open.c 25.4 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>
Linus Torvalds's avatar
Linus Torvalds committed
33

34 35
#include "internal.h"

36 37
int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
	struct file *filp)
Linus Torvalds's avatar
Linus Torvalds committed
38
{
39
	int ret;
Linus Torvalds's avatar
Linus Torvalds committed
40 41 42 43 44 45 46
	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;
47
	newattrs.ia_valid = ATTR_SIZE | time_attrs;
48 49 50 51
	if (filp) {
		newattrs.ia_file = filp;
		newattrs.ia_valid |= ATTR_FILE;
	}
Linus Torvalds's avatar
Linus Torvalds committed
52

53
	/* Remove suid/sgid on truncate too */
54 55 56
	ret = should_remove_suid(dentry);
	if (ret)
		newattrs.ia_valid |= ret | ATTR_FORCE;
57

58
	mutex_lock(&dentry->d_inode->i_mutex);
59
	ret = notify_change(dentry, &newattrs);
60
	mutex_unlock(&dentry->d_inode->i_mutex);
61
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
62 63
}

64
static long do_sys_truncate(const char __user *pathname, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
65
{
66 67
	struct path path;
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
68 69 70 71 72 73
	int error;

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

74
	error = user_path(pathname, &path);
Linus Torvalds's avatar
Linus Torvalds committed
75 76
	if (error)
		goto out;
77
	inode = path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
78 79 80 81 82 83 84 85 86 87

	/* 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;

88
	error = mnt_want_write(path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
89 90 91
	if (error)
		goto dput_and_out;

92
	error = inode_permission(inode, MAY_WRITE);
93 94
	if (error)
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
95 96

	error = -EPERM;
97
	if (IS_APPEND(inode))
98
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
99

100
	error = get_write_access(inode);
Linus Torvalds's avatar
Linus Torvalds committed
101
	if (error)
102
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
103

104 105 106 107
	/*
	 * Make sure that there are no leases.  get_write_access() protects
	 * against the truncate racing with a lease-granting setlease().
	 */
108
	error = break_lease(inode, O_WRONLY);
Linus Torvalds's avatar
Linus Torvalds committed
109
	if (error)
110
		goto put_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
111 112

	error = locks_verify_truncate(inode, NULL, length);
113
	if (!error)
114
		error = security_path_truncate(&path);
115
	if (!error)
116
		error = do_truncate(path.dentry, length, 0, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
117

118 119
put_write_and_out:
	put_write_access(inode);
120
mnt_drop_write_and_out:
121
	mnt_drop_write(path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
122
dput_and_out:
123
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
124 125 126 127
out:
	return error;
}

128
SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds's avatar
Linus Torvalds committed
129
{
130
	return do_sys_truncate(path, length);
Linus Torvalds's avatar
Linus Torvalds committed
131 132
}

133
static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds's avatar
Linus Torvalds committed
134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151
{
	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;

152
	dentry = file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
153 154 155 156 157 158 159 160 161 162 163 164 165 166
	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;

167
	sb_start_write(inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
168
	error = locks_verify_truncate(inode, file, length);
169
	if (!error)
170
		error = security_path_truncate(&file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
171
	if (!error)
172
		error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, file);
173
	sb_end_write(inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
174 175 176 177 178 179
out_putf:
	fput(file);
out:
	return error;
}

180
SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds's avatar
Linus Torvalds committed
181
{
182
	long ret = do_sys_ftruncate(fd, length, 1);
183
	/* avoid REGPARM breakage on x86: */
184
	asmlinkage_protect(2, ret, fd, length);
185
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
186 187 188 189
}

/* LFS versions of truncate are only needed on 32 bit machines */
#if BITS_PER_LONG == 32
190
SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
191 192 193
{
	return do_sys_truncate(path, length);
}
194 195 196 197 198 199 200
#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
201

202
SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
203
{
204
	long ret = do_sys_ftruncate(fd, length, 0);
205
	/* avoid REGPARM breakage on x86: */
206
	asmlinkage_protect(2, ret, fd, length);
207
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
208
}
209 210 211 212 213 214
#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
215
#endif
216
#endif /* BITS_PER_LONG == 32 */
Linus Torvalds's avatar
Linus Torvalds committed
217

218 219

int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
220
{
221 222
	struct inode *inode = file->f_path.dentry->d_inode;
	long ret;
223 224

	if (offset < 0 || len <= 0)
225
		return -EINVAL;
226 227

	/* Return error if mode is not supported */
228 229 230 231 232 233
	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))
234
		return -EOPNOTSUPP;
235 236

	if (!(file->f_mode & FMODE_WRITE))
237
		return -EBADF;
238 239 240 241 242 243 244 245

