xref: /linux/fs/ext4/ioctl.c (revision 74ce1896c6c65b2f8cccbf59162d542988835835)
1 /*
2  * linux/fs/ext4/ioctl.c
3  *
4  * Copyright (C) 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  */
9 
10 #include <linux/fs.h>
11 #include <linux/capability.h>
12 #include <linux/time.h>
13 #include <linux/compat.h>
14 #include <linux/mount.h>
15 #include <linux/file.h>
16 #include <linux/quotaops.h>
17 #include <linux/uuid.h>
18 #include <linux/uaccess.h>
19 #include <linux/delay.h>
20 #include "ext4_jbd2.h"
21 #include "ext4.h"
22 #include <linux/fsmap.h>
23 #include "fsmap.h"
24 #include <trace/events/ext4.h>
25 
26 /**
27  * Swap memory between @a and @b for @len bytes.
28  *
29  * @a:          pointer to first memory area
30  * @b:          pointer to second memory area
31  * @len:        number of bytes to swap
32  *
33  */
34 static void memswap(void *a, void *b, size_t len)
35 {
36 	unsigned char *ap, *bp;
37 
38 	ap = (unsigned char *)a;
39 	bp = (unsigned char *)b;
40 	while (len-- > 0) {
41 		swap(*ap, *bp);
42 		ap++;
43 		bp++;
44 	}
45 }
46 
47 /**
48  * Swap i_data and associated attributes between @inode1 and @inode2.
49  * This function is used for the primary swap between inode1 and inode2
50  * and also to revert this primary swap in case of errors.
51  *
52  * Therefore you have to make sure, that calling this method twice
53  * will revert all changes.
54  *
55  * @inode1:     pointer to first inode
56  * @inode2:     pointer to second inode
57  */
58 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
59 {
60 	loff_t isize;
61 	struct ext4_inode_info *ei1;
62 	struct ext4_inode_info *ei2;
63 
64 	ei1 = EXT4_I(inode1);
65 	ei2 = EXT4_I(inode2);
66 
67 	swap(inode1->i_flags, inode2->i_flags);
68 	swap(inode1->i_version, inode2->i_version);
69 	swap(inode1->i_blocks, inode2->i_blocks);
70 	swap(inode1->i_bytes, inode2->i_bytes);
71 	swap(inode1->i_atime, inode2->i_atime);
72 	swap(inode1->i_mtime, inode2->i_mtime);
73 
74 	memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
75 	swap(ei1->i_flags, ei2->i_flags);
76 	swap(ei1->i_disksize, ei2->i_disksize);
77 	ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
78 	ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
79 
80 	isize = i_size_read(inode1);
81 	i_size_write(inode1, i_size_read(inode2));
82 	i_size_write(inode2, isize);
83 }
84 
85 /**
86  * Swap the information from the given @inode and the inode
87  * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
88  * important fields of the inodes.
89  *
90  * @sb:         the super block of the filesystem
91  * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
92  *
93  */
94 static long swap_inode_boot_loader(struct super_block *sb,
95 				struct inode *inode)
96 {
97 	handle_t *handle;
98 	int err;
99 	struct inode *inode_bl;
100 	struct ext4_inode_info *ei_bl;
101 	struct ext4_sb_info *sbi = EXT4_SB(sb);
102 
103 	if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode))
104 		return -EINVAL;
105 
106 	if (!inode_owner_or_capable(inode) || !capable(CAP_SYS_ADMIN))
107 		return -EPERM;
108 
109 	inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO);
110 	if (IS_ERR(inode_bl))
111 		return PTR_ERR(inode_bl);
112 	ei_bl = EXT4_I(inode_bl);
113 
114 	filemap_flush(inode->i_mapping);
115 	filemap_flush(inode_bl->i_mapping);
116 
117 	/* Protect orig inodes against a truncate and make sure,
118 	 * that only 1 swap_inode_boot_loader is running. */
119 	lock_two_nondirectories(inode, inode_bl);
120 
121 	truncate_inode_pages(&inode->i_data, 0);
