]>
Commit | Line | Data |
---|---|---|
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> | |
9 | #include <linux/file.h> | |
10 | #include <linux/fdtable.h> | |
11 | #include <linux/quotaops.h> | |
12 | #include <linux/fsnotify.h> | |
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> | |
18 | #include <linux/capability.h> | |
19 | #include <linux/securebits.h> | |
20 | #include <linux/security.h> | |
21 | #include <linux/mount.h> | |
22 | #include <linux/vfs.h> | |
23 | #include <linux/fcntl.h> | |
24 | #include <asm/uaccess.h> | |
25 | #include <linux/fs.h> | |
26 | #include <linux/personality.h> | |
27 | #include <linux/pagemap.h> | |
28 | #include <linux/syscalls.h> | |
29 | #include <linux/rcupdate.h> | |
30 | #include <linux/audit.h> | |
31 | #include <linux/falloc.h> | |
32 | #include <linux/fs_struct.h> | |
33 | ||
34 | int vfs_statfs(struct dentry *dentry, struct kstatfs *buf) | |
35 | { | |
36 | int retval = -ENODEV; | |
37 | ||
38 | if (dentry) { | |
39 | retval = -ENOSYS; | |
40 | if (dentry->d_sb->s_op->statfs) { | |
41 | memset(buf, 0, sizeof(*buf)); | |
42 | retval = security_sb_statfs(dentry); | |
43 | if (retval) | |
44 | return retval; | |
45 | retval = dentry->d_sb->s_op->statfs(dentry, buf); | |
46 | if (retval == 0 && buf->f_frsize == 0) | |
47 | buf->f_frsize = buf->f_bsize; | |
48 | } | |
49 | } | |
50 | return retval; | |
51 | } | |
52 | ||
53 | EXPORT_SYMBOL(vfs_statfs); | |
54 | ||
55 | static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf) | |
56 | { | |
57 | struct kstatfs st; | |
58 | int retval; | |
59 | ||
60 | retval = vfs_statfs(dentry, &st); | |
61 | if (retval) | |
62 | return retval; | |
63 | ||
64 | if (sizeof(*buf) == sizeof(st)) | |
65 | memcpy(buf, &st, sizeof(st)); | |
66 | else { | |
67 | if (sizeof buf->f_blocks == 4) { | |
68 | if ((st.f_blocks | st.f_bfree | st.f_bavail | | |
69 | st.f_bsize | st.f_frsize) & | |
70 | 0xffffffff00000000ULL) | |
71 | return -EOVERFLOW; | |
72 | /* | |
73 | * f_files and f_ffree may be -1; it's okay to stuff | |
74 | * that into 32 bits | |
75 | */ | |
76 | if (st.f_files != -1 && | |
77 | (st.f_files & 0xffffffff00000000ULL)) | |
78 | return -EOVERFLOW; | |
79 | if (st.f_ffree != -1 && | |
80 | (st.f_ffree & 0xffffffff00000000ULL)) | |
81 | return -EOVERFLOW; | |
82 | } | |
83 | ||
84 | buf->f_type = st.f_type; | |
85 | buf->f_bsize = st.f_bsize; | |
86 | buf->f_blocks = st.f_blocks; | |
87 | buf->f_bfree = st.f_bfree; | |
88 | buf->f_bavail = st.f_bavail; | |
89 | buf->f_files = st.f_files; | |
90 | buf->f_ffree = st.f_ffree; | |
91 | buf->f_fsid = st.f_fsid; | |
92 | buf->f_namelen = st.f_namelen; | |
93 | buf->f_frsize = st.f_frsize; | |
94 | memset(buf->f_spare, 0, sizeof(buf->f_spare)); | |
95 | } | |
96 | return 0; | |
97 | } | |
98 | ||
99 | static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf) | |
100 | { | |
101 | struct kstatfs st; | |
102 | int retval; | |
103 | ||
104 | retval = vfs_statfs(dentry, &st); | |
105 | if (retval) | |
106 | return retval; | |
107 | ||
108 | if (sizeof(*buf) == sizeof(st)) | |
109 | memcpy(buf, &st, sizeof(st)); | |
110 | else { | |
111 | buf->f_type = st.f_type; | |
112 | buf->f_bsize = st.f_bsize; | |
113 | buf->f_blocks = st.f_blocks; | |
114 | buf->f_bfree = st.f_bfree; | |
115 | buf->f_bavail = st.f_bavail; | |
116 | buf->f_files = st.f_files; | |
117 | buf->f_ffree = st.f_ffree; | |
118 | buf->f_fsid = st.f_fsid; | |
119 | buf->f_namelen = st.f_namelen; | |
120 | buf->f_frsize = st.f_frsize; | |
121 | memset(buf->f_spare, 0, sizeof(buf->f_spare)); | |
122 | } | |
123 | return 0; | |
124 | } | |
125 | ||
126 | SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct statfs __user *, buf) | |
127 | { | |
128 | struct path path; | |
129 | int error; | |
130 | ||
131 | error = user_path(pathname, &path); | |
132 | if (!error) { | |
133 | struct statfs tmp; | |
134 | error = vfs_statfs_native(path.dentry, &tmp); | |
135 | if (!error && copy_to_user(buf, &tmp, sizeof(tmp))) | |
136 | error = -EFAULT; | |
137 | path_put(&path); | |
138 | } | |
139 | return error; | |
140 | } | |
141 | ||
142 | SYSCALL_DEFINE3(statfs64, const char __user *, pathname, size_t, sz, struct statfs64 __user *, buf) | |
