open.c 27 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, unsigned long, length)
Linus Torvalds's avatar
Linus Torvalds committed
294 295 296 297 298
{
	/* on 32-bit boxen it will cut the range 2^31--2^32-1 off */
	return do_sys_truncate(path, (long)length);
}

299
static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds's avatar
Linus Torvalds committed
300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317
{
	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;

318
	dentry = file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
	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);
334 335 336
	if (!error)
		error = security_path_truncate(&file->f_path, length,
					       ATTR_MTIME|ATTR_CTIME);
Linus Torvalds's avatar
Linus Torvalds committed
337
	if (!error)
338
		error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
Linus Torvalds's avatar
Linus Torvalds committed
339 340 341 342 343 344
out_putf:
	fput(file);
out:
	return error;
}

345
SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds's avatar
Linus Torvalds committed
346
{
347
	long ret = do_sys_ftruncate(fd, length, 1);
348
	/* avoid REGPARM breakage on x86: */
349
	asmlinkage_protect(2, ret, fd, length);
350
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
351 352 353 354
}

/* LFS versions of truncate are only needed on 32 bit machines */
#if BITS_PER_LONG == 32
355
SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
356 357 358
{
	return do_sys_truncate(path, length);
}
359 360 361 362 363 364 365
#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
366

367
SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
368
{
369
	long ret = do_sys_ftruncate(fd, length, 0);
370
	/* avoid REGPARM breakage on x86: */
371
	asmlinkage_protect(2, ret, fd, length);
372
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
373
}
374 375 376 377 378 379
#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
380
#endif
381
#endif /* BITS_PER_LONG == 32 */
Linus Torvalds's avatar
Linus Torvalds committed
382

383 384

int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
385
{
386 387
	struct inode *inode = file->f_path.dentry->d_inode;
	long ret;
388 389

	if (offset < 0 || len <= 0)
390
		return -EINVAL;
391 392 393

	/* Return error if mode is not supported */
	if (mode && !(mode & FALLOC_FL_KEEP_SIZE))
394
		return -EOPNOTSUPP;
395 396

	if (!(file->f_mode & FMODE_WRITE))
397
		return -EBADF;
398 399 400 401 402 403
	/*
	 * Revalidate the write permissions, in case security policy has
	 * changed since the files were opened.
	 */
	ret = security_file_permission(file, MAY_WRITE);
	if (ret)
404
		return ret;
405 406

	if (S_ISFIFO(inode->i_mode))
407
		return -ESPIPE;
408 409 410 411 412 413

	/*
	 * Let individual file system decide if it supports preallocation
	 * for directories or not.
	 */
	if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
414
		return -ENODEV;
415 416 417

	/* Check for wrap through zero too */
	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
418
		return -EFBIG;
419

420 421
	if (!inode->i_op->fallocate)
		return -EOPNOTSUPP;
422

423 424 425 426 427 428 429 430 431 432 433 434 435 436 437
	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;
438
}
439

440 441 442 443 444 445 446
#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
447

Linus Torvalds's avatar
Linus Torvalds committed
448 449 450 451 452
/*
 * 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.
 */
453
SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds's avatar
Linus Torvalds committed
454
{
455 456
	const struct cred *old_cred;
	struct cred *override_cred;
457
	struct path path;
458
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
459 460 461 462 463
	int res;

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

464 465 466
	override_cred = prepare_creds();
	if (!override_cred)
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
467

468 469
	override_cred->fsuid = override_cred->uid;
	override_cred->fsgid = override_cred->gid;
Linus Torvalds's avatar
Linus Torvalds committed
470

471
	if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells's avatar
David Howells committed
472
		/* Clear the capabilities if we switch to a non-root user */
473 474
		if (override_cred->uid)
			cap_clear(override_cred->cap_effective);
475
		else
476 477
			override_cred->cap_effective =
				override_cred->cap_permitted;
478
	}
Linus Torvalds's avatar
Linus Torvalds committed
479

480 481
	old_cred = override_creds(override_cred);

482
	res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
483 484 485
	if (res)
		goto out;

486
	inode = path.dentry->d_inode;
487 488

	if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
489 490 491 492 493
		/*
		 * MAY_EXEC on regular files is denied if the fs is mounted
		 * with the "noexec" flag.
		 */
		res = -EACCES;
494
		if (path.mnt->mnt_flags & MNT_NOEXEC)
495 496 497
			goto out_path_release;
	}

