open.c 24.4 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>
Robert Love's avatar
Robert Love committed
11
#include <linux/fsnotify.h>
Linus Torvalds's avatar
Linus Torvalds committed
12 13 14 15
#include <linux/module.h>
#include <linux/tty.h>
#include <linux/namei.h>
#include <linux/backing-dev.h>
16
#include <linux/capability.h>
17
#include <linux/securebits.h>
Linus Torvalds's avatar
Linus Torvalds committed
18 19
#include <linux/security.h>
#include <linux/mount.h>
20
#include <linux/fcntl.h>
21
#include <linux/slab.h>
Linus Torvalds's avatar
Linus Torvalds committed
22 23
#include <asm/uaccess.h>
#include <linux/fs.h>
24
#include <linux/personality.h>
Linus Torvalds's avatar
Linus Torvalds committed
25 26
#include <linux/pagemap.h>
#include <linux/syscalls.h>
27
#include <linux/rcupdate.h>
28
#include <linux/audit.h>
29
#include <linux/falloc.h>
30
#include <linux/fs_struct.h>
31
#include <linux/ima.h>
32
#include <linux/dnotify.h>
Al Viro's avatar
Al Viro committed
33
#include <linux/compat.h>
Linus Torvalds's avatar
Linus Torvalds committed
34

35 36
#include "internal.h"

37 38
int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
	struct file *filp)
Linus Torvalds's avatar
Linus Torvalds committed
39
{
40
	int ret;
Linus Torvalds's avatar
Linus Torvalds committed
41 42 43 44 45 46 47
	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;
48
	newattrs.ia_valid = ATTR_SIZE | time_attrs;
49 50 51 52
	if (filp) {
		newattrs.ia_file = filp;
		newattrs.ia_valid |= ATTR_FILE;
	}
Linus Torvalds's avatar
Linus Torvalds committed
53

54
	/* Remove suid/sgid on truncate too */
55 56 57
	ret = should_remove_suid(dentry);
	if (ret)
		newattrs.ia_valid |= ret | ATTR_FORCE;
58

59
	mutex_lock(&dentry->d_inode->i_mutex);
60
	ret = notify_change(dentry, &newattrs);
61
	mutex_unlock(&dentry->d_inode->i_mutex);
62
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
63
}
64
EXPORT_SYMBOL(do_truncate);
Linus Torvalds's avatar
Linus Torvalds committed
65

66
long vfs_truncate(struct path *path, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
67
{
68
	struct inode *inode;
69
	long error;
Linus Torvalds's avatar
Linus Torvalds committed
70

71
	inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
72 73 74

	/* For directories it's -EISDIR, for other non-regulars - -EINVAL */
	if (S_ISDIR(inode->i_mode))
75
		return -EISDIR;
Linus Torvalds's avatar
Linus Torvalds committed
76
	if (!S_ISREG(inode->i_mode))
77
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
78

79
	error = mnt_want_write(path->mnt);
Linus Torvalds's avatar
Linus Torvalds committed
80
	if (error)
81
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
82

83
	error = inode_permission(inode, MAY_WRITE);
84 85
	if (error)
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
86 87

	error = -EPERM;
88
	if (IS_APPEND(inode))
89
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
90

91
	error = get_write_access(inode);
Linus Torvalds's avatar
Linus Torvalds committed
92
	if (error)
93
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
94

95 96 97 98
	/*
	 * Make sure that there are no leases.  get_write_access() protects
	 * against the truncate racing with a lease-granting setlease().
	 */
99
	error = break_lease(inode, O_WRONLY);
Linus Torvalds's avatar
Linus Torvalds committed
100
	if (error)
101
		goto put_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
102 103

	error = locks_verify_truncate(inode, NULL, length);
104
	if (!error)
105
		error = security_path_truncate(path);
106
	if (!error)
107
		error = do_truncate(path->dentry, length, 0, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
108

109 110
put_write_and_out:
	put_write_access(inode);
111
mnt_drop_write_and_out:
112
	mnt_drop_write(path->mnt);
Linus Torvalds's avatar
Linus Torvalds committed
113 114 115
out:
	return error;
}
116 117 118 119
EXPORT_SYMBOL_GPL(vfs_truncate);

static long do_sys_truncate(const char __user *pathname, loff_t length)
{
120
	unsigned int lookup_flags = LOOKUP_FOLLOW;
121 122 123 124 125 126
	struct path path;
	int error;

	if (length < 0)	/* sorry, but loff_t says... */
		return -EINVAL;

