xref: /linux/fs/ext4/ioctl.c (revision bba2c3615bd6cfee7456d1130f2e6b01b3f4e9ba)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * linux/fs/ext4/ioctl.c
4  *
5  * Copyright (C) 1993, 1994, 1995
6  * Remy Card (card@masi.ibp.fr)
7  * Laboratoire MASI - Institut Blaise Pascal
8  * Universite Pierre et Marie Curie (Paris VI)
9  */
10 
11 #include <linux/fs.h>
12 #include <linux/capability.h>
13 #include <linux/time.h>
14 #include <linux/compat.h>
15 #include <linux/mount.h>
16 #include <linux/file.h>
17 #include <linux/quotaops.h>
18 #include <linux/random.h>
19 #include <linux/uaccess.h>
20 #include <linux/delay.h>
21 #include <linux/iversion.h>
22 #include <linux/fileattr.h>
23 #include <linux/uuid.h>
24 #include "ext4_jbd2.h"
25 #include "ext4.h"
26 #include <linux/fsmap.h>
27 #include "fsmap.h"
28 #include <trace/events/ext4.h>
29 #include <linux/fserror.h>
30 
31 typedef void ext4_update_sb_callback(struct ext4_sb_info *sbi,
32 				     struct ext4_super_block *es,
33 				     const void *arg);
34 
35 /*
36  * Superblock modification callback function for changing file system
37  * label
38  */
39 static void ext4_sb_setlabel(struct ext4_sb_info *sbi,
40 			     struct ext4_super_block *es, const void *arg)
41 {
42 	/* Sanity check, this should never happen */
43 	BUILD_BUG_ON(sizeof(es->s_volume_name) < EXT4_LABEL_MAX);
44 
45 	memcpy(es->s_volume_name, (char *)arg, EXT4_LABEL_MAX);
46 }
47 
48 /*
49  * Superblock modification callback function for changing file system
50  * UUID.
51  */
52 static void ext4_sb_setuuid(struct ext4_sb_info *sbi,
53 			    struct ext4_super_block *es, const void *arg)
54 {
55 	memcpy(es->s_uuid, (__u8 *)arg, UUID_SIZE);
56 }
57 
58 static
59 int ext4_update_primary_sb(struct super_block *sb, handle_t *handle,
60 			   ext4_update_sb_callback func,
61 			   const void *arg)
62 {
63 	int err = 0;
64 	struct ext4_sb_info *sbi = EXT4_SB(sb);
65 	struct buffer_head *bh = sbi->s_sbh;
66 	struct ext4_super_block *es = sbi->s_es;
67 
68 	trace_ext4_update_sb(sb, bh->b_blocknr, 1);
69 
70 	BUFFER_TRACE(bh, "get_write_access");
71 	err = ext4_journal_get_write_access(handle, sb,
72 					    bh,
73 					    EXT4_JTR_NONE);
74 	if (err)
75 		goto out_err;
76 
77 	lock_buffer(bh);
78 	func(sbi, es, arg);
79 	ext4_superblock_csum_set(sb);
80 	unlock_buffer(bh);
81 
82 	if (buffer_write_io_error(bh) || !buffer_uptodate(bh)) {
83 		ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
84 			 "superblock detected");
85 		clear_buffer_write_io_error(bh);
86 		set_buffer_uptodate(bh);
87 	}
88 
89 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
90 	if (err)
91 		goto out_err;
92 	err = sync_dirty_buffer(bh);
93 out_err:
94 	ext4_std_error(sb, err);
95 	return err;
96 }
97 
98 /*
99  * Update one backup superblock in the group 'grp' using the callback
100  * function 'func' and argument 'arg'. If the handle is NULL the
101  * modification is not journalled.
102  *
103  * Returns: 0 when no modification was done (no superblock in the group)
104  *	    1 when the modification was successful
105  *	   <0 on error
106  */
107 static int ext4_update_backup_sb(struct super_block *sb,
108 				 handle_t *handle, ext4_group_t grp,
109 				 ext4_update_sb_callback func, const void *arg)
110 {
111 	int err = 0;
112 	ext4_fsblk_t sb_block;
113 	struct buffer_head *bh;
114 	unsigned long offset = 0;
115 	struct ext4_super_block *es;
116 
117 	if (!ext4_bg_has_super(sb, grp))
118 		return 0;
119 
120 	/*
121 	 * For the group 0 there is always 1k padding, so we have
122 	 * either adjust offset, or sb_block depending on blocksize
123 	 */
124 	if (grp == 0) {
125 		sb_block = 1 * EXT4_MIN_BLOCK_SIZE;
126 		offset = do_div(sb_block, sb->s_blocksize);
127 	} else {
128 		sb_block = ext4_group_first_block_no(sb, grp);
129 		offset = 0;
130 	}
131 
132 	trace_ext4_update_sb(sb, sb_block, handle ? 1 : 0);
133 
134 	bh = ext4_sb_bread(sb, sb_block, 0);
135 	if (IS_ERR(bh))
136 		return PTR_ERR(bh);
137 
138 	if (handle) {
139 		BUFFER_TRACE(bh, "get_write_access");
140 		err = ext4_journal_get_write_access(handle, sb,
141 						    bh,
142 						    EXT4_JTR_NONE);
143 		if (err)
144 			goto out_bh;
145 	}
146 
147 	es = (struct ext4_super_block *) (bh->b_data + offset);
148 	lock_buffer(bh);
149 	if (ext4_has_feature_metadata_csum(sb) &&
150 	    es->s_checksum != ext4_superblock_csum(es)) {
151 		ext4_msg(sb, KERN_ERR, "Invalid checksum for backup "
152 		"superblock %llu", sb_block);
153 		unlock_buffer(bh);
154 		goto out_bh;
155 	}
156 	func(EXT4_SB(sb), es, arg);
157 	if (ext4_has_feature_metadata_csum(sb))
158 		es->s_checksum = ext4_superblock_csum(es);
159 	set_buffer_uptodate(bh);
160 	unlock_buffer(bh);
161 
162 	if (handle) {
163 		err = ext4_handle_dirty_metadata(handle, NULL, bh);
164 		if (err)
165 			goto out_bh;
166 	} else {
167 		BUFFER_TRACE(bh, "marking dirty");
168 		mark_buffer_dirty(bh);
169 	}
170 	err = sync_dirty_buffer(bh);
171 
172 out_bh:
173 	brelse(bh);
174 	ext4_std_error(sb, err);
175 	return (err) ? err : 1;
176 }
177 
178 /*
179  * Update primary and backup superblocks using the provided function
180  * func and argument arg.
181  *
182  * Only the primary superblock and at most two backup superblock
183  * modifications are journalled; the rest is modified without journal.
184  * This is safe because e2fsck will re-write them if there is a problem,
185  * and we're very unlikely to ever need more than two backups.
186  */
187 static
188 int ext4_update_superblocks_fn(struct super_block *sb,
189 			       ext4_update_sb_callback func,
190 			       const void *arg)
191 {
192 	handle_t *handle;
193 	ext4_group_t ngroups;
194 	unsigned int three = 1;
195 	unsigned int five = 5;
196 	unsigned int seven = 7;
197 	int err = 0, ret, i;
198 	ext4_group_t grp, primary_grp;
199 	struct ext4_sb_info *sbi = EXT4_SB(sb);
200 
201 	/*
202 	 * We can't update superblocks while the online resize is running
203 	 */
204 	if (test_and_set_bit_lock(EXT4_FLAGS_RESIZING,
205 				  &sbi->s_ext4_flags)) {
206 		ext4_msg(sb, KERN_ERR, "Can't modify superblock while"
207 			 "performing online resize");
208 		return -EBUSY;
209 	}
210 
211 	/*
212 	 * We're only going to update primary superblock and two
213 	 * backup superblocks in this transaction.
214 	 */
215 	handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 3);
216 	if (IS_ERR(handle)) {
217 		err = PTR_ERR(handle);
218 		goto out;
219 	}
220 
221 	/* Update primary superblock */
222 	err = ext4_update_primary_sb(sb, handle, func, arg);
223 	if (err) {
224 		ext4_msg(sb, KERN_ERR, "Failed to update primary "
225 			 "superblock");
226 		goto out_journal;
227 	}
228 
229 	primary_grp = ext4_get_group_number(sb, sbi->s_sbh->b_blocknr);
230 	ngroups = ext4_get_groups_count(sb);
231 
232 	/*
233 	 * Update backup superblocks. We have to start from group 0
234 	 * because it might not be where the primary superblock is
235 	 * if the fs is mounted with -o sb=<backup_sb_block>
236 	 */
237 	i = 0;
238 	grp = 0;
239 	while (grp < ngroups) {
240 		/* Skip primary superblock */
241 		if (grp == primary_grp)
242 			goto next_grp;
243 
244 		ret = ext4_update_backup_sb(sb, handle, grp, func, arg);
245 		if (ret < 0) {
246 			/* Ignore bad checksum; try to update next sb */
247 			if (ret == -EFSBADCRC)
248 				goto next_grp;
249 			err = ret;
250 			goto out_journal;
251 		}
252 
253 		i += ret;
254 		if (handle && i > 1) {
255 			/*
256 			 * We're only journalling primary superblock and
257 			 * two backup superblocks; the rest is not
258 			 * journalled.
259 			 */
260 			err = ext4_journal_stop(handle);
261 			if (err)
262 				goto out;
263 			handle = NULL;
264 		}
265 next_grp:
266 		grp = ext4_list_backups(sb, &three, &five, &seven);
267 	}
268 
269 out_journal:
270 	if (handle) {
271 		ret = ext4_journal_stop(handle);
272 		if (ret && !err)
273 			err = ret;
274 	}
275 out:
276 	clear_bit_unlock(EXT4_FLAGS_RESIZING, &sbi->s_ext4_flags);
277 	smp_mb__after_atomic();
278 	return err ? err : 0;
279 }
280 
281 /*
282  * Swap memory between @a and @b for @len bytes.
