open.c 24.5 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6 7 8 9
/*
 *  linux/fs/open.c
 *
 *  Copyright (C) 1991, 1992  Linus Torvalds
 */

#include <linux/string.h>
#include <linux/mm.h>
#include <linux/file.h>
Al Viro's avatar
Al Viro committed
10
#include <linux/fdtable.h>
Robert Love's avatar
Robert Love committed
11
#include <linux/fsnotify.h>
Linus Torvalds's avatar
Linus Torvalds committed
12 13 14 15
#include <linux/module.h>
#include <linux/tty.h>
#include <linux/namei.h>
#include <linux/backing-dev.h>
16
#include <linux/capability.h>
17
#include <linux/securebits.h>
Linus Torvalds's avatar
Linus Torvalds committed
18 19
#include <linux/security.h>
#include <linux/mount.h>
20
#include <linux/fcntl.h>
21
#include <linux/slab.h>
Linus Torvalds's avatar
Linus Torvalds committed
22 23
#include <asm/uaccess.h>
#include <linux/fs.h>
24
#include <linux/personality.h>
Linus Torvalds's avatar
Linus Torvalds committed
25 26
#include <linux/pagemap.h>
#include <linux/syscalls.h>
27
#include <linux/rcupdate.h>
28
#include <linux/audit.h>
29
#include <linux/falloc.h>
30
#include <linux/fs_struct.h>
31
#include <linux/ima.h>
32
#include <linux/dnotify.h>
Al Viro's avatar
Al Viro committed
33
#include <linux/compat.h>
Linus Torvalds's avatar
Linus Torvalds committed
34

35 36
#include "internal.h"

37 38
int do_truncate(struct dentry *dentry, loff_t length, unsigned int time_attrs,
	struct file *filp)
Linus Torvalds's avatar
Linus Torvalds committed
39
{
40
	int ret;
Linus Torvalds's avatar
Linus Torvalds committed
41 42 43 44 45 46 47
	struct iattr newattrs;

	/* Not pretty: "inode->i_size" shouldn't really be signed. But it is. */
	if (length < 0)
		return -EINVAL;

	newattrs.ia_size = length;
48
	newattrs.ia_valid = ATTR_SIZE | time_attrs;
49 50 51 52
	if (filp) {
		newattrs.ia_file = filp;
		newattrs.ia_valid |= ATTR_FILE;
	}
Linus Torvalds's avatar
Linus Torvalds committed
53

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

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

65
long vfs_truncate(struct path *path, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
66
{
67
	struct inode *inode;
68
	long error;
Linus Torvalds's avatar
Linus Torvalds committed
69

70
	inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
71 72 73

	/* For directories it's -EISDIR, for other non-regulars - -EINVAL */
	if (S_ISDIR(inode->i_mode))
74
		return -EISDIR;
Linus Torvalds's avatar
Linus Torvalds committed
75
	if (!S_ISREG(inode->i_mode))
76
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
77

78
	error = mnt_want_write(path->mnt);
Linus Torvalds's avatar
Linus Torvalds committed
79
	if (error)
80
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
81

82
	error = inode_permission(inode, MAY_WRITE);
83 84
	if (error)
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
85 86

	error = -EPERM;
87
	if (IS_APPEND(inode))
88
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
89

90
	error = get_write_access(inode);
Linus Torvalds's avatar
Linus Torvalds committed
91
	if (error)
92
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
93

94 95 96 97
	/*
	 * Make sure that there are no leases.  get_write_access() protects
	 * against the truncate racing with a lease-granting setlease().
	 */
98
	error = break_lease(inode, O_WRONLY);
Linus Torvalds's avatar
Linus Torvalds committed
99
	if (error)
100
		goto put_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
101 102

	error = locks_verify_truncate(inode, NULL, length);
103
	if (!error)
104
		error = security_path_truncate(path);
105
	if (!error)
106
		error = do_truncate(path->dentry, length, 0, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
107

