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> |
2dfc1cae | 32 | #include <linux/dnotify.h> |
3f6d078d | 33 | #include <linux/compat.h> |
a793d79e | 34 | #include <linux/mnt_idmapping.h> |
5970e15d | 35 | #include <linux/filelock.h> |
1da177e4 | 36 | |
e81e3f4d EP |
37 | #include "internal.h" |
38 | ||
abf08576 | 39 | int do_truncate(struct mnt_idmap *idmap, struct dentry *dentry, |
643fe55a | 40 | loff_t length, unsigned int time_attrs, struct file *filp) |
1da177e4 | 41 | { |
939a9421 | 42 | int ret; |
1da177e4 LT |
43 | struct iattr newattrs; |
44 | ||
45 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
46 | if (length < 0) | |
47 | return -EINVAL; | |
48 | ||
49 | newattrs.ia_size = length; | |
4a30131e | 50 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
51 | if (filp) { |
52 | newattrs.ia_file = filp; | |
53 | newattrs.ia_valid |= ATTR_FILE; | |
54 | } | |
1da177e4 | 55 | |
45f147a1 | 56 | /* Remove suid, sgid, and file capabilities on truncate too */ |
9452e93e | 57 | ret = dentry_needs_remove_privs(idmap, dentry); |
45f147a1 JK |
58 | if (ret < 0) |
59 | return ret; | |
939a9421 AW |
60 | if (ret) |
61 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 62 | |
5955102c | 63 | inode_lock(dentry->d_inode); |
27ac0ffe | 64 | /* Note any delegations or leases have already been broken: */ |
abf08576 | 65 | ret = notify_change(idmap, dentry, &newattrs, NULL); |
5955102c | 66 | inode_unlock(dentry->d_inode); |
939a9421 | 67 | return ret; |
1da177e4 LT |
68 | } |
69 | ||
7df818b2 | 70 | long vfs_truncate(const struct path *path, loff_t length) |
1da177e4 | 71 | { |
abf08576 | 72 | struct mnt_idmap *idmap; |
2d8f3038 | 73 | struct inode *inode; |
a02de960 | 74 | long error; |
1da177e4 | 75 | |
a02de960 | 76 | inode = path->dentry->d_inode; |
1da177e4 LT |
77 | |
78 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
1da177e4 | 79 | if (S_ISDIR(inode->i_mode)) |
a02de960 | 80 | return -EISDIR; |
1da177e4 | 81 | if (!S_ISREG(inode->i_mode)) |
a02de960 | 82 | return -EINVAL; |
1da177e4 | 83 | |
a02de960 | 84 | error = mnt_want_write(path->mnt); |
1da177e4 | 85 | if (error) |
a02de960 | 86 | goto out; |
1da177e4 | 87 | |
abf08576 | 88 | idmap = mnt_idmap(path->mnt); |
4609e1f1 | 89 | error = inode_permission(idmap, inode, MAY_WRITE); |
9ac9b847 DH |
90 | if (error) |
91 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
92 | |
93 | error = -EPERM; | |
c82e42da | 94 | if (IS_APPEND(inode)) |
9ac9b847 | 95 | goto mnt_drop_write_and_out; |
1da177e4 | 96 | |
8cf9ee50 | 97 | error = get_write_access(inode); |
1da177e4 | 98 | if (error) |
9ac9b847 | 99 | goto mnt_drop_write_and_out; |
1da177e4 | 100 | |
9700382c | 101 | /* |
102 | * Make sure that there are no leases. get_write_access() protects | |
103 | * against the truncate racing with a lease-granting setlease(). | |
104 | */ | |
8737c930 | 105 | error = break_lease(inode, O_WRONLY); |
1da177e4 | 106 | if (error) |
9700382c | 107 | goto put_write_and_out; |
1da177e4 | 108 | |
f7e33bdb | 109 | error = security_path_truncate(path); |
907f4554 | 110 | if (!error) |
abf08576 | 111 | error = do_truncate(idmap, path->dentry, length, 0, NULL); |
1da177e4 | 112 | |
9700382c | 113 | put_write_and_out: |
8cf9ee50 | 114 | put_write_access(inode); |
9ac9b847 | 115 | mnt_drop_write_and_out: |
a02de960 | 116 | mnt_drop_write(path->mnt); |
1da177e4 LT |
117 | out: |
118 | return error; | |
119 | } | |
a02de960 DH |
120 | EXPORT_SYMBOL_GPL(vfs_truncate); |
121 | ||
df260e21 | 122 | long do_sys_truncate(const char __user *pathname, loff_t length) |
a02de960 | 123 | { |
48f7530d | 124 | unsigned int lookup_flags = LOOKUP_FOLLOW; |
a02de960 DH |
125 | struct path path; |
126 | int error; | |
127 | ||
128 | if (length < 0) /* sorry, but loff_t says... */ | |
129 | return -EINVAL; | |
130 | ||
48f7530d JL |
131 | retry: |
132 | error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path); | |
a02de960 DH |
133 | if (!error) { |
134 | error = vfs_truncate(&path, length); | |
135 | path_put(&path); | |
136 | } | |
48f7530d JL |
137 | if (retry_estale(error, lookup_flags)) { |
138 | lookup_flags |= LOOKUP_REVAL; | |
139 | goto retry; | |
140 | } | |
a02de960 DH |
141 | return error; |
142 | } | |
1da177e4 | 143 | |
4fd8da8d | 144 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 145 | { |
4fd8da8d | 146 | return do_sys_truncate(path, length); |
1da177e4 LT |
147 | } |
148 | ||
3f6d078d AV |
149 | #ifdef CONFIG_COMPAT |
150 | COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length) | |
151 | { | |
152 | return do_sys_truncate(path, length); | |
153 | } | |
154 | #endif | |
155 | ||
5f0d594c | 156 | long do_ftruncate(struct file *file, loff_t length, int small) |
1da177e4 | 157 | { |
bf2965d5 | 158 | struct inode *inode; |
1da177e4 | 159 | struct dentry *dentry; |
1da177e4 LT |
160 | int error; |
161 | ||
1da177e4 | 162 | /* explicitly opened as large or we are on 64-bit box */ |
5f0d594c | 163 | if (file->f_flags & O_LARGEFILE) |
1da177e4 LT |
164 | small = 0; |
165 | ||
5f0d594c | 166 | dentry = file->f_path.dentry; |
1da177e4 | 167 | inode = dentry->d_inode; |
5f0d594c TS |
168 | if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE)) |
169 | return -EINVAL; | |
1da177e4 | 170 | |
1da177e4 LT |
171 | /* Cannot ftruncate over 2^31 bytes without large file support */ |
172 | if (small && length > MAX_NON_LFS) | |
5f0d594c | 173 | return -EINVAL; |
1da177e4 | 174 | |
78757af6 | 175 | /* Check IS_APPEND on real upper inode */ |
5f0d594c TS |
176 | if (IS_APPEND(file_inode(file))) |
177 | return -EPERM; | |
14da9200 | 178 | sb_start_write(inode->i_sb); |
5f0d594c | 179 | error = security_file_truncate(file); |
1da177e4 | 180 | if (!error) |
5f0d594c TS |
181 | error = do_truncate(file_mnt_idmap(file), dentry, length, |
182 | ATTR_MTIME | ATTR_CTIME, file); | |
14da9200 | 183 | sb_end_write(inode->i_sb); |
5f0d594c TS |
184 | |
185 | return error; | |
186 | } | |
187 | ||
188 | long do_sys_ftruncate(unsigned int fd, loff_t length, int small) | |
189 | { | |
190 | struct fd f; | |
191 | int error; | |
192 | ||
193 | if (length < 0) | |
194 | return -EINVAL; | |
195 | f = fdget(fd); | |
196 | if (!f.file) | |
197 | return -EBADF; | |
198 | ||
199 | error = do_ftruncate(f.file, length, small); | |
200 | ||
2903ff01 | 201 | fdput(f); |
1da177e4 LT |
202 | return error; |
203 | } | |
204 | ||
4b8e88e5 | 205 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, off_t, length) |
1da177e4 | 206 | { |
2cf09666 | 207 | return do_sys_ftruncate(fd, length, 1); |
1da177e4 LT |
208 | } |
209 | ||
3f6d078d | 210 | #ifdef CONFIG_COMPAT |
4b8e88e5 | 211 | COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_off_t, length) |
3f6d078d AV |
212 | { |
213 | return do_sys_ftruncate(fd, length, 1); | |
214 | } | |
215 | #endif | |
216 | ||
1da177e4 LT |
217 | /* LFS versions of truncate are only needed on 32 bit machines */ |
218 | #if BITS_PER_LONG == 32 | |
4a0fd5bf | 219 | SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length) |
1da177e4 LT |
220 | { |
221 | return do_sys_truncate(path, length); | |
222 | } | |
223 | ||
4a0fd5bf | 224 | SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length) |
1da177e4 | 225 | { |
2cf09666 | 226 | return do_sys_ftruncate(fd, length, 0); |
1da177e4 | 227 | } |
6673e0c3 | 228 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 229 | |
59c10c52 GR |
230 | #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_TRUNCATE64) |
231 | COMPAT_SYSCALL_DEFINE3(truncate64, const char __user *, pathname, | |
232 | compat_arg_u64_dual(length)) | |
233 | { | |
234 | return ksys_truncate(pathname, compat_arg_u64_glue(length)); | |
235 | } | |
