open.c 24.3 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
{
Al Viro's avatar
Al Viro committed
135
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
136
	struct dentry *dentry;
137
	struct fd f;
Linus Torvalds's avatar
Linus Torvalds committed
138 139 140 141 142 143
	int error;

	error = -EINVAL;
	if (length < 0)
		goto out;
	error = -EBADF;
144 145
	f = fdget(fd);
	if (!f.file)
Linus Torvalds's avatar
Linus Torvalds committed
146 147 148
		goto out;

	/* explicitly opened as large or we are on 64-bit box */
149
	if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds's avatar
Linus Torvalds committed
150 151
		small = 0;

152
	dentry = f.file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
153 154
	inode = dentry->d_inode;
	error = -EINVAL;
155
	if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds's avatar
Linus Torvalds committed
156 157 158 159 160 161 162 163 164 165 166
		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);
168
	error = locks_verify_truncate(inode, f.file, length);
169
	if (!error)
170
		error = security_path_truncate(&f.file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
171
	if (!error)
172
		error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
173
	sb_end_write(inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
174
out_putf:
175
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
176 177 178 179
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
}

SYSCALL_DEFINE(fallocate)(int fd, int mode, loff_t offset, loff_t len)
{
279
	struct fd f = fdget(fd);
280 281
	int error = -EBADF;

282 283 284
	if (f.file) {
		error = do_fallocate(f.file, mode, offset, len);
		fdput(f);
285 286
	}
	return error;
287
}
288

289 290 291 292 293 294 295
#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
296

Linus Torvalds's avatar
Linus Torvalds committed
297 298 299 300 301
/*
 * 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.
 */
302
SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
Linus Torvalds's avatar
Linus Torvalds committed
303
{
304 305
	const struct cred *old_cred;
	struct cred *override_cred;
306
	struct path path;
307
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
308 309 310 311 312
	int res;

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

313 314 315
	override_cred = prepare_creds();
	if (!override_cred)
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
316

317 318
	override_cred->fsuid = override_cred->uid;
	override_cred->fsgid = override_cred->gid;
Linus Torvalds's avatar
Linus Torvalds committed
319

320
	if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells's avatar
David Howells committed
321
		/* Clear the capabilities if we switch to a non-root user */
322 323
		kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
		if (!uid_eq(override_cred->uid, root_uid))
324
			cap_clear(override_cred->cap_effective);
325
		else
326 327
			override_cred->cap_effective =
				override_cred->cap_permitted;
328
	}
Linus Torvalds's avatar
Linus Torvalds committed
329

330 331
	old_cred = override_creds(override_cred);

332
	res = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
333 334 335
	if (res)
		goto out;

336
	inode = path.dentry->d_inode;
337 338

	if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
339 340 341 342 343
		/*
		 * MAY_EXEC on regular files is denied if the fs is mounted
		 * with the "noexec" flag.
		 */
		res = -EACCES;
344
		if (path.mnt->mnt_flags & MNT_NOEXEC)
345 346 347
			goto out_path_release;
	}

348
	res = inode_permission(inode, mode | MAY_ACCESS);
349
	/* SuS v2 requires we report a read only fs too */
350
	if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
351
		goto out_path_release;
352 353 354 355 356 357 358 359 360 361
	/*
	 * 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.
	 */
362
	if (__mnt_is_readonly(path.mnt))
363
		res = -EROFS;
Linus Torvalds's avatar
Linus Torvalds committed
364

365
out_path_release:
366
	path_put(&path);
367
out:
368 369
	revert_creds(old_cred);
	put_cred(override_cred);
Linus Torvalds's avatar
Linus Torvalds committed
370 371 372
	return res;
}

373
SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
374 375 376 377
{
	return sys_faccessat(AT_FDCWD, filename, mode);
}

378
SYSCALL_DEFINE1(chdir, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
379
{
380
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
381 382
	int error;

383
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
384 385 386
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
387
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
388 389 390
	if (error)
		goto dput_and_out;

391
	set_fs_pwd(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
392 393

dput_and_out:
394
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
395 396 397 398
out:
	return error;
}

399
SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds's avatar
Linus Torvalds committed
400
{
401
	struct fd f = fdget_raw(fd);
Linus Torvalds's avatar
Linus Torvalds committed
402
	struct inode *inode;
403
	int error = -EBADF;
Linus Torvalds's avatar
Linus Torvalds committed
404 405

	error = -EBADF;
406
	if (!f.file)
Linus Torvalds's avatar
Linus Torvalds committed
407 408
		goto out;