108 109
put_write_and_out:
	put_write_access(inode);
110
mnt_drop_write_and_out:
111
	mnt_drop_write(path->mnt);
Linus Torvalds's avatar
Linus Torvalds committed
112 113 114
out:
	return error;
}
115 116 117 118
EXPORT_SYMBOL_GPL(vfs_truncate);

static long do_sys_truncate(const char __user *pathname, loff_t length)
{
119
	unsigned int lookup_flags = LOOKUP_FOLLOW;
120 121 122 123 124 125
	struct path path;
	int error;

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

126 127
retry:
	error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
128 129 130 131
	if (!error) {
		error = vfs_truncate(&path, length);
		path_put(&path);
	}
132 133 134 135
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
136 137
	return error;
}
Linus Torvalds's avatar
Linus Torvalds committed
138

139
SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds's avatar
Linus Torvalds committed
140
{
141
	return do_sys_truncate(path, length);
Linus Torvalds's avatar
Linus Torvalds committed
142 143
}

Al Viro's avatar
Al Viro committed
144 145 146 147 148 149 150
#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
{
	return do_sys_truncate(path, length);
}
#endif

151
static long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds's avatar
Linus Torvalds committed
152
{
Al Viro's avatar
Al Viro committed
153
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
154
	struct dentry *dentry;
155
	struct fd f;
Linus Torvalds's avatar
Linus Torvalds committed
156 157 158 159 160 161
	int error;

	error = -EINVAL;
	if (length < 0)
		goto out;
	error = -EBADF;
162 163
	f = fdget(fd);
	if (!f.file)
Linus Torvalds's avatar
Linus Torvalds committed
164 165 166
		goto out;

	/* explicitly opened as large or we are on 64-bit box */
167
	if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds's avatar
Linus Torvalds committed
168 169
		small = 0;

170
	dentry = f.file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
171 172
	inode = dentry->d_inode;
	error = -EINVAL;
173
	if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds's avatar
Linus Torvalds committed
174 175 176 177 178 179 180 181 182 183 184
		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;

185
	sb_start_write(inode->i_sb);
186
	error = locks_verify_truncate(inode, f.file, length);
187
	if (!error)
188
		error = security_path_truncate(&f.file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
189
	if (!error)
190
		error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
191
	sb_end_write(inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
192
out_putf:
193
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
194 195 196 197
out:
	return error;
}

198
SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds's avatar
Linus Torvalds committed
199
{
200
	return do_sys_ftruncate(fd, length, 1);
Linus Torvalds's avatar
Linus Torvalds committed
201 202
}

Al Viro's avatar
Al Viro committed
203 204 205 206 207 208 209
#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(ftruncate, unsigned int, fd, compat_ulong_t, length)
{
	return do_sys_ftruncate(fd, length, 1);
}
#endif

Linus Torvalds's avatar
Linus Torvalds committed
210 211
/* LFS versions of truncate are only needed on 32 bit machines */
#if BITS_PER_LONG == 32
212
SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
Linus Torvalds's avatar
Linus Torvalds committed
213 214 215 216
{
	return do_sys_truncate(path, length);
}

217
SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
Linus Torvalds's avatar
Linus Torvalds committed
218
{
219
	return do_sys_ftruncate(fd, length, 0);
Linus Torvalds's avatar
Linus Torvalds committed
220
}
221
#endif /* BITS_PER_LONG == 32 */
Linus Torvalds's avatar
Linus Torvalds committed
222

223 224

int do_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
225
{
Al Viro's avatar
Al Viro committed
226
	struct inode *inode = file_inode(file);
227
	long ret;
228 229

	if (offset < 0 || len <= 0)
230
		return -EINVAL;
231 232

	/* Return error if mode is not supported */
233 234 235 236 237 238
	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))
239
		return -EOPNOTSUPP;
240 241

	if (!(file->f_mode & FMODE_WRITE))
242
		return -EBADF;
243 244 245 246 247 248 249 250

	/* 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;

251 252 253 254 255 256
	/*
	 * Revalidate the write permissions, in case security policy has
	 * changed since the files were opened.
	 */
	ret = security_file_permission(file, MAY_WRITE);
	if (ret)
257
		return ret;
258 259

	if (S_ISFIFO(inode->i_mode))
260
		return -ESPIPE;
261 262 263 264 265 266

	/*
	 * Let individual file system decide if it supports preallocation
	 * for directories or not.
	 */
	if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
267
		return -ENODEV;
268 269 270

