Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/open.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | #include <linux/string.h> | |
8 | #include <linux/mm.h> | |
1da177e4 | 9 | #include <linux/file.h> |
9f3acc31 | 10 | #include <linux/fdtable.h> |
1da177e4 | 11 | #include <linux/quotaops.h> |
0eeca283 | 12 | #include <linux/fsnotify.h> |
1da177e4 LT |
13 | #include <linux/module.h> |
14 | #include <linux/slab.h> | |
15 | #include <linux/tty.h> | |
16 | #include <linux/namei.h> | |
17 | #include <linux/backing-dev.h> | |
16f7e0fe | 18 | #include <linux/capability.h> |
086f7316 | 19 | #include <linux/securebits.h> |
1da177e4 LT |
20 | #include <linux/security.h> |
21 | #include <linux/mount.h> | |
22 | #include <linux/vfs.h> | |
5590ff0d | 23 | #include <linux/fcntl.h> |
1da177e4 LT |
24 | #include <asm/uaccess.h> |
25 | #include <linux/fs.h> | |
ef3daeda | 26 | #include <linux/personality.h> |
1da177e4 LT |
27 | #include <linux/pagemap.h> |
28 | #include <linux/syscalls.h> | |
ab2af1f5 | 29 | #include <linux/rcupdate.h> |
73241ccc | 30 | #include <linux/audit.h> |
97ac7350 | 31 | #include <linux/falloc.h> |
5ad4e53b | 32 | #include <linux/fs_struct.h> |
b65a9cfc | 33 | #include <linux/ima.h> |
1da177e4 | 34 | |
e81e3f4d EP |
35 | #include "internal.h" |
36 | ||
726c3342 | 37 | int vfs_statfs(struct dentry *dentry, struct kstatfs *buf) |
1da177e4 LT |
38 | { |
39 | int retval = -ENODEV; | |
40 | ||
726c3342 | 41 | if (dentry) { |
1da177e4 | 42 | retval = -ENOSYS; |
726c3342 | 43 | if (dentry->d_sb->s_op->statfs) { |
1da177e4 | 44 | memset(buf, 0, sizeof(*buf)); |
726c3342 | 45 | retval = security_sb_statfs(dentry); |
1da177e4 LT |
46 | if (retval) |
47 | return retval; | |
726c3342 | 48 | retval = dentry->d_sb->s_op->statfs(dentry, buf); |
1da177e4 LT |
49 | if (retval == 0 && buf->f_frsize == 0) |
50 | buf->f_frsize = buf->f_bsize; | |
51 | } | |
52 | } | |
53 | return retval; | |
54 | } | |
55 | ||
56 | EXPORT_SYMBOL(vfs_statfs); | |
57 | ||
726c3342 | 58 | static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf) |
1da177e4 LT |
59 | { |
60 | struct kstatfs st; | |
61 | int retval; | |
62 | ||
726c3342 | 63 | retval = vfs_statfs(dentry, &st); |
1da177e4 LT |
64 | if (retval) |
65 | return retval; | |
66 | ||
67 | if (sizeof(*buf) == sizeof(st)) | |
68 | memcpy(buf, &st, sizeof(st)); | |
69 | else { | |
70 | if (sizeof buf->f_blocks == 4) { | |
f4a67cce JT |
71 | if ((st.f_blocks | st.f_bfree | st.f_bavail | |
72 | st.f_bsize | st.f_frsize) & | |
1da177e4 LT |
73 | 0xffffffff00000000ULL) |
74 | return -EOVERFLOW; | |
75 | /* | |
76 | * f_files and f_ffree may be -1; it's okay to stuff | |
77 | * that into 32 bits | |
78 | */ | |
79 | if (st.f_files != -1 && | |
80 | (st.f_files & 0xffffffff00000000ULL)) | |
81 | return -EOVERFLOW; | |
82 | if (st.f_ffree != -1 && | |
83 | (st.f_ffree & 0xffffffff00000000ULL)) | |
84 | return -EOVERFLOW; | |
85 | } | |
86 | ||
87 | buf->f_type = st.f_type; | |
88 | buf->f_bsize = st.f_bsize; | |
89 | buf->f_blocks = st.f_blocks; | |
90 | buf->f_bfree = st.f_bfree; | |
91 | buf->f_bavail = st.f_bavail; | |
92 | buf->f_files = st.f_files; | |
93 | buf->f_ffree = st.f_ffree; | |
94 | buf->f_fsid = st.f_fsid; | |
95 | buf->f_namelen = st.f_namelen; | |
96 | buf->f_frsize = st.f_frsize; | |
97 | memset(buf->f_spare, 0, sizeof(buf->f_spare)); | |
98 | } | |
99 | return 0; | |
100 | } | |
101 | ||
726c3342 | 102 | static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf) |
1da177e4 LT |
103 | { |
104 | struct kstatfs st; | |
105 | int retval; | |
106 | ||
726c3342 | 107 | retval = vfs_statfs(dentry, &st); |
1da177e4 LT |
108 | if (retval) |
109 | return retval; | |
110 | ||
111 | if (sizeof(*buf) == sizeof(st)) | |
112 | memcpy(buf, &st, sizeof(st)); | |
113 | else { | |
114 | buf->f_type = st.f_type; | |
115 | buf->f_bsize = st.f_bsize; | |
116 | buf->f_blocks = st.f_blocks; | |
117 | buf->f_bfree = st.f_bfree; | |
118 | buf->f_bavail = st.f_bavail; | |
119 | buf->f_files = st.f_files; | |
120 | buf->f_ffree = st.f_ffree; | |
121 | buf->f_fsid = st.f_fsid; | |
122 | buf->f_namelen = st.f_namelen; | |
123 | buf->f_frsize = st.f_frsize; | |
124 | memset(buf->f_spare, 0, sizeof(buf->f_spare)); | |
125 | } | |
126 | return 0; | |
127 | } | |
128 | ||
bdc480e3 | 129 | SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct statfs __user *, buf) |
1da177e4 | 130 | { |
2d8f3038 | 131 | struct path path; |
1da177e4 LT |
132 | int error; |
133 | ||
2d8f3038 | 134 | error = user_path(pathname, &path); |
1da177e4 LT |
135 | if (!error) { |
136 | struct statfs tmp; | |
2d8f3038 | 137 | error = vfs_statfs_native(path.dentry, &tmp); |
1da177e4 LT |
138 | if (!error && copy_to_user(buf, &tmp, sizeof(tmp))) |
139 | error = -EFAULT; | |
2d8f3038 | 140 | path_put(&path); |
1da177e4 LT |
141 | } |
142 | return error; | |
143 | } | |
144 | ||
bdc480e3 | 145 | SYSCALL_DEFINE3(statfs64, const char __user *, pathname, size_t, sz, struct statfs64 __user *, buf) |
1da177e4 | 146 | { |
2d8f3038 | 147 | struct path path; |
1da177e4 LT |
148 | long error; |
149 | ||
150 | if (sz != sizeof(*buf)) | |
151 | return -EINVAL; | |
2d8f3038 | 152 | error = user_path(pathname, &path); |
1da177e4 LT |
153 | if (!error) { |
154 | struct statfs64 tmp; | |
2d8f3038 | 155 | error = vfs_statfs64(path.dentry, &tmp); |
1da177e4 LT |
156 | if (!error && copy_to_user(buf, &tmp, sizeof(tmp))) |
