]>
Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/fs/open.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
8 | #include <linux/string.h> | |
9 | #include <linux/mm.h> | |
1da177e4 | 10 | #include <linux/file.h> |
9f3acc31 | 11 | #include <linux/fdtable.h> |
0eeca283 | 12 | #include <linux/fsnotify.h> |
1da177e4 | 13 | #include <linux/module.h> |
1da177e4 LT |
14 | #include <linux/tty.h> |
15 | #include <linux/namei.h> | |
16 | #include <linux/backing-dev.h> | |
16f7e0fe | 17 | #include <linux/capability.h> |
086f7316 | 18 | #include <linux/securebits.h> |
1da177e4 LT |
19 | #include <linux/security.h> |
20 | #include <linux/mount.h> | |
5590ff0d | 21 | #include <linux/fcntl.h> |
5a0e3ad6 | 22 | #include <linux/slab.h> |
7c0f6ba6 | 23 | #include <linux/uaccess.h> |
1da177e4 | 24 | #include <linux/fs.h> |
ef3daeda | 25 | #include <linux/personality.h> |
1da177e4 LT |
26 | #include <linux/pagemap.h> |
27 | #include <linux/syscalls.h> | |
ab2af1f5 | 28 | #include <linux/rcupdate.h> |
73241ccc | 29 | #include <linux/audit.h> |
97ac7350 | 30 | #include <linux/falloc.h> |
5ad4e53b | 31 | #include <linux/fs_struct.h> |
b65a9cfc | 32 | #include <linux/ima.h> |
2dfc1cae | 33 | #include <linux/dnotify.h> |
3f6d078d | 34 | #include <linux/compat.h> |
1da177e4 | 35 | |
e81e3f4d EP |
36 | #include "internal.h" |
37 | ||
fcc38bd3 SJR |
38 | #define CREATE_TRACE_POINTS |
39 | #include <trace/events/fs.h> | |
40 | ||
4a30131e N |
41 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, |
42 | struct file *filp) | |
1da177e4 | 43 | { |
939a9421 | 44 | int ret; |
1da177e4 LT |
45 | struct iattr newattrs; |
46 | ||
47 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
48 | if (length < 0) | |
49 | return -EINVAL; | |
50 | ||
51 | newattrs.ia_size = length; | |
4a30131e | 52 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
53 | if (filp) { |
54 | newattrs.ia_file = filp; | |
55 | newattrs.ia_valid |= ATTR_FILE; | |
56 | } | |
1da177e4 | 57 | |
45f147a1 JK |
58 | /* Remove suid, sgid, and file capabilities on truncate too */ |
59 | ret = dentry_needs_remove_privs(dentry); | |
60 | if (ret < 0) | |
61 | return ret; | |
939a9421 AW |
62 | if (ret) |
63 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 64 | |
5955102c | 65 | inode_lock(dentry->d_inode); |
27ac0ffe BF |
66 | /* Note any delegations or leases have already been broken: */ |
67 | ret = notify_change(dentry, &newattrs, NULL); | |
5955102c | 68 | inode_unlock(dentry->d_inode); |
939a9421 | 69 | return ret; |
1da177e4 | 70 | } |
a3a49a17 | 71 | EXPORT_SYMBOL_GPL(do_truncate); |
1da177e4 | 72 | |
7df818b2 | 73 | long vfs_truncate(const struct path *path, loff_t length) |
1da177e4 | 74 | { |
2d8f3038 | 75 | struct inode *inode; |
a02de960 | 76 | long error; |
1da177e4 | 77 | |
a02de960 | 78 | inode = path->dentry->d_inode; |
1da177e4 LT |
79 | |
80 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
1da177e4 | 81 | if (S_ISDIR(inode->i_mode)) |
a02de960 | 82 | return -EISDIR; |
1da177e4 | 83 | if (!S_ISREG(inode->i_mode)) |
a02de960 | 84 | return -EINVAL; |
1da177e4 | 85 | |
a02de960 | 86 | error = mnt_want_write(path->mnt); |
1da177e4 | 87 | if (error) |
a02de960 | 88 | goto out; |
1da177e4 | 89 | |
256984a8 | 90 | error = inode_permission(inode, MAY_WRITE); |
9ac9b847 DH |
91 | if (error) |
92 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
93 | |
94 | error = -EPERM; | |
c82e42da | 95 | if (IS_APPEND(inode)) |
9ac9b847 | 96 | goto mnt_drop_write_and_out; |
1da177e4 | 97 | |
8cf9ee50 | 98 | error = get_write_access(inode); |
1da177e4 | 99 | if (error) |
9ac9b847 | 100 | goto mnt_drop_write_and_out; |
1da177e4 | 101 | |
9700382c | 102 | /* |
103 | * Make sure that there are no leases. get_write_access() protects | |
104 | * against the truncate racing with a lease-granting setlease(). | |
105 | */ | |
8737c930 | 106 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 107 | if (error) |
9700382c | 108 | goto put_write_and_out; |
1da177e4 LT |
109 | |
110 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 | 111 | if (!error) |
a02de960 | 112 | error = security_path_truncate(path); |
907f4554 | 113 | if (!error) |
a02de960 | 114 | error = do_truncate(path->dentry, length, 0, NULL); |
1da177e4 | 115 | |
9700382c | 116 | put_write_and_out: |
8cf9ee50 | 117 | put_write_access(inode); |
9ac9b847 | 118 | mnt_drop_write_and_out: |
a02de960 | 119 | mnt_drop_write(path->mnt); |
1da177e4 LT |
120 | out: |
121 | return error; | |
122 | } | |
a02de960 DH |
123 | EXPORT_SYMBOL_GPL(vfs_truncate); |
124 | ||
df260e21 | 125 | long do_sys_truncate(const char __user *pathname, loff_t length) |
a02de960 | 126 | { |
48f7530d | 127 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
128 | struct path path; |
129 | int error; | |
130 | ||
131 | if (length < 0) /* sorry, but loff_t says... */ | |
132 | return -EINVAL; | |
133 | ||
48f7530d JL |
134 | retry: |
135 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
136 | if (!error) { |
137 | error = vfs_truncate(&path, length); | |
138 | path_put(&path); | |
139 | } | |
48f7530d JL |
140 | if (retry_estale(error, lookup_flags)) { |
141 | lookup_flags |= LOOKUP_REVAL; | |
142 | goto retry; | |
143 | } | |
a02de960 DH |
144 | return error; |
145 | } | |
1da177e4 | 146 | |
4fd8da8d | 147 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 148 | { |
4fd8da8d | 149 | return do_sys_truncate(path, length); |
1da177e4 LT |
150 | } |
151 | ||
3f6d078d AV |
152 | #ifdef CONFIG_COMPAT |
153 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
154 | { | |
155 | return do_sys_truncate(path, length); | |
156 | } | |
157 | #endif | |
158 | ||
411d9475 | 159 | long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 | 160 | { |
bf2965d5 | 161 | struct inode *inode; |
1da177e4 | 162 | struct dentry *dentry; |