	/* Check for wrap through zero too */
	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
271
		return -EFBIG;
272

273
	if (!file->f_op->fallocate)
274
		return -EOPNOTSUPP;
275

276 277 278 279
	sb_start_write(inode->i_sb);
	ret = file->f_op->fallocate(file, mode, offset, len);
	sb_end_write(inode->i_sb);
	return ret;
280 281
}

282
SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
283
{
284
	struct fd f = fdget(fd);
285 286
	int error = -EBADF;

287 288 289
	if (f.file) {
		error = do_fallocate(f.file, mode, offset, len);
		fdput(f);
290 291
	}
	return error;
292
}
293

Linus Torvalds's avatar
Linus Torvalds committed
294 295 296 297 298
/*
 * 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.
 */
299
SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds's avatar
Linus Torvalds committed
300
{
301 302
	const struct cred *old_cred;
	struct cred *override_cred;
303
	struct path path;
304
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
305
	int res;
306
	unsigned int lookup_flags = LOOKUP_FOLLOW;
Linus Torvalds's avatar
Linus Torvalds committed
307 308 309 310

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

311 312 313
	override_cred = prepare_creds();
	if (!override_cred)
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
314

315 316
	override_cred->fsuid = override_cred->uid;
	override_cred->fsgid = override_cred->gid;
Linus Torvalds's avatar
Linus Torvalds committed
317

318
	if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells's avatar
David Howells committed
319
		/* Clear the capabilities if we switch to a non-root user */
320 321
		kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
		if (!uid_eq(override_cred->uid, root_uid))
322
			cap_clear(override_cred->cap_effective);
323
		else
324 325
			override_cred->cap_effective =
				override_cred->cap_permitted;
326
	}
Linus Torvalds's avatar
Linus Torvalds committed
327

328
	old_cred = override_creds(override_cred);
329 330
retry:
	res = user_path_at(dfd, filename, lookup_flags, &path);
331 332 333
	if (res)
		goto out;

334
	inode = path.dentry->d_inode;
335 336

	if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
337 338 339 340 341
		/*
		 * MAY_EXEC on regular files is denied if the fs is mounted
		 * with the "noexec" flag.
		 */
		res = -EACCES;
342
		if (path.mnt->mnt_flags & MNT_NOEXEC)
343 344 345
			goto out_path_release;
	}

346
	res = inode_permission(inode, mode | MAY_ACCESS);
347
	/* SuS v2 requires we report a read only fs too */
348
	if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
349
		goto out_path_release;
350 351 352 353 354 355 356 357 358 359
	/*
	 * 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.
	 */
360
	if (__mnt_is_readonly(path.mnt))
361
		res = -EROFS;
Linus Torvalds's avatar
Linus Torvalds committed
362

363
out_path_release:
364
	path_put(&path);
365 366 367 368
	if (retry_estale(res, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
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
	int error;
384 385 386
	unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
retry:
	error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds's avatar
Linus Torvalds committed
387 388 389
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
390
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
391 392 393
	if (error)
		goto dput_and_out;

394
	set_fs_pwd(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
395 396

dput_and_out:
397
	path_put(&path);
398 399 400 401
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
Linus Torvalds's avatar
Linus Torvalds committed
402 403 404 405
out:
	return error;
}

406
SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds's avatar
Linus Torvalds committed
407
{
408
	struct fd f = fdget_raw(fd);
Linus Torvalds's avatar
Linus Torvalds committed
409
	struct inode *inode;
410
	int error = -EBADF;
Linus Torvalds's avatar
Linus Torvalds committed
411 412

	error = -EBADF;
413
	if (!f.file)
Linus Torvalds's avatar
Linus Torvalds committed
414 415
		goto out;

Al Viro's avatar
Al Viro committed
416
	inode = file_inode(f.file);
Linus Torvalds's avatar
Linus Torvalds committed
417 418 419 420 421

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

Eric Paris's avatar
Eric Paris committed
422
	error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
423
	if (!error)
424
		set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
425
out_putf:
426
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
427 428 429 430
out:
	return error;
}