122 	truncate_inode_pages(&inode_bl->i_data, 0);
123 
124 	/* Wait for all existing dio workers */
125 	ext4_inode_block_unlocked_dio(inode);
126 	ext4_inode_block_unlocked_dio(inode_bl);
127 	inode_dio_wait(inode);
128 	inode_dio_wait(inode_bl);
129 
130 	handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
131 	if (IS_ERR(handle)) {
132 		err = -EINVAL;
133 		goto journal_err_out;
134 	}
135 
136 	/* Protect extent tree against block allocations via delalloc */
137 	ext4_double_down_write_data_sem(inode, inode_bl);
138 
139 	if (inode_bl->i_nlink == 0) {
140 		/* this inode has never been used as a BOOT_LOADER */
141 		set_nlink(inode_bl, 1);
142 		i_uid_write(inode_bl, 0);
143 		i_gid_write(inode_bl, 0);
144 		inode_bl->i_flags = 0;
145 		ei_bl->i_flags = 0;
146 		inode_bl->i_version = 1;
147 		i_size_write(inode_bl, 0);
148 		inode_bl->i_mode = S_IFREG;
149 		if (ext4_has_feature_extents(sb)) {
150 			ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
151 			ext4_ext_tree_init(handle, inode_bl);
152 		} else
153 			memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
154 	}
155 
156 	swap_inode_data(inode, inode_bl);
157 
158 	inode->i_ctime = inode_bl->i_ctime = current_time(inode);
159 
160 	spin_lock(&sbi->s_next_gen_lock);
161 	inode->i_generation = sbi->s_next_generation++;
162 	inode_bl->i_generation = sbi->s_next_generation++;
163 	spin_unlock(&sbi->s_next_gen_lock);
164 
165 	ext4_discard_preallocations(inode);
166 
167 	err = ext4_mark_inode_dirty(handle, inode);
168 	if (err < 0) {
169 		ext4_warning(inode->i_sb,
170 			"couldn't mark inode #%lu dirty (err %d)",
171 			inode->i_ino, err);
172 		/* Revert all changes: */
173 		swap_inode_data(inode, inode_bl);
174 	} else {
175 		err = ext4_mark_inode_dirty(handle, inode_bl);
176 		if (err < 0) {
177 			ext4_warning(inode_bl->i_sb,
178 				"couldn't mark inode #%lu dirty (err %d)",
179 				inode_bl->i_ino, err);
180 			/* Revert all changes: */
181 			swap_inode_data(inode, inode_bl);
182 			ext4_mark_inode_dirty(handle, inode);
183 		}
184 	}
185 	ext4_journal_stop(handle);
186 	ext4_double_up_write_data_sem(inode, inode_bl);
187 
188 journal_err_out:
189 	ext4_inode_resume_unlocked_dio(inode);
190 	ext4_inode_resume_unlocked_dio(inode_bl);
191 	unlock_two_nondirectories(inode, inode_bl);
192 	iput(inode_bl);
193 	return err;
194 }
195 
196 #ifdef CONFIG_EXT4_FS_ENCRYPTION
197 static int uuid_is_zero(__u8 u[16])
198 {
199 	int	i;
200 
201 	for (i = 0; i < 16; i++)
202 		if (u[i])
203 			return 0;
204 	return 1;
205 }
206 #endif
207 
208 static int ext4_ioctl_setflags(struct inode *inode,
209 			       unsigned int flags)
210 {
211 	struct ext4_inode_info *ei = EXT4_I(inode);
212 	handle_t *handle = NULL;
213 	int err = -EPERM, migrate = 0;
214 	struct ext4_iloc iloc;
215 	unsigned int oldflags, mask, i;
216 	unsigned int jflag;
217 
218 	/* Is it quota file? Do not allow user to mess with it */
219 	if (ext4_is_quota_file(inode))
220 		goto flags_out;
221 
222 	oldflags = ei->i_flags;
223 
224 	/* The JOURNAL_DATA flag is modifiable only by root */
225 	jflag = flags & EXT4_JOURNAL_DATA_FL;
226 
227 	/*
228 	 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
229 	 * the relevant capability.
230 	 *
231 	 * This test looks nicer. Thanks to Pauline Middelink
232 	 */
233 	if ((flags ^ oldflags) & (EXT4_APPEND_FL | EXT4_IMMUTABLE_FL)) {
234 		if (!capable(CAP_LINUX_IMMUTABLE))
235 			goto flags_out;
236 	}
237 
238 	/*
239 	 * The JOURNAL_DATA flag can only be changed by
240 	 * the relevant capability.