143 | { | |
144 | struct path path; | |
145 | long error; | |
146 | ||
147 | if (sz != sizeof(*buf)) | |
148 | return -EINVAL; | |
149 | error = user_path(pathname, &path); | |
150 | if (!error) { | |
151 | struct statfs64 tmp; | |
152 | error = vfs_statfs64(path.dentry, &tmp); | |
153 | if (!error && copy_to_user(buf, &tmp, sizeof(tmp))) | |
154 | error = -EFAULT; | |
155 | path_put(&path); | |
156 | } | |
157 | return error; | |
158 | } | |
159 | ||
160 | SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct statfs __user *, buf) | |
161 | { | |
162 | struct file * file; | |
163 | struct statfs tmp; | |
164 | int error; | |
165 | ||
166 | error = -EBADF; | |
167 | file = fget(fd); | |
168 | if (!file) | |
169 | goto out; | |
170 | error = vfs_statfs_native(file->f_path.dentry, &tmp); | |
171 | if (!error && copy_to_user(buf, &tmp, sizeof(tmp))) | |
172 | error = -EFAULT; | |
173 | fput(file); | |
174 | out: | |
175 | return error; | |
176 | } | |
177 | ||
178 | SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, size_t, sz, struct statfs64 __user *, buf) | |
179 | { | |
180 | struct file * file; | |
181 | struct statfs64 tmp; | |
182 | int error; | |
183 | ||
184 | if (sz != sizeof(*buf)) | |
185 | return -EINVAL; | |
186 | ||
187 | error = -EBADF; | |
188 | file = fget(fd); | |
189 | if (!file) | |
190 | goto out; | |
191 | error = vfs_statfs64(file->f_path.dentry, &tmp); | |
192 | if (!error && copy_to_user(buf, &tmp, sizeof(tmp))) | |
193 | error = -EFAULT; | |
194 | fput(file); | |
195 | out: | |
196 | return error; | |
197 | } | |
198 | ||
199 | int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs, | |
200 | struct file *filp) | |
201 | { | |
202 | int ret; | |
203 | struct iattr newattrs; | |
204 | ||
205 | /* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */ | |
206 | if (length < 0) | |
207 | return -EINVAL; | |
208 | ||
209 | newattrs.ia_size = length; | |
210 | newattrs.ia_valid = ATTR_SIZE | time_attrs; | |
211 | if (filp) { | |
212 | newattrs.ia_file = filp; | |
213 | newattrs.ia_valid |= ATTR_FILE; | |
214 | } | |
215 | ||
216 | /* Remove suid/sgid on truncate too */ | |
217 | ret = should_remove_suid(dentry); | |
218 | if (ret) | |
219 | newattrs.ia_valid |= ret | ATTR_FORCE; | |
220 | ||
221 | mutex_lock(&dentry->d_inode->i_mutex); | |
222 | ret = notify_change(dentry, &newattrs); | |
223 | mutex_unlock(&dentry->d_inode->i_mutex); | |
224 | return ret; | |
225 | } | |
226 | ||
227 | static long do_sys_truncate(const char __user *pathname, loff_t length) | |
228 | { | |
229 | struct path path; | |
230 | struct inode *inode; | |
231 | int error; | |
232 | ||
233 | error = -EINVAL; | |
234 | if (length < 0) /* sorry, but loff_t says... */ | |
235 | goto out; | |
236 | ||
237 | error = user_path(pathname, &path); | |
238 | if (error) | |
239 | goto out; | |
240 | inode = path.dentry->d_inode; | |
241 | ||
242 | /* For directories it's -EISDIR, for other non-regulars - -EINVAL */ | |
243 | error = -EISDIR; | |
244 | if (S_ISDIR(inode->i_mode)) | |
245 | goto dput_and_out; | |
246 | ||
247 | error = -EINVAL; | |
248 | if (!S_ISREG(inode->i_mode)) | |
249 | goto dput_and_out; | |
250 | ||
251 | error = mnt_want_write(path.mnt); | |
252 | if (error) | |
253 | goto dput_and_out; | |
254 | ||
255 | error = inode_permission(inode, MAY_WRITE); | |
256 | if (error) | |
257 | goto mnt_drop_write_and_out; | |
258 | ||
259 | error = -EPERM; | |
260 | if (IS_APPEND(inode)) | |
261 | goto mnt_drop_write_and_out; | |
262 | ||
263 | error = get_write_access(inode); | |
264 | if (error) | |
265 | goto mnt_drop_write_and_out; | |
266 | ||
267 | /* | |
268 | * Make sure that there are no leases. get_write_access() protects | |
269 | * against the truncate racing with a lease-granting setlease(). | |
270 | */ | |
271 | error = break_lease(inode, FMODE_WRITE); | |
272 | if (error) | |
273 | goto put_write_and_out; | |
274 | ||
275 | error = locks_verify_truncate(inode, NULL, length); | |
276 | if (!error) | |
277 | error = security_path_truncate(&path, length, 0); | |
278 | if (!error) { | |
279 | vfs_dq_init(inode); | |