431
SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
432
{
433
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
434
	int error;
435 436 437
	unsigned int lookup_flags = LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
retry:
	error = user_path_at(AT_FDCWD, filename, lookup_flags, &path);
Linus Torvalds's avatar
Linus Torvalds committed
438 439 440
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
441
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
442 443 444 445
	if (error)
		goto dput_and_out;

	error = -EPERM;
446
	if (!nsown_capable(CAP_SYS_CHROOT))
Linus Torvalds's avatar
Linus Torvalds committed
447
		goto dput_and_out;
448 449 450
	error = security_path_chroot(&path);
	if (error)
		goto dput_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
451

452
	set_fs_root(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
453 454
	error = 0;
dput_and_out:
455
	path_put(&path);
456 457 458 459
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
Linus Torvalds's avatar
Linus Torvalds committed
460 461 462 463
out:
	return error;
}

464
static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
465
{
466
	struct inode *inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
467
	struct iattr newattrs;
468
	int error;
Linus Torvalds's avatar
Linus Torvalds committed
469

470 471 472
	error = mnt_want_write(path->mnt);
	if (error)
		return error;
473
	mutex_lock(&inode->i_mutex);
474
	error = security_path_chmod(path, mode);
475
	if (error)
476
		goto out_unlock;
Linus Torvalds's avatar
Linus Torvalds committed
477 478
	newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
	newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
479
	error = notify_change(path->dentry, &newattrs);
480
out_unlock:
481
	mutex_unlock(&inode->i_mutex);
482 483 484 485
	mnt_drop_write(path->mnt);
	return error;
}

486
SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
487 488 489 490 491 492
{
	struct file * file;
	int err = -EBADF;

	file = fget(fd);
	if (file) {
493
		audit_inode(NULL, file->f_path.dentry, 0);
494 495 496
		err = chmod_common(&file->f_path, mode);
		fput(file);
	}
Linus Torvalds's avatar
Linus Torvalds committed
497 498 499
	return err;
}

500
SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
501
{
502
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
503
	int error;
504 505 506
	unsigned int lookup_flags = LOOKUP_FOLLOW;
retry:
	error = user_path_at(dfd, filename, lookup_flags, &path);
507 508 509
	if (!error) {
		error = chmod_common(&path, mode);
		path_put(&path);
510 511 512 513
		if (retry_estale(error, lookup_flags)) {
			lookup_flags |= LOOKUP_REVAL;
			goto retry;
		}
514
	}
Linus Torvalds's avatar
Linus Torvalds committed
515 516 517
	return error;
}

518
SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
519 520 521 522
{
	return sys_fchmodat(AT_FDCWD, filename, mode);
}

523
static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds's avatar
Linus Torvalds committed
524
{
525
	struct inode *inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
526 527
	int error;
	struct iattr newattrs;
528 529 530 531 532
	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
533 534 535

	newattrs.ia_valid =  ATTR_CTIME;
	if (user != (uid_t) -1) {
536 537
		if (!uid_valid(uid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
538
		newattrs.ia_valid |= ATTR_UID;
539
		newattrs.ia_uid = uid;
Linus Torvalds's avatar
Linus Torvalds committed
540 541
	}
	if (group != (gid_t) -1) {
542 543
		if (!gid_valid(gid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
544
		newattrs.ia_valid |= ATTR_GID;
545
		newattrs.ia_gid = gid;
Linus Torvalds's avatar
Linus Torvalds committed
546 547
	}
	if (!S_ISDIR(inode->i_mode))
548 549
		newattrs.ia_valid |=
			ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
550
	mutex_lock(&inode->i_mutex);
551
	error = security_path_chown(path, uid, gid);
552 553
	if (!error)
		error = notify_change(path->dentry, &newattrs);
554
	mutex_unlock(&inode->i_mutex);
555

Linus Torvalds's avatar
Linus Torvalds committed
556 557 558
	return error;
}

559 560
SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
		gid_t, group, int, flag)
561
{
562
	struct path path;
563
	int error = -EINVAL;
564
	int lookup_flags;
565

566
	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
567 568
		goto out;

