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> |
1da177e4 LT |
22 | #include <asm/uaccess.h> |
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 | ||
4a30131e N |
37 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, |
38 | struct file *filp) | |
1da177e4 | 39 | { |
939a9421 | 40 | int ret; |
1da177e4 LT |
41 | struct iattr newattrs; |
42 | ||
43 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
44 | if (length < 0) | |
45 | return -EINVAL; | |
46 | ||
47 | newattrs.ia_size = length; | |
4a30131e | 48 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
49 | if (filp) { |
50 | newattrs.ia_file = filp; | |
51 | newattrs.ia_valid |= ATTR_FILE; | |
52 | } | |
1da177e4 | 53 | |
45f147a1 JK |
54 | /* Remove suid, sgid, and file capabilities on truncate too */ |
55 | ret = dentry_needs_remove_privs(dentry); | |
56 | if (ret < 0) | |
57 | return ret; | |
939a9421 AW |
58 | if (ret) |
59 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 60 | |
5955102c | 61 | inode_lock(dentry->d_inode); |
27ac0ffe BF |
62 | /* Note any delegations or leases have already been broken: */ |
63 | ret = notify_change(dentry, &newattrs, NULL); | |
5955102c | 64 | inode_unlock(dentry->d_inode); |
939a9421 | 65 | return ret; |
1da177e4 LT |
66 | } |
67 | ||
7df818b2 | 68 | long vfs_truncate(const struct path *path, loff_t length) |
1da177e4 | 69 | { |
2d8f3038 | 70 | struct inode *inode; |
e45612b3 | 71 | struct dentry *upperdentry; |
a02de960 | 72 | long error; |
1da177e4 | 73 | |
a02de960 | 74 | inode = path->dentry->d_inode; |
1da177e4 LT |
75 | |
76 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
1da177e4 | 77 | if (S_ISDIR(inode->i_mode)) |
a02de960 | 78 | return -EISDIR; |
1da177e4 | 79 | if (!S_ISREG(inode->i_mode)) |
a02de960 | 80 | return -EINVAL; |
1da177e4 | 81 | |
a02de960 | 82 | error = mnt_want_write(path->mnt); |
1da177e4 | 83 | if (error) |
a02de960 | 84 | goto out; |
1da177e4 | 85 | |
256984a8 | 86 | error = inode_permission(inode, MAY_WRITE); |
9ac9b847 DH |
87 | if (error) |
88 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
89 | |
90 | error = -EPERM; | |
c82e42da | 91 | if (IS_APPEND(inode)) |
9ac9b847 | 92 | goto mnt_drop_write_and_out; |
1da177e4 | 93 | |
e45612b3 MS |
94 | /* |
95 | * If this is an overlayfs then do as if opening the file so we get | |
96 | * write access on the upper inode, not on the overlay inode. For | |
97 | * non-overlay filesystems d_real() is an identity function. | |
98 | */ | |
99 | upperdentry = d_real(path->dentry, NULL, O_WRONLY); | |
100 | error = PTR_ERR(upperdentry); | |
101 | if (IS_ERR(upperdentry)) | |
102 | goto mnt_drop_write_and_out; | |
103 | ||
104 | error = get_write_access(upperdentry->d_inode); | |
1da177e4 | 105 | if (error) |
9ac9b847 | 106 | goto mnt_drop_write_and_out; |
1da177e4 | 107 | |
9700382c | 108 | /* |
109 | * Make sure that there are no leases. get_write_access() protects | |
110 | * against the truncate racing with a lease-granting setlease(). | |
111 | */ | |
8737c930 | 112 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 113 | if (error) |
9700382c | 114 | goto put_write_and_out; |
1da177e4 LT |
115 | |
116 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 | 117 | if (!error) |
a02de960 | 118 | error = security_path_truncate(path); |
907f4554 | 119 | if (!error) |
a02de960 | 120 | error = do_truncate(path->dentry, length, 0, NULL); |
1da177e4 | 121 | |
9700382c | 122 | put_write_and_out: |
e45612b3 | 123 | put_write_access(upperdentry->d_inode); |
9ac9b847 | 124 | mnt_drop_write_and_out: |
a02de960 | 125 | mnt_drop_write(path->mnt); |
1da177e4 LT |
126 | out: |
127 | return error; | |
128 | } | |
a02de960 DH |
129 | EXPORT_SYMBOL_GPL(vfs_truncate); |
130 | ||
131 | static long do_sys_truncate(const char __user *pathname, loff_t length) | |
132 | { | |
48f7530d | 133 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
134 | struct path path; |
135 | int error; | |
136 | ||
137 | if (length < 0) /* sorry, but loff_t says... */ | |
138 | return -EINVAL; | |
139 | ||
48f7530d JL |
140 | retry: |
141 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
142 | if (!error) { |
143 | error = vfs_truncate(&path, length); | |
144 | path_put(&path); | |
145 | } | |
48f7530d JL |
146 | if (retry_estale(error, lookup_flags)) { |
147 | lookup_flags |= LOOKUP_REVAL; | |
148 | goto retry; | |
149 | } | |
a02de960 DH |
150 | return error; |
151 | } | |
1da177e4 | 152 | |
4fd8da8d | 153 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 154 | { |
4fd8da8d | 155 | return do_sys_truncate(path, length); |
1da177e4 LT |
156 | } |
157 | ||
3f6d078d AV |
158 | #ifdef CONFIG_COMPAT |
159 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
160 | { | |
161 | return do_sys_truncate(path, length); | |
162 | } | |
163 | #endif | |
164 | ||
b01ec0ef | 165 | static long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 | 166 | { |
bf2965d5 | 167 | struct inode *inode; |
1da177e4 | 168 | struct dentry *dentry; |
2903ff01 | 169 | struct fd f; |
1da177e4 LT |
170 | int error; |
171 | ||
172 | error = -EINVAL; | |
173 | if (length < 0) | |
174 | goto out; | |
