xref: /linux/fs/ext4/ioctl.c (revision f4738f56d1dc62aaba69b33702a5ab098f1b8c63)
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 
30 typedef void ext4_update_sb_callback(struct ext4_super_block *es,
31 				       const void *arg);
32 
33 /*
34  * Superblock modification callback function for changing file system
35  * label
36  */
37 static void ext4_sb_setlabel(struct ext4_super_block *es, const void *arg)
38 {
39 	/* Sanity check, this should never happen */
40 	BUILD_BUG_ON(sizeof(es->s_volume_name) < EXT4_LABEL_MAX);
41 
42 	memcpy(es->s_volume_name, (char *)arg, EXT4_LABEL_MAX);
43 }
44 
45 /*
46  * Superblock modification callback function for changing file system
47  * UUID.
48  */
49 static void ext4_sb_setuuid(struct ext4_super_block *es, const void *arg)
50 {
51 	memcpy(es->s_uuid, (__u8 *)arg, UUID_SIZE);
52 }
53 
54 static
55 int ext4_update_primary_sb(struct super_block *sb, handle_t *handle,
56 			   ext4_update_sb_callback func,
57 			   const void *arg)
58 {
59 	int err = 0;
60 	struct ext4_sb_info *sbi = EXT4_SB(sb);
61 	struct buffer_head *bh = sbi->s_sbh;
62 	struct ext4_super_block *es = sbi->s_es;
63 
64 	trace_ext4_update_sb(sb, bh->b_blocknr, 1);
65 
66 	BUFFER_TRACE(bh, "get_write_access");
67 	err = ext4_journal_get_write_access(handle, sb,
68 					    bh,
69 					    EXT4_JTR_NONE);
70 	if (err)
71 		goto out_err;
72 
73 	lock_buffer(bh);
74 	func(es, arg);
75 	ext4_superblock_csum_set(sb);
76 	unlock_buffer(bh);
77 
78 	if (buffer_write_io_error(bh) || !buffer_uptodate(bh)) {
79 		ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
80 			 "superblock detected");
81 		clear_buffer_write_io_error(bh);
82 		set_buffer_uptodate(bh);
83 	}
84 
85 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
86 	if (err)
87 		goto out_err;
88 	err = sync_dirty_buffer(bh);
89 out_err:
90 	ext4_std_error(sb, err);
91 	return err;
92 }
93 
94 /*
95  * Update one backup superblock in the group 'grp' using the callback
96  * function 'func' and argument 'arg'. If the handle is NULL the
97  * modification is not journalled.
98  *
99  * Returns: 0 when no modification was done (no superblock in the group)
100  *	    1 when the modification was successful
101  *	   <0 on error
102  */
103 static int ext4_update_backup_sb(struct super_block *sb,
104 				 handle_t *handle, ext4_group_t grp,
105 				 ext4_update_sb_callback func, const void *arg)
106 {
107 	int err = 0;
108 	ext4_fsblk_t sb_block;
109 	struct buffer_head *bh;
110 	unsigned long offset = 0;
111 	struct ext4_super_block *es;
112 
113 	if (!ext4_bg_has_super(sb, grp))
114 		return 0;
115 
116 	/*
117 	 * For the group 0 there is always 1k padding, so we have
118 	 * either adjust offset, or sb_block depending on blocksize
119 	 */
120 	if (grp == 0) {
121 		sb_block = 1 * EXT4_MIN_BLOCK_SIZE;
122 		offset = do_div(sb_block, sb->s_blocksize);
123 	} else {
124 		sb_block = ext4_group_first_block_no(sb, grp);
125 		offset = 0;
126 	}
127 
128 	trace_ext4_update_sb(sb, sb_block, handle ? 1 : 0);
129 
130 	bh = ext4_sb_bread(sb, sb_block, 0);
131 	if (IS_ERR(bh))
132 		return PTR_ERR(bh);
133 
134 	if (handle) {
135 		BUFFER_TRACE(bh, "get_write_access");
136 		err = ext4_journal_get_write_access(handle, sb,
137 						    bh,
138 						    EXT4_JTR_NONE);
139 		if (err)
140 			goto out_bh;
141 	}
142 
143 	es = (struct ext4_super_block *) (bh->b_data + offset);
144 	lock_buffer(bh);
145 	if (ext4_has_metadata_csum(sb) &&
146 	    es->s_checksum != ext4_superblock_csum(sb, es)) {
147 		ext4_msg(sb, KERN_ERR, "Invalid checksum for backup "
148 		"superblock %llu", sb_block);
149 		unlock_buffer(bh);
150 		goto out_bh;
151 	}
152 	func(es, arg);
153 	if (ext4_has_metadata_csum(sb))
154 		es->s_checksum = ext4_superblock_csum(sb, es);
155 	set_buffer_uptodate(bh);
156 	unlock_buffer(bh);
157 
158 	if (handle) {
159 		err = ext4_handle_dirty_metadata(handle, NULL, bh);
160 		if (err)
161 			goto out_bh;
162 	} else {
163 		BUFFER_TRACE(bh, "marking dirty");
164 		mark_buffer_dirty(bh);
165 	}
166 	err = sync_dirty_buffer(bh);
167 
168 out_bh:
169 	brelse(bh);
170 	ext4_std_error(sb, err);
171 	return (err) ? err : 1;
172 }
173 
174 /*
175  * Update primary and backup superblocks using the provided function
176  * func and argument arg.
177  *
178  * Only the primary superblock and at most two backup superblock
179  * modifications are journalled; the rest is modified without journal.
180  * This is safe because e2fsck will re-write them if there is a problem,
181  * and we're very unlikely to ever need more than two backups.