	/* 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;

246 247 248 249 250 251
	/*
	 * Revalidate the write permissions, in case security policy has
	 * changed since the files were opened.
	 */
	ret = security_file_permission(file, MAY_WRITE);
	if (ret)
252
		return ret;
253 254

	if (S_ISFIFO(inode->i_mode))
255
		return -ESPIPE;
256 257 258 259 260 261

	/*
	 * Let individual file system decide if it supports preallocation
	 * for directories or not.
	 */
	if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
262
		return -ENODEV;
263 264 265

	/* Check for wrap through zero too */
	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
266
		return -EFBIG;
267

268
	if (!file->f_op->fallocate)
269
		return -EOPNOTSUPP;
270

271 272 273 274
	sb_start_write(inode->i_sb);
	ret = file->f_op->fallocate(file, mode, offset, len);
	sb_end_write(inode->i_sb);
	return ret;
275 276 277 278 279 280 281 282 283 284 285 286 287 288
}

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;
289
}
290

291 292 293 294 295 296 297
#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
298

Linus Torvalds's avatar
Linus Torvalds committed
299 300 301 302 303
/*
 * 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.
 */
304
SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds's avatar
Linus Torvalds committed
305
{
306 307
	const struct cred *old_cred;
	struct cred *override_cred;
308
	struct path path;
309
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
310 311 312 313 314
	int res;

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

315 316 317
	override_cred = prepare_creds();
	if (!override_cred)
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
318

319 320
	override_cred->fsuid = override_cred->uid;
	override_cred->fsgid = override_cred->gid;
Linus Torvalds's avatar
Linus Torvalds committed
321

322
	if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells's avatar
David Howells committed
323
		/* Clear the capabilities if we switch to a non-root user */
324 325
		kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
		if (!uid_eq(override_cred->uid, root_uid))
326
			cap_clear(override_cred->cap_effective);
327
		else
328 329
			override_cred->cap_effective =
				override_cred->cap_permitted;
330
	}
Linus Torvalds's avatar
Linus Torvalds committed
331

332 333
	old_cred = override_creds(override_cred);

334
	res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
335 336 337
	if (res)
		goto out;

338
	inode = path.dentry->d_inode;
339 340

	if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
341 342 343 344 345
		/*
		 * MAY_EXEC on regular files is denied if the fs is mounted
		 * with the "noexec" flag.
		 */
		res = -EACCES;
346
		if (path.mnt->mnt_flags & MNT_NOEXEC)
347 348 349
			goto out_path_release;
	}

350
	res = inode_permission(inode, mode | MAY_ACCESS);
351
	/* SuS v2 requires we report a read only fs too */
352
	if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
353
		goto out_path_release;
354 355 356 357 358 359 360 361 362 363
	/*
	 * 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.
	 */
364
	if (__mnt_is_readonly(path.mnt))
365
		res = -EROFS;
Linus Torvalds's avatar
Linus Torvalds committed
366

367
out_path_release:
368
	path_put(&path);
369
out:
370 371
	revert_creds(old_cred);
	put_cred(override_cred);
Linus Torvalds's avatar
Linus Torvalds committed
372 373 374
	return res;
}

375
SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
376 377 378 379
{
	return sys_faccessat(AT_FDCWD, filename, mode);
}

380
SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
381
{
382
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
383 384
	int error;

385
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
386 387 388
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
389
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
390 391 392
	if (error)
		goto dput_and_out;

393
	set_fs_pwd(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
394 395

dput_and_out:
396
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
397 398 399 400
out:
	return error;
}

401
SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds's avatar
Linus Torvalds committed
402 403 404
{
	struct file *file;
	struct inode *inode;
405
	int error, fput_needed;
Linus Torvalds's avatar
Linus Torvalds committed
406 407

	error = -EBADF;
408
	file = fget_raw_light(fd, &fput_needed);
Linus Torvalds's avatar
Linus Torvalds committed
409 410 411
	if (!file)
		goto out;

412
	inode = file->f_path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
413 414 415 416 417

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

Eric Paris's avatar
Eric Paris committed
418
	error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
419
	if (!error)
420
		set_fs_pwd(current->fs, &file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
421
out_putf:
422
	fput_light(file, fput_needed);
Linus Torvalds's avatar
Linus Torvalds committed
423 424 425 426
out:
	return error;
}

427
SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
428
{
429
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
430 431
	int error;

432
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
433 434 435
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
436
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
437 438 439 440 441 442
	if (error)
		goto dput_and_out;

	error = -EPERM;
	if (!capable(CAP_SYS_CHROOT))
		goto dput_and_out;
443 444 445
	error = security_path_chroot(&path);
	if (error)
		goto dput_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
446