283  *
284  * @a:          pointer to first memory area
285  * @b:          pointer to second memory area
286  * @len:        number of bytes to swap
287  *
288  */
289 static void memswap(void *a, void *b, size_t len)
290 {
291 	unsigned char *ap, *bp;
292 
293 	ap = (unsigned char *)a;
294 	bp = (unsigned char *)b;
295 	while (len-- > 0) {
296 		swap(*ap, *bp);
297 		ap++;
298 		bp++;
299 	}
300 }
301 
302 /*
303  * Swap i_data and associated attributes between @inode1 and @inode2.
304  * This function is used for the primary swap between inode1 and inode2
305  * and also to revert this primary swap in case of errors.
306  *
307  * Therefore you have to make sure, that calling this method twice
308  * will revert all changes.
309  *
310  * @inode1:     pointer to first inode
311  * @inode2:     pointer to second inode
312  */
313 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
314 {
315 	loff_t isize;
316 	struct ext4_inode_info *ei1;
317 	struct ext4_inode_info *ei2;
318 	unsigned long tmp;
319 	struct timespec64 ts1, ts2;
320 
321 	ei1 = EXT4_I(inode1);
322 	ei2 = EXT4_I(inode2);
323 
324 	swap(inode1->i_version, inode2->i_version);
325 
326 	ts1 = inode_get_atime(inode1);
327 	ts2 = inode_get_atime(inode2);
328 	inode_set_atime_to_ts(inode1, ts2);
329 	inode_set_atime_to_ts(inode2, ts1);
330 
331 	ts1 = inode_get_mtime(inode1);
332 	ts2 = inode_get_mtime(inode2);
333 	inode_set_mtime_to_ts(inode1, ts2);
334 	inode_set_mtime_to_ts(inode2, ts1);
335 
336 	memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
337 	tmp = ei1->i_flags & EXT4_FL_SHOULD_SWAP;
338 	ei1->i_flags = (ei2->i_flags & EXT4_FL_SHOULD_SWAP) |
339 		(ei1->i_flags & ~EXT4_FL_SHOULD_SWAP);
340 	ei2->i_flags = tmp | (ei2->i_flags & ~EXT4_FL_SHOULD_SWAP);
341 	swap(ei1->i_disksize, ei2->i_disksize);
342 	ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
343 	ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
344 
345 	isize = i_size_read(inode1);
346 	i_size_write(inode1, i_size_read(inode2));
347 	i_size_write(inode2, isize);
348 }
349 
350 void ext4_reset_inode_seed(struct inode *inode)
351 {
352 	struct ext4_inode_info *ei = EXT4_I(inode);
353 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
354 	__le32 inum = cpu_to_le32(inode->i_ino);
355 	__le32 gen = cpu_to_le32(inode->i_generation);
356 	__u32 csum;
357 
358 	if (!ext4_has_feature_metadata_csum(inode->i_sb))
359 		return;
360 
361 	csum = ext4_chksum(sbi->s_csum_seed, (__u8 *)&inum, sizeof(inum));
362 	ei->i_csum_seed = ext4_chksum(csum, (__u8 *)&gen, sizeof(gen));
363 }
364 
365 /*
366  * Swap the information from the given @inode and the inode
367  * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
368  * important fields of the inodes.
369  *
370  * @sb:         the super block of the filesystem
371  * @idmap:	idmap of the mount the inode was found from
372  * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
373  *
374  */
375 static long swap_inode_boot_loader(struct super_block *sb,
376 				struct mnt_idmap *idmap,
377 				struct inode *inode)
378 {
379 	handle_t *handle;
380 	int err;
381 	struct inode *inode_bl;
382 	struct ext4_inode_info *ei_bl;
383 	qsize_t size, size_bl, diff;
384 	blkcnt_t blocks;
385 	unsigned short bytes;
386 
387 	inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO,
388 			EXT4_IGET_SPECIAL | EXT4_IGET_BAD);
389 	if (IS_ERR(inode_bl))
390 		return PTR_ERR(inode_bl);
391 	ei_bl = EXT4_I(inode_bl);
392 
393 	/* Protect orig inodes against a truncate and make sure,
394 	 * that only 1 swap_inode_boot_loader is running. */
395 	lock_two_nondirectories(inode, inode_bl);
396 
397 	if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode) ||
398 	    IS_SWAPFILE(inode) || IS_ENCRYPTED(inode) ||
399 	    (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL) ||
400 	    ext4_has_inline_data(inode)) {
401 		err = -EINVAL;
402 		goto journal_err_out;
403 	}
404 
405 	if (IS_RDONLY(inode) || IS_APPEND(inode) || IS_IMMUTABLE(inode) ||
406 	    !inode_owner_or_capable(idmap, inode) ||
407 	    !capable(CAP_SYS_ADMIN)) {
408 		err = -EPERM;
409 		goto journal_err_out;
410 	}
411 
412 	filemap_invalidate_lock(inode->i_mapping);
413 	err = filemap_write_and_wait(inode->i_mapping);
414 	if (err)
415 		goto err_out;
416 
417 	err = filemap_write_and_wait(inode_bl->i_mapping);
418 	if (err)
419 		goto err_out;
420 
421 	/* Wait for all existing dio workers */
422 	inode_dio_wait(inode);
423 	inode_dio_wait(inode_bl);
424 
425 	truncate_inode_pages(&inode->i_data, 0);
426 	truncate_inode_pages(&inode_bl->i_data, 0);
427 
428 	handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
429 	if (IS_ERR(handle)) {
430 		err = -EINVAL;
431 		goto err_out;
432 	}
433 	ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_SWAP_BOOT, handle);
434 
435 	/* Protect extent tree against block allocations via delalloc */
436 	ext4_double_down_write_data_sem(inode, inode_bl);
437 
438 	if (is_bad_inode(inode_bl) || !S_ISREG(inode_bl->i_mode)) {
439 		/* this inode has never been used as a BOOT_LOADER */
440 		set_nlink(inode_bl, 1);
441 		i_uid_write(inode_bl, 0);
442 		i_gid_write(inode_bl, 0);
443 		inode_bl->i_flags = 0;
444 		ei_bl->i_flags = 0;
445 		inode_set_iversion(inode_bl, 1);
446 		i_size_write(inode_bl, 0);
447 		EXT4_I(inode_bl)->i_disksize = inode_bl->i_size;
448 		inode_bl->i_mode = S_IFREG;
449 		if (ext4_has_feature_extents(sb)) {
450 			ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
451 			ext4_ext_tree_init(handle, inode_bl);
452 		} else
453 			memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
454 	}
455 
456 	err = dquot_initialize(inode);
457 	if (err)
458 		goto err_out1;
459 
460 	size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
461 	size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
462 	diff = size - size_bl;
463 	swap_inode_data(inode, inode_bl);
464 
465 	inode_set_ctime_current(inode);
466 	inode_set_ctime_current(inode_bl);
467 	inode_inc_iversion(inode);
468 
469 	inode->i_generation = get_random_u32();
470 	inode_bl->i_generation = get_random_u32();
471 	ext4_reset_inode_seed(inode);
472 	ext4_reset_inode_seed(inode_bl);
473 
474 	ext4_discard_preallocations(inode);
475 
476 	err = ext4_mark_inode_dirty(handle, inode);
477 	if (err < 0) {
478 		/* No need to update quota information. */
479 		ext4_warning(inode->i_sb,
480 			"couldn't mark inode #%llu dirty (err %d)",
481 			inode->i_ino, err);
482 		/* Revert all changes: */
483 		swap_inode_data(inode, inode_bl);
484 		ext4_mark_inode_dirty(handle, inode);
485 		goto err_out1;
486 	}
487 
488 	blocks = inode_bl->i_blocks;
489 	bytes = inode_bl->i_bytes;
490 	inode_bl->i_blocks = inode->i_blocks;
491 	inode_bl->i_bytes = inode->i_bytes;
492 	err = ext4_mark_inode_dirty(handle, inode_bl);
493 	if (err < 0) {
494 		/* No need to update quota information. */
495 		ext4_warning(inode_bl->i_sb,
496 			"couldn't mark inode #%llu dirty (err %d)",
497 			inode_bl->i_ino, err);
498 		goto revert;
499 	}
500 
501 	/* Bootloader inode should not be counted into quota information. */
502 	if (diff > 0)
503 		dquot_free_space(inode, diff);
504 	else
505 		err = dquot_alloc_space(inode, -1 * diff);
506 
507 	if (err < 0) {
508 revert:
509 		/* Revert all changes: */
510 		inode_bl->i_blocks = blocks;
511 		inode_bl->i_bytes = bytes;
512 		swap_inode_data(inode, inode_bl);
513 		ext4_mark_inode_dirty(handle, inode);
514 		ext4_mark_inode_dirty(handle, inode_bl);
515 	}
516 
517 err_out1:
518 	ext4_journal_stop(handle);
519 	ext4_double_up_write_data_sem(inode, inode_bl);
520 
521 err_out:
522 	filemap_invalidate_unlock(inode->i_mapping);
523 journal_err_out:
524 	unlock_two_nondirectories(inode, inode_bl);
525 	iput(inode_bl);
526 	return err;
527 }
528 
529 /*
530  * If immutable is set and we are not clearing it, we're not allowed to change
531  * anything else in the inode.  Don't error out if we're only trying to set
532  * immutable on an immutable file.