409
	inode = f.file->f_path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
410 411 412 413 414

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

Eric Paris's avatar
Eric Paris committed
415
	error = inode_permission(inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
416
	if (!error)
417
		set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
418
out_putf:
419
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
420 421 422 423
out:
	return error;
}

424
SYSCALL_DEFINE1(chroot, const char __user *, filename)
Linus Torvalds's avatar
Linus Torvalds committed
425
{
426
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
427 428
	int error;

429
	error = user_path_dir(filename, &path);
Linus Torvalds's avatar
Linus Torvalds committed
430 431 432
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
433
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
434 435 436 437
	if (error)
		goto dput_and_out;

	error = -EPERM;
438
	if (!nsown_capable(CAP_SYS_CHROOT))
Linus Torvalds's avatar
Linus Torvalds committed
439
		goto dput_and_out;
440 441 442
	error = security_path_chroot(&path);
	if (error)
		goto dput_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
443

444
	set_fs_root(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
445 446
	error = 0;
dput_and_out:
447
	path_put(&path);
Linus Torvalds's avatar
Linus Torvalds committed
448 449 450 451
out:
	return error;
}

452
static int chmod_common(struct path *path, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
453
{
454
	struct inode *inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
455
	struct iattr newattrs;
456
	int error;
Linus Torvalds's avatar
Linus Torvalds committed
457

458 459 460
	error = mnt_want_write(path->mnt);
	if (error)
		return error;
461
	mutex_lock(&inode->i_mutex);
462
	error = security_path_chmod(path, mode);
463
	if (error)
464
		goto out_unlock;
Linus Torvalds's avatar
Linus Torvalds committed
465 466
	newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
	newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
467
	error = notify_change(path->dentry, &newattrs);
468
out_unlock:
469
	mutex_unlock(&inode->i_mutex);
470 471 472 473
	mnt_drop_write(path->mnt);
	return error;
}

474
SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
475 476 477 478 479 480
{
	struct file * file;
	int err = -EBADF;

	file = fget(fd);
	if (file) {
481
		audit_inode(NULL, file->f_path.dentry, 0);
482 483 484
		err = chmod_common(&file->f_path, mode);
		fput(file);
	}
Linus Torvalds's avatar
Linus Torvalds committed
485 486 487
	return err;
}

488
SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename, umode_t, mode)
Linus Torvalds's avatar
Linus Torvalds committed
489
{
490
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
491 492
	int error;

493
	error = user_path_at(dfd, filename, LOOKUP_FOLLOW, &path);
494 495 496 497
	if (!error) {
		error = chmod_common(&path, mode);
		path_put(&path);
	}
Linus Torvalds's avatar
Linus Torvalds committed
498 499 500
	return error;
}

501
SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
502 503 504 505
{
	return sys_fchmodat(AT_FDCWD, filename, mode);
}

506
static int chown_common(struct path *path, uid_t user, gid_t group)
Linus Torvalds's avatar
Linus Torvalds committed
507
{
508
	struct inode *inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
509 510
	int error;
	struct iattr newattrs;
511 512 513 514 515
	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
516 517 518

	newattrs.ia_valid =  ATTR_CTIME;
	if (user != (uid_t) -1) {
519 520
		if (!uid_valid(uid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
521
		newattrs.ia_valid |= ATTR_UID;
522
		newattrs.ia_uid = uid;
Linus Torvalds's avatar
Linus Torvalds committed
523 524
	}
	if (group != (gid_t) -1) {
525 526
		if (!gid_valid(gid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
527
		newattrs.ia_valid |= ATTR_GID;
528
		newattrs.ia_gid = gid;
Linus Torvalds's avatar
Linus Torvalds committed
529 530
	}
	if (!S_ISDIR(inode->i_mode))
531 532
		newattrs.ia_valid |=
			ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
533
	mutex_lock(&inode->i_mutex);
534
	error = security_path_chown(path, uid, gid);
535 536
	if (!error)
		error = notify_change(path->dentry, &newattrs);
537
	mutex_unlock(&inode->i_mutex);
538

Linus Torvalds's avatar
Linus Torvalds committed
539 540 541
	return error;
}

542 543
SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
		gid_t, group, int, flag)
544
{
545
	struct path path;
546
	int error = -EINVAL;
547
	int lookup_flags;
548

549
	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
550 551
		goto out;

552 553 554 555
	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);
556 557
	if (error)
		goto out;
558
	error = mnt_want_write(path.mnt);
559 560
	if (error)
		goto out_release;
561
	error = chown_common(&path, user, group);
562
	mnt_drop_write(path.mnt);
563
out_release:
564
	path_put(&path);
565 566 567 568
out:
	return error;
}