127 128
retry:
	error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
129 130 131 132
	if (!error) {
		error = vfs_truncate(&path, length);
		path_put(&path);
	}
133 134 135 136
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
137 138
	return error;
}
Linus Torvalds's avatar
Linus Torvalds committed
139

140
SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds's avatar
Linus Torvalds committed
141
{
142
	return do_sys_truncate(path, length);
Linus Torvalds's avatar
Linus Torvalds committed
143 144
}

Al Viro's avatar
Al Viro committed
145 146 147 148 149 150 151
#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
{
	return do_sys_truncate(path, length);
}
#endif

152
static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds's avatar
Linus Torvalds committed
153
{
154
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
155
	struct dentry *dentry;
156
	struct fd f;
Linus Torvalds's avatar
Linus Torvalds committed
157 158 159 160 161 162
	int error;

	error = -EINVAL;
	if (length < 0)
		goto out;
	error = -EBADF;
163 164
	f = fdget(fd);
	if (!f.file)
Linus Torvalds's avatar
Linus Torvalds committed
165 166 167
		goto out;

	/* explicitly opened as large or we are on 64-bit box */
168
	if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds's avatar
Linus Torvalds committed
169 170
		small = 0;

171
	dentry = f.file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
172 173
	inode = dentry->d_inode;
	error = -EINVAL;
174
	if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds's avatar
Linus Torvalds committed
175 176 177 178 179 180 181 182 183 184 185
		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;

186
	sb_start_write(inode->i_sb);
187
	error = locks_verify_truncate(inode, f.file, length);
188
	if (!error)
189
		error = security_path_truncate(&f.file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
190
	if (!error)
191
		error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
192
	sb_end_write(inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
193
out_putf:
194
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
195 196 197 198
out:
	return error;
}

199
SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds's avatar
Linus Torvalds committed
200
{
201
	return do_sys_ftruncate(fd, length, 1);
Linus Torvalds's avatar
Linus Torvalds committed
202 203
}

Al Viro's avatar
Al Viro committed
204 205 206 207 208 209 210
#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
{
	return do_sys_ftruncate(fd, length, 1);
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
211 212
/* LFS versions of truncate are only needed on 32 bit machines */
#if BITS_PER_LONG == 32
213
SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
Linus Torvalds's avatar
Linus Torvalds committed
214 215 216 217
{
	return do_sys_truncate(path, length);
}

218
SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
Linus Torvalds's avatar
Linus Torvalds committed
219
{
220
	return do_sys_ftruncate(fd, length, 0);
Linus Torvalds's avatar
Linus Torvalds committed
221
}
222
#endif /* BITS_PER_LONG == 32 */
Linus Torvalds's avatar
Linus Torvalds committed
223

224 225

int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
226
{
Al Viro's avatar
Al Viro committed
227
	struct inode *inode = file_inode(file);
228
	long ret;
229 230

	if (offset < 0 || len <= 0)
231
		return -EINVAL;
232 233

	/* Return error if mode is not supported */
234 235 236 237 238 239
	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
		return -EOPNOTSUPP;

	/* Punch hole must have keep size set */
	if ((mode & FALLOC_FL_PUNCH_HOLE) &&
	    !(mode & FALLOC_FL_KEEP_SIZE))
240
		return -EOPNOTSUPP;
241 242

	if (!(file->f_mode & FMODE_WRITE))
243
		return -EBADF;
244 245 246 247 248 249 250 251

	/* It's not possible punch hole on append only file */
	if (mode & FALLOC_FL_PUNCH_HOLE && IS_APPEND(inode))
		return -EPERM;

	if (IS_IMMUTABLE(inode))
		return -EPERM;

252 253 254 255 256 257
	/*
	 * Revalidate the write permissions, in case security policy has
	 * changed since the files were opened.
	 */
	ret = security_file_permission(file, MAY_WRITE);
	if (ret)
258
		return ret;
259 260

	if (S_ISFIFO(inode->i_mode))
261
		return -ESPIPE;
262 263 264 265 266 267

	/*
	 * Let individual file system decide if it supports preallocation
	 * for directories or not.
	 */
	if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
268
		return -ENODEV;
269 270 271