447
	set_fs_root(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
448 449
	error = 0;
dput_and_out:
450
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
451 452 453 454
out:
	return error;
}

455
static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
456
{
457
	struct inode *inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
458
	struct iattr newattrs;
459
	int error;
Linus Torvalds's avatar
Linus Torvalds committed
460

461 462 463
	error = mnt_want_write(path->mnt);
	if (error)
		return error;
464
	mutex_lock(&inode->i_mutex);
465
	error = security_path_chmod(path, mode);
466
	if (error)
467
		goto out_unlock;
Linus Torvalds's avatar
Linus Torvalds committed
468 469
	newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
	newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
470
	error = notify_change(path->dentry, &newattrs);
471
out_unlock:
472
	mutex_unlock(&inode->i_mutex);
473 474 475 476
	mnt_drop_write(path->mnt);
	return error;
}

477
SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
478 479 480 481 482 483 484 485 486 487
{
	struct file * file;
	int err = -EBADF;

	file = fget(fd);
	if (file) {
		audit_inode(NULL, file->f_path.dentry);
		err = chmod_common(&file->f_path, mode);
		fput(file);
	}
Linus Torvalds's avatar
Linus Torvalds committed
488 489 490
	return err;
}

491
SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
492
{
493
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
494 495
	int error;

496
	error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
497 498 499 500
	if (!error) {
		error = chmod_common(&path, mode);
		path_put(&path);
	}
Linus Torvalds's avatar
Linus Torvalds committed
501 502 503
	return error;
}

504
SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
505 506 507 508
{
	return sys_fchmodat(AT_FDCWD, filename, mode);
}

509
static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds's avatar
Linus Torvalds committed
510
{
511
	struct inode *inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
512 513
	int error;
	struct iattr newattrs;
514 515 516 517 518
	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
519 520 521

	newattrs.ia_valid =  ATTR_CTIME;
	if (user != (uid_t) -1) {
522 523
		if (!uid_valid(uid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
524
		newattrs.ia_valid |= ATTR_UID;
525
		newattrs.ia_uid = uid;
Linus Torvalds's avatar
Linus Torvalds committed
526 527
	}
	if (group != (gid_t) -1) {
528 529
		if (!gid_valid(gid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
530
		newattrs.ia_valid |= ATTR_GID;
531
		newattrs.ia_gid = gid;
Linus Torvalds's avatar
Linus Torvalds committed
532 533
	}
	if (!S_ISDIR(inode->i_mode))
534 535
		newattrs.ia_valid |=
			ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
536
	mutex_lock(&inode->i_mutex);
537 538 539
	error = security_path_chown(path, user, group);
	if (!error)
		error = notify_change(path->dentry, &newattrs);
540
	mutex_unlock(&inode->i_mutex);
541

Linus Torvalds's avatar
Linus Torvalds committed
542 543 544
	return error;
}

545 546
SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
		gid_t, group, int, flag)
547
{
548
	struct path path;
549
	int error = -EINVAL;
550
	int lookup_flags;
551

552
	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
553 554
		goto out;

555 556 557 558
	lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
	if (flag & AT_EMPTY_PATH)
		lookup_flags |= LOOKUP_EMPTY;
	error = user_path_at(dfd, filename, lookup_flags, &path);
559 560
	if (error)
		goto out;
561
	error = mnt_want_write(path.mnt);
562 563
	if (error)
		goto out_release;
564
	error = chown_common(&path, user, group);
565
	mnt_drop_write(path.mnt);
566
out_release:
567
	path_put(&path);
568 569 570 571
out:
	return error;
}

572
SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
573
{
574 575
	return sys_fchownat(AT_FDCWD, filename, user, group, 0);
}
Linus Torvalds's avatar
Linus Torvalds committed
576

577 578 579 580
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
581 582
}

583
SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
584 585 586
{
	struct file * file;
	int error = -EBADF;
587
	struct dentry * dentry;
Linus Torvalds's avatar
Linus Torvalds committed
588 589

	file = fget(fd);
590 591 592
	if (!file)
		goto out;

593
	error = mnt_want_write_file(file);
594 595
	if (error)
		goto out_fput;
596
	dentry = file->f_path.dentry;
597
	audit_inode(NULL, dentry);
598
	error = chown_common(&file->f_path, user, group);
Al Viro's avatar
Al Viro committed
599
	mnt_drop_write_file(file);
600
out_fput:
601 602
	fput(file);
out:
Linus Torvalds's avatar
Linus Torvalds committed
603 604 605
	return error;
}