175 | error = -EBADF; | |
2903ff01 AV |
176 | f = fdget(fd); |
177 | if (!f.file) | |
1da177e4 LT |
178 | goto out; |
179 | ||
180 | /* explicitly opened as large or we are on 64-bit box */ | |
2903ff01 | 181 | if (f.file->f_flags & O_LARGEFILE) |
1da177e4 LT |
182 | small = 0; |
183 | ||
2903ff01 | 184 | dentry = f.file->f_path.dentry; |
1da177e4 LT |
185 | inode = dentry->d_inode; |
186 | error = -EINVAL; | |
2903ff01 | 187 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
188 | goto out_putf; |
189 | ||
190 | error = -EINVAL; | |
191 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
192 | if (small && length > MAX_NON_LFS) | |
193 | goto out_putf; | |
194 | ||
195 | error = -EPERM; | |
196 | if (IS_APPEND(inode)) | |
197 | goto out_putf; | |
198 | ||
14da9200 | 199 | sb_start_write(inode->i_sb); |
2903ff01 | 200 | error = locks_verify_truncate(inode, f.file, length); |
be6d3e56 | 201 | if (!error) |
2903ff01 | 202 | error = security_path_truncate(&f.file->f_path); |
1da177e4 | 203 | if (!error) |
2903ff01 | 204 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file); |
14da9200 | 205 | sb_end_write(inode->i_sb); |
1da177e4 | 206 | out_putf: |
2903ff01 | 207 | fdput(f); |
1da177e4 LT |
208 | out: |
209 | return error; | |
210 | } | |
211 | ||
bdc480e3 | 212 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 213 | { |
2cf09666 | 214 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
215 | } |
216 | ||
3f6d078d AV |
217 | #ifdef CONFIG_COMPAT |
218 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
219 | { | |
220 | return do_sys_ftruncate(fd, length, 1); | |
221 | } | |
222 | #endif | |
223 | ||
1da177e4 LT |
224 | /* LFS versions of truncate are only needed on 32 bit machines */ |
225 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 226 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
227 | { |
228 | return do_sys_truncate(path, length); | |
229 | } | |
230 | ||
4a0fd5bf | 231 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 232 | { |
2cf09666 | 233 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 234 | } |
6673e0c3 | 235 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 236 | |
3e63cbb1 | 237 | |
72c72bdf | 238 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 239 | { |
496ad9aa | 240 | struct inode *inode = file_inode(file); |
3e63cbb1 | 241 | long ret; |
97ac7350 AA |
242 | |
243 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 244 | return -EINVAL; |
97ac7350 AA |
245 | |
246 | /* Return error if mode is not supported */ | |
dd46c787 | 247 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) |
409332b6 LC |
248 | return -EOPNOTSUPP; |
249 | ||
250 | /* Punch hole and zero range are mutually exclusive */ | |
251 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
252 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
253 | return -EOPNOTSUPP; |
254 | ||
255 | /* Punch hole must have keep size set */ | |
256 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
257 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 258 | return -EOPNOTSUPP; |
97ac7350 | 259 | |
00f5e619 NJ |
260 | /* Collapse range should only be used exclusively. */ |
261 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
262 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
263 | return -EINVAL; | |
264 | ||
dd46c787 NJ |
265 | /* Insert range should only be used exclusively. */ |
266 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
267 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
268 | return -EINVAL; | |
269 | ||
97ac7350 | 270 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 271 | return -EBADF; |
1ca551c6 | 272 | |
00f5e619 | 273 | /* |
8fc61d92 | 274 | * We can only allow pure fallocate on append only files |
00f5e619 | 275 | */ |
8fc61d92 | 276 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
277 | return -EPERM; |
278 | ||
279 | if (IS_IMMUTABLE(inode)) | |
280 | return -EPERM; | |
281 | ||
0790b31b | 282 | /* |
6d2b6170 | 283 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
284 | */ |
285 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 286 | return -ETXTBSY; |
0790b31b | 287 | |
97ac7350 AA |
288 | /* |
289 | * Revalidate the write permissions, in case security policy has | |
290 | * changed since the files were opened. | |
291 | */ | |
292 | ret = security_file_permission(file, MAY_WRITE); | |
293 | if (ret) | |
3e63cbb1 | 294 | return ret; |
97ac7350 | 295 | |
97ac7350 | 296 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 297 | return -ESPIPE; |
97ac7350 | 298 | |
97ac7350 AA |
299 | /* |
300 | * Let individual file system decide if it supports preallocation | |
301 | * for directories or not. | |
302 | */ | |
303 | if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode)) | |
3e63cbb1 | 304 | return -ENODEV; |
97ac7350 | 305 | |
97ac7350 AA |
306 | /* Check for wrap through zero too */ |
307 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 308 | return -EFBIG; |
97ac7350 | 309 | |
2fe17c10 | 310 | if (!file->f_op->fallocate) |
3e63cbb1 | 311 | return -EOPNOTSUPP; |
97ac7350 | 312 | |
14da9200 JK |
313 | sb_start_write(inode->i_sb); |
