]>
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 | ||
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; |
4d0c5ba2 | 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 | |
4d0c5ba2 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 | */ | |
495e6429 | 99 | upperdentry = d_real(path->dentry, NULL, O_WRONLY, 0); |
4d0c5ba2 MS |
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: |
4d0c5ba2 | 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 | ||
411d9475 | 165 | 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; | |
78757af6 AG |
196 | /* Check IS_APPEND on real upper inode */ |
197 | if (IS_APPEND(file_inode(f.file))) | |
1da177e4 LT |
198 | goto out_putf; |
199 | ||
14da9200 | 200 | sb_start_write(inode->i_sb); |
2903ff01 | 201 | error = locks_verify_truncate(inode, f.file, length); |
be6d3e56 | 202 | if (!error) |
2903ff01 | 203 | error = security_path_truncate(&f.file->f_path); |
1da177e4 | 204 | if (!error) |
2903ff01 | 205 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file); |
14da9200 | 206 | sb_end_write(inode->i_sb); |
1da177e4 | 207 | out_putf: |
2903ff01 | 208 | fdput(f); |
1da177e4 LT |
209 | out: |
210 | return error; | |
211 | } | |
212 | ||
bdc480e3 | 213 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 214 | { |
2cf09666 | 215 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
216 | } |
217 | ||
3f6d078d AV |
218 | #ifdef CONFIG_COMPAT |
219 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
220 | { | |
221 | return do_sys_ftruncate(fd, length, 1); | |
222 | } | |
223 | #endif | |
224 | ||
1da177e4 LT |
225 | /* LFS versions of truncate are only needed on 32 bit machines */ |
226 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 227 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
228 | { |
229 | return do_sys_truncate(path, length); | |
230 | } | |
231 | ||
4a0fd5bf | 232 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 233 | { |
2cf09666 | 234 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 235 | } |
6673e0c3 | 236 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 237 | |
3e63cbb1 | 238 | |
72c72bdf | 239 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 240 | { |
496ad9aa | 241 | struct inode *inode = file_inode(file); |
3e63cbb1 | 242 | long ret; |
97ac7350 AA |
243 | |
244 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 245 | return -EINVAL; |
97ac7350 AA |
246 | |
247 | /* Return error if mode is not supported */ | |
dd46c787 | 248 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) |
409332b6 LC |
249 | return -EOPNOTSUPP; |
250 | ||
251 | /* Punch hole and zero range are mutually exclusive */ | |
252 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
253 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
254 | return -EOPNOTSUPP; |
255 | ||
256 | /* Punch hole must have keep size set */ | |
257 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
258 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 259 | return -EOPNOTSUPP; |
97ac7350 | 260 | |
00f5e619 NJ |
261 | /* Collapse range should only be used exclusively. */ |
262 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
263 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
264 | return -EINVAL; | |
265 | ||
dd46c787 NJ |
266 | /* Insert range should only be used exclusively. */ |
267 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
268 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
269 | return -EINVAL; | |
270 | ||
71be6b49 DW |
271 | /* Unshare range should only be used with allocate mode. */ |
272 | if ((mode & FALLOC_FL_UNSHARE_RANGE) && | |
273 | (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE))) | |
274 | return -EINVAL; | |
275 | ||
97ac7350 | 276 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 277 | return -EBADF; |
1ca551c6 | 278 | |
00f5e619 | 279 | /* |
8fc61d92 | 280 | * We can only allow pure fallocate on append only files |
00f5e619 | 281 | */ |
8fc61d92 | 282 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
283 | return -EPERM; |
284 | ||
285 | if (IS_IMMUTABLE(inode)) | |
286 | return -EPERM; | |
287 | ||
0790b31b | 288 | /* |
6d2b6170 | 289 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
290 | */ |
291 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 292 | return -ETXTBSY; |
0790b31b | 293 | |
97ac7350 AA |
294 | /* |
295 | * Revalidate the write permissions, in case security policy has | |
296 | * changed since the files were opened. | |
297 | */ | |
298 | ret = security_file_permission(file, MAY_WRITE); | |
299 | if (ret) | |
3e63cbb1 | 300 | return ret; |
97ac7350 | 301 | |
97ac7350 | 302 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 303 | return -ESPIPE; |
97ac7350 | 304 | |
9e79b132 AG |
305 | if (S_ISDIR(inode->i_mode)) |
306 | return -EISDIR; | |
307 | ||
