]>
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; |
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 | */ | |
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: |
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 | ||
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 | ||
71be6b49 DW |
270 | /* Unshare range should only be used with allocate mode. */ |
271 | if ((mode & FALLOC_FL_UNSHARE_RANGE) && | |
272 | (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE))) | |
273 | return -EINVAL; | |
274 | ||
97ac7350 | 275 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 276 | return -EBADF; |
1ca551c6 | 277 | |
00f5e619 | 278 | /* |
8fc61d92 | 279 | * We can only allow pure fallocate on append only files |
00f5e619 | 280 | */ |
8fc61d92 | 281 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
282 | return -EPERM; |
283 | ||
284 | if (IS_IMMUTABLE(inode)) | |
285 | return -EPERM; | |
286 | ||
0790b31b | 287 | /* |
6d2b6170 | 288 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
289 | */ |
290 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 291 | return -ETXTBSY; |
0790b31b | 292 | |
97ac7350 AA |
293 | /* |
294 | * Revalidate the write permissions, in case security policy has | |
295 | * changed since the files were opened. | |
296 | */ | |
297 | ret = security_file_permission(file, MAY_WRITE); | |
298 | if (ret) | |
3e63cbb1 | 299 | return ret; |
97ac7350 | 300 | |
97ac7350 | 301 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 302 | return -ESPIPE; |
97ac7350 | 303 | |
97ac7350 AA |
304 | /* |
305 | * Let individual file system decide if it supports preallocation | |
306 | * for directories or not. | |
307 | */ | |
25f4c414 DW |
308 | if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode) && |
309 | !S_ISBLK(inode->i_mode)) | |
3e63cbb1 | 310 | return -ENODEV; |
97ac7350 | 311 | |
97ac7350 AA |
312 | /* Check for wrap through zero too */ |
313 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 314 | return -EFBIG; |
97ac7350 | 315 | |
2fe17c10 | 316 | if (!file->f_op->fallocate) |
3e63cbb1 | 317 | return -EOPNOTSUPP; |
97ac7350 | 318 | |
14da9200 JK |
319 | sb_start_write(inode->i_sb); |
320 | ret = file->f_op->fallocate(file, mode, offset, len); | |
820c12d5 HS |
321 | |
322 | /* | |
323 | * Create inotify and fanotify events. | |
324 | * | |
325 | * To keep the logic simple always create events if fallocate succeeds. | |
326 | * This implies that events are even created if the file size remains | |
327 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
328 | */ | |
329 | if (ret == 0) | |
330 | fsnotify_modify(file); | |
331 | ||
14da9200 JK |
332 | sb_end_write(inode->i_sb); |
333 | return ret; | |
3e63cbb1 | 334 | } |
72c72bdf | 335 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 336 | |
4a0fd5bf | 337 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
3e63cbb1 | 338 | { |
2903ff01 | 339 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
340 | int error = -EBADF; |
341 | ||
2903ff01 | 342 | if (f.file) { |
72c72bdf | 343 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 344 | fdput(f); |
3e63cbb1 | 345 | } |
3e63cbb1 | 346 | return error; |
97ac7350 | 347 | } |
3e63cbb1 | 348 | |
1da177e4 LT |
349 | /* |
350 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
351 | * We do this by temporarily clearing all FS-related capabilities and | |
352 | * switching the fsuid/fsgid around to the real ones. | |
353 | */ | |
6559eed8 | 354 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
1da177e4 | 355 | { |
d84f4f99 DH |
356 | const struct cred *old_cred; |
357 | struct cred *override_cred; | |
2d8f3038 | 358 | struct path path; |
256984a8 | 359 | struct inode *inode; |
1da177e4 | 360 | int res; |
87fa5595 | 361 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
1da177e4 LT |
362 | |
363 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
364 | return -EINVAL; | |
365 | ||
d84f4f99 DH |
366 | override_cred = prepare_creds(); |
367 | if (!override_cred) | |
368 | return -ENOMEM; | |
1da177e4 | 369 | |
d84f4f99 DH |
370 | override_cred->fsuid = override_cred->uid; |
371 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 372 | |