314 | ret = file->f_op->fallocate(file, mode, offset, len); | |
820c12d5 HS |
315 | |
316 | /* | |
317 | * Create inotify and fanotify events. | |
318 | * | |
319 | * To keep the logic simple always create events if fallocate succeeds. | |
320 | * This implies that events are even created if the file size remains | |
321 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
322 | */ | |
323 | if (ret == 0) | |
324 | fsnotify_modify(file); | |
325 | ||
14da9200 JK |
326 | sb_end_write(inode->i_sb); |
327 | return ret; | |
3e63cbb1 | 328 | } |
72c72bdf | 329 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 330 | |
4a0fd5bf | 331 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
3e63cbb1 | 332 | { |
2903ff01 | 333 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
334 | int error = -EBADF; |
335 | ||
2903ff01 | 336 | if (f.file) { |
72c72bdf | 337 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 338 | fdput(f); |
3e63cbb1 | 339 | } |
3e63cbb1 | 340 | return error; |
97ac7350 | 341 | } |
3e63cbb1 | 342 | |
1da177e4 LT |
343 | /* |
344 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
345 | * We do this by temporarily clearing all FS-related capabilities and | |
346 | * switching the fsuid/fsgid around to the real ones. | |
347 | */ | |
6559eed8 | 348 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
1da177e4 | 349 | { |
d84f4f99 DH |
350 | const struct cred *old_cred; |
351 | struct cred *override_cred; | |
2d8f3038 | 352 | struct path path; |
256984a8 | 353 | struct inode *inode; |
1da177e4 | 354 | int res; |
87fa5595 | 355 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
1da177e4 LT |
356 | |
357 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
358 | return -EINVAL; | |
359 | ||
d84f4f99 DH |
360 | override_cred = prepare_creds(); |
361 | if (!override_cred) | |
362 | return -ENOMEM; | |
1da177e4 | 363 | |
d84f4f99 DH |
364 | override_cred->fsuid = override_cred->uid; |
365 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 366 | |
086f7316 | 367 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 368 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
369 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
370 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 371 | cap_clear(override_cred->cap_effective); |
086f7316 | 372 | else |
d84f4f99 DH |
373 | override_cred->cap_effective = |
374 | override_cred->cap_permitted; | |
086f7316 | 375 | } |
1da177e4 | 376 | |
d84f4f99 | 377 | old_cred = override_creds(override_cred); |
87fa5595 JL |
378 | retry: |
379 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
380 | if (res) |
381 | goto out; | |
382 | ||
63afdfc7 | 383 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
384 | |
385 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
386 | /* |
387 | * MAY_EXEC on regular files is denied if the fs is mounted | |
388 | * with the "noexec" flag. | |
389 | */ | |
390 | res = -EACCES; | |
90f8572b | 391 | if (path_noexec(&path)) |
30524472 AV |
392 | goto out_path_release; |
393 | } | |
394 | ||
256984a8 | 395 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 396 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 397 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 398 | goto out_path_release; |
2f676cbc DH |
399 | /* |
400 | * This is a rare case where using __mnt_is_readonly() | |
401 | * is OK without a mnt_want/drop_write() pair. Since | |
402 | * no actual write to the fs is performed here, we do | |
403 | * not need to telegraph to that to anyone. | |
404 | * | |
405 | * By doing this, we accept that this access is | |
406 | * inherently racy and know that the fs may change | |
407 | * state before we even see this result. | |
408 | */ | |
2d8f3038 | 409 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 410 | res = -EROFS; |
1da177e4 | 411 | |
6902d925 | 412 | out_path_release: |
2d8f3038 | 413 | path_put(&path); |
87fa5595 JL |
414 | if (retry_estale(res, lookup_flags)) { |
415 | lookup_flags |= LOOKUP_REVAL; | |
416 | goto retry; | |
417 | } | |
6902d925 | 418 | out: |
d84f4f99 DH |
419 | revert_creds(old_cred); |
420 | put_cred(override_cred); | |
1da177e4 LT |
421 | return res; |
422 | } | |
423 | ||
ca013e94 | 424 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d UD |
425 | { |
426 | return sys_faccessat(AT_FDCWD, filename, mode); | |
427 | } | |
428 | ||
3cdad428 | 429 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 430 | { |
2d8f3038 | 431 | struct path path; |
1da177e4 | 432 | int error; |
0291c0a5 JL |
433 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
434 | retry: | |
435 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
436 | if (error) |
437 | goto out; | |
438 | ||
9cfcac81 | 439 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
440 | if (error) |
441 | goto dput_and_out; | |
442 | ||
2d8f3038 | 443 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
444 | |
445 | dput_and_out: | |
2d8f3038 | 446 | path_put(&path); |
0291c0a5 JL |
447 | if (retry_estale(error, lookup_flags)) { |
448 | lookup_flags |= LOOKUP_REVAL; | |
449 | goto retry; | |
450 | } | |
1da177e4 LT |
451 | out: |
452 | return error; | |
453 | } | |