236 | #endif | |
237 | ||
238 | #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_FTRUNCATE64) | |
239 | COMPAT_SYSCALL_DEFINE3(ftruncate64, unsigned int, fd, | |
240 | compat_arg_u64_dual(length)) | |
241 | { | |
242 | return ksys_ftruncate(fd, compat_arg_u64_glue(length)); | |
243 | } | |
244 | #endif | |
3e63cbb1 | 245 | |
72c72bdf | 246 | int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len) |
97ac7350 | 247 | { |
496ad9aa | 248 | struct inode *inode = file_inode(file); |
3e63cbb1 | 249 | long ret; |
23cc6ef6 | 250 | loff_t sum; |
97ac7350 AA |
251 | |
252 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 253 | return -EINVAL; |
97ac7350 | 254 | |
57413d8e | 255 | if (mode & ~(FALLOC_FL_MODE_MASK | FALLOC_FL_KEEP_SIZE)) |
409332b6 LC |
256 | return -EOPNOTSUPP; |
257 | ||
57413d8e CH |
258 | /* |
259 | * Modes are exclusive, even if that is not obvious from the encoding | |
260 | * as bit masks and the mix with the flag in the same namespace. | |
261 | * | |
262 | * To make things even more complicated, FALLOC_FL_ALLOCATE_RANGE is | |
263 | * encoded as no bit set. | |
264 | */ | |
265 | switch (mode & FALLOC_FL_MODE_MASK) { | |
266 | case FALLOC_FL_ALLOCATE_RANGE: | |
267 | case FALLOC_FL_UNSHARE_RANGE: | |
268 | case FALLOC_FL_ZERO_RANGE: | |
269 | break; | |
270 | case FALLOC_FL_PUNCH_HOLE: | |
271 | if (!(mode & FALLOC_FL_KEEP_SIZE)) | |
272 | return -EOPNOTSUPP; | |
273 | break; | |
274 | case FALLOC_FL_COLLAPSE_RANGE: | |
275 | case FALLOC_FL_INSERT_RANGE: | |
276 | if (mode & FALLOC_FL_KEEP_SIZE) | |
277 | return -EOPNOTSUPP; | |
278 | break; | |
279 | default: | |
3e63cbb1 | 280 | return -EOPNOTSUPP; |
57413d8e | 281 | } |
71be6b49 | 282 | |
97ac7350 | 283 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 284 | return -EBADF; |
1ca551c6 | 285 | |
00f5e619 | 286 | /* |
57413d8e | 287 | * On append-only files only space preallocation is supported. |
00f5e619 | 288 | */ |
8fc61d92 | 289 | if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode)) |
1ca551c6 MS |
290 | return -EPERM; |
291 | ||
292 | if (IS_IMMUTABLE(inode)) | |
293 | return -EPERM; | |
294 | ||
0790b31b | 295 | /* |
6d2b6170 | 296 | * We cannot allow any fallocate operation on an active swapfile |
0790b31b LC |
297 | */ |
298 | if (IS_SWAPFILE(inode)) | |
6d2b6170 | 299 | return -ETXTBSY; |
0790b31b | 300 | |
97ac7350 AA |
301 | /* |
302 | * Revalidate the write permissions, in case security policy has | |
303 | * changed since the files were opened. | |
304 | */ | |
305 | ret = security_file_permission(file, MAY_WRITE); | |
306 | if (ret) | |
3e63cbb1 | 307 | return ret; |
97ac7350 | 308 | |
d9e5d310 AG |
309 | ret = fsnotify_file_area_perm(file, MAY_WRITE, &offset, len); |
310 | if (ret) | |
311 | return ret; | |
312 | ||
97ac7350 | 313 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 314 | return -ESPIPE; |
97ac7350 | 315 | |
9e79b132 AG |
316 | if (S_ISDIR(inode->i_mode)) |
317 | return -EISDIR; | |
318 | ||
319 | if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode)) | |
3e63cbb1 | 320 | return -ENODEV; |
97ac7350 | 321 | |
23cc6ef6 JS |
322 | /* Check for wraparound */ |
323 | if (check_add_overflow(offset, len, &sum)) | |
324 | return -EFBIG; | |
325 | ||
326 | if (sum > inode->i_sb->s_maxbytes) | |
3e63cbb1 | 327 | return -EFBIG; |
97ac7350 | 328 | |
2fe17c10 | 329 | if (!file->f_op->fallocate) |
3e63cbb1 | 330 | return -EOPNOTSUPP; |
97ac7350 | 331 | |
bfe219d3 | 332 | file_start_write(file); |
14da9200 | 333 | ret = file->f_op->fallocate(file, mode, offset, len); |
820c12d5 HS |
334 | |
335 | /* | |
336 | * Create inotify and fanotify events. | |
337 | * | |
338 | * To keep the logic simple always create events if fallocate succeeds. | |
339 | * This implies that events are even created if the file size remains | |
340 | * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE. | |
341 | */ | |
342 | if (ret == 0) | |
343 | fsnotify_modify(file); | |
344 | ||
bfe219d3 | 345 | file_end_write(file); |
14da9200 | 346 | return ret; |
3e63cbb1 | 347 | } |
72c72bdf | 348 | EXPORT_SYMBOL_GPL(vfs_fallocate); |
3e63cbb1 | 349 | |
edf292c7 | 350 | int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len) |
3e63cbb1 | 351 | { |
2903ff01 | 352 | struct fd f = fdget(fd); |
3e63cbb1 AJ |
353 | int error = -EBADF; |
354 | ||
2903ff01 | 355 | if (f.file) { |
72c72bdf | 356 | error = vfs_fallocate(f.file, mode, offset, len); |
2903ff01 | 357 | fdput(f); |
3e63cbb1 | 358 | } |
3e63cbb1 | 359 | return error; |
97ac7350 | 360 | } |
3e63cbb1 | 361 | |
edf292c7 DB |
362 | SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len) |
363 | { | |
364 | return ksys_fallocate(fd, mode, offset, len); | |
365 | } | |
366 | ||
59c10c52 GR |
367 | #if defined(CONFIG_COMPAT) && defined(__ARCH_WANT_COMPAT_FALLOCATE) |
368 | COMPAT_SYSCALL_DEFINE6(fallocate, int, fd, int, mode, compat_arg_u64_dual(offset), | |
369 | compat_arg_u64_dual(len)) | |
370 | { | |
371 | return ksys_fallocate(fd, mode, compat_arg_u64_glue(offset), | |
372 | compat_arg_u64_glue(len)); | |
373 | } | |
374 | #endif | |
375 | ||
1da177e4 LT |
376 | /* |
377 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
378 | * We do this by temporarily clearing all FS-related capabilities and | |
379 | * switching the fsuid/fsgid around to the real ones. | |
981ee95c MG |
380 | * |
381 | * Creating new credentials is expensive, so we try to skip doing it, | |
382 | * which we can if the result would match what we already got. | |
1da177e4 | 383 | */ |
981ee95c MG |
384 | static bool access_need_override_creds(int flags) |
385 | { | |
386 | const struct cred *cred; | |
387 | ||
388 | if (flags & AT_EACCESS) | |
389 | return false; | |
390 | ||
391 | cred = current_cred(); | |
392 | if (!uid_eq(cred->fsuid, cred->uid) || | |
393 | !gid_eq(cred->fsgid, cred->gid)) | |
394 | return true; | |
395 | ||
396 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { | |
397 | kuid_t root_uid = make_kuid(cred->user_ns, 0); | |
398 | if (!uid_eq(cred->uid, root_uid)) { | |
399 | if (!cap_isclear(cred->cap_effective)) | |
400 | return true; | |
401 | } else { | |
402 | if (!cap_isidentical(cred->cap_effective, | |
403 | cred->cap_permitted)) | |
404 | return true; | |
405 | } | |
406 | } | |
407 | ||
408 | return false; | |
409 | } | |
410 | ||
94704515 | 411 | static const struct cred *access_override_creds(void) |
1da177e4 | 412 | { |
d84f4f99 DH |
413 | const struct cred *old_cred; |
414 | struct cred *override_cred; | |
1da177e4 | 415 | |
d84f4f99 DH |
416 | override_cred = prepare_creds(); |
417 | if (!override_cred) | |
94704515 | 418 | return NULL; |
1da177e4 | 419 | |
981ee95c MG |
420 | /* |
421 | * XXX access_need_override_creds performs checks in hopes of skipping | |
422 | * this work. Make sure it stays in sync if making any changes in this | |
423 | * routine. | |
424 | */ | |
425 | ||
d84f4f99 DH |
426 | override_cred->fsuid = override_cred->uid; |
427 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 428 | |
086f7316 | 429 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 430 | /* Clear the capabilities if we switch to a non-root user */ |
18815a18 EB |
431 | kuid_t root_uid = make_kuid(override_cred->user_ns, 0); |
432 | if (!uid_eq(override_cred->uid, root_uid)) | |
d84f4f99 | 433 | cap_clear(override_cred->cap_effective); |
086f7316 | 434 | else |
d84f4f99 DH |
435 | override_cred->cap_effective = |
436 | override_cred->cap_permitted; | |
086f7316 | 437 | } |
1da177e4 | 438 | |
d7852fbd LT |
439 | /* |
440 | * The new set of credentials can *only* be used in | |
441 | * task-synchronous circumstances, and does not need | |
442 | * RCU freeing, unless somebody then takes a separate | |
