open.c 24.6 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
long vfs_truncate(struct path *path, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
65
{
66
	struct inode *inode;
67
	long error;
Linus Torvalds's avatar
Linus Torvalds committed
68

69
	inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
70 71 72

	/* For directories it's -EISDIR, for other non-regulars - -EINVAL */
	if (S_ISDIR(inode->i_mode))
73
		return -EISDIR;
Linus Torvalds's avatar
Linus Torvalds committed
74
	if (!S_ISREG(inode->i_mode))
75
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
76

77
	error = mnt_want_write(path->mnt);
Linus Torvalds's avatar
Linus Torvalds committed
78
	if (error)
79
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
80

81
	error = inode_permission(inode, MAY_WRITE);
82 83
	if (error)
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
84 85

	error = -EPERM;
86
	if (IS_APPEND(inode))
87
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
88

89
	error = get_write_access(inode);
Linus Torvalds's avatar
Linus Torvalds committed
90
	if (error)
91
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
92

93 94 95 96
	/*
	 * Make sure that there are no leases.  get_write_access() protects
	 * against the truncate racing with a lease-granting setlease().
	 */
97
	error = break_lease(inode, O_WRONLY);
Linus Torvalds's avatar
Linus Torvalds committed
98
	if (error)
99
		goto put_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
100 101

	error = locks_verify_truncate(inode, NULL, length);
102
	if (!error)
103
		error = security_path_truncate(path);
104
	if (!error)
105
		error = do_truncate(path->dentry, length, 0, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
106

107 108
put_write_and_out:
	put_write_access(inode);
109
mnt_drop_write_and_out:
110
	mnt_drop_write(path->mnt);
Linus Torvalds's avatar
Linus Torvalds committed
111 112 113
out:
	return error;
}
114 115 116 117
EXPORT_SYMBOL_GPL(vfs_truncate);

static long do_sys_truncate(const char __user *pathname, loff_t length)
{
118
	unsigned int lookup_flags = LOOKUP_FOLLOW;
119 120 121 122 123 124
	struct path path;
	int error;

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

125 126
retry:
	error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
127 128 129 130
	if (!error) {
		error = vfs_truncate(&path, length);
		path_put(&path);
	}
131 132 133 134
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
135 136
	return error;
}
Linus Torvalds's avatar
Linus Torvalds committed
137

138
SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds's avatar
Linus Torvalds committed
139
{
140
	return do_sys_truncate(path, length);
Linus Torvalds's avatar
Linus Torvalds committed
141 142
}

143
static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds's avatar
Linus Torvalds committed
144
{
Al Viro's avatar
Al Viro committed
145
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
146
	struct dentry *dentry;
147
	struct fd f;
Linus Torvalds's avatar
Linus Torvalds committed
148 149 150 151 152 153
	int error;

	error = -EINVAL;
	if (length < 0)
		goto out;
	error = -EBADF;
154 155
	f = fdget(fd);
	if (!f.file)
Linus Torvalds's avatar
Linus Torvalds committed
156 157 158
		goto out;

	/* explicitly opened as large or we are on 64-bit box */
159
	if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds's avatar
Linus Torvalds committed
160 161
		small = 0;

162
	dentry = f.file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
163 164
	inode = dentry->d_inode;
	error = -EINVAL;
165
	if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds's avatar
Linus Torvalds committed
166 167 168 169 170 171 172 173 174 175 176
		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;

177
	sb_start_write(inode->i_sb);
178
	error = locks_verify_truncate(inode, f.file, length);
179
	if (!error)
180
		error = security_path_truncate(&f.file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
181
	if (!error)
182
		error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
183
	sb_end_write(inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
184
out_putf:
185
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
186 187 188 189
out:
	return error;
}

190
SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds's avatar
Linus Torvalds committed
191
{
192
	long ret = do_sys_ftruncate(fd, length, 1);
193
	/* avoid REGPARM breakage on x86: */
194
	asmlinkage_protect(2, ret, fd, length);
195
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
196 197 198 199
}

/* LFS versions of truncate are only needed on 32 bit machines */
#if BITS_PER_LONG == 32
200
SYSCALL_DEFINE(truncate64)(const char __user * path, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
201 202 203
{
	return do_sys_truncate(path, length);
}
204 205 206 207 208 209 210
#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
211