2903ff01 | 163 | struct fd f; |
1da177e4 LT |
164 | int error; |
165 | ||
166 | error = -EINVAL; | |
167 | if (length < 0) | |
168 | goto out; | |
169 | error = -EBADF; | |
2903ff01 AV |
170 | f = fdget(fd); |
171 | if (!f.file) | |
1da177e4 LT |
172 | goto out; |
173 | ||
174 | /* explicitly opened as large or we are on 64-bit box */ | |
2903ff01 | 175 | if (f.file->f_flags & O_LARGEFILE) |
1da177e4 LT |
176 | small = 0; |
177 | ||
2903ff01 | 178 | dentry = f.file->f_path.dentry; |
1da177e4 LT |
179 | inode = dentry->d_inode; |
180 | error = -EINVAL; | |
2903ff01 | 181 | if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE)) |
1da177e4 LT |
182 | goto out_putf; |
183 | ||
184 | error = -EINVAL; | |
185 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
186 | if (small && length > MAX_NON_LFS) | |
187 | goto out_putf; | |
188 | ||
189 | error = -EPERM; | |
78757af6 AG |
190 | /* Check IS_APPEND on real upper inode */ |
191 | if (IS_APPEND(file_inode(f.file))) | |
1da177e4 LT |
192 | goto out_putf; |
193 | ||
14da9200 | 194 | sb_start_write(inode->i_sb); |
2903ff01 | 195 | error = locks_verify_truncate(inode, f.file, length); |
be6d3e56 | 196 | if (!error) |
2903ff01 | 197 | error = security_path_truncate(&f.file->f_path); |
1da177e4 | 198 | if (!error) |
2903ff01 | 199 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file); |
14da9200 | 200 | sb_end_write(inode->i_sb); |
1da177e4 | 201 | out_putf: |
2903ff01 | 202 | fdput(f); |
1da177e4 LT |
203 | out: |
204 | return error; | |
205 | } | |
206 | ||
bdc480e3 | 207 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 208 | { |
2cf09666 | 209 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
210 | } |
211 | ||
3f6d078d AV |
212 | #ifdef CONFIG_COMPAT |
213 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length) | |
214 | { | |
215 | return do_sys_ftruncate(fd, length, 1); | |
216 | } | |
217 | #endif | |
218 | ||
1da177e4 LT |
219 | /* LFS versions of truncate are only needed on 32 bit machines */ |
220 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 221 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
222 | { |
223 | return do_sys_truncate(path, length); | |
224 | } | |
225 | ||
4a0fd5bf | 226 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 227 | { |
2cf09666 | 228 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 229 | } |
6673e0c3 | 230 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 231 | |
3e63cbb1 | 232 | |
72c72bdf | 233 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 234 | { |
496ad9aa | 235 | struct inode *inode = file_inode(file); |
3e63cbb1 | 236 | long ret; |
97ac7350 AA |
237 | |
238 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 239 | return -EINVAL; |
97ac7350 AA |
240 | |
241 | /* Return error if mode is not supported */ | |
dd46c787 | 242 | if (mode & ~FALLOC_FL_SUPPORTED_MASK) |
409332b6 LC |
243 | return -EOPNOTSUPP; |
244 | ||
245 | /* Punch hole and zero range are mutually exclusive */ | |
246 | if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) == | |
247 | (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) | |
79124f18 JB |
248 | return -EOPNOTSUPP; |
249 | ||
250 | /* Punch hole must have keep size set */ | |
251 | if ((mode & FALLOC_FL_PUNCH_HOLE) && | |
252 | !(mode & FALLOC_FL_KEEP_SIZE)) | |
3e63cbb1 | 253 | return -EOPNOTSUPP; |
97ac7350 | 254 | |
00f5e619 NJ |
255 | /* Collapse range should only be used exclusively. */ |
256 | if ((mode & FALLOC_FL_COLLAPSE_RANGE) && | |
257 | (mode & ~FALLOC_FL_COLLAPSE_RANGE)) | |
258 | return -EINVAL; | |
259 | ||
dd46c787 NJ |
260 | /* Insert range should only be used exclusively. */ |
261 | if ((mode & FALLOC_FL_INSERT_RANGE) && | |
262 | (mode & ~FALLOC_FL_INSERT_RANGE)) | |
263 | return -EINVAL; | |
264 | ||
71be6b49 DW |
265 | /* Unshare range should only be used with allocate mode. */ |
266 | if ((mode & FALLOC_FL_UNSHARE_RANGE) && | |
267 | (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE))) | |
268 | return -EINVAL; | |
269 | ||
97ac7350 | 270 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 271 | return -EBADF; |
1ca551c6 | 272 | |
00f5e619 | 273 | /* |
8fc61d92 | 274 | * We can only allow pure fallocate on append only files |
00f5e619 | 275 | */ |
8fc61d92 | 276 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
277 | return -EPERM; |
278 | ||
279 | if (IS_IMMUTABLE(inode)) | |
280 | return -EPERM; | |
281 | ||
0790b31b | 282 | /* |
6d2b6170 | 283 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
284 | */ |
285 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 286 | return -ETXTBSY; |
0790b31b | 287 | |
97ac7350 AA |
288 | /* |
289 | * Revalidate the write permissions, in case security policy has | |
290 | * changed since the files were opened. | |
291 | */ | |
292 | ret = security_file_permission(file, MAY_WRITE); | |
293 | if (ret) | |
3e63cbb1 | 294 | return ret; |
97ac7350 | 295 | |
97ac7350 | 296 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 297 | return -ESPIPE; |
97ac7350 | 298 | |
9e79b132 AG |
299 | if (S_ISDIR(inode->i_mode)) |
300 | return -EISDIR; | |
301 | ||
302 | if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode)) | |
3e63cbb1 | 303 | return -ENODEV; |
97ac7350 | 304 | |
97ac7350 AA |
305 | /* Check for wrap through zero too */ |
306 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 307 | return -EFBIG; |
97ac7350 | 308 | |
2fe17c10 | 309 | if (!file->f_op->fallocate) |
3e63cbb1 | 310 | return -EOPNOTSUPP; |
97ac7350 | 311 | |
bfe219d3 | 312 | file_start_write(file); |
14da9200 | 313 | ret = file->f_op->fallocate(file, mode, offset, len); |
820c12d5 HS |
314 | |
315 | /* | |
316 | * Create inotify and fanotify events. | |
317 | * | |
318 | * To keep the logic simple always create events if fallocate succeeds. | |
319 | * This implies that events are even created if the file size remains | |
