open.c 29 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>
22
#include <linux/uaccess.h>
Linus Torvalds's avatar
Linus Torvalds committed
23
#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 55 56 57
	/* Remove suid, sgid, and file capabilities on truncate too */
	ret = dentry_needs_remove_privs(dentry);
	if (ret < 0)
		return ret;
58 59
	if (ret)
		newattrs.ia_valid |= ret | ATTR_FORCE;
60

Al Viro's avatar
Al Viro committed
61
	inode_lock(dentry->d_inode);
62 63
	/* Note any delegations or leases have already been broken: */
	ret = notify_change(dentry, &newattrs, NULL);
Al Viro's avatar
Al Viro committed
64
	inode_unlock(dentry->d_inode);
65
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
66 67
}

Al Viro's avatar
Al Viro committed
68
long vfs_truncate(const struct path *path, loff_t length)
Linus Torvalds's avatar
Linus Torvalds committed
69
{
70
	struct inode *inode;
71
	struct dentry *upperdentry;
72
	long error;
Linus Torvalds's avatar
Linus Torvalds committed
73

74
	inode = path->dentry->d_inode;
Linus Torvalds's avatar
Linus Torvalds committed
75 76 77

	/* For directories it's -EISDIR, for other non-regulars - -EINVAL */
	if (S_ISDIR(inode->i_mode))
78
		return -EISDIR;
Linus Torvalds's avatar
Linus Torvalds committed
79
	if (!S_ISREG(inode->i_mode))
80
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
81

82
	error = mnt_want_write(path->mnt);
Linus Torvalds's avatar
Linus Torvalds committed
83
	if (error)
84
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
85

86
	error = inode_permission(inode, MAY_WRITE);
87 88
	if (error)
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
89 90

	error = -EPERM;
91
	if (IS_APPEND(inode))
92
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
93

94 95 96 97 98
	/*
	 * If this is an overlayfs then do as if opening the file so we get
	 * write access on the upper inode, not on the overlay inode.  For
	 * non-overlay filesystems d_real() is an identity function.
	 */
Miklos Szeredi's avatar
Miklos Szeredi committed
99
	upperdentry = d_real(path->dentry, NULL, O_WRONLY, 0);
100 101 102 103 104
	error = PTR_ERR(upperdentry);
	if (IS_ERR(upperdentry))
		goto mnt_drop_write_and_out;

	error = get_write_access(upperdentry->d_inode);
Linus Torvalds's avatar
Linus Torvalds committed
105
	if (error)
106
		goto mnt_drop_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
107

108 109 110 111
	/*
	 * Make sure that there are no leases.  get_write_access() protects
	 * against the truncate racing with a lease-granting setlease().
	 */
112
	error = break_lease(inode, O_WRONLY);
Linus Torvalds's avatar
Linus Torvalds committed
113
	if (error)
114
		goto put_write_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
115 116

	error = locks_verify_truncate(inode, NULL, length);
117
	if (!error)
118
		error = security_path_truncate(path);
119
	if (!error)
120
		error = do_truncate(path->dentry, length, 0, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
121

122
put_write_and_out:
123
	put_write_access(upperdentry->d_inode);
124
mnt_drop_write_and_out:
125
	mnt_drop_write(path->mnt);
Linus Torvalds's avatar
Linus Torvalds committed
126 127 128
out:
	return error;
}
129 130
EXPORT_SYMBOL_GPL(vfs_truncate);

131
long do_sys_truncate(const char __user *pathname, loff_t length)
132
{
133
	unsigned int lookup_flags = LOOKUP_FOLLOW;
134 135 136 137 138 139
	struct path path;
	int error;

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

140 141
retry:
	error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
142 143 144 145
	if (!error) {
		error = vfs_truncate(&path, length);
		path_put(&path);
	}
146 147 148 149
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
150 151
	return error;
}
Linus Torvalds's avatar
Linus Torvalds committed
152

153
SYSCALL_DEFINE2(truncate, const char __user *, path, long, length)
Linus Torvalds's avatar
Linus Torvalds committed
154
{
155
	return do_sys_truncate(path, length);
Linus Torvalds's avatar
Linus Torvalds committed
156 157
}

Al Viro's avatar
Al Viro committed
158 159 160 161 162 163 164
#ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(truncate, const char __user *, path, compat_off_t, length)
{
	return do_sys_truncate(path, length);
}
#endif