443 | * reference to it. | |
444 | * | |
445 | * NOTE! This is _only_ true because this credential | |
446 | * is used purely for override_creds() that installs | |
447 | * it as the subjective cred. Other threads will be | |
448 | * accessing ->real_cred, not the subjective cred. | |
449 | * | |
450 | * If somebody _does_ make a copy of this (using the | |
451 | * 'get_current_cred()' function), that will clear the | |
452 | * non_rcu field, because now that other user may be | |
453 | * expecting RCU freeing. But normal thread-synchronous | |
d2185690 BS |
454 | * cred accesses will keep things non-racy to avoid RCU |
455 | * freeing. | |
d7852fbd LT |
456 | */ |
457 | override_cred->non_rcu = 1; | |
458 | ||
d84f4f99 | 459 | old_cred = override_creds(override_cred); |
94704515 MS |
460 | |
461 | /* override_cred() gets its own ref */ | |
462 | put_cred(override_cred); | |
463 | ||
464 | return old_cred; | |
465 | } | |
466 | ||
eb9d7d39 | 467 | static long do_faccessat(int dfd, const char __user *filename, int mode, int flags) |
94704515 MS |
468 | { |
469 | struct path path; | |
470 | struct inode *inode; | |
471 | int res; | |
472 | unsigned int lookup_flags = LOOKUP_FOLLOW; | |
c8ffd8bc | 473 | const struct cred *old_cred = NULL; |
94704515 MS |
474 | |
475 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
476 | return -EINVAL; | |
477 | ||
c8ffd8bc MS |
478 | if (flags & ~(AT_EACCESS | AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) |
479 | return -EINVAL; | |
480 | ||
481 | if (flags & AT_SYMLINK_NOFOLLOW) | |
482 | lookup_flags &= ~LOOKUP_FOLLOW; | |
483 | if (flags & AT_EMPTY_PATH) | |
484 | lookup_flags |= LOOKUP_EMPTY; | |
485 | ||
981ee95c | 486 | if (access_need_override_creds(flags)) { |
c8ffd8bc MS |
487 | old_cred = access_override_creds(); |
488 | if (!old_cred) | |
489 | return -ENOMEM; | |
490 | } | |
94704515 | 491 | |
87fa5595 JL |
492 | retry: |
493 | res = user_path_at(dfd, filename, lookup_flags, &path); | |
6902d925 DH |
494 | if (res) |
495 | goto out; | |
496 | ||
63afdfc7 | 497 | inode = d_backing_inode(path.dentry); |
256984a8 AV |
498 | |
499 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
500 | /* |
501 | * MAY_EXEC on regular files is denied if the fs is mounted | |
502 | * with the "noexec" flag. | |
503 | */ | |
504 | res = -EACCES; | |
90f8572b | 505 | if (path_noexec(&path)) |
30524472 AV |
506 | goto out_path_release; |
507 | } | |
508 | ||
4609e1f1 | 509 | res = inode_permission(mnt_idmap(path.mnt), inode, mode | MAY_ACCESS); |
6902d925 | 510 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 511 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 512 | goto out_path_release; |
2f676cbc DH |
513 | /* |
514 | * This is a rare case where using __mnt_is_readonly() | |
515 | * is OK without a mnt_want/drop_write() pair. Since | |
516 | * no actual write to the fs is performed here, we do | |
517 | * not need to telegraph to that to anyone. | |
518 | * | |
519 | * By doing this, we accept that this access is | |
520 | * inherently racy and know that the fs may change | |
521 | * state before we even see this result. | |
522 | */ | |
2d8f3038 | 523 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 524 | res = -EROFS; |
1da177e4 | 525 | |
6902d925 | 526 | out_path_release: |
2d8f3038 | 527 | path_put(&path); |
87fa5595 JL |
528 | if (retry_estale(res, lookup_flags)) { |
529 | lookup_flags |= LOOKUP_REVAL; | |
530 | goto retry; | |
531 | } | |
6902d925 | 532 | out: |
c8ffd8bc MS |
533 | if (old_cred) |
534 | revert_creds(old_cred); | |
535 | ||
1da177e4 LT |
536 | return res; |
537 | } | |
538 | ||
cbfe20f5 DB |
539 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
540 | { | |
c8ffd8bc MS |
541 | return do_faccessat(dfd, filename, mode, 0); |
542 | } | |
543 | ||
544 | SYSCALL_DEFINE4(faccessat2, int, dfd, const char __user *, filename, int, mode, | |
545 | int, flags) | |
546 | { | |
547 | return do_faccessat(dfd, filename, mode, flags); | |
cbfe20f5 DB |
548 | } |
549 | ||
ca013e94 | 550 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d | 551 | { |
c8ffd8bc | 552 | return do_faccessat(AT_FDCWD, filename, mode, 0); |
5590ff0d UD |
553 | } |
554 | ||
db63f1e3 | 555 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 556 | { |
2d8f3038 | 557 | struct path path; |
1da177e4 | 558 | int error; |
0291c0a5 JL |
559 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
560 | retry: | |
561 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
562 | if (error) |
563 | goto out; | |
564 | ||
02f92b38 | 565 | error = path_permission(&path, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
566 | if (error) |
567 | goto dput_and_out; | |
568 | ||
2d8f3038 | 569 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
570 | |
571 | dput_and_out: | |
2d8f3038 | 572 | path_put(&path); |
0291c0a5 JL |
573 | if (retry_estale(error, lookup_flags)) { |
574 | lookup_flags |= LOOKUP_REVAL; | |
575 | goto retry; | |
576 | } | |
1da177e4 LT |
577 | out: |
578 | return error; | |
579 | } | |
580 | ||
3cdad428 | 581 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 | 582 | { |
2903ff01 | 583 | struct fd f = fdget_raw(fd); |
159b0956 | 584 | int error; |
1da177e4 LT |
585 | |
586 | error = -EBADF; | |
2903ff01 | 587 | if (!f.file) |
1da177e4 LT |
588 | goto out; |
589 | ||
1da177e4 | 590 | error = -ENOTDIR; |
159b0956 | 591 | if (!d_can_lookup(f.file->f_path.dentry)) |
1da177e4 LT |
592 | goto out_putf; |
593 | ||
02f92b38 | 594 | error = file_permission(f.file, MAY_EXEC | MAY_CHDIR); |
1da177e4 | 595 | if (!error) |
2903ff01 | 596 | set_fs_pwd(current->fs, &f.file->f_path); |
1da177e4 | 597 | out_putf: |
2903ff01 | 598 | fdput(f); |
1da177e4 LT |
599 | out: |
600 | return error; | |
601 | } | |
602 | ||
4b7ca501 | 603 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 604 | { |
2d8f3038 | 605 | struct path path; |
1da177e4 | 606 | int error; |
2771261e JL |
607 | unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY; |
608 | retry: | |
609 | error = user_path_at(AT_FDCWD, filename, lookup_flags, &path); | |
1da177e4 LT |
610 | if (error) |
611 | goto out; | |
612 | ||
02f92b38 | 613 | error = path_permission(&path, MAY_EXEC | MAY_CHDIR); |
1da177e4 LT |
614 | if (error) |
615 | goto dput_and_out; | |
616 | ||
617 | error = -EPERM; | |
c7b96acf | 618 | if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT)) |
1da177e4 | 619 | goto dput_and_out; |
8b8efb44 TH |
620 | error = security_path_chroot(&path); |
621 | if (error) | |
622 | goto dput_and_out; | |
1da177e4 | 623 | |
2d8f3038 | 624 | set_fs_root(current->fs, &path); |
1da177e4 LT |
625 | error = 0; |
626 | dput_and_out: | |
2d8f3038 | 627 | path_put(&path); |
2771261e JL |
628 | if (retry_estale(error, lookup_flags)) { |
629 | lookup_flags |= LOOKUP_REVAL; | |
630 | goto retry; | |
631 | } | |
1da177e4 LT |
632 | out: |
633 | return error; | |
634 | } | |
635 | ||
1097742e | 636 | int chmod_common(const struct path *path, umode_t mode) |
1da177e4 | 637 | { |
e57712eb | 638 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 639 | struct inode *delegated_inode = NULL; |
1da177e4 | 640 | struct iattr newattrs; |
e57712eb | 641 | int error; |
1da177e4 | 642 | |
e57712eb AV |
643 | error = mnt_want_write(path->mnt); |
644 | if (error) | |
645 | return error; | |
27ac0ffe | 646 | retry_deleg: |
5955102c | 647 | inode_lock(inode); |
cdcf116d | 648 | error = security_path_chmod(path, mode); |
e57712eb | 649 | if (error) |
fe542cf5 | 650 | goto out_unlock; |
1da177e4 LT |
651 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); |
652 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