320 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
321 | */ | |
322 | if (ret == 0) | |
323 | fsnotify_modify(file); | |
324 | ||
bfe219d3 | 325 | file_end_write(file); |
14da9200 | 326 | return ret; |
3e63cbb1 | 327 | } |
72c72bdf | 328 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 329 | |
edf292c7 | 330 | int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len) |
3e63cbb1 | 331 | { |
2903ff01 | 332 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
333 | int error = -EBADF; |
334 | ||
2903ff01 | 335 | if (f.file) { |
72c72bdf | 336 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 337 | fdput(f); |
3e63cbb1 | 338 | } |
3e63cbb1 | 339 | return error; |
97ac7350 | 340 | } |
3e63cbb1 | 341 | |
edf292c7 DB |
342 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
343 | { | |
344 | return ksys_fallocate(fd, mode, offset, len); | |
345 | } | |
346 | ||
1da177e4 LT |
347 | /* |
348 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
349 | * We do this by temporarily clearing all FS-related capabilities and | |
350 | * switching the fsuid/fsgid around to the real ones. | |
351 | */ | |
cbfe20f5 | 352 | long do_faccessat(int dfd, const char __user *filename, int mode) |
1da177e4 | 353 | { |
d84f4f99 DH |
354 | const struct cred *old_cred; |
355 | struct cred *override_cred; | |
2d8f3038 | 356 | struct path path; |
256984a8 | 357 | struct inode *inode; |
1da177e4 | 358 | int res; |
87fa5595 | 359 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
1da177e4 LT |
360 | |
361 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
362 | return -EINVAL; | |
363 | ||
d84f4f99 DH |
364 | override_cred = prepare_creds(); |
365 | if (!override_cred) | |
366 | return -ENOMEM; | |
1da177e4 | 367 | |
d84f4f99 DH |
368 | override_cred->fsuid = override_cred->uid; |
369 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 370 | |
086f7316 | 371 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 372 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
373 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
374 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 375 | cap_clear(override_cred->cap_effective); |
086f7316 | 376 | else |
d84f4f99 DH |
377 | override_cred->cap_effective = |
378 | override_cred->cap_permitted; | |
086f7316 | 379 | } |
1da177e4 | 380 | |
d7852fbd LT |
381 | /* |
382 | * The new set of credentials can *only* be used in | |
383 | * task-synchronous circumstances, and does not need | |
384 | * RCU freeing, unless somebody then takes a separate | |
385 | * reference to it. | |
386 | * | |
387 | * NOTE! This is _only_ true because this credential | |
388 | * is used purely for override_creds() that installs | |
389 | * it as the subjective cred. Other threads will be | |
390 | * accessing ->real_cred, not the subjective cred. | |
391 | * | |
392 | * If somebody _does_ make a copy of this (using the | |
393 | * 'get_current_cred()' function), that will clear the | |
394 | * non_rcu field, because now that other user may be | |
395 | * expecting RCU freeing. But normal thread-synchronous | |
396 | * cred accesses will keep things non-RCY. | |
397 | */ | |
398 | override_cred->non_rcu = 1; | |
399 | ||
d84f4f99 | 400 | old_cred = override_creds(override_cred); |
87fa5595 JL |
401 | retry: |
402 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
403 | if (res) |
404 | goto out; | |
405 | ||
63afdfc7 | 406 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
407 | |
408 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
409 | /* |
410 | * MAY_EXEC on regular files is denied if the fs is mounted | |
411 | * with the "noexec" flag. | |
412 | */ | |
413 | res = -EACCES; | |
90f8572b | 414 | if (path_noexec(&path)) |
30524472 AV |
415 | goto out_path_release; |
416 | } | |
417 | ||
256984a8 | 418 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 419 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 420 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 421 | goto out_path_release; |
2f676cbc DH |
422 | /* |
423 | * This is a rare case where using __mnt_is_readonly() | |
424 | * is OK without a mnt_want/drop_write() pair. Since | |
425 | * no actual write to the fs is performed here, we do | |
426 | * not need to telegraph to that to anyone. | |
427 | * | |
428 | * By doing this, we accept that this access is | |
429 | * inherently racy and know that the fs may change | |
430 | * state before we even see this result. | |
431 | */ | |
2d8f3038 | 432 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 433 | res = -EROFS; |
1da177e4 | 434 | |
6902d925 | 435 | out_path_release: |
2d8f3038 | 436 | path_put(&path); |
87fa5595 JL |
437 | if (retry_estale(res, lookup_flags)) { |
438 | lookup_flags |= LOOKUP_REVAL; | |
439 | goto retry; | |
440 | } | |
6902d925 | 441 | out: |
d84f4f99 DH |
442 | revert_creds(old_cred); |
443 | put_cred(override_cred); | |
1da177e4 LT |
444 | return res; |
445 | } | |
446 | ||
cbfe20f5 DB |
447 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
448 | { | |
449 | return do_faccessat(dfd, filename, mode); | |
450 | } | |
451 | ||
ca013e94 | 452 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d | 453 | { |
cbfe20f5 | 454 | return do_faccessat(AT_FDCWD, filename, mode); |
5590ff0d UD |
455 | } |
456 | ||
447016e9 | 457 | int ksys_chdir(const char __user *filename) |
1da177e4 | 458 | { |
2d8f3038 | 459 | struct path path; |
1da177e4 | 460 | int error; |
0291c0a5 JL |
461 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
462 | retry: | |
463 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
464 | if (error) |
465 | goto out; | |
466 | ||
9cfcac81 | 467 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
468 | if (error) |
469 | goto dput_and_out; | |
470 | ||