086f7316 | 373 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 374 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
375 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
376 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 377 | cap_clear(override_cred->cap_effective); |
086f7316 | 378 | else |
d84f4f99 DH |
379 | override_cred->cap_effective = |
380 | override_cred->cap_permitted; | |
086f7316 | 381 | } |
1da177e4 | 382 | |
d84f4f99 | 383 | old_cred = override_creds(override_cred); |
87fa5595 JL |
384 | retry: |
385 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
386 | if (res) |
387 | goto out; | |
388 | ||
63afdfc7 | 389 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
390 | |
391 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
392 | /* |
393 | * MAY_EXEC on regular files is denied if the fs is mounted | |
394 | * with the "noexec" flag. | |
395 | */ | |
396 | res = -EACCES; | |
90f8572b | 397 | if (path_noexec(&path)) |
30524472 AV |
398 | goto out_path_release; |
399 | } | |
400 | ||
256984a8 | 401 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 402 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 403 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 404 | goto out_path_release; |
2f676cbc DH |
405 | /* |
406 | * This is a rare case where using __mnt_is_readonly() | |
407 | * is OK without a mnt_want/drop_write() pair. Since | |
408 | * no actual write to the fs is performed here, we do | |
409 | * not need to telegraph to that to anyone. | |
410 | * | |
411 | * By doing this, we accept that this access is | |
412 | * inherently racy and know that the fs may change | |
413 | * state before we even see this result. | |
414 | */ | |
2d8f3038 | 415 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 416 | res = -EROFS; |
1da177e4 | 417 | |
6902d925 | 418 | out_path_release: |
2d8f3038 | 419 | path_put(&path); |
87fa5595 JL |
420 | if (retry_estale(res, lookup_flags)) { |
421 | lookup_flags |= LOOKUP_REVAL; | |
422 | goto retry; | |
423 | } | |
6902d925 | 424 | out: |
d84f4f99 DH |
425 | revert_creds(old_cred); |
426 | put_cred(override_cred); | |
1da177e4 LT |
427 | return res; |
428 | } | |
429 | ||
ca013e94 | 430 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d UD |
431 | { |
432 | return sys_faccessat(AT_FDCWD, filename, mode); | |
433 | } | |
434 | ||
3cdad428 | 435 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 436 | { |
2d8f3038 | 437 | struct path path; |
1da177e4 | 438 | int error; |
0291c0a5 JL |
439 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
440 | retry: | |
441 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
442 | if (error) |
443 | goto out; | |
444 | ||
9cfcac81 | 445 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
446 | if (error) |
447 | goto dput_and_out; | |
448 | ||
2d8f3038 | 449 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
450 | |
451 | dput_and_out: | |
2d8f3038 | 452 | path_put(&path); |
0291c0a5 JL |
453 | if (retry_estale(error, lookup_flags)) { |
454 | lookup_flags |= LOOKUP_REVAL; | |
455 | goto retry; | |
456 | } | |
1da177e4 LT |
457 | out: |
458 | return error; | |
459 | } | |
460 | ||
3cdad428 | 461 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 462 | { |
2903ff01 | 463 | struct fd f = fdget_raw(fd); |
1da177e4 | 464 | struct inode *inode; |
2903ff01 | 465 | int error = -EBADF; |
1da177e4 LT |
466 | |
467 | error = -EBADF; | |
2903ff01 | 468 | if (!f.file) |
1da177e4 LT |
469 | goto out; |
470 | ||
496ad9aa | 471 | inode = file_inode(f.file); |
1da177e4 LT |
472 | |
473 | error = -ENOTDIR; | |
474 | if (!S_ISDIR(inode->i_mode)) | |
475 | goto out_putf; | |
476 | ||
9cfcac81 | 477 | error = inode_permission(inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 | 478 | if (!error) |
2903ff01 | 479 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 480 | out_putf: |
2903ff01 | 481 | fdput(f); |
1da177e4 LT |
482 | out: |
483 | return error; | |
484 | } | |
485 | ||
3480b257 | 486 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 487 | { |
2d8f3038 | 488 | struct path path; |
1da177e4 | 489 | int error; |