533  */
534 static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
535 				      unsigned int flags)
536 {
537 	struct ext4_inode_info *ei = EXT4_I(inode);
538 	unsigned int oldflags = ei->i_flags;
539 
540 	if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
541 		return 0;
542 
543 	if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
544 		return -EPERM;
545 	if (ext4_has_feature_project(inode->i_sb) &&
546 	    __kprojid_val(ei->i_projid) != new_projid)
547 		return -EPERM;
548 
549 	return 0;
550 }
551 
552 static void ext4_dax_dontcache(struct inode *inode, unsigned int flags)
553 {
554 	struct ext4_inode_info *ei = EXT4_I(inode);
555 
556 	if (S_ISDIR(inode->i_mode))
557 		return;
558 
559 	if (test_opt2(inode->i_sb, DAX_NEVER) ||
560 	    test_opt(inode->i_sb, DAX_ALWAYS))
561 		return;
562 
563 	if ((ei->i_flags ^ flags) & EXT4_DAX_FL)
564 		d_mark_dontcache(inode);
565 }
566 
567 static bool dax_compatible(struct inode *inode, unsigned int oldflags,
568 			   unsigned int flags)
569 {
570 	/* Allow the DAX flag to be changed on inline directories */
571 	if (S_ISDIR(inode->i_mode)) {
572 		flags &= ~EXT4_INLINE_DATA_FL;
573 		oldflags &= ~EXT4_INLINE_DATA_FL;
574 	}
575 
576 	if (flags & EXT4_DAX_FL) {
577 		if ((oldflags & EXT4_DAX_MUT_EXCL) ||
578 		     ext4_test_inode_state(inode,
579 					  EXT4_STATE_VERITY_IN_PROGRESS)) {
580 			return false;
581 		}
582 	}
583 
584 	if ((flags & EXT4_DAX_MUT_EXCL) && (oldflags & EXT4_DAX_FL))
585 			return false;
586 
587 	return true;
588 }
589 
590 static int ext4_ioctl_setflags(struct inode *inode,
591 			       unsigned int flags)
592 {
593 	struct ext4_inode_info *ei = EXT4_I(inode);
594 	handle_t *handle = NULL;
595 	int err = -EPERM, migrate = 0;
596 	struct ext4_iloc iloc;
597 	unsigned int oldflags, mask, i;
598 	struct super_block *sb = inode->i_sb;
599 
600 	/* Is it quota file? Do not allow user to mess with it */
601 	if (ext4_is_quota_file(inode))
602 		goto flags_out;
603 
604 	oldflags = ei->i_flags;
605 	/*
606 	 * The JOURNAL_DATA flag can only be changed by
607 	 * the relevant capability.
608 	 */
609 	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
610 		if (!capable(CAP_SYS_RESOURCE))
611 			goto flags_out;
612 	}
613 
614 	if (!dax_compatible(inode, oldflags, flags)) {
615 		err = -EOPNOTSUPP;
616 		goto flags_out;
617 	}
618 
619 	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
620 		migrate = 1;
621 
622 	if ((flags ^ oldflags) & EXT4_CASEFOLD_FL) {
623 		if (!ext4_has_feature_casefold(sb)) {
624 			err = -EOPNOTSUPP;
625 			goto flags_out;
626 		}
627 
628 		if (!S_ISDIR(inode->i_mode)) {
629 			err = -ENOTDIR;
630 			goto flags_out;
631 		}
632 
633 		if (!ext4_empty_dir(inode)) {
634 			err = -ENOTEMPTY;
635 			goto flags_out;
636 		}
637 	}
638 
639 	/*
640 	 * Wait for all pending directio and then flush all the dirty pages
641 	 * for this file.  The flush marks all the pages readonly, so any
642 	 * subsequent attempt to write to the file (particularly mmap pages)
643 	 * will come through the filesystem and fail.
644 	 */
645 	if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
646 	    (flags & EXT4_IMMUTABLE_FL)) {
647 		inode_dio_wait(inode);
648 		err = filemap_write_and_wait(inode->i_mapping);
649 		if (err)
650 			goto flags_out;
651 	}
652 
653 	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
654 	if (IS_ERR(handle)) {
655 		err = PTR_ERR(handle);
656 		goto flags_out;
657 	}
658 	if (IS_SYNC(inode))
659 		ext4_handle_sync(handle);
660 	err = ext4_reserve_inode_write(handle, inode, &iloc);
661 	if (err)
662 		goto flags_err;
663 
664 	ext4_dax_dontcache(inode, flags);
665 
666 	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
667 		if (!(mask & EXT4_FL_USER_MODIFIABLE))
668 			continue;
669 		/* These flags get special treatment later */
670 		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
671 			continue;
672 		if (mask & flags)
673 			ext4_set_inode_flag(inode, i);
674 		else
675 			ext4_clear_inode_flag(inode, i);
676 	}
677 
678 	ext4_set_inode_flags(inode, false);
679 
680 	inode_set_ctime_current(inode);
681 	inode_inc_iversion(inode);
682 
683 	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
684 flags_err:
685 	ext4_journal_stop(handle);
686 	if (err)
687 		goto flags_out;
688 
689 	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
690 		/*
691 		 * Changes to the journaling mode can cause unsafe changes to
692 		 * S_DAX if the inode is DAX
693 		 */
694 		if (IS_DAX(inode)) {
695 			err = -EBUSY;
696 			goto flags_out;
697 		}
698 
699 		err = ext4_change_inode_journal_flag(inode,
700 						     flags & EXT4_JOURNAL_DATA_FL);
701 		if (err)
702 			goto flags_out;
703 	}
704 	if (migrate) {
705 		if (flags & EXT4_EXTENTS_FL)
706 			err = ext4_ext_migrate(inode);
707 		else
708 			err = ext4_ind_migrate(inode);
709 	}
710 
711 flags_out:
712 	return err;
713 }
714 
715 #ifdef CONFIG_QUOTA
716 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
717 {
718 	struct super_block *sb = inode->i_sb;
719 	struct ext4_inode_info *ei = EXT4_I(inode);
720 	int err, rc;
721 	handle_t *handle;
722 	kprojid_t kprojid;
723 	struct ext4_iloc iloc;
724 	struct ext4_inode *raw_inode;
725 	struct dquot *transfer_to[MAXQUOTAS] = { };
726 
727 	if (!ext4_has_feature_project(sb)) {
728 		if (projid != EXT4_DEF_PROJID)
729 			return -EOPNOTSUPP;
730 		else
731 			return 0;
732 	}
733 
734 	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
735 		return -EOPNOTSUPP;
736 
737 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
738 
739 	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
740 		return 0;
741 
742 	err = -EPERM;
743 	/* Is it quota file? Do not allow user to mess with it */
744 	if (ext4_is_quota_file(inode))
745 		return err;
746 
747 	err = dquot_initialize(inode);
748 	if (err)
749 		return err;
750 
751 	err = ext4_get_inode_loc(inode, &iloc);
752 	if (err)
753 		return err;
754 
755 	raw_inode = ext4_raw_inode(&iloc);
756 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
757 		err = ext4_expand_extra_isize(inode,
758 					      EXT4_SB(sb)->s_want_extra_isize,
759 					      &iloc);
760 		if (err)
761 			return err;
762 	} else {
763 		brelse(iloc.bh);
764 	}
765 
766 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
767 		EXT4_QUOTA_INIT_BLOCKS(sb) +
768 		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
769 	if (IS_ERR(handle))
770 		return PTR_ERR(handle);
771 
772 	err = ext4_reserve_inode_write(handle, inode, &iloc);
773 	if (err)
774 		goto out_stop;
775 
776 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
777 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
778 
779 		/* __dquot_transfer() calls back ext4_get_inode_usage() which
780 		 * counts xattr inode references.
781 		 */
782 		down_read(&EXT4_I(inode)->xattr_sem);
783 		err = __dquot_transfer(inode, transfer_to);
784 		up_read(&EXT4_I(inode)->xattr_sem);
785 		dqput(transfer_to[PRJQUOTA]);
786 		if (err)
787 			goto out_dirty;
788 	}
789 
790 	EXT4_I(inode)->i_projid = kprojid;
791 	inode_set_ctime_current(inode);
792 	inode_inc_iversion(inode);
793 out_dirty:
794 	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
795 	if (!err)
796 		err = rc;
797 out_stop:
798 	ext4_journal_stop(handle);
799 	return err;
800 }
801 #else
802 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
803 {
804 	if (projid != EXT4_DEF_PROJID)
805 		return -EOPNOTSUPP;
806 	return 0;
807 }
808 #endif
809 
810 int ext4_force_shutdown(struct super_block *sb, u32 flags)
811 {
812 	struct ext4_sb_info *sbi = EXT4_SB(sb);
813 	int ret;
814 
815 	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
816 		return -EINVAL;
817 
818 	if (ext4_forced_shutdown(sb))
819 		return 0;
820 
821 	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
822 	trace_ext4_shutdown(sb, flags);
823 
824 	switch (flags) {
825 	case EXT4_GOING_FLAGS_DEFAULT:
826 		ret = bdev_freeze(sb->s_bdev);
827 		if (ret)
828 			return ret;
829 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
830 		bdev_thaw(sb->s_bdev);
831 		break;
832 	case EXT4_GOING_FLAGS_LOGFLUSH:
833 		/*
834 		 * Call ext4_force_commit() before setting EXT4_FLAGS_SHUTDOWN.
835 		 * This is because in data=ordered mode, journal commit
836 		 * triggers data writeback which fails if shutdown is already
837 		 * set, causing the journal to be aborted prematurely before
838 		 * the commit succeeds.