182  */
183 static
184 int ext4_update_superblocks_fn(struct super_block *sb,
185 			       ext4_update_sb_callback func,
186 			       const void *arg)
187 {
188 	handle_t *handle;
189 	ext4_group_t ngroups;
190 	unsigned int three = 1;
191 	unsigned int five = 5;
192 	unsigned int seven = 7;
193 	int err = 0, ret, i;
194 	ext4_group_t grp, primary_grp;
195 	struct ext4_sb_info *sbi = EXT4_SB(sb);
196 
197 	/*
198 	 * We can't update superblocks while the online resize is running
199 	 */
200 	if (test_and_set_bit_lock(EXT4_FLAGS_RESIZING,
201 				  &sbi->s_ext4_flags)) {
202 		ext4_msg(sb, KERN_ERR, "Can't modify superblock while"
203 			 "performing online resize");
204 		return -EBUSY;
205 	}
206 
207 	/*
208 	 * We're only going to update primary superblock and two
209 	 * backup superblocks in this transaction.
210 	 */
211 	handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 3);
212 	if (IS_ERR(handle)) {
213 		err = PTR_ERR(handle);
214 		goto out;
215 	}
216 
217 	/* Update primary superblock */
218 	err = ext4_update_primary_sb(sb, handle, func, arg);
219 	if (err) {
220 		ext4_msg(sb, KERN_ERR, "Failed to update primary "
221 			 "superblock");
222 		goto out_journal;
223 	}
224 
225 	primary_grp = ext4_get_group_number(sb, sbi->s_sbh->b_blocknr);
226 	ngroups = ext4_get_groups_count(sb);
227 
228 	/*
229 	 * Update backup superblocks. We have to start from group 0
230 	 * because it might not be where the primary superblock is
231 	 * if the fs is mounted with -o sb=<backup_sb_block>
232 	 */
233 	i = 0;
234 	grp = 0;
235 	while (grp < ngroups) {
236 		/* Skip primary superblock */
237 		if (grp == primary_grp)
238 			goto next_grp;
239 
240 		ret = ext4_update_backup_sb(sb, handle, grp, func, arg);
241 		if (ret < 0) {
242 			/* Ignore bad checksum; try to update next sb */
243 			if (ret == -EFSBADCRC)
244 				goto next_grp;
245 			err = ret;
246 			goto out_journal;
247 		}
248 
249 		i += ret;
250 		if (handle && i > 1) {
251 			/*
252 			 * We're only journalling primary superblock and
253 			 * two backup superblocks; the rest is not
254 			 * journalled.
255 			 */
256 			err = ext4_journal_stop(handle);
257 			if (err)
258 				goto out;
259 			handle = NULL;
260 		}
261 next_grp:
262 		grp = ext4_list_backups(sb, &three, &five, &seven);
263 	}
264 
265 out_journal:
266 	if (handle) {
267 		ret = ext4_journal_stop(handle);
268 		if (ret && !err)
269 			err = ret;
270 	}
271 out:
272 	clear_bit_unlock(EXT4_FLAGS_RESIZING, &sbi->s_ext4_flags);
273 	smp_mb__after_atomic();
274 	return err ? err : 0;
275 }
276 
277 /*
278  * Swap memory between @a and @b for @len bytes.
279  *
280  * @a:          pointer to first memory area
281  * @b:          pointer to second memory area
282  * @len:        number of bytes to swap
283  *
284  */
285 static void memswap(void *a, void *b, size_t len)
286 {
287 	unsigned char *ap, *bp;
288 
289 	ap = (unsigned char *)a;
290 	bp = (unsigned char *)b;
291 	while (len-- > 0) {
292 		swap(*ap, *bp);
293 		ap++;
294 		bp++;
295 	}
296 }
297 
298 /*
299  * Swap i_data and associated attributes between @inode1 and @inode2.
300  * This function is used for the primary swap between inode1 and inode2
301  * and also to revert this primary swap in case of errors.
302  *
303  * Therefore you have to make sure, that calling this method twice
304  * will revert all changes.
305  *
306  * @inode1:     pointer to first inode
307  * @inode2:     pointer to second inode
308  */
309 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
310 {
311 	loff_t isize;
312 	struct ext4_inode_info *ei1;
313 	struct ext4_inode_info *ei2;
314 	unsigned long tmp;
315 
316 	ei1 = EXT4_I(inode1);
317 	ei2 = EXT4_I(inode2);
318 
319 	swap(inode1->i_version, inode2->i_version);
320 	swap(inode1->i_atime, inode2->i_atime);
321 	swap(inode1->i_mtime, inode2->i_mtime);
322 
323 	memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
324 	tmp = ei1->i_flags & EXT4_FL_SHOULD_SWAP;
325 	ei1->i_flags = (ei2->i_flags & EXT4_FL_SHOULD_SWAP) |
326 		(ei1->i_flags & ~EXT4_FL_SHOULD_SWAP);
327 	ei2->i_flags = tmp | (ei2->i_flags & ~EXT4_FL_SHOULD_SWAP);
328 	swap(ei1->i_disksize, ei2->i_disksize);
329 	ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
330 	ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
331 
332 	isize = i_size_read(inode1);
333 	i_size_write(inode1, i_size_read(inode2));
334 	i_size_write(inode2, isize);
335 }
336 
337 void ext4_reset_inode_seed(struct inode *inode)
338 {
339 	struct ext4_inode_info *ei = EXT4_I(inode);
340 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
341 	__le32 inum = cpu_to_le32(inode->i_ino);
342 	__le32 gen = cpu_to_le32(inode->i_generation);
343 	__u32 csum;
344 
345 	if (!ext4_has_metadata_csum(inode->i_sb))
346 		return;
347 
348 	csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum, sizeof(inum));
349 	ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen, sizeof(gen));
350 }
351 
352 /*
353  * Swap the information from the given @inode and the inode
354  * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
355  * important fields of the inodes.
