open.c 27.5 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
2
3
4
5
6
7
8
9
/*
 *  linux/fs/open.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 */

#include <linux/string.h>
#include <linux/mm.h>
#include <linux/file.h>
Al Viro's avatar
Al Viro committed
10
#include <linux/fdtable.h>
Linus Torvalds's avatar
Linus Torvalds committed
11
#include <linux/quotaops.h>
Robert Love's avatar
Robert Love committed
12
#include <linux/fsnotify.h>
Linus Torvalds's avatar
Linus Torvalds committed
13
14
15
16
17
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/tty.h>
#include <linux/namei.h>
#include <linux/backing-dev.h>
18
#include <linux/capability.h>
19
#include <linux/securebits.h>
Linus Torvalds's avatar
Linus Torvalds committed
20
21
22
#include <linux/security.h>
#include <linux/mount.h>
#include <linux/vfs.h>
23
#include <linux/fcntl.h>
Linus Torvalds's avatar
Linus Torvalds committed
24
25
#include <asm/uaccess.h>
#include <linux/fs.h>
26
#include <linux/personality.h>
Linus Torvalds's avatar
Linus Torvalds committed
27
28
#include <linux/pagemap.h>
#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>
Linus Torvalds's avatar
Linus Torvalds committed
33

34
int vfs_statfs(struct dentry *dentry, struct kstatfs *buf)
Linus Torvalds's avatar
Linus Torvalds committed
35
36
37
{
	int retval = -ENODEV;

38
	if (dentry) {
Linus Torvalds's avatar
Linus Torvalds committed
39
		retval = -ENOSYS;
40
		if (dentry->d_sb->s_op->statfs) {
Linus Torvalds's avatar
Linus Torvalds committed
41
			memset(buf, 0, sizeof(*buf));
42
			retval = security_sb_statfs(dentry);
Linus Torvalds's avatar
Linus Torvalds committed
43
44
			if (retval)
				return retval;
45
			retval = dentry->d_sb->s_op->statfs(dentry, buf);
Linus Torvalds's avatar
Linus Torvalds committed
46
47
48
49
50
51
52
53
54
			if (retval == 0 && buf->f_frsize == 0)
				buf->f_frsize = buf->f_bsize;
		}
	}
	return retval;
}

EXPORT_SYMBOL(vfs_statfs);

55
static int vfs_statfs_native(struct dentry *dentry, struct statfs *buf)
Linus Torvalds's avatar
Linus Torvalds committed
56
57
58
59
{
	struct kstatfs st;
	int retval;

60
	retval = vfs_statfs(dentry, &st);
Linus Torvalds's avatar
Linus Torvalds committed
61
62
63
64
65
66
67
	if (retval)
		return retval;

	if (sizeof(*buf) == sizeof(st))
		memcpy(buf, &st, sizeof(st));
	else {
		if (sizeof buf->f_blocks == 4) {
Jon Tollefson's avatar
Jon Tollefson committed
68
69
			if ((st.f_blocks | st.f_bfree | st.f_bavail |
			     st.f_bsize | st.f_frsize) &
Linus Torvalds's avatar
Linus Torvalds committed
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
			    0xffffffff00000000ULL)
				return -EOVERFLOW;
			/*
			 * f_files and f_ffree may be -1; it's okay to stuff
			 * that into 32 bits
			 */
			if (st.f_files != -1 &&
			    (st.f_files & 0xffffffff00000000ULL))
				return -EOVERFLOW;
			if (st.f_ffree != -1 &&
			    (st.f_ffree & 0xffffffff00000000ULL))
				return -EOVERFLOW;
		}

		buf->f_type = st.f_type;
		buf->f_bsize = st.f_bsize;
		buf->f_blocks = st.f_blocks;
		buf->f_bfree = st.f_bfree;
		buf->f_bavail = st.f_bavail;
		buf->f_files = st.f_files;
		buf->f_ffree = st.f_ffree;
		buf->f_fsid = st.f_fsid;
		buf->f_namelen = st.f_namelen;
		buf->f_frsize = st.f_frsize;
		memset(buf->f_spare, 0, sizeof(buf->f_spare));
	}
	return 0;
}

99
static int vfs_statfs64(struct dentry *dentry, struct statfs64 *buf)
Linus Torvalds's avatar
Linus Torvalds committed
100
101
102
103
{
	struct kstatfs st;
	int retval;

