]>
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 | ||
227e7746 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; | |
199 | if (IS_APPEND(inode)) | |
200 | goto out_putf; | |
201 | ||
14da9200 | 202 | sb_start_write(inode->i_sb); |
2903ff01 | 203 | error = locks_verify_truncate(inode, f.file, length); |
be6d3e56 | 204 | if (!error) |
2903ff01 | 205 | error = security_path_truncate(&f.file->f_path); |
1da177e4 | 206 | if (!error) |
2903ff01 | 207 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file); |
14da9200 | 208 | sb_end_write(inode->i_sb); |
1da177e4 | 209 | out_putf: |
2903ff01 | 210 | fdput(f); |
1da177e4 LT |
211 | out: |
212 | return error; | |
213 | } | |
214 | ||
bdc480e3 | 215 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 216 | { |
2cf09666 | 217 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
218 | } |
219 | ||
3f6d078d AV |
220 | #ifdef CONFIG_COMPAT |
221 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
222 | { | |
223 | return do_sys_ftruncate(fd, length, 1); | |
224 | } | |
225 | #endif | |
226 | ||
1da177e4 LT |
227 | /* LFS versions of truncate are only needed on 32 bit machines */ |
228 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 229 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
230 | { |
231 | return do_sys_truncate(path, length); | |
232 | } | |
233 | ||
4a0fd5bf | 234 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 235 | { |
2cf09666 | 236 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 237 | } |
6673e0c3 | 238 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 239 | |
3e63cbb1 | 240 | |
72c72bdf | 241 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 242 | { |
496ad9aa | 243 | struct inode *inode = file_inode(file); |
3e63cbb1 | 244 | long ret; |
97ac7350 AA |
245 | |
246 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 247 | return -EINVAL; |
97ac7350 AA |
248 | |
249 | /* Return error if mode is not supported */ | |
dd46c787 | 250 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) |
409332b6 LC |
251 | return -EOPNOTSUPP; |
252 | ||
253 | /* Punch hole and zero range are mutually exclusive */ | |
254 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
255 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
256 | return -EOPNOTSUPP; |
257 | ||
258 | /* Punch hole must have keep size set */ | |
259 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
260 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 261 | return -EOPNOTSUPP; |
97ac7350 | 262 | |
00f5e619 NJ |
263 | /* Collapse range should only be used exclusively. */ |
264 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
265 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
266 | return -EINVAL; | |
267 | ||
dd46c787 NJ |
268 | /* Insert range should only be used exclusively. */ |
269 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
270 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
271 | return -EINVAL; | |
272 | ||
71be6b49 DW |
273 | /* Unshare range should only be used with allocate mode. */ |
274 | if ((mode & FALLOC_FL_UNSHARE_RANGE) && | |
275 | (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE))) | |
276 | return -EINVAL; | |
277 | ||
97ac7350 | 278 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 279 | return -EBADF; |
1ca551c6 | 280 | |
00f5e619 | 281 | /* |
8fc61d92 | 282 | * We can only allow pure fallocate on append only files |
00f5e619 | 283 | */ |
8fc61d92 | 284 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
285 | return -EPERM; |
286 | ||
287 | if (IS_IMMUTABLE(inode)) | |
288 | return -EPERM; | |
289 | ||
0790b31b | 290 | /* |
6d2b6170 | 291 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
292 | */ |
293 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 294 | return -ETXTBSY; |
0790b31b | 295 | |
97ac7350 AA |
296 | /* |