606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627
/*
 * 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()
		 */
628
		error = __mnt_want_write(mnt);
629 630 631 632 633 634
		if (error)
			put_write_access(inode);
	}
	return error;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
635 636 637 638 639 640 641 642 643 644 645 646 647
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;
}

648
static int do_dentry_open(struct file *f,
649 650
			  int (*open)(struct inode *, struct file *),
			  const struct cred *cred)
Linus Torvalds's avatar
Linus Torvalds committed
651
{
652
	static const struct file_operations empty_fops = {};
Linus Torvalds's avatar
Linus Torvalds committed
653 654 655
	struct inode *inode;
	int error;

Al Viro's avatar
Al Viro committed
656
	f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
657
				FMODE_PREAD | FMODE_PWRITE;
658 659 660 661

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

662
	path_get(&f->f_path);
663
	inode = f->f_path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
664
	if (f->f_mode & FMODE_WRITE) {
665
		error = __get_file_write_access(inode, f->f_path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
666 667
		if (error)
			goto cleanup_file;
668 669
		if (!special_file(inode->i_mode))
			file_take_write(f);
Linus Torvalds's avatar
Linus Torvalds committed
670 671 672 673
	}

	f->f_mapping = inode->i_mapping;
	f->f_pos = 0;
Nick Piggin's avatar
Nick Piggin committed
674
	file_sb_list_add(f, inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
675

676 677
	if (unlikely(f->f_mode & FMODE_PATH)) {
		f->f_op = &empty_fops;
678
		return 0;
679 680 681 682
	}

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

683
	error = security_file_open(f, cred);
684 685 686
	if (error)
		goto cleanup_all;

687 688 689 690
	error = break_lease(inode, f->f_flags);
	if (error)
		goto cleanup_all;

691 692 693 694
	if (!open && f->f_op)
		open = f->f_op->open;
	if (open) {
		error = open(inode, f);
Linus Torvalds's avatar
Linus Torvalds committed
695 696 697
		if (error)
			goto cleanup_all;
	}
698 699
	if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
		i_readcount_inc(inode);
700

Linus Torvalds's avatar
Linus Torvalds committed
701 702 703 704
	f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);

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

705
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
706 707 708

cleanup_all:
	fops_put(f->f_op);
709
	file_sb_list_del(f);
710
	if (f->f_mode & FMODE_WRITE) {
Linus Torvalds's avatar
Linus Torvalds committed
711
		put_write_access(inode);
712 713 714 715 716 717 718 719
		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);
720
			mnt_drop_write(f->f_path.mnt);
721
		}
722
	}
Linus Torvalds's avatar
Linus Torvalds committed
723
cleanup_file:
724 725 726
	path_put(&f->f_path);
	f->f_path.mnt = NULL;
	f->f_path.dentry = NULL;
727
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
728 729
}

Miklos Szeredi's avatar
Miklos Szeredi committed
730 731 732 733 734 735 736 737 738 739 740
/**
 * 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
741 742 743
int finish_open(struct file *file, struct dentry *dentry,
		int (*open)(struct inode *, struct file *),
		int *opened)
Miklos Szeredi's avatar
Miklos Szeredi committed
744
{
745
	int error;
Al Viro's avatar
Al Viro committed
746
	BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredi's avatar
Miklos Szeredi committed
747

748
	file->f_path.dentry = dentry;
749
	error = do_dentry_open(file, open, current_cred());
750
	if (!error)
751
		*opened |= FILE_OPENED;
Miklos Szeredi's avatar
Miklos Szeredi committed
752

753
	return error;
Miklos Szeredi's avatar
Miklos Szeredi committed
754 755 756 757 758 759 760 761 762 763 764 765
}
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.
 */
Al Viro's avatar
Al Viro committed
766
int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredi's avatar
Miklos Szeredi committed
767
{
Al Viro's avatar
Al Viro committed
768
	file->f_path.dentry = dentry;
Al Viro's avatar
Al Viro committed
769
	return 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
770 771 772
}
EXPORT_SYMBOL(finish_no_open);

773
struct file *dentry_open(const struct path *path, int flags,
774
			 const struct cred *cred)
775 776 777 778
{
	int error;
	struct file *f;

779 780
	validate_creds(cred);

781
	/* We must always pass in a valid mount pointer. */
782
	BUG_ON(!path->mnt);
783

784 785
	error = -ENFILE;
	f = get_empty_filp();
786
	if (f == NULL)
787 788
		return ERR_PTR(error);