356  *
357  * @sb:         the super block of the filesystem
358  * @idmap:	idmap of the mount the inode was found from
359  * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
360  *
361  */
362 static long swap_inode_boot_loader(struct super_block *sb,
363 				struct mnt_idmap *idmap,
364 				struct inode *inode)
365 {
366 	handle_t *handle;
367 	int err;
368 	struct inode *inode_bl;
369 	struct ext4_inode_info *ei_bl;
370 	qsize_t size, size_bl, diff;
371 	blkcnt_t blocks;
372 	unsigned short bytes;
373 
374 	inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO,
375 			EXT4_IGET_SPECIAL | EXT4_IGET_BAD);
376 	if (IS_ERR(inode_bl))
377 		return PTR_ERR(inode_bl);
378 	ei_bl = EXT4_I(inode_bl);
379 
380 	/* Protect orig inodes against a truncate and make sure,
381 	 * that only 1 swap_inode_boot_loader is running. */
382 	lock_two_nondirectories(inode, inode_bl);
383 
384 	if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode) ||
385 	    IS_SWAPFILE(inode) || IS_ENCRYPTED(inode) ||
386 	    (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL) ||
387 	    ext4_has_inline_data(inode)) {
388 		err = -EINVAL;
389 		goto journal_err_out;
390 	}
391 
392 	if (IS_RDONLY(inode) || IS_APPEND(inode) || IS_IMMUTABLE(inode) ||
393 	    !inode_owner_or_capable(idmap, inode) ||
394 	    !capable(CAP_SYS_ADMIN)) {
395 		err = -EPERM;
396 		goto journal_err_out;
397 	}
398 
399 	filemap_invalidate_lock(inode->i_mapping);
400 	err = filemap_write_and_wait(inode->i_mapping);
401 	if (err)
402 		goto err_out;
403 
404 	err = filemap_write_and_wait(inode_bl->i_mapping);
405 	if (err)
406 		goto err_out;
407 
408 	/* Wait for all existing dio workers */
409 	inode_dio_wait(inode);
410 	inode_dio_wait(inode_bl);
411 
412 	truncate_inode_pages(&inode->i_data, 0);
413 	truncate_inode_pages(&inode_bl->i_data, 0);
414 
415 	handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
416 	if (IS_ERR(handle)) {
417 		err = -EINVAL;
418 		goto err_out;
419 	}
420 	ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_SWAP_BOOT, handle);
421 
422 	/* Protect extent tree against block allocations via delalloc */
423 	ext4_double_down_write_data_sem(inode, inode_bl);
424 
425 	if (is_bad_inode(inode_bl) || !S_ISREG(inode_bl->i_mode)) {
426 		/* this inode has never been used as a BOOT_LOADER */
427 		set_nlink(inode_bl, 1);
428 		i_uid_write(inode_bl, 0);
429 		i_gid_write(inode_bl, 0);
430 		inode_bl->i_flags = 0;
431 		ei_bl->i_flags = 0;
432 		inode_set_iversion(inode_bl, 1);
433 		i_size_write(inode_bl, 0);
434 		EXT4_I(inode_bl)->i_disksize = inode_bl->i_size;
435 		inode_bl->i_mode = S_IFREG;
436 		if (ext4_has_feature_extents(sb)) {
437 			ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
438 			ext4_ext_tree_init(handle, inode_bl);
439 		} else
440 			memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
441 	}
442 
443 	err = dquot_initialize(inode);
444 	if (err)
445 		goto err_out1;
446 
447 	size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
448 	size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
449 	diff = size - size_bl;
450 	swap_inode_data(inode, inode_bl);
451 
452 	inode_set_ctime_current(inode);
453 	inode_set_ctime_current(inode_bl);
454 	inode_inc_iversion(inode);
455 
456 	inode->i_generation = get_random_u32();
457 	inode_bl->i_generation = get_random_u32();
458 	ext4_reset_inode_seed(inode);
459 	ext4_reset_inode_seed(inode_bl);
460 
461 	ext4_discard_preallocations(inode, 0);
462 
463 	err = ext4_mark_inode_dirty(handle, inode);
464 	if (err < 0) {
465 		/* No need to update quota information. */
466 		ext4_warning(inode->i_sb,
467 			"couldn't mark inode #%lu dirty (err %d)",
468 			inode->i_ino, err);
469 		/* Revert all changes: */
470 		swap_inode_data(inode, inode_bl);
471 		ext4_mark_inode_dirty(handle, inode);
472 		goto err_out1;
473 	}
474 
475 	blocks = inode_bl->i_blocks;
476 	bytes = inode_bl->i_bytes;
477 	inode_bl->i_blocks = inode->i_blocks;
478 	inode_bl->i_bytes = inode->i_bytes;
479 	err = ext4_mark_inode_dirty(handle, inode_bl);
480 	if (err < 0) {
481 		/* No need to update quota information. */
482 		ext4_warning(inode_bl->i_sb,
483 			"couldn't mark inode #%lu dirty (err %d)",
484 			inode_bl->i_ino, err);
485 		goto revert;
486 	}
487 
488 	/* Bootloader inode should not be counted into quota information. */
489 	if (diff > 0)
490 		dquot_free_space(inode, diff);
491 	else
492 		err = dquot_alloc_space(inode, -1 * diff);
493 
494 	if (err < 0) {
495 revert:
496 		/* Revert all changes: */
497 		inode_bl->i_blocks = blocks;
498 		inode_bl->i_bytes = bytes;
499 		swap_inode_data(inode, inode_bl);
500 		ext4_mark_inode_dirty(handle, inode);
501 		ext4_mark_inode_dirty(handle, inode_bl);
502 	}
503 
504 err_out1:
505 	ext4_journal_stop(handle);
506 	ext4_double_up_write_data_sem(inode, inode_bl);
507 
508 err_out:
509 	filemap_invalidate_unlock(inode->i_mapping);
510 journal_err_out:
511 	unlock_two_nondirectories(inode, inode_bl);
512 	iput(inode_bl);
513 	return err;
514 }
515 
516 /*
517  * If immutable is set and we are not clearing it, we're not allowed to change
518  * anything else in the inode.  Don't error out if we're only trying to set
519  * immutable on an immutable file.