241 	 */
242 	if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
243 		if (!capable(CAP_SYS_RESOURCE))
244 			goto flags_out;
245 	}
246 	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
247 		migrate = 1;
248 
249 	if (flags & EXT4_EOFBLOCKS_FL) {
250 		/* we don't support adding EOFBLOCKS flag */
251 		if (!(oldflags & EXT4_EOFBLOCKS_FL)) {
252 			err = -EOPNOTSUPP;
253 			goto flags_out;
254 		}
255 	} else if (oldflags & EXT4_EOFBLOCKS_FL) {
256 		err = ext4_truncate(inode);
257 		if (err)
258 			goto flags_out;
259 	}
260 
261 	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
262 	if (IS_ERR(handle)) {
263 		err = PTR_ERR(handle);
264 		goto flags_out;
265 	}
266 	if (IS_SYNC(inode))
267 		ext4_handle_sync(handle);
268 	err = ext4_reserve_inode_write(handle, inode, &iloc);
269 	if (err)
270 		goto flags_err;
271 
272 	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
273 		if (!(mask & EXT4_FL_USER_MODIFIABLE))
274 			continue;
275 		/* These flags get special treatment later */
276 		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
277 			continue;
278 		if (mask & flags)
279 			ext4_set_inode_flag(inode, i);
280 		else
281 			ext4_clear_inode_flag(inode, i);
282 	}
283 
284 	ext4_set_inode_flags(inode);
285 	inode->i_ctime = current_time(inode);
286 
287 	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
288 flags_err:
289 	ext4_journal_stop(handle);
290 	if (err)
291 		goto flags_out;
292 
293 	if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL))
294 		err = ext4_change_inode_journal_flag(inode, jflag);
295 	if (err)
296 		goto flags_out;
297 	if (migrate) {
298 		if (flags & EXT4_EXTENTS_FL)
299 			err = ext4_ext_migrate(inode);
300 		else
301 			err = ext4_ind_migrate(inode);
302 	}
303 
304 flags_out:
305 	return err;
306 }
307 
308 #ifdef CONFIG_QUOTA
309 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
310 {
311 	struct inode *inode = file_inode(filp);
312 	struct super_block *sb = inode->i_sb;
313 	struct ext4_inode_info *ei = EXT4_I(inode);
314 	int err, rc;
315 	handle_t *handle;
316 	kprojid_t kprojid;
317 	struct ext4_iloc iloc;
318 	struct ext4_inode *raw_inode;
319 	struct dquot *transfer_to[MAXQUOTAS] = { };
320 
321 	if (!ext4_has_feature_project(sb)) {
322 		if (projid != EXT4_DEF_PROJID)
323 			return -EOPNOTSUPP;
324 		else
325 			return 0;
326 	}
327 
328 	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
329 		return -EOPNOTSUPP;
330 
331 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
332 
333 	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
334 		return 0;
335 
336 	err = mnt_want_write_file(filp);
337 	if (err)
338 		return err;
339 
340 	err = -EPERM;
341 	inode_lock(inode);
342 	/* Is it quota file? Do not allow user to mess with it */
343 	if (ext4_is_quota_file(inode))
344 		goto out_unlock;
345 
346 	err = ext4_get_inode_loc(inode, &iloc);
347 	if (err)
348 		goto out_unlock;
349 
350 	raw_inode = ext4_raw_inode(&iloc);
351 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
352 		err = ext4_expand_extra_isize(inode,
353 					      EXT4_SB(sb)->s_want_extra_isize,
354 					      &iloc);
355 		if (err)
356 			goto out_unlock;
357 	} else {
358 		brelse(iloc.bh);
359 	}
360 
361 	dquot_initialize(inode);
362 
363 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
364 		EXT4_QUOTA_INIT_BLOCKS(sb) +
365 		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
366 	if (IS_ERR(handle)) {
367 		err = PTR_ERR(handle);
368 		goto out_unlock;
369 	}
370 
371 	err = ext4_reserve_inode_write(handle, inode, &iloc);
372 	if (err)
373 		goto out_stop;
374 
375 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
376 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
377 
378 		/* __dquot_transfer() calls back ext4_get_inode_usage() which
379 		 * counts xattr inode references.