abf08576 | 653 | error = notify_change(mnt_idmap(path->mnt), path->dentry, |
b8b546a0 | 654 | &newattrs, &delegated_inode); |
fe542cf5 | 655 | out_unlock: |
5955102c | 656 | inode_unlock(inode); |
27ac0ffe BF |
657 | if (delegated_inode) { |
658 | error = break_deleg_wait(&delegated_inode); | |
659 | if (!error) | |
660 | goto retry_deleg; | |
661 | } | |
e57712eb AV |
662 | mnt_drop_write(path->mnt); |
663 | return error; | |
664 | } | |
665 | ||
9e96c8c0 CH |
666 | int vfs_fchmod(struct file *file, umode_t mode) |
667 | { | |
668 | audit_file(file); | |
669 | return chmod_common(&file->f_path, mode); | |
670 | } | |
671 | ||
b25ba7c3 | 672 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode) |
e57712eb | 673 | { |
173c8401 | 674 | struct fd f = fdget(fd); |
e57712eb AV |
675 | int err = -EBADF; |
676 | ||
173c8401 | 677 | if (f.file) { |
9e96c8c0 | 678 | err = vfs_fchmod(f.file, mode); |
173c8401 | 679 | fdput(f); |
e57712eb | 680 | } |
1da177e4 LT |
681 | return err; |
682 | } | |
683 | ||
09da082b AG |
684 | static int do_fchmodat(int dfd, const char __user *filename, umode_t mode, |
685 | unsigned int flags) | |
1da177e4 | 686 | { |
2d8f3038 | 687 | struct path path; |
1da177e4 | 688 | int error; |
09da082b AG |
689 | unsigned int lookup_flags; |
690 | ||
5daeb41a | 691 | if (unlikely(flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH))) |
09da082b AG |
692 | return -EINVAL; |
693 | ||
694 | lookup_flags = (flags & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; | |
5daeb41a AS |
695 | if (flags & AT_EMPTY_PATH) |
696 | lookup_flags |= LOOKUP_EMPTY; | |
09da082b | 697 | |
14ff690c JL |
698 | retry: |
699 | error = user_path_at(dfd, filename, lookup_flags, &path); | |
e57712eb AV |
700 | if (!error) { |
701 | error = chmod_common(&path, mode); | |
702 | path_put(&path); | |
14ff690c JL |
703 | if (retry_estale(error, lookup_flags)) { |
704 | lookup_flags |= LOOKUP_REVAL; | |
705 | goto retry; | |
706 | } | |
e57712eb | 707 | } |
1da177e4 LT |
708 | return error; |
709 | } | |
710 | ||
09da082b AG |
711 | SYSCALL_DEFINE4(fchmodat2, int, dfd, const char __user *, filename, |
712 | umode_t, mode, unsigned int, flags) | |
713 | { | |
714 | return do_fchmodat(dfd, filename, mode, flags); | |
715 | } | |
716 | ||
03450e27 DB |
717 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, |
718 | umode_t, mode) | |
719 | { | |
09da082b | 720 | return do_fchmodat(dfd, filename, mode, 0); |
03450e27 DB |
721 | } |
722 | ||
49f0a076 | 723 | SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode) |
5590ff0d | 724 | { |
09da082b | 725 | return do_fchmodat(AT_FDCWD, filename, mode, 0); |
5590ff0d UD |
726 | } |
727 | ||
55650b2f | 728 | /* |
b27c82e1 CB |
729 | * Check whether @kuid is valid and if so generate and set vfsuid_t in |
730 | * ia_vfsuid. | |
731 | * | |
732 | * Return: true if @kuid is valid, false if not. | |
733 | */ | |
734 | static inline bool setattr_vfsuid(struct iattr *attr, kuid_t kuid) | |
735 | { | |
736 | if (!uid_valid(kuid)) | |
737 | return false; | |
738 | attr->ia_valid |= ATTR_UID; | |
739 | attr->ia_vfsuid = VFSUIDT_INIT(kuid); | |
740 | return true; | |
741 | } | |
742 | ||
55650b2f | 743 | /* |
b27c82e1 CB |
744 | * Check whether @kgid is valid and if so generate and set vfsgid_t in |
745 | * ia_vfsgid. | |
746 | * | |
747 | * Return: true if @kgid is valid, false if not. | |
748 | */ | |
749 | static inline bool setattr_vfsgid(struct iattr *attr, kgid_t kgid) | |
750 | { | |
751 | if (!gid_valid(kgid)) | |
752 | return false; | |
753 | attr->ia_valid |= ATTR_GID; | |
754 | attr->ia_vfsgid = VFSGIDT_INIT(kgid); | |
755 | return true; | |
756 | } | |
757 | ||
b873498f | 758 | int chown_common(const struct path *path, uid_t user, gid_t group) |
1da177e4 | 759 | { |
abf08576 | 760 | struct mnt_idmap *idmap; |
4d7ca409 | 761 | struct user_namespace *fs_userns; |
fe542cf5 | 762 | struct inode *inode = path->dentry->d_inode; |
27ac0ffe | 763 | struct inode *delegated_inode = NULL; |
1da177e4 LT |
764 | int error; |
765 | struct iattr newattrs; | |
52137abe EB |
766 | kuid_t uid; |
767 | kgid_t gid; | |
768 | ||
769 | uid = make_kuid(current_user_ns(), user); | |
770 | gid = make_kgid(current_user_ns(), group); | |
1da177e4 | 771 | |
abf08576 | 772 | idmap = mnt_idmap(path->mnt); |
bd303368 | 773 | fs_userns = i_user_ns(inode); |
b8b546a0 | 774 | |
c1b8940b | 775 | retry_deleg: |
f52d74b1 TH |
776 | newattrs.ia_vfsuid = INVALID_VFSUID; |
777 | newattrs.ia_vfsgid = INVALID_VFSGID; | |
1da177e4 | 778 | newattrs.ia_valid = ATTR_CTIME; |
b27c82e1 CB |
779 | if ((user != (uid_t)-1) && !setattr_vfsuid(&newattrs, uid)) |
780 | return -EINVAL; | |
781 | if ((group != (gid_t)-1) && !setattr_vfsgid(&newattrs, gid)) | |
782 | return -EINVAL; | |
5955102c | 783 | inode_lock(inode); |
ed5a7047 CB |
784 | if (!S_ISDIR(inode->i_mode)) |
785 | newattrs.ia_valid |= ATTR_KILL_SUID | ATTR_KILL_PRIV | | |
9452e93e | 786 | setattr_should_drop_sgid(idmap, inode); |
b27c82e1 CB |
787 | /* Continue to send actual fs values, not the mount values. */ |
788 | error = security_path_chown( | |
789 | path, | |
4d7ca409 CB |
790 | from_vfsuid(idmap, fs_userns, newattrs.ia_vfsuid), |
791 | from_vfsgid(idmap, fs_userns, newattrs.ia_vfsgid)); | |
fe542cf5 | 792 | if (!error) |
abf08576 | 793 | error = notify_change(idmap, path->dentry, &newattrs, |
2f221d6f | 794 | &delegated_inode); |
5955102c | 795 | inode_unlock(inode); |
27ac0ffe BF |
796 | if (delegated_inode) { |
797 | error = break_deleg_wait(&delegated_inode); | |
798 | if (!error) | |
799 | goto retry_deleg; | |
800 | } | |
1da177e4 LT |
801 | return error; |
802 | } | |
803 | ||
55731b3c DB |
804 | int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group, |
805 | int flag) | |
5590ff0d | 806 | { |
2d8f3038 | 807 | struct path path; |
5590ff0d | 808 | int error = -EINVAL; |
65cfc672 | 809 | int lookup_flags; |
5590ff0d | 810 | |
65cfc672 | 811 | if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
5590ff0d UD |
812 | goto out; |
813 | ||
65cfc672 AV |
814 | lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; |
815 | if (flag & AT_EMPTY_PATH) | |
816 | lookup_flags |= LOOKUP_EMPTY; | |
99a5df37 | 817 | retry: |
65cfc672 | 818 | error = user_path_at(dfd, filename, lookup_flags, &path); |
6902d925 DH |
819 | if (error) |
820 | goto out; | |
2d8f3038 | 821 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
822 | if (error) |
823 | goto out_release; | |
fe542cf5 | 824 | error = chown_common(&path, user, group); |
2d8f3038 | 825 | mnt_drop_write(path.mnt); |
2af482a7 | 826 | out_release: |
2d8f3038 | 827 | path_put(&path); |
99a5df37 JL |
828 | if (retry_estale(error, lookup_flags)) { |
829 | lookup_flags |= LOOKUP_REVAL; | |
830 | goto retry; | |
831 | } | |
5590ff0d UD |
832 | out: |
833 | return error; | |
834 | } | |
835 | ||
55731b3c DB |
836 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
837 | gid_t, group, int, flag) | |
838 | { | |
839 | return do_fchownat(dfd, filename, user, group, flag); | |
840 | } | |
841 | ||
55e4def0 | 842 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 843 | { |
55731b3c | 844 | return do_fchownat(AT_FDCWD, filename, user, group, 0); |
55e4def0 | 845 | } |
1da177e4 | 846 | |
55e4def0 DH |
847 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
848 | { | |
55731b3c DB |
849 | return do_fchownat(AT_FDCWD, filename, user, group, |
850 | AT_SYMLINK_NOFOLLOW); | |
1da177e4 LT |
851 | } |
852 | ||
c04011fe CH |
853 | int vfs_fchown(struct file *file, uid_t user, gid_t group) |
854 | { | |
855 | int error; | |
856 | ||
857 | error = mnt_want_write_file(file); | |
858 | if (error) | |
859 | return error; | |