212
SYSCALL_DEFINE(ftruncate64)(unsigned int fd, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
213
{
214
	long ret = do_sys_ftruncate(fd, length, 0);
215
	/* avoid REGPARM breakage on x86: */
216
	asmlinkage_protect(2, ret, fd, length);
217
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
218
}
219 220 221 222 223 224
#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
225
#endif
226
#endif /* BITS_PER_LONG == 32 */
Linus Torvalds's avatar
Linus Torvalds committed
227

228 229

int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
230
{
231 232
	struct inode *inode = file->f_path.dentry->d_inode;
	long ret;
233 234

	if (offset < 0 || len <= 0)
235
		return -EINVAL;
236 237

	/* Return error if mode is not supported */
238 239 240 241 242 243
	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))
244
		return -EOPNOTSUPP;
245 246

	if (!(file->f_mode & FMODE_WRITE))
247
		return -EBADF;
248 249 250 251 252 253 254 255

	/* 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;

256 257 258 259 260 261
	/*
	 * Revalidate the write permissions, in case security policy has
	 * changed since the files were opened.
	 */
	ret = security_file_permission(file, MAY_WRITE);
	if (ret)
262
		return ret;
263 264

	if (S_ISFIFO(inode->i_mode))
265
		return -ESPIPE;
266 267 268 269 270 271

	/*
	 * Let individual file system decide if it supports preallocation
	 * for directories or not.
	 */
	if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
272
		return -ENODEV;
273 274 275

	/* Check for wrap through zero too */
	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
276
		return -EFBIG;
277

278
	if (!file->f_op->fallocate)
279
		return -EOPNOTSUPP;
280

281 282 283 284
	sb_start_write(inode->i_sb);
	ret = file->f_op->fallocate(file, mode, offset, len);
	sb_end_write(inode->i_sb);
	return ret;
285 286 287 288
}

SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
{
289
	struct fd f = fdget(fd);
290 291
	int error = -EBADF;

292 293 294
	if (f.file) {
		error = do_fallocate(f.file, mode, offset, len);
		fdput(f);
295 296
	}
	return error;
297
}
298

299 300 301 302 303 304 305
#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
306

Linus Torvalds's avatar
Linus Torvalds committed
307 308 309 310 311
/*
 * 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.
 */
312
SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds's avatar
Linus Torvalds committed
313
{
314 315
	const struct cred *old_cred;
	struct cred *override_cred;
316
	struct path path;
317
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
318
	int res;
319
	unsigned int lookup_flags = LOOKUP_FOLLOW;
Linus Torvalds's avatar
Linus Torvalds committed
320 321 322 323

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

324 325 326
	override_cred = prepare_creds();
	if (!override_cred)
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
327

328 329
	override_cred->fsuid = override_cred->uid;
	override_cred->fsgid = override_cred->gid;
Linus Torvalds's avatar
Linus Torvalds committed
330

331
	if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells's avatar
David Howells committed
332
		/* Clear the capabilities if we switch to a non-root user */
333 334
		kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
		if (!uid_eq(override_cred->uid, root_uid))
335
			cap_clear(override_cred->cap_effective);
336
		else
337 338
			override_cred->cap_effective =
				override_cred->cap_permitted;
339
	}
Linus Torvalds's avatar
Linus Torvalds committed
340

341
	old_cred = override_creds(override_cred);
342 343
retry:
	res = user_path_at(dfd, filename, lookup_flags, &path);
344 345 346
	if (res)
		goto out;

347
	inode = path.dentry->d_inode;
348 349

	if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
350 351 352 353 354
		/*
		 * MAY_EXEC on regular files is denied if the fs is mounted
		 * with the "noexec" flag.
		 */
		res = -EACCES;
355
		if (path.mnt->mnt_flags & MNT_NOEXEC)
356 357 358
			goto out_path_release;
	}

359
	res = inode_permission(inode, mode | MAY_ACCESS);
360
	/* SuS v2 requires we report a read only fs too */
361
	if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
362
		goto out_path_release;
363 364 365 366 367 368 369 370 371 372
	/*
	 * 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.
	 */
373
	if (__mnt_is_readonly(path.mnt))
374
		res = -EROFS;
Linus Torvalds's avatar
Linus Torvalds committed
375

376
out_path_release:
377
	path_put(&path);
378 379 380 381
	if (retry_estale(res, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
382
out:
383 384
	revert_creds(old_cred);
	put_cred(override_cred);
Linus Torvalds's avatar
Linus Torvalds committed
385 386 387
	return res;
}