308 | if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode)) | |
3e63cbb1 | 309 | return -ENODEV; |
97ac7350 | 310 | |
97ac7350 AA |
311 | /* Check for wrap through zero too */ |
312 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 313 | return -EFBIG; |
97ac7350 | 314 | |
2fe17c10 | 315 | if (!file->f_op->fallocate) |
3e63cbb1 | 316 | return -EOPNOTSUPP; |
97ac7350 | 317 | |
bfe219d3 | 318 | file_start_write(file); |
14da9200 | 319 | ret = file->f_op->fallocate(file, mode, offset, len); |
820c12d5 HS |
320 | |
321 | /* | |
322 | * Create inotify and fanotify events. | |
323 | * | |
324 | * To keep the logic simple always create events if fallocate succeeds. | |
325 | * This implies that events are even created if the file size remains | |
326 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
327 | */ | |
328 | if (ret == 0) | |
329 | fsnotify_modify(file); | |
330 | ||
bfe219d3 | 331 | file_end_write(file); |
14da9200 | 332 | return ret; |
3e63cbb1 | 333 | } |
72c72bdf | 334 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 335 | |
4a0fd5bf | 336 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
3e63cbb1 | 337 | { |
2903ff01 | 338 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
339 | int error = -EBADF; |
340 | ||
2903ff01 | 341 | if (f.file) { |
72c72bdf | 342 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 343 | fdput(f); |
3e63cbb1 | 344 | } |
3e63cbb1 | 345 | return error; |
97ac7350 | 346 | } |
3e63cbb1 | 347 | |
1da177e4 LT |
348 | /* |
349 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
350 | * We do this by temporarily clearing all FS-related capabilities and | |
351 | * switching the fsuid/fsgid around to the real ones. | |
352 | */ | |
cbfe20f5 | 353 | long do_faccessat(int dfd, const char __user *filename, int mode) |
1da177e4 | 354 | { |
d84f4f99 DH |
355 | const struct cred *old_cred; |
356 | struct cred *override_cred; | |
2d8f3038 | 357 | struct path path; |
256984a8 | 358 | struct inode *inode; |
1da177e4 | 359 | int res; |
87fa5595 | 360 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
1da177e4 LT |
361 | |
362 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
363 | return -EINVAL; | |
364 | ||
d84f4f99 DH |
365 | override_cred = prepare_creds(); |
366 | if (!override_cred) | |
367 | return -ENOMEM; | |
1da177e4 | 368 | |
d84f4f99 DH |
369 | override_cred->fsuid = override_cred->uid; |
370 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 371 | |
086f7316 | 372 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 373 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
374 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
375 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 376 | cap_clear(override_cred->cap_effective); |
086f7316 | 377 | else |
d84f4f99 DH |
378 | override_cred->cap_effective = |
379 | override_cred->cap_permitted; | |
086f7316 | 380 | } |
1da177e4 | 381 | |
d84f4f99 | 382 | old_cred = override_creds(override_cred); |
87fa5595 JL |
383 | retry: |
384 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
385 | if (res) |
386 | goto out; | |
387 | ||
63afdfc7 | 388 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
389 | |
390 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
391 | /* |
392 | * MAY_EXEC on regular files is denied if the fs is mounted | |
393 | * with the "noexec" flag. | |
394 | */ | |
395 | res = -EACCES; | |
90f8572b | 396 | if (path_noexec(&path)) |
30524472 AV |
397 | goto out_path_release; |
398 | } | |
399 | ||
256984a8 | 400 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 401 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 402 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 403 | goto out_path_release; |
2f676cbc DH |
404 | /* |
405 | * This is a rare case where using __mnt_is_readonly() | |
406 | * is OK without a mnt_want/drop_write() pair. Since | |
407 | * no actual write to the fs is performed here, we do | |
408 | * not need to telegraph to that to anyone. | |
409 | * | |
410 | * By doing this, we accept that this access is | |
411 | * inherently racy and know that the fs may change | |
412 | * state before we even see this result. | |
413 | */ | |
2d8f3038 | 414 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 415 | res = -EROFS; |
1da177e4 | 416 | |
6902d925 | 417 | out_path_release: |
2d8f3038 | 418 | path_put(&path); |
87fa5595 JL |
419 | if (retry_estale(res, lookup_flags)) { |
420 | lookup_flags |= LOOKUP_REVAL; | |
421 | goto retry; | |
422 | } | |
6902d925 | 423 | out: |
d84f4f99 DH |
424 | revert_creds(old_cred); |
425 | put_cred(override_cred); | |
1da177e4 LT |
426 | return res; |
427 | } | |
428 | ||
cbfe20f5 DB |
429 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
430 | { | |
431 | return do_faccessat(dfd, filename, mode); | |
432 | } | |
433 | ||