	/* Check for wrap through zero too */
	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
272
		return -EFBIG;
273

274
	if (!file->f_op->fallocate)
275
		return -EOPNOTSUPP;
276

277 278 279 280
	sb_start_write(inode->i_sb);
	ret = file->f_op->fallocate(file, mode, offset, len);
	sb_end_write(inode->i_sb);
	return ret;
281
}
282
EXPORT_SYMBOL(do_fallocate);
283

284
SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
285
{
286
	struct fd f = fdget(fd);
287 288
	int error = -EBADF;

289 290 291
	if (f.file) {
		error = do_fallocate(f.file, mode, offset, len);
		fdput(f);
292 293
	}
	return error;
294
}
295

Linus Torvalds's avatar
Linus Torvalds committed
296 297 298 299 300
/*
 * 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.
 */
301
SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds's avatar
Linus Torvalds committed
302
{
303 304
	const struct cred *old_cred;
	struct cred *override_cred;
305
	struct path path;
306
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
307
	int res;
308
	unsigned int lookup_flags = LOOKUP_FOLLOW;
Linus Torvalds's avatar
Linus Torvalds committed
309 310 311 312

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

313 314 315
	override_cred = prepare_creds();
	if (!override_cred)
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
316

317 318
	override_cred->fsuid = override_cred->uid;
	override_cred->fsgid = override_cred->gid;
Linus Torvalds's avatar
Linus Torvalds committed
319

320
	if (!issecure(SECURE_NO_SETUID_FIXUP)) {
321
		/* Clear the capabilities if we switch to a non-root user */
322 323
		kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
		if (!uid_eq(override_cred->uid, root_uid))
324
			cap_clear(override_cred->cap_effective);
325
		else
326 327
			override_cred->cap_effective =
				override_cred->cap_permitted;
328
	}
Linus Torvalds's avatar
Linus Torvalds committed
329

330
	old_cred = override_creds(override_cred);
331 332
retry:
	res = user_path_at(dfd, filename, lookup_flags, &path);
333 334 335
	if (res)
		goto out;

336
	inode = path.dentry->d_inode;
337 338

	if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
339 340 341 342 343
		/*
		 * MAY_EXEC on regular files is denied if the fs is mounted
		 * with the "noexec" flag.
		 */
		res = -EACCES;
344
		if (path.mnt->mnt_flags & MNT_NOEXEC)
345 346 347
			goto out_path_release;
	}

348
	res = inode_permission(inode, mode | MAY_ACCESS);
349
	/* SuS v2 requires we report a read only fs too */
350
	if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
351
		goto out_path_release;
352 353 354 355 356 357 358 359 360 361
	/*
	 * 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.
	 */
362
	if (__mnt_is_readonly(path.mnt))
363
		res = -EROFS;
Linus Torvalds's avatar
Linus Torvalds committed
364

365
out_path_release:
366
	path_put(&path);
367 368 369 370
	if (retry_estale(res, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
371
out:
372 373
	revert_creds(old_cred);
	put_cred(override_cred);
Linus Torvalds's avatar
Linus Torvalds committed
374 375 376
	return res;
}

377
SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
378 379 380 381
{
	return sys_faccessat(AT_FDCWD, filename, mode);
}

382
SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
383
{
384
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
385
	int error;
386 387 388
	unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
retry:
	error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds's avatar
Linus Torvalds committed
389 390 391
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
392
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
393 394 395
	if (error)
		goto dput_and_out;

396
	set_fs_pwd(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
397 398

dput_and_out:
399
	path_put(&path);
400 401 402 403
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
Linus Torvalds's avatar
Linus Torvalds committed
404 405 406 407
out:
	return error;
}

408
SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds's avatar
Linus Torvalds committed
409
{
410
	struct fd f = fdget_raw(fd);
Linus Torvalds's avatar
Linus Torvalds committed
411
	struct inode *inode;
412
	int error = -EBADF;
Linus Torvalds's avatar
Linus Torvalds committed
413 414

	error = -EBADF;
415
	if (!f.file)
Linus Torvalds's avatar
Linus Torvalds committed
416 417
		goto out;