380 		 */
381 		down_read(&EXT4_I(inode)->xattr_sem);
382 		err = __dquot_transfer(inode, transfer_to);
383 		up_read(&EXT4_I(inode)->xattr_sem);
384 		dqput(transfer_to[PRJQUOTA]);
385 		if (err)
386 			goto out_dirty;
387 	}
388 
389 	EXT4_I(inode)->i_projid = kprojid;
390 	inode->i_ctime = current_time(inode);
391 out_dirty:
392 	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
393 	if (!err)
394 		err = rc;
395 out_stop:
396 	ext4_journal_stop(handle);
397 out_unlock:
398 	inode_unlock(inode);
399 	mnt_drop_write_file(filp);
400 	return err;
401 }
402 #else
403 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
404 {
405 	if (projid != EXT4_DEF_PROJID)
406 		return -EOPNOTSUPP;
407 	return 0;
408 }
409 #endif
410 
411 /* Transfer internal flags to xflags */
412 static inline __u32 ext4_iflags_to_xflags(unsigned long iflags)
413 {
414 	__u32 xflags = 0;
415 
416 	if (iflags & EXT4_SYNC_FL)
417 		xflags |= FS_XFLAG_SYNC;
418 	if (iflags & EXT4_IMMUTABLE_FL)
419 		xflags |= FS_XFLAG_IMMUTABLE;
420 	if (iflags & EXT4_APPEND_FL)
421 		xflags |= FS_XFLAG_APPEND;
422 	if (iflags & EXT4_NODUMP_FL)
423 		xflags |= FS_XFLAG_NODUMP;
424 	if (iflags & EXT4_NOATIME_FL)
425 		xflags |= FS_XFLAG_NOATIME;
426 	if (iflags & EXT4_PROJINHERIT_FL)
427 		xflags |= FS_XFLAG_PROJINHERIT;
428 	return xflags;
429 }
430 
431 #define EXT4_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
432 				  FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
433 				  FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
434 
435 /* Transfer xflags flags to internal */
436 static inline unsigned long ext4_xflags_to_iflags(__u32 xflags)
437 {
438 	unsigned long iflags = 0;
439 
440 	if (xflags & FS_XFLAG_SYNC)
441 		iflags |= EXT4_SYNC_FL;
442 	if (xflags & FS_XFLAG_IMMUTABLE)
443 		iflags |= EXT4_IMMUTABLE_FL;
444 	if (xflags & FS_XFLAG_APPEND)
445 		iflags |= EXT4_APPEND_FL;
446 	if (xflags & FS_XFLAG_NODUMP)
447 		iflags |= EXT4_NODUMP_FL;
448 	if (xflags & FS_XFLAG_NOATIME)
449 		iflags |= EXT4_NOATIME_FL;
450 	if (xflags & FS_XFLAG_PROJINHERIT)
451 		iflags |= EXT4_PROJINHERIT_FL;
452 
453 	return iflags;
454 }
455 
456 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
457 {
458 	struct ext4_sb_info *sbi = EXT4_SB(sb);
459 	__u32 flags;
460 
461 	if (!capable(CAP_SYS_ADMIN))
462 		return -EPERM;
463 
464 	if (get_user(flags, (__u32 __user *)arg))
465 		return -EFAULT;
466 
467 	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
468 		return -EINVAL;
469 
470 	if (ext4_forced_shutdown(sbi))
471 		return 0;
472 
473 	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
474 
475 	switch (flags) {
476 	case EXT4_GOING_FLAGS_DEFAULT:
477 		freeze_bdev(sb->s_bdev);
478 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
479 		thaw_bdev(sb->s_bdev, sb);
480 		break;
481 	case EXT4_GOING_FLAGS_LOGFLUSH:
482 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
483 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
484 			(void) ext4_force_commit(sb);
485 			jbd2_journal_abort(sbi->s_journal, 0);
486 		}
487 		break;
488 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
489 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
490 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
491 			msleep(100);
492 			jbd2_journal_abort(sbi->s_journal, 0);
493 		}
494 		break;
495 	default:
496 		return -EINVAL;
497 	}
498 	clear_opt(sb, DISCARD);
499 	return 0;
500 }
501 
502 struct getfsmap_info {
503 	struct super_block	*gi_sb;
504 	struct fsmap_head __user *gi_data;
505 	unsigned int		gi_idx;
506 	__u32			gi_last_flags;
507 };
508 
509 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
510 {
511 	struct getfsmap_info *info = priv;
512 	struct fsmap fm;
513 
514 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
515 
516 	info->gi_last_flags = xfm->fmr_flags;
517 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
518 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
519 			sizeof(struct fsmap)))
520 		return -EFAULT;
521 
522 	return 0;
523 }
524 
525 static int ext4_ioc_getfsmap(struct super_block *sb,
526 			     struct fsmap_head __user *arg)
527 {
528 	struct getfsmap_info info = {0};
529 	struct ext4_fsmap_head xhead = {0};
530 	struct fsmap_head head;
531 	bool aborted = false;
532 	int error;
533 
534 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
535 		return -EFAULT;
536 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
537 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
538 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
539 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
540 		       sizeof(head.fmh_keys[1].fmr_reserved)))
541 		return -EINVAL;
542 	/*
543 	 * ext4 doesn't report file extents at all, so the only valid
544 	 * file offsets are the magic ones (all zeroes or all ones).