297 | * Revalidate the write permissions, in case security policy has | |
298 | * changed since the files were opened. | |
299 | */ | |
300 | ret = security_file_permission(file, MAY_WRITE); | |
301 | if (ret) | |
3e63cbb1 | 302 | return ret; |
97ac7350 | 303 | |
97ac7350 | 304 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 305 | return -ESPIPE; |
97ac7350 | 306 | |
97ac7350 AA |
307 | /* |
308 | * Let individual file system decide if it supports preallocation | |
309 | * for directories or not. | |
310 | */ | |
25f4c414 DW |
311 | if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode) && |
312 | !S_ISBLK(inode->i_mode)) | |
3e63cbb1 | 313 | return -ENODEV; |
97ac7350 | 314 | |
97ac7350 AA |
315 | /* Check for wrap through zero too */ |
316 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 317 | return -EFBIG; |
97ac7350 | 318 | |
2fe17c10 | 319 | if (!file->f_op->fallocate) |
3e63cbb1 | 320 | return -EOPNOTSUPP; |
97ac7350 | 321 | |
14da9200 JK |
322 | sb_start_write(inode->i_sb); |
323 | ret = file->f_op->fallocate(file, mode, offset, len); | |
820c12d5 HS |
324 | |
325 | /* | |
326 | * Create inotify and fanotify events. | |
327 | * | |
328 | * To keep the logic simple always create events if fallocate succeeds. | |
329 | * This implies that events are even created if the file size remains | |
330 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
331 | */ | |
332 | if (ret == 0) | |
333 | fsnotify_modify(file); | |
334 | ||
14da9200 JK |
335 | sb_end_write(inode->i_sb); |
336 | return ret; | |
3e63cbb1 | 337 | } |
72c72bdf | 338 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 339 | |
4a0fd5bf | 340 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
3e63cbb1 | 341 | { |
2903ff01 | 342 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
343 | int error = -EBADF; |
344 | ||
2903ff01 | 345 | if (f.file) { |
72c72bdf | 346 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 347 | fdput(f); |
3e63cbb1 | 348 | } |
3e63cbb1 | 349 | return error; |
97ac7350 | 350 | } |
3e63cbb1 | 351 | |
1da177e4 LT |
352 | /* |
353 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
354 | * We do this by temporarily clearing all FS-related capabilities and | |
355 | * switching the fsuid/fsgid around to the real ones. | |
356 | */ | |
6559eed8 | 357 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
1da177e4 | 358 | { |
d84f4f99 DH |
359 | const struct cred *old_cred; |
360 | struct cred *override_cred; | |
2d8f3038 | 361 | struct path path; |
256984a8 | 362 | struct inode *inode; |
1da177e4 | 363 | int res; |
87fa5595 | 364 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
1da177e4 LT |
365 | |
366 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
367 | return -EINVAL; | |
368 | ||
d84f4f99 DH |
369 | override_cred = prepare_creds(); |
370 | if (!override_cred) | |
371 | return -ENOMEM; | |
1da177e4 | 372 | |
d84f4f99 DH |
373 | override_cred->fsuid = override_cred->uid; |
374 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 375 | |
086f7316 | 376 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 377 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
378 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
379 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 380 | cap_clear(override_cred->cap_effective); |
086f7316 | 381 | else |
d84f4f99 DH |
382 | override_cred->cap_effective = |
383 | override_cred->cap_permitted; | |
086f7316 | 384 | } |
1da177e4 | 385 | |
d84f4f99 | 386 | old_cred = override_creds(override_cred); |
87fa5595 JL |
387 | retry: |
388 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
389 | if (res) |
390 | goto out; | |
391 | ||
63afdfc7 | 392 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
393 | |
394 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
395 | /* |
396 | * MAY_EXEC on regular files is denied if the fs is mounted | |
