open.c 26.9 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
201 202 203 204 205 206 207 208 209
{
	int err;
	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

216 217 218
	/* Remove suid/sgid on truncate too */
	newattrs.ia_valid |= should_remove_suid(dentry);

219
	mutex_lock(&dentry->d_inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
220
	err = notify_change(dentry, &newattrs);
221
	mutex_unlock(&dentry->d_inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
222 223 224
	return err;
}

225
static long do_sys_truncate(const char __user *pathname, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
226
{
227 228
	struct path path;
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
229 230 231 232 233 234
	int error;

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

235
	error = user_path(pathname, &path);
Linus Torvalds's avatar
Linus Torvalds committed
236 237
	if (error)
		goto out;
238
	inode = path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
239 240 241 242 243 244 245 246 247 248

	/* 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;

249
	error = mnt_want_write(path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
250 251 252
	if (error)
		goto dput_and_out;

253
	error = inode_permission(inode, MAY_WRITE);
254 255
	if (error)
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
256 257

	error = -EPERM;
258
	if (IS_APPEND(inode))
259
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
260

261
	error = get_write_access(inode);
Linus Torvalds's avatar
Linus Torvalds committed
262
	if (error)
263
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
264

265 266 267 268 269
	/*
	 * 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
270
	if (error)
271
		goto put_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
272 273

	error = locks_verify_truncate(inode, NULL, length);
274 275
	if (!error)
		error = security_path_truncate(&path, length, 0);
Linus Torvalds's avatar
Linus Torvalds committed
276
	if (!error) {
277
		vfs_dq_init(inode);
278
		error = do_truncate(path.dentry, length, 0, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
279 280
	}

281 282
put_write_and_out:
	put_write_access(inode);
283
mnt_drop_write_and_out:
284
	mnt_drop_write(path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
285
dput_and_out:
286
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
287 288 289 290
out:
	return error;
}

291
SYSCALL_DEFINE2(truncate, const char __user *, path, unsigned long, length)
Linus Torvalds's avatar
Linus Torvalds committed
292 293 294 295 296
{
	/* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
	return do_sys_truncate(path, (long)length);
}

297
static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds's avatar
Linus Torvalds committed
298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
{
	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;

316
	dentry = file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
317 318 319 320 321 322 323 324 325 326 327 328 329 330 331
	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);
332 333 334
	if (!error)
		error = security_path_truncate(&file->f_path, length,
					       ATTR_MTIME|ATTR_CTIME);
Linus Torvalds's avatar
Linus Torvalds committed
335
	if (!error)
336
		error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds's avatar
Linus Torvalds committed
337 338 339 340 341 342
out_putf:
	fput(file);
out:
	return error;
}

343
SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds's avatar
Linus Torvalds committed
344
{
345
	long ret = do_sys_ftruncate(fd, length, 1);
346
	/* avoid REGPARM breakage on x86: */
347
	asmlinkage_protect(2, ret, fd, length);
348
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
349 350 351 352
}

/* LFS versions of truncate are only needed on 32 bit machines */
#if BITS_PER_LONG == 32
353
SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
354 355 356
{
	return do_sys_truncate(path, length);
}
357 358 359 360 361 362 363
#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
364

365
SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
366
{
367
	long ret = do_sys_ftruncate(fd, length, 0);
368
	/* avoid REGPARM breakage on x86: */
369
	asmlinkage_protect(2, ret, fd, length);
370
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
371
}
372 373 374 375 376 377
#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
378
#endif
379
#endif /* BITS_PER_LONG == 32 */
Linus Torvalds's avatar
Linus Torvalds committed
380

381
SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427
{
	struct file *file;
	struct inode *inode;
	long ret = -EINVAL;

	if (offset < 0 || len <= 0)
		goto out;