388
SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
389 390 391 392
{
	return sys_faccessat(AT_FDCWD, filename, mode);
}

393
SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
394
{
395
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
396 397
	int error;

398
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
399 400 401
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
402
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
403 404 405
	if (error)
		goto dput_and_out;

406
	set_fs_pwd(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
407 408

dput_and_out:
409
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
410 411 412 413
out:
	return error;
}

414
SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds's avatar
Linus Torvalds committed
415
{
416
	struct fd f = fdget_raw(fd);
Linus Torvalds's avatar
Linus Torvalds committed
417
	struct inode *inode;
418
	int error = -EBADF;
Linus Torvalds's avatar
Linus Torvalds committed
419 420

	error = -EBADF;
421
	if (!f.file)
Linus Torvalds's avatar
Linus Torvalds committed
422 423
		goto out;

424
	inode = f.file->f_path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
425 426 427 428 429

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

Eric Paris's avatar
Eric Paris committed
430
	error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
431
	if (!error)
432
		set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
433
out_putf:
434
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
435 436 437 438
out:
	return error;
}

439
SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
440
{
441
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
442 443
	int error;

444
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
445 446 447
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
448
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
449 450 451 452
	if (error)
		goto dput_and_out;

	error = -EPERM;
453
	if (!nsown_capable(CAP_SYS_CHROOT))
Linus Torvalds's avatar
Linus Torvalds committed
454
		goto dput_and_out;
455 456 457
	error = security_path_chroot(&path);
	if (error)
		goto dput_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
458

459
	set_fs_root(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
460 461
	error = 0;
dput_and_out:
462
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
463 464 465 466
out:
	return error;
}

467
static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
468
{
469
	struct inode *inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
470
	struct iattr newattrs;
471
	int error;
Linus Torvalds's avatar
Linus Torvalds committed
472

473 474 475
	error = mnt_want_write(path->mnt);
	if (error)
		return error;
476
	mutex_lock(&inode->i_mutex);
477
	error = security_path_chmod(path, mode);
478
	if (error)
479
		goto out_unlock;
Linus Torvalds's avatar
Linus Torvalds committed
480 481
	newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
	newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
482
	error = notify_change(path->dentry, &newattrs);
483
out_unlock:
484
	mutex_unlock(&inode->i_mutex);
485 486 487 488
	mnt_drop_write(path->mnt);
	return error;
}

489
SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
490 491 492 493 494 495
{
	struct file * file;
	int err = -EBADF;

	file = fget(fd);
	if (file) {
496
		audit_inode(NULL, file->f_path.dentry, 0);
497 498 499
		err = chmod_common(&file->f_path, mode);
		fput(file);
	}
Linus Torvalds's avatar
Linus Torvalds committed
500 501 502
	return err;
}

503
SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
504
{
505
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
506 507
	int error;

508
	error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
509 510 511 512
	if (!error) {
		error = chmod_common(&path, mode);
		path_put(&path);
	}
Linus Torvalds's avatar
Linus Torvalds committed
513 514 515
	return error;
}

516
SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
517 518 519 520
{
	return sys_fchmodat(AT_FDCWD, filename, mode);
}

521
static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds's avatar
Linus Torvalds committed
522
{
523
	struct inode *inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
524 525
	int error;
	struct iattr newattrs;
526 527 528 529 530
	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
531 532 533

	newattrs.ia_valid =  ATTR_CTIME;
	if (user != (uid_t) -1) {
534 535
		if (!uid_valid(uid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
536
		newattrs.ia_valid |= ATTR_UID;
537
		newattrs.ia_uid = uid;
Linus Torvalds's avatar
Linus Torvalds committed
538 539
	}
	if (group != (gid_t) -1) {
540 541
		if (!gid_valid(gid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
542
		newattrs.ia_valid |= ATTR_GID;
543
		newattrs.ia_gid = gid;
Linus Torvalds's avatar
Linus Torvalds committed
544 545
	}
	if (!S_ISDIR(inode->i_mode))
546 547
		newattrs.ia_valid |=
			ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
548
	mutex_lock(&inode->i_mutex);
549
	error = security_path_chown(path, uid, gid);
550 551
	if (!error)
		error = notify_change(path->dentry, &newattrs);
552
	mutex_unlock(&inode->i_mutex);
553

Linus Torvalds's avatar
Linus Torvalds committed
554 555 556
	return error;
}

557 558
SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
		gid_t, group, int, flag)
559
{
560
	struct path path;
561
	int error = -EINVAL;
562
	int lookup_flags;
563