520  */
521 static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
522 				      unsigned int flags)
523 {
524 	struct ext4_inode_info *ei = EXT4_I(inode);
525 	unsigned int oldflags = ei->i_flags;
526 
527 	if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
528 		return 0;
529 
530 	if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
531 		return -EPERM;
532 	if (ext4_has_feature_project(inode->i_sb) &&
533 	    __kprojid_val(ei->i_projid) != new_projid)
534 		return -EPERM;
535 
536 	return 0;
537 }
538 
539 static void ext4_dax_dontcache(struct inode *inode, unsigned int flags)
540 {
541 	struct ext4_inode_info *ei = EXT4_I(inode);
542 
543 	if (S_ISDIR(inode->i_mode))
544 		return;
545 
546 	if (test_opt2(inode->i_sb, DAX_NEVER) ||
547 	    test_opt(inode->i_sb, DAX_ALWAYS))
548 		return;
549 
550 	if ((ei->i_flags ^ flags) & EXT4_DAX_FL)
551 		d_mark_dontcache(inode);
552 }
553 
554 static bool dax_compatible(struct inode *inode, unsigned int oldflags,
555 			   unsigned int flags)
556 {
557 	/* Allow the DAX flag to be changed on inline directories */
558 	if (S_ISDIR(inode->i_mode)) {
559 		flags &= ~EXT4_INLINE_DATA_FL;
560 		oldflags &= ~EXT4_INLINE_DATA_FL;
561 	}
562 
563 	if (flags & EXT4_DAX_FL) {
564 		if ((oldflags & EXT4_DAX_MUT_EXCL) ||
565 		     ext4_test_inode_state(inode,
566 					  EXT4_STATE_VERITY_IN_PROGRESS)) {
567 			return false;
568 		}
569 	}
570 
571 	if ((flags & EXT4_DAX_MUT_EXCL) && (oldflags & EXT4_DAX_FL))
572 			return false;
573 
574 	return true;
575 }
576 
577 static int ext4_ioctl_setflags(struct inode *inode,
578 			       unsigned int flags)
579 {
580 	struct ext4_inode_info *ei = EXT4_I(inode);
581 	handle_t *handle = NULL;
582 	int err = -EPERM, migrate = 0;
583 	struct ext4_iloc iloc;
584 	unsigned int oldflags, mask, i;
585 	struct super_block *sb = inode->i_sb;
586 
587 	/* Is it quota file? Do not allow user to mess with it */
588 	if (ext4_is_quota_file(inode))
589 		goto flags_out;
590 
591 	oldflags = ei->i_flags;
592 	/*
593 	 * The JOURNAL_DATA flag can only be changed by
594 	 * the relevant capability.
595 	 */
596 	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
597 		if (!capable(CAP_SYS_RESOURCE))
598 			goto flags_out;
599 	}
600 
601 	if (!dax_compatible(inode, oldflags, flags)) {
602 		err = -EOPNOTSUPP;
603 		goto flags_out;
604 	}
605 
606 	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
607 		migrate = 1;
608 
609 	if ((flags ^ oldflags) & EXT4_CASEFOLD_FL) {
610 		if (!ext4_has_feature_casefold(sb)) {
611 			err = -EOPNOTSUPP;
612 			goto flags_out;
613 		}
614 
615 		if (!S_ISDIR(inode->i_mode)) {
616 			err = -ENOTDIR;
617 			goto flags_out;
618 		}
619 
620 		if (!ext4_empty_dir(inode)) {
621 			err = -ENOTEMPTY;
622 			goto flags_out;
623 		}
624 	}
625 
626 	/*
627 	 * Wait for all pending directio and then flush all the dirty pages
628 	 * for this file.  The flush marks all the pages readonly, so any
629 	 * subsequent attempt to write to the file (particularly mmap pages)
630 	 * will come through the filesystem and fail.
631 	 */
632 	if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
633 	    (flags & EXT4_IMMUTABLE_FL)) {
634 		inode_dio_wait(inode);
635 		err = filemap_write_and_wait(inode->i_mapping);
636 		if (err)
637 			goto flags_out;
638 	}
639 
640 	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
641 	if (IS_ERR(handle)) {
642 		err = PTR_ERR(handle);
643 		goto flags_out;
644 	}
645 	if (IS_SYNC(inode))
646 		ext4_handle_sync(handle);
647 	err = ext4_reserve_inode_write(handle, inode, &iloc);
648 	if (err)
649 		goto flags_err;
650 
651 	ext4_dax_dontcache(inode, flags);
652 
653 	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
654 		if (!(mask & EXT4_FL_USER_MODIFIABLE))
655 			continue;
656 		/* These flags get special treatment later */
657 		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
658 			continue;
659 		if (mask & flags)
660 			ext4_set_inode_flag(inode, i);
661 		else
662 			ext4_clear_inode_flag(inode, i);
663 	}
664 
665 	ext4_set_inode_flags(inode, false);
666 
667 	inode_set_ctime_current(inode);
668 	inode_inc_iversion(inode);
669 
670 	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
671 flags_err:
672 	ext4_journal_stop(handle);
673 	if (err)
674 		goto flags_out;
675 
676 	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
677 		/*
678 		 * Changes to the journaling mode can cause unsafe changes to
679 		 * S_DAX if the inode is DAX
680 		 */
681 		if (IS_DAX(inode)) {
682 			err = -EBUSY;
683 			goto flags_out;
684 		}
685 
686 		err = ext4_change_inode_journal_flag(inode,
687 						     flags & EXT4_JOURNAL_DATA_FL);
688 		if (err)
689 			goto flags_out;
690 	}
691 	if (migrate) {
692 		if (flags & EXT4_EXTENTS_FL)
693 			err = ext4_ext_migrate(inode);
694 		else
695 			err = ext4_ind_migrate(inode);
696 	}
697 
698 flags_out:
699 	return err;
700 }
701 
702 #ifdef CONFIG_QUOTA
703 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
704 {
705 	struct super_block *sb = inode->i_sb;
706 	struct ext4_inode_info *ei = EXT4_I(inode);
707 	int err, rc;
708 	handle_t *handle;
709 	kprojid_t kprojid;
710 	struct ext4_iloc iloc;