280 | error = do_truncate(path.dentry, length, 0, NULL); | |
281 | } | |
282 | ||
283 | put_write_and_out: | |
284 | put_write_access(inode); | |
285 | mnt_drop_write_and_out: | |
286 | mnt_drop_write(path.mnt); | |
287 | dput_and_out: | |
288 | path_put(&path); | |
289 | out: | |
290 | return error; | |
291 | } | |
292 | ||
293 | SYSCALL_DEFINE2(truncate, const char __user *, path, long, length) | |
294 | { | |
295 | return do_sys_truncate(path, length); | |
296 | } | |
297 | ||
298 | static long do_sys_ftruncate(unsigned int fd, loff_t length, int small) | |
299 | { | |
300 | struct inode * inode; | |
301 | struct dentry *dentry; | |
302 | struct file * file; | |
303 | int error; | |
304 | ||
305 | error = -EINVAL; | |
306 | if (length < 0) | |
307 | goto out; | |
308 | error = -EBADF; | |
309 | file = fget(fd); | |
310 | if (!file) | |
311 | goto out; | |
312 | ||
313 | /* explicitly opened as large or we are on 64-bit box */ | |
314 | if (file->f_flags & O_LARGEFILE) | |
315 | small = 0; | |
316 | ||
317 | dentry = file->f_path.dentry; | |
318 | inode = dentry->d_inode; | |
319 | error = -EINVAL; | |
320 | if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE)) | |
321 | goto out_putf; | |
322 | ||
323 | error = -EINVAL; | |
324 | /* Cannot ftruncate over 2^31 bytes without large file support */ | |
325 | if (small && length > MAX_NON_LFS) | |
326 | goto out_putf; | |
327 | ||
328 | error = -EPERM; | |
329 | if (IS_APPEND(inode)) | |
330 | goto out_putf; | |
331 | ||
332 | error = locks_verify_truncate(inode, file, length); | |
333 | if (!error) | |
334 | error = security_path_truncate(&file->f_path, length, | |
335 | ATTR_MTIME|ATTR_CTIME); | |
336 | if (!error) | |
337 | error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file); | |
338 | out_putf: | |
339 | fput(file); | |
340 | out: | |
341 | return error; | |
342 | } | |
343 | ||
344 | SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length) | |
345 | { | |
346 | long ret = do_sys_ftruncate(fd, length, 1); | |
347 | /* avoid REGPARM breakage on x86: */ | |
348 | asmlinkage_protect(2, ret, fd, length); | |
349 | return ret; | |
350 | } | |
351 | ||
352 | /* LFS versions of truncate are only needed on 32 bit machines */ | |
353 | #if BITS_PER_LONG == 32 | |
354 | SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length) | |
355 | { | |
356 | return do_sys_truncate(path, length); | |
357 | } | |
358 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS | |
359 | asmlinkage long SyS_truncate64(long path, loff_t length) | |
360 | { | |
361 | return SYSC_truncate64((const char __user *) path, length); | |
362 | } | |
363 | SYSCALL_ALIAS(sys_truncate64, SyS_truncate64); | |
364 | #endif | |
365 | ||
366 | SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length) | |
367 | { | |
368 | long ret = do_sys_ftruncate(fd, length, 0); | |
369 | /* avoid REGPARM breakage on x86: */ | |
370 | asmlinkage_protect(2, ret, fd, length); | |
371 | return ret; | |
372 | } | |
373 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS | |
374 | asmlinkage long SyS_ftruncate64(long fd, loff_t length) | |
375 | { | |
376 | return SYSC_ftruncate64((unsigned int) fd, length); | |
377 | } | |
378 | SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64); | |
379 | #endif | |
380 | #endif /* BITS_PER_LONG == 32 */ | |
381 | ||
382 | ||
383 | int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len) | |
384 | { | |
385 | struct inode *inode = file->f_path.dentry->d_inode; | |
386 | long ret; | |
387 | ||
388 | if (offset < 0 || len <= 0) | |
389 | return -EINVAL; | |
390 | ||
391 | /* Return error if mode is not supported */ | |
392 | if (mode && !(mode & FALLOC_FL_KEEP_SIZE)) | |
393 | return -EOPNOTSUPP; | |
394 | ||
395 | if (!(file->f_mode & FMODE_WRITE)) | |
396 | return -EBADF; | |
397 | /* | |
398 | * Revalidate the write permissions, in case security policy has | |
399 | * changed since the files were opened. | |
400 | */ | |
401 | ret = security_file_permission(file, MAY_WRITE); | |
402 | if (ret) | |
403 | return ret; | |
404 | ||
405 | if (S_ISFIFO(inode->i_mode)) | |
406 | return -ESPIPE; | |
407 | ||
408 | /* | |