157 | error = -EFAULT; | |
2d8f3038 | 158 | path_put(&path); |
1da177e4 LT |
159 | } |
160 | return error; | |
161 | } | |
162 | ||
bdc480e3 | 163 | SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct statfs __user *, buf) |
1da177e4 LT |
164 | { |
165 | struct file * file; | |
166 | struct statfs tmp; | |
167 | int error; | |
168 | ||
169 | error = -EBADF; | |
170 | file = fget(fd); | |
171 | if (!file) | |
172 | goto out; | |
0f7fc9e4 | 173 | error = vfs_statfs_native(file->f_path.dentry, &tmp); |
1da177e4 LT |
174 | if (!error && copy_to_user(buf, &tmp, sizeof(tmp))) |
175 | error = -EFAULT; | |
176 | fput(file); | |
177 | out: | |
178 | return error; | |
179 | } | |
180 | ||
257ac264 | 181 | SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, size_t, sz, struct statfs64 __user *, buf) |
1da177e4 LT |
182 | { |
183 | struct file * file; | |
184 | struct statfs64 tmp; | |
185 | int error; | |
186 | ||
187 | if (sz != sizeof(*buf)) | |
188 | return -EINVAL; | |
189 | ||
190 | error = -EBADF; | |
191 | file = fget(fd); | |
192 | if (!file) | |
193 | goto out; | |
0f7fc9e4 | 194 | error = vfs_statfs64(file->f_path.dentry, &tmp); |
1da177e4 LT |
195 | if (!error && copy_to_user(buf, &tmp, sizeof(tmp))) |
196 | error = -EFAULT; | |
197 | fput(file); | |
198 | out: | |
199 | return error; | |
200 | } | |
201 | ||
4a30131e N |
202 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, |
203 | struct file *filp) | |
1da177e4 | 204 | { |
939a9421 | 205 | int ret; |
1da177e4 LT |
206 | struct iattr newattrs; |
207 | ||
208 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
209 | if (length < 0) | |
210 | return -EINVAL; | |
211 | ||
212 | newattrs.ia_size = length; | |
4a30131e | 213 | newattrs.ia_valid = ATTR_SIZE | time_attrs; |
cc4e69de MS |
214 | if (filp) { |
215 | newattrs.ia_file = filp; | |
216 | newattrs.ia_valid |= ATTR_FILE; | |
217 | } | |
1da177e4 | 218 | |
7b82dc0e | 219 | /* Remove suid/sgid on truncate too */ |
939a9421 AW |
220 | ret = should_remove_suid(dentry); |
221 | if (ret) | |
222 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
7b82dc0e | 223 | |
1b1dcc1b | 224 | mutex_lock(&dentry->d_inode->i_mutex); |
939a9421 | 225 | ret = notify_change(dentry, &newattrs); |
1b1dcc1b | 226 | mutex_unlock(&dentry->d_inode->i_mutex); |
939a9421 | 227 | return ret; |
1da177e4 LT |
228 | } |
229 | ||
2d8f3038 | 230 | static long do_sys_truncate(const char __user *pathname, loff_t length) |
1da177e4 | 231 | { |
2d8f3038 AV |
232 | struct path path; |
233 | struct inode *inode; | |
1da177e4 LT |
234 | int error; |
235 | ||
236 | error = -EINVAL; | |
237 | if (length < 0) /* sorry, but loff_t says... */ | |
238 | goto out; | |
239 | ||
2d8f3038 | 240 | error = user_path(pathname, &path); |
1da177e4 LT |
241 | if (error) |
242 | goto out; | |
2d8f3038 | 243 | inode = path.dentry->d_inode; |
1da177e4 LT |
244 | |
245 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
246 | error = -EISDIR; | |
247 | if (S_ISDIR(inode->i_mode)) | |
248 | goto dput_and_out; | |
249 | ||
250 | error = -EINVAL; | |
251 | if (!S_ISREG(inode->i_mode)) | |
252 | goto dput_and_out; | |
253 | ||
2d8f3038 | 254 | error = mnt_want_write(path.mnt); |
1da177e4 LT |
255 | if (error) |
256 | goto dput_and_out; | |
257 | ||
256984a8 | 258 | error = inode_permission(inode, MAY_WRITE); |
9ac9b847 DH |
259 | if (error) |
260 | goto mnt_drop_write_and_out; | |
1da177e4 LT |
261 | |
262 | error = -EPERM; | |
c82e42da | 263 | if (IS_APPEND(inode)) |
9ac9b847 | 264 | goto mnt_drop_write_and_out; |
1da177e4 | 265 | |
9700382c | 266 | error = get_write_access(inode); |
1da177e4 | 267 | if (error) |
9ac9b847 | 268 | goto mnt_drop_write_and_out; |
1da177e4 | 269 | |
9700382c | 270 | /* |
271 | * Make sure that there are no leases. get_write_access() protects | |
272 | * against the truncate racing with a lease-granting setlease(). | |
273 | */ | |
274 | error = break_lease(inode, FMODE_WRITE); | |
1da177e4 | 275 | if (error) |
9700382c | 276 | goto put_write_and_out; |
1da177e4 LT |
277 | |
278 | error = locks_verify_truncate(inode, NULL, length); | |
be6d3e56 KT |
279 | if (!error) |
280 | error = security_path_truncate(&path, length, 0); | |
1da177e4 | 281 | if (!error) { |
9e3509e2 | 282 | vfs_dq_init(inode); |
2d8f3038 | 283 | error = do_truncate(path.dentry, length, 0, NULL); |
1da177e4 | 284 | } |
1da177e4 | 285 | |
9700382c | 286 | put_write_and_out: |
287 | put_write_access(inode); | |
9ac9b847 | 288 | mnt_drop_write_and_out: |
2d8f3038 | 289 | mnt_drop_write(path.mnt); |
1da177e4 | 290 | dput_and_out: |
2d8f3038 | 291 | path_put(&path); |
1da177e4 LT |
292 | out: |
293 | return error; | |
294 | } | |
295 | ||
4fd8da8d | 296 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) |
1da177e4 | 297 | { |
4fd8da8d | 298 | return do_sys_truncate(path, length); |
1da177e4 LT |
299 | } |
300 | ||
b01ec0ef | 301 | static long do_sys_ftruncate(unsigned int fd, loff_t length, int small) |
1da177e4 LT |
302 | { |
303 | struct inode * inode; | |
304 | struct dentry *dentry; | |
305 | struct file * file; | |
306 | int error; | |
307 | ||
308 | error = -EINVAL; | |
309 | if (length < 0) | |
310 | goto out; | |
311 | error = -EBADF; | |
312 | file = fget(fd); | |
313 | if (!file) | |
314 | goto out; | |
315 | ||
316 | /* explicitly opened as large or we are on 64-bit box */ | |
317 | if (file->f_flags & O_LARGEFILE) | |
318 | small = 0; | |
319 | ||