ca013e94 | 434 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d | 435 | { |
cbfe20f5 | 436 | return do_faccessat(AT_FDCWD, filename, mode); |
5590ff0d UD |
437 | } |
438 | ||
447016e9 | 439 | int ksys_chdir(const char __user *filename) |
1da177e4 | 440 | { |
2d8f3038 | 441 | struct path path; |
1da177e4 | 442 | int error; |
0291c0a5 JL |
443 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
444 | retry: | |
445 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
446 | if (error) |
447 | goto out; | |
448 | ||
9cfcac81 | 449 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
450 | if (error) |
451 | goto dput_and_out; | |
452 | ||
2d8f3038 | 453 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
454 | |
455 | dput_and_out: | |
2d8f3038 | 456 | path_put(&path); |
0291c0a5 JL |
457 | if (retry_estale(error, lookup_flags)) { |
458 | lookup_flags |= LOOKUP_REVAL; | |
459 | goto retry; | |
460 | } | |
1da177e4 LT |
461 | out: |
462 | return error; | |
463 | } | |
464 | ||
447016e9 DB |
465 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
466 | { | |
467 | return ksys_chdir(filename); | |
468 | } | |
469 | ||
3cdad428 | 470 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 471 | { |
2903ff01 | 472 | struct fd f = fdget_raw(fd); |
159b0956 | 473 | int error; |
1da177e4 LT |
474 | |
475 | error = -EBADF; | |
2903ff01 | 476 | if (!f.file) |
1da177e4 LT |
477 | goto out; |
478 | ||
1da177e4 | 479 | error = -ENOTDIR; |
159b0956 | 480 | if (!d_can_lookup(f.file->f_path.dentry)) |
1da177e4 LT |
481 | goto out_putf; |
482 | ||
159b0956 | 483 | error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR); |
1da177e4 | 484 | if (!error) |
2903ff01 | 485 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 486 | out_putf: |
2903ff01 | 487 | fdput(f); |
1da177e4 LT |
488 | out: |
489 | return error; | |
490 | } | |
491 | ||
a16fe33a | 492 | int ksys_chroot(const char __user *filename) |
1da177e4 | 493 | { |
2d8f3038 | 494 | struct path path; |
1da177e4 | 495 | int error; |
2771261e JL |
496 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
497 | retry: | |
498 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
499 | if (error) |
500 | goto out; | |
501 | ||
9cfcac81 | 502 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
503 | if (error) |
504 | goto dput_and_out; | |
505 | ||
506 | error = -EPERM; | |
c7b96acf | 507 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 508 | goto dput_and_out; |
8b8efb44 TH |
509 | error = security_path_chroot(&path); |
510 | if (error) | |
511 | goto dput_and_out; | |
1da177e4 | 512 | |
2d8f3038 | 513 | set_fs_root(current->fs, &path); |
1da177e4 LT |
514 | error = 0; |
515 | dput_and_out: | |
2d8f3038 | 516 | path_put(&path); |
2771261e JL |
517 | if (retry_estale(error, lookup_flags)) { |
518 | lookup_flags |= LOOKUP_REVAL; | |
519 | goto retry; | |
520 | } | |
1da177e4 LT |
521 | out: |
522 | return error; | |
523 | } | |
524 | ||
a16fe33a DB |
525 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
526 | { | |
527 | return ksys_chroot(filename); | |
528 | } | |
529 | ||
be01f9f2 | 530 | static int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 531 | { |
e57712eb | 532 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 533 | struct inode *delegated_inode = NULL; |
1da177e4 | 534 | struct iattr newattrs; |
e57712eb | 535 | int error; |
1da177e4 | 536 | |
e57712eb AV |
537 | error = mnt_want_write(path->mnt); |
538 | if (error) | |
539 | return error; | |
27ac0ffe | 540 | retry_deleg: |
5955102c | 541 | inode_lock(inode); |
cdcf116d | 542 | error = security_path_chmod(path, mode); |
e57712eb | 543 | if (error) |
fe542cf5 | 544 | goto out_unlock; |
1da177e4 LT |
545 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
546 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 547 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 548 | out_unlock: |
5955102c | 549 | inode_unlock(inode); |
27ac0ffe BF |
550 | if (delegated_inode) { |
551 | error = break_deleg_wait(&delegated_inode); | |
552 | if (!error) | |
553 | goto retry_deleg; | |
554 | } | |
e57712eb AV |
555 | mnt_drop_write(path->mnt); |
556 | return error; | |
557 | } | |
558 | ||
03450e27 | 559 | int ksys_fchmod(unsigned int fd, umode_t mode) |
e57712eb | 560 | { |
173c8401 | 561 | struct fd f = fdget(fd); |
e57712eb AV |
562 | int err = -EBADF; |
563 | ||
173c8401 | 564 | if (f.file) { |
9f45f5bf | 565 | audit_file(f.file); |
173c8401 AV |
566 | err = chmod_common(&f.file->f_path, mode); |
567 | fdput(f); | |
e57712eb | 568 | } |
1da177e4 LT |
569 | return err; |
570 | } | |
571 | ||
03450e27 DB |
572 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
573 | { | |
574 | return ksys_fchmod(fd, mode); | |
575 | } | |
576 | ||
577 | int do_fchmodat(int dfd, const char __user *filename, umode_t mode) | |
1da177e4 | 578 | { |