839 		 */
840 		(void) ext4_force_commit(sb);
841 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
842 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
843 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
844 		break;
845 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
846 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
847 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
848 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
849 		break;
850 	default:
851 		return -EINVAL;
852 	}
853 	clear_opt(sb, DISCARD);
854 	fserror_report_shutdown(sb, GFP_KERNEL);
855 	return 0;
856 }
857 
858 static int ext4_ioctl_shutdown(struct super_block *sb, unsigned long arg)
859 {
860 	u32 flags;
861 
862 	if (!capable(CAP_SYS_ADMIN))
863 		return -EPERM;
864 
865 	if (get_user(flags, (__u32 __user *)arg))
866 		return -EFAULT;
867 
868 	return ext4_force_shutdown(sb, flags);
869 }
870 
871 struct getfsmap_info {
872 	struct super_block	*gi_sb;
873 	struct fsmap_head __user *gi_data;
874 	unsigned int		gi_idx;
875 	__u32			gi_last_flags;
876 };
877 
878 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
879 {
880 	struct getfsmap_info *info = priv;
881 	struct fsmap fm;
882 
883 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
884 
885 	info->gi_last_flags = xfm->fmr_flags;
886 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
887 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
888 			sizeof(struct fsmap)))
889 		return -EFAULT;
890 
891 	return 0;
892 }
893 
894 static int ext4_ioc_getfsmap(struct super_block *sb,
895 			     struct fsmap_head __user *arg)
896 {
897 	struct getfsmap_info info = { NULL };
898 	struct ext4_fsmap_head xhead = {0};
899 	struct fsmap_head head;
900 	bool aborted = false;
901 	int error;
902 
903 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
904 		return -EFAULT;
905 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
906 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
907 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
908 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
909 		       sizeof(head.fmh_keys[1].fmr_reserved)))
910 		return -EINVAL;
911 	/*
912 	 * ext4 doesn't report file extents at all, so the only valid
913 	 * file offsets are the magic ones (all zeroes or all ones).
914 	 */
915 	if (head.fmh_keys[0].fmr_offset ||
916 	    (head.fmh_keys[1].fmr_offset != 0 &&
917 	     head.fmh_keys[1].fmr_offset != -1ULL))
918 		return -EINVAL;
919 
920 	xhead.fmh_iflags = head.fmh_iflags;
921 	xhead.fmh_count = head.fmh_count;
922 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
923 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
924 
925 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
926 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
927 
928 	info.gi_sb = sb;
929 	info.gi_data = arg;
930 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
931 	if (error == EXT4_QUERY_RANGE_ABORT)
932 		aborted = true;
933 	else if (error)
934 		return error;
935 
936 	/* If we didn't abort, set the "last" flag in the last fmx */
937 	if (!aborted && info.gi_idx) {
938 		info.gi_last_flags |= FMR_OF_LAST;
939 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
940 				 &info.gi_last_flags,
941 				 sizeof(info.gi_last_flags)))
942 			return -EFAULT;
943 	}
944 
945 	/* copy back header */
946 	head.fmh_entries = xhead.fmh_entries;
947 	head.fmh_oflags = xhead.fmh_oflags;
948 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
949 		return -EFAULT;
950 
951 	return 0;
952 }
953 
954 static long ext4_ioctl_group_add(struct file *file,
955 				 struct ext4_new_group_data *input)
956 {
957 	struct super_block *sb = file_inode(file)->i_sb;
958 	int err, err2=0;
959 
960 	err = ext4_resize_begin(sb);
961 	if (err)
962 		return err;
963 
964 	if (ext4_has_feature_bigalloc(sb)) {
965 		ext4_msg(sb, KERN_ERR,
966 			 "Online resizing not supported with bigalloc");
967 		err = -EOPNOTSUPP;
968 		goto group_add_out;
969 	}
970 
971 	err = mnt_want_write_file(file);
972 	if (err)
973 		goto group_add_out;
974 
975 	err = ext4_group_add(sb, input);
976 	if (EXT4_SB(sb)->s_journal) {
977 		ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE, NULL);
978 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
979 		err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
980 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
981 	}
982 	if (err == 0)
983 		err = err2;
984 	mnt_drop_write_file(file);
985 	if (!err && ext4_has_group_desc_csum(sb) &&
986 	    test_opt(sb, INIT_INODE_TABLE))
987 		err = ext4_register_li_request(sb, input->group);
988 group_add_out:
989 	err2 = ext4_resize_end(sb, false);
990 	if (err == 0)
991 		err = err2;
992 	return err;
993 }
994 
995 int ext4_fileattr_get(struct dentry *dentry, struct file_kattr *fa)
996 {
997 	struct inode *inode = d_inode(dentry);
998 	struct ext4_inode_info *ei = EXT4_I(inode);
999 	u32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
1000 
1001 	if (S_ISREG(inode->i_mode))
1002 		flags &= ~FS_PROJINHERIT_FL;
1003 
1004 	fileattr_fill_flags(fa, flags);
1005 	if (ext4_has_feature_project(inode->i_sb))
1006 		fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
1007 
1008 	return 0;
1009 }
1010 
1011 int ext4_fileattr_set(struct mnt_idmap *idmap,
1012 		      struct dentry *dentry, struct file_kattr *fa)
1013 {
1014 	struct inode *inode = d_inode(dentry);
1015 	u32 flags = fa->flags;
1016 	int err = -EOPNOTSUPP;
1017 
1018 	if (flags & ~EXT4_FL_USER_VISIBLE)
1019 		goto out;
1020 
1021 	/*
1022 	 * chattr(1) grabs flags via GETFLAGS, modifies the result and
1023 	 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
1024 	 * more restrictive than just silently masking off visible but
1025 	 * not settable flags as we always did.
1026 	 */
1027 	flags &= EXT4_FL_USER_MODIFIABLE;
1028 	if (ext4_mask_flags(inode->i_mode, flags) != flags)
1029 		goto out;
1030 	err = ext4_ioctl_check_immutable(inode, fa->fsx_projid, flags);
1031 	if (err)
1032 		goto out;
1033 	err = ext4_ioctl_setflags(inode, flags);
1034 	if (err)
1035 		goto out;
1036 	err = ext4_ioctl_setproject(inode, fa->fsx_projid);
1037 out:
1038 	return err;
1039 }
1040 
1041 /* So that the fiemap access checks can't overflow on 32 bit machines. */
1042 #define FIEMAP_MAX_EXTENTS	(UINT_MAX / sizeof(struct fiemap_extent))
1043 
1044 static int ext4_ioctl_get_es_cache(struct file *filp, unsigned long arg)
1045 {
1046 	struct fiemap fiemap;
1047 	struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
1048 	struct fiemap_extent_info fieinfo = { 0, };
1049 	struct inode *inode = file_inode(filp);
1050 	int error;
1051 
1052 	if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
1053 		return -EFAULT;
1054 
1055 	if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
1056 		return -EINVAL;
1057 
1058 	fieinfo.fi_flags = fiemap.fm_flags;
1059 	fieinfo.fi_extents_max = fiemap.fm_extent_count;
1060 	fieinfo.fi_extents_start = ufiemap->fm_extents;
1061 
1062 	error = ext4_get_es_cache(inode, &fieinfo, fiemap.fm_start,
1063 			fiemap.fm_length);
1064 	fiemap.fm_flags = fieinfo.fi_flags;
1065 	fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
1066 	if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
1067 		error = -EFAULT;
1068 
1069 	return error;
1070 }
1071 
1072 static int ext4_ioctl_checkpoint(struct file *filp, unsigned long arg)
1073 {
1074 	int err = 0;
1075 	__u32 flags = 0;
1076 	unsigned int flush_flags = 0;
1077 	struct super_block *sb = file_inode(filp)->i_sb;
1078 
1079 	if (copy_from_user(&flags, (__u32 __user *)arg,
1080 				sizeof(__u32)))
1081 		return -EFAULT;
1082 
1083 	if (!capable(CAP_SYS_ADMIN))
1084 		return -EPERM;
1085 
1086 	/* check for invalid bits set */
1087 	if ((flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID) ||
1088 				((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1089 				(flags & JBD2_JOURNAL_FLUSH_ZEROOUT)))
1090 		return -EINVAL;
1091 
1092 	if (!EXT4_SB(sb)->s_journal)
1093 		return -ENODEV;
1094 
1095 	if ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1096 	    !bdev_max_discard_sectors(EXT4_SB(sb)->s_journal->j_dev))
1097 		return -EOPNOTSUPP;
1098 
1099 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
1100 		return 0;
1101 
1102 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DISCARD)
1103 		flush_flags |= JBD2_JOURNAL_FLUSH_DISCARD;
1104 
1105 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT) {
1106 		flush_flags |= JBD2_JOURNAL_FLUSH_ZEROOUT;
1107 		pr_info_ratelimited("warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow");
1108 	}
1109 
1110 	jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1111 	err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, flush_flags);
1112 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1113 
1114 	return err;
1115 }
1116 
1117 static int ext4_ioctl_setlabel(struct file *filp, const char __user *user_label)
1118 {
1119 	size_t len;
1120 	int ret = 0;
1121 	char new_label[EXT4_LABEL_MAX + 1];
1122 	struct super_block *sb = file_inode(filp)->i_sb;
1123 
1124 	if (!capable(CAP_SYS_ADMIN))
1125 		return -EPERM;
1126 
1127 	/*
1128 	 * Copy the maximum length allowed for ext4 label with one more to
1129 	 * find the required terminating null byte in order to test the
1130 	 * label length. The on disk label doesn't need to be null terminated.
1131 	 */
1132 	if (copy_from_user(new_label, user_label, EXT4_LABEL_MAX + 1))
1133 		return -EFAULT;
1134 
1135 	len = strnlen(new_label, EXT4_LABEL_MAX + 1);
1136 	if (len > EXT4_LABEL_MAX)
1137 		return -EINVAL;
1138 
1139 	/*
1140 	 * Clear the buffer after the new label
1141 	 */
1142 	memset(new_label + len, 0, EXT4_LABEL_MAX - len);
1143 
1144 	ret = mnt_want_write_file(filp);
1145 	if (ret)
1146 		return ret;
1147 
1148 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setlabel, new_label);
1149 
1150 	mnt_drop_write_file(filp);
1151 	return ret;
1152 }
1153 
1154 static int ext4_ioctl_getlabel(struct ext4_sb_info *sbi, char __user *user_label)
1155 {
1156 	char label[EXT4_LABEL_MAX + 1];
1157 
1158 	/*
1159 	 * EXT4_LABEL_MAX must always be smaller than FSLABEL_MAX because
1160 	 * FSLABEL_MAX must include terminating null byte, while s_volume_name
1161 	 * does not have to.