409 | * Let individual file system decide if it supports preallocation | |
410 | * for directories or not. | |
411 | */ | |
412 | if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode)) | |
413 | return -ENODEV; | |
414 | ||
415 | /* Check for wrap through zero too */ | |
416 | if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0)) | |
417 | return -EFBIG; | |
418 | ||
419 | if (!inode->i_op->fallocate) | |
420 | return -EOPNOTSUPP; | |
421 | ||
422 | return inode->i_op->fallocate(inode, mode, offset, len); | |
423 | } | |
424 | ||
425 | SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len) | |
426 | { | |
427 | struct file *file; | |
428 | int error = -EBADF; | |
429 | ||
430 | file = fget(fd); | |
431 | if (file) { | |
432 | error = do_fallocate(file, mode, offset, len); | |
433 | fput(file); | |
434 | } | |
435 | ||
436 | return error; | |
437 | } | |
438 | ||
439 | #ifdef CONFIG_HAVE_SYSCALL_WRAPPERS | |
440 | asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len) | |
441 | { | |
442 | return SYSC_fallocate((int)fd, (int)mode, offset, len); | |
443 | } | |
444 | SYSCALL_ALIAS(sys_fallocate, SyS_fallocate); | |
445 | #endif | |
446 | ||
447 | /* | |
448 | * access() needs to use the real uid/gid, not the effective uid/gid. | |
449 | * We do this by temporarily clearing all FS-related capabilities and | |
450 | * switching the fsuid/fsgid around to the real ones. | |
451 | */ | |
452 | SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode) | |
453 | { | |
454 | const struct cred *old_cred; | |
455 | struct cred *override_cred; | |
456 | struct path path; | |
457 | struct inode *inode; | |
458 | int res; | |
459 | ||
460 | if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ | |
461 | return -EINVAL; | |
462 | ||
463 | override_cred = prepare_creds(); | |
464 | if (!override_cred) | |
465 | return -ENOMEM; | |
466 | ||
467 | override_cred->fsuid = override_cred->uid; | |
468 | override_cred->fsgid = override_cred->gid; | |
469 | ||
470 | if (!issecure(SECURE_NO_SETUID_FIXUP)) { | |
471 | /* Clear the capabilities if we switch to a non-root user */ | |
472 | if (override_cred->uid) | |
473 | cap_clear(override_cred->cap_effective); | |
474 | else | |
475 | override_cred->cap_effective = | |
476 | override_cred->cap_permitted; | |
477 | } | |
478 | ||
479 | old_cred = override_creds(override_cred); | |
480 | ||
481 | res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path); | |
482 | if (res) | |
483 | goto out; | |
484 | ||
485 | inode = path.dentry->d_inode; | |
486 | ||
487 | if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) { | |
488 | /* | |
489 | * MAY_EXEC on regular files is denied if the fs is mounted | |
490 | * with the "noexec" flag. | |
491 | */ | |
492 | res = -EACCES; | |
493 | if (path.mnt->mnt_flags & MNT_NOEXEC) | |
494 | goto out_path_release; | |
495 | } | |
496 | ||
497 | res = inode_permission(inode, mode | MAY_ACCESS); | |
498 | /* SuS v2 requires we report a read only fs too */ | |
499 | if (res || !(mode & S_IWOTH) || special_file(inode->i_mode)) | |
500 | goto out_path_release; | |
501 | /* | |
502 | * This is a rare case where using __mnt_is_readonly() | |
503 | * is OK without a mnt_want/drop_write() pair. Since | |
504 | * no actual write to the fs is performed here, we do | |
505 | * not need to telegraph to that to anyone. | |
506 | * | |
507 | * By doing this, we accept that this access is | |
508 | * inherently racy and know that the fs may change | |
509 | * state before we even see this result. | |
510 | */ | |
511 | if (__mnt_is_readonly(path.mnt)) | |
512 | res = -EROFS; | |
513 | ||
514 | out_path_release: | |
515 | path_put(&path); | |
516 | out: | |
517 | revert_creds(old_cred); | |
518 | put_cred(override_cred); | |
519 | return res; | |
520 | } | |
521 | ||
522 | SYSCALL_DEFINE2(access, const char __user *, filename, int, mode) | |
523 | { | |
524 | return sys_faccessat(AT_FDCWD, filename, mode); | |
525 | } | |
526 | ||
527 | SYSCALL_DEFINE1(chdir, const char __user *, filename) | |
528 | { | |
529 | struct path path; | |
530 | int error; | |
531 | ||
532 | error = user_path_dir(filename, &path); | |