454 | ||
3cdad428 | 455 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 456 | { |
2903ff01 | 457 | struct fd f = fdget_raw(fd); |
1da177e4 | 458 | struct inode *inode; |
2903ff01 | 459 | int error = -EBADF; |
1da177e4 LT |
460 | |
461 | error = -EBADF; | |
2903ff01 | 462 | if (!f.file) |
1da177e4 LT |
463 | goto out; |
464 | ||
496ad9aa | 465 | inode = file_inode(f.file); |
1da177e4 LT |
466 | |
467 | error = -ENOTDIR; | |
468 | if (!S_ISDIR(inode->i_mode)) | |
469 | goto out_putf; | |
470 | ||
9cfcac81 | 471 | error = inode_permission(inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 | 472 | if (!error) |
2903ff01 | 473 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 474 | out_putf: |
2903ff01 | 475 | fdput(f); |
1da177e4 LT |
476 | out: |
477 | return error; | |
478 | } | |
479 | ||
3480b257 | 480 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 481 | { |
2d8f3038 | 482 | struct path path; |
1da177e4 | 483 | int error; |
2771261e JL |
484 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
485 | retry: | |
486 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
487 | if (error) |
488 | goto out; | |
489 | ||
9cfcac81 | 490 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
491 | if (error) |
492 | goto dput_and_out; | |
493 | ||
494 | error = -EPERM; | |
c7b96acf | 495 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 496 | goto dput_and_out; |
8b8efb44 TH |
497 | error = security_path_chroot(&path); |
498 | if (error) | |
499 | goto dput_and_out; | |
1da177e4 | 500 | |
2d8f3038 | 501 | set_fs_root(current->fs, &path); |
1da177e4 LT |
502 | error = 0; |
503 | dput_and_out: | |
2d8f3038 | 504 | path_put(&path); |
2771261e 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 | ||
be01f9f2 | 513 | static int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 514 | { |
e57712eb | 515 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 516 | struct inode *delegated_inode = NULL; |
1da177e4 | 517 | struct iattr newattrs; |
e57712eb | 518 | int error; |
1da177e4 | 519 | |
e57712eb AV |
520 | error = mnt_want_write(path->mnt); |
521 | if (error) | |
522 | return error; | |
27ac0ffe | 523 | retry_deleg: |
5955102c | 524 | inode_lock(inode); |
cdcf116d | 525 | error = security_path_chmod(path, mode); |
e57712eb | 526 | if (error) |
fe542cf5 | 527 | goto out_unlock; |
1da177e4 LT |
528 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
529 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 530 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 531 | out_unlock: |
5955102c | 532 | inode_unlock(inode); |
27ac0ffe BF |
533 | if (delegated_inode) { |
534 | error = break_deleg_wait(&delegated_inode); | |
535 | if (!error) | |
536 | goto retry_deleg; | |
537 | } | |
e57712eb AV |
538 | mnt_drop_write(path->mnt); |
539 | return error; | |
540 | } | |
541 | ||
49f0a076 | 542 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 543 | { |
173c8401 | 544 | struct fd f = fdget(fd); |
e57712eb AV |
545 | int err = -EBADF; |
546 | ||
173c8401 | 547 | if (f.file) { |
9f45f5bf | 548 | audit_file(f.file); |
173c8401 AV |
549 | err = chmod_common(&f.file->f_path, mode); |
550 | fdput(f); | |
e57712eb | 551 | } |
1da177e4 LT |
552 | return err; |
553 | } | |
554 | ||
49f0a076 | 555 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode) |
1da177e4 | 556 | { |
2d8f3038 | 557 | struct path path; |
1da177e4 | 558 | int error; |
14ff690c JL |
559 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
560 | retry: | |
561 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
562 | if (!error) { |
563 | error = chmod_common(&path, mode); | |
564 | path_put(&path); | |
14ff690c JL |
565 | if (retry_estale(error, lookup_flags)) { |
566 | lookup_flags |= LOOKUP_REVAL; | |
567 | goto retry; | |
568 | } | |
e57712eb | 569 | } |
1da177e4 LT |
570 | return error; |
571 | } | |
572 | ||
49f0a076 | 573 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d UD |
574 | { |
575 | return sys_fchmodat(AT_FDCWD, filename, mode); | |
576 | } | |
577 | ||
7fd25dac | 578 | static int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 579 | { |
fe542cf5 | 580 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 581 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
582 | int error; |
583 | struct iattr newattrs; | |
52137abe EB |
584 | kuid_t uid; |
585 | kgid_t gid; | |
586 | ||
587 | uid = make_kuid(current_user_ns(), user); | |
588 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 589 | |
c1b8940b | 590 | retry_deleg: |
1da177e4 LT |
591 | newattrs.ia_valid = ATTR_CTIME; |
592 | if (user != (uid_t) -1) { | |
52137abe EB |
593 | if (!uid_valid(uid)) |
594 | return -EINVAL; | |
1da177e4 | 595 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 596 | newattrs.ia_uid = uid; |
1da177e4 LT |
597 | } |
598 | if (group != (gid_t) -1) { | |
52137abe EB |
599 | if (!gid_valid(gid)) |
600 | return -EINVAL; | |
1da177e4 | 601 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 602 | newattrs.ia_gid = gid; |
1da177e4 LT |
603 | } |
604 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
605 | newattrs.ia_valid |= |