	/* Return error if mode is not supported */
	ret = -EOPNOTSUPP;
	if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
		goto out;

	ret = -EBADF;
	file = fget(fd);
	if (!file)
		goto out;
	if (!(file->f_mode & FMODE_WRITE))
		goto out_fput;
	/*
	 * Revalidate the write permissions, in case security policy has
	 * changed since the files were opened.
	 */
	ret = security_file_permission(file, MAY_WRITE);
	if (ret)
		goto out_fput;

	inode = file->f_path.dentry->d_inode;

	ret = -ESPIPE;
	if (S_ISFIFO(inode->i_mode))
		goto out_fput;

	ret = -ENODEV;
	/*
	 * Let individual file system decide if it supports preallocation
	 * for directories or not.
	 */
	if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
		goto out_fput;

	ret = -EFBIG;
	/* Check for wrap through zero too */
	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
		goto out_fput;

Al Viro's avatar
Al Viro committed
428
	if (inode->i_op->fallocate)
429 430
		ret = inode->i_op->fallocate(inode, mode, offset, len);
	else
431
		ret = -EOPNOTSUPP;
432 433 434 435 436 437

out_fput:
	fput(file);
out:
	return ret;
}
438 439 440 441 442 443 444
#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
445

Linus Torvalds's avatar
Linus Torvalds committed
446 447 448 449 450
/*
 * 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.
 */
451
SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds's avatar
Linus Torvalds committed
452
{
453 454
	const struct cred *old_cred;
	struct cred *override_cred;
455
	struct path path;
456
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
457 458 459 460 461
	int res;

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

462 463 464
	override_cred = prepare_creds();
	if (!override_cred)
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
465

466 467
	override_cred->fsuid = override_cred->uid;
	override_cred->fsgid = override_cred->gid;
Linus Torvalds's avatar
Linus Torvalds committed
468

469
	if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells's avatar
David Howells committed
470
		/* Clear the capabilities if we switch to a non-root user */
471 472
		if (override_cred->uid)
			cap_clear(override_cred->cap_effective);
473
		else
474 475
			override_cred->cap_effective =
				override_cred->cap_permitted;
476
	}
Linus Torvalds's avatar
Linus Torvalds committed
477

478 479
	old_cred = override_creds(override_cred);

480
	res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
481 482 483
	if (res)
		goto out;

484
	inode = path.dentry->d_inode;
485 486

	if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
487 488 489 490 491
		/*
		 * MAY_EXEC on regular files is denied if the fs is mounted
		 * with the "noexec" flag.
		 */
		res = -EACCES;
492
		if (path.mnt->mnt_flags & MNT_NOEXEC)
493 494 495
			goto out_path_release;
	}

496
	res = inode_permission(inode, mode | MAY_ACCESS);
497
	/* SuS v2 requires we report a read only fs too */
498
	if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
499
		goto out_path_release;
500 501 502 503 504 505 506 507 508 509
	/*
	 * 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.
	 */
510
	if (__mnt_is_readonly(path.mnt))
511
		res = -EROFS;
Linus Torvalds's avatar
Linus Torvalds committed
512

513
out_path_release:
514
	path_put(&path);
515
out:
516 517
	revert_creds(old_cred);
	put_cred(override_cred);
Linus Torvalds's avatar
Linus Torvalds committed
518 519 520
	return res;
}

521
SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
522 523 524 525
{
	return sys_faccessat(AT_FDCWD, filename, mode);
}

526
SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
527
{
528
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
529 530
	int error;

531
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
532 533 534
	if (error)
		goto out;

535
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds's avatar
Linus Torvalds committed
536 537 538
	if (error)
		goto dput_and_out;

539
	set_fs_pwd(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
540 541

dput_and_out:
542
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
543 544 545 546
out:
	return error;
}

547
SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds's avatar
Linus Torvalds committed
548 549 550 551 552 553 554 555 556 557
{
	struct file *file;
	struct inode *inode;
	int error;

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

558
	inode = file->f_path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
559 560 561 562 563

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

564
	error = inode_permission(inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds's avatar
Linus Torvalds committed
565
	if (!error)
566
		set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
567 568 569 570 571 572
out_putf:
	fput(file);
out:
	return error;
}

573
SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
574
{
575
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
576 577
	int error;

578
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
579 580 581
	if (error)
		goto out;