564
	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
565 566
		goto out;

567 568 569 570
	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);
571 572
	if (error)
		goto out;
573
	error = mnt_want_write(path.mnt);
574 575
	if (error)
		goto out_release;
576
	error = chown_common(&path, user, group);
577
	mnt_drop_write(path.mnt);
578
out_release:
579
	path_put(&path);
580 581 582 583
out:
	return error;
}

584
SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
585
{
586 587
	return sys_fchownat(AT_FDCWD, filename, user, group, 0);
}
Linus Torvalds's avatar
Linus Torvalds committed
588

589 590 591 592
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
593 594
}

595
SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
596
{
597
	struct fd f = fdget(fd);
Linus Torvalds's avatar
Linus Torvalds committed
598 599
	int error = -EBADF;

600
	if (!f.file)
601 602
		goto out;

603
	error = mnt_want_write_file(f.file);
604 605
	if (error)
		goto out_fput;
606
	audit_inode(NULL, f.file->f_path.dentry, 0);
607 608
	error = chown_common(&f.file->f_path, user, group);
	mnt_drop_write_file(f.file);
609
out_fput:
610
	fdput(f);
611
out:
Linus Torvalds's avatar
Linus Torvalds committed
612 613 614
	return error;
}

615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636
/*
 * 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()
		 */
637
		error = __mnt_want_write(mnt);
638 639 640 641 642 643
		if (error)
			put_write_access(inode);
	}
	return error;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
644 645 646 647 648 649 650 651 652 653 654 655 656
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;
}

657
static int do_dentry_open(struct file *f,
658 659
			  int (*open)(struct inode *, struct file *),
			  const struct cred *cred)
Linus Torvalds's avatar
Linus Torvalds committed
660
{
661
	static const struct file_operations empty_fops = {};
Linus Torvalds's avatar
Linus Torvalds committed
662 663 664
	struct inode *inode;
	int error;

Al Viro's avatar
Al Viro committed
665
	f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
666
				FMODE_PREAD | FMODE_PWRITE;
667 668 669 670

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

671
	path_get(&f->f_path);
672
	inode = f->f_path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
673
	if (f->f_mode & FMODE_WRITE) {
674
		error = __get_file_write_access(inode, f->f_path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
675 676
		if (error)
			goto cleanup_file;
677 678
		if (!special_file(inode->i_mode))
			file_take_write(f);
Linus Torvalds's avatar
Linus Torvalds committed
679 680 681 682
	}

	f->f_mapping = inode->i_mapping;
	f->f_pos = 0;
Nick Piggin's avatar
Nick Piggin committed
683
	file_sb_list_add(f, inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
684

685 686
	if (unlikely(f->f_mode & FMODE_PATH)) {
		f->f_op = &empty_fops;
687
		return 0;
688 689 690 691
	}

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

692
	error = security_file_open(f, cred);
693 694 695
	if (error)
		goto cleanup_all;

696 697 698 699
	error = break_lease(inode, f->f_flags);
	if (error)
		goto cleanup_all;

700 701 702 703
	if (!open && f->f_op)
		open = f->f_op->open;
	if (open) {
		error = open(inode, f);
Linus Torvalds's avatar
Linus Torvalds committed
704 705 706
		if (error)
			goto cleanup_all;
	}
707 708
	if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
		i_readcount_inc(inode);
709

Linus Torvalds's avatar
Linus Torvalds committed
710 711 712 713
	f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);

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

714
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
715 716 717

cleanup_all:
	fops_put(f->f_op);
718
	file_sb_list_del(f);
719
	if (f->f_mode & FMODE_WRITE) {
Linus Torvalds's avatar
Linus Torvalds committed
720
		put_write_access(inode);
721 722 723 724 725 726 727 728
		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);
729
			__mnt_drop_write(f->f_path.mnt);
730
		}
731
	}
Linus Torvalds's avatar
Linus Torvalds committed
732
cleanup_file:
733 734 735
	path_put(&f->f_path);
	f->f_path.mnt = NULL;
	f->f_path.dentry = NULL;
736
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
737 738
}

Miklos Szeredi's avatar
Miklos Szeredi committed
739 740 741 742 743 744 745 746 747 748 749
/**
 * 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
750 751 752
int finish_open(struct file *file, struct dentry *dentry,
		int (*open)(struct inode *, struct file *),
		int *opened)
Miklos Szeredi's avatar
Miklos Szeredi committed
753
{
754
	int error;
Al Viro's avatar
Al Viro committed
755
	BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredi's avatar
Miklos Szeredi committed
756