165
long do_sys_ftruncate(unsigned int fd, loff_t length, int small)
Linus Torvalds's avatar
Linus Torvalds committed
166
{
Al Viro's avatar
Al Viro committed
167
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
168
	struct dentry *dentry;
169
	struct fd f;
Linus Torvalds's avatar
Linus Torvalds committed
170 171 172 173 174 175
	int error;

	error = -EINVAL;
	if (length < 0)
		goto out;
	error = -EBADF;
176 177
	f = fdget(fd);
	if (!f.file)
Linus Torvalds's avatar
Linus Torvalds committed
178 179 180
		goto out;

	/* explicitly opened as large or we are on 64-bit box */
181
	if (f.file->f_flags & O_LARGEFILE)
Linus Torvalds's avatar
Linus Torvalds committed
182 183
		small = 0;

184
	dentry = f.file->f_path.dentry;
Linus Torvalds's avatar
Linus Torvalds committed
185 186
	inode = dentry->d_inode;
	error = -EINVAL;
187
	if (!S_ISREG(inode->i_mode) || !(f.file->f_mode & FMODE_WRITE))
Linus Torvalds's avatar
Linus Torvalds committed
188 189 190 191 192 193 194 195
		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;
196 197
	/* Check IS_APPEND on real upper inode */
	if (IS_APPEND(file_inode(f.file)))
Linus Torvalds's avatar
Linus Torvalds committed
198 199
		goto out_putf;

200
	sb_start_write(inode->i_sb);
201
	error = locks_verify_truncate(inode, f.file, length);
202
	if (!error)
203
		error = security_path_truncate(&f.file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
204
	if (!error)
205
		error = do_truncate(dentry, length, ATTR_MTIME|ATTR_CTIME, f.file);
206
	sb_end_write(inode->i_sb);
Linus Torvalds's avatar
Linus Torvalds committed
207
out_putf:
208
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
209 210 211 212
out:
	return error;
}

213
SYSCALL_DEFINE2(ftruncate, unsigned int, fd, unsigned long, length)
Linus Torvalds's avatar
Linus Torvalds committed
214
{
215
	return do_sys_ftruncate(fd, length, 1);
Linus Torvalds's avatar
Linus Torvalds committed
216 217
}

Al Viro's avatar
Al Viro committed
218 219 220 221 222 223 224
#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
225 226
/* LFS versions of truncate are only needed on 32 bit machines */
#if BITS_PER_LONG == 32
227
SYSCALL_DEFINE2(truncate64, const char __user *, path, loff_t, length)
Linus Torvalds's avatar
Linus Torvalds committed
228 229 230 231
{
	return do_sys_truncate(path, length);
}

232
SYSCALL_DEFINE2(ftruncate64, unsigned int, fd, loff_t, length)
Linus Torvalds's avatar
Linus Torvalds committed
233
{
234
	return do_sys_ftruncate(fd, length, 0);
Linus Torvalds's avatar
Linus Torvalds committed
235
}
236
#endif /* BITS_PER_LONG == 32 */
Linus Torvalds's avatar
Linus Torvalds committed
237

238

239
int vfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
240
{
Al Viro's avatar
Al Viro committed
241
	struct inode *inode = file_inode(file);
242
	long ret;
243 244

	if (offset < 0 || len <= 0)
245
		return -EINVAL;
246 247

	/* Return error if mode is not supported */
248
	if (mode & ~FALLOC_FL_SUPPORTED_MASK)
249 250 251 252 253
		return -EOPNOTSUPP;

	/* Punch hole and zero range are mutually exclusive */
	if ((mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE)) ==
	    (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE))
254 255 256 257 258
		return -EOPNOTSUPP;

	/* Punch hole must have keep size set */
	if ((mode & FALLOC_FL_PUNCH_HOLE) &&
	    !(mode & FALLOC_FL_KEEP_SIZE))
259
		return -EOPNOTSUPP;
260

261 262 263 264 265
	/* Collapse range should only be used exclusively. */
	if ((mode & FALLOC_FL_COLLAPSE_RANGE) &&
	    (mode & ~FALLOC_FL_COLLAPSE_RANGE))
		return -EINVAL;

266 267 268 269 270
	/* Insert range should only be used exclusively. */
	if ((mode & FALLOC_FL_INSERT_RANGE) &&
	    (mode & ~FALLOC_FL_INSERT_RANGE))
		return -EINVAL;