711 	struct ext4_inode *raw_inode;
712 	struct dquot *transfer_to[MAXQUOTAS] = { };
713 
714 	if (!ext4_has_feature_project(sb)) {
715 		if (projid != EXT4_DEF_PROJID)
716 			return -EOPNOTSUPP;
717 		else
718 			return 0;
719 	}
720 
721 	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
722 		return -EOPNOTSUPP;
723 
724 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
725 
726 	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
727 		return 0;
728 
729 	err = -EPERM;
730 	/* Is it quota file? Do not allow user to mess with it */
731 	if (ext4_is_quota_file(inode))
732 		return err;
733 
734 	err = dquot_initialize(inode);
735 	if (err)
736 		return err;
737 
738 	err = ext4_get_inode_loc(inode, &iloc);
739 	if (err)
740 		return err;
741 
742 	raw_inode = ext4_raw_inode(&iloc);
743 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
744 		err = ext4_expand_extra_isize(inode,
745 					      EXT4_SB(sb)->s_want_extra_isize,
746 					      &iloc);
747 		if (err)
748 			return err;
749 	} else {
750 		brelse(iloc.bh);
751 	}
752 
753 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
754 		EXT4_QUOTA_INIT_BLOCKS(sb) +
755 		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
756 	if (IS_ERR(handle))
757 		return PTR_ERR(handle);
758 
759 	err = ext4_reserve_inode_write(handle, inode, &iloc);
760 	if (err)
761 		goto out_stop;
762 
763 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
764 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
765 
766 		/* __dquot_transfer() calls back ext4_get_inode_usage() which
767 		 * counts xattr inode references.
768 		 */
769 		down_read(&EXT4_I(inode)->xattr_sem);
770 		err = __dquot_transfer(inode, transfer_to);
771 		up_read(&EXT4_I(inode)->xattr_sem);
772 		dqput(transfer_to[PRJQUOTA]);
773 		if (err)
774 			goto out_dirty;
775 	}
776 
777 	EXT4_I(inode)->i_projid = kprojid;
778 	inode_set_ctime_current(inode);
779 	inode_inc_iversion(inode);
780 out_dirty:
781 	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
782 	if (!err)
783 		err = rc;
784 out_stop:
785 	ext4_journal_stop(handle);
786 	return err;
787 }
788 #else
789 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
790 {
791 	if (projid != EXT4_DEF_PROJID)
792 		return -EOPNOTSUPP;
793 	return 0;
794 }
795 #endif
796 
797 int ext4_force_shutdown(struct super_block *sb, u32 flags)
798 {
799 	struct ext4_sb_info *sbi = EXT4_SB(sb);
800 	int ret;
801 
802 	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
803 		return -EINVAL;
804 
805 	if (ext4_forced_shutdown(sb))
806 		return 0;
807 
808 	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
809 	trace_ext4_shutdown(sb, flags);
810 
811 	switch (flags) {
812 	case EXT4_GOING_FLAGS_DEFAULT:
813 		ret = freeze_bdev(sb->s_bdev);
814 		if (ret)
815 			return ret;
816 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
817 		thaw_bdev(sb->s_bdev);
818 		break;
819 	case EXT4_GOING_FLAGS_LOGFLUSH:
820 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
821 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
822 			(void) ext4_force_commit(sb);
823 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
824 		}
825 		break;
826 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
827 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
828 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
829 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
830 		break;
831 	default:
832 		return -EINVAL;
833 	}
834 	clear_opt(sb, DISCARD);
835 	return 0;
836 }
837 
838 static int ext4_ioctl_shutdown(struct super_block *sb, unsigned long arg)
839 {
840 	u32 flags;
841 
842 	if (!capable(CAP_SYS_ADMIN))
843 		return -EPERM;
844 
845 	if (get_user(flags, (__u32 __user *)arg))
846 		return -EFAULT;
847 
848 	return ext4_force_shutdown(sb, flags);
849 }
850 
851 struct getfsmap_info {
852 	struct super_block	*gi_sb;
853 	struct fsmap_head __user *gi_data;
854 	unsigned int		gi_idx;
855 	__u32			gi_last_flags;
856 };
857 
858 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
859 {
860 	struct getfsmap_info *info = priv;
861 	struct fsmap fm;
862 
863 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
864 
865 	info->gi_last_flags = xfm->fmr_flags;
866 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
867 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
868 			sizeof(struct fsmap)))
869 		return -EFAULT;
870 
871 	return 0;
872 }
873 
874 static int ext4_ioc_getfsmap(struct super_block *sb,
875 			     struct fsmap_head __user *arg)
876 {
877 	struct getfsmap_info info = { NULL };
878 	struct ext4_fsmap_head xhead = {0};
879 	struct fsmap_head head;
880 	bool aborted = false;
881 	int error;
882 
883 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
884 		return -EFAULT;
885 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
886 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
887 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
888 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
889 		       sizeof(head.fmh_keys[1].fmr_reserved)))
890 		return -EINVAL;
891 	/*
892 	 * ext4 doesn't report file extents at all, so the only valid
893 	 * file offsets are the magic ones (all zeroes or all ones).