104
	retval = vfs_statfs(dentry, &st);
Linus Torvalds's avatar
Linus Torvalds committed
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
	if (retval)
		return retval;

	if (sizeof(*buf) == sizeof(st))
		memcpy(buf, &st, sizeof(st));
	else {
		buf->f_type = st.f_type;
		buf->f_bsize = st.f_bsize;
		buf->f_blocks = st.f_blocks;
		buf->f_bfree = st.f_bfree;
		buf->f_bavail = st.f_bavail;
		buf->f_files = st.f_files;
		buf->f_ffree = st.f_ffree;
		buf->f_fsid = st.f_fsid;
		buf->f_namelen = st.f_namelen;
		buf->f_frsize = st.f_frsize;
		memset(buf->f_spare, 0, sizeof(buf->f_spare));
	}
	return 0;
}

126
SYSCALL_DEFINE2(statfs, const char __user *, pathname, struct statfs __user *, buf)
Linus Torvalds's avatar
Linus Torvalds committed
127
{
128
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
129
130
	int error;

131
	error = user_path(pathname, &path);
Linus Torvalds's avatar
Linus Torvalds committed
132
133
	if (!error) {
		struct statfs tmp;
134
		error = vfs_statfs_native(path.dentry, &tmp);
Linus Torvalds's avatar
Linus Torvalds committed
135
136
		if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
			error = -EFAULT;
137
		path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
138
139
140
141
	}
	return error;
}

142
SYSCALL_DEFINE3(statfs64, const char __user *, pathname, size_t, sz, struct statfs64 __user *, buf)
Linus Torvalds's avatar
Linus Torvalds committed
143
{
144
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
145
146
147
148
	long error;

	if (sz != sizeof(*buf))
		return -EINVAL;
149
	error = user_path(pathname, &path);
Linus Torvalds's avatar
Linus Torvalds committed
150
151
	if (!error) {
		struct statfs64 tmp;
152
		error = vfs_statfs64(path.dentry, &tmp);
Linus Torvalds's avatar
Linus Torvalds committed
153
154
		if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
			error = -EFAULT;
155
		path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
156
157
158
159
	}
	return error;
}

160
SYSCALL_DEFINE2(fstatfs, unsigned int, fd, struct statfs __user *, buf)
Linus Torvalds's avatar
Linus Torvalds committed
161
162
163
164
165
166
167
168
169
{
	struct file * file;
	struct statfs tmp;
	int error;

	error = -EBADF;
	file = fget(fd);
	if (!file)
		goto out;
170
	error = vfs_statfs_native(file->f_path.dentry, &tmp);
Linus Torvalds's avatar
Linus Torvalds committed
171
172
173
174
175
176
177
	if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
		error = -EFAULT;
	fput(file);
out:
	return error;
}

178
SYSCALL_DEFINE3(fstatfs64, unsigned int, fd, size_t, sz, struct statfs64 __user *, buf)
Linus Torvalds's avatar
Linus Torvalds committed
179
180
181
182
183
184
185
186
187
188
189
190
{
	struct file * file;
	struct statfs64 tmp;
	int error;

	if (sz != sizeof(*buf))
		return -EINVAL;

	error = -EBADF;
	file = fget(fd);
	if (!file)
		goto out;
191
	error = vfs_statfs64(file->f_path.dentry, &tmp);
Linus Torvalds's avatar
Linus Torvalds committed
192
193
194
195
196
197
198
	if (!error && copy_to_user(buf, &tmp, sizeof(tmp)))
		error = -EFAULT;
	fput(file);
out:
	return error;
}

199
200
int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
	struct file *filp)
Linus Torvalds's avatar
Linus Torvalds committed
201
{
202
	int ret;
Linus Torvalds's avatar
Linus Torvalds committed
203
204
205
206
207
208
209
	struct iattr newattrs;

	/* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
	if (length < 0)
		return -EINVAL;

	newattrs.ia_size = length;
210
	newattrs.ia_valid = ATTR_SIZE | time_attrs;
211
212
213
214
	if (filp) {
		newattrs.ia_file = filp;
		newattrs.ia_valid |= ATTR_FILE;
	}
Linus Torvalds's avatar
Linus Torvalds committed
215

216
	/* Remove suid/sgid on truncate too */
217
218
219
	ret = should_remove_suid(dentry);
	if (ret)
		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;
Linus Torvalds's avatar
Linus Torvalds committed
225
226
}

227
static long do_sys_truncate(const char __user *pathname, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
228
{
229
230
	struct path path;
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
231
232
233
234
235
236
	int error;

	error = -EINVAL;
	if (length < 0)	/* sorry, but loff_t says... */
		goto out;

237
	error = user_path(pathname, &path);
Linus Torvalds's avatar
Linus Torvalds committed
238
239
	if (error)
		goto out;
240
	inode = path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
241
242
243
244
245
246
247
248
249
250