2d8f3038 | 471 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
472 | |
473 | dput_and_out: | |
2d8f3038 | 474 | path_put(&path); |
0291c0a5 JL |
475 | if (retry_estale(error, lookup_flags)) { |
476 | lookup_flags |= LOOKUP_REVAL; | |
477 | goto retry; | |
478 | } | |
1da177e4 LT |
479 | out: |
480 | return error; | |
481 | } | |
482 | ||
447016e9 DB |
483 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
484 | { | |
485 | return ksys_chdir(filename); | |
486 | } | |
487 | ||
3cdad428 | 488 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 489 | { |
2903ff01 | 490 | struct fd f = fdget_raw(fd); |
159b0956 | 491 | int error; |
1da177e4 LT |
492 | |
493 | error = -EBADF; | |
2903ff01 | 494 | if (!f.file) |
1da177e4 LT |
495 | goto out; |
496 | ||
1da177e4 | 497 | error = -ENOTDIR; |
159b0956 | 498 | if (!d_can_lookup(f.file->f_path.dentry)) |
1da177e4 LT |
499 | goto out_putf; |
500 | ||
159b0956 | 501 | error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR); |
1da177e4 | 502 | if (!error) |
2903ff01 | 503 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 504 | out_putf: |
2903ff01 | 505 | fdput(f); |
1da177e4 LT |
506 | out: |
507 | return error; | |
508 | } | |
509 | ||
a16fe33a | 510 | int ksys_chroot(const char __user *filename) |
1da177e4 | 511 | { |
2d8f3038 | 512 | struct path path; |
1da177e4 | 513 | int error; |
2771261e JL |
514 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
515 | retry: | |
516 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
517 | if (error) |
518 | goto out; | |
519 | ||
9cfcac81 | 520 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
521 | if (error) |
522 | goto dput_and_out; | |
523 | ||
524 | error = -EPERM; | |
c7b96acf | 525 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 526 | goto dput_and_out; |
8b8efb44 TH |
527 | error = security_path_chroot(&path); |
528 | if (error) | |
529 | goto dput_and_out; | |
1da177e4 | 530 | |
2d8f3038 | 531 | set_fs_root(current->fs, &path); |
1da177e4 LT |
532 | error = 0; |
533 | dput_and_out: | |
2d8f3038 | 534 | path_put(&path); |
2771261e JL |
535 | if (retry_estale(error, lookup_flags)) { |
536 | lookup_flags |= LOOKUP_REVAL; | |
537 | goto retry; | |
538 | } | |
1da177e4 LT |
539 | out: |
540 | return error; | |
541 | } | |
542 | ||
a16fe33a DB |
543 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
544 | { | |
545 | return ksys_chroot(filename); | |
546 | } | |
547 | ||
be01f9f2 | 548 | static int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 549 | { |
e57712eb | 550 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 551 | struct inode *delegated_inode = NULL; |
1da177e4 | 552 | struct iattr newattrs; |
e57712eb | 553 | int error; |
1da177e4 | 554 | |
e57712eb AV |
555 | error = mnt_want_write(path->mnt); |
556 | if (error) | |
557 | return error; | |
27ac0ffe | 558 | retry_deleg: |
5955102c | 559 | inode_lock(inode); |
cdcf116d | 560 | error = security_path_chmod(path, mode); |
e57712eb | 561 | if (error) |
fe542cf5 | 562 | goto out_unlock; |
1da177e4 LT |
563 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
564 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
27ac0ffe | 565 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
fe542cf5 | 566 | out_unlock: |
5955102c | 567 | inode_unlock(inode); |
27ac0ffe BF |
568 | if (delegated_inode) { |
569 | error = break_deleg_wait(&delegated_inode); | |
570 | if (!error) | |
571 | goto retry_deleg; | |
572 | } | |
e57712eb AV |
573 | mnt_drop_write(path->mnt); |
574 | return error; | |
575 | } | |
576 | ||
03450e27 | 577 | int ksys_fchmod(unsigned int fd, umode_t mode) |
e57712eb | 578 | { |
173c8401 | 579 | struct fd f = fdget(fd); |
e57712eb AV |
580 | int err = -EBADF; |
581 | ||
173c8401 | 582 | if (f.file) { |
9f45f5bf | 583 | audit_file(f.file); |
173c8401 AV |
584 | err = chmod_common(&f.file->f_path, mode); |
585 | fdput(f); | |
e57712eb | 586 | } |
1da177e4 LT |
587 | return err; |
588 | } | |
589 | ||
03450e27 DB |
590 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
591 | { | |
592 | return ksys_fchmod(fd, mode); | |
593 | } | |
594 | ||
595 | int do_fchmodat(int dfd, const char __user *filename, umode_t mode) | |
1da177e4 | 596 | { |
2d8f3038 | 597 | struct path path; |
1da177e4 | 598 | int error; |
14ff690c JL |
599 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
600 | retry: | |
601 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
602 | if (!error) { |
603 | error = chmod_common(&path, mode); | |
604 | path_put(&path); | |
14ff690c JL |
605 | if (retry_estale(error, lookup_flags)) { |
606 | lookup_flags |= LOOKUP_REVAL; | |
607 | goto retry; | |
608 | } | |
e57712eb | 609 | } |
1da177e4 LT |
610 | return error; |
611 | } | |
612 | ||
03450e27 DB |
613 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, |
614 | umode_t, mode) | |
615 | { | |
616 | return do_fchmodat(dfd, filename, mode); | |
617 | } | |
618 | ||
49f0a076 | 619 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d | 620 | { |
03450e27 | 621 | return do_fchmodat(AT_FDCWD, filename, mode); |
5590ff0d UD |
622 | } |
623 | ||
7fd25dac | 624 | static int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 625 | { |
fe542cf5 | 626 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 627 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
628 | int error; |
629 | struct iattr newattrs; | |
52137abe EB |
630 | kuid_t uid; |
631 | kgid_t gid; | |
632 | ||
633 | uid = make_kuid(current_user_ns(), user); | |
634 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 635 | |
c1b8940b | 636 | retry_deleg: |
1da177e4 LT |
637 | newattrs.ia_valid = ATTR_CTIME; |
638 | if (user != (uid_t) -1) { | |
52137abe EB |
639 | if (!uid_valid(uid)) |
640 | return -EINVAL; | |
1da177e4 | 641 | newattrs.ia_valid |= ATTR_UID; |
52137abe | 642 | newattrs.ia_uid = uid; |