894 	 */
895 	if (head.fmh_keys[0].fmr_offset ||
896 	    (head.fmh_keys[1].fmr_offset != 0 &&
897 	     head.fmh_keys[1].fmr_offset != -1ULL))
898 		return -EINVAL;
899 
900 	xhead.fmh_iflags = head.fmh_iflags;
901 	xhead.fmh_count = head.fmh_count;
902 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
903 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
904 
905 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
906 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
907 
908 	info.gi_sb = sb;
909 	info.gi_data = arg;
910 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
911 	if (error == EXT4_QUERY_RANGE_ABORT)
912 		aborted = true;
913 	else if (error)
914 		return error;
915 
916 	/* If we didn't abort, set the "last" flag in the last fmx */
917 	if (!aborted && info.gi_idx) {
918 		info.gi_last_flags |= FMR_OF_LAST;
919 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
920 				 &info.gi_last_flags,
921 				 sizeof(info.gi_last_flags)))
922 			return -EFAULT;
923 	}
924 
925 	/* copy back header */
926 	head.fmh_entries = xhead.fmh_entries;
927 	head.fmh_oflags = xhead.fmh_oflags;
928 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
929 		return -EFAULT;
930 
931 	return 0;
932 }
933 
934 static long ext4_ioctl_group_add(struct file *file,
935 				 struct ext4_new_group_data *input)
936 {
937 	struct super_block *sb = file_inode(file)->i_sb;
938 	int err, err2=0;
939 
940 	err = ext4_resize_begin(sb);
941 	if (err)
942 		return err;
943 
944 	if (ext4_has_feature_bigalloc(sb)) {
945 		ext4_msg(sb, KERN_ERR,
946 			 "Online resizing not supported with bigalloc");
947 		err = -EOPNOTSUPP;
948 		goto group_add_out;
949 	}
950 
951 	err = mnt_want_write_file(file);
952 	if (err)
953 		goto group_add_out;
954 
955 	err = ext4_group_add(sb, input);
956 	if (EXT4_SB(sb)->s_journal) {
957 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
958 		err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
959 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
960 	}
961 	if (err == 0)
962 		err = err2;
963 	mnt_drop_write_file(file);
964 	if (!err && ext4_has_group_desc_csum(sb) &&
965 	    test_opt(sb, INIT_INODE_TABLE))
966 		err = ext4_register_li_request(sb, input->group);
967 group_add_out:
968 	err2 = ext4_resize_end(sb, false);
969 	if (err == 0)
970 		err = err2;
971 	return err;
972 }
973 
974 int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa)
975 {
976 	struct inode *inode = d_inode(dentry);
977 	struct ext4_inode_info *ei = EXT4_I(inode);
978 	u32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
979 
980 	if (S_ISREG(inode->i_mode))
981 		flags &= ~FS_PROJINHERIT_FL;
982 
983 	fileattr_fill_flags(fa, flags);
984 	if (ext4_has_feature_project(inode->i_sb))
985 		fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
986 
987 	return 0;
988 }
989 
990 int ext4_fileattr_set(struct mnt_idmap *idmap,
991 		      struct dentry *dentry, struct fileattr *fa)
992 {
993 	struct inode *inode = d_inode(dentry);
994 	u32 flags = fa->flags;
995 	int err = -EOPNOTSUPP;
996 
997 	if (flags & ~EXT4_FL_USER_VISIBLE)
998 		goto out;
999 
1000 	/*
1001 	 * chattr(1) grabs flags via GETFLAGS, modifies the result and
1002 	 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
1003 	 * more restrictive than just silently masking off visible but
1004 	 * not settable flags as we always did.
1005 	 */
1006 	flags &= EXT4_FL_USER_MODIFIABLE;
1007 	if (ext4_mask_flags(inode->i_mode, flags) != flags)
1008 		goto out;
1009 	err = ext4_ioctl_check_immutable(inode, fa->fsx_projid, flags);
1010 	if (err)
1011 		goto out;
1012 	err = ext4_ioctl_setflags(inode, flags);
1013 	if (err)
1014 		goto out;
1015 	err = ext4_ioctl_setproject(inode, fa->fsx_projid);
1016 out:
1017 	return err;
1018 }
1019 
1020 /* So that the fiemap access checks can't overflow on 32 bit machines. */
1021 #define FIEMAP_MAX_EXTENTS	(UINT_MAX / sizeof(struct fiemap_extent))
1022 
1023 static int ext4_ioctl_get_es_cache(struct file *filp, unsigned long arg)
1024 {
1025 	struct fiemap fiemap;
1026 	struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
1027 	struct fiemap_extent_info fieinfo = { 0, };
1028 	struct inode *inode = file_inode(filp);
1029 	int error;
1030 
1031 	if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
1032 		return -EFAULT;
1033 
1034 	if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
1035 		return -EINVAL;
1036 
1037 	fieinfo.fi_flags = fiemap.fm_flags;
1038 	fieinfo.fi_extents_max = fiemap.fm_extent_count;
1039 	fieinfo.fi_extents_start = ufiemap->fm_extents;
1040 
1041 	error = ext4_get_es_cache(inode, &fieinfo, fiemap.fm_start,
1042 			fiemap.fm_length);
1043 	fiemap.fm_flags = fieinfo.fi_flags;
1044 	fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
1045 	if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
1046 		error = -EFAULT;
1047 
1048 	return error;
1049 }
1050 
1051 static int ext4_ioctl_checkpoint(struct file *filp, unsigned long arg)
1052 {
1053 	int err = 0;
1054 	__u32 flags = 0;
1055 	unsigned int flush_flags = 0;
1056 	struct super_block *sb = file_inode(filp)->i_sb;
1057 
1058 	if (copy_from_user(&flags, (__u32 __user *)arg,
1059 				sizeof(__u32)))
1060 		return -EFAULT;
1061 
1062 	if (!capable(CAP_SYS_ADMIN))
1063 		return -EPERM;
1064 
1065 	/* check for invalid bits set */
1066 	if ((flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID) ||
1067 				((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1068 				(flags & JBD2_JOURNAL_FLUSH_ZEROOUT)))
1069 		return -EINVAL;
1070 
1071 	if (!EXT4_SB(sb)->s_journal)
1072 		return -ENODEV;
1073 
1074 	if ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1075 	    !bdev_max_discard_sectors(EXT4_SB(sb)->s_journal->j_dev))
1076 		return -EOPNOTSUPP;
1077 
1078 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
1079 		return 0;
1080 
1081 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DISCARD)
1082 		flush_flags |= JBD2_JOURNAL_FLUSH_DISCARD;
1083 
1084 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT) {
1085 		flush_flags |= JBD2_JOURNAL_FLUSH_ZEROOUT;
1086 		pr_info_ratelimited("warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow");
1087 	}
1088 
1089 	jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1090 	err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, flush_flags);
1091 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1092 
1093 	return err;
1094 }
1095 
1096 static int ext4_ioctl_setlabel(struct file *filp, const char __user *user_label)
1097 {
1098 	size_t len;
1099 	int ret = 0;
1100 	char new_label[EXT4_LABEL_MAX + 1];
1101 	struct super_block *sb = file_inode(filp)->i_sb;
1102 
1103 	if (!capable(CAP_SYS_ADMIN))
1104 		return -EPERM;
1105 
1106 	/*
1107 	 * Copy the maximum length allowed for ext4 label with one more to
1108 	 * find the required terminating null byte in order to test the
1109 	 * label length. The on disk label doesn't need to be null terminated.