397 | * with the "noexec" flag. | |
398 | */ | |
399 | res = -EACCES; | |
90f8572b | 400 | if (path_noexec(&path)) |
30524472 AV |
401 | goto out_path_release; |
402 | } | |
403 | ||
256984a8 | 404 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 405 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 406 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 407 | goto out_path_release; |
2f676cbc DH |
408 | /* |
409 | * This is a rare case where using __mnt_is_readonly() | |
410 | * is OK without a mnt_want/drop_write() pair. Since | |
411 | * no actual write to the fs is performed here, we do | |
412 | * not need to telegraph to that to anyone. | |
413 | * | |
414 | * By doing this, we accept that this access is | |
415 | * inherently racy and know that the fs may change | |
416 | * state before we even see this result. | |
417 | */ | |
2d8f3038 | 418 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 419 | res = -EROFS; |
1da177e4 | 420 | |
6902d925 | 421 | out_path_release: |
2d8f3038 | 422 | path_put(&path); |
87fa5595 JL |
423 | if (retry_estale(res, lookup_flags)) { |
424 | lookup_flags |= LOOKUP_REVAL; | |
425 | goto retry; | |
426 | } | |
6902d925 | 427 | out: |
d84f4f99 DH |
428 | revert_creds(old_cred); |
429 | put_cred(override_cred); | |
1da177e4 LT |
430 | return res; |
431 | } | |
432 | ||
ca013e94 | 433 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d UD |
434 | { |
435 | return sys_faccessat(AT_FDCWD, filename, mode); | |
436 | } | |
437 | ||
3cdad428 | 438 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 439 | { |
2d8f3038 | 440 | struct path path; |
1da177e4 | 441 | int error; |
0291c0a5 JL |
442 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
443 | retry: | |
444 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
445 | if (error) |
446 | goto out; | |
447 | ||
9cfcac81 | 448 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
449 | if (error) |
450 | goto dput_and_out; | |
451 | ||
2d8f3038 | 452 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
453 | |
454 | dput_and_out: | |
2d8f3038 | 455 | path_put(&path); |
0291c0a5 JL |
456 | if (retry_estale(error, lookup_flags)) { |
457 | lookup_flags |= LOOKUP_REVAL; | |
458 | goto retry; | |
459 | } | |
1da177e4 LT |
460 | out: |
461 | return error; | |
462 | } | |
463 | ||
3cdad428 | 464 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 465 | { |
2903ff01 | 466 | struct fd f = fdget_raw(fd); |
1da177e4 | 467 | struct inode *inode; |
2903ff01 | 468 | int error = -EBADF; |
1da177e4 LT |
469 | |
470 | error = -EBADF; | |
2903ff01 | 471 | if (!f.file) |
1da177e4 LT |
472 | goto out; |
473 | ||
496ad9aa | 474 | inode = file_inode(f.file); |
1da177e4 LT |
475 | |
476 | error = -ENOTDIR; | |
477 | if (!S_ISDIR(inode->i_mode)) | |
478 | goto out_putf; | |
479 | ||
9cfcac81 | 480 | error = inode_permission(inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 | 481 | if (!error) |
2903ff01 | 482 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 483 | out_putf: |
2903ff01 | 484 | fdput(f); |
1da177e4 LT |
485 | out: |
486 | return error; | |
487 | } | |
488 | ||
3480b257 | 489 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 490 | { |
2d8f3038 | 491 | struct path path; |
1da177e4 | 492 | int error; |
2771261e JL |
493 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
494 | retry: | |
495 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
496 | if (error) |
497 | goto out; | |
498 | ||
9cfcac81 | 499 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
500 | if (error) |
501 | goto dput_and_out; | |
502 | ||
503 | error = -EPERM; | |
c7b96acf | 504 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 505 | goto dput_and_out; |
8b8efb44 TH |
506 | error = security_path_chroot(&path); |
507 | if (error) | |
508 | goto dput_and_out; | |
1da177e4 | 509 | |
2d8f3038 | 510 | set_fs_root(current->fs, &path); |