2d8f3038 | 579 | struct path path; |
1da177e4 | 580 | int error; |
14ff690c JL |
581 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
582 | retry: | |
583 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
584 | if (!error) { |
585 | error = chmod_common(&path, mode); | |
586 | path_put(&path); | |
14ff690c JL |
587 | if (retry_estale(error, lookup_flags)) { |
588 | lookup_flags |= LOOKUP_REVAL; | |
589 | goto retry; | |
590 | } | |
e57712eb | 591 | } |
1da177e4 LT |
592 | return error; |
593 | } | |
594 | ||
03450e27 DB |
595 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, |
596 | umode_t, mode) | |
597 | { | |
598 | return do_fchmodat(dfd, filename, mode); | |
599 | } | |
600 | ||
49f0a076 | 601 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d | 602 | { |
03450e27 | 603 | return do_fchmodat(AT_FDCWD, filename, mode); |
5590ff0d UD |
604 | } |
605 | ||
7fd25dac | 606 | static int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 607 | { |
fe542cf5 | 608 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 609 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
610 | int error; |
611 | struct iattr newattrs; | |
52137abe EB |
612 | kuid_t uid; |
613 | kgid_t gid; | |
614 | ||
615 | uid = make_kuid(current_user_ns(), user); | |
616 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 617 | |
c1b8940b | 618 | retry_deleg: |
1da177e4 LT |
619 | newattrs.ia_valid = ATTR_CTIME; |
620 | if (user != (uid_t) -1) { | |
52137abe EB |
621 | if (!uid_valid(uid)) |
622 | return -EINVAL; | |
1da177e4 | 623 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 624 | newattrs.ia_uid = uid; |
1da177e4 LT |
625 | } |
626 | if (group != (gid_t) -1) { | |
52137abe EB |
627 | if (!gid_valid(gid)) |
628 | return -EINVAL; | |
1da177e4 | 629 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 630 | newattrs.ia_gid = gid; |
1da177e4 LT |
631 | } |
632 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
633 | newattrs.ia_valid |= |
634 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
5955102c | 635 | inode_lock(inode); |
d2b31ca6 | 636 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 637 | if (!error) |
27ac0ffe | 638 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
5955102c | 639 | inode_unlock(inode); |
27ac0ffe BF |
640 | if (delegated_inode) { |
641 | error = break_deleg_wait(&delegated_inode); | |
642 | if (!error) | |
643 | goto retry_deleg; | |
644 | } | |
1da177e4 LT |
645 | return error; |
646 | } | |
647 | ||
55731b3c DB |
648 | int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group, |
649 | int flag) | |
5590ff0d | 650 | { |
2d8f3038 | 651 | struct path path; |
5590ff0d | 652 | int error = -EINVAL; |
65cfc672 | 653 | int lookup_flags; |
5590ff0d | 654 | |
65cfc672 | 655 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
656 | goto out; |
657 | ||
65cfc672 AV |
658 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
659 | if (flag & AT_EMPTY_PATH) | |
660 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 661 | retry: |
65cfc672 | 662 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
663 | if (error) |
664 | goto out; | |
2d8f3038 | 665 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
666 | if (error) |
667 | goto out_release; | |
fe542cf5 | 668 | error = chown_common(&path, user, group); |
2d8f3038 | 669 | mnt_drop_write(path.mnt); |
2af482a7 | 670 | out_release: |
2d8f3038 | 671 | path_put(&path); |
99a5df37 JL |
672 | if (retry_estale(error, lookup_flags)) { |
673 | lookup_flags |= LOOKUP_REVAL; | |
674 | goto retry; | |
675 | } | |
5590ff0d UD |
676 | out: |
677 | return error; | |
678 | } | |
679 | ||
55731b3c DB |
680 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
681 | gid_t, group, int, flag) | |
682 | { | |
683 | return do_fchownat(dfd, filename, user, group, flag); | |
684 | } | |
685 | ||
55e4def0 | 686 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 687 | { |
55731b3c | 688 | return do_fchownat(AT_FDCWD, filename, user, group, 0); |
55e4def0 | 689 | } |
1da177e4 | 690 | |
55e4def0 DH |
691 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
692 | { | |
55731b3c DB |
693 | return do_fchownat(AT_FDCWD, filename, user, group, |
694 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
695 | } |
696 | ||
55731b3c | 697 | int ksys_fchown(unsigned int fd, uid_t user, gid_t group) |
1da177e4 | 698 | { |
2903ff01 | 699 | struct fd f = fdget(fd); |
1da177e4 LT |
700 | int error = -EBADF; |
701 | ||
2903ff01 | 702 | if (!f.file) |
6902d925 DH |
703 | goto out; |
704 | ||
7c6893e3 | 705 | error = mnt_want_write_file_path(f.file); |
2af482a7 DH |
706 | if (error) |
707 | goto out_fput; | |
9f45f5bf | 708 | audit_file(f.file); |
2903ff01 | 709 | error = chown_common(&f.file->f_path, user, group); |
7c6893e3 | 710 | mnt_drop_write_file_path(f.file); |