533 | if (error) | |
534 | goto out; | |
535 | ||
536 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS); | |
537 | if (error) | |
538 | goto dput_and_out; | |
539 | ||
540 | set_fs_pwd(current->fs, &path); | |
541 | ||
542 | dput_and_out: | |
543 | path_put(&path); | |
544 | out: | |
545 | return error; | |
546 | } | |
547 | ||
548 | SYSCALL_DEFINE1(fchdir, unsigned int, fd) | |
549 | { | |
550 | struct file *file; | |
551 | struct inode *inode; | |
552 | int error; | |
553 | ||
554 | error = -EBADF; | |
555 | file = fget(fd); | |
556 | if (!file) | |
557 | goto out; | |
558 | ||
559 | inode = file->f_path.dentry->d_inode; | |
560 | ||
561 | error = -ENOTDIR; | |
562 | if (!S_ISDIR(inode->i_mode)) | |
563 | goto out_putf; | |
564 | ||
565 | error = inode_permission(inode, MAY_EXEC | MAY_ACCESS); | |
566 | if (!error) | |
567 | set_fs_pwd(current->fs, &file->f_path); | |
568 | out_putf: | |
569 | fput(file); | |
570 | out: | |
571 | return error; | |
572 | } | |
573 | ||
574 | SYSCALL_DEFINE1(chroot, const char __user *, filename) | |
575 | { | |
576 | struct path path; | |
577 | int error; | |
578 | ||
579 | error = user_path_dir(filename, &path); | |
580 | if (error) | |
581 | goto out; | |
582 | ||
583 | error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS); | |
584 | if (error) | |
585 | goto dput_and_out; | |
586 | ||
587 | error = -EPERM; | |
588 | if (!capable(CAP_SYS_CHROOT)) | |
589 | goto dput_and_out; | |
590 | ||
591 | set_fs_root(current->fs, &path); | |
592 | error = 0; | |
593 | dput_and_out: | |
594 | path_put(&path); | |
595 | out: | |
596 | return error; | |
597 | } | |
598 | ||
599 | SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode) | |
600 | { | |
601 | struct inode * inode; | |
602 | struct dentry * dentry; | |
603 | struct file * file; | |
604 | int err = -EBADF; | |
605 | struct iattr newattrs; | |
606 | ||
607 | file = fget(fd); | |
608 | if (!file) | |
609 | goto out; | |
610 | ||
611 | dentry = file->f_path.dentry; | |
612 | inode = dentry->d_inode; | |
613 | ||
614 | audit_inode(NULL, dentry); | |
615 | ||
616 | err = mnt_want_write_file(file); | |
617 | if (err) | |
618 | goto out_putf; | |
619 | err = security_path_chmod(dentry, file->f_vfsmnt, mode); | |
620 | if (err) | |
621 | goto out_drop_write; | |
622 | mutex_lock(&inode->i_mutex); | |
623 | if (mode == (mode_t) -1) | |
624 | mode = inode->i_mode; | |
625 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); | |
626 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
627 | err = notify_change(dentry, &newattrs); | |
628 | mutex_unlock(&inode->i_mutex); | |
629 | out_drop_write: | |
630 | mnt_drop_write(file->f_path.mnt); | |
631 | out_putf: | |
632 | fput(file); | |
633 | out: | |
634 | return err; | |
635 | } | |
636 | ||
637 | SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode) | |
638 | { | |
639 | struct path path; | |
640 | struct inode *inode; | |
641 | int error; | |
642 | struct iattr newattrs; | |
643 | ||
644 | error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path); | |
645 | if (error) | |
646 | goto out; | |
647 | inode = path.dentry->d_inode; | |
648 | ||
649 | error = mnt_want_write(path.mnt); | |
650 | if (error) | |
651 | goto dput_and_out; | |
652 | error = security_path_chmod(path.dentry, path.mnt, mode); | |
653 | if (error) | |
654 | goto out_drop_write; | |
655 | mutex_lock(&inode->i_mutex); | |
656 | if (mode == (mode_t) -1) | |
657 | mode = inode->i_mode; | |
658 | newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); | |
659 | newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; | |
660 | error = notify_change(path.dentry, &newattrs); | |
661 | mutex_unlock(&inode->i_mutex); | |
662 | out_drop_write: | |
663 | mnt_drop_write(path.mnt); | |
664 | dput_and_out: | |
665 | path_put(&path); | |
666 | out: | |
667 | return error; | |
668 | } | |
669 | ||
670 | SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode) | |
671 | { | |
672 | return sys_fchmodat(AT_FDCWD, filename, mode); | |
673 | } | |
674 | ||
675 | static int chown_common(struct dentry * dentry, uid_t user, gid_t group) | |
676 | { | |