1da177e4 LT |
511 | error = 0; |
512 | dput_and_out: | |
2d8f3038 | 513 | path_put(&path); |
2771261e JL |
514 | if (retry_estale(error, lookup_flags)) { |
515 | lookup_flags |= LOOKUP_REVAL; | |
516 | goto retry; | |
517 | } | |
1da177e4 LT |
518 | out: |
519 | return error; | |
520 | } | |
521 | ||
be01f9f2 | 522 | static int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 523 | { |
e57712eb | 524 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 525 | struct inode *delegated_inode = NULL; |
1da177e4 | 526 | struct iattr newattrs; |
e57712eb | 527 | int error; |
1da177e4 | 528 | |
e57712eb AV |
529 | error = mnt_want_write(path->mnt); |
530 | if (error) | |
531 | return error; | |
27ac0ffe | 532 | retry_deleg: |
5955102c | 533 | inode_lock(inode); |
cdcf116d | 534 | error = security_path_chmod(path, mode); |
e57712eb | 535 | if (error) |
fe542cf5 | 536 | goto out_unlock; |
1da177e4 LT |
537 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
538 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 539 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 540 | out_unlock: |
5955102c | 541 | inode_unlock(inode); |
27ac0ffe BF |
542 | if (delegated_inode) { |
543 | error = break_deleg_wait(&delegated_inode); | |
544 | if (!error) | |
545 | goto retry_deleg; | |
546 | } | |
e57712eb AV |
547 | mnt_drop_write(path->mnt); |
548 | return error; | |
549 | } | |
550 | ||
49f0a076 | 551 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 552 | { |
173c8401 | 553 | struct fd f = fdget(fd); |
e57712eb AV |
554 | int err = -EBADF; |
555 | ||
173c8401 | 556 | if (f.file) { |
9f45f5bf | 557 | audit_file(f.file); |
173c8401 AV |
558 | err = chmod_common(&f.file->f_path, mode); |
559 | fdput(f); | |
e57712eb | 560 | } |
1da177e4 LT |
561 | return err; |
562 | } | |
563 | ||
49f0a076 | 564 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode) |
1da177e4 | 565 | { |
2d8f3038 | 566 | struct path path; |
1da177e4 | 567 | int error; |
14ff690c JL |
568 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
569 | retry: | |
570 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
571 | if (!error) { |
572 | error = chmod_common(&path, mode); | |
573 | path_put(&path); | |
14ff690c JL |
574 | if (retry_estale(error, lookup_flags)) { |
575 | lookup_flags |= LOOKUP_REVAL; | |
576 | goto retry; | |
577 | } | |
e57712eb | 578 | } |
1da177e4 LT |
579 | return error; |
580 | } | |
581 | ||
49f0a076 | 582 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d UD |
583 | { |
584 | return sys_fchmodat(AT_FDCWD, filename, mode); | |
585 | } | |
586 | ||
7fd25dac | 587 | static int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 588 | { |
fe542cf5 | 589 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 590 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
591 | int error; |
592 | struct iattr newattrs; | |
52137abe EB |
593 | kuid_t uid; |
594 | kgid_t gid; | |
595 | ||
596 | uid = make_kuid(current_user_ns(), user); | |
597 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 598 | |
c1b8940b | 599 | retry_deleg: |
1da177e4 LT |
600 | newattrs.ia_valid = ATTR_CTIME; |
601 | if (user != (uid_t) -1) { | |
52137abe EB |
602 | if (!uid_valid(uid)) |
603 | return -EINVAL; | |
1da177e4 | 604 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 605 | newattrs.ia_uid = uid; |
1da177e4 LT |
606 | } |
607 | if (group != (gid_t) -1) { | |
52137abe EB |
608 | if (!gid_valid(gid)) |
609 | return -EINVAL; | |
1da177e4 | 610 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 611 | newattrs.ia_gid = gid; |
1da177e4 LT |
612 | } |
613 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
614 | newattrs.ia_valid |= |
615 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
5955102c | 616 | inode_lock(inode); |
d2b31ca6 | 617 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 618 | if (!error) |