1da177e4 LT |
643 | } |
644 | if (group != (gid_t) -1) { | |
52137abe EB |
645 | if (!gid_valid(gid)) |
646 | return -EINVAL; | |
1da177e4 | 647 | newattrs.ia_valid |= ATTR_GID; |
52137abe | 648 | newattrs.ia_gid = gid; |
1da177e4 LT |
649 | } |
650 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
651 | newattrs.ia_valid |= |
652 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
5955102c | 653 | inode_lock(inode); |
d2b31ca6 | 654 | error = security_path_chown(path, uid, gid); |
fe542cf5 | 655 | if (!error) |
27ac0ffe | 656 | error = notify_change(path->dentry, &newattrs, &delegated_inode); |
5955102c | 657 | inode_unlock(inode); |
27ac0ffe BF |
658 | if (delegated_inode) { |
659 | error = break_deleg_wait(&delegated_inode); | |
660 | if (!error) | |
661 | goto retry_deleg; | |
662 | } | |
1da177e4 LT |
663 | return error; |
664 | } | |
665 | ||
55731b3c DB |
666 | int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group, |
667 | int flag) | |
5590ff0d | 668 | { |
2d8f3038 | 669 | struct path path; |
5590ff0d | 670 | int error = -EINVAL; |
65cfc672 | 671 | int lookup_flags; |
5590ff0d | 672 | |
65cfc672 | 673 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
674 | goto out; |
675 | ||
65cfc672 AV |
676 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
677 | if (flag & AT_EMPTY_PATH) | |
678 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 679 | retry: |
65cfc672 | 680 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
681 | if (error) |
682 | goto out; | |
2d8f3038 | 683 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
684 | if (error) |
685 | goto out_release; | |
fe542cf5 | 686 | error = chown_common(&path, user, group); |
2d8f3038 | 687 | mnt_drop_write(path.mnt); |
2af482a7 | 688 | out_release: |
2d8f3038 | 689 | path_put(&path); |
99a5df37 JL |
690 | if (retry_estale(error, lookup_flags)) { |
691 | lookup_flags |= LOOKUP_REVAL; | |
692 | goto retry; | |
693 | } | |
5590ff0d UD |
694 | out: |
695 | return error; | |
696 | } | |
697 | ||
55731b3c DB |
698 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
699 | gid_t, group, int, flag) | |
700 | { | |
701 | return do_fchownat(dfd, filename, user, group, flag); | |
702 | } | |
703 | ||
55e4def0 | 704 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 705 | { |
55731b3c | 706 | return do_fchownat(AT_FDCWD, filename, user, group, 0); |
55e4def0 | 707 | } |
1da177e4 | 708 | |
55e4def0 DH |
709 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
710 | { | |
55731b3c DB |
711 | return do_fchownat(AT_FDCWD, filename, user, group, |
712 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
713 | } |
714 | ||
55731b3c | 715 | int ksys_fchown(unsigned int fd, uid_t user, gid_t group) |
1da177e4 | 716 | { |
2903ff01 | 717 | struct fd f = fdget(fd); |
1da177e4 LT |
718 | int error = -EBADF; |
719 | ||
2903ff01 | 720 | if (!f.file) |
6902d925 DH |
721 | goto out; |
722 | ||
6742cee0 | 723 | error = mnt_want_write_file(f.file); |
2af482a7 DH |
724 | if (error) |
725 | goto out_fput; | |
9f45f5bf | 726 | audit_file(f.file); |
2903ff01 | 727 | error = chown_common(&f.file->f_path, user, group); |
6742cee0 | 728 | mnt_drop_write_file(f.file); |
2af482a7 | 729 | out_fput: |
2903ff01 | 730 | fdput(f); |
6902d925 | 731 | out: |
1da177e4 LT |
732 | return error; |
733 | } | |
734 | ||
55731b3c DB |
735 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
736 | { | |
737 | return ksys_fchown(fd, user, group); | |
738 | } | |
739 | ||
02e5180d | 740 | static int do_dentry_open(struct file *f, |
4bacc9c9 | 741 | struct inode *inode, |
ae2bb293 | 742 | int (*open)(struct inode *, struct file *)) |
1da177e4 | 743 | { |
1abf0c71 | 744 | static const struct file_operations empty_fops = {}; |
1da177e4 LT |
745 | int error; |
746 | ||
b5bcdda3 | 747 | path_get(&f->f_path); |
4bacc9c9 | 748 | f->f_inode = inode; |
1da177e4 | 749 | f->f_mapping = inode->i_mapping; |
1da177e4 | 750 | |
5660e13d JL |
751 | /* Ensure that we skip any errors that predate opening of the file */ |
752 | f->f_wb_err = filemap_sample_wb_err(f->f_mapping); | |
753 | ||
3f4d5a00 | 754 | if (unlikely(f->f_flags & O_PATH)) { |
f5d11409 | 755 | f->f_mode = FMODE_PATH | FMODE_OPENED; |
1abf0c71 | 756 | f->f_op = &empty_fops; |
af04fadc | 757 | return 0; |
1abf0c71 AV |
758 | } |
759 | ||
73601ea5 TH |
760 | /* Any file opened for execve()/uselib() has to be a regular file. */ |
761 | if (unlikely(f->f_flags & FMODE_EXEC && !S_ISREG(inode->i_mode))) { | |
762 | error = -EACCES; | |
763 | goto cleanup_file; | |
764 | } | |
765 | ||
dd20908a | 766 | if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { |
0ccb2863 | 767 | error = get_write_access(inode); |
3f4d5a00 | 768 | if (unlikely(error)) |
1da177e4 | 769 | goto cleanup_file; |
0ccb2863 | 770 | error = __mnt_want_write(f->f_path.mnt); |
3f4d5a00 | 771 | if (unlikely(error)) { |
0ccb2863 AV |
772 | put_write_access(inode); |
773 | goto cleanup_file; | |
774 | } | |
83f936c7 | 775 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
776 | } |
777 | ||
9c225f26 | 778 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
63b6df14 | 779 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) |
9c225f26 LT |
780 | f->f_mode |= FMODE_ATOMIC_POS; |
781 | ||
1abf0c71 | 782 | f->f_op = fops_get(inode->i_fop); |
7159d544 | 783 | if (WARN_ON(!f->f_op)) { |
72c2d531 AV |
784 | error = -ENODEV; |
785 | goto cleanup_all; | |
786 | } | |
1abf0c71 | 787 | |
e3f20ae2 | 788 | error = security_file_open(f); |
788e7dd4 YN |
789 | if (error) |
790 | goto cleanup_all; | |
791 | ||
c568d683 | 792 | error = break_lease(locks_inode(f), f->f_flags); |
f3c7691e BF |
793 | if (error) |
794 | goto cleanup_all; | |
795 | ||
ea73ea72 AV |
796 | /* normally all 3 are set; ->open() can clear them if needed */ |
797 | f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE; | |