757
	file->f_path.dentry = dentry;
758
	error = do_dentry_open(file, open, current_cred());
759
	if (!error)
760
		*opened |= FILE_OPENED;
Miklos Szeredi's avatar
Miklos Szeredi committed
761

762
	return error;
Miklos Szeredi's avatar
Miklos Szeredi committed
763 764 765 766 767 768 769 770 771 772 773 774
}
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
775
int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredi's avatar
Miklos Szeredi committed
776
{
Al Viro's avatar
Al Viro committed
777
	file->f_path.dentry = dentry;
Al Viro's avatar
Al Viro committed
778
	return 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
779 780 781
}
EXPORT_SYMBOL(finish_no_open);

782
struct file *dentry_open(const struct path *path, int flags,
783
			 const struct cred *cred)
784 785 786 787
{
	int error;
	struct file *f;

788 789
	validate_creds(cred);

790
	/* We must always pass in a valid mount pointer. */
791
	BUG_ON(!path->mnt);
792

793 794
	error = -ENFILE;
	f = get_empty_filp();
795
	if (f == NULL)
796 797
		return ERR_PTR(error);

798
	f->f_flags = flags;
799
	f->f_path = *path;
800
	error = do_dentry_open(f, NULL, cred);
801 802 803 804 805 806 807 808 809 810 811
	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;
812
}
Linus Torvalds's avatar
Linus Torvalds committed
813 814
EXPORT_SYMBOL(dentry_open);

815
static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
816 817 818 819
{
	int lookup_flags = 0;
	int acc_mode;

820 821 822 823
	if (flags & O_CREAT)
		op->mode = (mode & S_IALLUGO) | S_IFREG;
	else
		op->mode = 0;
824 825

	/* Must never be set by userspace */
Al Viro's avatar
Al Viro committed
826
	flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
827 828 829 830 831 832 833 834 835 836

	/*
	 * 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;

837 838 839 840 841 842 843 844 845 846
	/*
	 * 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);
	}
847

848
	op->open_flag = flags;
849 850 851 852 853 854 855 856 857 858 859 860

	/* 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;

861 862
	op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;

863 864 865 866 867 868 869 870 871 872 873 874 875
	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;
}

876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893
/**
 * file_open_name - open file and return file pointer
 *
 * @name:	struct filename containing path to open
 * @flags:	open flags as per the open(2) second argument
 * @mode:	mode for the new file if O_CREAT is set, else ignored
 *
 * This is the helper to open a file from kernelspace if you really
 * have to.  But in generally you should not do this, so please move
 * along, nothing to see here..
 */
struct file *file_open_name(struct filename *name, int flags, umode_t mode)
{
	struct open_flags op;
	int lookup = build_open_flags(flags, mode, &op);
	return do_filp_open(AT_FDCWD, name, &op, lookup);
}

894 895 896 897 898 899 900 901 902 903 904
/**
 * 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..
 */
905
struct file *filp_open(const char *filename, int flags, umode_t mode)
906
{
907 908
	struct filename name = {.name = filename};
	return file_open_name(&name, flags, mode);
909 910 911
}
EXPORT_SYMBOL(filp_open);

912 913 914 915 916 917 918 919 920 921 922 923 924 925
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);

926
long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
927
{
928 929
	struct open_flags op;
	int lookup = build_open_flags(flags, mode, &op);
930
	struct filename *tmp = getname(filename);
931
	int fd = PTR_ERR(tmp);
Linus Torvalds's avatar
Linus Torvalds committed
932 933

	if (!IS_ERR(tmp)) {
Ulrich Drepper's avatar
Ulrich Drepper committed
934
		fd = get_unused_fd_flags(flags);
Linus Torvalds's avatar
Linus Torvalds committed
935
		if (fd >= 0) {
936
			struct file *f = do_filp_open(dfd, tmp, &op, lookup);
937 938 939 940
			if (IS_ERR(f)) {
				put_unused_fd(fd);
				fd = PTR_ERR(f);
			} else {
941
				fsnotify_open(f);
942 943
				fd_install(fd, f);
			}
Linus Torvalds's avatar
Linus Torvalds committed
944 945 946 947 948
		}
		putname(tmp);
	}
	return fd;
}