582
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds's avatar
Linus Torvalds committed
583 584 585 586 587 588 589
	if (error)
		goto dput_and_out;

	error = -EPERM;
	if (!capable(CAP_SYS_CHROOT))
		goto dput_and_out;

590
	set_fs_root(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
591 592
	error = 0;
dput_and_out:
593
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
594 595 596 597
out:
	return error;
}

598
SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
599 600 601 602 603 604 605 606 607 608 609
{
	struct inode * inode;
	struct dentry * dentry;
	struct file * file;
	int err = -EBADF;
	struct iattr newattrs;

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

610
	dentry = file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
611 612
	inode = dentry->d_inode;

613
	audit_inode(NULL, dentry);
614

615
	err = mnt_want_write_file(file);
616
	if (err)
Linus Torvalds's avatar
Linus Torvalds committed
617
		goto out_putf;
618
	mutex_lock(&inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
619 620 621 622 623
	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);
624
	mutex_unlock(&inode->i_mutex);
625
	mnt_drop_write(file->f_path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
626 627 628 629 630 631
out_putf:
	fput(file);
out:
	return err;
}

632
SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
633
{
634 635
	struct path path;
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
636 637 638
	int error;
	struct iattr newattrs;

639
	error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Linus Torvalds's avatar
Linus Torvalds committed
640 641
	if (error)
		goto out;
642
	inode = path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
643

644
	error = mnt_want_write(path.mnt);
645
	if (error)
Linus Torvalds's avatar
Linus Torvalds committed
646
		goto dput_and_out;
647
	mutex_lock(&inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
648 649 650 651
	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;
652
	error = notify_change(path.dentry, &newattrs);
653
	mutex_unlock(&inode->i_mutex);
654
	mnt_drop_write(path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
655
dput_and_out:
656
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
657 658 659 660
out:
	return error;
}

661
SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
662 663 664 665
{
	return sys_fchmodat(AT_FDCWD, filename, mode);
}

Linus Torvalds's avatar
Linus Torvalds committed
666 667
static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
{
668
	struct inode *inode = dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
669 670 671 672 673 674 675 676 677 678 679 680 681
	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))
682 683
		newattrs.ia_valid |=
			ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
684
	mutex_lock(&inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
685
	error = notify_change(dentry, &newattrs);
686
	mutex_unlock(&inode->i_mutex);
687

Linus Torvalds's avatar
Linus Torvalds committed
688 689 690
	return error;
}

691
SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
692
{
693
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
694 695
	int error;

696
	error = user_path(filename, &path);
697 698
	if (error)
		goto out;
699
	error = mnt_want_write(path.mnt);
700 701
	if (error)
		goto out_release;
702 703
	error = chown_common(path.dentry, user, group);
	mnt_drop_write(path.mnt);
704
out_release:
705
	path_put(&path);
706
out:
Linus Torvalds's avatar
Linus Torvalds committed
707 708 709
	return error;
}

710 711
SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
		gid_t, group, int, flag)
712
{
713
	struct path path;
714 715 716 717 718 719 720
	int error = -EINVAL;
	int follow;

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

	follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
721
	error = user_path_at(dfd, filename, follow, &path);
722 723
	if (error)
		goto out;
724
	error = mnt_want_write(path.mnt);
725 726
	if (error)
		goto out_release;
727 728
	error = chown_common(path.dentry, user, group);
	mnt_drop_write(path.mnt);
729
out_release:
730
	path_put(&path);
731 732 733 734
out:
	return error;
}

735
SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
736
{
737
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
738 739
	int error;

740
	error = user_lpath(filename, &path);
741 742
	if (error)
		goto out;
743
	error = mnt_want_write(path.mnt);
744 745
	if (error)
		goto out_release;
746 747
	error = chown_common(path.dentry, user, group);
	mnt_drop_write(path.mnt);
748
out_release:
749
	path_put(&path);
750
out:
Linus Torvalds's avatar
Linus Torvalds committed
751 752 753
	return error;
}

754
SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
755 756 757
{
	struct file * file;
	int error = -EBADF;
758
	struct dentry * dentry;
Linus Torvalds's avatar
Linus Torvalds committed
759 760

	file = fget(fd);
761 762 763
	if (!file)
		goto out;