1162 	 */
1163 	BUILD_BUG_ON(EXT4_LABEL_MAX >= FSLABEL_MAX);
1164 
1165 	lock_buffer(sbi->s_sbh);
1166 	memtostr_pad(label, sbi->s_es->s_volume_name);
1167 	unlock_buffer(sbi->s_sbh);
1168 
1169 	if (copy_to_user(user_label, label, sizeof(label)))
1170 		return -EFAULT;
1171 	return 0;
1172 }
1173 
1174 static int ext4_ioctl_getuuid(struct ext4_sb_info *sbi,
1175 			struct fsuuid __user *ufsuuid)
1176 {
1177 	struct fsuuid fsuuid;
1178 	__u8 uuid[UUID_SIZE];
1179 
1180 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1181 		return -EFAULT;
1182 
1183 	if (fsuuid.fsu_len == 0) {
1184 		fsuuid.fsu_len = UUID_SIZE;
1185 		if (copy_to_user(&ufsuuid->fsu_len, &fsuuid.fsu_len,
1186 					sizeof(fsuuid.fsu_len)))
1187 			return -EFAULT;
1188 		return 0;
1189 	}
1190 
1191 	if (fsuuid.fsu_len < UUID_SIZE || fsuuid.fsu_flags != 0)
1192 		return -EINVAL;
1193 
1194 	lock_buffer(sbi->s_sbh);
1195 	memcpy(uuid, sbi->s_es->s_uuid, UUID_SIZE);
1196 	unlock_buffer(sbi->s_sbh);
1197 
1198 	fsuuid.fsu_len = UUID_SIZE;
1199 	if (copy_to_user(ufsuuid, &fsuuid, sizeof(fsuuid)) ||
1200 	    copy_to_user(&ufsuuid->fsu_uuid[0], uuid, UUID_SIZE))
1201 		return -EFAULT;
1202 	return 0;
1203 }
1204 
1205 static int ext4_ioctl_setuuid(struct file *filp,
1206 			const struct fsuuid __user *ufsuuid)
1207 {
1208 	int ret = 0;
1209 	struct super_block *sb = file_inode(filp)->i_sb;
1210 	struct fsuuid fsuuid;
1211 	__u8 uuid[UUID_SIZE];
1212 
1213 	if (!capable(CAP_SYS_ADMIN))
1214 		return -EPERM;
1215 
1216 	/*
1217 	 * If any checksums (group descriptors or metadata) are being used
1218 	 * then the checksum seed feature is required to change the UUID.
1219 	 */
1220 	if (((ext4_has_feature_gdt_csum(sb) ||
1221 	      ext4_has_feature_metadata_csum(sb))
1222 			&& !ext4_has_feature_csum_seed(sb))
1223 		|| ext4_has_feature_stable_inodes(sb))
1224 		return -EOPNOTSUPP;
1225 
1226 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1227 		return -EFAULT;
1228 
1229 	if (fsuuid.fsu_len != UUID_SIZE || fsuuid.fsu_flags != 0)
1230 		return -EINVAL;
1231 
1232 	if (copy_from_user(uuid, &ufsuuid->fsu_uuid[0], UUID_SIZE))
1233 		return -EFAULT;
1234 
1235 	ret = mnt_want_write_file(filp);
1236 	if (ret)
1237 		return ret;
1238 
1239 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setuuid, &uuid);
1240 	mnt_drop_write_file(filp);
1241 
1242 	return ret;
1243 }
1244 
1245 
1246 #define TUNE_OPS_SUPPORTED (EXT4_TUNE_FL_ERRORS_BEHAVIOR |    \
1247 	EXT4_TUNE_FL_MNT_COUNT | EXT4_TUNE_FL_MAX_MNT_COUNT | \
1248 	EXT4_TUNE_FL_CHECKINTRVAL | EXT4_TUNE_FL_LAST_CHECK_TIME | \
1249 	EXT4_TUNE_FL_RESERVED_BLOCKS | EXT4_TUNE_FL_RESERVED_UID | \
1250 	EXT4_TUNE_FL_RESERVED_GID | EXT4_TUNE_FL_DEFAULT_MNT_OPTS | \
1251 	EXT4_TUNE_FL_DEF_HASH_ALG | EXT4_TUNE_FL_RAID_STRIDE | \
1252 	EXT4_TUNE_FL_RAID_STRIPE_WIDTH | EXT4_TUNE_FL_MOUNT_OPTS | \
1253 	EXT4_TUNE_FL_FEATURES | EXT4_TUNE_FL_EDIT_FEATURES | \
1254 	EXT4_TUNE_FL_FORCE_FSCK | EXT4_TUNE_FL_ENCODING | \
1255 	EXT4_TUNE_FL_ENCODING_FLAGS)
1256 
1257 #define EXT4_TUNE_SET_COMPAT_SUPP \
1258 		(EXT4_FEATURE_COMPAT_DIR_INDEX |	\
1259 		 EXT4_FEATURE_COMPAT_STABLE_INODES)
1260 #define EXT4_TUNE_SET_INCOMPAT_SUPP \
1261 		(EXT4_FEATURE_INCOMPAT_EXTENTS |	\
1262 		 EXT4_FEATURE_INCOMPAT_EA_INODE |	\
1263 		 EXT4_FEATURE_INCOMPAT_ENCRYPT |	\
1264 		 EXT4_FEATURE_INCOMPAT_CSUM_SEED |	\
1265 		 EXT4_FEATURE_INCOMPAT_LARGEDIR |	\
1266 		 EXT4_FEATURE_INCOMPAT_CASEFOLD)
1267 #define EXT4_TUNE_SET_RO_COMPAT_SUPP \
1268 		(EXT4_FEATURE_RO_COMPAT_LARGE_FILE |	\
1269 		 EXT4_FEATURE_RO_COMPAT_DIR_NLINK |	\
1270 		 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE |	\
1271 		 EXT4_FEATURE_RO_COMPAT_PROJECT |	\
1272 		 EXT4_FEATURE_RO_COMPAT_VERITY)
1273 
1274 #define EXT4_TUNE_CLEAR_COMPAT_SUPP (0)
1275 #define EXT4_TUNE_CLEAR_INCOMPAT_SUPP (0)
1276 #define EXT4_TUNE_CLEAR_RO_COMPAT_SUPP (0)
1277 
1278 #define SB_ENC_SUPP_MASK (SB_ENC_STRICT_MODE_FL |	\
1279 			  SB_ENC_NO_COMPAT_FALLBACK_FL)
1280 
1281 static int ext4_ioctl_get_tune_sb(struct ext4_sb_info *sbi,
1282 				  struct ext4_tune_sb_params __user *params)
1283 {
1284 	struct ext4_tune_sb_params ret;
1285 	struct ext4_super_block *es = sbi->s_es;
1286 
1287 	memset(&ret, 0, sizeof(ret));
1288 	ret.set_flags = TUNE_OPS_SUPPORTED;
1289 	ret.errors_behavior = le16_to_cpu(es->s_errors);
1290 	ret.mnt_count = le16_to_cpu(es->s_mnt_count);
1291 	ret.max_mnt_count = le16_to_cpu(es->s_max_mnt_count);
1292 	ret.checkinterval = le32_to_cpu(es->s_checkinterval);
1293 	ret.last_check_time = le32_to_cpu(es->s_lastcheck);
1294 	ret.reserved_blocks = ext4_r_blocks_count(es);
1295 	ret.blocks_count = ext4_blocks_count(es);
1296 	ret.reserved_uid = ext4_get_resuid(es);
1297 	ret.reserved_gid = ext4_get_resgid(es);
1298 	ret.default_mnt_opts = le32_to_cpu(es->s_default_mount_opts);
1299 	ret.def_hash_alg = es->s_def_hash_version;
1300 	ret.raid_stride = le16_to_cpu(es->s_raid_stride);
1301 	ret.raid_stripe_width = le32_to_cpu(es->s_raid_stripe_width);
1302 	ret.encoding = le16_to_cpu(es->s_encoding);
1303 	ret.encoding_flags = le16_to_cpu(es->s_encoding_flags);
1304 	strscpy_pad(ret.mount_opts, es->s_mount_opts);
1305 	ret.feature_compat = le32_to_cpu(es->s_feature_compat);
1306 	ret.feature_incompat = le32_to_cpu(es->s_feature_incompat);
1307 	ret.feature_ro_compat = le32_to_cpu(es->s_feature_ro_compat);
1308 	ret.set_feature_compat_mask = EXT4_TUNE_SET_COMPAT_SUPP;
1309 	ret.set_feature_incompat_mask = EXT4_TUNE_SET_INCOMPAT_SUPP;
1310 	ret.set_feature_ro_compat_mask = EXT4_TUNE_SET_RO_COMPAT_SUPP;
1311 	ret.clear_feature_compat_mask = EXT4_TUNE_CLEAR_COMPAT_SUPP;
1312 	ret.clear_feature_incompat_mask = EXT4_TUNE_CLEAR_INCOMPAT_SUPP;
1313 	ret.clear_feature_ro_compat_mask = EXT4_TUNE_CLEAR_RO_COMPAT_SUPP;
1314 	if (copy_to_user(params, &ret, sizeof(ret)))
1315 		return -EFAULT;
1316 	return 0;
1317 }
1318 
1319 static void ext4_sb_setparams(struct ext4_sb_info *sbi,
1320 			      struct ext4_super_block *es, const void *arg)
1321 {
1322 	const struct ext4_tune_sb_params *params = arg;
1323 