271 272 273 274 275
	/* Unshare range should only be used with allocate mode. */
	if ((mode & FALLOC_FL_UNSHARE_RANGE) &&
	    (mode & ~(FALLOC_FL_UNSHARE_RANGE | FALLOC_FL_KEEP_SIZE)))
		return -EINVAL;

276
	if (!(file->f_mode & FMODE_WRITE))
277
		return -EBADF;
278

279
	/*
280
	 * We can only allow pure fallocate on append only files
281
	 */
282
	if ((mode & ~FALLOC_FL_KEEP_SIZE) && IS_APPEND(inode))
283 284 285 286 287
		return -EPERM;

	if (IS_IMMUTABLE(inode))
		return -EPERM;

288
	/*
289
	 * We cannot allow any fallocate operation on an active swapfile
290 291
	 */
	if (IS_SWAPFILE(inode))
292
		return -ETXTBSY;
293

294 295 296 297 298 299
	/*
	 * Revalidate the write permissions, in case security policy has
	 * changed since the files were opened.
	 */
	ret = security_file_permission(file, MAY_WRITE);
	if (ret)
300
		return ret;
301 302

	if (S_ISFIFO(inode->i_mode))
303
		return -ESPIPE;
304

305 306 307 308
	if (S_ISDIR(inode->i_mode))
		return -EISDIR;

	if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
309
		return -ENODEV;
310 311 312

	/* Check for wrap through zero too */
	if (((offset + len) > inode->i_sb->s_maxbytes) || ((offset + len) < 0))
313
		return -EFBIG;
314

315
	if (!file->f_op->fallocate)
316
		return -EOPNOTSUPP;
317

318
	file_start_write(file);
319
	ret = file->f_op->fallocate(file, mode, offset, len);
320 321 322 323 324 325 326 327 328 329 330

	/*
	 * Create inotify and fanotify events.
	 *
	 * To keep the logic simple always create events if fallocate succeeds.
	 * This implies that events are even created if the file size remains
	 * unchanged, e.g. when using flag FALLOC_FL_KEEP_SIZE.
	 */
	if (ret == 0)
		fsnotify_modify(file);

331
	file_end_write(file);
332
	return ret;
333
}
334
EXPORT_SYMBOL_GPL(vfs_fallocate);
335

336
int ksys_fallocate(int fd, int mode, loff_t offset, loff_t len)
337
{
338
	struct fd f = fdget(fd);
339 340
	int error = -EBADF;

341
	if (f.file) {
342
		error = vfs_fallocate(f.file, mode, offset, len);
343
		fdput(f);
344 345
	}
	return error;
346
}
347

348 349 350 351 352
SYSCALL_DEFINE4(fallocate, int, fd, int, mode, loff_t, offset, loff_t, len)
{
	return ksys_fallocate(fd, mode, offset, len);
}

Linus Torvalds's avatar
Linus Torvalds committed
353 354 355 356 357
/*
 * 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.
 */
358
long do_faccessat(int dfd, const char __user *filename, int mode)
Linus Torvalds's avatar
Linus Torvalds committed
359
{
360 361
	const struct cred *old_cred;
	struct cred *override_cred;
362
	struct path path;
363
	struct inode *inode;
Linus Torvalds's avatar
Linus Torvalds committed
364
	int res;
365
	unsigned int lookup_flags = LOOKUP_FOLLOW;
Linus Torvalds's avatar
Linus Torvalds committed
366 367 368 369

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

370 371 372
	override_cred = prepare_creds();
	if (!override_cred)
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
373

374 375
	override_cred->fsuid = override_cred->uid;
	override_cred->fsgid = override_cred->gid;
Linus Torvalds's avatar
Linus Torvalds committed
376

377
	if (!issecure(SECURE_NO_SETUID_FIXUP)) {
David Howells's avatar
David Howells committed
378
		/* Clear the capabilities if we switch to a non-root user */
379 380
		kuid_t root_uid = make_kuid(override_cred->user_ns, 0);
		if (!uid_eq(override_cred->uid, root_uid))
381
			cap_clear(override_cred->cap_effective);
382
		else
383 384
			override_cred->cap_effective =
				override_cred->cap_permitted;
385
	}
Linus Torvalds's avatar
Linus Torvalds committed
386

387
	old_cred = override_creds(override_cred);
388 389
retry:
	res = user_path_at(dfd, filename, lookup_flags, &path);
390 391 392
	if (res)
		goto out;