72c2d531 | 798 | if (!open) |
834f2a4a TM |
799 | open = f->f_op->open; |
800 | if (open) { | |
801 | error = open(inode, f); | |
1da177e4 LT |
802 | if (error) |
803 | goto cleanup_all; | |
804 | } | |
f5d11409 | 805 | f->f_mode |= FMODE_OPENED; |
890275b5 MZ |
806 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) |
807 | i_readcount_inc(inode); | |
293bc982 | 808 | if ((f->f_mode & FMODE_READ) && |
84363182 | 809 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 810 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 811 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 812 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 813 | f->f_mode |= FMODE_CAN_WRITE; |
834f2a4a | 814 | |
c75b1d94 | 815 | f->f_write_hint = WRITE_LIFE_NOT_SET; |
1da177e4 LT |
816 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
817 | ||
818 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
af04fadc | 819 | |
69527c55 AV |
820 | /* NB: we're sure to have correct a_ops only after f_op->open */ |
821 | if (f->f_flags & O_DIRECT) { | |
822 | if (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO) | |
823 | return -EINVAL; | |
824 | } | |
09d91cda SL |
825 | |
826 | /* | |
827 | * XXX: Huge page cache doesn't support writing yet. Drop all page | |
828 | * cache for this file before processing writes. | |
829 | */ | |
830 | if ((f->f_mode & FMODE_WRITE) && filemap_nr_thps(inode->i_mapping)) | |
831 | truncate_pagecache(inode, 0); | |
832 | ||
96b7e579 | 833 | return 0; |
1da177e4 LT |
834 | |
835 | cleanup_all: | |
6b4e8085 AV |
836 | if (WARN_ON_ONCE(error > 0)) |
837 | error = -EINVAL; | |
1da177e4 | 838 | fops_put(f->f_op); |
83f936c7 | 839 | if (f->f_mode & FMODE_WRITER) { |
1da177e4 | 840 | put_write_access(inode); |
83f936c7 | 841 | __mnt_drop_write(f->f_path.mnt); |
4a3fd211 | 842 | } |
1da177e4 | 843 | cleanup_file: |
02e5180d AV |
844 | path_put(&f->f_path); |
845 | f->f_path.mnt = NULL; | |
846 | f->f_path.dentry = NULL; | |
dd37978c | 847 | f->f_inode = NULL; |
96b7e579 | 848 | return error; |
1da177e4 LT |
849 | } |
850 | ||
d18e9008 MS |
851 | /** |
852 | * finish_open - finish opening a file | |
0854d450 | 853 | * @file: file pointer |
d18e9008 MS |
854 | * @dentry: pointer to dentry |
855 | * @open: open callback | |
0854d450 | 856 | * @opened: state of open |
d18e9008 MS |
857 | * |
858 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
859 | * | |
860 | * If the open callback is set to NULL, then the standard f_op->open() | |
861 | * filesystem callback is substituted. | |
0854d450 MS |
862 | * |
863 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
864 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
865 | * on it after finish_open(). | |
866 | * | |
0854d450 | 867 | * Returns zero on success or -errno if the open failed. |
d18e9008 | 868 | */ |
30d90494 | 869 | int finish_open(struct file *file, struct dentry *dentry, |
be12af3e | 870 | int (*open)(struct inode *, struct file *)) |
d18e9008 | 871 | { |
aad888f8 | 872 | BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 873 | |
b5bcdda3 | 874 | file->f_path.dentry = dentry; |
aad888f8 | 875 | return do_dentry_open(file, d_backing_inode(dentry), open); |
d18e9008 MS |
876 | } |
877 | EXPORT_SYMBOL(finish_open); | |
878 | ||
879 | /** | |
880 | * finish_no_open - finish ->atomic_open() without opening the file | |
881 | * | |
0854d450 | 882 | * @file: file pointer |
d18e9008 MS |
883 | * @dentry: dentry or NULL (as returned from ->lookup()) |
884 | * | |
885 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
886 | * |
887 | * NB: unlike finish_open() this function does consume the dentry reference and | |
888 | * the caller need not dput() it. | |
889 | * | |
64e1ac4d | 890 | * Returns "0" which must be the return value of ->atomic_open() after having |
0854d450 | 891 | * called this function. |
d18e9008 | 892 | */ |
e45198a6 | 893 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 894 | { |
30d90494 | 895 | file->f_path.dentry = dentry; |
64e1ac4d | 896 | return 0; |
d18e9008 MS |
897 | } |
898 | EXPORT_SYMBOL(finish_no_open); | |
899 | ||
9bf39ab2 MS |
900 | char *file_path(struct file *filp, char *buf, int buflen) |
901 | { | |
902 | return d_path(&filp->f_path, buf, buflen); | |
903 | } | |
904 | EXPORT_SYMBOL(file_path); | |
905 | ||
4bacc9c9 DH |
906 | /** |
907 | * vfs_open - open the file at the given path | |
908 | * @path: path to open | |
909 | * @file: newly allocated file with f_flag initialized | |
910 | * @cred: credentials to use | |
911 | */ | |
ae2bb293 | 912 | int vfs_open(const struct path *path, struct file *file) |
4bacc9c9 | 913 | { |
54d5ca87 | 914 | file->f_path = *path; |
a6518f73 | 915 | return do_dentry_open(file, d_backing_inode(path->dentry), NULL); |
4bacc9c9 DH |
916 | } |
917 | ||
765927b2 | 918 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 919 | const struct cred *cred) |
a1a5b3d9 PS |
920 | { |
921 | int error; | |
922 | struct file *f; | |
923 | ||
e0e81739 DH |
924 | validate_creds(cred); |
925 | ||
c212f9aa | 926 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 927 | BUG_ON(!path->mnt); |
322ee5b3 | 928 | |
ea73ea72 | 929 | f = alloc_empty_file(flags, cred); |
af04fadc | 930 | if (!IS_ERR(f)) { |
ae2bb293 | 931 | error = vfs_open(path, f); |
4d27f326 AV |
932 | if (error) { |
933 | fput(f); | |
af04fadc AV |
934 | f = ERR_PTR(error); |
935 | } | |
2a027e7a AV |
936 | } |
937 | return f; | |
a1a5b3d9 | 938 | } |
1da177e4 LT |
939 | EXPORT_SYMBOL(dentry_open); |
940 | ||
2abc77af AV |
941 | struct file *open_with_fake_path(const struct path *path, int flags, |
942 | struct inode *inode, const struct cred *cred) | |
943 | { | |
d3b1084d | 944 | struct file *f = alloc_empty_file_noaccount(flags, cred); |
2abc77af AV |
945 | if (!IS_ERR(f)) { |