	/* For directories it's -EISDIR, for other non-regulars - -EINVAL */
	error = -EISDIR;
	if (S_ISDIR(inode->i_mode))
		goto dput_and_out;

	error = -EINVAL;
	if (!S_ISREG(inode->i_mode))
		goto dput_and_out;

251
	error = mnt_want_write(path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
252
253
254
	if (error)
		goto dput_and_out;

255
	error = inode_permission(inode, MAY_WRITE);
256
257
	if (error)
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
258
259

	error = -EPERM;
260
	if (IS_APPEND(inode))
261
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
262

263
	error = get_write_access(inode);
Linus Torvalds's avatar
Linus Torvalds committed
264
	if (error)
265
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
266

267
268
269
270
271
	/*
	 * Make sure that there are no leases.  get_write_access() protects
	 * against the truncate racing with a lease-granting setlease().
	 */
	error = break_lease(inode, FMODE_WRITE);
Linus Torvalds's avatar
Linus Torvalds committed
272
	if (error)
273
		goto put_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
274
275

	error = locks_verify_truncate(inode, NULL, length);
276
277
	if (!error)
		error = security_path_truncate(&path, length, 0);
Linus Torvalds's avatar
Linus Torvalds committed
278
	if (!error) {
279
		vfs_dq_init(inode);
280
		error = do_truncate(path.dentry, length, 0, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
281
282
	}

283
284
put_write_and_out:
	put_write_access(inode);
285
mnt_drop_write_and_out:
286
	mnt_drop_write(path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
287
dput_and_out:
288
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
289
290
291
292
out:
	return error;
}

293
SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds's avatar
Linus Torvalds committed
294
{
295
	return do_sys_truncate(path, length);
Linus Torvalds's avatar
Linus Torvalds committed
296
297
}

298
static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds's avatar
Linus Torvalds committed
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
{
	struct inode * inode;
	struct dentry *dentry;
	struct file * file;
	int error;

	error = -EINVAL;
	if (length < 0)
		goto out;
	error = -EBADF;
	file = fget(fd);
	if (!file)
		goto out;

	/* explicitly opened as large or we are on 64-bit box */
	if (file->f_flags & O_LARGEFILE)
		small = 0;

317
	dentry = file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
	inode = dentry->d_inode;
	error = -EINVAL;
	if (!S_ISREG(inode->i_mode) || !(file->f_mode & FMODE_WRITE))
		goto out_putf;

	error = -EINVAL;
	/* Cannot ftruncate over 2^31 bytes without large file support */
	if (small && length > MAX_NON_LFS)
		goto out_putf;

	error = -EPERM;
	if (IS_APPEND(inode))
		goto out_putf;

	error = locks_verify_truncate(inode, file, length);
333
334
335
	if (!error)
		error = security_path_truncate(&file->f_path, length,
					       ATTR_MTIME|ATTR_CTIME);
Linus Torvalds's avatar
Linus Torvalds committed
336
	if (!error)
337
		error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds's avatar
Linus Torvalds committed
338
339
340
341
342
343
out_putf:
	fput(file);
out:
	return error;
}

344
SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds's avatar
Linus Torvalds committed
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;
Linus Torvalds's avatar
Linus Torvalds committed
350
351
352
353
}

/* LFS versions of truncate are only needed on 32 bit machines */
#if BITS_PER_LONG == 32
354
SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
355
356
357
{
	return do_sys_truncate(path, length);
}
358
359
360
361
362
363
364
#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
asmlinkage long SyS_truncate64(long path, loff_t length)
{
	return SYSC_truncate64((const char __user *) path, length);
}
SYSCALL_ALIAS(sys_truncate64, SyS_truncate64);
#endif
Linus Torvalds's avatar
Linus Torvalds committed
365

366
SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
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;
Linus Torvalds's avatar
Linus Torvalds committed
372
}
373
374
375
376
377
378
#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
asmlinkage long SyS_ftruncate64(long fd, loff_t length)
{
	return SYSC_ftruncate64((unsigned int) fd, length);
}
SYSCALL_ALIAS(sys_ftruncate64, SyS_ftruncate64);
Linus Torvalds's avatar
Linus Torvalds committed
379
#endif
380
#endif /* BITS_PER_LONG == 32 */
Linus Torvalds's avatar
Linus Torvalds committed
381