Al Viro's avatar
Al Viro committed
418
	inode = file_inode(f.file);
Linus Torvalds's avatar
Linus Torvalds committed
419 420 421 422 423

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

Eric Paris's avatar
Eric Paris committed
424
	error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
425
	if (!error)
426
		set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
427
out_putf:
428
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
429 430 431 432
out:
	return error;
}

433
SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
434
{
435
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
436
	int error;
437 438 439
	unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
retry:
	error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds's avatar
Linus Torvalds committed
440 441 442
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
443
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
444 445 446 447
	if (error)
		goto dput_and_out;

	error = -EPERM;
448
	if (!nsown_capable(CAP_SYS_CHROOT))
Linus Torvalds's avatar
Linus Torvalds committed
449
		goto dput_and_out;
450 451 452
	error = security_path_chroot(&path);
	if (error)
		goto dput_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
453

454
	set_fs_root(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
455 456
	error = 0;
dput_and_out:
457
	path_put(&path);
458 459 460 461
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
Linus Torvalds's avatar
Linus Torvalds committed
462 463 464 465
out:
	return error;
}

466
static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
467
{
468
	struct inode *inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
469
	struct iattr newattrs;
470
	int error;
Linus Torvalds's avatar
Linus Torvalds committed
471

472 473 474
	error = mnt_want_write(path->mnt);
	if (error)
		return error;
475
	mutex_lock(&inode->i_mutex);
476
	error = security_path_chmod(path, mode);
477
	if (error)
478
		goto out_unlock;
Linus Torvalds's avatar
Linus Torvalds committed
479 480
	newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
	newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
481
	error = notify_change(path->dentry, &newattrs);
482
out_unlock:
483
	mutex_unlock(&inode->i_mutex);
484 485 486 487
	mnt_drop_write(path->mnt);
	return error;
}

488
SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
489 490 491 492 493 494
{
	struct file * file;
	int err = -EBADF;

	file = fget(fd);
	if (file) {
495
		audit_inode(NULL, file->f_path.dentry, 0);
496 497 498
		err = chmod_common(&file->f_path, mode);
		fput(file);
	}
Linus Torvalds's avatar
Linus Torvalds committed
499 500 501
	return err;
}

502
SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
503
{
504
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
505
	int error;
506 507 508
	unsigned int lookup_flags = LOOKUP_FOLLOW;
retry:
	error = user_path_at(dfd, filename, lookup_flags, &path);
509 510 511
	if (!error) {
		error = chmod_common(&path, mode);
		path_put(&path);
512 513 514 515
		if (retry_estale(error, lookup_flags)) {
			lookup_flags |= LOOKUP_REVAL;
			goto retry;
		}
516
	}
Linus Torvalds's avatar
Linus Torvalds committed
517 518 519
	return error;
}

520
SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
521 522 523 524
{
	return sys_fchmodat(AT_FDCWD, filename, mode);
}

525
static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds's avatar
Linus Torvalds committed
526
{
527
	struct inode *inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
528 529
	int error;
	struct iattr newattrs;
530 531 532 533 534
	kuid_t uid;
	kgid_t gid;

	uid = make_kuid(current_user_ns(), user);
	gid = make_kgid(current_user_ns(), group);
Linus Torvalds's avatar
Linus Torvalds committed
535 536 537

	newattrs.ia_valid =  ATTR_CTIME;
	if (user != (uid_t) -1) {
538 539
		if (!uid_valid(uid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
540
		newattrs.ia_valid |= ATTR_UID;
541
		newattrs.ia_uid = uid;
Linus Torvalds's avatar
Linus Torvalds committed
542 543
	}
	if (group != (gid_t) -1) {
544 545
		if (!gid_valid(gid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
546
		newattrs.ia_valid |= ATTR_GID;
547
		newattrs.ia_gid = gid;
Linus Torvalds's avatar
Linus Torvalds committed
548 549
	}
	if (!S_ISDIR(inode->i_mode))
550 551
		newattrs.ia_valid |=
			ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
552
	mutex_lock(&inode->i_mutex);
553
	error = security_path_chown(path, uid, gid);
554 555
	if (!error)
		error = notify_change(path->dentry, &newattrs);
556
	mutex_unlock(&inode->i_mutex);
557