764
	error = mnt_want_write_file(file);
765 766
	if (error)
		goto out_fput;
767
	dentry = file->f_path.dentry;
768
	audit_inode(NULL, dentry);
769
	error = chown_common(dentry, user, group);
770 771
	mnt_drop_write(file->f_path.mnt);
out_fput:
772 773
	fput(file);
out:
Linus Torvalds's avatar
Linus Torvalds committed
774 775 776
	return error;
}

777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805
/*
 * 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;
}

806
static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
807
					int flags, struct file *f,
808 809
					int (*open)(struct inode *, struct file *),
					const struct cred *cred)
Linus Torvalds's avatar
Linus Torvalds committed
810 811 812 813 814
{
	struct inode *inode;
	int error;

	f->f_flags = flags;
815
	f->f_mode = (__force fmode_t)((flags+1) & O_ACCMODE) | FMODE_LSEEK |
816
				FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds's avatar
Linus Torvalds committed
817 818
	inode = dentry->d_inode;
	if (f->f_mode & FMODE_WRITE) {
819
		error = __get_file_write_access(inode, mnt);
Linus Torvalds's avatar
Linus Torvalds committed
820 821
		if (error)
			goto cleanup_file;
822 823
		if (!special_file(inode->i_mode))
			file_take_write(f);
Linus Torvalds's avatar
Linus Torvalds committed
824 825 826
	}

	f->f_mapping = inode->i_mapping;
827 828
	f->f_path.dentry = dentry;
	f->f_path.mnt = mnt;
Linus Torvalds's avatar
Linus Torvalds committed
829 830 831 832
	f->f_pos = 0;
	f->f_op = fops_get(inode->i_fop);
	file_move(f, &inode->i_sb->s_files);

833
	error = security_dentry_open(f, cred);
834 835 836
	if (error)
		goto cleanup_all;

837 838 839 840
	if (!open && f->f_op)
		open = f->f_op->open;
	if (open) {
		error = open(inode, f);
Linus Torvalds's avatar
Linus Torvalds committed
841 842 843
		if (error)
			goto cleanup_all;
	}
844

Linus Torvalds's avatar
Linus Torvalds committed
845 846 847 848 849 850
	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) {
851 852
		if (!f->f_mapping->a_ops ||
		    ((!f->f_mapping->a_ops->direct_IO) &&
853
		    (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds's avatar
Linus Torvalds committed
854 855 856 857 858 859 860 861 862
			fput(f);
			f = ERR_PTR(-EINVAL);
		}
	}

	return f;

cleanup_all:
	fops_put(f->f_op);
863
	if (f->f_mode & FMODE_WRITE) {
Linus Torvalds's avatar
Linus Torvalds committed
864
		put_write_access(inode);
865 866 867 868 869 870 871 872
		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);
873
			mnt_drop_write(mnt);
874
		}
875
	}
Linus Torvalds's avatar
Linus Torvalds committed
876
	file_kill(f);
877 878
	f->f_path.dentry = NULL;
	f->f_path.mnt = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
879 880 881 882 883 884 885
cleanup_file:
	put_filp(f);
	dput(dentry);
	mntput(mnt);
	return ERR_PTR(error);
}

886 887 888 889 890 891 892 893 894 895
/**
 * 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.
896 897 898 899
 * 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).
900 901 902 903 904 905 906 907
 * 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 *))
{
908 909
	const struct cred *cred = current_cred();

910 911 912 913
	if (IS_ERR(nd->intent.open.file))
		goto out;
	if (IS_ERR(dentry))
		goto out_err;
914
	nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
915 916
					     nd->intent.open.flags - 1,
					     nd->intent.open.file,
917
					     open, cred);
918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935
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)
{
936
	const struct cred *cred = current_cred();
937 938 939 940 941
	struct file *filp;

	/* Pick up the filp from the open intent */
	filp = nd->intent.open.file;
	/* Has the filesystem initialised the file for us? */
942
	if (filp->f_path.dentry == NULL)
943
		filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
944
				     NULL, cred);
945
	else
Jan Blunck's avatar
Jan Blunck committed
946
		path_put(&nd->path);
947 948 949
	return filp;
}