393
	inode = d_backing_inode(path.dentry);
394 395

	if ((mode & MAY_EXEC) && S_ISREG(inode->i_mode)) {
396 397 398 399 400
		/*
		 * MAY_EXEC on regular files is denied if the fs is mounted
		 * with the "noexec" flag.
		 */
		res = -EACCES;
401
		if (path_noexec(&path))
402 403 404
			goto out_path_release;
	}

405
	res = inode_permission(inode, mode | MAY_ACCESS);
406
	/* SuS v2 requires we report a read only fs too */
407
	if (res || !(mode & S_IWOTH) || special_file(inode->i_mode))
408
		goto out_path_release;
409 410 411 412 413 414 415 416 417 418
	/*
	 * 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.
	 */
419
	if (__mnt_is_readonly(path.mnt))
420
		res = -EROFS;
Linus Torvalds's avatar
Linus Torvalds committed
421

422
out_path_release:
423
	path_put(&path);
424 425 426 427
	if (retry_estale(res, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
428
out:
429 430
	revert_creds(old_cred);
	put_cred(override_cred);
Linus Torvalds's avatar
Linus Torvalds committed
431 432 433
	return res;
}

434 435 436 437 438
SYSCALL_DEFINE3(faccessat, int, dfd, const char __user *, filename, int, mode)
{
	return do_faccessat(dfd, filename, mode);
}

439
SYSCALL_DEFINE2(access, const char __user *, filename, int, mode)
440
{
441
	return do_faccessat(AT_FDCWD, filename, mode);
442 443
}

444
int ksys_chdir(const char __user *filename)
Linus Torvalds's avatar
Linus Torvalds committed
445
{
446
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
447
	int error;
448 449 450
	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
451 452 453
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
454
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
455 456 457
	if (error)
		goto dput_and_out;

458
	set_fs_pwd(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
459 460

dput_and_out:
461
	path_put(&path);
462 463 464 465
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
Linus Torvalds's avatar
Linus Torvalds committed
466 467 468 469
out:
	return error;
}

470 471 472 473 474
SYSCALL_DEFINE1(chdir, const char __user *, filename)
{
	return ksys_chdir(filename);
}

475
SYSCALL_DEFINE1(fchdir, unsigned int, fd)
Linus Torvalds's avatar
Linus Torvalds committed
476
{
477
	struct fd f = fdget_raw(fd);
478
	int error;
Linus Torvalds's avatar
Linus Torvalds committed
479 480

	error = -EBADF;
481
	if (!f.file)
Linus Torvalds's avatar
Linus Torvalds committed
482 483 484
		goto out;

	error = -ENOTDIR;
485
	if (!d_can_lookup(f.file->f_path.dentry))
Linus Torvalds's avatar
Linus Torvalds committed
486 487
		goto out_putf;

488
	error = inode_permission(file_inode(f.file), MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
489
	if (!error)
490
		set_fs_pwd(current->fs, &f.file->f_path);
Linus Torvalds's avatar
Linus Torvalds committed
491
out_putf:
492
	fdput(f);
Linus Torvalds's avatar
Linus Torvalds committed
493 494 495 496
out:
	return error;
}

497
int ksys_chroot(const char __user *filename)
Linus Torvalds's avatar
Linus Torvalds committed
498
{
499
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
500
	int error;
501 502 503
	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
504 505 506
	if (error)
		goto out;

Eric Paris's avatar
Eric Paris committed
507
	error = inode_permission(path.dentry->d_inode, MAY_EXEC | MAY_CHDIR);
Linus Torvalds's avatar
Linus Torvalds committed
508 509 510 511
	if (error)
		goto dput_and_out;

	error = -EPERM;
512
	if (!ns_capable(current_user_ns(), CAP_SYS_CHROOT))
Linus Torvalds's avatar
Linus Torvalds committed
513
		goto dput_and_out;
514 515 516
	error = security_path_chroot(&path);
	if (error)
		goto dput_and_out;
Linus Torvalds's avatar
Linus Torvalds committed
517

518
	set_fs_root(current->fs, &path);
Linus Torvalds's avatar
Linus Torvalds committed
519 520
	error = 0;
dput_and_out:
521
	path_put(&path);
522 523 524 525
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
Linus Torvalds's avatar
Linus Torvalds committed
526 527 528 529
out:
	return error;
}