545 	 */
546 	if (head.fmh_keys[0].fmr_offset ||
547 	    (head.fmh_keys[1].fmr_offset != 0 &&
548 	     head.fmh_keys[1].fmr_offset != -1ULL))
549 		return -EINVAL;
550 
551 	xhead.fmh_iflags = head.fmh_iflags;
552 	xhead.fmh_count = head.fmh_count;
553 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
554 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
555 
556 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
557 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
558 
559 	info.gi_sb = sb;
560 	info.gi_data = arg;
561 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
562 	if (error == EXT4_QUERY_RANGE_ABORT) {
563 		error = 0;
564 		aborted = true;
565 	} else if (error)
566 		return error;
567 
568 	/* If we didn't abort, set the "last" flag in the last fmx */
569 	if (!aborted && info.gi_idx) {
570 		info.gi_last_flags |= FMR_OF_LAST;
571 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
572 				 &info.gi_last_flags,
573 				 sizeof(info.gi_last_flags)))
574 			return -EFAULT;
575 	}
576 
577 	/* copy back header */
578 	head.fmh_entries = xhead.fmh_entries;
579 	head.fmh_oflags = xhead.fmh_oflags;
580 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
581 		return -EFAULT;
582 
583 	return 0;
584 }
585 
586 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
587 {
588 	struct inode *inode = file_inode(filp);
589 	struct super_block *sb = inode->i_sb;
590 	struct ext4_inode_info *ei = EXT4_I(inode);
591 	unsigned int flags;
592 
593 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
594 
595 	switch (cmd) {
596 	case FS_IOC_GETFSMAP:
597 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
598 	case EXT4_IOC_GETFLAGS:
599 		flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
600 		return put_user(flags, (int __user *) arg);
601 	case EXT4_IOC_SETFLAGS: {
602 		int err;
603 
604 		if (!inode_owner_or_capable(inode))
605 			return -EACCES;
606 
607 		if (get_user(flags, (int __user *) arg))
608 			return -EFAULT;
609 
610 		if (flags & ~EXT4_FL_USER_VISIBLE)
611 			return -EOPNOTSUPP;
612 		/*
613 		 * chattr(1) grabs flags via GETFLAGS, modifies the result and
614 		 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
615 		 * more restrictive than just silently masking off visible but
616 		 * not settable flags as we always did.
617 		 */
618 		flags &= EXT4_FL_USER_MODIFIABLE;
619 		if (ext4_mask_flags(inode->i_mode, flags) != flags)
620 			return -EOPNOTSUPP;
621 
622 		err = mnt_want_write_file(filp);
623 		if (err)
624 			return err;
625 
626 		inode_lock(inode);
627 		err = ext4_ioctl_setflags(inode, flags);
628 		inode_unlock(inode);
629 		mnt_drop_write_file(filp);
630 		return err;
631 	}
632 	case EXT4_IOC_GETVERSION:
633 	case EXT4_IOC_GETVERSION_OLD:
634 		return put_user(inode->i_generation, (int __user *) arg);
635 	case EXT4_IOC_SETVERSION:
636 	case EXT4_IOC_SETVERSION_OLD: {
637 		handle_t *handle;
638 		struct ext4_iloc iloc;
639 		__u32 generation;
640 		int err;
641 
642 		if (!inode_owner_or_capable(inode))
643 			return -EPERM;
644 
645 		if (ext4_has_metadata_csum(inode->i_sb)) {
646 			ext4_warning(sb, "Setting inode version is not "
647 				     "supported with metadata_csum enabled.");
648 			return -ENOTTY;
649 		}
650 
651 		err = mnt_want_write_file(filp);
652 		if (err)
653 			return err;
654 		if (get_user(generation, (int __user *) arg)) {
655 			err = -EFAULT;
656 			goto setversion_out;
657 		}
658 
659 		inode_lock(inode);
660 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
661 		if (IS_ERR(handle)) {
662 			err = PTR_ERR(handle);
663 			goto unlock_out;
664 		}
665 		err = ext4_reserve_inode_write(handle, inode, &iloc);
666 		if (err == 0) {
667 			inode->i_ctime = current_time(inode);
668 			inode->i_generation = generation;
669 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
670 		}
671 		ext4_journal_stop(handle);
672 
673 unlock_out:
674 		inode_unlock(inode);