2771261e JL |
490 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
491 | retry: | |
492 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
493 | if (error) |
494 | goto out; | |
495 | ||
9cfcac81 | 496 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
497 | if (error) |
498 | goto dput_and_out; | |
499 | ||
500 | error = -EPERM; | |
c7b96acf | 501 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 502 | goto dput_and_out; |
8b8efb44 TH |
503 | error = security_path_chroot(&path); |
504 | if (error) | |
505 | goto dput_and_out; | |
1da177e4 | 506 | |
2d8f3038 | 507 | set_fs_root(current->fs, &path); |
1da177e4 LT |
508 | error = 0; |
509 | dput_and_out: | |
2d8f3038 | 510 | path_put(&path); |
2771261e JL |
511 | if (retry_estale(error, lookup_flags)) { |
512 | lookup_flags |= LOOKUP_REVAL; | |
513 | goto retry; | |
514 | } | |
1da177e4 LT |
515 | out: |
516 | return error; | |
517 | } | |
518 | ||
be01f9f2 | 519 | static int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 520 | { |
e57712eb | 521 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 522 | struct inode *delegated_inode = NULL; |
1da177e4 | 523 | struct iattr newattrs; |
e57712eb | 524 | int error; |
1da177e4 | 525 | |
e57712eb AV |
526 | error = mnt_want_write(path->mnt); |
527 | if (error) | |
528 | return error; | |
27ac0ffe | 529 | retry_deleg: |
5955102c | 530 | inode_lock(inode); |
cdcf116d | 531 | error = security_path_chmod(path, mode); |
e57712eb | 532 | if (error) |
fe542cf5 | 533 | goto out_unlock; |
1da177e4 LT |
534 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
535 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 536 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 537 | out_unlock: |
5955102c | 538 | inode_unlock(inode); |
27ac0ffe BF |
539 | if (delegated_inode) { |
540 | error = break_deleg_wait(&delegated_inode); | |
541 | if (!error) | |
542 | goto retry_deleg; | |
543 | } | |
e57712eb AV |
544 | mnt_drop_write(path->mnt); |
545 | return error; | |
546 | } | |
547 | ||
49f0a076 | 548 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 549 | { |
173c8401 | 550 | struct fd f = fdget(fd); |
e57712eb AV |
551 | int err = -EBADF; |
552 | ||
173c8401 | 553 | if (f.file) { |
9f45f5bf | 554 | audit_file(f.file); |
173c8401 AV |
555 | err = chmod_common(&f.file->f_path, mode); |
556 | fdput(f); | |
e57712eb | 557 | } |
1da177e4 LT |
558 | return err; |
559 | } | |
560 | ||
49f0a076 | 561 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode) |
1da177e4 | 562 | { |
2d8f3038 | 563 | struct path path; |
1da177e4 | 564 | int error; |
14ff690c JL |
565 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
566 | retry: | |
567 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
568 | if (!error) { |
569 | error = chmod_common(&path, mode); | |
570 | path_put(&path); | |
14ff690c JL |
571 | if (retry_estale(error, lookup_flags)) { |
572 | lookup_flags |= LOOKUP_REVAL; | |
573 | goto retry; | |
574 | } | |
e57712eb | 575 | } |
1da177e4 LT |
576 | return error; |
577 | } | |
578 | ||
49f0a076 | 579 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d UD |
580 | { |
581 | return sys_fchmodat(AT_FDCWD, filename, mode); | |
582 | } | |
583 | ||
7fd25dac | 584 | static int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 585 | { |
fe542cf5 | 586 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 587 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
588 | int error; |
589 | struct iattr newattrs; | |
52137abe EB |
590 | kuid_t uid; |
591 | kgid_t gid; | |
592 | ||
593 | uid = make_kuid(current_user_ns(), user); | |
594 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 595 | |
c1b8940b | 596 | retry_deleg: |
1da177e4 LT |
597 | newattrs.ia_valid = ATTR_CTIME; |
598 | if (user != (uid_t) -1) { | |
52137abe EB |
599 | if (!uid_valid(uid)) |
600 | return -EINVAL; | |
1da177e4 | 601 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 602 | newattrs.ia_uid = uid; |
1da177e4 LT |
603 | } |
604 | if (group != (gid_t) -1) { | |
52137abe EB |
605 | if (!gid_valid(gid)) |
606 | return -EINVAL; | |
1da177e4 | 607 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 608 | newattrs.ia_gid = gid; |