382
383

int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
384
{
385
386
	struct inode *inode = file->f_path.dentry->d_inode;
	long ret;
387
388

	if (offset < 0 || len <= 0)
389
		return -EINVAL;
390
391
392

	/* Return error if mode is not supported */
	if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
393
		return -EOPNOTSUPP;
394
395

	if (!(file->f_mode & FMODE_WRITE))
396
		return -EBADF;
397
398
399
400
401
402
	/*
	 * Revalidate the write permissions, in case security policy has
	 * changed since the files were opened.
	 */
	ret = security_file_permission(file, MAY_WRITE);
	if (ret)
403
		return ret;
404
405

	if (S_ISFIFO(inode->i_mode))
406
		return -ESPIPE;
407
408
409
410
411
412

	/*
	 * Let individual file system decide if it supports preallocation
	 * for directories or not.
	 */
	if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
413
		return -ENODEV;
414
415
416

	/* Check for wrap through zero too */
	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
417
		return -EFBIG;
418

419
420
	if (!inode->i_op->fallocate)
		return -EOPNOTSUPP;
421

422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
	return inode->i_op->fallocate(inode, mode, offset, len);
}

SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
{
	struct file *file;
	int error = -EBADF;

	file = fget(fd);
	if (file) {
		error = do_fallocate(file, mode, offset, len);
		fput(file);
	}

	return error;
437
}
438

439
440
441
442
443
444
445
#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
asmlinkage long SyS_fallocate(long fd, long mode, loff_t offset, loff_t len)
{
	return SYSC_fallocate((int)fd, (int)mode, offset, len);
}
SYSCALL_ALIAS(sys_fallocate, SyS_fallocate);
#endif
446

Linus Torvalds's avatar
Linus Torvalds committed
447
448
449
450
451
/*
 * access() needs to use the real uid/gid, not the effective uid/gid.
 * We do this by temporarily clearing all FS-related capabilities and
 * switching the fsuid/fsgid around to the real ones.
 */
452
SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds's avatar
Linus Torvalds committed
453
{
454
455
	const struct cred *old_cred;
	struct cred *override_cred;
456
	struct path path;
457
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
458
459
460
461
462
	int res;

	if (mode & ~S_IRWXO)	/* where's F_OK, X_OK, W_OK, R_OK? */
		return -EINVAL;

463
464
465
	override_cred = prepare_creds();
	if (!override_cred)
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
466

467
468
	override_cred->fsuid = override_cred->uid;
	override_cred->fsgid = override_cred->gid;
Linus Torvalds's avatar
Linus Torvalds committed
469

470
	if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells's avatar
David Howells committed
471
		/* Clear the capabilities if we switch to a non-root user */
472
473
		if (override_cred->uid)
			cap_clear(override_cred->cap_effective);
474
		else
475
476
			override_cred->cap_effective =
				override_cred->cap_permitted;
477
	}
Linus Torvalds's avatar
Linus Torvalds committed
478

479
480
	old_cred = override_creds(override_cred);

481
	res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
482
483
484
	if (res)
		goto out;

485
	inode = path.dentry->d_inode;
486
487

	if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
488
489
490
491
492
		/*
		 * MAY_EXEC on regular files is denied if the fs is mounted
		 * with the "noexec" flag.
		 */
		res = -EACCES;
493
		if (path.mnt->mnt_flags & MNT_NOEXEC)
494
495
496
			goto out_path_release;
	}

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
503
504
505
506
507
508
509
510
	/*
	 * This is a rare case where using __mnt_is_readonly()
	 * is OK without a mnt_want/drop_write() pair.  Since
	 * no actual write to the fs is performed here, we do
	 * not need to telegraph to that to anyone.
	 *
	 * By doing this, we accept that this access is
	 * inherently racy and know that the fs may change
	 * state before we even see this result.
	 */
511
	if (__mnt_is_readonly(path.mnt))
512
		res = -EROFS;
Linus Torvalds's avatar
Linus Torvalds committed
513

514
out_path_release:
515
	path_put(&path);
516
out:
517
518
	revert_creds(old_cred);
	put_cred(override_cred);
Linus Torvalds's avatar
Linus Torvalds committed
519
520
521
	return res;
}

522
SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
523
524
525
526
{
	return sys_faccessat(AT_FDCWD, filename, mode);
}

527
SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
528
{
529
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
530
531
	int error;

532
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
533
534
535
	if (error)
		goto out;

536
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds's avatar
Linus Torvalds committed
537
538
539
	if (error)
		goto dput_and_out;

540
	set_fs_pwd(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
541
542

dput_and_out:
543
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
544
545
546
547
out:
	return error;
}

548
SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds's avatar
Linus Torvalds committed
549
550
551
552
553
554
555
556
557
558
{
	struct file *file;
	struct inode *inode;
	int error;

	error = -EBADF;
	file = fget(fd);
	if (!file)
		goto out;