530 531 532 533 534
SYSCALL_DEFINE1(chroot, const char __user *, filename)
{
	return ksys_chroot(filename);
}

535
static int chmod_common(const struct path *path, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
536
{
537
	struct inode *inode = path->dentry->d_inode;
538
	struct inode *delegated_inode = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
539
	struct iattr newattrs;
540
	int error;
Linus Torvalds's avatar
Linus Torvalds committed
541

542 543 544
	error = mnt_want_write(path->mnt);
	if (error)
		return error;
545
retry_deleg:
Al Viro's avatar
Al Viro committed
546
	inode_lock(inode);
547
	error = security_path_chmod(path, mode);
548
	if (error)
549
		goto out_unlock;
Linus Torvalds's avatar
Linus Torvalds committed
550 551
	newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
	newattrs.ia_valid = ATTR_MODE | ATTR_CTIME;
552
	error = notify_change(path->dentry, &newattrs, &delegated_inode);
553
out_unlock:
Al Viro's avatar
Al Viro committed
554
	inode_unlock(inode);
555 556 557 558 559
	if (delegated_inode) {
		error = break_deleg_wait(&delegated_inode);
		if (!error)
			goto retry_deleg;
	}
560 561 562 563
	mnt_drop_write(path->mnt);
	return error;
}

564
int ksys_fchmod(unsigned int fd, umode_t mode)
565
{
Al Viro's avatar
Al Viro committed
566
	struct fd f = fdget(fd);
567 568
	int err = -EBADF;

Al Viro's avatar
Al Viro committed
569
	if (f.file) {
Al Viro's avatar
Al Viro committed
570
		audit_file(f.file);
Al Viro's avatar
Al Viro committed
571 572
		err = chmod_common(&f.file->f_path, mode);
		fdput(f);
573
	}
Linus Torvalds's avatar
Linus Torvalds committed
574 575 576
	return err;
}

577 578 579 580 581 582
SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
{
	return ksys_fchmod(fd, mode);
}

int do_fchmodat(int dfd, const char __user *filename, umode_t mode)
Linus Torvalds's avatar
Linus Torvalds committed
583
{
584
	struct path path;
Linus Torvalds's avatar
Linus Torvalds committed
585
	int error;
586 587 588
	unsigned int lookup_flags = LOOKUP_FOLLOW;
retry:
	error = user_path_at(dfd, filename, lookup_flags, &path);
589 590 591
	if (!error) {
		error = chmod_common(&path, mode);
		path_put(&path);
592 593 594 595
		if (retry_estale(error, lookup_flags)) {
			lookup_flags |= LOOKUP_REVAL;
			goto retry;
		}
596
	}
Linus Torvalds's avatar
Linus Torvalds committed
597 598 599
	return error;
}

600 601 602 603 604 605
SYSCALL_DEFINE3(fchmodat, int, dfd, const char __user *, filename,
		umode_t, mode)
{
	return do_fchmodat(dfd, filename, mode);
}

606
SYSCALL_DEFINE2(chmod, const char __user *, filename, umode_t, mode)
607
{
608
	return do_fchmodat(AT_FDCWD, filename, mode);
609 610
}

611
static int chown_common(const struct path *path, uid_t user, gid_t group)
Linus Torvalds's avatar
Linus Torvalds committed
612
{
613
	struct inode *inode = path->dentry->d_inode;
614
	struct inode *delegated_inode = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
615 616
	int error;
	struct iattr newattrs;
617 618 619 620 621
	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
622

623
retry_deleg:
Linus Torvalds's avatar
Linus Torvalds committed
624 625
	newattrs.ia_valid =  ATTR_CTIME;
	if (user != (uid_t) -1) {
626 627
		if (!uid_valid(uid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
628
		newattrs.ia_valid |= ATTR_UID;
629
		newattrs.ia_uid = uid;
Linus Torvalds's avatar
Linus Torvalds committed
630 631
	}
	if (group != (gid_t) -1) {
632 633
		if (!gid_valid(gid))
			return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
634
		newattrs.ia_valid |= ATTR_GID;
635
		newattrs.ia_gid = gid;
Linus Torvalds's avatar
Linus Torvalds committed
636 637
	}
	if (!S_ISDIR(inode->i_mode))
638 639
		newattrs.ia_valid |=
			ATTR_KILL_SUID | ATTR_KILL_SGID | ATTR_KILL_PRIV;
Al Viro's avatar
Al Viro committed
640
	inode_lock(inode);
641
	error = security_path_chown(path, uid, gid);
642
	if (!error)
643
		error = notify_change(path->dentry, &newattrs, &delegated_inode);
Al Viro's avatar
Al Viro committed
644
	inode_unlock(inode);
645 646 647 648 649
	if (delegated_inode) {
		error = break_deleg_wait(&delegated_inode);
		if (!error)
			goto retry_deleg;
	}
Linus Torvalds's avatar
Linus Torvalds committed
650 651 652
	return error;
}