677 | struct inode *inode = dentry->d_inode; | |
678 | int error; | |
679 | struct iattr newattrs; | |
680 | ||
681 | newattrs.ia_valid = ATTR_CTIME; | |
682 | if (user != (uid_t) -1) { | |
683 | newattrs.ia_valid |= ATTR_UID; | |
684 | newattrs.ia_uid = user; | |
685 | } | |
686 | if (group != (gid_t) -1) { | |
687 | newattrs.ia_valid |= ATTR_GID; | |
688 | newattrs.ia_gid = group; | |
689 | } | |
690 | if (!S_ISDIR(inode->i_mode)) | |
691 | newattrs.ia_valid |= | |
692 | ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV; | |
693 | mutex_lock(&inode->i_mutex); | |
694 | error = notify_change(dentry, &newattrs); | |
695 | mutex_unlock(&inode->i_mutex); | |
696 | ||
697 | return error; | |
698 | } | |
699 | ||
700 | SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group) | |
701 | { | |
702 | struct path path; | |
703 | int error; | |
704 | ||
705 | error = user_path(filename, &path); | |
706 | if (error) | |
707 | goto out; | |
708 | error = mnt_want_write(path.mnt); | |
709 | if (error) | |
710 | goto out_release; | |
711 | error = security_path_chown(&path, user, group); | |
712 | if (!error) | |
713 | error = chown_common(path.dentry, user, group); | |
714 | mnt_drop_write(path.mnt); | |
715 | out_release: | |
716 | path_put(&path); | |
717 | out: | |
718 | return error; | |
719 | } | |
720 | ||
721 | SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user, | |
722 | gid_t, group, int, flag) | |
723 | { | |
724 | struct path path; | |
725 | int error = -EINVAL; | |
726 | int follow; | |
727 | ||
728 | if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0) | |
729 | goto out; | |
730 | ||
731 | follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW; | |
732 | error = user_path_at(dfd, filename, follow, &path); | |
733 | if (error) | |
734 | goto out; | |
735 | error = mnt_want_write(path.mnt); | |
736 | if (error) | |
737 | goto out_release; | |
738 | error = security_path_chown(&path, user, group); | |
739 | if (!error) | |
740 | error = chown_common(path.dentry, user, group); | |
741 | mnt_drop_write(path.mnt); | |
742 | out_release: | |
743 | path_put(&path); | |
744 | out: | |
745 | return error; | |
746 | } | |
747 | ||
748 | SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group) | |
749 | { | |
750 | struct path path; | |
751 | int error; | |
752 | ||
753 | error = user_lpath(filename, &path); | |
754 | if (error) | |
755 | goto out; | |
756 | error = mnt_want_write(path.mnt); | |
757 | if (error) | |
758 | goto out_release; | |
759 | error = security_path_chown(&path, user, group); | |
760 | if (!error) | |
761 | error = chown_common(path.dentry, user, group); | |
762 | mnt_drop_write(path.mnt); | |
763 | out_release: | |
764 | path_put(&path); | |
765 | out: | |
766 | return error; | |
767 | } | |
768 | ||
769 | SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group) | |
770 | { | |
771 | struct file * file; | |
772 | int error = -EBADF; | |
773 | struct dentry * dentry; | |
774 | ||
775 | file = fget(fd); | |
776 | if (!file) | |
777 | goto out; | |
778 | ||
779 | error = mnt_want_write_file(file); | |
780 | if (error) | |
781 | goto out_fput; | |
782 | dentry = file->f_path.dentry; | |
783 | audit_inode(NULL, dentry); | |
784 | error = security_path_chown(&file->f_path, user, group); | |
785 | if (!error) | |
786 | error = chown_common(dentry, user, group); | |
787 | mnt_drop_write(file->f_path.mnt); | |
788 | out_fput: | |
789 | fput(file); | |
790 | out: | |
791 | return error; | |
792 | } | |
793 | ||
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 | ||
823 | static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt, | |
824 | int flags, struct file *f, | |
825 | int (*open)(struct inode *, struct file *), | |
826 | const struct cred *cred) | |
827 | { | |
828 | struct inode *inode; | |
829 | int error; | |
830 | ||
831 | f->f_flags = flags; | |
832 | f->f_mode = (__force fmode_t)((flags+1) & O_ACCMODE) | FMODE_LSEEK | | |
833 | FMODE_PREAD | FMODE_PWRITE; | |
834 | inode = dentry->d_inode; | |
835 | if (f->f_mode & FMODE_WRITE) { | |
836 | error = __get_file_write_access(inode, mnt); | |
837 | if (error) | |
838 | goto cleanup_file; | |
