | acc_mode |= MAY_APPEND; | |
955 | ||
956 | op->acc_mode = acc_mode; | |
957 | ||
1abf0c71 AV |
958 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
959 | ||
47c805dc AV |
960 | if (flags & O_CREAT) { |
961 | op->intent |= LOOKUP_CREATE; | |
962 | if (flags & O_EXCL) | |
963 | op->intent |= LOOKUP_EXCL; | |
964 | } | |
965 | ||
966 | if (flags & O_DIRECTORY) | |
967 | lookup_flags |= LOOKUP_DIRECTORY; | |
968 | if (!(flags & O_NOFOLLOW)) | |
969 | lookup_flags |= LOOKUP_FOLLOW; | |
f9652e10 AV |
970 | op->lookup_flags = lookup_flags; |
971 | return 0; | |
47c805dc AV |
972 | } |
973 | ||
669abf4e JL |
974 | /** |
975 | * file_open_name - open file and return file pointer | |
976 | * | |
977 | * @name: struct filename containing path to open | |
978 | * @flags: open flags as per the open(2) second argument | |
979 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
980 | * | |
981 | * This is the helper to open a file from kernelspace if you really | |
982 | * have to. But in generally you should not do this, so please move | |
983 | * along, nothing to see here.. | |
984 | */ | |
985 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
986 | { | |
987 | struct open_flags op; | |
f9652e10 AV |
988 | int err = build_open_flags(flags, mode, &op); |
989 | return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
990 | } |
991 | ||
47c805dc AV |
992 | /** |
993 | * filp_open - open file and return file pointer | |
994 | * | |
995 | * @filename: path to open | |
996 | * @flags: open flags as per the open(2) second argument | |
997 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
998 | * | |
999 | * This is the helper to open a file from kernelspace if you really | |
1000 | * have to. But in generally you should not do this, so please move | |
1001 | * along, nothing to see here.. | |
1002 | */ | |
a218d0fd | 1003 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 1004 | { |
51689104 PM |
1005 | struct filename *name = getname_kernel(filename); |
1006 | struct file *file = ERR_CAST(name); | |
1007 | ||
1008 | if (!IS_ERR(name)) { | |
1009 | file = file_open_name(name, flags, mode); | |
1010 | putname(name); | |
1011 | } | |
1012 | return file; | |
47c805dc AV |
1013 | } |
1014 | EXPORT_SYMBOL(filp_open); | |
1015 | ||
73d049a4 | 1016 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
378c6520 | 1017 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1018 | { |
1019 | struct open_flags op; | |
378c6520 | 1020 | int err = build_open_flags(flags, mode, &op); |
f9652e10 AV |
1021 | if (err) |
1022 | return ERR_PTR(err); | |
f9652e10 | 1023 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
1024 | } |
1025 | EXPORT_SYMBOL(file_open_root); | |
1026 | ||
9a08c352 JB |
1027 | struct file *filp_clone_open(struct file *oldfile) |
1028 | { | |
1029 | struct file *file; | |
1030 | int retval; | |
1031 | ||
1032 | file = get_empty_filp(); | |
1033 | if (IS_ERR(file)) | |
1034 | return file; | |
1035 | ||
1036 | file->f_flags = oldfile->f_flags; | |
1037 | retval = vfs_open(&oldfile->f_path, file, oldfile->f_cred); | |
1038 | if (retval) { | |
1039 | put_filp(file); | |
1040 | return ERR_PTR(retval); | |
1041 | } | |
1042 | ||
1043 | return file; | |
1044 | } | |
1045 | EXPORT_SYMBOL(filp_clone_open); | |
1046 | ||
a218d0fd | 1047 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
1da177e4 | 1048 | { |
47c805dc | 1049 | struct open_flags op; |
f9652e10 AV |
1050 | int fd = build_open_flags(flags, mode, &op); |
1051 | struct filename *tmp; | |
1052 | ||
1053 | if (fd) | |
1054 | return fd; | |
1055 | ||
1056 | tmp = getname(filename); | |
1057 | if (IS_ERR(tmp)) | |
1058 | return PTR_ERR(tmp); | |
1059 | ||
1060 | fd = get_unused_fd_flags(flags); | |
1061 | if (fd >= 0) { | |
1062 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1063 | if (IS_ERR(f)) { | |
1064 | put_unused_fd(fd); | |
1065 | fd = PTR_ERR(f); | |
1066 | } else { | |
1067 | fsnotify_open(f); | |
1068 | fd_install(fd, f); | |
308a1eef | 1069 | trace_do_sys_open(tmp->name, flags, mode); |
1da177e4 | 1070 | } |
1da177e4 | 1071 | } |
f9652e10 | 1072 | putname(tmp); |
1da177e4 | 1073 | return fd; |
1da177e4 | 1074 | } |
e922efc3 | 1075 | |
a218d0fd | 1076 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) |
e922efc3 MS |
1077 | { |
1078 | if (force_o_largefile()) | |
1079 | flags |= O_LARGEFILE; | |
1080 | ||
2cf09666 | 1081 | return do_sys_open(AT_FDCWD, filename, flags, mode); |
e922efc3 | 1082 | } |