1324 	if (params->set_flags & EXT4_TUNE_FL_ERRORS_BEHAVIOR)
1325 		es->s_errors = cpu_to_le16(params->errors_behavior);
1326 	if (params->set_flags & EXT4_TUNE_FL_MNT_COUNT)
1327 		es->s_mnt_count = cpu_to_le16(params->mnt_count);
1328 	if (params->set_flags & EXT4_TUNE_FL_MAX_MNT_COUNT)
1329 		es->s_max_mnt_count = cpu_to_le16(params->max_mnt_count);
1330 	if (params->set_flags & EXT4_TUNE_FL_CHECKINTRVAL)
1331 		es->s_checkinterval = cpu_to_le32(params->checkinterval);
1332 	if (params->set_flags & EXT4_TUNE_FL_LAST_CHECK_TIME)
1333 		es->s_lastcheck = cpu_to_le32(params->last_check_time);
1334 	if (params->set_flags & EXT4_TUNE_FL_RESERVED_BLOCKS) {
1335 		ext4_fsblk_t blk = params->reserved_blocks;
1336 
1337 		es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1338 		es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1339 	}
1340 	if (params->set_flags & EXT4_TUNE_FL_RESERVED_UID) {
1341 		int uid = params->reserved_uid;
1342 
1343 		es->s_def_resuid = cpu_to_le16(uid & 0xFFFF);
1344 		es->s_def_resuid_hi = cpu_to_le16(uid >> 16);
1345 	}
1346 	if (params->set_flags & EXT4_TUNE_FL_RESERVED_GID) {
1347 		int gid = params->reserved_gid;
1348 
1349 		es->s_def_resgid = cpu_to_le16(gid & 0xFFFF);
1350 		es->s_def_resgid_hi = cpu_to_le16(gid >> 16);
1351 	}
1352 	if (params->set_flags & EXT4_TUNE_FL_DEFAULT_MNT_OPTS)
1353 		es->s_default_mount_opts = cpu_to_le32(params->default_mnt_opts);
1354 	if (params->set_flags & EXT4_TUNE_FL_DEF_HASH_ALG)
1355 		es->s_def_hash_version = params->def_hash_alg;
1356 	if (params->set_flags & EXT4_TUNE_FL_RAID_STRIDE)
1357 		es->s_raid_stride = cpu_to_le16(params->raid_stride);
1358 	if (params->set_flags & EXT4_TUNE_FL_RAID_STRIPE_WIDTH)
1359 		es->s_raid_stripe_width =
1360 			cpu_to_le32(params->raid_stripe_width);
1361 	if (params->set_flags & EXT4_TUNE_FL_ENCODING)
1362 		es->s_encoding = cpu_to_le16(params->encoding);
1363 	if (params->set_flags & EXT4_TUNE_FL_ENCODING_FLAGS)
1364 		es->s_encoding_flags = cpu_to_le16(params->encoding_flags);
1365 	strscpy_pad(es->s_mount_opts, params->mount_opts);
1366 	if (params->set_flags & EXT4_TUNE_FL_EDIT_FEATURES) {
1367 		es->s_feature_compat |=
1368 			cpu_to_le32(params->set_feature_compat_mask);
1369 		es->s_feature_incompat |=
1370 			cpu_to_le32(params->set_feature_incompat_mask);
1371 		es->s_feature_ro_compat |=
1372 			cpu_to_le32(params->set_feature_ro_compat_mask);
1373 		es->s_feature_compat &=
1374 			~cpu_to_le32(params->clear_feature_compat_mask);
1375 		es->s_feature_incompat &=
1376 			~cpu_to_le32(params->clear_feature_incompat_mask);
1377 		es->s_feature_ro_compat &=
1378 			~cpu_to_le32(params->clear_feature_ro_compat_mask);
1379 		if (params->set_feature_compat_mask &
1380 		    EXT4_FEATURE_COMPAT_DIR_INDEX)
1381 			es->s_def_hash_version = sbi->s_def_hash_version;
1382 		if (params->set_feature_incompat_mask &
1383 		    EXT4_FEATURE_INCOMPAT_CSUM_SEED)
1384 			es->s_checksum_seed = cpu_to_le32(sbi->s_csum_seed);
1385 	}
1386 	if (params->set_flags & EXT4_TUNE_FL_FORCE_FSCK)
1387 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
1388 }
1389 
1390 static int ext4_ioctl_set_tune_sb(struct file *filp,
1391 				  struct ext4_tune_sb_params __user *in)
1392 {
1393 	struct ext4_tune_sb_params params;
1394 	struct super_block *sb = file_inode(filp)->i_sb;
1395 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1396 	struct ext4_super_block *es = sbi->s_es;
1397 	int enabling_casefold = 0;
1398 	int ret;
1399 
1400 	if (!capable(CAP_SYS_ADMIN))
1401 		return -EPERM;
1402 
1403 	if (copy_from_user(&params, in, sizeof(params)))
1404 		return -EFAULT;
1405 
1406 	if (strnlen(params.mount_opts, sizeof(params.mount_opts)) ==
1407 	    sizeof(params.mount_opts))
1408 		return -E2BIG;
1409 
1410 	if ((params.set_flags & ~TUNE_OPS_SUPPORTED) != 0)
1411 		return -EOPNOTSUPP;
1412 
1413 	if ((params.set_flags & EXT4_TUNE_FL_ERRORS_BEHAVIOR) &&
1414 	    (params.errors_behavior > EXT4_ERRORS_PANIC))
1415 		return -EINVAL;
1416 
1417 	if ((params.set_flags & EXT4_TUNE_FL_RESERVED_BLOCKS) &&
1418 	    (params.reserved_blocks > ext4_blocks_count(sbi->s_es) / 2))
1419 		return -EINVAL;
1420 	if ((params.set_flags & EXT4_TUNE_FL_DEF_HASH_ALG) &&
1421 	    ((params.def_hash_alg > DX_HASH_LAST) ||
1422 	     (params.def_hash_alg == DX_HASH_SIPHASH)))
1423 		return -EINVAL;
1424 	if ((params.set_flags & EXT4_TUNE_FL_FEATURES) &&
1425 	    (params.set_flags & EXT4_TUNE_FL_EDIT_FEATURES))
1426 		return -EINVAL;
1427 
1428 	if (params.set_flags & EXT4_TUNE_FL_FEATURES) {
1429 		params.set_feature_compat_mask =
1430 			params.feature_compat &
1431 			~le32_to_cpu(es->s_feature_compat);
1432 		params.set_feature_incompat_mask =
1433 			params.feature_incompat &
1434 			~le32_to_cpu(es->s_feature_incompat);
1435 		params.set_feature_ro_compat_mask =
1436 			params.feature_ro_compat &
1437 			~le32_to_cpu(es->s_feature_ro_compat);
1438 		params.clear_feature_compat_mask =
1439 			~params.feature_compat &
1440 			le32_to_cpu(es->s_feature_compat);
1441 		params.clear_feature_incompat_mask =
1442 			~params.feature_incompat &
1443 			le32_to_cpu(es->s_feature_incompat);
1444 		params.clear_feature_ro_compat_mask =
1445 			~params.feature_ro_compat &
1446 			le32_to_cpu(es->s_feature_ro_compat);
1447 		params.set_flags |= EXT4_TUNE_FL_EDIT_FEATURES;
1448 	}
1449 	if (params.set_flags & EXT4_TUNE_FL_EDIT_FEATURES) {
1450 		if ((params.set_feature_compat_mask &
1451 		     ~EXT4_TUNE_SET_COMPAT_SUPP) ||
1452 		    (params.set_feature_incompat_mask &
1453 		     ~EXT4_TUNE_SET_INCOMPAT_SUPP) ||
1454 		    (params.set_feature_ro_compat_mask &
1455 		     ~EXT4_TUNE_SET_RO_COMPAT_SUPP) ||
1456 		    (params.clear_feature_compat_mask &
1457 		     ~EXT4_TUNE_CLEAR_COMPAT_SUPP) ||
1458 		    (params.clear_feature_incompat_mask &
1459 		     ~EXT4_TUNE_CLEAR_INCOMPAT_SUPP) ||
1460 		    (params.clear_feature_ro_compat_mask &
1461 		     ~EXT4_TUNE_CLEAR_RO_COMPAT_SUPP))
1462 			return -EOPNOTSUPP;
1463 
1464 		/*
1465 		 * Filter out the features that are already set from
1466 		 * the set_mask.