606 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
5955102c | 607 | inode_lock(inode); |
d2b31ca6 | 608 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 609 | if (!error) |
27ac0ffe | 610 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
5955102c | 611 | inode_unlock(inode); |
27ac0ffe BF |
612 | if (delegated_inode) { |
613 | error = break_deleg_wait(&delegated_inode); | |
614 | if (!error) | |
615 | goto retry_deleg; | |
616 | } | |
1da177e4 LT |
617 | return error; |
618 | } | |
619 | ||
6559eed8 HC |
620 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
621 | gid_t, group, int, flag) | |
5590ff0d | 622 | { |
2d8f3038 | 623 | struct path path; |
5590ff0d | 624 | int error = -EINVAL; |
65cfc672 | 625 | int lookup_flags; |
5590ff0d | 626 | |
65cfc672 | 627 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
628 | goto out; |
629 | ||
65cfc672 AV |
630 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
631 | if (flag & AT_EMPTY_PATH) | |
632 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 633 | retry: |
65cfc672 | 634 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
635 | if (error) |
636 | goto out; | |
2d8f3038 | 637 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
638 | if (error) |
639 | goto out_release; | |
fe542cf5 | 640 | error = chown_common(&path, user, group); |
2d8f3038 | 641 | mnt_drop_write(path.mnt); |
2af482a7 | 642 | out_release: |
2d8f3038 | 643 | path_put(&path); |
99a5df37 JL |
644 | if (retry_estale(error, lookup_flags)) { |
645 | lookup_flags |= LOOKUP_REVAL; | |
646 | goto retry; | |
647 | } | |
5590ff0d UD |
648 | out: |
649 | return error; | |
650 | } | |
651 | ||
55e4def0 | 652 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 653 | { |
55e4def0 DH |
654 | return sys_fchownat(AT_FDCWD, filename, user, group, 0); |
655 | } | |
1da177e4 | 656 | |
55e4def0 DH |
657 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
658 | { | |
659 | return sys_fchownat(AT_FDCWD, filename, user, group, | |
660 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
661 | } |
662 | ||
ca013e94 | 663 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
1da177e4 | 664 | { |
2903ff01 | 665 | struct fd f = fdget(fd); |
1da177e4 LT |
666 | int error = -EBADF; |
667 | ||
2903ff01 | 668 | if (!f.file) |
6902d925 DH |
669 | goto out; |
670 | ||
2903ff01 | 671 | error = mnt_want_write_file(f.file); |
2af482a7 DH |
672 | if (error) |
673 | goto out_fput; | |
9f45f5bf | 674 | audit_file(f.file); |
2903ff01 AV |
675 | error = chown_common(&f.file->f_path, user, group); |
676 | mnt_drop_write_file(f.file); | |
2af482a7 | 677 | out_fput: |
2903ff01 | 678 | fdput(f); |
6902d925 | 679 | out: |
1da177e4 LT |
680 | return error; |
681 | } | |
682 | ||
90ad1a8e MS |
683 | int open_check_o_direct(struct file *f) |
684 | { | |
685 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
686 | if (f->f_flags & O_DIRECT) { | |
e748dcd0 | 687 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) |
90ad1a8e | 688 | return -EINVAL; |
90ad1a8e MS |
689 | } |
690 | return 0; | |
691 | } | |
692 | ||
02e5180d | 693 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 694 | struct inode *inode, |
96b7e579 AV |
695 | int (*open)(struct inode *, struct file *), |
696 | const struct cred *cred) | |
1da177e4 | 697 | { |
1abf0c71 | 698 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
699 | int error; |
700 | ||
5300990c | 701 | f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK | |
a1a5b3d9 | 702 | FMODE_PREAD | FMODE_PWRITE; |
1abf0c71 | 703 | |
b5bcdda3 | 704 | path_get(&f->f_path); |
4bacc9c9 | 705 | f->f_inode = inode; |
1da177e4 | 706 | f->f_mapping = inode->i_mapping; |
1da177e4 | 707 | |
3f4d5a00 AV |
708 | if (unlikely(f->f_flags & O_PATH)) { |
709 | f->f_mode = FMODE_PATH; | |
1abf0c71 | 710 | f->f_op = &empty_fops; |
96b7e579 | 711 | return 0; |
1abf0c71 AV |
712 | } |
713 | ||
dd20908a | 714 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 715 | error = get_write_access(inode); |
3f4d5a00 | 716 | if (unlikely(error)) |
1da177e4 | 717 | goto cleanup_file; |
0ccb2863 | 718 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 719 | if (unlikely(error)) { |
0ccb2863 AV |
720 | put_write_access(inode); |
721 | goto cleanup_file; | |
722 | } | |
83f936c7 | 723 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
724 | } |
725 | ||
9c225f26 | 726 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
63b6df14 | 727 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) |
9c225f26 LT |
728 | f->f_mode |= FMODE_ATOMIC_POS; |
729 | ||
1abf0c71 | 730 | f->f_op = fops_get(inode->i_fop); |
72c2d531 AV |
731 | if (unlikely(WARN_ON(!f->f_op))) { |
732 | error = -ENODEV; | |
733 | goto cleanup_all; | |
734 | } | |
1abf0c71 | 735 | |
83d49856 | 736 | error = security_file_open(f, cred); |
788e7dd4 YN |
737 | if (error) |
738 | goto cleanup_all; | |
739 | ||
56f6af74 | 740 | error = break_lease(locks_inode(f), f->f_flags); |
f3c7691e BF |
741 | if (error) |
742 | goto cleanup_all; | |
743 | ||
72c2d531 | 744 | if (!open) |
834f2a4a TM |
745 | open = f->f_op->open; |
746 | if (open) { | |
747 | error = open(inode, f); | |
1da177e4 LT |