653 654
int do_fchownat(int dfd, const char __user *filename, uid_t user, gid_t group,
		int flag)
655
{
656
	struct path path;
657
	int error = -EINVAL;
658
	int lookup_flags;
659

660
	if ((flag & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
661 662
		goto out;

663 664 665
	lookup_flags = (flag & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
	if (flag & AT_EMPTY_PATH)
		lookup_flags |= LOOKUP_EMPTY;
666
retry:
667
	error = user_path_at(dfd, filename, lookup_flags, &path);
668 669
	if (error)
		goto out;
670
	error = mnt_want_write(path.mnt);
671 672
	if (error)
		goto out_release;
673
	error = chown_common(&path, user, group);
674
	mnt_drop_write(path.mnt);
675
out_release:
676
	path_put(&path);
677 678 679 680
	if (retry_estale(error, lookup_flags)) {
		lookup_flags |= LOOKUP_REVAL;
		goto retry;
	}
681 682 683 684
out:
	return error;
}

685 686 687 688 689 690
SYSCALL_DEFINE5(fchownat, int, dfd, const char __user *, filename, uid_t, user,
		gid_t, group, int, flag)
{
	return do_fchownat(dfd, filename, user, group, flag);
}

691
SYSCALL_DEFINE3(chown, const char __user *, filename, uid_t, user, gid_t, group)
Linus Torvalds's avatar
Linus Torvalds committed
692
{
693
	return do_fchownat(AT_FDCWD, filename, user, group, 0);
694
}
Linus Torvalds's avatar
Linus Torvalds committed
695

696 697
SYSCALL_DEFINE3(lchown, const char __user *, filename, uid_t, user, gid_t, group)
{
698 699
	return do_fchownat(AT_FDCWD, filename, user, group,
			   AT_SYMLINK_NOFOLLOW);
Linus Torvalds's avatar
Linus Torvalds committed
700 701
}

702
int ksys_fchown(unsigned int fd, uid_t user, gid_t group)
Linus Torvalds's avatar
Linus Torvalds committed
703
{
704
	struct fd f = fdget(fd);
Linus Torvalds's avatar
Linus Torvalds committed
705 706
	int error = -EBADF;

707
	if (!f.file)
708 709
		goto out;

710
	error = mnt_want_write_file_path(f.file);
711 712
	if (error)
		goto out_fput;
Al Viro's avatar
Al Viro committed
713
	audit_file(f.file);
714
	error = chown_common(&f.file->f_path, user, group);
715
	mnt_drop_write_file_path(f.file);
716
out_fput:
717
	fdput(f);
718
out:
Linus Torvalds's avatar
Linus Torvalds committed
719 720 721
	return error;
}

722 723 724 725 726
SYSCALL_DEFINE3(fchown, unsigned int, fd, uid_t, user, gid_t, group)
{
	return ksys_fchown(fd, user, group);
}

727
static int do_dentry_open(struct file *f,
728
			  struct inode *inode,
729 730
			  int (*open)(struct inode *, struct file *),
			  const struct cred *cred)
Linus Torvalds's avatar
Linus Torvalds committed
731
{
732
	static const struct file_operations empty_fops = {};
Linus Torvalds's avatar
Linus Torvalds committed
733 734
	int error;

Al Viro's avatar
Al Viro committed
735
	f->f_mode = OPEN_FMODE(f->f_flags) | FMODE_LSEEK |
736
				FMODE_PREAD | FMODE_PWRITE;
737

738
	path_get(&f->f_path);
739
	f->f_inode = inode;
Linus Torvalds's avatar
Linus Torvalds committed
740 741
	f->f_mapping = inode->i_mapping;

742 743 744
	/* Ensure that we skip any errors that predate opening of the file */
	f->f_wb_err = filemap_sample_wb_err(f->f_mapping);