27ac0ffe | 619 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
5955102c | 620 | inode_unlock(inode); |
27ac0ffe BF |
621 | if (delegated_inode) { |
622 | error = break_deleg_wait(&delegated_inode); | |
623 | if (!error) | |
624 | goto retry_deleg; | |
625 | } | |
1da177e4 LT |
626 | return error; |
627 | } | |
628 | ||
6559eed8 HC |
629 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
630 | gid_t, group, int, flag) | |
5590ff0d | 631 | { |
2d8f3038 | 632 | struct path path; |
5590ff0d | 633 | int error = -EINVAL; |
65cfc672 | 634 | int lookup_flags; |
5590ff0d | 635 | |
65cfc672 | 636 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
637 | goto out; |
638 | ||
65cfc672 AV |
639 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
640 | if (flag & AT_EMPTY_PATH) | |
641 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 642 | retry: |
65cfc672 | 643 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
644 | if (error) |
645 | goto out; | |
2d8f3038 | 646 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
647 | if (error) |
648 | goto out_release; | |
fe542cf5 | 649 | error = chown_common(&path, user, group); |
2d8f3038 | 650 | mnt_drop_write(path.mnt); |
2af482a7 | 651 | out_release: |
2d8f3038 | 652 | path_put(&path); |
99a5df37 JL |
653 | if (retry_estale(error, lookup_flags)) { |
654 | lookup_flags |= LOOKUP_REVAL; | |
655 | goto retry; | |
656 | } | |
5590ff0d UD |
657 | out: |
658 | return error; | |
659 | } | |
660 | ||
55e4def0 | 661 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 662 | { |
55e4def0 DH |
663 | return sys_fchownat(AT_FDCWD, filename, user, group, 0); |
664 | } | |
1da177e4 | 665 | |
55e4def0 DH |
666 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
667 | { | |
668 | return sys_fchownat(AT_FDCWD, filename, user, group, | |
669 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
670 | } |
671 | ||
ca013e94 | 672 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
1da177e4 | 673 | { |
2903ff01 | 674 | struct fd f = fdget(fd); |
1da177e4 LT |
675 | int error = -EBADF; |
676 | ||
2903ff01 | 677 | if (!f.file) |
6902d925 DH |
678 | goto out; |
679 | ||
2903ff01 | 680 | error = mnt_want_write_file(f.file); |
2af482a7 DH |
681 | if (error) |
682 | goto out_fput; | |
9f45f5bf | 683 | audit_file(f.file); |
2903ff01 AV |
684 | error = chown_common(&f.file->f_path, user, group); |
685 | mnt_drop_write_file(f.file); | |
2af482a7 | 686 | out_fput: |
2903ff01 | 687 | fdput(f); |
6902d925 | 688 | out: |
1da177e4 LT |
689 | return error; |
690 | } | |
691 | ||
90ad1a8e MS |
692 | int open_check_o_direct(struct file *f) |
693 | { | |
694 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
695 | if (f->f_flags & O_DIRECT) { | |
e748dcd0 | 696 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) |
90ad1a8e | 697 | return -EINVAL; |
90ad1a8e MS |
698 | } |
699 | return 0; | |
700 | } | |
701 | ||
02e5180d | 702 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 703 | struct inode *inode, |
96b7e579 AV |
704 | int (*open)(struct inode *, struct file *), |
705 | const struct cred *cred) | |
1da177e4 | 706 | { |
1abf0c71 | 707 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
708 | int error; |
709 | ||
5300990c | 710 | f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK | |
a1a5b3d9 | 711 | FMODE_PREAD | FMODE_PWRITE; |
1abf0c71 | 712 | |
b5bcdda3 | 713 | path_get(&f->f_path); |
4bacc9c9 | 714 | f->f_inode = inode; |
1da177e4 | 715 | f->f_mapping = inode->i_mapping; |
1da177e4 | 716 | |
3f4d5a00 AV |
717 | if (unlikely(f->f_flags & O_PATH)) { |
718 | f->f_mode = FMODE_PATH; | |
1abf0c71 | 719 | f->f_op = &empty_fops; |
96b7e579 | 720 | return 0; |
1abf0c71 AV |
721 | } |
722 | ||
dd20908a | 723 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 724 | error = get_write_access(inode); |
3f4d5a00 | 725 | if (unlikely(error)) |