569
SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
570
{
571 572
	return sys_fchownat(AT_FDCWD, filename, user, group, 0);
}
Linus Torvalds's avatar
Linus Torvalds committed
573

574 575 576 577
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
578 579
}

580
SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
581
{
582
	struct fd f = fdget(fd);
Linus Torvalds's avatar
Linus Torvalds committed
583 584
	int error = -EBADF;

585
	if (!f.file)
586 587
		goto out;

588
	error = mnt_want_write_file(f.file);
589 590
	if (error)
		goto out_fput;
591
	audit_inode(NULL, f.file->f_path.dentry, 0);
592 593
	error = chown_common(&f.file->f_path, user, group);
	mnt_drop_write_file(f.file);
594
out_fput:
595
	fdput(f);
596
out:
Linus Torvalds's avatar
Linus Torvalds committed
597 598 599
	return error;
}

600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621
/*
 * 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()
		 */
622
		error = __mnt_want_write(mnt);
623 624 625 626 627 628
		if (error)
			put_write_access(inode);
	}
	return error;
}

Miklos Szeredi's avatar
Miklos Szeredi committed
629 630 631 632 633 634 635 636 637 638 639 640 641
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;
}

642
static int do_dentry_open(struct file *f,
643 644
			  int (*open)(struct inode *, struct file *),
			  const struct cred *cred)
Linus Torvalds's avatar
Linus Torvalds committed
645
{
646
	static const struct file_operations empty_fops = {};
Linus Torvalds's avatar
Linus Torvalds committed
647 648 649
	struct inode *inode;
	int error;

Al Viro's avatar
Al Viro committed
650
	f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
651
				FMODE_PREAD | FMODE_PWRITE;
652 653 654 655

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

656
	path_get(&f->f_path);
657
	inode = f->f_path.dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
658
	if (f->f_mode & FMODE_WRITE) {
659
		error = __get_file_write_access(inode, f->f_path.mnt);
Linus Torvalds's avatar
Linus Torvalds committed
660 661
		if (error)
			goto cleanup_file;
662 663
		if (!special_file(inode->i_mode))
			file_take_write(f);
Linus Torvalds's avatar
Linus Torvalds committed
664 665 666 667
	}

	f->f_mapping = inode->i_mapping;
	f->f_pos = 0;
Nick Piggin's avatar
Nick Piggin committed
668
	file_sb_list_add(f, inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
669

670 671
	if (unlikely(f->f_mode & FMODE_PATH)) {
		f->f_op = &empty_fops;
672
		return 0;
673 674 675 676
	}

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

677
	error = security_file_open(f, cred);
678 679 680
	if (error)
		goto cleanup_all;

681 682 683 684
	error = break_lease(inode, f->f_flags);
	if (error)
		goto cleanup_all;

685 686 687 688
	if (!open && f->f_op)
		open = f->f_op->open;
	if (open) {
		error = open(inode, f);
Linus Torvalds's avatar
Linus Torvalds committed
689 690 691
		if (error)
			goto cleanup_all;
	}
692 693
	if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
		i_readcount_inc(inode);
694

Linus Torvalds's avatar
Linus Torvalds committed
695 696 697 698
	f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);

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

699
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
700 701 702

cleanup_all:
	fops_put(f->f_op);
703
	file_sb_list_del(f);
704
	if (f->f_mode & FMODE_WRITE) {
Linus Torvalds's avatar
Linus Torvalds committed
705
		put_write_access(inode);
706 707 708 709 710 711 712 713
		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);
714
			__mnt_drop_write(f->f_path.mnt);
715
		}
716
	}
Linus Torvalds's avatar
Linus Torvalds committed
717
cleanup_file:
718 719 720
	path_put(&f->f_path);
	f->f_path.mnt = NULL;
	f->f_path.dentry = NULL;
721
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
722 723
}

Miklos Szeredi's avatar
Miklos Szeredi committed
724 725 726 727 728 729 730 731 732 733 734
/**
 * 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
735 736 737
int finish_open(struct file *file, struct dentry *dentry,
		int (*open)(struct inode *, struct file *),
		int *opened)
Miklos Szeredi's avatar
Miklos Szeredi committed
738
{
739
	int error;
Al Viro's avatar
Al Viro committed
740
	BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
Miklos Szeredi's avatar
Miklos Szeredi committed
741