1110 	 */
1111 	if (copy_from_user(new_label, user_label, EXT4_LABEL_MAX + 1))
1112 		return -EFAULT;
1113 
1114 	len = strnlen(new_label, EXT4_LABEL_MAX + 1);
1115 	if (len > EXT4_LABEL_MAX)
1116 		return -EINVAL;
1117 
1118 	/*
1119 	 * Clear the buffer after the new label
1120 	 */
1121 	memset(new_label + len, 0, EXT4_LABEL_MAX - len);
1122 
1123 	ret = mnt_want_write_file(filp);
1124 	if (ret)
1125 		return ret;
1126 
1127 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setlabel, new_label);
1128 
1129 	mnt_drop_write_file(filp);
1130 	return ret;
1131 }
1132 
1133 static int ext4_ioctl_getlabel(struct ext4_sb_info *sbi, char __user *user_label)
1134 {
1135 	char label[EXT4_LABEL_MAX + 1];
1136 
1137 	/*
1138 	 * EXT4_LABEL_MAX must always be smaller than FSLABEL_MAX because
1139 	 * FSLABEL_MAX must include terminating null byte, while s_volume_name
1140 	 * does not have to.
1141 	 */
1142 	BUILD_BUG_ON(EXT4_LABEL_MAX >= FSLABEL_MAX);
1143 
1144 	memset(label, 0, sizeof(label));
1145 	lock_buffer(sbi->s_sbh);
1146 	strncpy(label, sbi->s_es->s_volume_name, EXT4_LABEL_MAX);
1147 	unlock_buffer(sbi->s_sbh);
1148 
1149 	if (copy_to_user(user_label, label, sizeof(label)))
1150 		return -EFAULT;
1151 	return 0;
1152 }
1153 
1154 static int ext4_ioctl_getuuid(struct ext4_sb_info *sbi,
1155 			struct fsuuid __user *ufsuuid)
1156 {
1157 	struct fsuuid fsuuid;
1158 	__u8 uuid[UUID_SIZE];
1159 
1160 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1161 		return -EFAULT;
1162 
1163 	if (fsuuid.fsu_len == 0) {
1164 		fsuuid.fsu_len = UUID_SIZE;
1165 		if (copy_to_user(&ufsuuid->fsu_len, &fsuuid.fsu_len,
1166 					sizeof(fsuuid.fsu_len)))
1167 			return -EFAULT;
1168 		return 0;
1169 	}
1170 
1171 	if (fsuuid.fsu_len < UUID_SIZE || fsuuid.fsu_flags != 0)
1172 		return -EINVAL;
1173 
1174 	lock_buffer(sbi->s_sbh);
1175 	memcpy(uuid, sbi->s_es->s_uuid, UUID_SIZE);
1176 	unlock_buffer(sbi->s_sbh);
1177 
1178 	fsuuid.fsu_len = UUID_SIZE;
1179 	if (copy_to_user(ufsuuid, &fsuuid, sizeof(fsuuid)) ||
1180 	    copy_to_user(&ufsuuid->fsu_uuid[0], uuid, UUID_SIZE))
1181 		return -EFAULT;
1182 	return 0;
1183 }
1184 
1185 static int ext4_ioctl_setuuid(struct file *filp,
1186 			const struct fsuuid __user *ufsuuid)
1187 {
1188 	int ret = 0;
1189 	struct super_block *sb = file_inode(filp)->i_sb;
1190 	struct fsuuid fsuuid;
1191 	__u8 uuid[UUID_SIZE];
1192 
1193 	if (!capable(CAP_SYS_ADMIN))
1194 		return -EPERM;
1195 
1196 	/*
1197 	 * If any checksums (group descriptors or metadata) are being used
1198 	 * then the checksum seed feature is required to change the UUID.