498
	res = inode_permission(inode, mode | MAY_ACCESS);
499
	/* SuS v2 requires we report a read only fs too */
500
	if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
501
		goto out_path_release;
502 503 504 505 506 507 508 509 510 511
	/*
	 * 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.
	 */
512
	if (__mnt_is_readonly(path.mnt))
513
		res = -EROFS;
Linus Torvalds's avatar
Linus Torvalds committed
514

515
out_path_release:
516
	path_put(&path);
517
out:
518 519
	revert_creds(old_cred);
	put_cred(override_cred);
Linus Torvalds's avatar
Linus Torvalds committed
520 521 522
	return res;
}

523
SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
524 525 526 527
{
	return sys_faccessat(AT_FDCWD, filename, mode);
}

528
SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
529
{
530
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
531 532
	int error;

533
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
534 535 536
	if (error)
		goto out;

537
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds's avatar
Linus Torvalds committed
538 539 540
	if (error)
		goto dput_and_out;

541
	set_fs_pwd(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
542 543

dput_and_out:
544
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
545 546 547 548
out:
	return error;
}

549
SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds's avatar
Linus Torvalds committed
550 551 552 553 554 555 556 557 558 559
{
	struct file *file;
	struct inode *inode;
	int error;

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

560
	inode = file->f_path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
561 562 563 564 565

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

566
	error = inode_permission(inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds's avatar
Linus Torvalds committed
567
	if (!error)
568
		set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
569 570 571 572 573 574
out_putf:
	fput(file);
out:
	return error;
}

575
SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
576
{
577
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
578 579
	int error;

580
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
581 582 583
	if (error)
		goto out;

584
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_ACCESS);
Linus Torvalds's avatar
Linus Torvalds committed
585 586 587 588 589 590 591
	if (error)
		goto dput_and_out;

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

592
	set_fs_root(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
593 594
	error = 0;
dput_and_out:
595
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
596 597 598 599
out:
	return error;
}

600
SYSCALL_DEFINE2(fchmod, unsigned int, fd, mode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
601 602 603 604 605 606 607 608 609 610 611
{
	struct inode * inode;
	struct dentry * dentry;
	struct file * file;
	int err = -EBADF;
	struct iattr newattrs;

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

612
	dentry = file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
613 614
	inode = dentry->d_inode;

615
	audit_inode(NULL, dentry);
616

617
	err = mnt_want_write_file(file);
618
	if (err)
Linus Torvalds's avatar
Linus Torvalds committed
619
		goto out_putf;
620
	mutex_lock(&inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
621 622 623 624 625
	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);
626
	mutex_unlock(&inode->i_mutex);
627
	mnt_drop_write(file->f_path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
628 629 630 631 632 633
out_putf:
	fput(file);
out:
	return err;
}

634
SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, mode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
635
{
636 637
	struct path path;
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
638 639 640
	int error;
	struct iattr newattrs;

641
	error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
Linus Torvalds's avatar
Linus Torvalds committed
642 643
	if (error)
		goto out;
644
	inode = path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
645

646
	error = mnt_want_write(path.mnt);
647
	if (error)
Linus Torvalds's avatar
Linus Torvalds committed
648
		goto dput_and_out;
649
	mutex_lock(&inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
650 651 652 653
	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;
654
	error = notify_change(path.dentry, &newattrs);
655
	mutex_unlock(&inode->i_mutex);
656
	mnt_drop_write(path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
657
dput_and_out:
658
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
659 660 661 662
out:
	return error;
}

663
SYSCALL_DEFINE2(chmod, const char __user *, filename, mode_t, mode)
664 665 666 667
{
	return sys_fchmodat(AT_FDCWD, filename, mode);
}

Linus Torvalds's avatar
Linus Torvalds committed
668 669
static int chown_common(struct dentry * dentry, uid_t user, gid_t group)
{
670
	struct inode *inode = dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
671 672 673 674 675 676 677 678 679 680 681 682 683
	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))
684 685
		newattrs.ia_valid |=
			ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
686
	mutex_lock(&inode->i_mutex);
Linus Torvalds's avatar
Linus Torvalds committed
687
	error = notify_change(dentry, &newattrs);
688
	mutex_unlock(&inode->i_mutex);
689

Linus Torvalds's avatar
Linus Torvalds committed
690 691 692
	return error;
}

693
SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
694
{
695
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
696 697
	int error;

698
	error = user_path(filename, &path);
699 700
	if (error)
		goto out;
701
	error = mnt_want_write(path.mnt);
702 703
	if (error)
		goto out_release;
704 705
	error = chown_common(path.dentry, user, group);
	mnt_drop_write(path.mnt);
706
out_release:
707
	path_put(&path);
708
out:
Linus Torvalds's avatar
Linus Torvalds committed
709 710 711
	return error;
}