860 | audit_file(file); | |
861 | error = chown_common(&file->f_path, user, group); | |
862 | mnt_drop_write_file(file); | |
863 | return error; | |
864 | } | |
865 | ||
55731b3c | 866 | int ksys_fchown(unsigned int fd, uid_t user, gid_t group) |
1da177e4 | 867 | { |
2903ff01 | 868 | struct fd f = fdget(fd); |
1da177e4 LT |
869 | int error = -EBADF; |
870 | ||
c04011fe CH |
871 | if (f.file) { |
872 | error = vfs_fchown(f.file, user, group); | |
873 | fdput(f); | |
874 | } | |
1da177e4 LT |
875 | return error; |
876 | } | |
877 | ||
55731b3c DB |
878 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
879 | { | |
880 | return ksys_fchown(fd, user, group); | |
881 | } | |
882 | ||
83bc1d29 AG |
883 | static inline int file_get_write_access(struct file *f) |
884 | { | |
885 | int error; | |
886 | ||
887 | error = get_write_access(f->f_inode); | |
888 | if (unlikely(error)) | |
889 | return error; | |
890 | error = mnt_get_write_access(f->f_path.mnt); | |
891 | if (unlikely(error)) | |
892 | goto cleanup_inode; | |
893 | if (unlikely(f->f_mode & FMODE_BACKING)) { | |
def3ae83 | 894 | error = mnt_get_write_access(backing_file_user_path(f)->mnt); |
83bc1d29 AG |
895 | if (unlikely(error)) |
896 | goto cleanup_mnt; | |
897 | } | |
898 | return 0; | |
899 | ||
900 | cleanup_mnt: | |
901 | mnt_put_write_access(f->f_path.mnt); | |
902 | cleanup_inode: | |
903 | put_write_access(f->f_inode); | |
904 | return error; | |
905 | } | |
906 | ||
02e5180d | 907 | static int do_dentry_open(struct file *f, |
ae2bb293 | 908 | int (*open)(struct inode *, struct file *)) |
1da177e4 | 909 | { |
1abf0c71 | 910 | static const struct file_operations empty_fops = {}; |
0f4a2ceb | 911 | struct inode *inode = f->f_path.dentry->d_inode; |
1da177e4 LT |
912 | int error; |
913 | ||
b5bcdda3 | 914 | path_get(&f->f_path); |
4bacc9c9 | 915 | f->f_inode = inode; |
1da177e4 | 916 | f->f_mapping = inode->i_mapping; |
5660e13d | 917 | f->f_wb_err = filemap_sample_wb_err(f->f_mapping); |
735e4ae5 | 918 | f->f_sb_err = file_sample_sb_err(f); |
5660e13d | 919 | |
3f4d5a00 | 920 | if (unlikely(f->f_flags & O_PATH)) { |
f5d11409 | 921 | f->f_mode = FMODE_PATH | FMODE_OPENED; |
1abf0c71 | 922 | f->f_op = &empty_fops; |
af04fadc | 923 | return 0; |
1abf0c71 AV |
924 | } |
925 | ||
d6da19c9 AG |
926 | if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ) { |
927 | i_readcount_inc(inode); | |
928 | } else if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) { | |
83bc1d29 | 929 | error = file_get_write_access(f); |
3f4d5a00 | 930 | if (unlikely(error)) |
1da177e4 | 931 | goto cleanup_file; |
83f936c7 | 932 | f->f_mode |= FMODE_WRITER; |
1da177e4 LT |
933 | } |
934 | ||
2be7d348 LT |
935 | /* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */ |
936 | if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)) | |
937 | f->f_mode |= FMODE_ATOMIC_POS; | |
938 | ||
1abf0c71 | 939 | f->f_op = fops_get(inode->i_fop); |
7159d544 | 940 | if (WARN_ON(!f->f_op)) { |
72c2d531 AV |
941 | error = -ENODEV; |
942 | goto cleanup_all; | |
943 | } | |
1abf0c71 | 944 | |
e3f20ae2 | 945 | error = security_file_open(f); |
788e7dd4 YN |
946 | if (error) |
947 | goto cleanup_all; | |
948 | ||
c65454a9 | 949 | error = break_lease(file_inode(f), f->f_flags); |
f3c7691e BF |
950 | if (error) |
951 | goto cleanup_all; | |
952 | ||
ea73ea72 AV |
953 | /* normally all 3 are set; ->open() can clear them if needed */ |
954 | f->f_mode |= FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE; | |
72c2d531 | 955 | if (!open) |
834f2a4a TM |
956 | open = f->f_op->open; |
957 | if (open) { | |
958 | error = open(inode, f); | |
1da177e4 LT |
959 | if (error) |
960 | goto cleanup_all; | |
961 | } | |
f5d11409 | 962 | f->f_mode |= FMODE_OPENED; |
293bc982 | 963 | if ((f->f_mode & FMODE_READ) && |
84363182 | 964 | likely(f->f_op->read || f->f_op->read_iter)) |
7f7f25e8 | 965 | f->f_mode |= FMODE_CAN_READ; |
293bc982 | 966 | if ((f->f_mode & FMODE_WRITE) && |
84363182 | 967 | likely(f->f_op->write || f->f_op->write_iter)) |
7f7f25e8 | 968 | f->f_mode |= FMODE_CAN_WRITE; |
e7478158 JD |
969 | if ((f->f_mode & FMODE_LSEEK) && !f->f_op->llseek) |
970 | f->f_mode &= ~FMODE_LSEEK; | |
a2ad63da N |
971 | if (f->f_mapping->a_ops && f->f_mapping->a_ops->direct_IO) |
972 | f->f_mode |= FMODE_CAN_ODIRECT; | |
834f2a4a | 973 | |
1da177e4 | 974 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
164f4064 | 975 | f->f_iocb_flags = iocb_flags(f); |
1da177e4 LT |
976 | |
977 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
af04fadc | 978 | |
a2ad63da N |
979 | if ((f->f_flags & O_DIRECT) && !(f->f_mode & FMODE_CAN_ODIRECT)) |
980 | return -EINVAL; | |
09d91cda SL |
981 | |
982 | /* | |
983 | * XXX: Huge page cache doesn't support writing yet. Drop all page | |
984 | * cache for this file before processing writes. | |
985 | */ | |
eb6ecbed CF |
986 | if (f->f_mode & FMODE_WRITE) { |
987 | /* | |
8e344782 MG |
988 | * Depends on full fence from get_write_access() to synchronize |
989 | * against collapse_file() regarding i_writecount and nr_thps | |
990 | * updates. Ensures subsequent insertion of THPs into the page | |
991 | * cache will fail. | |
eb6ecbed | 992 | */ |
55fc0d91 | 993 | if (filemap_nr_thps(inode->i_mapping)) { |
8468e937 RW |
994 | struct address_space *mapping = inode->i_mapping; |
995 | ||
55fc0d91 | 996 | filemap_invalidate_lock(inode->i_mapping); |
8468e937 RW |
997 | /* |
998 | * unmap_mapping_range just need to be called once | |
999 | * here, because the private pages is not need to be | |
1000 | * unmapped mapping (e.g. data segment of dynamic | |
1001 | * shared libraries here). | |
1002 | */ | |
1003 | unmap_mapping_range(mapping, 0, 0, 0); | |
1004 | truncate_inode_pages(mapping, 0); | |
55fc0d91 RW |
1005 | filemap_invalidate_unlock(inode->i_mapping); |
1006 | } | |
eb6ecbed | 1007 | } |
09d91cda | 1008 | |
96b7e579 | 1009 | return 0; |
1da177e4 LT |
1010 | |
1011 | cleanup_all: | |
6b4e8085 AV |
1012 | if (WARN_ON_ONCE(error > 0)) |
1013 | error = -EINVAL; | |
1da177e4 | 1014 | fops_put(f->f_op); |
d6da19c9 | 1015 | put_file_access(f); |
1da177e4 | 1016 | cleanup_file: |
02e5180d AV |
1017 | path_put(&f->f_path); |
1018 | f->f_path.mnt = NULL; | |
1019 | f->f_path.dentry = NULL; | |
dd37978c | 1020 | f->f_inode = NULL; |
96b7e579 | 1021 | return error; |
1da177e4 LT |
1022 | } |
1023 | ||
d18e9008 MS |
1024 | /** |
1025 | * finish_open - finish opening a file | |
0854d450 | 1026 | * @file: file pointer |
d18e9008 MS |
1027 | * @dentry: pointer to dentry |
1028 | * @open: open callback | |
1029 | * | |
1030 | * This can be used to finish opening a file passed to i_op->atomic_open(). | |
1031 | * | |
1032 | * If the open callback is set to NULL, then the standard f_op->open() | |
1033 | * filesystem callback is substituted. | |
0854d450 MS |
1034 | * |
1035 | * NB: the dentry reference is _not_ consumed. If, for example, the dentry is | |
1036 | * the return value of d_splice_alias(), then the caller needs to perform dput() | |
1037 | * on it after finish_open(). | |
1038 | * | |
0854d450 | 1039 | * Returns zero on success or -errno if the open failed. |
d18e9008 | 1040 | */ |
30d90494 | 1041 | int finish_open(struct file *file, struct dentry *dentry, |
be12af3e | 1042 | int (*open)(struct inode *, struct file *)) |
d18e9008 | 1043 | { |
aad888f8 | 1044 | BUG_ON(file->f_mode & FMODE_OPENED); /* once it's opened, it's opened */ |
d18e9008 | 1045 | |
b5bcdda3 | 1046 | file->f_path.dentry = dentry; |
0f4a2ceb | 1047 | return do_dentry_open(file, open); |
d18e9008 MS |
1048 | } |
1049 | EXPORT_SYMBOL(finish_open); | |
1050 | ||
1051 | /** | |