1da177e4 LT |
609 | } |
610 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
611 | newattrs.ia_valid |= |
612 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
5955102c | 613 | inode_lock(inode); |
d2b31ca6 | 614 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 615 | if (!error) |
27ac0ffe | 616 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
5955102c | 617 | inode_unlock(inode); |
27ac0ffe BF |
618 | if (delegated_inode) { |
619 | error = break_deleg_wait(&delegated_inode); | |
620 | if (!error) | |
621 | goto retry_deleg; | |
622 | } | |
1da177e4 LT |
623 | return error; |
624 | } | |
625 | ||
6559eed8 HC |
626 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
627 | gid_t, group, int, flag) | |
5590ff0d | 628 | { |
2d8f3038 | 629 | struct path path; |
5590ff0d | 630 | int error = -EINVAL; |
65cfc672 | 631 | int lookup_flags; |
5590ff0d | 632 | |
65cfc672 | 633 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
634 | goto out; |
635 | ||
65cfc672 AV |
636 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
637 | if (flag & AT_EMPTY_PATH) | |
638 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 639 | retry: |
65cfc672 | 640 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
641 | if (error) |
642 | goto out; | |
2d8f3038 | 643 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
644 | if (error) |
645 | goto out_release; | |
fe542cf5 | 646 | error = chown_common(&path, user, group); |
2d8f3038 | 647 | mnt_drop_write(path.mnt); |
2af482a7 | 648 | out_release: |
2d8f3038 | 649 | path_put(&path); |
99a5df37 JL |
650 | if (retry_estale(error, lookup_flags)) { |
651 | lookup_flags |= LOOKUP_REVAL; | |
652 | goto retry; | |
653 | } | |
5590ff0d UD |
654 | out: |
655 | return error; | |
656 | } | |
657 | ||
55e4def0 | 658 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 659 | { |
55e4def0 DH |
660 | return sys_fchownat(AT_FDCWD, filename, user, group, 0); |
661 | } | |
1da177e4 | 662 | |
55e4def0 DH |
663 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
664 | { | |
665 | return sys_fchownat(AT_FDCWD, filename, user, group, | |
666 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
667 | } |
668 | ||
ca013e94 | 669 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
1da177e4 | 670 | { |
2903ff01 | 671 | struct fd f = fdget(fd); |
1da177e4 LT |
672 | int error = -EBADF; |
673 | ||
2903ff01 | 674 | if (!f.file) |
6902d925 DH |
675 | goto out; |
676 | ||
2903ff01 | 677 | error = mnt_want_write_file(f.file); |
2af482a7 DH |
678 | if (error) |
679 | goto out_fput; | |
9f45f5bf | 680 | audit_file(f.file); |
2903ff01 AV |
681 | error = chown_common(&f.file->f_path, user, group); |
682 | mnt_drop_write_file(f.file); | |
2af482a7 | 683 | out_fput: |
2903ff01 | 684 | fdput(f); |
6902d925 | 685 | out: |
1da177e4 LT |
686 | return error; |
687 | } | |
688 | ||
90ad1a8e MS |
689 | int open_check_o_direct(struct file *f) |
690 | { | |
691 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
692 | if (f->f_flags & O_DIRECT) { | |
e748dcd0 | 693 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) |
90ad1a8e | 694 | return -EINVAL; |
90ad1a8e MS |
695 | } |
696 | return 0; | |
697 | } | |
698 | ||
02e5180d | 699 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 700 | struct inode *inode, |
96b7e579 AV |
701 | int (*open)(struct inode *, struct file *), |
702 | const struct cred *cred) | |
1da177e4 | 703 | { |
1abf0c71 | 704 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
705 | int error; |
706 | ||
5300990c | 707 | f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK | |
a1a5b3d9 | 708 | FMODE_PREAD | FMODE_PWRITE; |
1abf0c71 | 709 | |
b5bcdda3 | 710 | path_get(&f->f_path); |
4bacc9c9 | 711 | f->f_inode = inode; |
1da177e4 | 712 | f->f_mapping = inode->i_mapping; |
1da177e4 | 713 | |
3f4d5a00 AV |
714 | if (unlikely(f->f_flags & O_PATH)) { |
715 | f->f_mode = FMODE_PATH; | |
1abf0c71 | 716 | f->f_op = &empty_fops; |
96b7e579 | 717 | return 0; |
1abf0c71 AV |
718 | } |
719 | ||
dd20908a | 720 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 721 | error = get_write_access(inode); |