712 713
SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
		gid_t, group, int, flag)
714
{
715
	struct path path;
716 717 718 719 720 721 722
	int error = -EINVAL;
	int follow;

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

	follow = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
723
	error = user_path_at(dfd, filename, follow, &path);
724 725
	if (error)
		goto out;
726
	error = mnt_want_write(path.mnt);
727 728
	if (error)
		goto out_release;
729 730
	error = chown_common(path.dentry, user, group);
	mnt_drop_write(path.mnt);
731
out_release:
732
	path_put(&path);
733 734 735 736
out:
	return error;
}

737
SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
738
{
739
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
740 741
	int error;

742
	error = user_lpath(filename, &path);
743 744
	if (error)
		goto out;
745
	error = mnt_want_write(path.mnt);
746 747
	if (error)
		goto out_release;
748 749
	error = chown_common(path.dentry, user, group);
	mnt_drop_write(path.mnt);
750
out_release:
751
	path_put(&path);
752
out:
Linus Torvalds's avatar
Linus Torvalds committed
753 754 755
	return error;
}

756
SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
757 758 759
{
	struct file * file;
	int error = -EBADF;
760
	struct dentry * dentry;
Linus Torvalds's avatar
Linus Torvalds committed
761 762

	file = fget(fd);
763 764 765
	if (!file)
		goto out;

766
	error = mnt_want_write_file(file);
767 768
	if (error)
		goto out_fput;
769
	dentry = file->f_path.dentry;
770
	audit_inode(NULL, dentry);
771
	error = chown_common(dentry, user, group);
772 773
	mnt_drop_write(file->f_path.mnt);
out_fput:
774 775
	fput(file);
out:
Linus Torvalds's avatar
Linus Torvalds committed
776 777 778
	return error;
}

779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807
/*
 * 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;
}

808
static struct file *__dentry_open(struct dentry *dentry, struct vfsmount *mnt,
809
					int flags, struct file *f,
810 811
					int (*open)(struct inode *, struct file *),
					const struct cred *cred)
Linus Torvalds's avatar
Linus Torvalds committed
812 813 814 815 816
{
	struct inode *inode;
	int error;

	f->f_flags = flags;
817
	f->f_mode = (__force fmode_t)((flags+1) & O_ACCMODE) | FMODE_LSEEK |
818
				FMODE_PREAD | FMODE_PWRITE;
Linus Torvalds's avatar
Linus Torvalds committed
819 820
	inode = dentry->d_inode;
	if (f->f_mode & FMODE_WRITE) {
821
		error = __get_file_write_access(inode, mnt);
Linus Torvalds's avatar
Linus Torvalds committed
822 823
		if (error)
			goto cleanup_file;
824 825
		if (!special_file(inode->i_mode))
			file_take_write(f);
Linus Torvalds's avatar
Linus Torvalds committed
826 827 828
	}

	f->f_mapping = inode->i_mapping;
829 830
	f->f_path.dentry = dentry;
	f->f_path.mnt = mnt;
Linus Torvalds's avatar
Linus Torvalds committed
831 832 833 834
	f->f_pos = 0;
	f->f_op = fops_get(inode->i_fop);
	file_move(f, &inode->i_sb->s_files);

835
	error = security_dentry_open(f, cred);
836 837 838
	if (error)
		goto cleanup_all;

839 840 841 842
	if (!open && f->f_op)
		open = f->f_op->open;
	if (open) {
		error = open(inode, f);
Linus Torvalds's avatar
Linus Torvalds committed
843 844 845
		if (error)
			goto cleanup_all;
	}
846

Linus Torvalds's avatar
Linus Torvalds committed
847 848 849 850 851 852
	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) {
853 854
		if (!f->f_mapping->a_ops ||
		    ((!f->f_mapping->a_ops->direct_IO) &&
855
		    (!f->f_mapping->a_ops->get_xip_mem))) {
Linus Torvalds's avatar
Linus Torvalds committed
856 857 858 859 860 861 862 863 864
			fput(f);
			f = ERR_PTR(-EINVAL);
		}
	}

	return f;

cleanup_all:
	fops_put(f->f_op);
865
	if (f->f_mode & FMODE_WRITE) {
Linus Torvalds's avatar
Linus Torvalds committed
866
		put_write_access(inode);
867 868 869 870 871 872 873 874
		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);
875
			mnt_drop_write(mnt);
876
		}
877
	}
Linus Torvalds's avatar
Linus Torvalds committed
878
	file_kill(f);
879 880
	f->f_path.dentry = NULL;
	f->f_path.mnt = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
881 882 883 884 885 886 887
cleanup_file:
	put_filp(f);
	dput(dentry);
	mntput(mnt);
	return ERR_PTR(error);
}