0f7fc9e4 | 320 | dentry = file->f_path.dentry; |
1da177e4 LT |
321 | inode = dentry->d_inode; |
322 | error = -EINVAL; | |
323 | if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE)) | |
324 | goto out_putf; | |
325 | ||
326 | error = -EINVAL; | |
327 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
328 | if (small && length > MAX_NON_LFS) | |
329 | goto out_putf; | |
330 | ||
331 | error = -EPERM; | |
332 | if (IS_APPEND(inode)) | |
333 | goto out_putf; | |
334 | ||
335 | error = locks_verify_truncate(inode, file, length); | |
be6d3e56 KT |
336 | if (!error) |
337 | error = security_path_truncate(&file->f_path, length, | |
338 | ATTR_MTIME|ATTR_CTIME); | |
1da177e4 | 339 | if (!error) |
6e656be8 | 340 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file); |
1da177e4 LT |
341 | out_putf: |
342 | fput(file); | |
343 | out: | |
344 | return error; | |
345 | } | |
346 | ||
bdc480e3 | 347 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) |
1da177e4 | 348 | { |
0a489cb3 | 349 | long ret = do_sys_ftruncate(fd, length, 1); |
385910f2 | 350 | /* avoid REGPARM breakage on x86: */ |
54a01510 | 351 | asmlinkage_protect(2, ret, fd, length); |
0a489cb3 | 352 | return ret; |
1da177e4 LT |
353 | } |
354 | ||
355 | /* LFS versions of truncate are only needed on 32 bit machines */ | |
356 | #if BITS_PER_LONG == 32 | |
6673e0c3 | 357 | SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length) |
1da177e4 LT |
358 | { |
359 | return do_sys_truncate(path, length); | |
360 | } | |
6673e0c3 HC |
361 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS |
362 | asmlinkage long SyS_truncate64(long path, loff_t length) | |
363 | { | |
364 | return SYSC_truncate64((const char __user *) path, length); | |
365 | } | |
366 | SYSCALL_ALIAS(sys_truncate64, SyS_truncate64); | |
367 | #endif | |
1da177e4 | 368 | |
6673e0c3 | 369 | SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length) |
1da177e4 | 370 | { |
0a489cb3 | 371 | long ret = do_sys_ftruncate(fd, length, 0); |
385910f2 | 372 | /* avoid REGPARM breakage on x86: */ |
54a01510 | 373 | asmlinkage_protect(2, ret, fd, length); |
0a489cb3 | 374 | return ret; |
1da177e4 | 375 | } |
6673e0c3 HC |
376 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS |
377 | asmlinkage long SyS_ftruncate64(long fd, loff_t length) | |
378 | { | |
379 | return SYSC_ftruncate64((unsigned int) fd, length); | |
380 | } | |
381 | SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64); | |
1da177e4 | 382 | #endif |
6673e0c3 | 383 | #endif /* BITS_PER_LONG == 32 */ |
1da177e4 | 384 | |
3e63cbb1 AJ |
385 | |
386 | int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len) | |
97ac7350 | 387 | { |
3e63cbb1 AJ |
388 | struct inode *inode = file->f_path.dentry->d_inode; |
389 | long ret; | |
97ac7350 AA |
390 | |
391 | if (offset < 0 || len <= 0) | |
3e63cbb1 | 392 | return -EINVAL; |
97ac7350 AA |
393 | |
394 | /* Return error if mode is not supported */ | |
97ac7350 | 395 | if (mode && !(mode & FALLOC_FL_KEEP_SIZE)) |
3e63cbb1 | 396 | return -EOPNOTSUPP; |
97ac7350 | 397 | |
97ac7350 | 398 | if (!(file->f_mode & FMODE_WRITE)) |
3e63cbb1 | 399 | return -EBADF; |
97ac7350 AA |
400 | /* |
401 | * Revalidate the write permissions, in case security policy has | |
402 | * changed since the files were opened. | |
403 | */ | |
404 | ret = security_file_permission(file, MAY_WRITE); | |
405 | if (ret) | |
3e63cbb1 | 406 | return ret; |
97ac7350 | 407 | |
97ac7350 | 408 | if (S_ISFIFO(inode->i_mode)) |
3e63cbb1 | 409 | return -ESPIPE; |
97ac7350 | 410 | |
97ac7350 AA |
411 | /* |
412 | * Let individual file system decide if it supports preallocation | |
413 | * for directories or not. | |
414 | */ | |
415 | if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode)) | |
3e63cbb1 | 416 | return -ENODEV; |
97ac7350 | 417 | |
97ac7350 AA |
418 | /* Check for wrap through zero too */ |
419 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
3e63cbb1 | 420 | return -EFBIG; |
97ac7350 | 421 | |
3e63cbb1 AJ |
422 | if (!inode->i_op->fallocate) |
423 | return -EOPNOTSUPP; | |
97ac7350 | 424 | |
3e63cbb1 AJ |
425 | return inode->i_op->fallocate(inode, mode, offset, len); |
426 | } | |
427 | ||
428 | SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len) | |
429 | { | |
430 | struct file *file; | |
431 | int error = -EBADF; | |
432 | ||
433 | file = fget(fd); | |
434 | if (file) { | |
435 | error = do_fallocate(file, mode, offset, len); | |
436 | fput(file); | |
437 | } | |
438 | ||
439 | return error; | |
97ac7350 | 440 | } |
3e63cbb1 | 441 | |
6673e0c3 HC |
442 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS |
443 | asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len) | |
444 | { | |
445 | return SYSC_fallocate((int)fd, (int)mode, offset, len); | |
446 | } | |
447 | SYSCALL_ALIAS(sys_fallocate, SyS_fallocate); | |
448 | #endif | |
97ac7350 | 449 | |
1da177e4 LT |
450 | /* |
451 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
452 | * We do this by temporarily clearing all FS-related capabilities and | |
453 | * switching the fsuid/fsgid around to the real ones. | |
454 | */ | |
6559eed8 | 455 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) |
1da177e4 | 456 | { |
d84f4f99 DH |
457 | const struct cred *old_cred; |
458 | struct cred *override_cred; | |
2d8f3038 | 459 | struct path path; |
256984a8 | 460 | struct inode *inode; |
1da177e4 LT |
461 | int res; |
462 | ||