2af482a7 | 711 | out_fput: |
2903ff01 | 712 | fdput(f); |
6902d925 | 713 | out: |
1da177e4 LT |
714 | return error; |
715 | } | |
716 | ||
55731b3c DB |
717 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
718 | { | |
719 | return ksys_fchown(fd, user, group); | |
720 | } | |
721 | ||
90ad1a8e MS |
722 | int open_check_o_direct(struct file *f) |
723 | { | |
724 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
725 | if (f->f_flags & O_DIRECT) { | |
e748dcd0 | 726 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) |
90ad1a8e | 727 | return -EINVAL; |
90ad1a8e MS |
728 | } |
729 | return 0; | |
730 | } | |
731 | ||
02e5180d | 732 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 733 | struct inode *inode, |
96b7e579 AV |
734 | int (*open)(struct inode *, struct file *), |
735 | const struct cred *cred) | |
1da177e4 | 736 | { |
1abf0c71 | 737 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
738 | int error; |
739 | ||
5300990c | 740 | f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK | |
a1a5b3d9 | 741 | FMODE_PREAD | FMODE_PWRITE; |
1abf0c71 | 742 | |
b5bcdda3 | 743 | path_get(&f->f_path); |
4bacc9c9 | 744 | f->f_inode = inode; |
1da177e4 | 745 | f->f_mapping = inode->i_mapping; |
1da177e4 | 746 | |
5660e13d JL |
747 | /* Ensure that we skip any errors that predate opening of the file */ |
748 | f->f_wb_err = filemap_sample_wb_err(f->f_mapping); | |
749 | ||
3f4d5a00 AV |
750 | if (unlikely(f->f_flags & O_PATH)) { |
751 | f->f_mode = FMODE_PATH; | |
1abf0c71 | 752 | f->f_op = &empty_fops; |
96b7e579 | 753 | return 0; |
1abf0c71 AV |
754 | } |
755 | ||
dd20908a | 756 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 757 | error = get_write_access(inode); |
3f4d5a00 | 758 | if (unlikely(error)) |
1da177e4 | 759 | goto cleanup_file; |
0ccb2863 | 760 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 761 | if (unlikely(error)) { |
0ccb2863 AV |
762 | put_write_access(inode); |
763 | goto cleanup_file; | |
764 | } | |
83f936c7 | 765 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
766 | } |
767 | ||
9c225f26 | 768 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
63b6df14 | 769 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) |
9c225f26 LT |
770 | f->f_mode |= FMODE_ATOMIC_POS; |
771 | ||
1abf0c71 | 772 | f->f_op = fops_get(inode->i_fop); |
72c2d531 AV |
773 | if (unlikely(WARN_ON(!f->f_op))) { |
774 | error = -ENODEV; | |
775 | goto cleanup_all; | |
776 | } | |
1abf0c71 | 777 | |
83d49856 | 778 | error = security_file_open(f, cred); |
788e7dd4 YN |
779 | if (error) |
780 | goto cleanup_all; | |
781 | ||
c568d683 | 782 | error = break_lease(locks_inode(f), f->f_flags); |
f3c7691e BF |
783 | if (error) |
784 | goto cleanup_all; | |
785 | ||
72c2d531 | 786 | if (!open) |
834f2a4a TM |
787 | open = f->f_op->open; |
788 | if (open) { | |
789 | error = open(inode, f); | |
1da177e4 LT |
790 | if (error) |
791 | goto cleanup_all; | |
792 | } | |
890275b5 MZ |
793 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
794 | i_readcount_inc(inode); | |
293bc982 | 795 | if ((f->f_mode & FMODE_READ) && |
84363182 | 796 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 797 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 798 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 799 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 800 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 801 | |
c75b1d94 | 802 | f->f_write_hint = WRITE_LIFE_NOT_SET; |
1da177e4 LT |
803 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
804 | ||
805 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
806 | ||
96b7e579 | 807 | return 0; |
1da177e4 LT |
808 | |
809 | cleanup_all: | |
810 | fops_put(f->f_op); | |
83f936c7 | 811 | if (f->f_mode & FMODE_WRITER) { |
1da177e4 | 812 | put_write_access(inode); |
83f936c7 | 813 | __mnt_drop_write(f->f_path.mnt); |
4a3fd211 | 814 | } |
1da177e4 | 815 | cleanup_file: |
02e5180d AV |
816 | path_put(&f->f_path); |
817 | f->f_path.mnt = NULL; | |
818 | f->f_path.dentry = NULL; | |
dd37978c | 819 | f->f_inode = NULL; |
96b7e579 | 820 | return error; |
1da177e4 LT |
821 | } |
822 | ||
d18e9008 MS |
823 | /** |
824 | * finish_open - finish opening a file | |
0854d450 | 825 | * @file: file pointer |
d18e9008 MS |
826 | * @dentry: pointer to dentry |
827 | * @open: open callback | |
0854d450 | 828 | * @opened: state of open |
d18e9008 MS |
829 | * |
830 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
831 | * | |
832 | * If the open callback is set to NULL, then the standard f_op->open() | |
833 | * filesystem callback is substituted. | |
0854d450 MS |
834 | * |
835 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
836 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
837 | * on it after finish_open(). | |
838 | * | |