1052 | * finish_no_open - finish ->atomic_open() without opening the file | |
1053 | * | |
0854d450 | 1054 | * @file: file pointer |
d18e9008 MS |
1055 | * @dentry: dentry or NULL (as returned from ->lookup()) |
1056 | * | |
1057 | * This can be used to set the result of a successful lookup in ->atomic_open(). | |
0854d450 MS |
1058 | * |
1059 | * NB: unlike finish_open() this function does consume the dentry reference and | |
1060 | * the caller need not dput() it. | |
1061 | * | |
64e1ac4d | 1062 | * Returns "0" which must be the return value of ->atomic_open() after having |
0854d450 | 1063 | * called this function. |
d18e9008 | 1064 | */ |
e45198a6 | 1065 | int finish_no_open(struct file *file, struct dentry *dentry) |
d18e9008 | 1066 | { |
30d90494 | 1067 | file->f_path.dentry = dentry; |
64e1ac4d | 1068 | return 0; |
d18e9008 MS |
1069 | } |
1070 | EXPORT_SYMBOL(finish_no_open); | |
1071 | ||
9bf39ab2 MS |
1072 | char *file_path(struct file *filp, char *buf, int buflen) |
1073 | { | |
1074 | return d_path(&filp->f_path, buf, buflen); | |
1075 | } | |
1076 | EXPORT_SYMBOL(file_path); | |
1077 | ||
4bacc9c9 DH |
1078 | /** |
1079 | * vfs_open - open the file at the given path | |
1080 | * @path: path to open | |
1081 | * @file: newly allocated file with f_flag initialized | |
4bacc9c9 | 1082 | */ |
ae2bb293 | 1083 | int vfs_open(const struct path *path, struct file *file) |
4bacc9c9 | 1084 | { |
7d1cf5e6 N |
1085 | int ret; |
1086 | ||
54d5ca87 | 1087 | file->f_path = *path; |
7d1cf5e6 N |
1088 | ret = do_dentry_open(file, NULL); |
1089 | if (!ret) { | |
1090 | /* | |
1091 | * Once we return a file with FMODE_OPENED, __fput() will call | |
1092 | * fsnotify_close(), so we need fsnotify_open() here for | |
1093 | * symmetry. | |
1094 | */ | |
1095 | fsnotify_open(file); | |
1096 | } | |
1097 | return ret; | |
4bacc9c9 DH |
1098 | } |
1099 | ||
765927b2 | 1100 | struct file *dentry_open(const struct path *path, int flags, |
745ca247 | 1101 | const struct cred *cred) |
a1a5b3d9 PS |
1102 | { |
1103 | int error; | |
1104 | struct file *f; | |
1105 | ||
c212f9aa | 1106 | /* We must always pass in a valid mount pointer. */ |
765927b2 | 1107 | BUG_ON(!path->mnt); |
322ee5b3 | 1108 | |
ea73ea72 | 1109 | f = alloc_empty_file(flags, cred); |
af04fadc | 1110 | if (!IS_ERR(f)) { |
ae2bb293 | 1111 | error = vfs_open(path, f); |
4d27f326 AV |
1112 | if (error) { |
1113 | fput(f); | |
af04fadc AV |
1114 | f = ERR_PTR(error); |
1115 | } | |
2a027e7a AV |
1116 | } |
1117 | return f; | |
a1a5b3d9 | 1118 | } |
1da177e4 LT |
1119 | EXPORT_SYMBOL(dentry_open); |
1120 | ||
fb70bf12 CL |
1121 | /** |
1122 | * dentry_create - Create and open a file | |
1123 | * @path: path to create | |
1124 | * @flags: O_ flags | |
1125 | * @mode: mode bits for new file | |
1126 | * @cred: credentials to use | |
1127 | * | |
1128 | * Caller must hold the parent directory's lock, and have prepared | |
1129 | * a negative dentry, placed in @path->dentry, for the new file. | |
1130 | * | |
1131 | * Caller sets @path->mnt to the vfsmount of the filesystem where | |
1132 | * the new file is to be created. The parent directory and the | |
1133 | * negative dentry must reside on the same filesystem instance. | |
1134 | * | |
1135 | * On success, returns a "struct file *". Otherwise a ERR_PTR | |
1136 | * is returned. | |
1137 | */ | |
1138 | struct file *dentry_create(const struct path *path, int flags, umode_t mode, | |
1139 | const struct cred *cred) | |
1140 | { | |
1141 | struct file *f; | |
1142 | int error; | |
1143 | ||
fb70bf12 CL |
1144 | f = alloc_empty_file(flags, cred); |
1145 | if (IS_ERR(f)) | |
1146 | return f; | |
1147 | ||
abf08576 | 1148 | error = vfs_create(mnt_idmap(path->mnt), |
fb70bf12 CL |
1149 | d_inode(path->dentry->d_parent), |
1150 | path->dentry, mode, true); | |
1151 | if (!error) | |
1152 | error = vfs_open(path, f); | |
1153 | ||
1154 | if (unlikely(error)) { | |
1155 | fput(f); | |
1156 | return ERR_PTR(error); | |
1157 | } | |
1158 | return f; | |
1159 | } | |
1160 | EXPORT_SYMBOL(dentry_create); | |
1161 | ||
cbb0b9d4 AG |
1162 | /** |
1163 | * kernel_file_open - open a file for kernel internal use | |
1164 | * @path: path of the file to open | |
1165 | * @flags: open flags | |
cbb0b9d4 AG |
1166 | * @cred: credentials for open |
1167 | * | |
1168 | * Open a file for use by in-kernel consumers. The file is not accounted | |
1169 | * against nr_files and must not be installed into the file descriptor | |
1170 | * table. | |
1171 | * | |
1172 | * Return: Opened file on success, an error pointer on failure. | |
1173 | */ | |
1174 | struct file *kernel_file_open(const struct path *path, int flags, | |
af58dc1f | 1175 | const struct cred *cred) |
2abc77af | 1176 | { |
cbb0b9d4 AG |
1177 | struct file *f; |
1178 | int error; | |
2abc77af | 1179 | |
cbb0b9d4 AG |
1180 | f = alloc_empty_file_noaccount(flags, cred); |
1181 | if (IS_ERR(f)) | |
1182 | return f; | |
1183 | ||
1184 | f->f_path = *path; | |
0f4a2ceb | 1185 | error = do_dentry_open(f, NULL); |
cbb0b9d4 AG |
1186 | if (error) { |
1187 | fput(f); | |
7d1cf5e6 | 1188 | return ERR_PTR(error); |
2abc77af | 1189 | } |
7d1cf5e6 N |
1190 | |
1191 | fsnotify_open(f); | |
2abc77af AV |
1192 | return f; |
1193 | } | |
cbb0b9d4 AG |
1194 | EXPORT_SYMBOL_GPL(kernel_file_open); |
1195 | ||
fddb5d43 AS |
1196 | #define WILL_CREATE(flags) (flags & (O_CREAT | __O_TMPFILE)) |
1197 | #define O_PATH_FLAGS (O_DIRECTORY | O_NOFOLLOW | O_PATH | O_CLOEXEC) | |
1198 | ||
35cb6d54 | 1199 | inline struct open_how build_open_how(int flags, umode_t mode) |
fddb5d43 AS |
1200 | { |
1201 | struct open_how how = { | |
1202 | .flags = flags & VALID_OPEN_FLAGS, | |
1203 | .mode = mode & S_IALLUGO, | |
1204 | }; | |
1205 | ||
1206 | /* O_PATH beats everything else. */ | |
1207 | if (how.flags & O_PATH) | |
1208 | how.flags &= O_PATH_FLAGS; | |
1209 | /* Modes should only be set for create-like flags. */ | |
1210 | if (!WILL_CREATE(how.flags)) | |
1211 | how.mode = 0; | |
1212 | return how; | |
1213 | } | |
1214 | ||
35cb6d54 | 1215 | inline int build_open_flags(const struct open_how *how, struct open_flags *op) |
47c805dc | 1216 | { |
cfe80306 | 1217 | u64 flags = how->flags; |
cedd0bdc | 1218 | u64 strip = __FMODE_NONOTIFY | O_CLOEXEC; |
47c805dc | 1219 | int lookup_flags = 0; |
62fb4a15 | 1220 | int acc_mode = ACC_MODE(flags); |
47c805dc | 1221 | |
cfe80306 CB |
1222 | BUILD_BUG_ON_MSG(upper_32_bits(VALID_OPEN_FLAGS), |
1223 | "struct open_flags doesn't yet handle flags > 32 bits"); | |
1224 | ||
1225 | /* | |
1226 | * Strip flags that either shouldn't be set by userspace like | |
1227 | * FMODE_NONOTIFY or that aren't relevant in determining struct | |
1228 | * open_flags like O_CLOEXEC. | |
1229 | */ | |
1230 | flags &= ~strip; | |
fddb5d43 | 1231 | |
629e014b | 1232 | /* |
fddb5d43 AS |
1233 | * Older syscalls implicitly clear all of the invalid flags or argument |
1234 | * values before calling build_open_flags(), but openat2(2) checks all | |
1235 | * of its arguments. | |
629e014b | 1236 | */ |
fddb5d43 AS |
1237 | if (flags & ~VALID_OPEN_FLAGS) |
1238 | return -EINVAL; | |
1239 | if (how->resolve & ~VALID_RESOLVE_FLAGS) | |
1240 | return -EINVAL; | |
629e014b | 1241 | |
398840f8 AS |
1242 | /* Scoping flags are mutually exclusive. */ |
1243 | if ((how->resolve & RESOLVE_BENEATH) && (how->resolve & RESOLVE_IN_ROOT)) | |
1244 | return -EINVAL; | |
1245 | ||
fddb5d43 AS |
1246 | /* Deal with the mode. */ |
1247 | if (WILL_CREATE(flags)) { | |
1248 | if (how->mode & ~S_IALLUGO) | |
1249 | return -EINVAL; | |
1250 | op->mode = how->mode | S_IFREG; | |
1251 | } else { | |
1252 | if (how->mode != 0) | |
1253 | return -EINVAL; | |