675 setversion_out:
676 		mnt_drop_write_file(filp);
677 		return err;
678 	}
679 	case EXT4_IOC_GROUP_EXTEND: {
680 		ext4_fsblk_t n_blocks_count;
681 		int err, err2=0;
682 
683 		err = ext4_resize_begin(sb);
684 		if (err)
685 			return err;
686 
687 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
688 			err = -EFAULT;
689 			goto group_extend_out;
690 		}
691 
692 		if (ext4_has_feature_bigalloc(sb)) {
693 			ext4_msg(sb, KERN_ERR,
694 				 "Online resizing not supported with bigalloc");
695 			err = -EOPNOTSUPP;
696 			goto group_extend_out;
697 		}
698 
699 		err = mnt_want_write_file(filp);
700 		if (err)
701 			goto group_extend_out;
702 
703 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
704 		if (EXT4_SB(sb)->s_journal) {
705 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
706 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
707 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
708 		}
709 		if (err == 0)
710 			err = err2;
711 		mnt_drop_write_file(filp);
712 group_extend_out:
713 		ext4_resize_end(sb);
714 		return err;
715 	}
716 
717 	case EXT4_IOC_MOVE_EXT: {
718 		struct move_extent me;
719 		struct fd donor;
720 		int err;
721 
722 		if (!(filp->f_mode & FMODE_READ) ||
723 		    !(filp->f_mode & FMODE_WRITE))
724 			return -EBADF;
725 
726 		if (copy_from_user(&me,
727 			(struct move_extent __user *)arg, sizeof(me)))
728 			return -EFAULT;
729 		me.moved_len = 0;
730 
731 		donor = fdget(me.donor_fd);
732 		if (!donor.file)
733 			return -EBADF;
734 
735 		if (!(donor.file->f_mode & FMODE_WRITE)) {
736 			err = -EBADF;
737 			goto mext_out;
738 		}
739 
740 		if (ext4_has_feature_bigalloc(sb)) {
741 			ext4_msg(sb, KERN_ERR,
742 				 "Online defrag not supported with bigalloc");
743 			err = -EOPNOTSUPP;
744 			goto mext_out;
745 		} else if (IS_DAX(inode)) {
746 			ext4_msg(sb, KERN_ERR,
747 				 "Online defrag not supported with DAX");
748 			err = -EOPNOTSUPP;
749 			goto mext_out;
750 		}
751 
752 		err = mnt_want_write_file(filp);
753 		if (err)
754 			goto mext_out;
755 
756 		err = ext4_move_extents(filp, donor.file, me.orig_start,
757 					me.donor_start, me.len, &me.moved_len);
758 		mnt_drop_write_file(filp);
759 
760 		if (copy_to_user((struct move_extent __user *)arg,
761 				 &me, sizeof(me)))
762 			err = -EFAULT;
763 mext_out:
764 		fdput(donor);
765 		return err;
766 	}
767 
768 	case EXT4_IOC_GROUP_ADD: {
769 		struct ext4_new_group_data input;
770 		int err, err2=0;
771 
772 		err = ext4_resize_begin(sb);
773 		if (err)
774 			return err;
775 
776 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
777 				sizeof(input))) {
778 			err = -EFAULT;
779 			goto group_add_out;
780 		}
781 
782 		if (ext4_has_feature_bigalloc(sb)) {
783 			ext4_msg(sb, KERN_ERR,
784 				 "Online resizing not supported with bigalloc");
785 			err = -EOPNOTSUPP;
786 			goto group_add_out;
787 		}
788 
789 		err = mnt_want_write_file(filp);
790 		if (err)
791 			goto group_add_out;
792 
793 		err = ext4_group_add(sb, &input);
794 		if (EXT4_SB(sb)->s_journal) {
795 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
796 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
797 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
798 		}
799 		if (err == 0)
800 			err = err2;
801 		mnt_drop_write_file(filp);
802 		if (!err && ext4_has_group_desc_csum(sb) &&
803 		    test_opt(sb, INIT_INODE_TABLE))
804 			err = ext4_register_li_request(sb, input.group);
805 group_add_out:
806 		ext4_resize_end(sb);
807 		return err;
808 	}
809 
810 	case EXT4_IOC_MIGRATE:
811 	{
812 		int err;
813 		if (!inode_owner_or_capable(inode))
814 			return -EACCES;
815 
816 		err = mnt_want_write_file(filp);
817 		if (err)
818 			return err;
819 		/*
820 		 * inode_mutex prevent write and truncate on the file.
821 		 * Read still goes through. We take i_data_sem in
822 		 * ext4_ext_swap_inode_data before we switch the
823 		 * inode format to prevent read.