946 | int error; | |
947 | ||
948 | f->f_path = *path; | |
949 | error = do_dentry_open(f, inode, NULL); | |
950 | if (error) { | |
951 | fput(f); | |
952 | f = ERR_PTR(error); | |
953 | } | |
954 | } | |
955 | return f; | |
956 | } | |
957 | EXPORT_SYMBOL(open_with_fake_path); | |
958 | ||
a218d0fd | 959 | static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op) |
47c805dc AV |
960 | { |
961 | int lookup_flags = 0; | |
62fb4a15 | 962 | int acc_mode = ACC_MODE(flags); |
47c805dc | 963 | |
629e014b CH |
964 | /* |
965 | * Clear out all open flags we don't know about so that we don't report | |
966 | * them in fcntl(F_GETFD) or similar interfaces. | |
967 | */ | |
968 | flags &= VALID_OPEN_FLAGS; | |
969 | ||
e305f48b | 970 | if (flags & (O_CREAT | __O_TMPFILE)) |
e68726ff MS |
971 | op->mode = (mode & S_IALLUGO) | S_IFREG; |
972 | else | |
973 | op->mode = 0; | |
47c805dc AV |
974 | |
975 | /* Must never be set by userspace */ | |
c6f3d811 | 976 | flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC; |
47c805dc AV |
977 | |
978 | /* | |
979 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
980 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
981 | * always set instead of having to deal with possibly weird behaviour | |
982 | * for malicious applications setting only __O_SYNC. | |
983 | */ | |
984 | if (flags & __O_SYNC) | |
985 | flags |= O_DSYNC; | |
986 | ||
bb458c64 AV |
987 | if (flags & __O_TMPFILE) { |
988 | if ((flags & O_TMPFILE_MASK) != O_TMPFILE) | |
60545d0d | 989 | return -EINVAL; |
ba57ea64 AV |
990 | if (!(acc_mode & MAY_WRITE)) |
991 | return -EINVAL; | |
60545d0d AV |
992 | } else if (flags & O_PATH) { |
993 | /* | |
994 | * If we have O_PATH in the open flag. Then we | |
995 | * cannot have anything other than the below set of flags | |
996 | */ | |
1abf0c71 AV |
997 | flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH; |
998 | acc_mode = 0; | |
1abf0c71 | 999 | } |
47c805dc | 1000 | |
1abf0c71 | 1001 | op->open_flag = flags; |
47c805dc AV |
1002 | |
1003 | /* O_TRUNC implies we need access checks for write permissions */ | |
1004 | if (flags & O_TRUNC) | |
1005 | acc_mode |= MAY_WRITE; | |
1006 | ||
1007 | /* Allow the LSM permission hook to distinguish append | |
1008 | access from general write access. */ | |
1009 | if (flags & O_APPEND) | |
1010 | acc_mode |= MAY_APPEND; | |
1011 | ||
1012 | op->acc_mode = acc_mode; | |
1013 | ||
1abf0c71 AV |
1014 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
1015 | ||
47c805dc AV |
1016 | if (flags & O_CREAT) { |
1017 | op->intent |= LOOKUP_CREATE; | |
1018 | if (flags & O_EXCL) | |
1019 | op->intent |= LOOKUP_EXCL; | |
1020 | } | |
1021 | ||
1022 | if (flags & O_DIRECTORY) | |
1023 | lookup_flags |= LOOKUP_DIRECTORY; | |
1024 | if (!(flags & O_NOFOLLOW)) | |
1025 | lookup_flags |= LOOKUP_FOLLOW; | |
f9652e10 AV |
1026 | op->lookup_flags = lookup_flags; |
1027 | return 0; | |
47c805dc AV |
1028 | } |
1029 | ||
669abf4e JL |
1030 | /** |
1031 | * file_open_name - open file and return file pointer | |
1032 | * | |
1033 | * @name: struct filename containing path to open | |
1034 | * @flags: open flags as per the open(2) second argument | |
1035 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1036 | * | |
1037 | * This is the helper to open a file from kernelspace if you really | |
1038 | * have to. But in generally you should not do this, so please move | |
1039 | * along, nothing to see here.. | |
1040 | */ | |
1041 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
1042 | { | |
1043 | struct open_flags op; | |
f9652e10 AV |
1044 | int err = build_open_flags(flags, mode, &op); |
1045 | return err ? ERR_PTR(err) : do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
1046 | } |
1047 | ||
47c805dc AV |
1048 | /** |
1049 | * filp_open - open file and return file pointer | |
1050 | * | |
1051 | * @filename: path to open | |
1052 | * @flags: open flags as per the open(2) second argument | |
1053 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1054 | * | |
1055 | * This is the helper to open a file from kernelspace if you really | |
1056 | * have to. But in generally you should not do this, so please move | |
1057 | * along, nothing to see here.. | |
1058 | */ | |
a218d0fd | 1059 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 1060 | { |
51689104 PM |
1061 | struct filename *name = getname_kernel(filename); |
1062 | struct file *file = ERR_CAST(name); | |
1063 | ||
1064 | if (!IS_ERR(name)) { | |
1065 | file = file_open_name(name, flags, mode); | |
1066 | putname(name); | |
1067 | } | |
1068 | return file; | |
47c805dc AV |
1069 | } |
1070 | EXPORT_SYMBOL(filp_open); | |
1071 | ||
73d049a4 | 1072 | struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt, |
378c6520 | 1073 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1074 | { |
1075 | struct open_flags op; | |
378c6520 | 1076 | int err = build_open_flags(flags, mode, &op); |
f9652e10 AV |
1077 | if (err) |
1078 | return ERR_PTR(err); | |
f9652e10 | 1079 | return do_file_open_root(dentry, mnt, filename, &op); |
73d049a4 AV |
1080 | } |
1081 | EXPORT_SYMBOL(file_open_root); | |
1082 | ||
a218d0fd | 1083 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
1da177e4 | 1084 | { |
47c805dc | 1085 | struct open_flags op; |
f9652e10 AV |
1086 | int fd = build_open_flags(flags, mode, &op); |
1087 | struct filename *tmp; | |
1088 | ||
1089 | if (fd) | |
1090 | return fd; | |
1091 | ||
1092 | tmp = getname(filename); | |
1093 | if (IS_ERR(tmp)) | |
1094 | return PTR_ERR(tmp); | |
1095 | ||
1096 | fd = get_unused_fd_flags(flags); | |
1097 | if (fd >= 0) { | |
1098 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1099 | if (IS_ERR(f)) { | |
1100 | put_unused_fd(fd); | |
1101 | fd = PTR_ERR(f); | |
1102 | } else { | |
1103 | fsnotify_open(f); | |