559
	inode = file->f_path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
560
561
562
563
564

	error = -ENOTDIR;
	if (!S_ISDIR(inode->i_mode))
		goto out_putf;

565
	error = inode_permission(inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds's avatar
Linus Torvalds committed
566
	if (!error)
567
		set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
568
569
570
571
572
573
out_putf:
	fput(file);
out:
	return error;
}

574
SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
575
{
576
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
577
578
	int error;

579
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
580
581
582
	if (error)
		goto out;

583
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds's avatar
Linus Torvalds committed
584
585
586
587
588
589
	if (error)
		goto dput_and_out;

	error = -EPERM;
	if (!capable(CAP_SYS_CHROOT))
		goto dput_and_out;
590
591
592
	error = security_path_chroot(&path);
	if (error)
		goto dput_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
593

594
	set_fs_root(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
595
596
	error = 0;
dput_and_out:
597
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
598
599
600
601
out:
	return error;
}

602
SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
603
604
605
606
607
608
609
610
611
612
613
{
	struct inode * inode;
	struct dentry * dentry;
	struct file * file;
	int err = -EBADF;
	struct iattr newattrs;

	file = fget(fd);
	if (!file)
		goto out;

614
	dentry = file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
615
616
	inode = dentry->d_inode;

617
	audit_inode(NULL, dentry);
618

619
	err = mnt_want_write_file(file);
620
	if (err)
Linus Torvalds's avatar
Linus Torvalds committed
621
		goto out_putf;
622
623
624
	err = security_path_chmod(dentry, file->f_vfsmnt, mode);
	if (err)
		goto out_drop_write;
625
	mutex_lock(&inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
626
627
628
629
630
	if (mode == (mode_t) -1)
		mode = inode->i_mode;
	newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
	newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
	err = notify_change(dentry, &newattrs);
631
	mutex_unlock(&inode->i_mutex);
632
out_drop_write:
633
	mnt_drop_write(file->f_path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
634
635
636
637
638
639
out_putf:
	fput(file);
out:
	return err;
}

640
SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
641
{
642
643
	struct path path;
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
644
645
646
	int error;
	struct iattr newattrs;

647
	error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Linus Torvalds's avatar
Linus Torvalds committed
648
649
	if (error)
		goto out;
650
	inode = path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
651

652
	error = mnt_want_write(path.mnt);
653
	if (error)
Linus Torvalds's avatar
Linus Torvalds committed
654
		goto dput_and_out;
655
656
657
	error = security_path_chmod(path.dentry, path.mnt, mode);
	if (error)
		goto out_drop_write;
658
	mutex_lock(&inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
659
660
661
662
	if (mode == (mode_t) -1)
		mode = inode->i_mode;
	newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
	newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
663
	error = notify_change(path.dentry, &newattrs);
664
	mutex_unlock(&inode->i_mutex);
665
out_drop_write:
666
	mnt_drop_write(path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
667
dput_and_out:
668
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
669
670
671
672
out:
	return error;
}

673
SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
674
675
676
677
{
	return sys_fchmodat(AT_FDCWD, filename, mode);
}

Linus Torvalds's avatar
Linus Torvalds committed
678
679
static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
{
680
	struct inode *inode = dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
681
682
683
684
685
686
687
688
689
690
691
692
693
	int error;
	struct iattr newattrs;

	newattrs.ia_valid =  ATTR_CTIME;
	if (user != (uid_t) -1) {
		newattrs.ia_valid |= ATTR_UID;
		newattrs.ia_uid = user;
	}
	if (group != (gid_t) -1) {
		newattrs.ia_valid |= ATTR_GID;
		newattrs.ia_gid = group;
	}
	if (!S_ISDIR(inode->i_mode))
694
695
		newattrs.ia_valid |=
			ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
696
	mutex_lock(&inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
697
	error = notify_change(dentry, &newattrs);
698
	mutex_unlock(&inode->i_mutex);
699

Linus Torvalds's avatar
Linus Torvalds committed
700
701
702
	return error;
}

703
SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
704
{
705
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
706
707
	int error;

708
	error = user_path(filename, &path);
709
710
	if (error)
		goto out;
711
	error = mnt_want_write(path.mnt);
712
713
	if (error)
		goto out_release;
714
715
716
	error = security_path_chown(&path, user, group);
	if (!error)
		error = chown_common(path.dentry, user, group);
717
	mnt_drop_write(path.mnt);
718
out_release:
719
	path_put(&path);
720
out:
Linus Torvalds's avatar
Linus Torvalds committed
721
722
723
	return error;
}