824 		 */
825 		inode_lock((inode));
826 		err = ext4_ext_migrate(inode);
827 		inode_unlock((inode));
828 		mnt_drop_write_file(filp);
829 		return err;
830 	}
831 
832 	case EXT4_IOC_ALLOC_DA_BLKS:
833 	{
834 		int err;
835 		if (!inode_owner_or_capable(inode))
836 			return -EACCES;
837 
838 		err = mnt_want_write_file(filp);
839 		if (err)
840 			return err;
841 		err = ext4_alloc_da_blocks(inode);
842 		mnt_drop_write_file(filp);
843 		return err;
844 	}
845 
846 	case EXT4_IOC_SWAP_BOOT:
847 	{
848 		int err;
849 		if (!(filp->f_mode & FMODE_WRITE))
850 			return -EBADF;
851 		err = mnt_want_write_file(filp);
852 		if (err)
853 			return err;
854 		err = swap_inode_boot_loader(sb, inode);
855 		mnt_drop_write_file(filp);
856 		return err;
857 	}
858 
859 	case EXT4_IOC_RESIZE_FS: {
860 		ext4_fsblk_t n_blocks_count;
861 		int err = 0, err2 = 0;
862 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
863 
864 		if (ext4_has_feature_bigalloc(sb)) {
865 			ext4_msg(sb, KERN_ERR,
866 				 "Online resizing not (yet) supported with bigalloc");
867 			return -EOPNOTSUPP;
868 		}
869 
870 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
871 				   sizeof(__u64))) {
872 			return -EFAULT;
873 		}
874 
875 		err = ext4_resize_begin(sb);
876 		if (err)
877 			return err;
878 
879 		err = mnt_want_write_file(filp);
880 		if (err)
881 			goto resizefs_out;
882 
883 		err = ext4_resize_fs(sb, n_blocks_count);
884 		if (EXT4_SB(sb)->s_journal) {
885 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
886 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
887 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
888 		}
889 		if (err == 0)
890 			err = err2;
891 		mnt_drop_write_file(filp);
892 		if (!err && (o_group > EXT4_SB(sb)->s_groups_count) &&
893 		    ext4_has_group_desc_csum(sb) &&
894 		    test_opt(sb, INIT_INODE_TABLE))
895 			err = ext4_register_li_request(sb, o_group);
896 
897 resizefs_out:
898 		ext4_resize_end(sb);
899 		return err;
900 	}
901 
902 	case FITRIM:
903 	{
904 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
905 		struct fstrim_range range;
906 		int ret = 0;
907 
908 		if (!capable(CAP_SYS_ADMIN))
909 			return -EPERM;
910 
911 		if (!blk_queue_discard(q))
912 			return -EOPNOTSUPP;
913 
914 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
915 		    sizeof(range)))
916 			return -EFAULT;
917 
918 		range.minlen = max((unsigned int)range.minlen,
919 				   q->limits.discard_granularity);
920 		ret = ext4_trim_fs(sb, &range);
921 		if (ret < 0)
922 			return ret;
923 
924 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
925 		    sizeof(range)))
926 			return -EFAULT;
927 
928 		return 0;
929 	}
930 	case EXT4_IOC_PRECACHE_EXTENTS:
931 		return ext4_ext_precache(inode);
932 
933 	case EXT4_IOC_SET_ENCRYPTION_POLICY:
934 		if (!ext4_has_feature_encrypt(sb))
935 			return -EOPNOTSUPP;
936 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
937 
938 	case EXT4_IOC_GET_ENCRYPTION_PWSALT: {
939 #ifdef CONFIG_EXT4_FS_ENCRYPTION
940 		int err, err2;
941 		struct ext4_sb_info *sbi = EXT4_SB(sb);
942 		handle_t *handle;
943 
944 		if (!ext4_has_feature_encrypt(sb))
945 			return -EOPNOTSUPP;
946 		if (uuid_is_zero(sbi->s_es->s_encrypt_pw_salt)) {
947 			err = mnt_want_write_file(filp);
948 			if (err)
949 				return err;
950 			handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
951 			if (IS_ERR(handle)) {
952 				err = PTR_ERR(handle);
953 				goto pwsalt_err_exit;
954 			}
955 			err = ext4_journal_get_write_access(handle, sbi->s_sbh);
956 			if (err)
957 				goto pwsalt_err_journal;
958 			generate_random_uuid(sbi->s_es->s_encrypt_pw_salt);
959 			err = ext4_handle_dirty_metadata(handle, NULL,
960 							 sbi->s_sbh);
961 		pwsalt_err_journal:
962 			err2 = ext4_journal_stop(handle);
963 			if (err2 && !err)
964 				err = err2;
965 		pwsalt_err_exit:
966 			mnt_drop_write_file(filp);
967 			if (err)
968 				return err;
969 		}
970 		if (copy_to_user((void __user *) arg,
971 				 sbi->s_es->s_encrypt_pw_salt, 16))