463 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
464 | return -EINVAL; | |
465 | ||
d84f4f99 DH |
466 | override_cred = prepare_creds(); |
467 | if (!override_cred) | |
468 | return -ENOMEM; | |
1da177e4 | 469 | |
d84f4f99 DH |
470 | override_cred->fsuid = override_cred->uid; |
471 | override_cred->fsgid = override_cred->gid; | |
1da177e4 | 472 | |
086f7316 | 473 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { |
1cdcbec1 | 474 | /* Clear the capabilities if we switch to a non-root user */ |
d84f4f99 DH |
475 | if (override_cred->uid) |
476 | cap_clear(override_cred->cap_effective); | |
086f7316 | 477 | else |
d84f4f99 DH |
478 | override_cred->cap_effective = |
479 | override_cred->cap_permitted; | |
086f7316 | 480 | } |
1da177e4 | 481 | |
d84f4f99 DH |
482 | old_cred = override_creds(override_cred); |
483 | ||
2d8f3038 | 484 | res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path); |
6902d925 DH |
485 | if (res) |
486 | goto out; | |
487 | ||
2d8f3038 | 488 | inode = path.dentry->d_inode; |
256984a8 AV |
489 | |
490 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
30524472 AV |
491 | /* |
492 | * MAY_EXEC on regular files is denied if the fs is mounted | |
493 | * with the "noexec" flag. | |
494 | */ | |
495 | res = -EACCES; | |
2d8f3038 | 496 | if (path.mnt->mnt_flags & MNT_NOEXEC) |
30524472 AV |
497 | goto out_path_release; |
498 | } | |
499 | ||
256984a8 | 500 | res = inode_permission(inode, mode | MAY_ACCESS); |
6902d925 | 501 | /* SuS v2 requires we report a read only fs too */ |
256984a8 | 502 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) |
6902d925 | 503 | goto out_path_release; |
2f676cbc DH |
504 | /* |
505 | * This is a rare case where using __mnt_is_readonly() | |
506 | * is OK without a mnt_want/drop_write() pair. Since | |
507 | * no actual write to the fs is performed here, we do | |
508 | * not need to telegraph to that to anyone. | |
509 | * | |
510 | * By doing this, we accept that this access is | |
511 | * inherently racy and know that the fs may change | |
512 | * state before we even see this result. | |
513 | */ | |
2d8f3038 | 514 | if (__mnt_is_readonly(path.mnt)) |
6902d925 | 515 | res = -EROFS; |
1da177e4 | 516 | |
6902d925 | 517 | out_path_release: |
2d8f3038 | 518 | path_put(&path); |
6902d925 | 519 | out: |
d84f4f99 DH |
520 | revert_creds(old_cred); |
521 | put_cred(override_cred); | |
1da177e4 LT |
522 | return res; |
523 | } | |
524 | ||
ca013e94 | 525 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) |
5590ff0d UD |
526 | { |
527 | return sys_faccessat(AT_FDCWD, filename, mode); | |
528 | } | |
529 | ||
3cdad428 | 530 | SYSCALL_DEFINE1(chdir, const char __user *, filename) |
1da177e4 | 531 | { |
2d8f3038 | 532 | struct path path; |
1da177e4 LT |
533 | int error; |
534 | ||
2d8f3038 | 535 | error = user_path_dir(filename, &path); |
1da177e4 LT |
536 | if (error) |
537 | goto out; | |
538 | ||
2d8f3038 | 539 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS); |
1da177e4 LT |
540 | if (error) |
541 | goto dput_and_out; | |
542 | ||
2d8f3038 | 543 | set_fs_pwd(current->fs, &path); |
1da177e4 LT |
544 | |
545 | dput_and_out: | |
2d8f3038 | 546 | path_put(&path); |
1da177e4 LT |
547 | out: |
548 | return error; | |
549 | } | |
550 | ||
3cdad428 | 551 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) |
1da177e4 LT |
552 | { |
553 | struct file *file; | |
1da177e4 | 554 | struct inode *inode; |
1da177e4 LT |
555 | int error; |
556 | ||
557 | error = -EBADF; | |
558 | file = fget(fd); | |
559 | if (!file) | |
560 | goto out; | |
561 | ||
ac748a09 | 562 | inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
563 | |
564 | error = -ENOTDIR; | |
565 | if (!S_ISDIR(inode->i_mode)) | |
566 | goto out_putf; | |
567 | ||
256984a8 | 568 | error = inode_permission(inode, MAY_EXEC | MAY_ACCESS); |
1da177e4 | 569 | if (!error) |
ac748a09 | 570 | set_fs_pwd(current->fs, &file->f_path); |
1da177e4 LT |
571 | out_putf: |
572 | fput(file); | |
573 | out: | |
574 | return error; | |
575 | } | |
576 | ||
3480b257 | 577 | SYSCALL_DEFINE1(chroot, const char __user *, filename) |
1da177e4 | 578 | { |
2d8f3038 | 579 | struct path path; |
1da177e4 LT |
580 | int error; |
581 | ||
2d8f3038 | 582 | error = user_path_dir(filename, &path); |
1da177e4 LT |
583 | if (error) |
584 | goto out; | |
585 | ||
2d8f3038 | 586 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS); |
1da177e4 LT |
587 | if (error) |
588 | goto dput_and_out; | |
589 | ||
590 | error = -EPERM; | |
591 | if (!capable(CAP_SYS_CHROOT)) | |
592 | goto dput_and_out; | |
8b8efb44 TH |
593 | error = security_path_chroot(&path); |
594 | if (error) | |
595 | goto dput_and_out; | |
1da177e4 | 596 | |
2d8f3038 | 597 | set_fs_root(current->fs, &path); |
1da177e4 LT |
598 | error = 0; |
599 | dput_and_out: | |
2d8f3038 | 600 | path_put(&path); |
1da177e4 LT |
601 | out: |
602 | return error; | |
603 | } | |
604 | ||
a26eab24 | 605 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode) |
1da177e4 LT |
606 | { |
607 | struct inode * inode; | |
608 | struct dentry * dentry; | |
609 | struct file * file; | |
610 | int err = -EBADF; | |
611 | struct iattr newattrs; | |
612 | ||
613 | file = fget(fd); | |
614 | if (!file) | |
615 | goto out; | |
616 | ||
0f7fc9e4 | 617 | dentry = file->f_path.dentry; |
1da177e4 LT |
618 | inode = dentry->d_inode; |
619 | ||
5a190ae6 | 620 | audit_inode(NULL, dentry); |