1da177e4 | 726 | goto cleanup_file; |
0ccb2863 | 727 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 728 | if (unlikely(error)) { |
0ccb2863 AV |
729 | put_write_access(inode); |
730 | goto cleanup_file; | |
731 | } | |
83f936c7 | 732 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
733 | } |
734 | ||
9c225f26 | 735 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
63b6df14 | 736 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) |
9c225f26 LT |
737 | f->f_mode |= FMODE_ATOMIC_POS; |
738 | ||
1abf0c71 | 739 | f->f_op = fops_get(inode->i_fop); |
72c2d531 AV |
740 | if (unlikely(WARN_ON(!f->f_op))) { |
741 | error = -ENODEV; | |
742 | goto cleanup_all; | |
743 | } | |
1abf0c71 | 744 | |
83d49856 | 745 | error = security_file_open(f, cred); |
788e7dd4 YN |
746 | if (error) |
747 | goto cleanup_all; | |
748 | ||
c568d683 | 749 | error = break_lease(locks_inode(f), f->f_flags); |
f3c7691e BF |
750 | if (error) |
751 | goto cleanup_all; | |
752 | ||
72c2d531 | 753 | if (!open) |
834f2a4a TM |
754 | open = f->f_op->open; |
755 | if (open) { | |
756 | error = open(inode, f); | |
1da177e4 LT |
757 | if (error) |
758 | goto cleanup_all; | |
759 | } | |
890275b5 MZ |
760 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
761 | i_readcount_inc(inode); | |
293bc982 | 762 | if ((f->f_mode & FMODE_READ) && |
84363182 | 763 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 764 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 765 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 766 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 767 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 768 | |
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 | |
e305f48b | 908 | if (flags & (O_CREAT | __O_TMPFILE)) |
e68726ff MS |
909 | op->mode = (mode & S_IALLUGO) | S_IFREG; |
910 | else | |
911 | op->mode = 0; | |
47c805dc AV |
912 | |
913 | /* Must never be set by userspace */ | |
c6f3d811 | 914 | flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC; |
47c805dc AV |
915 | |
916 | /* | |
917 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
918 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
919 | * always set instead of having to deal with possibly weird behaviour | |
920 | * for malicious applications setting only __O_SYNC. | |
921 | */ | |
922 | if (flags & __O_SYNC) | |
923 | flags |= O_DSYNC; | |
924 | ||
bb458c64 AV |
925 | if (flags & __O_TMPFILE) { |
926 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d | 927 | return -EINVAL; |
ba57ea64 AV |
928 | if (!(acc_mode & MAY_WRITE)) |
929 | return -EINVAL; | |
60545d0d AV |
930 | } else if (flags & O_PATH) { |
931 | /* | |
932 | * If we have O_PATH in the open flag. Then we | |
933 | * cannot have anything other than the below set of flags | |
934 | */ | |
1abf0c71 AV |
935 | flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH; |
936 | acc_mode = 0; | |
1abf0c71 | 937 | } |
47c805dc | 938 | |
1abf0c71 | 939 | op->open_flag = flags; |
47c805dc AV |
940 | |
941 | /* O_TRUNC implies we need access checks for write permissions */ | |
942 | if (flags & O_TRUNC) | |
943 | acc_mode |= MAY_WRITE; | |
944 | ||
945 | /* Allow the LSM permission hook to distinguish append | |
946 | access from general write access. */ | |
947 | if (flags & O_APPEND) | |
948 | acc_mode |= MAY_APPEND; | |
949 | ||
950 | op->acc_mode = acc_mode; | |
951 | ||
1abf0c71 AV |
952 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
953 | ||
47c805dc AV |
954 | if (flags & O_CREAT) { |
955 | op->intent |= LOOKUP_CREATE; | |
956 | if (flags & O_EXCL) | |
957 | op->intent |= LOOKUP_EXCL; | |
958 | } | |
959 | ||
960 | if (flags & O_DIRECTORY) | |
961 | lookup_flags |= LOOKUP_DIRECTORY; | |
962 | if (!(flags & O_NOFOLLOW)) | |
963 | lookup_flags |= LOOKUP_FOLLOW; | |
f9652e10 AV |
964 | op->lookup_flags = lookup_flags; |