742
	file->f_path.dentry = dentry;
743
	error = do_dentry_open(file, open, current_cred());
744
	if (!error)
745
		*opened |= FILE_OPENED;
Miklos Szeredi's avatar
Miklos Szeredi committed
746

747
	return error;
Miklos Szeredi's avatar
Miklos Szeredi committed
748 749 750 751 752 753 754 755 756 757 758 759
}
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
760
int finish_no_open(struct file *file, struct dentry *dentry)
Miklos Szeredi's avatar
Miklos Szeredi committed
761
{
Al Viro's avatar
Al Viro committed
762
	file->f_path.dentry = dentry;
Al Viro's avatar
Al Viro committed
763
	return 1;
Miklos Szeredi's avatar
Miklos Szeredi committed
764 765 766
}
EXPORT_SYMBOL(finish_no_open);

767
struct file *dentry_open(const struct path *path, int flags,
768
			 const struct cred *cred)
769 770 771 772
{
	int error;
	struct file *f;

773 774
	validate_creds(cred);

775
	/* We must always pass in a valid mount pointer. */
776
	BUG_ON(!path->mnt);
777

778 779
	error = -ENFILE;
	f = get_empty_filp();
780
	if (f == NULL)
781 782
		return ERR_PTR(error);

783
	f->f_flags = flags;
784
	f->f_path = *path;
785
	error = do_dentry_open(f, NULL, cred);
786 787 788 789 790 791 792 793 794 795 796
	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;
797
}
Linus Torvalds's avatar
Linus Torvalds committed
798 799
EXPORT_SYMBOL(dentry_open);

800
static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
801 802 803 804
{
	int lookup_flags = 0;
	int acc_mode;

805 806 807 808
	if (flags & O_CREAT)
		op->mode = (mode & S_IALLUGO) | S_IFREG;
	else
		op->mode = 0;
809 810

	/* Must never be set by userspace */
Al Viro's avatar
Al Viro committed
811
	flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
812 813 814 815 816 817 818 819 820 821

	/*
	 * 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;

822 823 824 825 826 827 828 829 830 831
	/*
	 * 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);
	}
832

833
	op->open_flag = flags;
834 835 836 837 838 839 840 841 842 843 844 845

	/* 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;

846 847
	op->intent = flags & O_PATH ? 0 : LOOKUP_OPEN;

848 849 850 851 852 853 854 855 856 857 858 859 860
	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;
}

861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878
/**
 * 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);
}

879 880 881 882 883 884 885 886 887 888 889
/**
 * 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..
 */
890
struct file *filp_open(const char *filename, int flags, umode_t mode)
891
{
892 893
	struct filename name = {.name = filename};
	return file_open_name(&name, flags, mode);
894 895 896
}
EXPORT_SYMBOL(filp_open);

897 898 899 900 901 902 903 904 905 906 907 908 909 910
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);

911
long do_sys_open(int dfd, const char __user *filename, int flags, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
912
{
913 914
	struct open_flags op;
	int lookup = build_open_flags(flags, mode, &op);
915
	struct filename *tmp = getname(filename);
916
	int fd = PTR_ERR(tmp);
Linus Torvalds's avatar
Linus Torvalds committed
917 918

	if (!IS_ERR(tmp)) {
Ulrich Drepper's avatar
Ulrich Drepper committed
919
		fd = get_unused_fd_flags(flags);
Linus Torvalds's avatar
Linus Torvalds committed
920
		if (fd >= 0) {
921
			struct file *f = do_filp_open(dfd, tmp, &op, lookup);
922 923 924 925
			if (IS_ERR(f)) {
				put_unused_fd(fd);
				fd = PTR_ERR(f);
			} else {
926
				fsnotify_open(f);
927 928
				fd_install(fd, f);
			}
Linus Torvalds's avatar
Linus Torvalds committed
929 930 931 932 933
		}
		putname(tmp);
	}
	return fd;
}
934

935
SYSCALL_DEFINE3(open, const char __user *, filename, int, flags, umode_t, mode)
936
{
937 938
	long ret;

939 940 941
	if (force_o_largefile())
		flags |= O_LARGEFILE;

942 943
	ret = do_sys_open(AT_FDCWD, filename, flags, mode);
	/* avoid REGPARM breakage on x86: */
944
	asmlinkage_protect(3, ret, filename, flags, mode);
945
	return ret;
946
}
Linus Torvalds's avatar
Linus Torvalds committed
947

948
SYSCALL_DEFINE4(openat, int, dfd, const char __user *, filename, int, flags,
949
		umode_t, mode)
950
{