Al Viro's avatar
Al Viro committed
745 746
	if (unlikely(f->f_flags & O_PATH)) {
		f->f_mode = FMODE_PATH;
747
		f->f_op = &empty_fops;
748
		goto done;
749 750
	}

751
	if (f->f_mode & FMODE_WRITE && !special_file(inode->i_mode)) {
752
		error = get_write_access(inode);
Al Viro's avatar
Al Viro committed
753
		if (unlikely(error))
Linus Torvalds's avatar
Linus Torvalds committed
754
			goto cleanup_file;
755
		error = __mnt_want_write(f->f_path.mnt);
Al Viro's avatar
Al Viro committed
756
		if (unlikely(error)) {
757 758 759
			put_write_access(inode);
			goto cleanup_file;
		}
760
		f->f_mode |= FMODE_WRITER;
Linus Torvalds's avatar
Linus Torvalds committed
761 762
	}

763
	/* POSIX.1-2008/SUSv4 Section XSI 2.9.7 */
764
	if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))
765 766
		f->f_mode |= FMODE_ATOMIC_POS;

767
	f->f_op = fops_get(inode->i_fop);
Al Viro's avatar
Al Viro committed
768 769 770 771
	if (unlikely(WARN_ON(!f->f_op))) {
		error = -ENODEV;
		goto cleanup_all;
	}
772

773
	error = security_file_open(f, cred);
774 775 776
	if (error)
		goto cleanup_all;

777
	error = break_lease(locks_inode(f), f->f_flags);
778 779 780
	if (error)
		goto cleanup_all;

Al Viro's avatar
Al Viro committed
781
	if (!open)
782 783 784
		open = f->f_op->open;
	if (open) {
		error = open(inode, f);
Linus Torvalds's avatar
Linus Torvalds committed
785 786 787
		if (error)
			goto cleanup_all;
	}
788 789
	if ((f->f_mode & (FMODE_READ | FMODE_WRITE)) == FMODE_READ)
		i_readcount_inc(inode);
790
	if ((f->f_mode & FMODE_READ) &&
Al Viro's avatar
Al Viro committed
791
	     likely(f->f_op->read || f->f_op->read_iter))
792
		f->f_mode |= FMODE_CAN_READ;
793
	if ((f->f_mode & FMODE_WRITE) &&
Al Viro's avatar
Al Viro committed
794
	     likely(f->f_op->write || f->f_op->write_iter))
795
		f->f_mode |= FMODE_CAN_WRITE;
796

797
	f->f_write_hint = WRITE_LIFE_NOT_SET;
Linus Torvalds's avatar
Linus Torvalds committed
798 799 800
	f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC);

	file_ra_state_init(&f->f_ra, f->f_mapping->host->i_mapping);
801 802 803 804 805 806
done:
	/* NB: we're sure to have correct a_ops only after f_op->open */
	error = -EINVAL;
	if ((f->f_flags & O_DIRECT) &&
	    (!f->f_mapping->a_ops || !f->f_mapping->a_ops->direct_IO))
	    	goto out_fput;
807
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
808 809 810

cleanup_all:
	fops_put(f->f_op);
811
	if (f->f_mode & FMODE_WRITER) {
Linus Torvalds's avatar
Linus Torvalds committed
812
		put_write_access(inode);
813
		__mnt_drop_write(f->f_path.mnt);
814
	}
Linus Torvalds's avatar
Linus Torvalds committed
815
cleanup_file:
816 817 818
	path_put(&f->f_path);
	f->f_path.mnt = NULL;
	f->f_path.dentry = NULL;
819
	f->f_inode = NULL;
820
	return error;
821 822 823
out_fput:
    	fput(f);
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
824 825
}

826 827
/**
 * finish_open - finish opening a file
828
 * @file: file pointer
829 830
 * @dentry: pointer to dentry
 * @open: open callback
831
 * @opened: state of open
832 833 834 835 836
 *
 * 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.
837 838 839 840 841 842 843 844 845
 *
 * NB: the dentry reference is _not_ consumed.  If, for example, the dentry is
 * the return value of d_splice_alias(), then the caller needs to perform dput()
 * on it after finish_open().
 *
 * On successful return @file is a fully instantiated open file.  After this, if
 * an error occurs in ->atomic_open(), it needs to clean up with fput().
 *
 * Returns zero on success or -errno if the open failed.
846
 */
Al Viro's avatar
Al Viro committed
847 848 849
int finish_open(struct file *file, struct dentry *dentry,
		int (*open)(struct inode *, struct file *),
		int *opened)
850
{
851
	int error;
Al Viro's avatar
Al Viro committed
852
	BUG_ON(*opened & FILE_OPENED); /* once it's opened, it's opened */
853