748 | if (error) |
749 | goto cleanup_all; | |
750 | } | |
890275b5 MZ |
751 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
752 | i_readcount_inc(inode); | |
293bc982 | 753 | if ((f->f_mode & FMODE_READ) && |
84363182 | 754 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 755 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 756 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 757 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 758 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 759 | |
1da177e4 LT |
760 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
761 | ||
762 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
763 | ||
96b7e579 | 764 | return 0; |
1da177e4 LT |
765 | |
766 | cleanup_all: | |
767 | fops_put(f->f_op); | |
83f936c7 | 768 | if (f->f_mode & FMODE_WRITER) { |
1da177e4 | 769 | put_write_access(inode); |
83f936c7 | 770 | __mnt_drop_write(f->f_path.mnt); |
4a3fd211 | 771 | } |
1da177e4 | 772 | cleanup_file: |
02e5180d AV |
773 | path_put(&f->f_path); |
774 | f->f_path.mnt = NULL; | |
775 | f->f_path.dentry = NULL; | |
dd37978c | 776 | f->f_inode = NULL; |
96b7e579 | 777 | return error; |
1da177e4 LT |
778 | } |
779 | ||
d18e9008 MS |
780 | /** |
781 | * finish_open - finish opening a file | |
0854d450 | 782 | * @file: file pointer |
d18e9008 MS |
783 | * @dentry: pointer to dentry |
784 | * @open: open callback | |
0854d450 | 785 | * @opened: state of open |
d18e9008 MS |
786 | * |
787 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
788 | * | |
789 | * If the open callback is set to NULL, then the standard f_op->open() | |
790 | * filesystem callback is substituted. | |
0854d450 MS |
791 | * |
792 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
793 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
794 | * on it after finish_open(). | |
795 | * | |
796 | * On successful return @file is a fully instantiated open file. After this, if | |
797 | * an error occurs in ->atomic_open(), it needs to clean up with fput(). | |
798 | * | |
799 | * Returns zero on success or -errno if the open failed. | |
d18e9008 | 800 | */ |
30d90494 AV |
801 | int finish_open(struct file *file, struct dentry *dentry, |
802 | int (*open)(struct inode *, struct file *), | |
803 | int *opened) | |
d18e9008 | 804 | { |
96b7e579 | 805 | int error; |
3d8a00d2 | 806 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 807 | |
b5bcdda3 | 808 | file->f_path.dentry = dentry; |
4bacc9c9 DH |
809 | error = do_dentry_open(file, d_backing_inode(dentry), open, |
810 | current_cred()); | |
96b7e579 | 811 | if (!error) |
47237687 | 812 | *opened |= FILE_OPENED; |
d18e9008 | 813 | |
96b7e579 | 814 | return error; |
d18e9008 MS |
815 | } |
816 | EXPORT_SYMBOL(finish_open); | |
817 | ||
818 | /** | |
819 | * finish_no_open - finish ->atomic_open() without opening the file | |
820 | * | |
0854d450 | 821 | * @file: file pointer |
d18e9008 MS |
822 | * @dentry: dentry or NULL (as returned from ->lookup()) |
823 | * | |
824 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
825 | * |
826 | * NB: unlike finish_open() this function does consume the dentry reference and | |
827 | * the caller need not dput() it. | |
828 | * | |
829 | * Returns "1" which must be the return value of ->atomic_open() after having | |
830 | * called this function. | |
d18e9008 | 831 | */ |
e45198a6 | 832 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 833 | { |
30d90494 | 834 | file->f_path.dentry = dentry; |
e45198a6 | 835 | return 1; |
d18e9008 MS |
836 | } |
837 | EXPORT_SYMBOL(finish_no_open); | |
838 | ||
9bf39ab2 MS |
839 | char *file_path(struct file *filp, char *buf, int buflen) |
840 | { | |
841 | return d_path(&filp->f_path, buf, buflen); | |
842 | } | |
843 | EXPORT_SYMBOL(file_path); | |
844 | ||
4bacc9c9 DH |
845 | /** |
846 | * vfs_open - open the file at the given path | |
847 | * @path: path to open | |
848 | * @file: newly allocated file with f_flag initialized | |
849 | * @cred: credentials to use | |
850 | */ | |
851 | int vfs_open(const struct path *path, struct file *file, | |
852 | const struct cred *cred) | |
853 | { | |
2d902671 | 854 | struct dentry *dentry = d_real(path->dentry, NULL, file->f_flags); |
4bacc9c9 | 855 | |
2d902671 MS |
856 | if (IS_ERR(dentry)) |
857 | return PTR_ERR(dentry); | |
4bacc9c9 | 858 | |
54d5ca87 | 859 | file->f_path = *path; |
2d902671 | 860 | return do_dentry_open(file, d_backing_inode(dentry), NULL, cred); |
4bacc9c9 DH |
861 | } |
862 | ||
765927b2 | 863 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 864 | const struct cred *cred) |
a1a5b3d9 PS |
865 | { |
866 | int error; | |
867 | struct file *f; | |
868 | ||
e0e81739 DH |
869 | validate_creds(cred); |
870 | ||
c212f9aa | 871 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 872 | BUG_ON(!path->mnt); |
322ee5b3 | 873 | |
a1a5b3d9 | 874 | f = get_empty_filp(); |
1afc99be AV |
875 | if (!IS_ERR(f)) { |
876 | f->f_flags = flags; | |
4aa7c634 | 877 | error = vfs_open(path, f, cred); |
1afc99be AV |
878 | if (!error) { |
879 | /* from now on we need fput() to dispose of f */ | |
880 | error = open_check_o_direct(f); | |
881 | if (error) { | |