724
725
SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
		gid_t, group, int, flag)
726
{
727
	struct path path;
728
729
730
731
732
733
734
	int error = -EINVAL;
	int follow;

	if ((flag & ~AT_SYMLINK_NOFOLLOW) != 0)
		goto out;

	follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
735
	error = user_path_at(dfd, filename, follow, &path);
736
737
	if (error)
		goto out;
738
	error = mnt_want_write(path.mnt);
739
740
	if (error)
		goto out_release;
741
742
743
	error = security_path_chown(&path, user, group);
	if (!error)
		error = chown_common(path.dentry, user, group);
744
	mnt_drop_write(path.mnt);
745
out_release:
746
	path_put(&path);
747
748
749
750
out:
	return error;
}

751
SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
752
{
753
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
754
755
	int error;

756
	error = user_lpath(filename, &path);
757
758
	if (error)
		goto out;
759
	error = mnt_want_write(path.mnt);
760
761
	if (error)
		goto out_release;
762
763
764
	error = security_path_chown(&path, user, group);
	if (!error)
		error = chown_common(path.dentry, user, group);
765
	mnt_drop_write(path.mnt);
766
out_release:
767
	path_put(&path);
768
out:
Linus Torvalds's avatar
Linus Torvalds committed
769
770
771
	return error;
}

772
SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
773
774
775
{
	struct file * file;
	int error = -EBADF;
776
	struct dentry * dentry;
Linus Torvalds's avatar
Linus Torvalds committed
777
778

	file = fget(fd);
779
780
781
	if (!file)
		goto out;

782
	error = mnt_want_write_file(file);
783
784
	if (error)
		goto out_fput;
785
	dentry = file->f_path.dentry;
786
	audit_inode(NULL, dentry);
787
788
789
	error = security_path_chown(&file->f_path, user, group);
	if (!error)
		error = chown_common(dentry, user, group);
790
791
	mnt_drop_write(file->f_path.mnt);
out_fput:
792
793
	fput(file);
out:
Linus Torvalds's avatar
Linus Torvalds committed
794
795
796
	return error;
}

797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
/*
 * You have to be very careful that these write
 * counts get cleaned up in error cases and
 * upon __fput().  This should probably never
 * be called outside of __dentry_open().
 */
static inline int __get_file_write_access(struct inode *inode,
					  struct vfsmount *mnt)
{
	int error;
	error = get_write_access(inode);
	if (error)
		return error;
	/*
	 * Do not take mount writer counts on
	 * special files since no writes to
	 * the mount itself will occur.
	 */
	if (!special_file(inode->i_mode)) {
		/*
		 * Balanced in __fput()
		 */
		error = mnt_want_write(mnt);
		if (error)
			put_write_access(inode);
	}
	return error;
}

826
static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
827
					int flags, struct file *f,
828
829
					int (*open)(struct inode *, struct file *),
					const struct cred *cred)
Linus Torvalds's avatar
Linus Torvalds committed
830
831
832
833
834
{
	struct inode *inode;
	int error;

	f->f_flags = flags;
835
	f->f_mode = (__force fmode_t)((flags+1) & O_ACCMODE) | FMODE_LSEEK |
836
				FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds's avatar
Linus Torvalds committed
837
838
	inode = dentry->d_inode;
	if (f->f_mode & FMODE_WRITE) {
839
		error = __get_file_write_access(inode, mnt);
Linus Torvalds's avatar
Linus Torvalds committed
840
841
		if (error)
			goto cleanup_file;
842
843
		if (!special_file(inode->i_mode))
			file_take_write(f);
Linus Torvalds's avatar
Linus Torvalds committed
844
845
846
	}

	f->f_mapping = inode->i_mapping;
847
848
	f->f_path.dentry = dentry;
	f->f_path.mnt = mnt;
Linus Torvalds's avatar
Linus Torvalds committed
849
850
851
852
	f->f_pos = 0;
	f->f_op = fops_get(inode->i_fop);
	file_move(f, &inode->i_sb->s_files);

853
	error = security_dentry_open(f, cred);
854
855
856
	if (error)
		goto cleanup_all;

857
858
859
860
	if (!open && f->f_op)
		open = f->f_op->open;
	if (open) {
		error = open(inode, f);
Linus Torvalds's avatar
Linus Torvalds committed
861
862
863
		if (error)
			goto cleanup_all;
	}
864

Linus Torvalds's avatar
Linus Torvalds committed
865
866
867
868
869
870
	f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);

	file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);