569 570 571
	lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
	if (flag & AT_EMPTY_PATH)
		lookup_flags |= LOOKUP_EMPTY;
572
retry:
573
	error = user_path_at(dfd, filename, lookup_flags, &path);
574 575
	if (error)
		goto out;
576
	error = mnt_want_write(path.mnt);
577 578
	if (error)
		goto out_release;
579
	error = chown_common(&path, user, group);
580
	mnt_drop_write(path.mnt);
581
out_release:
582
	path_put(&path);
583 584 585 586
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
587 588 589 590
out:
	return error;
}

591
SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
592
{
593 594
	return sys_fchownat(AT_FDCWD, filename, user, group, 0);
}
Linus Torvalds's avatar
Linus Torvalds committed
595

596 597 598 599
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
600 601
}

602
SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
603
{
604
	struct fd f = fdget(fd);
Linus Torvalds's avatar
Linus Torvalds committed
605 606
	int error = -EBADF;

607
	if (!f.file)
608 609
		goto out;

610
	error = mnt_want_write_file(f.file);
611 612
	if (error)
		goto out_fput;
613
	audit_inode(NULL, f.file->f_path.dentry, 0);
614 615
	error = chown_common(&f.file->f_path, user, group);
	mnt_drop_write_file(f.file);
616
out_fput:
617
	fdput(f);
618
out:
Linus Torvalds's avatar
Linus Torvalds committed
619 620 621
	return error;
}

622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643
/*
 * 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()
		 */
644
		error = __mnt_want_write(mnt);
645 646 647 648 649 650
		if (error)
			put_write_access(inode);
	}
	return error;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
651 652 653 654 655 656 657 658 659 660 661 662 663
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;
}

664
static int do_dentry_open(struct file *f,
665 666
			  int (*open)(struct inode *, struct file *),
			  const struct cred *cred)
Linus Torvalds's avatar
Linus Torvalds committed
667
{
668
	static const struct file_operations empty_fops = {};
Linus Torvalds's avatar
Linus Torvalds committed
669 670 671
	struct inode *inode;
	int error;

Al Viro's avatar
Al Viro committed
672
	f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
673
				FMODE_PREAD | FMODE_PWRITE;
674 675 676 677

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

678
	path_get(&f->f_path);
679
	inode = f->f_inode = f->f_path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
680
	if (f->f_mode & FMODE_WRITE) {
681
		error = __get_file_write_access(inode, f->f_path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
682 683
		if (error)
			goto cleanup_file;
684 685
		if (!special_file(inode->i_mode))
			file_take_write(f);
Linus Torvalds's avatar
Linus Torvalds committed
686 687 688
	}

	f->f_mapping = inode->i_mapping;
Nick Piggin's avatar
Nick Piggin committed
689
	file_sb_list_add(f, inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
690

691 692
	if (unlikely(f->f_mode & FMODE_PATH)) {
		f->f_op = &empty_fops;
693
		return 0;
694 695 696 697
	}

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

698
	error = security_file_open(f, cred);
699 700 701
	if (error)
		goto cleanup_all;

702 703 704 705
	error = break_lease(inode, f->f_flags);
	if (error)
		goto cleanup_all;

706 707 708 709
	if (!open && f->f_op)
		open = f->f_op->open;
	if (open) {
		error = open(inode, f);
Linus Torvalds's avatar
Linus Torvalds committed
710 711 712
		if (error)
			goto cleanup_all;
	}
713 714
	if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
		i_readcount_inc(inode);
715

Linus Torvalds's avatar
Linus Torvalds committed
716 717 718 719
	f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);

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

720
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
721 722 723

cleanup_all:
	fops_put(f->f_op);
724
	file_sb_list_del(f);
725
	if (f->f_mode & FMODE_WRITE) {
Linus Torvalds's avatar
Linus Torvalds committed
726
		put_write_access(inode);
727 728 729 730 731 732 733 734
		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);
735
			__mnt_drop_write(f->f_path.mnt);
736
		}
737
	}
Linus Torvalds's avatar
Linus Torvalds committed
738
cleanup_file:
739 740 741
	path_put(&f->f_path);
	f->f_path.mnt = NULL;
	f->f_path.dentry = NULL;
742
	f->f_inode = NULL;
743
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
744 745
}