882 | fput(f); | |
883 | f = ERR_PTR(error); | |
884 | } | |
885 | } else { | |
886 | put_filp(f); | |
2a027e7a AV |
887 | f = ERR_PTR(error); |
888 | } | |
2a027e7a AV |
889 | } |
890 | return f; | |
a1a5b3d9 | 891 | } |
1da177e4 LT |
892 | EXPORT_SYMBOL(dentry_open); |
893 | ||
a218d0fd | 894 | static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op) |
47c805dc AV |
895 | { |
896 | int lookup_flags = 0; | |
62fb4a15 | 897 | int acc_mode = ACC_MODE(flags); |
47c805dc | 898 | |
e305f48b | 899 | if (flags & (O_CREAT | __O_TMPFILE)) |
e68726ff MS |
900 | op->mode = (mode & S_IALLUGO) | S_IFREG; |
901 | else | |
902 | op->mode = 0; | |
47c805dc AV |
903 | |
904 | /* Must never be set by userspace */ | |
c6f3d811 | 905 | flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC; |
47c805dc AV |
906 | |
907 | /* | |
908 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
909 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
910 | * always set instead of having to deal with possibly weird behaviour | |
911 | * for malicious applications setting only __O_SYNC. | |
912 | */ | |
913 | if (flags & __O_SYNC) | |
914 | flags |= O_DSYNC; | |
915 | ||
bb458c64 AV |
916 | if (flags & __O_TMPFILE) { |
917 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d | 918 | return -EINVAL; |
ba57ea64 AV |
919 | if (!(acc_mode & MAY_WRITE)) |
920 | return -EINVAL; | |
60545d0d AV |
921 | } else if (flags & O_PATH) { |
922 | /* | |
923 | * If we have O_PATH in the open flag. Then we | |
924 | * cannot have anything other than the below set of flags | |
925 | */ | |
1abf0c71 AV |
926 | flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH; |
927 | acc_mode = 0; | |
1abf0c71 | 928 | } |
47c805dc | 929 | |
1abf0c71 | 930 | op->open_flag = flags; |
47c805dc AV |
931 | |
932 | /* O_TRUNC implies we need access checks for write permissions */ | |
933 | if (flags & O_TRUNC) | |
934 | acc_mode |= MAY_WRITE; | |
935 | ||
936 | /* Allow the LSM permission hook to distinguish append | |
937 | access from general write access. */ | |
938 | if (flags & O_APPEND) | |
939 | acc_mode |= MAY_APPEND; | |
940 | ||
941 | op->acc_mode = acc_mode; | |
942 | ||
1abf0c71 AV |
943 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
944 | ||
47c805dc AV |
945 | if (flags & O_CREAT) { |
946 | op->intent |= LOOKUP_CREATE; | |
947 | if (flags & O_EXCL) | |
948 | op->intent |= LOOKUP_EXCL; | |
949 | } | |
950 | ||
951 | if (flags & O_DIRECTORY) | |
952 | lookup_flags |= LOOKUP_DIRECTORY; | |
953 | if (!(flags & O_NOFOLLOW)) | |
954 | lookup_flags |= LOOKUP_FOLLOW; | |
f9652e10 AV |
955 | op->lookup_flags = lookup_flags; |
956 | return 0; | |
47c805dc AV |
957 | } |
958 | ||
669abf4e JL |
959 | /** |
960 | * file_open_name - open file and return file pointer | |
961 | * | |
962 | * @name: struct filename containing path to open | |
963 | * @flags: open flags as per the open(2) second argument | |
964 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
965 | * | |
966 | * This is the helper to open a file from kernelspace if you really | |
967 | * have to. But in generally you should not do this, so please move | |
968 | * along, nothing to see here.. | |
969 | */ | |
970 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
971 | { | |
972 | struct open_flags op; | |
f9652e10 AV |
973 | int err = build_open_flags(flags, mode, &op); |
974 | return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
975 | } |
976 | ||
47c805dc AV |
977 | /** |
978 | * filp_open - open file and return file pointer | |
979 | * | |
980 | * @filename: path to open | |
981 | * @flags: open flags as per the open(2) second argument | |
982 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
983 | * | |
984 | * This is the helper to open a file from kernelspace if you really | |
985 | * have to. But in generally you should not do this, so please move | |
986 | * along, nothing to see here.. | |
987 | */ | |
a218d0fd | 988 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 989 | { |
51689104 PM |
990 | struct filename *name = getname_kernel(filename); |
991 | struct file *file = ERR_CAST(name); | |
992 | ||
993 | if (!IS_ERR(name)) { | |
994 | file = file_open_name(name, flags, mode); | |
995 | putname(name); | |
996 | } | |
997 | return file; | |
47c805dc AV |
998 | } |
999 | EXPORT_SYMBOL(filp_open); | |
1000 | ||
73d049a4 | 1001 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
378c6520 | 1002 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1003 | { |
1004 | struct open_flags op; | |
378c6520 | 1005 | int err = build_open_flags(flags, mode, &op); |
f9652e10 AV |
1006 | if (err) |
1007 | return ERR_PTR(err); | |
f9652e10 | 1008 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
1009 | } |
1010 | EXPORT_SYMBOL(file_open_root); | |
1011 | ||
9a08c352 JB |
1012 | struct file *filp_clone_open(struct file *oldfile) |
1013 | { | |
1014 | struct file *file; | |
1015 | int retval; | |
1016 | ||
1017 | file = get_empty_filp(); | |
1018 | if (IS_ERR(file)) | |
1019 | return file; | |
1020 | ||
1021 | file->f_flags = oldfile->f_flags; | |
1022 | retval = vfs_open(&oldfile->f_path, file, oldfile->f_cred); | |