972 			return -EFAULT;
973 		return 0;
974 #else
975 		return -EOPNOTSUPP;
976 #endif
977 	}
978 	case EXT4_IOC_GET_ENCRYPTION_POLICY:
979 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
980 
981 	case EXT4_IOC_FSGETXATTR:
982 	{
983 		struct fsxattr fa;
984 
985 		memset(&fa, 0, sizeof(struct fsxattr));
986 		fa.fsx_xflags = ext4_iflags_to_xflags(ei->i_flags & EXT4_FL_USER_VISIBLE);
987 
988 		if (ext4_has_feature_project(inode->i_sb)) {
989 			fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
990 				EXT4_I(inode)->i_projid);
991 		}
992 
993 		if (copy_to_user((struct fsxattr __user *)arg,
994 				 &fa, sizeof(fa)))
995 			return -EFAULT;
996 		return 0;
997 	}
998 	case EXT4_IOC_FSSETXATTR:
999 	{
1000 		struct fsxattr fa;
1001 		int err;
1002 
1003 		if (copy_from_user(&fa, (struct fsxattr __user *)arg,
1004 				   sizeof(fa)))
1005 			return -EFAULT;
1006 
1007 		/* Make sure caller has proper permission */
1008 		if (!inode_owner_or_capable(inode))
1009 			return -EACCES;
1010 
1011 		if (fa.fsx_xflags & ~EXT4_SUPPORTED_FS_XFLAGS)
1012 			return -EOPNOTSUPP;
1013 
1014 		flags = ext4_xflags_to_iflags(fa.fsx_xflags);
1015 		if (ext4_mask_flags(inode->i_mode, flags) != flags)
1016 			return -EOPNOTSUPP;
1017 
1018 		err = mnt_want_write_file(filp);
1019 		if (err)
1020 			return err;
1021 
1022 		inode_lock(inode);
1023 		flags = (ei->i_flags & ~EXT4_FL_XFLAG_VISIBLE) |
1024 			 (flags & EXT4_FL_XFLAG_VISIBLE);
1025 		err = ext4_ioctl_setflags(inode, flags);
1026 		inode_unlock(inode);
1027 		mnt_drop_write_file(filp);
1028 		if (err)
1029 			return err;
1030 
1031 		err = ext4_ioctl_setproject(filp, fa.fsx_projid);
1032 		if (err)
1033 			return err;
1034 
1035 		return 0;
1036 	}
1037 	case EXT4_IOC_SHUTDOWN:
1038 		return ext4_shutdown(sb, arg);
1039 	default:
1040 		return -ENOTTY;
1041 	}
1042 }
1043 
1044 #ifdef CONFIG_COMPAT
1045 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1046 {
1047 	/* These are just misnamed, they actually get/put from/to user an int */
1048 	switch (cmd) {
1049 	case EXT4_IOC32_GETFLAGS:
1050 		cmd = EXT4_IOC_GETFLAGS;
1051 		break;
1052 	case EXT4_IOC32_SETFLAGS:
1053 		cmd = EXT4_IOC_SETFLAGS;
1054 		break;
1055 	case EXT4_IOC32_GETVERSION:
1056 		cmd = EXT4_IOC_GETVERSION;
1057 		break;
1058 	case EXT4_IOC32_SETVERSION:
1059 		cmd = EXT4_IOC_SETVERSION;
1060 		break;
1061 	case EXT4_IOC32_GROUP_EXTEND:
1062 		cmd = EXT4_IOC_GROUP_EXTEND;
1063 		break;
1064 	case EXT4_IOC32_GETVERSION_OLD:
1065 		cmd = EXT4_IOC_GETVERSION_OLD;
1066 		break;
1067 	case EXT4_IOC32_SETVERSION_OLD:
1068 		cmd = EXT4_IOC_SETVERSION_OLD;
1069 		break;
1070 	case EXT4_IOC32_GETRSVSZ:
1071 		cmd = EXT4_IOC_GETRSVSZ;
1072 		break;
1073 	case EXT4_IOC32_SETRSVSZ:
1074 		cmd = EXT4_IOC_SETRSVSZ;
1075 		break;
1076 	case EXT4_IOC32_GROUP_ADD: {
1077 		struct compat_ext4_new_group_input __user *uinput;
1078 		struct ext4_new_group_input input;
1079 		mm_segment_t old_fs;
1080 		int err;
1081 
1082 		uinput = compat_ptr(arg);
1083 		err = get_user(input.group, &uinput->group);
1084 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1085 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1086 		err |= get_user(input.inode_table, &uinput->inode_table);
1087 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1088 		err |= get_user(input.reserved_blocks,
1089 				&uinput->reserved_blocks);
1090 		if (err)
1091 			return -EFAULT;
1092 		old_fs = get_fs();
1093 		set_fs(KERNEL_DS);
1094 		err = ext4_ioctl(file, EXT4_IOC_GROUP_ADD,
1095 				 (unsigned long) &input);
1096 		set_fs(old_fs);
1097 		return err;
1098 	}
1099 	case EXT4_IOC_MOVE_EXT:
1100 	case EXT4_IOC_RESIZE_FS:
1101 	case EXT4_IOC_PRECACHE_EXTENTS:
1102 	case EXT4_IOC_SET_ENCRYPTION_POLICY:
1103 	case EXT4_IOC_GET_ENCRYPTION_PWSALT:
1104 	case EXT4_IOC_GET_ENCRYPTION_POLICY:
1105 	case EXT4_IOC_SHUTDOWN:
1106 	case FS_IOC_GETFSMAP:
1107 		break;
1108 	default:
1109 		return -ENOIOCTLCMD;
1110 	}
1111 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1112 }
1113 #endif
1114