73241ccc | 621 | |
96029c4e | 622 | err = mnt_want_write_file(file); |
2af482a7 | 623 | if (err) |
1da177e4 | 624 | goto out_putf; |
fe542cf5 | 625 | mutex_lock(&inode->i_mutex); |
89eda068 TH |
626 | err = security_path_chmod(dentry, file->f_vfsmnt, mode); |
627 | if (err) | |
fe542cf5 | 628 | goto out_unlock; |
1da177e4 LT |
629 | if (mode == (mode_t) -1) |
630 | mode = inode->i_mode; | |
631 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); | |
632 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
633 | err = notify_change(dentry, &newattrs); | |
fe542cf5 | 634 | out_unlock: |
1b1dcc1b | 635 | mutex_unlock(&inode->i_mutex); |
2af482a7 | 636 | mnt_drop_write(file->f_path.mnt); |
1da177e4 LT |
637 | out_putf: |
638 | fput(file); | |
639 | out: | |
640 | return err; | |
641 | } | |
642 | ||
6559eed8 | 643 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode) |
1da177e4 | 644 | { |
2d8f3038 AV |
645 | struct path path; |
646 | struct inode *inode; | |
1da177e4 LT |
647 | int error; |
648 | struct iattr newattrs; | |
649 | ||
2d8f3038 | 650 | error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path); |
1da177e4 LT |
651 | if (error) |
652 | goto out; | |
2d8f3038 | 653 | inode = path.dentry->d_inode; |
1da177e4 | 654 | |
2d8f3038 | 655 | error = mnt_want_write(path.mnt); |
2af482a7 | 656 | if (error) |
1da177e4 | 657 | goto dput_and_out; |
fe542cf5 | 658 | mutex_lock(&inode->i_mutex); |
89eda068 TH |
659 | error = security_path_chmod(path.dentry, path.mnt, mode); |
660 | if (error) | |
fe542cf5 | 661 | goto out_unlock; |
1da177e4 LT |
662 | if (mode == (mode_t) -1) |
663 | mode = inode->i_mode; | |
664 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); | |
665 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
2d8f3038 | 666 | error = notify_change(path.dentry, &newattrs); |
fe542cf5 | 667 | out_unlock: |
1b1dcc1b | 668 | mutex_unlock(&inode->i_mutex); |
2d8f3038 | 669 | mnt_drop_write(path.mnt); |
1da177e4 | 670 | dput_and_out: |
2d8f3038 | 671 | path_put(&path); |
1da177e4 LT |
672 | out: |
673 | return error; | |
674 | } | |
675 | ||
a26eab24 | 676 | SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode) |
5590ff0d UD |
677 | { |
678 | return sys_fchmodat(AT_FDCWD, filename, mode); | |
679 | } | |
680 | ||
fe542cf5 | 681 | static int chown_common(struct path *path, uid_t user, gid_t group) |
1da177e4 | 682 | { |
fe542cf5 | 683 | struct inode *inode = path->dentry->d_inode; |
1da177e4 LT |
684 | int error; |
685 | struct iattr newattrs; | |
686 | ||
1da177e4 LT |
687 | newattrs.ia_valid = ATTR_CTIME; |
688 | if (user != (uid_t) -1) { | |
689 | newattrs.ia_valid |= ATTR_UID; | |
690 | newattrs.ia_uid = user; | |
691 | } | |
692 | if (group != (gid_t) -1) { | |
693 | newattrs.ia_valid |= ATTR_GID; | |
694 | newattrs.ia_gid = group; | |
695 | } | |
696 | if (!S_ISDIR(inode->i_mode)) | |
b5376771 SH |
697 | newattrs.ia_valid |= |
698 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
1b1dcc1b | 699 | mutex_lock(&inode->i_mutex); |
fe542cf5 TH |
700 | error = security_path_chown(path, user, group); |
701 | if (!error) | |
702 | error = notify_change(path->dentry, &newattrs); | |
1b1dcc1b | 703 | mutex_unlock(&inode->i_mutex); |
beb29e05 | 704 | |
1da177e4 LT |
705 | return error; |
706 | } | |
707 | ||
ca013e94 | 708 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 709 | { |
2d8f3038 | 710 | struct path path; |
1da177e4 LT |
711 | int error; |
712 | ||
2d8f3038 | 713 | error = user_path(filename, &path); |
6902d925 DH |
714 | if (error) |
715 | goto out; | |
2d8f3038 | 716 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
717 | if (error) |
718 | goto out_release; | |
fe542cf5 | 719 | error = chown_common(&path, user, group); |
2d8f3038 | 720 | mnt_drop_write(path.mnt); |
2af482a7 | 721 | out_release: |
2d8f3038 | 722 | path_put(&path); |
6902d925 | 723 | out: |
1da177e4 LT |
724 | return error; |
725 | } | |
726 | ||
6559eed8 HC |
727 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, |
728 | gid_t, group, int, flag) | |
5590ff0d | 729 | { |
2d8f3038 | 730 | struct path path; |
5590ff0d UD |
731 | int error = -EINVAL; |
732 | int follow; | |
733 | ||
734 | if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0) | |
735 | goto out; | |
736 | ||
737 | follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; | |
2d8f3038 | 738 | error = user_path_at(dfd, filename, follow, &path); |
6902d925 DH |
739 | if (error) |
740 | goto out; | |
2d8f3038 | 741 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
742 | if (error) |
743 | goto out_release; | |
fe542cf5 | 744 | error = chown_common(&path, user, group); |
2d8f3038 | 745 | mnt_drop_write(path.mnt); |
2af482a7 | 746 | out_release: |
2d8f3038 | 747 | path_put(&path); |
5590ff0d UD |
748 | out: |
749 | return error; | |
750 | } | |
751 | ||
ca013e94 | 752 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) |
1da177e4 | 753 | { |
2d8f3038 | 754 | struct path path; |
1da177e4 LT |
755 | int error; |
756 | ||
2d8f3038 | 757 | error = user_lpath(filename, &path); |
6902d925 DH |
758 | if (error) |
759 | goto out; | |
2d8f3038 | 760 | error = mnt_want_write(path.mnt); |
2af482a7 DH |
761 | if (error) |
762 | goto out_release; | |
fe542cf5 | 763 | error = chown_common(&path, user, group); |
2d8f3038 | 764 | mnt_drop_write(path.mnt); |
2af482a7 | 765 | out_release: |
2d8f3038 | 766 | path_put(&path); |
6902d925 | 767 | out: |