e68726ff | 1254 | op->mode = 0; |
fddb5d43 | 1255 | } |
47c805dc AV |
1256 | |
1257 | /* | |
43b45063 CB |
1258 | * Block bugs where O_DIRECTORY | O_CREAT created regular files. |
1259 | * Note, that blocking O_DIRECTORY | O_CREAT here also protects | |
1260 | * O_TMPFILE below which requires O_DIRECTORY being raised. | |
47c805dc | 1261 | */ |
43b45063 CB |
1262 | if ((flags & (O_DIRECTORY | O_CREAT)) == (O_DIRECTORY | O_CREAT)) |
1263 | return -EINVAL; | |
1264 | ||
1265 | /* Now handle the creative implementation of O_TMPFILE. */ | |
bb458c64 | 1266 | if (flags & __O_TMPFILE) { |
43b45063 CB |
1267 | /* |
1268 | * In order to ensure programs get explicit errors when trying | |
1269 | * to use O_TMPFILE on old kernels we enforce that O_DIRECTORY | |
1270 | * is raised alongside __O_TMPFILE. | |
1271 | */ | |
1272 | if (!(flags & O_DIRECTORY)) | |
60545d0d | 1273 | return -EINVAL; |
ba57ea64 AV |
1274 | if (!(acc_mode & MAY_WRITE)) |
1275 | return -EINVAL; | |
fddb5d43 AS |
1276 | } |
1277 | if (flags & O_PATH) { | |
1278 | /* O_PATH only permits certain other flags to be set. */ | |
1279 | if (flags & ~O_PATH_FLAGS) | |
1280 | return -EINVAL; | |
1abf0c71 | 1281 | acc_mode = 0; |
1abf0c71 | 1282 | } |
47c805dc | 1283 | |
fddb5d43 AS |
1284 | /* |
1285 | * O_SYNC is implemented as __O_SYNC|O_DSYNC. As many places only | |
1286 | * check for O_DSYNC if the need any syncing at all we enforce it's | |
1287 | * always set instead of having to deal with possibly weird behaviour | |
1288 | * for malicious applications setting only __O_SYNC. | |
1289 | */ | |
1290 | if (flags & __O_SYNC) | |
1291 | flags |= O_DSYNC; | |
1292 | ||
1abf0c71 | 1293 | op->open_flag = flags; |
47c805dc AV |
1294 | |
1295 | /* O_TRUNC implies we need access checks for write permissions */ | |
1296 | if (flags & O_TRUNC) | |
1297 | acc_mode |= MAY_WRITE; | |
1298 | ||
1299 | /* Allow the LSM permission hook to distinguish append | |
1300 | access from general write access. */ | |
1301 | if (flags & O_APPEND) | |
1302 | acc_mode |= MAY_APPEND; | |
1303 | ||
1304 | op->acc_mode = acc_mode; | |
1305 | ||
1abf0c71 AV |
1306 | op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN; |
1307 | ||
47c805dc AV |
1308 | if (flags & O_CREAT) { |
1309 | op->intent |= LOOKUP_CREATE; | |
31d1726d | 1310 | if (flags & O_EXCL) { |
47c805dc | 1311 | op->intent |= LOOKUP_EXCL; |
31d1726d AV |
1312 | flags |= O_NOFOLLOW; |
1313 | } | |
47c805dc AV |
1314 | } |
1315 | ||
1316 | if (flags & O_DIRECTORY) | |
1317 | lookup_flags |= LOOKUP_DIRECTORY; | |
1318 | if (!(flags & O_NOFOLLOW)) | |
1319 | lookup_flags |= LOOKUP_FOLLOW; | |
fddb5d43 AS |
1320 | |
1321 | if (how->resolve & RESOLVE_NO_XDEV) | |
1322 | lookup_flags |= LOOKUP_NO_XDEV; | |
1323 | if (how->resolve & RESOLVE_NO_MAGICLINKS) | |
1324 | lookup_flags |= LOOKUP_NO_MAGICLINKS; | |
1325 | if (how->resolve & RESOLVE_NO_SYMLINKS) | |
1326 | lookup_flags |= LOOKUP_NO_SYMLINKS; | |
1327 | if (how->resolve & RESOLVE_BENEATH) | |
1328 | lookup_flags |= LOOKUP_BENEATH; | |
1329 | if (how->resolve & RESOLVE_IN_ROOT) | |
1330 | lookup_flags |= LOOKUP_IN_ROOT; | |
99668f61 JA |
1331 | if (how->resolve & RESOLVE_CACHED) { |
1332 | /* Don't bother even trying for create/truncate/tmpfile open */ | |
a0fc452a | 1333 | if (flags & (O_TRUNC | O_CREAT | __O_TMPFILE)) |
99668f61 JA |
1334 | return -EAGAIN; |
1335 | lookup_flags |= LOOKUP_CACHED; | |
1336 | } | |
fddb5d43 | 1337 | |
f9652e10 AV |
1338 | op->lookup_flags = lookup_flags; |
1339 | return 0; | |
47c805dc AV |
1340 | } |
1341 | ||
669abf4e JL |
1342 | /** |
1343 | * file_open_name - open file and return file pointer | |
1344 | * | |
1345 | * @name: struct filename containing path to open | |
1346 | * @flags: open flags as per the open(2) second argument | |
1347 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1348 | * | |
1349 | * This is the helper to open a file from kernelspace if you really | |
1350 | * have to. But in generally you should not do this, so please move | |
1351 | * along, nothing to see here.. | |
1352 | */ | |
1353 | struct file *file_open_name(struct filename *name, int flags, umode_t mode) | |
1354 | { | |
1355 | struct open_flags op; | |
fddb5d43 AS |
1356 | struct open_how how = build_open_how(flags, mode); |
1357 | int err = build_open_flags(&how, &op); | |
1358 | if (err) | |
1359 | return ERR_PTR(err); | |
1360 | return do_filp_open(AT_FDCWD, name, &op); | |
669abf4e JL |
1361 | } |
1362 | ||
47c805dc AV |
1363 | /** |
1364 | * filp_open - open file and return file pointer | |
1365 | * | |
1366 | * @filename: path to open | |
1367 | * @flags: open flags as per the open(2) second argument | |
1368 | * @mode: mode for the new file if O_CREAT is set, else ignored | |
1369 | * | |
1370 | * This is the helper to open a file from kernelspace if you really | |
1371 | * have to. But in generally you should not do this, so please move | |
1372 | * along, nothing to see here.. | |
1373 | */ | |
a218d0fd | 1374 | struct file *filp_open(const char *filename, int flags, umode_t mode) |
47c805dc | 1375 | { |
51689104 PM |
1376 | struct filename *name = getname_kernel(filename); |
1377 | struct file *file = ERR_CAST(name); | |
a69ce85e | 1378 | |
51689104 PM |
1379 | if (!IS_ERR(name)) { |
1380 | file = file_open_name(name, flags, mode); | |
1381 | putname(name); | |
1382 | } | |
1383 | return file; | |
47c805dc AV |
1384 | } |
1385 | EXPORT_SYMBOL(filp_open); | |
1386 | ||
ffb37ca3 | 1387 | struct file *file_open_root(const struct path *root, |
378c6520 | 1388 | const char *filename, int flags, umode_t mode) |
73d049a4 AV |
1389 | { |
1390 | struct open_flags op; | |
fddb5d43 AS |
1391 | struct open_how how = build_open_how(flags, mode); |
1392 | int err = build_open_flags(&how, &op); | |
f9652e10 AV |
1393 | if (err) |
1394 | return ERR_PTR(err); | |
ffb37ca3 | 1395 | return do_file_open_root(root, filename, &op); |
73d049a4 AV |
1396 | } |
1397 | EXPORT_SYMBOL(file_open_root); | |
1398 | ||
fddb5d43 AS |
1399 | static long do_sys_openat2(int dfd, const char __user *filename, |
1400 | struct open_how *how) | |
1da177e4 | 1401 | { |
47c805dc | 1402 | struct open_flags op; |
fddb5d43 | 1403 | int fd = build_open_flags(how, &op); |
f9652e10 AV |
1404 | struct filename *tmp; |
1405 | ||
1406 | if (fd) | |
1407 | return fd; | |
1408 | ||
1409 | tmp = getname(filename); | |
1410 | if (IS_ERR(tmp)) | |
1411 | return PTR_ERR(tmp); | |
1412 | ||
fddb5d43 | 1413 | fd = get_unused_fd_flags(how->flags); |
f9652e10 AV |
1414 | if (fd >= 0) { |
1415 | struct file *f = do_filp_open(dfd, tmp, &op); | |
1416 | if (IS_ERR(f)) { | |
1417 | put_unused_fd(fd); | |
1418 | fd = PTR_ERR(f); | |
1419 | } else { | |
f9652e10 | 1420 | fd_install(fd, f); |
1da177e4 | 1421 | } |
1da177e4 | 1422 | } |
f9652e10 | 1423 | putname(tmp); |
1da177e4 | 1424 | return fd; |
1da177e4 | 1425 | } |
e922efc3 | 1426 | |
fddb5d43 | 1427 | long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode) |
e922efc3 | 1428 | { |
fddb5d43 AS |
1429 | struct open_how how = build_open_how(flags, mode); |
1430 | return do_sys_openat2(dfd, filename, &how); | |
1431 | } | |
e922efc3 | 1432 | |
fddb5d43 AS |
1433 | |
1434 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1435 | { | |
166e07c3 CH |
1436 | if (force_o_largefile()) |
1437 | flags |= O_LARGEFILE; | |
1438 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
e922efc3 | 1439 | } |
1da177e4 | 1440 | |
6559eed8 | 1441 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
a218d0fd | 1442 | umode_t, mode) |
5590ff0d UD |
1443 | { |
1444 | if (force_o_largefile()) | |
1445 | flags |= O_LARGEFILE; | |
2cf09666 | 1446 | return do_sys_open(dfd, filename, flags, mode); |