965 | return 0; | |
47c805dc AV |
966 | } |
967 | ||
669abf4e JL |
968 | /** |
969 | * file_open_name - open file and return file pointer | |
970 | * | |
971 | * @name: struct filename containing path to open | |
972 | * @flags: open flags as per the open(2) second argument | |
973 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
974 | * | |
975 | * This is the helper to open a file from kernelspace if you really | |
976 | * have to. But in generally you should not do this, so please move | |
977 | * along, nothing to see here.. | |
978 | */ | |
979 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
980 | { | |
981 | struct open_flags op; | |
f9652e10 AV |
982 | int err = build_open_flags(flags, mode, &op); |
983 | return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
984 | } |
985 | ||
47c805dc AV |
986 | /** |
987 | * filp_open - open file and return file pointer | |
988 | * | |
989 | * @filename: path to open | |
990 | * @flags: open flags as per the open(2) second argument | |
991 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
992 | * | |
993 | * This is the helper to open a file from kernelspace if you really | |
994 | * have to. But in generally you should not do this, so please move | |
995 | * along, nothing to see here.. | |
996 | */ | |
a218d0fd | 997 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 998 | { |
51689104 PM |
999 | struct filename *name = getname_kernel(filename); |
1000 | struct file *file = ERR_CAST(name); | |
1001 | ||
1002 | if (!IS_ERR(name)) { | |
1003 | file = file_open_name(name, flags, mode); | |
1004 | putname(name); | |
1005 | } | |
1006 | return file; | |
47c805dc AV |
1007 | } |
1008 | EXPORT_SYMBOL(filp_open); | |
1009 | ||
73d049a4 | 1010 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
378c6520 | 1011 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1012 | { |
1013 | struct open_flags op; | |
378c6520 | 1014 | int err = build_open_flags(flags, mode, &op); |
f9652e10 AV |
1015 | if (err) |
1016 | return ERR_PTR(err); | |
f9652e10 | 1017 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
1018 | } |
1019 | EXPORT_SYMBOL(file_open_root); | |
1020 | ||
9a08c352 JB |
1021 | struct file *filp_clone_open(struct file *oldfile) |
1022 | { | |
1023 | struct file *file; | |
1024 | int retval; | |
1025 | ||
1026 | file = get_empty_filp(); | |
1027 | if (IS_ERR(file)) | |
1028 | return file; | |
1029 | ||
1030 | file->f_flags = oldfile->f_flags; | |
1031 | retval = vfs_open(&oldfile->f_path, file, oldfile->f_cred); | |
1032 | if (retval) { | |
1033 | put_filp(file); | |
1034 | return ERR_PTR(retval); | |
1035 | } | |
1036 | ||
1037 | return file; | |
1038 | } | |
1039 | EXPORT_SYMBOL(filp_clone_open); | |
1040 | ||
a218d0fd | 1041 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
1da177e4 | 1042 | { |
47c805dc | 1043 | struct open_flags op; |
f9652e10 AV |
1044 | int fd = build_open_flags(flags, mode, &op); |
1045 | struct filename *tmp; | |
1046 | ||
1047 | if (fd) | |
1048 | return fd; | |
1049 | ||
1050 | tmp = getname(filename); | |
1051 | if (IS_ERR(tmp)) | |
1052 | return PTR_ERR(tmp); | |
1053 | ||
1054 | fd = get_unused_fd_flags(flags); | |
1055 | if (fd >= 0) { | |
1056 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1057 | if (IS_ERR(f)) { | |
1058 | put_unused_fd(fd); | |
1059 | fd = PTR_ERR(f); | |
1060 | } else { | |
1061 | fsnotify_open(f); | |
1062 | fd_install(fd, f); | |
227e7746 | 1063 | trace_do_sys_open(tmp->name, flags, mode); |
1da177e4 | 1064 | } |
1da177e4 | 1065 | } |
f9652e10 | 1066 | putname(tmp); |
1da177e4 | 1067 | return fd; |
1da177e4 | 1068 | } |
e922efc3 | 1069 | |
a218d0fd | 1070 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) |
e922efc3 MS |
1071 | { |
1072 | if (force_o_largefile()) | |
1073 | flags |= O_LARGEFILE; | |
1074 | ||
2cf09666 | 1075 | return do_sys_open(AT_FDCWD, filename, flags, mode); |