1da177e4 LT |
768 | return error; |
769 | } | |
770 | ||
ca013e94 | 771 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) |
1da177e4 LT |
772 | { |
773 | struct file * file; | |
774 | int error = -EBADF; | |
6902d925 | 775 | struct dentry * dentry; |
1da177e4 LT |
776 | |
777 | file = fget(fd); | |
6902d925 DH |
778 | if (!file) |
779 | goto out; | |
780 | ||
96029c4e | 781 | error = mnt_want_write_file(file); |
2af482a7 DH |
782 | if (error) |
783 | goto out_fput; | |
0f7fc9e4 | 784 | dentry = file->f_path.dentry; |
5a190ae6 | 785 | audit_inode(NULL, dentry); |
fe542cf5 | 786 | error = chown_common(&file->f_path, user, group); |
2af482a7 DH |
787 | mnt_drop_write(file->f_path.mnt); |
788 | out_fput: | |
6902d925 DH |
789 | fput(file); |
790 | out: | |
1da177e4 LT |
791 | return error; |
792 | } | |
793 | ||
4a3fd211 DH |
794 | /* |
795 | * You have to be very careful that these write | |
796 | * counts get cleaned up in error cases and | |
797 | * upon __fput(). This should probably never | |
798 | * be called outside of __dentry_open(). | |
799 | */ | |
800 | static inline int __get_file_write_access(struct inode *inode, | |
801 | struct vfsmount *mnt) | |
802 | { | |
803 | int error; | |
804 | error = get_write_access(inode); | |
805 | if (error) | |
806 | return error; | |
807 | /* | |
808 | * Do not take mount writer counts on | |
809 | * special files since no writes to | |
810 | * the mount itself will occur. | |
811 | */ | |
812 | if (!special_file(inode->i_mode)) { | |
813 | /* | |
814 | * Balanced in __fput() | |
815 | */ | |
816 | error = mnt_want_write(mnt); | |
817 | if (error) | |
818 | put_write_access(inode); | |
819 | } | |
820 | return error; | |
821 | } | |
822 | ||
a1a5b3d9 | 823 | static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt, |
482928d5 | 824 | struct file *f, |
745ca247 DH |
825 | int (*open)(struct inode *, struct file *), |
826 | const struct cred *cred) | |
1da177e4 | 827 | { |
1da177e4 LT |
828 | struct inode *inode; |
829 | int error; | |
830 | ||
5300990c | 831 | f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK | |
a1a5b3d9 | 832 | FMODE_PREAD | FMODE_PWRITE; |
1da177e4 LT |
833 | inode = dentry->d_inode; |
834 | if (f->f_mode & FMODE_WRITE) { | |
4a3fd211 | 835 | error = __get_file_write_access(inode, mnt); |
1da177e4 LT |
836 | if (error) |
837 | goto cleanup_file; | |
ad775f5a DH |
838 | if (!special_file(inode->i_mode)) |
839 | file_take_write(f); | |
1da177e4 LT |
840 | } |
841 | ||
842 | f->f_mapping = inode->i_mapping; | |
0f7fc9e4 JJS |
843 | f->f_path.dentry = dentry; |
844 | f->f_path.mnt = mnt; | |
1da177e4 LT |
845 | f->f_pos = 0; |
846 | f->f_op = fops_get(inode->i_fop); | |
847 | file_move(f, &inode->i_sb->s_files); | |
848 | ||
745ca247 | 849 | error = security_dentry_open(f, cred); |
788e7dd4 YN |
850 | if (error) |
851 | goto cleanup_all; | |
852 | ||
834f2a4a TM |
853 | if (!open && f->f_op) |
854 | open = f->f_op->open; | |
855 | if (open) { | |
856 | error = open(inode, f); | |
1da177e4 LT |
857 | if (error) |
858 | goto cleanup_all; | |
859 | } | |
b65a9cfc | 860 | ima_counts_get(f); |
834f2a4a | 861 | |
1da177e4 LT |
862 | f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); |
863 | ||
864 | file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping); | |
865 | ||
866 | /* NB: we're sure to have correct a_ops only after f_op->open */ | |
867 | if (f->f_flags & O_DIRECT) { | |
ceffc078 CO |
868 | if (!f->f_mapping->a_ops || |
869 | ((!f->f_mapping->a_ops->direct_IO) && | |
70688e4d | 870 | (!f->f_mapping->a_ops->get_xip_mem))) { |
1da177e4 LT |
871 | fput(f); |
872 | f = ERR_PTR(-EINVAL); | |
873 | } | |
874 | } | |
875 | ||
876 | return f; | |
877 | ||
878 | cleanup_all: | |
879 | fops_put(f->f_op); | |
4a3fd211 | 880 | if (f->f_mode & FMODE_WRITE) { |
1da177e4 | 881 | put_write_access(inode); |
ad775f5a DH |
882 | if (!special_file(inode->i_mode)) { |
883 | /* | |
884 | * We don't consider this a real | |
885 | * mnt_want/drop_write() pair | |
886 | * because it all happenend right | |
887 | * here, so just reset the state. | |
888 | */ | |
889 | file_reset_write(f); | |
4a3fd211 | 890 | mnt_drop_write(mnt); |
ad775f5a | 891 | } |
4a3fd211 | 892 | } |
1da177e4 | 893 | file_kill(f); |
0f7fc9e4 JJS |
894 | f->f_path.dentry = NULL; |
895 | f->f_path.mnt = NULL; | |
1da177e4 LT |
896 | cleanup_file: |
897 | put_filp(f); | |
1da177e4 LT |
898 | dput(dentry); |
899 | mntput(mnt); | |
900 | return ERR_PTR(error); | |
901 | } | |
902 | ||
834f2a4a TM |
903 | /** |
904 | * lookup_instantiate_filp - instantiates the open intent filp | |
905 | * @nd: pointer to nameidata | |
906 | * @dentry: pointer to dentry | |
907 | * @open: open callback | |
908 | * | |
909 | * Helper for filesystems that want to use lookup open intents and pass back | |
910 | * a fully instantiated struct file to the caller. | |
911 | * This function is meant to be called from within a filesystem's | |
912 | * lookup method. | |
9a56c213 OD |
913 | * Beware of calling it for non-regular files! Those ->open methods might block |
914 | * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo, | |
915 | * leading to a deadlock, as nobody can open that fifo anymore, because | |
916 | * another process to open fifo will block on locked parent when doing lookup). | |
834f2a4a TM |
917 | * Note that in case of error, nd->intent.open.file is destroyed, but the |