3f4d5a00 | 722 | if (unlikely(error)) |
1da177e4 | 723 | goto cleanup_file; |
0ccb2863 | 724 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 725 | if (unlikely(error)) { |
0ccb2863 AV |
726 | put_write_access(inode); |
727 | goto cleanup_file; | |
728 | } | |
83f936c7 | 729 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
730 | } |
731 | ||
9c225f26 | 732 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
63b6df14 | 733 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) |
9c225f26 LT |
734 | f->f_mode |= FMODE_ATOMIC_POS; |
735 | ||
1abf0c71 | 736 | f->f_op = fops_get(inode->i_fop); |
72c2d531 AV |
737 | if (unlikely(WARN_ON(!f->f_op))) { |
738 | error = -ENODEV; | |
739 | goto cleanup_all; | |
740 | } | |
1abf0c71 | 741 | |
83d49856 | 742 | error = security_file_open(f, cred); |
788e7dd4 YN |
743 | if (error) |
744 | goto cleanup_all; | |
745 | ||
c568d683 | 746 | error = break_lease(locks_inode(f), f->f_flags); |
f3c7691e BF |
747 | if (error) |
748 | goto cleanup_all; | |
749 | ||
72c2d531 | 750 | if (!open) |
834f2a4a TM |
751 | open = f->f_op->open; |
752 | if (open) { | |
753 | error = open(inode, f); | |
1da177e4 LT |
754 | if (error) |
755 | goto cleanup_all; | |
756 | } | |
890275b5 MZ |
757 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
758 | i_readcount_inc(inode); | |
293bc982 | 759 | if ((f->f_mode & FMODE_READ) && |
84363182 | 760 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 761 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 762 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 763 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 764 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 765 | |
1da177e4 LT |
766 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
767 | ||
768 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
769 | ||
96b7e579 | 770 | return 0; |
1da177e4 LT |
771 | |
772 | cleanup_all: | |
773 | fops_put(f->f_op); | |
83f936c7 | 774 | if (f->f_mode & FMODE_WRITER) { |
1da177e4 | 775 | put_write_access(inode); |
83f936c7 | 776 | __mnt_drop_write(f->f_path.mnt); |
4a3fd211 | 777 | } |
1da177e4 | 778 | cleanup_file: |
02e5180d AV |
779 | path_put(&f->f_path); |
780 | f->f_path.mnt = NULL; | |
781 | f->f_path.dentry = NULL; | |
dd37978c | 782 | f->f_inode = NULL; |
96b7e579 | 783 | return error; |
1da177e4 LT |
784 | } |
785 | ||
d18e9008 MS |
786 | /** |
787 | * finish_open - finish opening a file | |
0854d450 | 788 | * @file: file pointer |
d18e9008 MS |
789 | * @dentry: pointer to dentry |
790 | * @open: open callback | |
0854d450 | 791 | * @opened: state of open |
d18e9008 MS |
792 | * |
793 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
794 | * | |
795 | * If the open callback is set to NULL, then the standard f_op->open() | |
796 | * filesystem callback is substituted. | |
0854d450 MS |
797 | * |
798 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
799 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
800 | * on it after finish_open(). | |
801 | * | |
802 | * On successful return @file is a fully instantiated open file. After this, if | |
803 | * an error occurs in ->atomic_open(), it needs to clean up with fput(). | |
804 | * | |
805 | * Returns zero on success or -errno if the open failed. | |
d18e9008 | 806 | */ |
30d90494 AV |
807 | int finish_open(struct file *file, struct dentry *dentry, |
808 | int (*open)(struct inode *, struct file *), | |
809 | int *opened) | |
d18e9008 | 810 | { |
96b7e579 | 811 | int error; |
3d8a00d2 | 812 | BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 813 | |
b5bcdda3 | 814 | file->f_path.dentry = dentry; |
4bacc9c9 DH |
815 | error = do_dentry_open(file, d_backing_inode(dentry), open, |
816 | current_cred()); | |
96b7e579 | 817 | if (!error) |
47237687 | 818 | *opened |= FILE_OPENED; |
d18e9008 | 819 | |
96b7e579 | 820 | return error; |
d18e9008 MS |
821 | } |
822 | EXPORT_SYMBOL(finish_open); | |
823 | ||
824 | /** | |
825 | * finish_no_open - finish ->atomic_open() without opening the file | |
826 | * | |
0854d450 | 827 | * @file: file pointer |
d18e9008 MS |
828 | * @dentry: dentry or NULL (as returned from ->lookup()) |