e922efc3 | 1076 | } |
1da177e4 | 1077 | |
6559eed8 | 1078 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1079 | umode_t, mode) |
5590ff0d UD |
1080 | { |
1081 | if (force_o_largefile()) | |
1082 | flags |= O_LARGEFILE; | |
1083 | ||
2cf09666 | 1084 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1085 | } |
5590ff0d | 1086 | |
1da177e4 LT |
1087 | #ifndef __alpha__ |
1088 | ||
1089 | /* | |
1090 | * For backward compatibility? Maybe this should be moved | |
1091 | * into arch/i386 instead? | |
1092 | */ | |
a218d0fd | 1093 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 LT |
1094 | { |
1095 | return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); | |
1096 | } | |
1097 | ||
1098 | #endif | |
1099 | ||
1100 | /* | |
1101 | * "id" is the POSIX thread ID. We use the | |
1102 | * files pointer for this.. | |
1103 | */ | |
1104 | int filp_close(struct file *filp, fl_owner_t id) | |
1105 | { | |
45778ca8 | 1106 | int retval = 0; |
1da177e4 LT |
1107 | |
1108 | if (!file_count(filp)) { | |
1109 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1110 | return 0; |
1da177e4 LT |
1111 | } |
1112 | ||
72c2d531 | 1113 | if (filp->f_op->flush) |
75e1fcc0 | 1114 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1115 | |
1abf0c71 AV |
1116 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1117 | dnotify_flush(filp, id); | |
1118 | locks_remove_posix(filp, id); | |
1119 | } | |
1da177e4 LT |
1120 | fput(filp); |
1121 | return retval; | |
1122 | } | |
1123 | ||
1124 | EXPORT_SYMBOL(filp_close); | |
1125 | ||
1126 | /* | |
1127 | * Careful here! We test whether the file pointer is NULL before | |
1128 | * releasing the fd. This ensures that one clone task can't release | |
1129 | * an fd while another clone is opening it. | |
1130 | */ | |
ca013e94 | 1131 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1132 | { |
483ce1d4 | 1133 | int retval = __close_fd(current->files, fd); |
ee731f4f EP |
1134 | |
1135 | /* can't restart close syscall because file table entry was cleared */ | |
1136 | if (unlikely(retval == -ERESTARTSYS || | |
1137 | retval == -ERESTARTNOINTR || | |
1138 | retval == -ERESTARTNOHAND || | |
1139 | retval == -ERESTART_RESTARTBLOCK)) | |
1140 | retval = -EINTR; | |
1141 | ||
1142 | return retval; | |
1da177e4 | 1143 | } |
1da177e4 LT |
1144 | EXPORT_SYMBOL(sys_close); |
1145 | ||
1146 | /* | |
1147 | * This routine simulates a hangup on the tty, to arrange that users | |
1148 | * are given clean terminals at login time. | |
1149 | */ | |
ca013e94 | 1150 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1151 | { |
1152 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1153 | tty_vhangup_self(); |
1da177e4 LT |
1154 | return 0; |
1155 | } | |
1156 | return -EPERM; | |
1157 | } | |
1158 | ||
1159 | /* | |
1160 | * Called when an inode is about to be open. | |
1161 | * We use this to disallow opening large files on 32bit systems if | |
1162 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1163 | * on this flag in sys_open. | |
1164 | */ | |
1165 | int generic_file_open(struct inode * inode, struct file * filp) | |
1166 | { | |
1167 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1168 | return -EOVERFLOW; |
1da177e4 LT |
1169 | return 0; |
1170 | } | |
1171 | ||
1172 | EXPORT_SYMBOL(generic_file_open); | |
1173 | ||
1174 | /* | |
1175 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1176 | * file descriptors. The function is not supposed to ever fail, the only |
1177 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1178 | * directly into file_operations structure. | |
1da177e4 LT |
1179 | */ |
1180 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1181 | { | |
1182 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1183 | return 0; | |
1184 | } | |
1185 | ||
1186 | EXPORT_SYMBOL(nonseekable_open); |