839 | if (!special_file(inode->i_mode)) | |
840 | file_take_write(f); | |
841 | } | |
842 | ||
843 | f->f_mapping = inode->i_mapping; | |
844 | f->f_path.dentry = dentry; | |
845 | f->f_path.mnt = mnt; | |
846 | f->f_pos = 0; | |
847 | f->f_op = fops_get(inode->i_fop); | |
848 | file_move(f, &inode->i_sb->s_files); | |
849 | ||
850 | error = security_dentry_open(f, cred); | |
851 | if (error) | |
852 | goto cleanup_all; | |
853 | ||
854 | if (!open && f->f_op) | |
855 | open = f->f_op->open; | |
856 | if (open) { | |
857 | error = open(inode, f); | |
858 | if (error) | |
859 | goto cleanup_all; | |
860 | } | |
861 | ||
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) { | |
868 | if (!f->f_mapping->a_ops || | |
869 | ((!f->f_mapping->a_ops->direct_IO) && | |
870 | (!f->f_mapping->a_ops->get_xip_mem))) { | |
871 | fput(f); | |
872 | f = ERR_PTR(-EINVAL); | |
873 | } | |
874 | } | |
875 | ||
876 | return f; | |
877 | ||
878 | cleanup_all: | |
879 | fops_put(f->f_op); | |
880 | if (f->f_mode & FMODE_WRITE) { | |
881 | put_write_access(inode); | |
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); | |
890 | mnt_drop_write(mnt); | |
891 | } | |
892 | } | |
893 | file_kill(f); | |
894 | f->f_path.dentry = NULL; | |
895 | f->f_path.mnt = NULL; | |
896 | cleanup_file: | |
897 | put_filp(f); | |
898 | dput(dentry); | |
899 | mntput(mnt); | |
900 | return ERR_PTR(error); | |
901 | } | |
902 | ||
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. | |
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). | |
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 | { | |
925 | const struct cred *cred = current_cred(); | |
926 | ||
927 | if (IS_ERR(nd->intent.open.file)) | |
928 | goto out; | |
929 | if (IS_ERR(dentry)) | |
930 | goto out_err; | |
931 | nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt), | |
932 | nd->intent.open.flags - 1, | |
933 | nd->intent.open.file, | |
934 | open, cred); | |
935 | out: | |
936 | return nd->intent.open.file; | |
937 | out_err: | |
938 | release_open_intent(nd); | |
939 | nd->intent.open.file = (struct file *)dentry; | |
940 | goto out; | |
941 | } | |
942 | EXPORT_SYMBOL_GPL(lookup_instantiate_filp); | |
943 | ||
944 | /** | |
945 | * nameidata_to_filp - convert a nameidata to an open filp. | |
946 | * @nd: pointer to nameidata | |
947 | * @flags: open flags | |
948 | * | |
949 | * Note that this function destroys the original nameidata | |
950 | */ | |
951 | struct file *nameidata_to_filp(struct nameidata *nd, int flags) | |
952 | { | |
953 | const struct cred *cred = current_cred(); | |
954 | struct file *filp; | |
955 | ||
956 | /* Pick up the filp from the open intent */ | |
957 | filp = nd->intent.open.file; | |
958 | /* Has the filesystem initialised the file for us? */ | |
959 | if (filp->f_path.dentry == NULL) | |
960 | filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp, | |
961 | NULL, cred); | |
962 | else | |
963 | path_put(&nd->path); | |
964 | return filp; | |
965 | } | |
966 | ||
967 | /* | |
968 | * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an | |
969 | * error. | |
970 | */ | |
971 | struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags, | |
972 | const struct cred *cred) | |
973 | { | |
974 | int error; | |
975 | struct file *f; | |
976 | ||
977 | validate_creds(cred); | |
978 | ||
979 | /* | |
980 | * We must always pass in a valid mount pointer. Historically | |
981 | * callers got away with not passing it, but we must enforce this at | |
982 | * the earliest possible point now to avoid strange problems deep in the | |
983 | * filesystem stack. | |
984 | */ | |
985 | if (!mnt) { | |
986 | printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__); | |
987 | dump_stack(); | |
988 | return ERR_PTR(-EINVAL); | |
989 | } | |
990 | ||
991 | error = -ENFILE; | |
992 | f = get_empty_filp(); | |
993 | if (f == NULL) { | |
994 | dput(dentry); | |
995 | mntput(mnt); | |
996 | return ERR_PTR(error); | |
997 | } | |
998 | ||
999 | return __dentry_open(dentry, mnt, flags, f, NULL, cred); | |
1000 | } | |
1001 | EXPORT_SYMBOL(dentry_open); | |
1002 | ||
1003 | static void __put_unused_fd(struct files_struct *files, unsigned int fd) | |