Linus Torvalds's avatar
Linus Torvalds committed
558 559 560
	return error;
}

561 562
SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
		gid_t, group, int, flag)
563
{
564
	struct path path;
565
	int error = -EINVAL;
566
	int lookup_flags;
567

568
	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
569 570
		goto out;

571 572 573
	lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
	if (flag & AT_EMPTY_PATH)
		lookup_flags |= LOOKUP_EMPTY;
574
retry:
575
	error = user_path_at(dfd, filename, lookup_flags, &path);
576 577
	if (error)
		goto out;
578
	error = mnt_want_write(path.mnt);
579 580
	if (error)
		goto out_release;
581
	error = chown_common(&path, user, group);
582
	mnt_drop_write(path.mnt);
583
out_release:
584
	path_put(&path);
585 586 587 588
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
589 590 591 592
out:
	return error;
}

593
SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
594
{
595 596
	return sys_fchownat(AT_FDCWD, filename, user, group, 0);
}
Linus Torvalds's avatar
Linus Torvalds committed
597

598 599 600 601
SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
{
	return sys_fchownat(AT_FDCWD, filename, user, group,
			    AT_SYMLINK_NOFOLLOW);
Linus Torvalds's avatar
Linus Torvalds committed
602 603
}

604
SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
605
{
606
	struct fd f = fdget(fd);
Linus Torvalds's avatar
Linus Torvalds committed
607 608
	int error = -EBADF;

609
	if (!f.file)
610 611
		goto out;

612
	error = mnt_want_write_file(f.file);
613 614
	if (error)
		goto out_fput;
615
	audit_inode(NULL, f.file->f_path.dentry, 0);
616 617
	error = chown_common(&f.file->f_path, user, group);
	mnt_drop_write_file(f.file);
618
out_fput:
619
	fdput(f);
620
out:
Linus Torvalds's avatar
Linus Torvalds committed
621 622 623
	return error;
}

624 625 626 627 628 629 630 631 632
/*
 * 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)
{
633
	int error = get_write_access(inode);
634 635
	if (error)
		return error;
636 637 638
	error = __mnt_want_write(mnt);
	if (error)
		put_write_access(inode);
639 640 641
	return error;
}

642 643 644 645 646 647 648 649 650 651 652 653 654
int open_check_o_direct(struct file *f)
{
	/* NB: we're sure to have correct a_ops only after f_op->open */
	if (f->f_flags & O_DIRECT) {
		if (!f->f_mapping->a_ops ||
		    ((!f->f_mapping->a_ops->direct_IO) &&
		    (!f->f_mapping->a_ops->get_xip_mem))) {
			return -EINVAL;
		}
	}
	return 0;
}

655
static int do_dentry_open(struct file *f,
656 657
			  int (*open)(struct inode *, struct file *),
			  const struct cred *cred)
Linus Torvalds's avatar
Linus Torvalds committed
658
{
659
	static const struct file_operations empty_fops = {};
Linus Torvalds's avatar
Linus Torvalds committed
660 661 662
	struct inode *inode;
	int error;

Al Viro's avatar
Al Viro committed
663
	f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
664
				FMODE_PREAD | FMODE_PWRITE;
665 666 667 668

	if (unlikely(f->f_flags & O_PATH))
		f->f_mode = FMODE_PATH;

669
	path_get(&f->f_path);
670
	inode = f->f_inode = f->f_path.dentry->d_inode;
671
	if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
672
		error = __get_file_write_access(inode, f->f_path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
673 674
		if (error)
			goto cleanup_file;
675
		file_take_write(f);
Linus Torvalds's avatar
Linus Torvalds committed
676 677 678
	}

	f->f_mapping = inode->i_mapping;
679
	file_sb_list_add(f, inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
680

681 682
	if (unlikely(f->f_mode & FMODE_PATH)) {
		f->f_op = &empty_fops;
683
		return 0;
684 685 686 687
	}

	f->f_op = fops_get(inode->i_fop);

688
	error = security_file_open(f, cred);
689 690 691
	if (error)
		goto cleanup_all;

692 693 694 695
	error = break_lease(inode, f->f_flags);
	if (error)
		goto cleanup_all;