829 | * | |
830 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
831 | * |
832 | * NB: unlike finish_open() this function does consume the dentry reference and | |
833 | * the caller need not dput() it. | |
834 | * | |
835 | * Returns "1" which must be the return value of ->atomic_open() after having | |
836 | * called this function. | |
d18e9008 | 837 | */ |
e45198a6 | 838 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 839 | { |
30d90494 | 840 | file->f_path.dentry = dentry; |
e45198a6 | 841 | return 1; |
d18e9008 MS |
842 | } |
843 | EXPORT_SYMBOL(finish_no_open); | |
844 | ||
9bf39ab2 MS |
845 | char *file_path(struct file *filp, char *buf, int buflen) |
846 | { | |
847 | return d_path(&filp->f_path, buf, buflen); | |
848 | } | |
849 | EXPORT_SYMBOL(file_path); | |
850 | ||
4bacc9c9 DH |
851 | /** |
852 | * vfs_open - open the file at the given path | |
853 | * @path: path to open | |
854 | * @file: newly allocated file with f_flag initialized | |
855 | * @cred: credentials to use | |
856 | */ | |
857 | int vfs_open(const struct path *path, struct file *file, | |
858 | const struct cred *cred) | |
859 | { | |
2d902671 | 860 | struct dentry *dentry = d_real(path->dentry, NULL, file->f_flags); |
4bacc9c9 | 861 | |
2d902671 MS |
862 | if (IS_ERR(dentry)) |
863 | return PTR_ERR(dentry); | |
4bacc9c9 | 864 | |
54d5ca87 | 865 | file->f_path = *path; |
2d902671 | 866 | return do_dentry_open(file, d_backing_inode(dentry), NULL, cred); |
4bacc9c9 DH |
867 | } |
868 | ||
765927b2 | 869 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 870 | const struct cred *cred) |
a1a5b3d9 PS |
871 | { |
872 | int error; | |
873 | struct file *f; | |
874 | ||
e0e81739 DH |
875 | validate_creds(cred); |
876 | ||
c212f9aa | 877 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 878 | BUG_ON(!path->mnt); |
322ee5b3 | 879 | |
a1a5b3d9 | 880 | f = get_empty_filp(); |
1afc99be AV |
881 | if (!IS_ERR(f)) { |
882 | f->f_flags = flags; | |
4aa7c634 | 883 | error = vfs_open(path, f, cred); |
1afc99be AV |
884 | if (!error) { |
885 | /* from now on we need fput() to dispose of f */ | |
886 | error = open_check_o_direct(f); | |
887 | if (error) { | |
888 | fput(f); | |
889 | f = ERR_PTR(error); | |
890 | } | |
891 | } else { | |
892 | put_filp(f); | |
2a027e7a AV |
893 | f = ERR_PTR(error); |
894 | } | |
2a027e7a AV |
895 | } |
896 | return f; | |
a1a5b3d9 | 897 | } |
1da177e4 LT |
898 | EXPORT_SYMBOL(dentry_open); |
899 | ||
a218d0fd | 900 | static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op) |
47c805dc AV |
901 | { |
902 | int lookup_flags = 0; | |
62fb4a15 | 903 | int acc_mode = ACC_MODE(flags); |
47c805dc | 904 | |
e305f48b | 905 | if (flags & (O_CREAT | __O_TMPFILE)) |
e68726ff MS |
906 | op->mode = (mode & S_IALLUGO) | S_IFREG; |
907 | else | |
908 | op->mode = 0; | |
47c805dc AV |
909 | |
910 | /* Must never be set by userspace */ | |
c6f3d811 | 911 | flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC; |
47c805dc AV |
912 | |
913 | /* | |
914 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
915 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
916 | * always set instead of having to deal with possibly weird behaviour | |
917 | * for malicious applications setting only __O_SYNC. | |
918 | */ | |
919 | if (flags & __O_SYNC) | |
920 | flags |= O_DSYNC; | |
921 | ||
bb458c64 AV |
922 | if (flags & __O_TMPFILE) { |
923 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d | 924 | return -EINVAL; |
ba57ea64 AV |
925 | if (!(acc_mode & MAY_WRITE)) |
926 | return -EINVAL; | |
60545d0d AV |
927 | } else if (flags & O_PATH) { |
928 | /* | |
929 | * If we have O_PATH in the open flag. Then we | |
930 | * cannot have anything other than the below set of flags | |
931 | */ | |
1abf0c71 AV |
932 | flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH; |
933 | acc_mode = 0; | |
1abf0c71 | 934 | } |
47c805dc | 935 | |
1abf0c71 | 936 | op->open_flag = flags; |
47c805dc AV |
937 | |
938 | /* O_TRUNC implies we need access checks for write permissions */ | |
939 | if (flags & O_TRUNC) | |
940 | acc_mode |= MAY_WRITE; | |
941 | ||
942 | /* Allow the LSM permission hook to distinguish append | |