1004 | { | |
1005 | struct fdtable *fdt = files_fdtable(files); | |
1006 | __FD_CLR(fd, fdt->open_fds); | |
1007 | if (fd < files->next_fd) | |
1008 | files->next_fd = fd; | |
1009 | } | |
1010 | ||
1011 | void put_unused_fd(unsigned int fd) | |
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 | /* | |
1022 | * Install a file pointer in the fd array. | |
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 | ||
1034 | void fd_install(unsigned int fd, struct file *file) | |
1035 | { | |
1036 | struct files_struct *files = current->files; | |
1037 | struct fdtable *fdt; | |
1038 | spin_lock(&files->file_lock); | |
1039 | fdt = files_fdtable(files); | |
1040 | BUG_ON(fdt->fd[fd] != NULL); | |
1041 | rcu_assign_pointer(fdt->fd[fd], file); | |
1042 | spin_unlock(&files->file_lock); | |
1043 | } | |
1044 | ||
1045 | EXPORT_SYMBOL(fd_install); | |
1046 | ||
1047 | long do_sys_open(int dfd, const char __user *filename, int flags, int mode) | |
1048 | { | |
1049 | char *tmp = getname(filename); | |
1050 | int fd = PTR_ERR(tmp); | |
1051 | ||
1052 | if (!IS_ERR(tmp)) { | |
1053 | fd = get_unused_fd_flags(flags); | |
1054 | if (fd >= 0) { | |
1055 | struct file *f = do_filp_open(dfd, tmp, flags, mode, 0); | |
1056 | if (IS_ERR(f)) { | |
1057 | put_unused_fd(fd); | |
1058 | fd = PTR_ERR(f); | |
1059 | } else { | |
1060 | fsnotify_open(f->f_path.dentry); | |
1061 | fd_install(fd, f); | |
1062 | } | |
1063 | } | |
1064 | putname(tmp); | |
1065 | } | |
1066 | return fd; | |
1067 | } | |
1068 | ||
1069 | SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode) | |
1070 | { | |
1071 | long ret; | |
1072 | ||
1073 | if (force_o_largefile()) | |
1074 | flags |= O_LARGEFILE; | |
1075 | ||
1076 | ret = do_sys_open(AT_FDCWD, filename, flags, mode); | |
1077 | /* avoid REGPARM breakage on x86: */ | |
1078 | asmlinkage_protect(3, ret, filename, flags, mode); | |
1079 | return ret; | |
1080 | } | |
1081 | ||
1082 | SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags, | |
1083 | int, mode) | |
1084 | { | |
1085 | long ret; | |
1086 | ||
1087 | if (force_o_largefile()) | |
1088 | flags |= O_LARGEFILE; | |
1089 | ||
1090 | ret = do_sys_open(dfd, filename, flags, mode); | |
1091 | /* avoid REGPARM breakage on x86: */ | |
1092 | asmlinkage_protect(4, ret, dfd, filename, flags, mode); | |
1093 | return ret; | |
1094 | } | |
1095 | ||
1096 | #ifndef __alpha__ | |
1097 | ||
1098 | /* | |
1099 | * For backward compatibility? Maybe this should be moved | |
1100 | * into arch/i386 instead? | |
1101 | */ | |
1102 | SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode) | |
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 | { | |
1115 | int retval = 0; | |
1116 | ||
1117 | if (!file_count(filp)) { | |
1118 | printk(KERN_ERR "VFS: Close: file count is 0\n"); | |
1119 | return 0; | |
1120 | } | |
1121 | ||
1122 | if (filp->f_op && filp->f_op->flush) | |
1123 | retval = filp->f_op->flush(filp, id); | |
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 | */ | |
1138 | SYSCALL_DEFINE1(close, unsigned int, fd) | |
1139 | { | |
1140 | struct file * filp; | |
1141 | struct files_struct *files = current->files; | |
1142 | struct fdtable *fdt; | |
1143 | int retval; | |
1144 | ||
1145 | spin_lock(&files->file_lock); | |
1146 | fdt = files_fdtable(files); | |
1147 | if (fd >= fdt->max_fds) | |
1148 | goto out_unlock; | |
1149 | filp = fdt->fd[fd]; | |
1150 | if (!filp) | |
1151 | goto out_unlock; | |
1152 | rcu_assign_pointer(fdt->fd[fd], NULL); | |
1153 | FD_CLR(fd, fdt->close_on_exec); | |
1154 | __put_unused_fd(files, fd); | |
1155 | spin_unlock(&files->file_lock); | |
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; | |
1166 | ||
1167 | out_unlock: | |
1168 | spin_unlock(&files->file_lock); | |
1169 | return -EBADF; | |
1170 | } | |
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 | */ | |
1177 | SYSCALL_DEFINE0(vhangup) | |
1178 | { | |
1179 | if (capable(CAP_SYS_TTY_CONFIG)) { | |
1180 | tty_vhangup_self(); | |
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) | |
1195 | return -EOVERFLOW; | |
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); |