696 697 698 699
	if (!open && f->f_op)
		open = f->f_op->open;
	if (open) {
		error = open(inode, f);
Linus Torvalds's avatar
Linus Torvalds committed
700 701 702
		if (error)
			goto cleanup_all;
	}
703 704
	if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
		i_readcount_inc(inode);
705

Linus Torvalds's avatar
Linus Torvalds committed
706 707 708 709
	f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);

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

710
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
711 712 713

cleanup_all:
	fops_put(f->f_op);
714
	file_sb_list_del(f);
715
	if (f->f_mode & FMODE_WRITE) {
716 717 718 719 720 721 722
		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.
			 */
723
			put_write_access(inode);
724
			file_reset_write(f);
725
			__mnt_drop_write(f->f_path.mnt);
726
		}
727
	}
Linus Torvalds's avatar
Linus Torvalds committed
728
cleanup_file:
729 730 731
	path_put(&f->f_path);
	f->f_path.mnt = NULL;
	f->f_path.dentry = NULL;
732
	f->f_inode = NULL;
733
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
734 735
}

736 737 738 739 740 741 742 743 744 745 746
/**
 * finish_open - finish opening a file
 * @od: opaque open data
 * @dentry: pointer to dentry
 * @open: open callback
 *
 * This can be used to finish opening a file passed to i_op->atomic_open().
 *
 * If the open callback is set to NULL, then the standard f_op->open()
 * filesystem callback is substituted.
 */
Al Viro's avatar
Al Viro committed
747 748 749
int finish_open(struct file *file, struct dentry *dentry,
		int (*open)(struct inode *, struct file *),
		int *opened)
750
{
751
	int error;
Al Viro's avatar
Al Viro committed
752
	BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
753

754
	file->f_path.dentry = dentry;
755
	error = do_dentry_open(file, open, current_cred());
756
	if (!error)
757
		*opened |= FILE_OPENED;
758

759
	return error;
760 761 762 763 764 765 766 767 768 769 770 771
}
EXPORT_SYMBOL(finish_open);

/**
 * finish_no_open - finish ->atomic_open() without opening the file
 *
 * @od: opaque open data
 * @dentry: dentry or NULL (as returned from ->lookup())
 *
 * This can be used to set the result of a successful lookup in ->atomic_open().
 * The filesystem's atomic_open() method shall return NULL after calling this.
 */
772
int finish_no_open(struct file *file, struct dentry *dentry)
773
{
Al Viro's avatar
Al Viro committed
774
	file->f_path.dentry = dentry;
775
	return 1;
776 777 778
}
EXPORT_SYMBOL(finish_no_open);

779
struct file *dentry_open(const struct path *path, int flags,
780
			 const struct cred *cred)
781 782 783 784
{
	int error;
	struct file *f;

785 786
	validate_creds(cred);

787
	/* We must always pass in a valid mount pointer. */
788
	BUG_ON(!path->mnt);
789

790
	f = get_empty_filp();
791 792 793 794 795 796 797 798 799 800 801 802 803
	if (!IS_ERR(f)) {
		f->f_flags = flags;
		f->f_path = *path;
		error = do_dentry_open(f, NULL, cred);
		if (!error) {
			/* from now on we need fput() to dispose of f */
			error = open_check_o_direct(f);
			if (error) {
				fput(f);
				f = ERR_PTR(error);
			}
		} else { 
			put_filp(f);
804 805 806 807
			f = ERR_PTR(error);
		}
	}
	return f;
808
}
Linus Torvalds's avatar
Linus Torvalds committed
809 810
EXPORT_SYMBOL(dentry_open);

811
static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
812 813 814 815
{
	int lookup_flags = 0;
	int acc_mode;

816 817 818 819
	if (flags & O_CREAT)
		op->mode = (mode & S_IALLUGO) | S_IFREG;
	else
		op->mode = 0;
820 821

	/* Must never be set by userspace */
822
	flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
823 824 825 826 827 828 829 830 831 832

	/*
	 * O_SYNC is implemented as __O_SYNC|O_DSYNC.  As many places only
	 * check for O_DSYNC if the need any syncing at all we enforce it's
	 * always set instead of having to deal with possibly weird behaviour
	 * for malicious applications setting only __O_SYNC.
	 */
	if (flags & __O_SYNC)
		flags |= O_DSYNC;