950 951 952 953
/*
 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
 * error.
 */
954 955
struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
			 const struct cred *cred)
956 957 958 959
{
	int error;
	struct file *f;

960 961 962 963 964 965 966 967 968 969 970 971
	/*
	 * 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);
	}

972 973
	error = -ENFILE;
	f = get_empty_filp();
974 975 976
	if (f == NULL) {
		dput(dentry);
		mntput(mnt);
977
		return ERR_PTR(error);
978
	}
979

980
	return __dentry_open(dentry, mnt, flags, f, NULL, cred);
981
}
Linus Torvalds's avatar
Linus Torvalds committed
982 983
EXPORT_SYMBOL(dentry_open);

984
static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds's avatar
Linus Torvalds committed
985
{
986 987
	struct fdtable *fdt = files_fdtable(files);
	__FD_CLR(fd, fdt->open_fds);
988 989
	if (fd < files->next_fd)
		files->next_fd = fd;
Linus Torvalds's avatar
Linus Torvalds committed
990 991
}

992
void put_unused_fd(unsigned int fd)
Linus Torvalds's avatar
Linus Torvalds committed
993 994 995 996 997 998 999 1000 1001 1002
{
	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);

/*
1003
 * Install a file pointer in the fd array.
Linus Torvalds's avatar
Linus Torvalds committed
1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
 *
 * 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.
 */

1015
void fd_install(unsigned int fd, struct file *file)
Linus Torvalds's avatar
Linus Torvalds committed
1016 1017
{
	struct files_struct *files = current->files;
1018
	struct fdtable *fdt;
Linus Torvalds's avatar
Linus Torvalds committed
1019
	spin_lock(&files->file_lock);
1020
	fdt = files_fdtable(files);
1021 1022
	BUG_ON(fdt->fd[fd] != NULL);
	rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds's avatar
Linus Torvalds committed
1023 1024 1025 1026 1027
	spin_unlock(&files->file_lock);
}

EXPORT_SYMBOL(fd_install);

1028
long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds's avatar
Linus Torvalds committed
1029
{
1030 1031
	char *tmp = getname(filename);
	int fd = PTR_ERR(tmp);
Linus Torvalds's avatar
Linus Torvalds committed
1032 1033

	if (!IS_ERR(tmp)) {
Ulrich Drepper's avatar
Ulrich Drepper committed
1034
		fd = get_unused_fd_flags(flags);
Linus Torvalds's avatar
Linus Torvalds committed
1035
		if (fd >= 0) {
1036
			struct file *f = do_filp_open(dfd, tmp, flags, mode, 0);
1037 1038 1039 1040
			if (IS_ERR(f)) {
				put_unused_fd(fd);
				fd = PTR_ERR(f);
			} else {
1041
				fsnotify_open(f->f_path.dentry);
1042 1043
				fd_install(fd, f);
			}
Linus Torvalds's avatar
Linus Torvalds committed
1044 1045 1046 1047 1048
		}
		putname(tmp);
	}
	return fd;
}
1049

1050
SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, int, mode)
1051
{
1052 1053
	long ret;

1054 1055 1056
	if (force_o_largefile())
		flags |= O_LARGEFILE;

1057 1058
	ret = do_sys_open(AT_FDCWD, filename, flags, mode);
	/* avoid REGPARM breakage on x86: */
1059
	asmlinkage_protect(3, ret, filename, flags, mode);
1060
	return ret;
1061
}
Linus Torvalds's avatar
Linus Torvalds committed
1062

1063 1064
SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
		int, mode)
1065
{
1066 1067
	long ret;

1068 1069 1070
	if (force_o_largefile())
		flags |= O_LARGEFILE;

1071 1072
	ret = do_sys_open(dfd, filename, flags, mode);
	/* avoid REGPARM breakage on x86: */
1073
	asmlinkage_protect(4, ret, dfd, filename, flags, mode);
1074
	return ret;