Miklos Szeredi's avatar
Miklos Szeredi committed
746 747 748 749 750 751 752 753 754 755 756
/**
 * 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
757 758 759
int finish_open(struct file *file, struct dentry *dentry,
		int (*open)(struct inode *, struct file *),
		int *opened)
Miklos Szeredi's avatar
Miklos Szeredi committed
760
{
761
	int error;
Al Viro's avatar
Al Viro committed
762
	BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredi's avatar
Miklos Szeredi committed
763

764
	file->f_path.dentry = dentry;
765
	error = do_dentry_open(file, open, current_cred());
766
	if (!error)
767
		*opened |= FILE_OPENED;
Miklos Szeredi's avatar
Miklos Szeredi committed
768

769
	return error;
Miklos Szeredi's avatar
Miklos Szeredi committed
770 771 772 773 774 775 776 777 778 779 780 781
}
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
782
int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredi's avatar
Miklos Szeredi committed
783
{
Al Viro's avatar
Al Viro committed
784
	file->f_path.dentry = dentry;
Al Viro's avatar
Al Viro committed
785
	return 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
786 787 788
}
EXPORT_SYMBOL(finish_no_open);

789
struct file *dentry_open(const struct path *path, int flags,
790
			 const struct cred *cred)
791 792 793 794
{
	int error;
	struct file *f;

795 796
	validate_creds(cred);

797
	/* We must always pass in a valid mount pointer. */
798
	BUG_ON(!path->mnt);
799

800
	f = get_empty_filp();
801 802 803 804 805 806 807 808 809 810 811 812 813
	if (!IS_ERR(f)) {
		f->f_flags = flags;
		f->f_path = *path;
		error = do_dentry_open(f, NULL, cred);
		if (!error) {
			/* from now on we need fput() to dispose of f */
			error = open_check_o_direct(f);
			if (error) {
				fput(f);
				f = ERR_PTR(error);
			}
		} else { 
			put_filp(f);
814 815 816 817
			f = ERR_PTR(error);
		}
	}
	return f;
818
}
Linus Torvalds's avatar
Linus Torvalds committed
819 820
EXPORT_SYMBOL(dentry_open);

821
static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
822 823 824 825
{
	int lookup_flags = 0;
	int acc_mode;

826 827 828 829
	if (flags & O_CREAT)
		op->mode = (mode & S_IALLUGO) | S_IFREG;
	else
		op->mode = 0;
830 831

	/* Must never be set by userspace */
Al Viro's avatar
Al Viro committed
832
	flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
833 834 835 836 837 838 839 840 841 842

	/*
	 * 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;

843 844 845 846 847 848 849 850 851 852
	/*
	 * 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);
	}
853

854
	op->open_flag = flags;
855 856 857 858 859 860 861 862 863 864 865 866

	/* 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;

867 868
	op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;

869 870 871 872 873 874 875 876 877 878 879 880 881
	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;
}

882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899
/**
 * 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);
}

900 901 902 903 904 905 906 907 908 909 910
/**
 * 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..
 */
911
struct file *filp_open(const char *filename, int flags, umode_t mode)
912
{
913 914
	struct filename name = {.name = filename};
	return file_open_name(&name, flags, mode);
915 916 917
}
EXPORT_SYMBOL(filp_open);

918 919 920 921 922 923 924 925 926 927 928 929 930 931
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);

932
long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
933
{
934 935
	struct open_flags op;
	int lookup = build_open_flags(flags, mode, &op);
936
	struct filename *tmp = getname(filename);
937
	int fd = PTR_ERR(tmp);
Linus Torvalds's avatar
Linus Torvalds committed
938 939

	if (!IS_ERR(tmp)) {
Ulrich Drepper's avatar
Ulrich Drepper committed
940
		fd = get_unused_fd_flags(flags);
Linus Torvalds's avatar
Linus Torvalds committed
941
		if (fd >= 0) {
942
			struct file *f = do_filp_open(dfd, tmp, &op, lookup);
943 944 945 946
			if (IS_ERR(f)) {
				put_unused_fd(fd);
				fd = PTR_ERR(f);
			} else {
947
				fsnotify_open(f);
948 949
				fd_install(fd, f);
			}
Linus Torvalds's avatar
Linus Torvalds committed
950 951 952 953 954
		}
		putname(tmp);
	}
	return fd;