	/* NB: we're sure to have correct a_ops only after f_op->open */
	if (f->f_flags & O_DIRECT) {
871
872
		if (!f->f_mapping->a_ops ||
		    ((!f->f_mapping->a_ops->direct_IO) &&
873
		    (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds's avatar
Linus Torvalds committed
874
875
876
877
878
879
880
881
882
			fput(f);
			f = ERR_PTR(-EINVAL);
		}
	}

	return f;

cleanup_all:
	fops_put(f->f_op);
883
	if (f->f_mode & FMODE_WRITE) {
Linus Torvalds's avatar
Linus Torvalds committed
884
		put_write_access(inode);
885
886
887
888
889
890
891
892
		if (!special_file(inode->i_mode)) {
			/*
			 * We don't consider this a real
			 * mnt_want/drop_write() pair
			 * because it all happenend right
			 * here, so just reset the state.
			 */
			file_reset_write(f);
893
			mnt_drop_write(mnt);
894
		}
895
	}
Linus Torvalds's avatar
Linus Torvalds committed
896
	file_kill(f);
897
898
	f->f_path.dentry = NULL;
	f->f_path.mnt = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
899
900
901
902
903
904
905
cleanup_file:
	put_filp(f);
	dput(dentry);
	mntput(mnt);
	return ERR_PTR(error);
}

906
907
908
909
910
911
912
913
914
915
/**
 * lookup_instantiate_filp - instantiates the open intent filp
 * @nd: pointer to nameidata
 * @dentry: pointer to dentry
 * @open: open callback
 *
 * Helper for filesystems that want to use lookup open intents and pass back
 * a fully instantiated struct file to the caller.
 * This function is meant to be called from within a filesystem's
 * lookup method.
916
917
918
919
 * Beware of calling it for non-regular files! Those ->open methods might block
 * (e.g. in fifo_open), leaving you with parent locked (and in case of fifo,
 * leading to a deadlock, as nobody can open that fifo anymore, because
 * another process to open fifo will block on locked parent when doing lookup).
920
921
922
923
924
925
926
927
 * Note that in case of error, nd->intent.open.file is destroyed, but the
 * path information remains valid.
 * If the open callback is set to NULL, then the standard f_op->open()
 * filesystem callback is substituted.
 */
struct file *lookup_instantiate_filp(struct nameidata *nd, struct dentry *dentry,
		int (*open)(struct inode *, struct file *))
{
928
929
	const struct cred *cred = current_cred();

930
931
932
933
	if (IS_ERR(nd->intent.open.file))
		goto out;
	if (IS_ERR(dentry))
		goto out_err;
934
	nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
935
936
					     nd->intent.open.flags - 1,
					     nd->intent.open.file,
937
					     open, cred);
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
out:
	return nd->intent.open.file;
out_err:
	release_open_intent(nd);
	nd->intent.open.file = (struct file *)dentry;
	goto out;
}
EXPORT_SYMBOL_GPL(lookup_instantiate_filp);

/**
 * nameidata_to_filp - convert a nameidata to an open filp.
 * @nd: pointer to nameidata
 * @flags: open flags
 *
 * Note that this function destroys the original nameidata
 */
struct file *nameidata_to_filp(struct nameidata *nd, int flags)
{
956
	const struct cred *cred = current_cred();
957
958
959
960
961
	struct file *filp;

	/* Pick up the filp from the open intent */
	filp = nd->intent.open.file;
	/* Has the filesystem initialised the file for us? */
962
	if (filp->f_path.dentry == NULL)
963
		filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
964
				     NULL, cred);
965
	else
Jan Blunck's avatar
Jan Blunck committed
966
		path_put(&nd->path);
967
968
969
	return filp;
}

970
971
972
973
/*
 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
 * error.
 */
974
975
struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
			 const struct cred *cred)
976
977
978
979
{
	int error;
	struct file *f;

980
981
	validate_creds(cred);

982
983
984
985
986
987
988
989
990
991
992
993
	/*
	 * We must always pass in a valid mount pointer.   Historically
	 * callers got away with not passing it, but we must enforce this at
	 * the earliest possible point now to avoid strange problems deep in the
	 * filesystem stack.
	 */
	if (!mnt) {
		printk(KERN_WARNING "%s called with NULL vfsmount\n", __func__);
		dump_stack();
		return ERR_PTR(-EINVAL);
	}

994
995
	error = -ENFILE;
	f = get_empty_filp();
996
997
998
	if (f == NULL) {
		dput(dentry);
		mntput(mnt);
999
		return ERR_PTR(error);
1000
	}
1001

1002
	return __dentry_open(dentry, mnt, flags, f, NULL, cred);
1003
}
Linus Torvalds's avatar
Linus Torvalds committed
1004
1005
EXPORT_SYMBOL(dentry_open);