1da177e4 | 1083 | |
6559eed8 | 1084 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1085 | umode_t, mode) |
5590ff0d UD |
1086 | { |
1087 | if (force_o_largefile()) | |
1088 | flags |= O_LARGEFILE; | |
1089 | ||
2cf09666 | 1090 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1091 | } |
5590ff0d | 1092 | |
e35d49f6 AV |
1093 | #ifdef CONFIG_COMPAT |
1094 | /* | |
1095 | * Exactly like sys_open(), except that it doesn't set the | |
1096 | * O_LARGEFILE flag. | |
1097 | */ | |
1098 | COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1099 | { | |
1100 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
1101 | } | |
1102 | ||
1103 | /* | |
1104 | * Exactly like sys_openat(), except that it doesn't set the | |
1105 | * O_LARGEFILE flag. | |
1106 | */ | |
1107 | COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode) | |
1108 | { | |
1109 | return do_sys_open(dfd, filename, flags, mode); | |
1110 | } | |
1111 | #endif | |
1112 | ||
1da177e4 LT |
1113 | #ifndef __alpha__ |
1114 | ||
1115 | /* | |
1116 | * For backward compatibility? Maybe this should be moved | |
1117 | * into arch/i386 instead? | |
1118 | */ | |
a218d0fd | 1119 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 LT |
1120 | { |
1121 | return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); | |
1122 | } | |
1123 | ||
1124 | #endif | |
1125 | ||
1126 | /* | |
1127 | * "id" is the POSIX thread ID. We use the | |
1128 | * files pointer for this.. | |
1129 | */ | |
1130 | int filp_close(struct file *filp, fl_owner_t id) | |
1131 | { | |
45778ca8 | 1132 | int retval = 0; |
1da177e4 LT |
1133 | |
1134 | if (!file_count(filp)) { | |
1135 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1136 | return 0; |
1da177e4 LT |
1137 | } |
1138 | ||
72c2d531 | 1139 | if (filp->f_op->flush) |
75e1fcc0 | 1140 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1141 | |
1abf0c71 AV |
1142 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1143 | dnotify_flush(filp, id); | |
1144 | locks_remove_posix(filp, id); | |
1145 | } | |
1da177e4 LT |
1146 | fput(filp); |
1147 | return retval; | |
1148 | } | |
1149 | ||
1150 | EXPORT_SYMBOL(filp_close); | |
1151 | ||
1152 | /* | |
1153 | * Careful here! We test whether the file pointer is NULL before | |
1154 | * releasing the fd. This ensures that one clone task can't release | |
1155 | * an fd while another clone is opening it. | |
1156 | */ | |
ca013e94 | 1157 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1158 | { |
483ce1d4 | 1159 | int retval = __close_fd(current->files, fd); |
ee731f4f EP |
1160 | |
1161 | /* can't restart close syscall because file table entry was cleared */ | |
1162 | if (unlikely(retval == -ERESTARTSYS || | |
1163 | retval == -ERESTARTNOINTR || | |
1164 | retval == -ERESTARTNOHAND || | |
1165 | retval == -ERESTART_RESTARTBLOCK)) | |
1166 | retval = -EINTR; | |
1167 | ||
1168 | return retval; | |
1da177e4 | 1169 | } |
1da177e4 LT |
1170 | EXPORT_SYMBOL(sys_close); |
1171 | ||
1172 | /* | |
1173 | * This routine simulates a hangup on the tty, to arrange that users | |
1174 | * are given clean terminals at login time. | |
1175 | */ | |
ca013e94 | 1176 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1177 | { |
1178 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1179 | tty_vhangup_self(); |
1da177e4 LT |
1180 | return 0; |
1181 | } | |
1182 | return -EPERM; | |
1183 | } | |
1184 | ||
1185 | /* | |
1186 | * Called when an inode is about to be open. | |
1187 | * We use this to disallow opening large files on 32bit systems if | |
1188 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1189 | * on this flag in sys_open. | |
1190 | */ | |
1191 | int generic_file_open(struct inode * inode, struct file * filp) | |
1192 | { | |
1193 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1194 | return -EOVERFLOW; |
1da177e4 LT |
1195 | return 0; |
1196 | } | |
1197 | ||
1198 | EXPORT_SYMBOL(generic_file_open); | |
1199 | ||
1200 | /* | |
1201 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1202 | * file descriptors. The function is not supposed to ever fail, the only |
1203 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1204 | * directly into file_operations structure. | |
1da177e4 LT |
1205 | */ |
1206 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1207 | { | |
1208 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1209 | return 0; | |
1210 | } | |
1211 | ||
1212 | EXPORT_SYMBOL(nonseekable_open); |