5590ff0d | 1447 | } |
5590ff0d | 1448 | |
fddb5d43 AS |
1449 | SYSCALL_DEFINE4(openat2, int, dfd, const char __user *, filename, |
1450 | struct open_how __user *, how, size_t, usize) | |
1451 | { | |
1452 | int err; | |
1453 | struct open_how tmp; | |
1454 | ||
1455 | BUILD_BUG_ON(sizeof(struct open_how) < OPEN_HOW_SIZE_VER0); | |
1456 | BUILD_BUG_ON(sizeof(struct open_how) != OPEN_HOW_SIZE_LATEST); | |
1457 | ||
1458 | if (unlikely(usize < OPEN_HOW_SIZE_VER0)) | |
1459 | return -EINVAL; | |
1460 | ||
1461 | err = copy_struct_from_user(&tmp, sizeof(tmp), how, usize); | |
1462 | if (err) | |
1463 | return err; | |
1464 | ||
571e5c0e RGB |
1465 | audit_openat2_how(&tmp); |
1466 | ||
fddb5d43 AS |
1467 | /* O_LARGEFILE is only allowed for non-O_PATH. */ |
1468 | if (!(tmp.flags & O_PATH) && force_o_largefile()) | |
1469 | tmp.flags |= O_LARGEFILE; | |
1470 | ||
1471 | return do_sys_openat2(dfd, filename, &tmp); | |
1472 | } | |
1473 | ||
e35d49f6 AV |
1474 | #ifdef CONFIG_COMPAT |
1475 | /* | |
1476 | * Exactly like sys_open(), except that it doesn't set the | |
1477 | * O_LARGEFILE flag. | |
1478 | */ | |
1479 | COMPAT_SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode) | |
1480 | { | |
1481 | return do_sys_open(AT_FDCWD, filename, flags, mode); | |
1482 | } | |
1483 | ||
1484 | /* | |
1485 | * Exactly like sys_openat(), except that it doesn't set the | |
1486 | * O_LARGEFILE flag. | |
1487 | */ | |
1488 | COMPAT_SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, umode_t, mode) | |
1489 | { | |
1490 | return do_sys_open(dfd, filename, flags, mode); | |
1491 | } | |
1492 | #endif | |
1493 | ||
1da177e4 LT |
1494 | #ifndef __alpha__ |
1495 | ||
1496 | /* | |
1497 | * For backward compatibility? Maybe this should be moved | |
1498 | * into arch/i386 instead? | |
1499 | */ | |
a218d0fd | 1500 | SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode) |
1da177e4 | 1501 | { |
166e07c3 | 1502 | int flags = O_CREAT | O_WRONLY | O_TRUNC; |
1da177e4 | 1503 | |
166e07c3 CH |
1504 | if (force_o_largefile()) |
1505 | flags |= O_LARGEFILE; | |
1506 | return do_sys_open(AT_FDCWD, pathname, flags, mode); | |
1507 | } | |
1da177e4 LT |
1508 | #endif |
1509 | ||
1510 | /* | |
1511 | * "id" is the POSIX thread ID. We use the | |
1512 | * files pointer for this.. | |
1513 | */ | |
021a160a | 1514 | static int filp_flush(struct file *filp, fl_owner_t id) |
1da177e4 | 1515 | { |
45778ca8 | 1516 | int retval = 0; |
1da177e4 | 1517 | |
47d58691 JH |
1518 | if (CHECK_DATA_CORRUPTION(file_count(filp) == 0, |
1519 | "VFS: Close: file count is 0 (f_op=%ps)", | |
1520 | filp->f_op)) { | |
45778ca8 | 1521 | return 0; |
1da177e4 LT |
1522 | } |
1523 | ||
72c2d531 | 1524 | if (filp->f_op->flush) |
75e1fcc0 | 1525 | retval = filp->f_op->flush(filp, id); |
1da177e4 | 1526 | |
1abf0c71 AV |
1527 | if (likely(!(filp->f_mode & FMODE_PATH))) { |
1528 | dnotify_flush(filp, id); | |
1529 | locks_remove_posix(filp, id); | |
1530 | } | |
1da177e4 LT |
1531 | return retval; |
1532 | } | |
1533 | ||
021a160a LT |
1534 | int filp_close(struct file *filp, fl_owner_t id) |
1535 | { | |
1536 | int retval; | |
1537 | ||
1538 | retval = filp_flush(filp, id); | |
1539 | fput(filp); | |
1540 | ||
1541 | return retval; | |
1542 | } | |
1da177e4 LT |
1543 | EXPORT_SYMBOL(filp_close); |
1544 | ||
1545 | /* | |
1546 | * Careful here! We test whether the file pointer is NULL before | |
1547 | * releasing the fd. This ensures that one clone task can't release | |
1548 | * an fd while another clone is opening it. | |
1549 | */ | |
ca013e94 | 1550 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 | 1551 | { |
021a160a LT |
1552 | int retval; |
1553 | struct file *file; | |
1554 | ||
a88c955f | 1555 | file = file_close_fd(fd); |
021a160a LT |
1556 | if (!file) |
1557 | return -EBADF; | |
1558 | ||
1559 | retval = filp_flush(file, current->files); | |
1560 | ||
1561 | /* | |
1562 | * We're returning to user space. Don't bother | |
1563 | * with any delayed fput() cases. | |
1564 | */ | |
1565 | __fput_sync(file); | |
ee731f4f EP |
1566 | |
1567 | /* can't restart close syscall because file table entry was cleared */ | |
1568 | if (unlikely(retval == -ERESTARTSYS || | |
1569 | retval == -ERESTARTNOINTR || | |
1570 | retval == -ERESTARTNOHAND || | |
1571 | retval == -ERESTART_RESTARTBLOCK)) | |
1572 | retval = -EINTR; | |
1573 | ||
1574 | return retval; | |
1da177e4 | 1575 | } |
1da177e4 | 1576 | |
278a5fba | 1577 | /** |
35931eb3 | 1578 | * sys_close_range() - Close all file descriptors in a given range. |
278a5fba CB |
1579 | * |
1580 | * @fd: starting file descriptor to close | |
1581 | * @max_fd: last file descriptor to close | |
1582 | * @flags: reserved for future extensions | |
1583 | * | |
1584 | * This closes a range of file descriptors. All file descriptors | |
1585 | * from @fd up to and including @max_fd are closed. | |
1586 | * Currently, errors to close a given file descriptor are ignored. | |
1587 | */ | |
1588 | SYSCALL_DEFINE3(close_range, unsigned int, fd, unsigned int, max_fd, | |
1589 | unsigned int, flags) | |
1590 | { | |
60997c3d | 1591 | return __close_range(fd, max_fd, flags); |
278a5fba CB |
1592 | } |
1593 | ||
1da177e4 LT |
1594 | /* |
1595 | * This routine simulates a hangup on the tty, to arrange that users | |
1596 | * are given clean terminals at login time. | |
1597 | */ | |
ca013e94 | 1598 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1599 | { |
1600 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1601 | tty_vhangup_self(); |
1da177e4 LT |
1602 | return 0; |
1603 | } | |
1604 | return -EPERM; | |
1605 | } | |
1606 | ||
1607 | /* | |
1608 | * Called when an inode is about to be open. | |
1609 | * We use this to disallow opening large files on 32bit systems if | |
1610 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1611 | * on this flag in sys_open. | |
1612 | */ | |
1613 | int generic_file_open(struct inode * inode, struct file * filp) | |
1614 | { | |
1615 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1616 | return -EOVERFLOW; |
1da177e4 LT |
1617 | return 0; |
1618 | } | |
1619 | ||
1620 | EXPORT_SYMBOL(generic_file_open); | |
1621 | ||
1622 | /* | |
1623 | * This is used by subsystems that don't want seekable | |
06b1e104 DT |
1624 | * file descriptors. The function is not supposed to ever fail, the only |
1625 | * reason it returns an 'int' and not 'void' is so that it can be plugged | |
1626 | * directly into file_operations structure. | |
1da177e4 LT |
1627 | */ |
1628 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1629 | { | |
1630 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1631 | return 0; | |
1632 | } | |
1633 | ||
1634 | EXPORT_SYMBOL(nonseekable_open); | |
10dce8af KS |
1635 | |
1636 | /* | |
1637 | * stream_open is used by subsystems that want stream-like file descriptors. | |
1638 | * Such file descriptors are not seekable and don't have notion of position | |
438ab720 KS |
1639 | * (file.f_pos is always 0 and ppos passed to .read()/.write() is always NULL). |
1640 | * Contrary to file descriptors of other regular files, .read() and .write() | |
1641 | * can run simultaneously. | |
10dce8af KS |
1642 | * |
1643 | * stream_open never fails and is marked to return int so that it could be | |
1644 | * directly used as file_operations.open . | |
1645 | */ | |
1646 | int stream_open(struct inode *inode, struct file *filp) | |
1647 | { | |
2be7d348 | 1648 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE | FMODE_ATOMIC_POS); |
10dce8af KS |
1649 | filp->f_mode |= FMODE_STREAM; |
1650 | return 0; | |
1651 | } | |
1652 | ||
1653 | EXPORT_SYMBOL(stream_open); |