943 | access from general write access. */ | |
944 | if (flags & O_APPEND) | |
945 | acc_mode |= MAY_APPEND; | |
946 | ||
947 | op->acc_mode = acc_mode; | |
948 | ||
1abf0c71 AV |
949 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
950 | ||
47c805dc AV |
951 | if (flags & O_CREAT) { |
952 | op->intent |= LOOKUP_CREATE; | |
953 | if (flags & O_EXCL) | |
954 | op->intent |= LOOKUP_EXCL; | |
955 | } | |
956 | ||
957 | if (flags & O_DIRECTORY) | |
958 | lookup_flags |= LOOKUP_DIRECTORY; | |
959 | if (!(flags & O_NOFOLLOW)) | |
960 | lookup_flags |= LOOKUP_FOLLOW; | |
f9652e10 AV |
961 | op->lookup_flags = lookup_flags; |
962 | return 0; | |
47c805dc AV |
963 | } |
964 | ||
669abf4e JL |
965 | /** |
966 | * file_open_name - open file and return file pointer | |
967 | * | |
968 | * @name: struct filename containing path to open | |
969 | * @flags: open flags as per the open(2) second argument | |
970 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
971 | * | |
972 | * This is the helper to open a file from kernelspace if you really | |
973 | * have to. But in generally you should not do this, so please move | |
974 | * along, nothing to see here.. | |
975 | */ | |
976 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
977 | { | |
978 | struct open_flags op; | |
f9652e10 AV |
979 | int err = build_open_flags(flags, mode, &op); |
980 | return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
981 | } |
982 | ||
47c805dc AV |
983 | /** |
984 | * filp_open - open file and return file pointer | |
985 | * | |
986 | * @filename: path to open | |
987 | * @flags: open flags as per the open(2) second argument | |
988 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
989 | * | |
990 | * This is the helper to open a file from kernelspace if you really | |
991 | * have to. But in generally you should not do this, so please move | |
992 | * along, nothing to see here.. | |
993 | */ | |
a218d0fd | 994 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 995 | { |
51689104 PM |
996 | struct filename *name = getname_kernel(filename); |
997 | struct file *file = ERR_CAST(name); | |
998 | ||
999 | if (!IS_ERR(name)) { | |
1000 | file = file_open_name(name, flags, mode); | |
1001 | putname(name); | |
1002 | } | |
1003 | return file; | |
47c805dc AV |
1004 | } |
1005 | EXPORT_SYMBOL(filp_open); | |
1006 | ||
73d049a4 | 1007 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
378c6520 | 1008 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1009 | { |
1010 | struct open_flags op; | |
378c6520 | 1011 | int err = build_open_flags(flags, mode, &op); |
f9652e10 AV |
1012 | if (err) |
1013 | return ERR_PTR(err); | |
f9652e10 | 1014 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
1015 | } |
1016 | EXPORT_SYMBOL(file_open_root); | |
1017 | ||
9a08c352 JB |
1018 | struct file *filp_clone_open(struct file *oldfile) |
1019 | { | |
1020 | struct file *file; | |
1021 | int retval; | |
1022 | ||
1023 | file = get_empty_filp(); | |
1024 | if (IS_ERR(file)) | |
1025 | return file; | |
1026 | ||
1027 | file->f_flags = oldfile->f_flags; | |
1028 | retval = vfs_open(&oldfile->f_path, file, oldfile->f_cred); | |
1029 | if (retval) { | |
1030 | put_filp(file); | |
1031 | return ERR_PTR(retval); | |
1032 | } | |
1033 | ||
1034 | return file; | |
1035 | } | |
1036 | EXPORT_SYMBOL(filp_clone_open); | |
1037 | ||
a218d0fd | 1038 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
1da177e4 | 1039 | { |
47c805dc | 1040 | struct open_flags op; |
f9652e10 AV |
1041 | int fd = build_open_flags(flags, mode, &op); |
1042 | struct filename *tmp; | |
1043 | ||
1044 | if (fd) | |
1045 | return fd; | |
1046 | ||
1047 | tmp = getname(filename); | |
1048 | if (IS_ERR(tmp)) | |
1049 | return PTR_ERR(tmp); | |
1050 | ||
1051 | fd = get_unused_fd_flags(flags); | |
1052 | if (fd >= 0) { | |
1053 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1054 | if (IS_ERR(f)) { | |
1055 | put_unused_fd(fd); | |
1056 | fd = PTR_ERR(f); | |
1057 | } else { | |
1058 | fsnotify_open(f); | |
1059 | fd_install(fd, f); | |
1da177e4 | 1060 | } |
1da177e4 | 1061 | } |
f9652e10 | 1062 | putname(tmp); |