833 834 835 836 837 838 839 840 841 842
	/*
	 * If we have O_PATH in the open flag. Then we
	 * cannot have anything other than the below set of flags
	 */
	if (flags & O_PATH) {
		flags &= O_DIRECTORY | O_NOFOLLOW | O_PATH;
		acc_mode = 0;
	} else {
		acc_mode = MAY_OPEN | ACC_MODE(flags);
	}
843

844
	op->open_flag = flags;
845 846 847 848 849 850 851 852 853 854 855 856

	/* O_TRUNC implies we need access checks for write permissions */
	if (flags & O_TRUNC)
		acc_mode |= MAY_WRITE;

	/* Allow the LSM permission hook to distinguish append
	   access from general write access. */
	if (flags & O_APPEND)
		acc_mode |= MAY_APPEND;

	op->acc_mode = acc_mode;

857 858
	op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;

859 860 861 862 863 864 865 866 867 868 869 870 871
	if (flags & O_CREAT) {
		op->intent |= LOOKUP_CREATE;
		if (flags & O_EXCL)
			op->intent |= LOOKUP_EXCL;
	}

	if (flags & O_DIRECTORY)
		lookup_flags |= LOOKUP_DIRECTORY;
	if (!(flags & O_NOFOLLOW))
		lookup_flags |= LOOKUP_FOLLOW;
	return lookup_flags;
}

872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889
/**
 * file_open_name - open file and return file pointer
 *
 * @name:	struct filename containing path to open
 * @flags:	open flags as per the open(2) second argument
 * @mode:	mode for the new file if O_CREAT is set, else ignored
 *
 * This is the helper to open a file from kernelspace if you really
 * have to.  But in generally you should not do this, so please move
 * along, nothing to see here..
 */
struct file *file_open_name(struct filename *name, int flags, umode_t mode)
{
	struct open_flags op;
	int lookup = build_open_flags(flags, mode, &op);
	return do_filp_open(AT_FDCWD, name, &op, lookup);
}

890 891 892 893 894 895 896 897 898 899 900
/**
 * filp_open - open file and return file pointer
 *
 * @filename:	path to open
 * @flags:	open flags as per the open(2) second argument
 * @mode:	mode for the new file if O_CREAT is set, else ignored
 *
 * This is the helper to open a file from kernelspace if you really
 * have to.  But in generally you should not do this, so please move
 * along, nothing to see here..
 */
901
struct file *filp_open(const char *filename, int flags, umode_t mode)
902
{
903 904
	struct filename name = {.name = filename};
	return file_open_name(&name, flags, mode);
905 906 907
}
EXPORT_SYMBOL(filp_open);

908 909 910 911 912 913 914 915 916 917 918 919 920 921
struct file *file_open_root(struct dentry *dentry, struct vfsmount *mnt,
			    const char *filename, int flags)
{
	struct open_flags op;
	int lookup = build_open_flags(flags, 0, &op);
	if (flags & O_CREAT)
		return ERR_PTR(-EINVAL);
	if (!filename && (flags & O_DIRECTORY))
		if (!dentry->d_inode->i_op->lookup)
			return ERR_PTR(-ENOTDIR);
	return do_file_open_root(dentry, mnt, filename, &op, lookup);
}
EXPORT_SYMBOL(file_open_root);

922
long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
923
{
924 925
	struct open_flags op;
	int lookup = build_open_flags(flags, mode, &op);
926
	struct filename *tmp = getname(filename);
927
	int fd = PTR_ERR(tmp);
Linus Torvalds's avatar
Linus Torvalds committed
928 929

	if (!IS_ERR(tmp)) {
Ulrich Drepper's avatar
Ulrich Drepper committed
930
		fd = get_unused_fd_flags(flags);
Linus Torvalds's avatar
Linus Torvalds committed
931
		if (fd >= 0) {
932
			struct file *f = do_filp_open(dfd, tmp, &op, lookup);
933 934 935 936
			if (IS_ERR(f)) {
				put_unused_fd(fd);
				fd = PTR_ERR(f);
			} else {
937
				fsnotify_open(f);
938 939
				fd_install(fd, f);
			}
Linus Torvalds's avatar
Linus Torvalds committed
940 941 942 943 944
		}
		putname(tmp);
	}
	return fd;
}
945