1023 | if (retval) { | |
1024 | put_filp(file); | |
1025 | return ERR_PTR(retval); | |
1026 | } | |
1027 | ||
1028 | return file; | |
1029 | } | |
1030 | EXPORT_SYMBOL(filp_clone_open); | |
1031 | ||
a218d0fd | 1032 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
1da177e4 | 1033 | { |
47c805dc | 1034 | struct open_flags op; |
f9652e10 AV |
1035 | int fd = build_open_flags(flags, mode, &op); |
1036 | struct filename *tmp; | |
1037 | ||
1038 | if (fd) | |
1039 | return fd; | |
1040 | ||
1041 | tmp = getname(filename); | |
1042 | if (IS_ERR(tmp)) | |
1043 | return PTR_ERR(tmp); | |
1044 | ||
1045 | fd = get_unused_fd_flags(flags); | |
1046 | if (fd >= 0) { | |
1047 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1048 | if (IS_ERR(f)) { | |
1049 | put_unused_fd(fd); | |
1050 | fd = PTR_ERR(f); | |
1051 | } else { | |
1052 | fsnotify_open(f); | |
1053 | fd_install(fd, f); | |
1da177e4 | 1054 | } |
1da177e4 | 1055 | } |
f9652e10 | 1056 | putname(tmp); |
1da177e4 | 1057 | return fd; |
1da177e4 | 1058 | } |
e922efc3 | 1059 | |
a218d0fd | 1060 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) |
e922efc3 MS |
1061 | { |
1062 | if (force_o_largefile()) | |
1063 | flags |= O_LARGEFILE; | |
1064 | ||
2cf09666 | 1065 | return do_sys_open(AT_FDCWD, filename, flags, mode); |
e922efc3 | 1066 | } |
1da177e4 | 1067 | |
6559eed8 | 1068 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1069 | umode_t, mode) |
5590ff0d UD |
1070 | { |
1071 | if (force_o_largefile()) | |
1072 | flags |= O_LARGEFILE; | |
1073 | ||
2cf09666 | 1074 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1075 | } |
5590ff0d | 1076 | |
1da177e4 LT |
1077 | #ifndef __alpha__ |
1078 | ||
1079 | /* | |
1080 | * For backward compatibility? Maybe this should be moved | |
1081 | * into arch/i386 instead? | |
1082 | */ | |
a218d0fd | 1083 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 LT |
1084 | { |
1085 | return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); | |
1086 | } | |
1087 | ||
1088 | #endif | |
1089 | ||
1090 | /* | |
1091 | * "id" is the POSIX thread ID. We use the | |
1092 | * files pointer for this.. | |
1093 | */ | |
1094 | int filp_close(struct file *filp, fl_owner_t id) | |
1095 | { | |
45778ca8 | 1096 | int retval = 0; |
1da177e4 LT |
1097 | |
1098 | if (!file_count(filp)) { | |
1099 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1100 | return 0; |
1da177e4 LT |
1101 | } |
1102 | ||
72c2d531 | 1103 | if (filp->f_op->flush) |
75e1fcc0 | 1104 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1105 | |
1abf0c71 AV |
1106 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1107 | dnotify_flush(filp, id); | |
1108 | locks_remove_posix(filp, id); | |
1109 | } | |
1da177e4 LT |
1110 | fput(filp); |
1111 | return retval; | |
1112 | } | |
1113 | ||
1114 | EXPORT_SYMBOL(filp_close); | |
1115 | ||
1116 | /* | |
1117 | * Careful here! We test whether the file pointer is NULL before | |
1118 | * releasing the fd. This ensures that one clone task can't release | |
1119 | * an fd while another clone is opening it. | |
1120 | */ | |
ca013e94 | 1121 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1122 | { |
483ce1d4 | 1123 | int retval = __close_fd(current->files, fd); |
ee731f4f EP |
1124 | |
1125 | /* can't restart close syscall because file table entry was cleared */ | |
1126 | if (unlikely(retval == -ERESTARTSYS || | |
1127 | retval == -ERESTARTNOINTR || | |
1128 | retval == -ERESTARTNOHAND || | |
1129 | retval == -ERESTART_RESTARTBLOCK)) | |
1130 | retval = -EINTR; | |
1131 | ||
1132 | return retval; | |
1da177e4 | 1133 | } |
1da177e4 LT |
1134 | EXPORT_SYMBOL(sys_close); |
1135 | ||
1136 | /* | |
1137 | * This routine simulates a hangup on the tty, to arrange that users | |
1138 | * are given clean terminals at login time. | |
1139 | */ | |
ca013e94 | 1140 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1141 | { |
1142 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1143 | tty_vhangup_self(); |
1da177e4 LT |
1144 | return 0; |
1145 | } | |
1146 | return -EPERM; | |
1147 | } | |
1148 | ||
1149 | /* | |
1150 | * Called when an inode is about to be open. | |
1151 | * We use this to disallow opening large files on 32bit systems if | |
1152 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1153 | * on this flag in sys_open. | |
1154 | */ | |
1155 | int generic_file_open(struct inode * inode, struct file * filp) | |
1156 | { | |
1157 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1158 | return -EOVERFLOW; |
1da177e4 LT |
1159 | return 0; |
1160 | } | |
1161 | ||
1162 | EXPORT_SYMBOL(generic_file_open); | |
1163 | ||
1164 | /* | |
1165 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1166 | * file descriptors. The function is not supposed to ever fail, the only |
1167 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1168 | * directly into file_operations structure. | |
1da177e4 LT |
1169 | */ |
1170 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1171 | { | |
1172 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1173 | return 0; | |
1174 | } | |
1175 | ||
1176 | EXPORT_SYMBOL(nonseekable_open); |