888 889 890 891 892 893 894 895 896 897
/**
 * 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.
898 899 900 901
 * 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).
902 903 904 905 906 907 908 909
 * 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 *))
{
910 911
	const struct cred *cred = current_cred();

912 913 914 915
	if (IS_ERR(nd->intent.open.file))
		goto out;
	if (IS_ERR(dentry))
		goto out_err;
916
	nd->intent.open.file = __dentry_open(dget(dentry), mntget(nd->path.mnt),
917 918
					     nd->intent.open.flags - 1,
					     nd->intent.open.file,
919
					     open, cred);
920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937
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)
{
938
	const struct cred *cred = current_cred();
939 940 941 942 943
	struct file *filp;

	/* Pick up the filp from the open intent */
	filp = nd->intent.open.file;
	/* Has the filesystem initialised the file for us? */
944
	if (filp->f_path.dentry == NULL)
945
		filp = __dentry_open(nd->path.dentry, nd->path.mnt, flags, filp,
946
				     NULL, cred);
947
	else
Jan Blunck's avatar
Jan Blunck committed
948
		path_put(&nd->path);
949 950 951
	return filp;
}

952 953 954 955
/*
 * dentry_open() will have done dput(dentry) and mntput(mnt) if it returns an
 * error.
 */
956 957
struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags,
			 const struct cred *cred)
958 959 960 961
{
	int error;
	struct file *f;

962 963 964 965 966 967 968 969 970 971 972 973
	/*
	 * 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);
	}

974 975
	error = -ENFILE;
	f = get_empty_filp();
976 977 978
	if (f == NULL) {
		dput(dentry);
		mntput(mnt);
979
		return ERR_PTR(error);
980
	}
981

982
	return __dentry_open(dentry, mnt, flags, f, NULL, cred);
983
}
Linus Torvalds's avatar
Linus Torvalds committed
984 985
EXPORT_SYMBOL(dentry_open);

986
static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds's avatar
Linus Torvalds committed
987
{
988 989
	struct fdtable *fdt = files_fdtable(files);
	__FD_CLR(fd, fdt->open_fds);
990 991
	if (fd < files->next_fd)
		files->next_fd = fd;
Linus Torvalds's avatar
Linus Torvalds committed
992 993
}

994
void put_unused_fd(unsigned int fd)
Linus Torvalds's avatar
Linus Torvalds committed
995 996 997 998 999 1000 1001 1002 1003 1004
{
	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);

/*
1005
 * Install a file pointer in the fd array.
Linus Torvalds's avatar
Linus Torvalds committed
1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016
 *
 * 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.
 */

1017
void fd_install(unsigned int fd, struct file *file)
Linus Torvalds's avatar
Linus Torvalds committed
1018 1019
{
	struct files_struct *files = current->files;
1020
	struct fdtable *fdt;
Linus Torvalds's avatar
Linus Torvalds committed
1021
	spin_lock(&files->file_lock);
1022
	fdt = files_fdtable(files);
1023 1024
	BUG_ON(fdt->fd[fd] != NULL);
	rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds's avatar
Linus Torvalds committed
1025 1026 1027 1028 1029
	spin_unlock(&files->file_lock);
}

EXPORT_SYMBOL(fd_install);

1030
long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
Linus Torvalds's avatar
Linus Torvalds committed
1031
{
1032 1033
	char *tmp = getname(filename);
	int fd = PTR_ERR(tmp);
Linus Torvalds's avatar
Linus Torvalds committed
1034 1035

	if (!IS_ERR(tmp)) {
Ulrich Drepper's avatar
Ulrich Drepper committed
1036
		fd = get_unused_fd_flags(flags);
Linus Torvalds's avatar
Linus Torvalds committed
1037
		if (fd >= 0) {
1038
			struct file *f = do_filp_open(dfd, tmp, flags, mode, 0);
1039 1040 1041 1042
			if (IS_ERR(f)) {
				put_unused_fd(fd);
				fd = PTR_ERR(f);
			} else {
1043
				fsnotify_open(f->f_path.dentry);
1044 1045
				fd_install(fd, f);
			}
Linus Torvalds's avatar
Linus Torvalds committed
1046 1047 1048 1049 1050
		}
		putname(tmp);
	}
	return fd;
}