1467 		 */
1468 		params.set_feature_compat_mask &=
1469 			~le32_to_cpu(es->s_feature_compat);
1470 		params.set_feature_incompat_mask &=
1471 			~le32_to_cpu(es->s_feature_incompat);
1472 		params.set_feature_ro_compat_mask &=
1473 			~le32_to_cpu(es->s_feature_ro_compat);
1474 		if ((params.set_feature_incompat_mask &
1475 		     EXT4_FEATURE_INCOMPAT_CASEFOLD)) {
1476 			enabling_casefold = 1;
1477 			if (!(params.set_flags & EXT4_TUNE_FL_ENCODING)) {
1478 				params.encoding = EXT4_ENC_UTF8_12_1;
1479 				params.set_flags |= EXT4_TUNE_FL_ENCODING;
1480 			}
1481 			if (!(params.set_flags & EXT4_TUNE_FL_ENCODING_FLAGS)) {
1482 				params.encoding_flags = 0;
1483 				params.set_flags |= EXT4_TUNE_FL_ENCODING_FLAGS;
1484 			}
1485 		}
1486 		if ((params.set_feature_compat_mask &
1487 		     EXT4_FEATURE_COMPAT_DIR_INDEX)) {
1488 			uuid_t	uu;
1489 
1490 			memcpy(&uu, sbi->s_hash_seed, UUID_SIZE);
1491 			if (uuid_is_null(&uu))
1492 				generate_random_uuid((char *)
1493 						     &sbi->s_hash_seed);
1494 			if (params.set_flags & EXT4_TUNE_FL_DEF_HASH_ALG)
1495 				sbi->s_def_hash_version = params.def_hash_alg;
1496 			else if (sbi->s_def_hash_version == 0)
1497 				sbi->s_def_hash_version = DX_HASH_HALF_MD4;
1498 			if (!(es->s_flags &
1499 			      cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH)) &&
1500 			    !(es->s_flags &
1501 			      cpu_to_le32(EXT2_FLAGS_SIGNED_HASH))) {
1502 #ifdef __CHAR_UNSIGNED__
1503 				sbi->s_hash_unsigned = 3;
1504 #else
1505 				sbi->s_hash_unsigned = 0;
1506 #endif
1507 			}
1508 		}
1509 	}
1510 	if (params.set_flags & EXT4_TUNE_FL_ENCODING) {
1511 		if (!enabling_casefold)
1512 			return -EINVAL;
1513 		if (params.encoding == 0)
1514 			params.encoding = EXT4_ENC_UTF8_12_1;
1515 		else if (params.encoding != EXT4_ENC_UTF8_12_1)
1516 			return -EINVAL;
1517 	}
1518 	if (params.set_flags & EXT4_TUNE_FL_ENCODING_FLAGS) {
1519 		if (!enabling_casefold)
1520 			return -EINVAL;
1521 		if (params.encoding_flags & ~SB_ENC_SUPP_MASK)
1522 			return -EINVAL;
1523 	}
1524 
1525 	ret = mnt_want_write_file(filp);
1526 	if (ret)
1527 		return ret;
1528 
1529 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setparams, &params);
1530 	mnt_drop_write_file(filp);
1531 
1532 	if (params.set_flags & EXT4_TUNE_FL_DEF_HASH_ALG)
1533 		sbi->s_def_hash_version = params.def_hash_alg;
1534 
1535 	return ret;
1536 }
1537 
1538 static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1539 {
1540 	struct inode *inode = file_inode(filp);
1541 	struct super_block *sb = inode->i_sb;
1542 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
1543 
1544 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
1545 
1546 	switch (cmd) {
1547 	case FS_IOC_GETFSMAP:
1548 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
1549 	case EXT4_IOC_GETVERSION:
1550 	case EXT4_IOC_GETVERSION_OLD:
1551 		return put_user(inode->i_generation, (int __user *) arg);
1552 	case EXT4_IOC_SETVERSION:
1553 	case EXT4_IOC_SETVERSION_OLD: {
1554 		handle_t *handle;
1555 		struct ext4_iloc iloc;
1556 		__u32 generation;
1557 		int err;
1558 
1559 		if (!inode_owner_or_capable(idmap, inode))
1560 			return -EPERM;
1561 
1562 		if (ext4_has_feature_metadata_csum(inode->i_sb)) {
1563 			ext4_warning(sb, "Setting inode version is not "
1564 				     "supported with metadata_csum enabled.");
1565 			return -ENOTTY;
1566 		}
1567 
1568 		err = mnt_want_write_file(filp);
1569 		if (err)
1570 			return err;
1571 		if (get_user(generation, (int __user *) arg)) {
1572 			err = -EFAULT;
1573 			goto setversion_out;
1574 		}
1575 
1576 		inode_lock(inode);
1577 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
1578 		if (IS_ERR(handle)) {
1579 			err = PTR_ERR(handle);
1580 			goto unlock_out;
1581 		}
1582 		err = ext4_reserve_inode_write(handle, inode, &iloc);
1583 		if (err == 0) {
1584 			inode_set_ctime_current(inode);
1585 			inode_inc_iversion(inode);
1586 			inode->i_generation = generation;
1587 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
1588 		}
1589 		ext4_journal_stop(handle);
1590 
1591 unlock_out:
1592 		inode_unlock(inode);
1593 setversion_out:
1594 		mnt_drop_write_file(filp);
1595 		return err;
1596 	}
1597 	case EXT4_IOC_GROUP_EXTEND: {
1598 		ext4_fsblk_t n_blocks_count;
1599 		int err, err2=0;
1600 
1601 		err = ext4_resize_begin(sb);
1602 		if (err)
1603 			return err;
1604 
1605 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
1606 			err = -EFAULT;
1607 			goto group_extend_out;
1608 		}
1609 
1610 		if (ext4_has_feature_bigalloc(sb)) {
1611 			ext4_msg(sb, KERN_ERR,
1612 				 "Online resizing not supported with bigalloc");
1613 			err = -EOPNOTSUPP;
1614 			goto group_extend_out;
1615 		}
1616 
1617 		err = mnt_want_write_file(filp);
1618 		if (err)
1619 			goto group_extend_out;
1620 
1621 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
1622 		if (EXT4_SB(sb)->s_journal) {
1623 			ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE,
1624 						NULL);
1625 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1626 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1627 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1628 		}
1629 		if (err == 0)
1630 			err = err2;
1631 		mnt_drop_write_file(filp);
1632 group_extend_out:
1633 		err2 = ext4_resize_end(sb, false);
1634 		if (err == 0)
1635 			err = err2;
1636 		return err;
1637 	}
1638 
1639 	case EXT4_IOC_MOVE_EXT: {
1640 		struct move_extent me;
1641 		int err;
1642 
1643 		if (!(filp->f_mode & FMODE_READ) ||
1644 		    !(filp->f_mode & FMODE_WRITE))
1645 			return -EBADF;
1646 
1647 		if (copy_from_user(&me,
1648 			(struct move_extent __user *)arg, sizeof(me)))
1649 			return -EFAULT;
1650 		me.moved_len = 0;
1651 
1652 		CLASS(fd, donor)(me.donor_fd);
1653 		if (fd_empty(donor))
1654 			return -EBADF;
1655 
1656 		if (!(fd_file(donor)->f_mode & FMODE_WRITE))
1657 			return -EBADF;
1658 
1659 		if (file_inode(filp)->i_sb != file_inode(fd_file(donor))->i_sb)
1660 			return -EXDEV;
1661 
1662 		err = mnt_want_write_file(filp);
1663 		if (err)
1664 			return err;
1665 
1666 		err = ext4_move_extents(filp, fd_file(donor), me.orig_start,
1667 					me.donor_start, me.len, &me.moved_len);
1668 		mnt_drop_write_file(filp);
1669 
1670 		if (copy_to_user((struct move_extent __user *)arg,
1671 				 &me, sizeof(me)))
1672 			err = -EFAULT;
1673 		return err;
1674 	}
1675 
1676 	case EXT4_IOC_GROUP_ADD: {
1677 		struct ext4_new_group_data input;
1678 
1679 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
1680 				sizeof(input)))
1681 			return -EFAULT;
1682 
1683 		return ext4_ioctl_group_add(filp, &input);
1684 	}
1685 
1686 	case EXT4_IOC_MIGRATE:
1687 	{
1688 		int err;
1689 		if (!inode_owner_or_capable(idmap, inode))
1690 			return -EACCES;
1691 
1692 		err = mnt_want_write_file(filp);
1693 		if (err)
1694 			return err;
1695 		/*
1696 		 * inode_mutex prevent write and truncate on the file.
1697 		 * Read still goes through. We take i_data_sem in
1698 		 * ext4_ext_swap_inode_data before we switch the
1699 		 * inode format to prevent read.
1700 		 */
1701 		inode_lock((inode));
1702 		err = ext4_ext_migrate(inode);
1703 		inode_unlock((inode));
1704 		mnt_drop_write_file(filp);
1705 		return err;
1706 	}
1707 
1708 	case EXT4_IOC_ALLOC_DA_BLKS:
1709 	{
1710 		int err;
1711 		if (!inode_owner_or_capable(idmap, inode))
1712 			return -EACCES;
1713 
1714 		err = mnt_want_write_file(filp);
1715 		if (err)
1716 			return err;
1717 		err = ext4_alloc_da_blocks(inode);
1718 		mnt_drop_write_file(filp);
1719 		return err;
1720 	}
1721 
1722 	case EXT4_IOC_SWAP_BOOT:
1723 	{
1724 		int err;
1725 		if (!(filp->f_mode & FMODE_WRITE))
1726 			return -EBADF;
1727 		err = mnt_want_write_file(filp);
1728 		if (err)
1729 			return err;
1730 		err = swap_inode_boot_loader(sb, idmap, inode);
1731 		mnt_drop_write_file(filp);
1732 		return err;
1733 	}
1734 
1735 	case EXT4_IOC_RESIZE_FS: {
1736 		ext4_fsblk_t n_blocks_count;
1737 		int err = 0, err2 = 0;
1738 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
1739 
1740 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1741 				   sizeof(__u64))) {
1742 			return -EFAULT;
1743 		}
1744 
1745 		err = ext4_resize_begin(sb);
1746 		if (err)
1747 			return err;
1748 
1749 		err = mnt_want_write_file(filp);
1750 		if (err)
1751 			goto resizefs_out;
1752 
1753 		err = ext4_resize_fs(sb, n_blocks_count);
1754 		if (EXT4_SB(sb)->s_journal) {
1755 			ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE, NULL);
1756 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1757 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1758 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1759 		}
1760 		if (err == 0)
1761 			err = err2;
1762 		mnt_drop_write_file(filp);
1763 		if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1764 		    ext4_has_group_desc_csum(sb) &&
1765 		    test_opt(sb, INIT_INODE_TABLE))
1766 			err = ext4_register_li_request(sb, o_group);
1767 
1768 resizefs_out:
1769 		err2 = ext4_resize_end(sb, true);
1770 		if (err == 0)
1771 			err = err2;
1772 		return err;
1773 	}
1774 
1775 	case FITRIM:
1776 	{
1777 		struct fstrim_range range;
1778 		int ret = 0;
1779 
1780 		if (!capable(CAP_SYS_ADMIN))
1781 			return -EPERM;
1782 
1783 		if (!bdev_max_discard_sectors(sb->s_bdev))
1784 			return -EOPNOTSUPP;
1785 
1786 		/*
1787 		 * We haven't replayed the journal, so we cannot use our
1788 		 * block-bitmap-guided storage zapping commands.