789
	f->f_flags = flags;
790
	f->f_path = *path;
791
	error = do_dentry_open(f, NULL, cred);
792 793 794 795 796 797 798 799 800 801 802
	if (!error) {
		error = open_check_o_direct(f);
		if (error) {
			fput(f);
			f = ERR_PTR(error);
		}
	} else { 
		put_filp(f);
		f = ERR_PTR(error);
	}
	return f;
803
}
Linus Torvalds's avatar
Linus Torvalds committed
804 805
EXPORT_SYMBOL(dentry_open);

806
static void __put_unused_fd(struct files_struct *files, unsigned int fd)
Linus Torvalds's avatar
Linus Torvalds committed
807
{
808
	struct fdtable *fdt = files_fdtable(files);
809
	__clear_open_fd(fd, fdt);
810 811
	if (fd < files->next_fd)
		files->next_fd = fd;
Linus Torvalds's avatar
Linus Torvalds committed
812 813
}

814
void put_unused_fd(unsigned int fd)
Linus Torvalds's avatar
Linus Torvalds committed
815 816 817 818 819 820 821 822 823 824
{
	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);

/*
825
 * Install a file pointer in the fd array.
Linus Torvalds's avatar
Linus Torvalds committed
826 827 828 829 830 831 832 833 834 835 836
 *
 * 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.
 */

837
void fd_install(unsigned int fd, struct file *file)
Linus Torvalds's avatar
Linus Torvalds committed
838 839
{
	struct files_struct *files = current->files;
840
	struct fdtable *fdt;
Linus Torvalds's avatar
Linus Torvalds committed
841
	spin_lock(&files->file_lock);
842
	fdt = files_fdtable(files);
843 844
	BUG_ON(fdt->fd[fd] != NULL);
	rcu_assign_pointer(fdt->fd[fd], file);
Linus Torvalds's avatar
Linus Torvalds committed
845 846 847 848 849
	spin_unlock(&files->file_lock);
}

EXPORT_SYMBOL(fd_install);

850
static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870
{
	int lookup_flags = 0;
	int acc_mode;

	if (!(flags & O_CREAT))
		mode = 0;
	op->mode = mode;

	/* Must never be set by userspace */
	flags &= ~FMODE_NONOTIFY;

	/*
	 * 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;

871 872 873 874 875 876 877 878 879 880
	/*
	 * 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);
	}
881

882
	op->open_flag = flags;
883 884 885 886 887 888 889 890 891 892 893 894

	/* 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;

895 896
	op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;

897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920
	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;
}

/**
 * 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..
 */
921
struct file *filp_open(const char *filename, int flags, umode_t mode)
922 923 924 925 926 927 928
{
	struct open_flags op;
	int lookup = build_open_flags(flags, mode, &op);
	return do_filp_open(AT_FDCWD, filename, &op, lookup);
}
EXPORT_SYMBOL(filp_open);

929 930 931 932 933 934 935 936 937 938 939 940 941 942
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);

943
long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
944
{
945 946
	struct open_flags op;
	int lookup = build_open_flags(flags, mode, &op);
947 948
	char *tmp = getname(filename);
	int fd = PTR_ERR(tmp);
Linus Torvalds's avatar
Linus Torvalds committed
949 950

	if (!IS_ERR(tmp)) {
Ulrich Drepper's avatar
Ulrich Drepper committed
951
		fd = get_unused_fd_flags(flags);
Linus Torvalds's avatar
Linus Torvalds committed
952
		if (fd >= 0) {
953
			struct file *f = do_filp_open(dfd, tmp, &op, lookup);
954 955 956 957
			if (IS_ERR(f)) {
				put_unused_fd(fd);
				fd = PTR_ERR(f);
			} else {
958
				fsnotify_open(f);
959 960
				fd_install(fd, f);
			}
Linus Torvalds's avatar
Linus Torvalds committed
961 962 963 964 965
		}
		putname(tmp);
	}
	return fd;
}
966

967
SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
968
{
969 970
	long ret;

971 972 973
	if (force_o_largefile())
		flags |= O_LARGEFILE;

974 975
	ret = do_sys_open(AT_FDCWD, filename, flags, mode);
	/* avoid REGPARM breakage on x86: */
976
	asmlinkage_protect(3, ret, filename, flags, mode);
977
	return ret;
978
}
Linus Torvalds's avatar
Linus Torvalds committed
979

980
SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
981
		umode_t, mode)
982
{
983 984
	long ret;

985 986 987
	if (force_o_largefile())
		flags |= O_LARGEFILE;

988 989
	ret = do_sys_open(dfd, filename, flags, mode);
	/* avoid REGPARM breakage on x86: */
990
	asmlinkage_protect(4, ret, dfd, filename, flags, mode);