839 | * On successful return @file is a fully instantiated open file. After this, if | |
840 | * an error occurs in ->atomic_open(), it needs to clean up with fput(). | |
841 | * | |
842 | * Returns zero on success or -errno if the open failed. | |
d18e9008 | 843 | */ |
30d90494 AV |
844 | int finish_open(struct file *file, struct dentry *dentry, |
845 | int (*open)(struct inode *, struct file *), | |
846 | int *opened) | |
d18e9008 | 847 | { |
96b7e579 | 848 | int error; |
3d8a00d2 | 849 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 850 | |
b5bcdda3 | 851 | file->f_path.dentry = dentry; |
4bacc9c9 DH |
852 | error = do_dentry_open(file, d_backing_inode(dentry), open, |
853 | current_cred()); | |
96b7e579 | 854 | if (!error) |
47237687 | 855 | *opened |= FILE_OPENED; |
d18e9008 | 856 | |
96b7e579 | 857 | return error; |
d18e9008 MS |
858 | } |
859 | EXPORT_SYMBOL(finish_open); | |
860 | ||
861 | /** | |
862 | * finish_no_open - finish ->atomic_open() without opening the file | |
863 | * | |
0854d450 | 864 | * @file: file pointer |
d18e9008 MS |
865 | * @dentry: dentry or NULL (as returned from ->lookup()) |
866 | * | |
867 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
868 | * |
869 | * NB: unlike finish_open() this function does consume the dentry reference and | |
870 | * the caller need not dput() it. | |
871 | * | |
872 | * Returns "1" which must be the return value of ->atomic_open() after having | |
873 | * called this function. | |
d18e9008 | 874 | */ |
e45198a6 | 875 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 876 | { |
30d90494 | 877 | file->f_path.dentry = dentry; |
e45198a6 | 878 | return 1; |
d18e9008 MS |
879 | } |
880 | EXPORT_SYMBOL(finish_no_open); | |
881 | ||
9bf39ab2 MS |
882 | char *file_path(struct file *filp, char *buf, int buflen) |
883 | { | |
884 | return d_path(&filp->f_path, buf, buflen); | |
885 | } | |
886 | EXPORT_SYMBOL(file_path); | |
887 | ||
4bacc9c9 DH |
888 | /** |
889 | * vfs_open - open the file at the given path | |
890 | * @path: path to open | |
891 | * @file: newly allocated file with f_flag initialized | |
892 | * @cred: credentials to use | |
893 | */ | |
894 | int vfs_open(const struct path *path, struct file *file, | |
895 | const struct cred *cred) | |
896 | { | |
495e6429 | 897 | struct dentry *dentry = d_real(path->dentry, NULL, file->f_flags, 0); |
4bacc9c9 | 898 | |
2d902671 MS |
899 | if (IS_ERR(dentry)) |
900 | return PTR_ERR(dentry); | |
4bacc9c9 | 901 | |
54d5ca87 | 902 | file->f_path = *path; |
2d902671 | 903 | return do_dentry_open(file, d_backing_inode(dentry), NULL, cred); |
4bacc9c9 DH |
904 | } |
905 | ||
765927b2 | 906 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 907 | const struct cred *cred) |
a1a5b3d9 PS |
908 | { |
909 | int error; | |
910 | struct file *f; | |
911 | ||
e0e81739 DH |
912 | validate_creds(cred); |
913 | ||
c212f9aa | 914 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 915 | BUG_ON(!path->mnt); |
322ee5b3 | 916 | |
a1a5b3d9 | 917 | f = get_empty_filp(); |
1afc99be AV |
918 | if (!IS_ERR(f)) { |
919 | f->f_flags = flags; | |
4aa7c634 | 920 | error = vfs_open(path, f, cred); |
1afc99be AV |
921 | if (!error) { |
922 | /* from now on we need fput() to dispose of f */ | |
923 | error = open_check_o_direct(f); | |
924 | if (error) { | |
925 | fput(f); | |
926 | f = ERR_PTR(error); | |
927 | } | |
928 | } else { | |
929 | put_filp(f); | |
2a027e7a AV |
930 | f = ERR_PTR(error); |
931 | } | |
2a027e7a AV |
932 | } |
933 | return f; | |
a1a5b3d9 | 934 | } |
1da177e4 LT |
935 | EXPORT_SYMBOL(dentry_open); |
936 | ||
a218d0fd | 937 | static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op) |
47c805dc AV |
938 | { |
939 | int lookup_flags = 0; | |
62fb4a15 | 940 | int acc_mode = ACC_MODE(flags); |
47c805dc | 941 | |
629e014b CH |
942 | /* |
943 | * Clear out all open flags we don't know about so that we don't report | |
944 | * them in fcntl(F_GETFD) or similar interfaces. | |
945 | */ | |
946 | flags &= VALID_OPEN_FLAGS; | |
947 | ||
e305f48b | 948 | if (flags & (O_CREAT | __O_TMPFILE)) |
e68726ff MS |
949 | op->mode = (mode & S_IALLUGO) | S_IFREG; |
950 | else | |
951 | op->mode = 0; | |
47c805dc AV |
952 | |
953 | /* Must never be set by userspace */ | |
c6f3d811 | 954 | flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC; |
47c805dc AV |
955 | |
956 | /* | |
957 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
958 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
959 | * always set instead of having to deal with possibly weird behaviour | |
960 | * for malicious applications setting only __O_SYNC. | |
961 | */ | |
962 | if (flags & __O_SYNC) | |
963 | flags |= O_DSYNC; | |
964 | ||
bb458c64 AV |
965 | if (flags & __O_TMPFILE) { |
966 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d | 967 | return -EINVAL; |
ba57ea64 AV |