918 | * path information remains valid. | |
919 | * If the open callback is set to NULL, then the standard f_op->open() | |
920 | * filesystem callback is substituted. | |
921 | */ | |
922 | struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry, | |
923 | int (*open)(struct inode *, struct file *)) | |
924 | { | |
745ca247 DH |
925 | const struct cred *cred = current_cred(); |
926 | ||
834f2a4a TM |
927 | if (IS_ERR(nd->intent.open.file)) |
928 | goto out; | |
929 | if (IS_ERR(dentry)) | |
930 | goto out_err; | |
4ac91378 | 931 | nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt), |
834f2a4a | 932 | nd->intent.open.file, |
745ca247 | 933 | open, cred); |
834f2a4a TM |
934 | out: |
935 | return nd->intent.open.file; | |
936 | out_err: | |
937 | release_open_intent(nd); | |
938 | nd->intent.open.file = (struct file *)dentry; | |
939 | goto out; | |
940 | } | |
941 | EXPORT_SYMBOL_GPL(lookup_instantiate_filp); | |
942 | ||
943 | /** | |
944 | * nameidata_to_filp - convert a nameidata to an open filp. | |
945 | * @nd: pointer to nameidata | |
946 | * @flags: open flags | |
947 | * | |
948 | * Note that this function destroys the original nameidata | |
949 | */ | |
482928d5 | 950 | struct file *nameidata_to_filp(struct nameidata *nd) |
834f2a4a | 951 | { |
745ca247 | 952 | const struct cred *cred = current_cred(); |
834f2a4a TM |
953 | struct file *filp; |
954 | ||
955 | /* Pick up the filp from the open intent */ | |
956 | filp = nd->intent.open.file; | |
957 | /* Has the filesystem initialised the file for us? */ | |
0f7fc9e4 | 958 | if (filp->f_path.dentry == NULL) |
482928d5 | 959 | filp = __dentry_open(nd->path.dentry, nd->path.mnt, filp, |
745ca247 | 960 | NULL, cred); |
834f2a4a | 961 | else |
1d957f9b | 962 | path_put(&nd->path); |
834f2a4a TM |
963 | return filp; |
964 | } | |
965 | ||
6fdcc216 PS |
966 | /* |
967 | * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an | |
968 | * error. | |
969 | */ | |
745ca247 DH |
970 | struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags, |
971 | const struct cred *cred) | |
a1a5b3d9 PS |
972 | { |
973 | int error; | |
974 | struct file *f; | |
975 | ||
e0e81739 DH |
976 | validate_creds(cred); |
977 | ||
322ee5b3 CH |
978 | /* |
979 | * We must always pass in a valid mount pointer. Historically | |
980 | * callers got away with not passing it, but we must enforce this at | |
981 | * the earliest possible point now to avoid strange problems deep in the | |
982 | * filesystem stack. | |
983 | */ | |
984 | if (!mnt) { | |
985 | printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__); | |
986 | dump_stack(); | |
987 | return ERR_PTR(-EINVAL); | |
988 | } | |
989 | ||
a1a5b3d9 PS |
990 | error = -ENFILE; |
991 | f = get_empty_filp(); | |
6fdcc216 PS |
992 | if (f == NULL) { |
993 | dput(dentry); | |
994 | mntput(mnt); | |
a1a5b3d9 | 995 | return ERR_PTR(error); |
6fdcc216 | 996 | } |
a1a5b3d9 | 997 | |
482928d5 AV |
998 | f->f_flags = flags; |
999 | return __dentry_open(dentry, mnt, f, NULL, cred); | |
a1a5b3d9 | 1000 | } |
1da177e4 LT |
1001 | EXPORT_SYMBOL(dentry_open); |
1002 | ||
b01ec0ef | 1003 | static void __put_unused_fd(struct files_struct *files, unsigned int fd) |
1da177e4 | 1004 | { |
badf1662 DS |
1005 | struct fdtable *fdt = files_fdtable(files); |
1006 | __FD_CLR(fd, fdt->open_fds); | |
0c9e63fd ED |
1007 | if (fd < files->next_fd) |
1008 | files->next_fd = fd; | |
1da177e4 LT |
1009 | } |
1010 | ||
fc9b52cd | 1011 | void put_unused_fd(unsigned int fd) |
1da177e4 LT |
1012 | { |
1013 | struct files_struct *files = current->files; | |
1014 | spin_lock(&files->file_lock); | |
1015 | __put_unused_fd(files, fd); | |
1016 | spin_unlock(&files->file_lock); | |
1017 | } | |
1018 | ||
1019 | EXPORT_SYMBOL(put_unused_fd); | |
1020 | ||
1021 | /* | |
5590ff0d | 1022 | * Install a file pointer in the fd array. |
1da177e4 LT |
1023 | * |
1024 | * The VFS is full of places where we drop the files lock between | |
1025 | * setting the open_fds bitmap and installing the file in the file | |
1026 | * array. At any such point, we are vulnerable to a dup2() race | |
1027 | * installing a file in the array before us. We need to detect this and | |
1028 | * fput() the struct file we are about to overwrite in this case. | |
1029 | * | |
1030 | * It should never happen - if we allow dup2() do it, _really_ bad things | |
1031 | * will follow. | |
1032 | */ | |
1033 | ||
fc9b52cd | 1034 | void fd_install(unsigned int fd, struct file *file) |
1da177e4 LT |
1035 | { |
1036 | struct files_struct *files = current->files; | |
badf1662 | 1037 | struct fdtable *fdt; |
1da177e4 | 1038 | spin_lock(&files->file_lock); |
badf1662 | 1039 | fdt = files_fdtable(files); |
ab2af1f5 DS |
1040 | BUG_ON(fdt->fd[fd] != NULL); |
1041 | rcu_assign_pointer(fdt->fd[fd], file); | |
1da177e4 LT |
1042 | spin_unlock(&files->file_lock); |
1043 | } | |
1044 | ||
1045 | EXPORT_SYMBOL(fd_install); | |
1046 | ||
5590ff0d | 1047 | long do_sys_open(int dfd, const char __user *filename, int flags, int mode) |
1da177e4 | 1048 | { |
e922efc3 MS |
1049 | char *tmp = getname(filename); |
1050 | int fd = PTR_ERR(tmp); | |
1da177e4 | 1051 | |
1da177e4 | 1052 | if (!IS_ERR(tmp)) { |
f23513e8 | 1053 | fd = get_unused_fd_flags(flags); |
1da177e4 | 1054 | if (fd >= 0) { |
6e8341a1 | 1055 | struct file *f = do_filp_open(dfd, tmp, flags, mode, 0); |
fed2fc18 TN |
1056 | if (IS_ERR(f)) { |
1057 | put_unused_fd(fd); | |