1104 | fd_install(fd, f); | |
fcc38bd3 | 1105 | trace_do_sys_open(tmp->name, flags, mode); |
1da177e4 | 1106 | } |
1da177e4 | 1107 | } |
f9652e10 | 1108 | putname(tmp); |
1da177e4 | 1109 | return fd; |
1da177e4 | 1110 | } |
e922efc3 | 1111 | |
a218d0fd | 1112 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) |
e922efc3 MS |
1113 | { |
1114 | if (force_o_largefile()) | |
1115 | flags |= O_LARGEFILE; | |
1116 | ||
2cf09666 | 1117 | return do_sys_open(AT_FDCWD, filename, flags, mode); |
e922efc3 | 1118 | } |
1da177e4 | 1119 | |
6559eed8 | 1120 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1121 | umode_t, mode) |
5590ff0d UD |
1122 | { |
1123 | if (force_o_largefile()) | |
1124 | flags |= O_LARGEFILE; | |
1125 | ||
2cf09666 | 1126 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1127 | } |
5590ff0d | 1128 | |
e35d49f6 AV |
1129 | #ifdef CONFIG_COMPAT |
1130 | /* | |
1131 | * Exactly like sys_open(), except that it doesn't set the | |
1132 | * O_LARGEFILE flag. | |
1133 | */ | |
1134 | COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1135 | { | |
1136 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
1137 | } | |
1138 | ||
1139 | /* | |
1140 | * Exactly like sys_openat(), except that it doesn't set the | |
1141 | * O_LARGEFILE flag. | |
1142 | */ | |
1143 | COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode) | |
1144 | { | |
1145 | return do_sys_open(dfd, filename, flags, mode); | |
1146 | } | |
1147 | #endif | |
1148 | ||
1da177e4 LT |
1149 | #ifndef __alpha__ |
1150 | ||
1151 | /* | |
1152 | * For backward compatibility? Maybe this should be moved | |
1153 | * into arch/i386 instead? | |
1154 | */ | |
a218d0fd | 1155 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 | 1156 | { |
bae217ea | 1157 | return ksys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); |
1da177e4 LT |
1158 | } |
1159 | ||
1160 | #endif | |
1161 | ||
1162 | /* | |
1163 | * "id" is the POSIX thread ID. We use the | |
1164 | * files pointer for this.. | |
1165 | */ | |
1166 | int filp_close(struct file *filp, fl_owner_t id) | |
1167 | { | |
45778ca8 | 1168 | int retval = 0; |
1da177e4 LT |
1169 | |
1170 | if (!file_count(filp)) { | |
1171 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1172 | return 0; |
1da177e4 LT |
1173 | } |
1174 | ||
72c2d531 | 1175 | if (filp->f_op->flush) |
75e1fcc0 | 1176 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1177 | |
1abf0c71 AV |
1178 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1179 | dnotify_flush(filp, id); | |
1180 | locks_remove_posix(filp, id); | |
1181 | } | |
1da177e4 LT |
1182 | fput(filp); |
1183 | return retval; | |
1184 | } | |
1185 | ||
1186 | EXPORT_SYMBOL(filp_close); | |
1187 | ||
1188 | /* | |
1189 | * Careful here! We test whether the file pointer is NULL before | |
1190 | * releasing the fd. This ensures that one clone task can't release | |
1191 | * an fd while another clone is opening it. | |
1192 | */ | |
ca013e94 | 1193 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1194 | { |
483ce1d4 | 1195 | int retval = __close_fd(current->files, fd); |
ee731f4f EP |
1196 | |
1197 | /* can't restart close syscall because file table entry was cleared */ | |
1198 | if (unlikely(retval == -ERESTARTSYS || | |
1199 | retval == -ERESTARTNOINTR || | |
1200 | retval == -ERESTARTNOHAND || | |
1201 | retval == -ERESTART_RESTARTBLOCK)) | |
1202 | retval = -EINTR; | |
1203 | ||
1204 | return retval; | |
1da177e4 | 1205 | } |
1da177e4 LT |
1206 | |
1207 | /* | |
1208 | * This routine simulates a hangup on the tty, to arrange that users | |
1209 | * are given clean terminals at login time. | |
1210 | */ | |
ca013e94 | 1211 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1212 | { |
1213 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1214 | tty_vhangup_self(); |
1da177e4 LT |
1215 | return 0; |
1216 | } | |
1217 | return -EPERM; | |
1218 | } | |
1219 | ||
1220 | /* | |
1221 | * Called when an inode is about to be open. | |
1222 | * We use this to disallow opening large files on 32bit systems if | |
1223 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1224 | * on this flag in sys_open. | |
1225 | */ | |
1226 | int generic_file_open(struct inode * inode, struct file * filp) | |
1227 | { | |
1228 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1229 | return -EOVERFLOW; |
1da177e4 LT |
1230 | return 0; |
1231 | } | |
1232 | ||
1233 | EXPORT_SYMBOL(generic_file_open); | |
1234 | ||
1235 | /* | |
1236 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1237 | * file descriptors. The function is not supposed to ever fail, the only |
1238 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1239 | * directly into file_operations structure. | |
1da177e4 LT |
1240 | */ |
1241 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1242 | { | |
1243 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1244 | return 0; | |
1245 | } | |
1246 | ||
1247 | EXPORT_SYMBOL(nonseekable_open); | |
10dce8af KS |
1248 | |
1249 | /* | |
1250 | * stream_open is used by subsystems that want stream-like file descriptors. | |
1251 | * Such file descriptors are not seekable and don't have notion of position | |
438ab720 KS |
1252 | * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL). |
1253 | * Contrary to file descriptors of other regular files, .read() and .write() | |
1254 | * can run simultaneously. | |
10dce8af KS |
1255 | * |
1256 | * stream_open never fails and is marked to return int so that it could be | |
1257 | * directly used as file_operations.open . | |
1258 | */ | |
1259 | int stream_open(struct inode *inode, struct file *filp) | |
1260 | { | |
1261 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS); | |
1262 | filp->f_mode |= FMODE_STREAM; | |
1263 | return 0; | |
1264 | } | |
1265 | ||
1266 | EXPORT_SYMBOL(stream_open); |