1da177e4 | 1063 | return fd; |
1da177e4 | 1064 | } |
e922efc3 | 1065 | |
a218d0fd | 1066 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) |
e922efc3 MS |
1067 | { |
1068 | if (force_o_largefile()) | |
1069 | flags |= O_LARGEFILE; | |
1070 | ||
2cf09666 | 1071 | return do_sys_open(AT_FDCWD, filename, flags, mode); |
e922efc3 | 1072 | } |
1da177e4 | 1073 | |
6559eed8 | 1074 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1075 | umode_t, mode) |
5590ff0d UD |
1076 | { |
1077 | if (force_o_largefile()) | |
1078 | flags |= O_LARGEFILE; | |
1079 | ||
2cf09666 | 1080 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1081 | } |
5590ff0d | 1082 | |
1da177e4 LT |
1083 | #ifndef __alpha__ |
1084 | ||
1085 | /* | |
1086 | * For backward compatibility? Maybe this should be moved | |
1087 | * into arch/i386 instead? | |
1088 | */ | |
a218d0fd | 1089 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 LT |
1090 | { |
1091 | return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); | |
1092 | } | |
1093 | ||
1094 | #endif | |
1095 | ||
1096 | /* | |
1097 | * "id" is the POSIX thread ID. We use the | |
1098 | * files pointer for this.. | |
1099 | */ | |
1100 | int filp_close(struct file *filp, fl_owner_t id) | |
1101 | { | |
45778ca8 | 1102 | int retval = 0; |
1da177e4 LT |
1103 | |
1104 | if (!file_count(filp)) { | |
1105 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1106 | return 0; |
1da177e4 LT |
1107 | } |
1108 | ||
72c2d531 | 1109 | if (filp->f_op->flush) |
75e1fcc0 | 1110 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1111 | |
1abf0c71 AV |
1112 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1113 | dnotify_flush(filp, id); | |
1114 | locks_remove_posix(filp, id); | |
1115 | } | |
1da177e4 LT |
1116 | fput(filp); |
1117 | return retval; | |
1118 | } | |
1119 | ||
1120 | EXPORT_SYMBOL(filp_close); | |
1121 | ||
1122 | /* | |
1123 | * Careful here! We test whether the file pointer is NULL before | |
1124 | * releasing the fd. This ensures that one clone task can't release | |
1125 | * an fd while another clone is opening it. | |
1126 | */ | |
ca013e94 | 1127 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1128 | { |
483ce1d4 | 1129 | int retval = __close_fd(current->files, fd); |
ee731f4f EP |
1130 | |
1131 | /* can't restart close syscall because file table entry was cleared */ | |
1132 | if (unlikely(retval == -ERESTARTSYS || | |
1133 | retval == -ERESTARTNOINTR || | |
1134 | retval == -ERESTARTNOHAND || | |
1135 | retval == -ERESTART_RESTARTBLOCK)) | |
1136 | retval = -EINTR; | |
1137 | ||
1138 | return retval; | |
1da177e4 | 1139 | } |
1da177e4 LT |
1140 | EXPORT_SYMBOL(sys_close); |
1141 | ||
1142 | /* | |
1143 | * This routine simulates a hangup on the tty, to arrange that users | |
1144 | * are given clean terminals at login time. | |
1145 | */ | |
ca013e94 | 1146 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1147 | { |
1148 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1149 | tty_vhangup_self(); |
1da177e4 LT |
1150 | return 0; |
1151 | } | |
1152 | return -EPERM; | |
1153 | } | |
1154 | ||
1155 | /* | |
1156 | * Called when an inode is about to be open. | |
1157 | * We use this to disallow opening large files on 32bit systems if | |
1158 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1159 | * on this flag in sys_open. | |
1160 | */ | |
1161 | int generic_file_open(struct inode * inode, struct file * filp) | |
1162 | { | |
1163 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1164 | return -EOVERFLOW; |
1da177e4 LT |
1165 | return 0; |
1166 | } | |
1167 | ||
1168 | EXPORT_SYMBOL(generic_file_open); | |
1169 | ||
1170 | /* | |
1171 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1172 | * file descriptors. The function is not supposed to ever fail, the only |
1173 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1174 | * directly into file_operations structure. | |
1da177e4 LT |
1175 | */ |
1176 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1177 | { | |
1178 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1179 | return 0; | |
1180 | } | |
1181 | ||
1182 | EXPORT_SYMBOL(nonseekable_open); |