1058 | fd = PTR_ERR(f); | |
1059 | } else { | |
0f7fc9e4 | 1060 | fsnotify_open(f->f_path.dentry); |
fed2fc18 TN |
1061 | fd_install(fd, f); |
1062 | } | |
1da177e4 | 1063 | } |
1da177e4 LT |
1064 | putname(tmp); |
1065 | } | |
1066 | return fd; | |
1da177e4 | 1067 | } |
e922efc3 | 1068 | |
ca013e94 | 1069 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode) |
e922efc3 | 1070 | { |
385910f2 LT |
1071 | long ret; |
1072 | ||
e922efc3 MS |
1073 | if (force_o_largefile()) |
1074 | flags |= O_LARGEFILE; | |
1075 | ||
385910f2 LT |
1076 | ret = do_sys_open(AT_FDCWD, filename, flags, mode); |
1077 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1078 | asmlinkage_protect(3, ret, filename, flags, mode); |
385910f2 | 1079 | return ret; |
e922efc3 | 1080 | } |
1da177e4 | 1081 | |
6559eed8 HC |
1082 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, |
1083 | int, mode) | |
5590ff0d | 1084 | { |
385910f2 LT |
1085 | long ret; |
1086 | ||
5590ff0d UD |
1087 | if (force_o_largefile()) |
1088 | flags |= O_LARGEFILE; | |
1089 | ||
385910f2 LT |
1090 | ret = do_sys_open(dfd, filename, flags, mode); |
1091 | /* avoid REGPARM breakage on x86: */ | |
54a01510 | 1092 | asmlinkage_protect(4, ret, dfd, filename, flags, mode); |
385910f2 | 1093 | return ret; |
5590ff0d | 1094 | } |
5590ff0d | 1095 | |
1da177e4 LT |
1096 | #ifndef __alpha__ |
1097 | ||
1098 | /* | |
1099 | * For backward compatibility? Maybe this should be moved | |
1100 | * into arch/i386 instead? | |
1101 | */ | |
002c8976 | 1102 | SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode) |
1da177e4 LT |
1103 | { |
1104 | return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode); | |
1105 | } | |
1106 | ||
1107 | #endif | |
1108 | ||
1109 | /* | |
1110 | * "id" is the POSIX thread ID. We use the | |
1111 | * files pointer for this.. | |
1112 | */ | |
1113 | int filp_close(struct file *filp, fl_owner_t id) | |
1114 | { | |
45778ca8 | 1115 | int retval = 0; |
1da177e4 LT |
1116 | |
1117 | if (!file_count(filp)) { | |
1118 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
45778ca8 | 1119 | return 0; |
1da177e4 LT |
1120 | } |
1121 | ||
45778ca8 | 1122 | if (filp->f_op && filp->f_op->flush) |
75e1fcc0 | 1123 | retval = filp->f_op->flush(filp, id); |
1da177e4 LT |
1124 | |
1125 | dnotify_flush(filp, id); | |
1126 | locks_remove_posix(filp, id); | |
1127 | fput(filp); | |
1128 | return retval; | |
1129 | } | |
1130 | ||
1131 | EXPORT_SYMBOL(filp_close); | |
1132 | ||
1133 | /* | |
1134 | * Careful here! We test whether the file pointer is NULL before | |
1135 | * releasing the fd. This ensures that one clone task can't release | |
1136 | * an fd while another clone is opening it. | |
1137 | */ | |
ca013e94 | 1138 | SYSCALL_DEFINE1(close, unsigned int, fd) |
1da177e4 LT |
1139 | { |
1140 | struct file * filp; | |
1141 | struct files_struct *files = current->files; | |
badf1662 | 1142 | struct fdtable *fdt; |
ee731f4f | 1143 | int retval; |
1da177e4 LT |
1144 | |
1145 | spin_lock(&files->file_lock); | |
badf1662 DS |
1146 | fdt = files_fdtable(files); |
1147 | if (fd >= fdt->max_fds) | |
1da177e4 | 1148 | goto out_unlock; |
badf1662 | 1149 | filp = fdt->fd[fd]; |
1da177e4 LT |
1150 | if (!filp) |
1151 | goto out_unlock; | |
ab2af1f5 | 1152 | rcu_assign_pointer(fdt->fd[fd], NULL); |
badf1662 | 1153 | FD_CLR(fd, fdt->close_on_exec); |
1da177e4 LT |
1154 | __put_unused_fd(files, fd); |
1155 | spin_unlock(&files->file_lock); | |
ee731f4f EP |
1156 | retval = filp_close(filp, files); |
1157 | ||
1158 | /* can't restart close syscall because file table entry was cleared */ | |
1159 | if (unlikely(retval == -ERESTARTSYS || | |
1160 | retval == -ERESTARTNOINTR || | |
1161 | retval == -ERESTARTNOHAND || | |
1162 | retval == -ERESTART_RESTARTBLOCK)) | |
1163 | retval = -EINTR; | |
1164 | ||
1165 | return retval; | |
1da177e4 LT |
1166 | |
1167 | out_unlock: | |
1168 | spin_unlock(&files->file_lock); | |
1169 | return -EBADF; | |
1170 | } | |
1da177e4 LT |
1171 | EXPORT_SYMBOL(sys_close); |
1172 | ||
1173 | /* | |
1174 | * This routine simulates a hangup on the tty, to arrange that users | |
1175 | * are given clean terminals at login time. | |
1176 | */ | |
ca013e94 | 1177 | SYSCALL_DEFINE0(vhangup) |
1da177e4 LT |
1178 | { |
1179 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
2cb5998b | 1180 | tty_vhangup_self(); |
1da177e4 LT |
1181 | return 0; |
1182 | } | |
1183 | return -EPERM; | |
1184 | } | |
1185 | ||
1186 | /* | |
1187 | * Called when an inode is about to be open. | |
1188 | * We use this to disallow opening large files on 32bit systems if | |
1189 | * the caller didn't specify O_LARGEFILE. On 64bit systems we force | |
1190 | * on this flag in sys_open. | |
1191 | */ | |
1192 | int generic_file_open(struct inode * inode, struct file * filp) | |
1193 | { | |
1194 | if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS) | |
a9c62a18 | 1195 | return -EOVERFLOW; |
1da177e4 LT |
1196 | return 0; |
1197 | } | |
1198 | ||
1199 | EXPORT_SYMBOL(generic_file_open); | |
1200 | ||
1201 | /* | |
1202 | * This is used by subsystems that don't want seekable | |
1203 | * file descriptors | |
1204 | */ | |
1205 | int nonseekable_open(struct inode *inode, struct file *filp) | |
1206 | { | |
1207 | filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE); | |
1208 | return 0; | |
1209 | } | |
1210 | ||
1211 | EXPORT_SYMBOL(nonseekable_open); |