854
	file->f_path.dentry = dentry;
855 856
	error = do_dentry_open(file, d_backing_inode(dentry), open,
			       current_cred());
857
	if (!error)
858
		*opened |= FILE_OPENED;
859

860
	return error;
861 862 863 864 865 866
}
EXPORT_SYMBOL(finish_open);

/**
 * finish_no_open - finish ->atomic_open() without opening the file
 *
867
 * @file: file pointer
868 869 870
 * @dentry: dentry or NULL (as returned from ->lookup())
 *
 * This can be used to set the result of a successful lookup in ->atomic_open().
871 872 873 874 875 876
 *
 * NB: unlike finish_open() this function does consume the dentry reference and
 * the caller need not dput() it.
 *
 * Returns "1" which must be the return value of ->atomic_open() after having
 * called this function.
877
 */
Al Viro's avatar
Al Viro committed
878
int finish_no_open(struct file *file, struct dentry *dentry)
879
{
Al Viro's avatar
Al Viro committed
880
	file->f_path.dentry = dentry;
Al Viro's avatar
Al Viro committed
881
	return 1;
882 883 884
}
EXPORT_SYMBOL(finish_no_open);

Miklos Szeredi's avatar
Miklos Szeredi committed
885 886 887 888 889 890
char *file_path(struct file *filp, char *buf, int buflen)
{
	return d_path(&filp->f_path, buf, buflen);
}
EXPORT_SYMBOL(file_path);

891 892 893 894 895 896 897 898 899
/**
 * vfs_open - open the file at the given path
 * @path: path to open
 * @file: newly allocated file with f_flag initialized
 * @cred: credentials to use
 */
int vfs_open(const struct path *path, struct file *file,
	     const struct cred *cred)
{
Miklos Szeredi's avatar
Miklos Szeredi committed
900
	struct dentry *dentry = d_real(path->dentry, NULL, file->f_flags, 0);
901

902 903
	if (IS_ERR(dentry))
		return PTR_ERR(dentry);
904

905
	file->f_path = *path;
906
	return do_dentry_open(file, d_backing_inode(dentry), NULL, cred);
907 908
}

909
struct file *dentry_open(const struct path *path, int flags,
910
			 const struct cred *cred)
911 912 913 914
{
	int error;
	struct file *f;

915 916
	validate_creds(cred);

917
	/* We must always pass in a valid mount pointer. */
918
	BUG_ON(!path->mnt);
919

920
	f = get_empty_filp();
921 922 923 924 925 926 927 928
	if (IS_ERR(f))
		return f;

	f->f_flags = flags;
	error = vfs_open(path, f, cred);
	if (error) {
		put_filp(f);
		return ERR_PTR(error);
929 930
	}
	return f;
931
}
Linus Torvalds's avatar
Linus Torvalds committed
932 933
EXPORT_SYMBOL(dentry_open);

934
static inline int build_open_flags(int flags, umode_t mode, struct open_flags *op)
935 936
{
	int lookup_flags = 0;
937
	int acc_mode = ACC_MODE(flags);
938

939 940 941 942 943 944
	/*
	 * Clear out all open flags we don't know about so that we don't report
	 * them in fcntl(F_GETFD) or similar interfaces.
	 */
	flags &= VALID_OPEN_FLAGS;

945
	if (flags & (O_CREAT | __O_TMPFILE))
946 947 948
		op->mode = (mode & S_IALLUGO) | S_IFREG;
	else
		op->mode = 0;
949 950

	/* Must never be set by userspace */
951
	flags &= ~FMODE_NONOTIFY & ~O_CLOEXEC;
952 953 954 955 956 957 958 959 960 961

	/*
	 * 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;

Al Viro's avatar
Al Viro committed
962 963
	if (flags & __O_TMPFILE) {
		if ((flags & O_TMPFILE_MASK) != O_TMPFILE)
964
			return -EINVAL;
965 966
		if (!(acc_mode & MAY_WRITE))
			return -EINVAL;
967 968 969 970 971
	} else if (flags & O_PATH) {
		/*
		 * If we have O_PATH in the open flag. Then we
		 * cannot have anything other than the below set of flags
		 */