968 | if (!(acc_mode & MAY_WRITE)) |
969 | return -EINVAL; | |
60545d0d AV |
970 | } else if (flags & O_PATH) { |
971 | /* | |
972 | * If we have O_PATH in the open flag. Then we | |
973 | * cannot have anything other than the below set of flags | |
974 | */ | |
1abf0c71 AV |
975 | flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH; |
976 | acc_mode = 0; | |
1abf0c71 | 977 | } |
47c805dc | 978 | |
1abf0c71 | 979 | op->open_flag = flags; |
47c805dc AV |
980 | |
981 | /* O_TRUNC implies we need access checks for write permissions */ | |
982 | if (flags & O_TRUNC) | |
983 | acc_mode |= MAY_WRITE; | |
984 | ||
985 | /* Allow the LSM permission hook to distinguish append | |
986 | access from general write access. */ | |
987 | if (flags & O_APPEND) | |
988 | acc_mode |= MAY_APPEND; | |
989 | ||
990 | op->acc_mode = acc_mode; | |
991 | ||
1abf0c71 AV |
992 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
993 | ||
47c805dc AV |
994 | if (flags & O_CREAT) { |
995 | op->intent |= LOOKUP_CREATE; | |
996 | if (flags & O_EXCL) | |
997 | op->intent |= LOOKUP_EXCL; | |
998 | } | |
999 | ||
1000 | if (flags & O_DIRECTORY) | |
1001 | lookup_flags |= LOOKUP_DIRECTORY; | |
1002 | if (!(flags & O_NOFOLLOW)) | |
1003 | lookup_flags |= LOOKUP_FOLLOW; | |
f9652e10 AV |
1004 | op->lookup_flags = lookup_flags; |
1005 | return 0; | |
47c805dc AV |
1006 | } |
1007 | ||
669abf4e JL |
1008 | /** |
1009 | * file_open_name - open file and return file pointer | |
1010 | * | |
1011 | * @name: struct filename containing path to open | |
1012 | * @flags: open flags as per the open(2) second argument | |
1013 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1014 | * | |
1015 | * This is the helper to open a file from kernelspace if you really | |
1016 | * have to. But in generally you should not do this, so please move | |
1017 | * along, nothing to see here.. | |
1018 | */ | |
1019 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
1020 | { | |
1021 | struct open_flags op; | |
f9652e10 AV |
1022 | int err = build_open_flags(flags, mode, &op); |
1023 | return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
1024 | } |
1025 | ||
47c805dc AV |
1026 | /** |
1027 | * filp_open - open file and return file pointer | |
1028 | * | |
1029 | * @filename: path to open | |
1030 | * @flags: open flags as per the open(2) second argument | |
1031 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1032 | * | |
1033 | * This is the helper to open a file from kernelspace if you really | |
1034 | * have to. But in generally you should not do this, so please move | |
1035 | * along, nothing to see here.. | |
1036 | */ | |
a218d0fd | 1037 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 1038 | { |
51689104 PM |
1039 | struct filename *name = getname_kernel(filename); |
1040 | struct file *file = ERR_CAST(name); | |
1041 | ||
1042 | if (!IS_ERR(name)) { | |
1043 | file = file_open_name(name, flags, mode); | |
1044 | putname(name); | |
1045 | } | |
1046 | return file; | |
47c805dc AV |
1047 | } |
1048 | EXPORT_SYMBOL(filp_open); | |
1049 | ||
73d049a4 | 1050 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
378c6520 | 1051 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1052 | { |
1053 | struct open_flags op; | |
378c6520 | 1054 | int err = build_open_flags(flags, mode, &op); |
f9652e10 AV |
1055 | if (err) |
1056 | return ERR_PTR(err); | |
f9652e10 | 1057 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
1058 | } |
1059 | EXPORT_SYMBOL(file_open_root); | |
1060 | ||
9a08c352 JB |
1061 | struct file *filp_clone_open(struct file *oldfile) |
1062 | { | |
1063 | struct file *file; | |
1064 | int retval; | |
1065 | ||
1066 | file = get_empty_filp(); | |
1067 | if (IS_ERR(file)) | |
1068 | return file; | |
1069 | ||
1070 | file->f_flags = oldfile->f_flags; | |
1071 | retval = vfs_open(&oldfile->f_path, file, oldfile->f_cred); | |
1072 | if (retval) { | |
1073 | put_filp(file); | |
1074 | return ERR_PTR(retval); | |
1075 | } | |
1076 | ||
1077 | return file; | |
1078 | } | |
1079 | EXPORT_SYMBOL(filp_clone_open); | |
1080 | ||
a218d0fd | 1081 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
1da177e4 | 1082 | { |
47c805dc | 1083 | struct open_flags op; |
f9652e10 AV |
1084 | int fd = build_open_flags(flags, mode, &op); |
1085 | struct filename *tmp; | |
1086 | ||
1087 | if (fd) | |
1088 | return fd; | |
1089 | ||
1090 | tmp = getname(filename); | |
1091 | if (IS_ERR(tmp)) | |
1092 | return PTR_ERR(tmp); | |
1093 | ||
1094 | fd = get_unused_fd_flags(flags); | |
1095 | if (fd >= 0) { | |
1096 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1097 | if (IS_ERR(f)) { | |
1098 | put_unused_fd(fd); | |
1099 | fd = PTR_ERR(f); | |
1100 | } else { | |
1101 | fsnotify_open(f); | |
1102 | fd_install(fd, f); | |
1da177e4 | 1103 | } |
1da177e4 | 1104 | } |