1006
static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds's avatar
Linus Torvalds committed
1007
{
1008
1009
	struct fdtable *fdt = files_fdtable(files);
	__FD_CLR(fd, fdt->open_fds);
1010
1011
	if (fd < files->next_fd)
		files->next_fd = fd;
Linus Torvalds's avatar
Linus Torvalds committed
1012
1013
}

1014
void put_unused_fd(unsigned int fd)
Linus Torvalds's avatar
Linus Torvalds committed
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
{
	struct files_struct *files = current->files;
	spin_lock(&files->file_lock);
	__put_unused_fd(files, fd);
	spin_unlock(&files->file_lock);
}

EXPORT_SYMBOL(put_unused_fd);

/*
1025
 * Install a file pointer in the fd array.
Linus Torvalds's avatar
Linus Torvalds committed
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
 *
 * The VFS is full of places where we drop the files lock between
 * setting the open_fds bitmap and installing the file in the file
 * array.  At any such point, we are vulnerable to a dup2() race
 * installing a file in the array before us.  We need to detect this and
 * fput() the struct file we are about to overwrite in this case.
 *
 * It should never happen - if we allow dup2() do it, _really_ bad things
 * will follow.
 */

1037
void fd_install(unsigned int fd, struct file *file)
Linus Torvalds's avatar
Linus Torvalds committed
1038
1039
{
	struct files_struct *files = current->files;
1040
	struct fdtable *fdt;
Linus Torvalds's avatar
Linus Torvalds committed
1041
	spin_lock(&files->file_lock);
1042
	fdt = files_fdtable(files);
1043
1044
	BUG_ON(fdt->fd[fd] != NULL);
	rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds's avatar
Linus Torvalds committed
1045
1046
1047
1048
1049
	spin_unlock(&files->file_lock);
}

EXPORT_SYMBOL(fd_install);

1050
long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds's avatar
Linus Torvalds committed
1051
{
1052
1053
	char *tmp = getname(filename);
	int fd = PTR_ERR(tmp);
Linus Torvalds's avatar
Linus Torvalds committed
1054
1055

	if (!IS_ERR(tmp)) {
Ulrich Drepper's avatar
Ulrich Drepper committed
1056
		fd = get_unused_fd_flags(flags);
Linus Torvalds's avatar
Linus Torvalds committed
1057
		if (fd >= 0) {
1058
			struct file *f = do_filp_open(dfd, tmp, flags, mode, 0);
1059
1060
1061
1062
			if (IS_ERR(f)) {
				put_unused_fd(fd);
				fd = PTR_ERR(f);
			} else {
1063
				fsnotify_open(f->f_path.dentry);
1064
1065
				fd_install(fd, f);
			}
Linus Torvalds's avatar
Linus Torvalds committed
1066
1067
1068
1069
1070
		}
		putname(tmp);
	}
	return fd;
}
1071

1072
SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode)
1073
{
1074
1075
	long ret;

1076
1077
1078
	if (force_o_largefile())
		flags |= O_LARGEFILE;

1079
1080
	ret = do_sys_open(AT_FDCWD, filename, flags, mode);
	/* avoid REGPARM breakage on x86: */
1081
	asmlinkage_protect(3, ret, filename, flags, mode);
1082
	return ret;
1083
}
Linus Torvalds's avatar
Linus Torvalds committed
1084

1085
1086
SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
		int, mode)
1087
{
1088
1089
	long ret;

1090
1091
1092
	if (force_o_largefile())
		flags |= O_LARGEFILE;

1093
1094
	ret = do_sys_open(dfd, filename, flags, mode);
	/* avoid REGPARM breakage on x86: */
1095
	asmlinkage_protect(4, ret, dfd, filename, flags, mode);
1096
	return ret;
1097
1098
}

Linus Torvalds's avatar
Linus Torvalds committed
1099
1100
1101
1102
1103
1104
#ifndef __alpha__

/*
 * For backward compatibility?  Maybe this should be moved
 * into arch/i386 instead?
 */
1105
SYSCALL_DEFINE2(creat, const char __user *, pathname, int, mode)
Linus Torvalds's avatar
Linus Torvalds committed
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
{
	return sys_open(pathname, O_CREAT | O_WRONLY | O_TRUNC, mode);
}

#endif

/*
 * "id" is the POSIX thread ID. We use the
 * files pointer for this..
 */
int filp_close(struct file *filp, fl_owner_t id)
{
1118
	int retval = 0;
Linus Torvalds's avatar
Linus Torvalds committed
1119
1120
1121

	if (!file_count(filp)) {
		printk(KERN_ERR "VFS: Close: file count is 0\n");
1122
		return 0;
Linus Torvalds's avatar