1789 		 */
1790 		if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1791 			return -EROFS;
1792 
1793 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1794 		    sizeof(range)))
1795 			return -EFAULT;
1796 
1797 		ret = ext4_trim_fs(sb, &range);
1798 		if (ret < 0)
1799 			return ret;
1800 
1801 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
1802 		    sizeof(range)))
1803 			return -EFAULT;
1804 
1805 		return 0;
1806 	}
1807 	case EXT4_IOC_PRECACHE_EXTENTS:
1808 	{
1809 		int ret;
1810 
1811 		inode_lock_shared(inode);
1812 		ret = ext4_ext_precache(inode);
1813 		inode_unlock_shared(inode);
1814 		return ret;
1815 	}
1816 	case FS_IOC_SET_ENCRYPTION_POLICY:
1817 		if (!ext4_has_feature_encrypt(sb))
1818 			return -EOPNOTSUPP;
1819 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1820 
1821 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1822 		return ext4_ioctl_get_encryption_pwsalt(filp, (void __user *)arg);
1823 
1824 	case FS_IOC_GET_ENCRYPTION_POLICY:
1825 		if (!ext4_has_feature_encrypt(sb))
1826 			return -EOPNOTSUPP;
1827 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1828 
1829 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1830 		if (!ext4_has_feature_encrypt(sb))
1831 			return -EOPNOTSUPP;
1832 		return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
1833 
1834 	case FS_IOC_ADD_ENCRYPTION_KEY:
1835 		if (!ext4_has_feature_encrypt(sb))
1836 			return -EOPNOTSUPP;
1837 		return fscrypt_ioctl_add_key(filp, (void __user *)arg);
1838 
1839 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1840 		if (!ext4_has_feature_encrypt(sb))
1841 			return -EOPNOTSUPP;
1842 		return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
1843 
1844 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1845 		if (!ext4_has_feature_encrypt(sb))
1846 			return -EOPNOTSUPP;
1847 		return fscrypt_ioctl_remove_key_all_users(filp,
1848 							  (void __user *)arg);
1849 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1850 		if (!ext4_has_feature_encrypt(sb))
1851 			return -EOPNOTSUPP;
1852 		return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
1853 
1854 	case FS_IOC_GET_ENCRYPTION_NONCE:
1855 		if (!ext4_has_feature_encrypt(sb))
1856 			return -EOPNOTSUPP;
1857 		return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
1858 
1859 	case EXT4_IOC_CLEAR_ES_CACHE:
1860 	{
1861 		if (!inode_owner_or_capable(idmap, inode))
1862 			return -EACCES;
1863 		ext4_clear_inode_es(inode);
1864 		return 0;
1865 	}
1866 
1867 	case EXT4_IOC_GETSTATE:
1868 	{
1869 		__u32	state = 0;
1870 
1871 		if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED))
1872 			state |= EXT4_STATE_FLAG_EXT_PRECACHED;
1873 		if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
1874 			state |= EXT4_STATE_FLAG_NEW;
1875 		if (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
1876 			state |= EXT4_STATE_FLAG_NEWENTRY;
1877 		if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE))
1878 			state |= EXT4_STATE_FLAG_DA_ALLOC_CLOSE;
1879 
1880 		return put_user(state, (__u32 __user *) arg);
1881 	}
1882 
1883 	case EXT4_IOC_GET_ES_CACHE:
1884 		return ext4_ioctl_get_es_cache(filp, arg);
1885 
1886 	case EXT4_IOC_SHUTDOWN:
1887 		return ext4_ioctl_shutdown(sb, arg);
1888 
1889 	case FS_IOC_ENABLE_VERITY:
1890 		if (!ext4_has_feature_verity(sb))
1891 			return -EOPNOTSUPP;
1892 		return fsverity_ioctl_enable(filp, (const void __user *)arg);
1893 
1894 	case FS_IOC_MEASURE_VERITY:
1895 		if (!ext4_has_feature_verity(sb))
1896 			return -EOPNOTSUPP;
1897 		return fsverity_ioctl_measure(filp, (void __user *)arg);
1898 
1899 	case FS_IOC_READ_VERITY_METADATA:
1900 		if (!ext4_has_feature_verity(sb))
1901 			return -EOPNOTSUPP;
1902 		return fsverity_ioctl_read_metadata(filp,
1903 						    (const void __user *)arg);
1904 
1905 	case EXT4_IOC_CHECKPOINT:
1906 		return ext4_ioctl_checkpoint(filp, arg);
1907 
1908 	case FS_IOC_GETFSLABEL:
1909 		return ext4_ioctl_getlabel(EXT4_SB(sb), (void __user *)arg);
1910 
1911 	case FS_IOC_SETFSLABEL:
1912 		return ext4_ioctl_setlabel(filp,
1913 					   (const void __user *)arg);
1914 
1915 	case EXT4_IOC_GETFSUUID:
1916 		return ext4_ioctl_getuuid(EXT4_SB(sb), (void __user *)arg);
1917 	case EXT4_IOC_SETFSUUID:
1918 		return ext4_ioctl_setuuid(filp, (const void __user *)arg);
1919 	case EXT4_IOC_GET_TUNE_SB_PARAM:
1920 		return ext4_ioctl_get_tune_sb(EXT4_SB(sb),
1921 					      (void __user *)arg);
1922 	case EXT4_IOC_SET_TUNE_SB_PARAM:
1923 		return ext4_ioctl_set_tune_sb(filp, (void __user *)arg);
1924 	default:
1925 		return -ENOTTY;
1926 	}
1927 }
1928 
1929 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1930 {
1931 	return __ext4_ioctl(filp, cmd, arg);
1932 }
1933 
1934 #ifdef CONFIG_COMPAT
1935 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1936 {
1937 	/* These are just misnamed, they actually get/put from/to user an int */
1938 	switch (cmd) {
1939 	case EXT4_IOC32_GETVERSION:
1940 		cmd = EXT4_IOC_GETVERSION;
1941 		break;
1942 	case EXT4_IOC32_SETVERSION:
1943 		cmd = EXT4_IOC_SETVERSION;
1944 		break;
1945 	case EXT4_IOC32_GROUP_EXTEND:
1946 		cmd = EXT4_IOC_GROUP_EXTEND;
1947 		break;
1948 	case EXT4_IOC32_GETVERSION_OLD:
1949 		cmd = EXT4_IOC_GETVERSION_OLD;
1950 		break;
1951 	case EXT4_IOC32_SETVERSION_OLD:
1952 		cmd = EXT4_IOC_SETVERSION_OLD;
1953 		break;
1954 	case EXT4_IOC32_GETRSVSZ:
1955 		cmd = EXT4_IOC_GETRSVSZ;
1956 		break;
1957 	case EXT4_IOC32_SETRSVSZ:
1958 		cmd = EXT4_IOC_SETRSVSZ;
1959 		break;
1960 	case EXT4_IOC32_GROUP_ADD: {
1961 		struct compat_ext4_new_group_input __user *uinput;
1962 		struct ext4_new_group_data input;
1963 		int err;
1964 
1965 		uinput = compat_ptr(arg);
1966 		err = get_user(input.group, &uinput->group);
1967 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1968 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1969 		err |= get_user(input.inode_table, &uinput->inode_table);
1970 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1971 		err |= get_user(input.reserved_blocks,
1972 				&uinput->reserved_blocks);
1973 		if (err)
1974 			return -EFAULT;
1975 		return ext4_ioctl_group_add(file, &input);
1976 	}
1977 	case EXT4_IOC_MOVE_EXT:
1978 	case EXT4_IOC_RESIZE_FS:
1979 	case FITRIM:
1980 	case EXT4_IOC_PRECACHE_EXTENTS:
1981 	case FS_IOC_SET_ENCRYPTION_POLICY:
1982 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1983 	case FS_IOC_GET_ENCRYPTION_POLICY:
1984 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1985 	case FS_IOC_ADD_ENCRYPTION_KEY:
1986 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1987 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1988 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1989 	case FS_IOC_GET_ENCRYPTION_NONCE:
1990 	case EXT4_IOC_SHUTDOWN:
1991 	case FS_IOC_GETFSMAP:
1992 	case FS_IOC_ENABLE_VERITY:
1993 	case FS_IOC_MEASURE_VERITY:
1994 	case FS_IOC_READ_VERITY_METADATA:
1995 	case EXT4_IOC_CLEAR_ES_CACHE:
1996 	case EXT4_IOC_GETSTATE:
1997 	case EXT4_IOC_GET_ES_CACHE:
1998 	case EXT4_IOC_CHECKPOINT:
1999 	case FS_IOC_GETFSLABEL:
2000 	case FS_IOC_SETFSLABEL:
2001 	case EXT4_IOC_GETFSUUID:
2002 	case EXT4_IOC_SETFSUUID:
2003 		break;
2004 	default:
2005 		return -ENOIOCTLCMD;
2006 	}
2007 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2008 }
2009 #endif
2010 
2011 static void set_overhead(struct ext4_sb_info *sbi,
2012 			 struct ext4_super_block *es, const void *arg)
2013 {
2014 	es->s_overhead_clusters = cpu_to_le32(*((unsigned long *) arg));
2015 }
2016 
2017 int ext4_update_overhead(struct super_block *sb, bool force)
2018 {
2019 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2020 
2021 	if (ext4_emergency_state(sb) || sb_rdonly(sb))
2022 		return 0;
2023 	if (!force &&
2024 	    (sbi->s_overhead == 0 ||
2025 	     sbi->s_overhead == le32_to_cpu(sbi->s_es->s_overhead_clusters)))
2026 		return 0;
2027 	return ext4_update_superblocks_fn(sb, set_overhead, &sbi->s_overhead);
2028 }
2029