f9652e10 | 1105 | putname(tmp); |
1da177e4 | 1106 | return fd; |
1da177e4 | 1107 | } |
e922efc3 | 1108 | |
a218d0fd | 1109 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) |
e922efc3 MS |
1110 | { |
1111 | if (force_o_largefile()) | |
1112 | flags |= O_LARGEFILE; | |
1113 | ||
2cf09666 | 1114 | return do_sys_open(AT_FDCWD, filename, flags, mode); |
e922efc3 | 1115 | } |
1da177e4 | 1116 | |
6559eed8 | 1117 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1118 | umode_t, mode) |
5590ff0d UD |
1119 | { |
1120 | if (force_o_largefile()) | |
1121 | flags |= O_LARGEFILE; | |
1122 | ||
2cf09666 | 1123 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1124 | } |
5590ff0d | 1125 | |
e35d49f6 AV |
1126 | #ifdef CONFIG_COMPAT |
1127 | /* | |
1128 | * Exactly like sys_open(), except that it doesn't set the | |
1129 | * O_LARGEFILE flag. | |
1130 | */ | |
1131 | COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1132 | { | |
1133 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
1134 | } | |
1135 | ||
1136 | /* | |
1137 | * Exactly like sys_openat(), except that it doesn't set the | |
1138 | * O_LARGEFILE flag. | |
1139 | */ | |
1140 | COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode) | |
1141 | { | |
1142 | return do_sys_open(dfd, filename, flags, mode); | |
1143 | } | |
1144 | #endif | |
1145 | ||
1da177e4 LT |
1146 | #ifndef __alpha__ |
1147 | ||
1148 | /* | |
1149 | * For backward compatibility? Maybe this should be moved | |
1150 | * into arch/i386 instead? | |
1151 | */ | |
a218d0fd | 1152 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 | 1153 | { |
bae217ea | 1154 | return ksys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); |
1da177e4 LT |
1155 | } |
1156 | ||
1157 | #endif | |
1158 | ||
1159 | /* | |
1160 | * "id" is the POSIX thread ID. We use the | |
1161 | * files pointer for this.. | |
1162 | */ | |
1163 | int filp_close(struct file *filp, fl_owner_t id) | |
1164 | { | |
45778ca8 | 1165 | int retval = 0; |
1da177e4 LT |
1166 | |
1167 | if (!file_count(filp)) { | |
1168 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1169 | return 0; |
1da177e4 LT |
1170 | } |
1171 | ||
72c2d531 | 1172 | if (filp->f_op->flush) |
75e1fcc0 | 1173 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1174 | |
1abf0c71 AV |
1175 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1176 | dnotify_flush(filp, id); | |
1177 | locks_remove_posix(filp, id); | |
1178 | } | |
1da177e4 LT |
1179 | fput(filp); |
1180 | return retval; | |
1181 | } | |
1182 | ||
1183 | EXPORT_SYMBOL(filp_close); | |
1184 | ||
1185 | /* | |
1186 | * Careful here! We test whether the file pointer is NULL before | |
1187 | * releasing the fd. This ensures that one clone task can't release | |
1188 | * an fd while another clone is opening it. | |
1189 | */ | |
ca013e94 | 1190 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1191 | { |
483ce1d4 | 1192 | int retval = __close_fd(current->files, fd); |
ee731f4f EP |
1193 | |
1194 | /* can't restart close syscall because file table entry was cleared */ | |
1195 | if (unlikely(retval == -ERESTARTSYS || | |
1196 | retval == -ERESTARTNOINTR || | |
1197 | retval == -ERESTARTNOHAND || | |
1198 | retval == -ERESTART_RESTARTBLOCK)) | |
1199 | retval = -EINTR; | |
1200 | ||
1201 | return retval; | |
1da177e4 | 1202 | } |
1da177e4 LT |
1203 | |
1204 | /* | |
1205 | * This routine simulates a hangup on the tty, to arrange that users | |
1206 | * are given clean terminals at login time. | |
1207 | */ | |
ca013e94 | 1208 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1209 | { |
1210 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1211 | tty_vhangup_self(); |
1da177e4 LT |
1212 | return 0; |
1213 | } | |
1214 | return -EPERM; | |
1215 | } | |
1216 | ||
1217 | /* | |
1218 | * Called when an inode is about to be open. | |
1219 | * We use this to disallow opening large files on 32bit systems if | |
1220 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1221 | * on this flag in sys_open. | |
1222 | */ | |
1223 | int generic_file_open(struct inode * inode, struct file * filp) | |
1224 | { | |
1225 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1226 | return -EOVERFLOW; |
1da177e4 LT |
1227 | return 0; |
1228 | } | |
1229 | ||
1230 | EXPORT_SYMBOL(generic_file_open); | |
1231 | ||
1232 | /* | |
1233 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1234 | * file descriptors. The function is not supposed to ever fail, the only |
1235 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1236 | * directly into file_operations structure. | |
1da177e4 LT |
1237 | */ |
1238 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1239 | { | |
1240 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1241 | return 0; | |
1242 | } | |
1243 | ||
1244 | EXPORT_SYMBOL(nonseekable_open); |