946
SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
947 948 949 950
{
	if (force_o_largefile())
		flags |= O_LARGEFILE;

951
	return do_sys_open(AT_FDCWD, filename, flags, mode);
952
}
Linus Torvalds's avatar
Linus Torvalds committed
953

954
SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
955
		umode_t, mode)
956 957 958 959
{
	if (force_o_largefile())
		flags |= O_LARGEFILE;

960
	return do_sys_open(dfd, filename, flags, mode);
961 962
}

Linus Torvalds's avatar
Linus Torvalds committed
963 964 965 966 967 968
#ifndef __alpha__

/*
 * For backward compatibility?  Maybe this should be moved
 * into arch/i386 instead?
 */
969
SYSCALL_DEFINE2(creat, const char __user *, pathname, umode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
970 971 972 973 974 975 976 977 978 979 980 981
{
	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)
{
982
	int retval = 0;
Linus Torvalds's avatar
Linus Torvalds committed
983 984 985

	if (!file_count(filp)) {
		printk(KERN_ERR "VFS: Close: file count is 0\n");
986
		return 0;
Linus Torvalds's avatar
Linus Torvalds committed
987 988
	}

989
	if (filp->f_op && filp->f_op->flush)
990
		retval = filp->f_op->flush(filp, id);
Linus Torvalds's avatar
Linus Torvalds committed
991

992 993 994 995
	if (likely(!(filp->f_mode & FMODE_PATH))) {
		dnotify_flush(filp, id);
		locks_remove_posix(filp, id);
	}
Linus Torvalds's avatar
Linus Torvalds committed
996 997 998 999 1000 1001 1002 1003 1004 1005 1006
	fput(filp);
	return retval;
}

EXPORT_SYMBOL(filp_close);

/*
 * Careful here! We test whether the file pointer is NULL before
 * releasing the fd. This ensures that one clone task can't release
 * an fd while another clone is opening it.
 */
1007
SYSCALL_DEFINE1(close, unsigned int, fd)
Linus Torvalds's avatar
Linus Torvalds committed
1008
{
1009
	int retval = __close_fd(current->files, fd);
1010 1011 1012 1013 1014 1015 1016 1017 1018

	/* can't restart close syscall because file table entry was cleared */
	if (unlikely(retval == -ERESTARTSYS ||
		     retval == -ERESTARTNOINTR ||
		     retval == -ERESTARTNOHAND ||
		     retval == -ERESTART_RESTARTBLOCK))
		retval = -EINTR;

	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
1019 1020 1021 1022 1023 1024 1025
}
EXPORT_SYMBOL(sys_close);

/*
 * This routine simulates a hangup on the tty, to arrange that users
 * are given clean terminals at login time.
 */
1026
SYSCALL_DEFINE0(vhangup)
Linus Torvalds's avatar
Linus Torvalds committed
1027 1028
{
	if (capable(CAP_SYS_TTY_CONFIG)) {
1029
		tty_vhangup_self();
Linus Torvalds's avatar
Linus Torvalds committed
1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
		return 0;
	}
	return -EPERM;
}

/*
 * Called when an inode is about to be open.
 * We use this to disallow opening large files on 32bit systems if
 * the caller didn't specify O_LARGEFILE.  On 64bit systems we force
 * on this flag in sys_open.
 */
int generic_file_open(struct inode * inode, struct file * filp)
{
	if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
1044
		return -EOVERFLOW;
Linus Torvalds's avatar
Linus Torvalds committed
1045 1046 1047 1048 1049 1050 1051
	return 0;
}

EXPORT_SYMBOL(generic_file_open);

/*
 * This is used by subsystems that don't want seekable
1052 1053 1054
 * file descriptors. The function is not supposed to ever fail, the only
 * reason it returns an 'int' and not 'void' is so that it can be plugged
 * directly into file_operations structure.
Linus Torvalds's avatar
Linus Torvalds committed
1055 1056 1057 1058 1059 1060 1061 1062
 */
int nonseekable_open(struct inode *inode, struct file *filp)
{
	filp->f_mode &= ~(FMODE_LSEEK | FMODE_PREAD | FMODE_PWRITE);
	return 0;
}

EXPORT_SYMBOL(nonseekable_open);