1199 	 */
1200 	if (((ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb))
1201 			&& !ext4_has_feature_csum_seed(sb))
1202 		|| ext4_has_feature_stable_inodes(sb))
1203 		return -EOPNOTSUPP;
1204 
1205 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1206 		return -EFAULT;
1207 
1208 	if (fsuuid.fsu_len != UUID_SIZE || fsuuid.fsu_flags != 0)
1209 		return -EINVAL;
1210 
1211 	if (copy_from_user(uuid, &ufsuuid->fsu_uuid[0], UUID_SIZE))
1212 		return -EFAULT;
1213 
1214 	ret = mnt_want_write_file(filp);
1215 	if (ret)
1216 		return ret;
1217 
1218 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setuuid, &uuid);
1219 	mnt_drop_write_file(filp);
1220 
1221 	return ret;
1222 }
1223 
1224 static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1225 {
1226 	struct inode *inode = file_inode(filp);
1227 	struct super_block *sb = inode->i_sb;
1228 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
1229 
1230 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
1231 
1232 	switch (cmd) {
1233 	case FS_IOC_GETFSMAP:
1234 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
1235 	case EXT4_IOC_GETVERSION:
1236 	case EXT4_IOC_GETVERSION_OLD:
1237 		return put_user(inode->i_generation, (int __user *) arg);
1238 	case EXT4_IOC_SETVERSION:
1239 	case EXT4_IOC_SETVERSION_OLD: {
1240 		handle_t *handle;
1241 		struct ext4_iloc iloc;
1242 		__u32 generation;
1243 		int err;
1244 
1245 		if (!inode_owner_or_capable(idmap, inode))
1246 			return -EPERM;
1247 
1248 		if (ext4_has_metadata_csum(inode->i_sb)) {
1249 			ext4_warning(sb, "Setting inode version is not "
1250 				     "supported with metadata_csum enabled.");
1251 			return -ENOTTY;
1252 		}
1253 
1254 		err = mnt_want_write_file(filp);
1255 		if (err)
1256 			return err;
1257 		if (get_user(generation, (int __user *) arg)) {
1258 			err = -EFAULT;
1259 			goto setversion_out;
1260 		}
1261 
1262 		inode_lock(inode);
1263 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
1264 		if (IS_ERR(handle)) {
1265 			err = PTR_ERR(handle);
1266 			goto unlock_out;
1267 		}
1268 		err = ext4_reserve_inode_write(handle, inode, &iloc);
1269 		if (err == 0) {
1270 			inode_set_ctime_current(inode);
1271 			inode_inc_iversion(inode);
1272 			inode->i_generation = generation;
1273 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
1274 		}
1275 		ext4_journal_stop(handle);
1276 
1277 unlock_out:
1278 		inode_unlock(inode);
1279 setversion_out:
1280 		mnt_drop_write_file(filp);
1281 		return err;
1282 	}
1283 	case EXT4_IOC_GROUP_EXTEND: {
1284 		ext4_fsblk_t n_blocks_count;
1285 		int err, err2=0;
1286 
1287 		err = ext4_resize_begin(sb);
1288 		if (err)
1289 			return err;
1290 
1291 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
1292 			err = -EFAULT;
1293 			goto group_extend_out;
1294 		}
1295 
1296 		if (ext4_has_feature_bigalloc(sb)) {
1297 			ext4_msg(sb, KERN_ERR,
1298 				 "Online resizing not supported with bigalloc");
1299 			err = -EOPNOTSUPP;
1300 			goto group_extend_out;
1301 		}
1302 
1303 		err = mnt_want_write_file(filp);
1304 		if (err)
1305 			goto group_extend_out;
1306 
1307 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
1308 		if (EXT4_SB(sb)->s_journal) {
1309 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1310 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1311 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1312 		}
1313 		if (err == 0)
1314 			err = err2;
1315 		mnt_drop_write_file(filp);
1316 group_extend_out:
1317 		err2 = ext4_resize_end(sb, false);
1318 		if (err == 0)
1319 			err = err2;
1320 		return err;
1321 	}
1322 
1323 	case EXT4_IOC_MOVE_EXT: {
1324 		struct move_extent me;
1325 		struct fd donor;
1326 		int err;
1327 
1328 		if (!(filp->f_mode & FMODE_READ) ||
1329 		    !(filp->f_mode & FMODE_WRITE))
1330 			return -EBADF;
1331 
1332 		if (copy_from_user(&me,
1333 			(struct move_extent __user *)arg, sizeof(me)))
1334 			return -EFAULT;
1335 		me.moved_len = 0;
1336 
1337 		donor = fdget(me.donor_fd);
1338 		if (!donor.file)
1339 			return -EBADF;
1340 
1341 		if (!(donor.file->f_mode & FMODE_WRITE)) {
1342 			err = -EBADF;
1343 			goto mext_out;
1344 		}
1345 
1346 		if (ext4_has_feature_bigalloc(sb)) {
1347 			ext4_msg(sb, KERN_ERR,
1348 				 "Online defrag not supported with bigalloc");
1349 			err = -EOPNOTSUPP;
1350 			goto mext_out;
1351 		} else if (IS_DAX(inode)) {
1352 			ext4_msg(sb, KERN_ERR,
1353 				 "Online defrag not supported with DAX");
1354 			err = -EOPNOTSUPP;
1355 			goto mext_out;
1356 		}
1357 
1358 		err = mnt_want_write_file(filp);
1359 		if (err)
1360 			goto mext_out;
1361 
1362 		err = ext4_move_extents(filp, donor.file, me.orig_start,
1363 					me.donor_start, me.len, &me.moved_len);
1364 		mnt_drop_write_file(filp);
1365 
1366 		if (copy_to_user((struct move_extent __user *)arg,
1367 				 &me, sizeof(me)))
1368 			err = -EFAULT;
1369 mext_out:
1370 		fdput(donor);
1371 		return err;
1372 	}
1373 
1374 	case EXT4_IOC_GROUP_ADD: {
1375 		struct ext4_new_group_data input;
1376 
1377 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
1378 				sizeof(input)))
1379 			return -EFAULT;
1380 
1381 		return ext4_ioctl_group_add(filp, &input);
1382 	}
1383 
1384 	case EXT4_IOC_MIGRATE:
1385 	{
1386 		int err;
1387 		if (!inode_owner_or_capable(idmap, inode))
1388 			return -EACCES;
1389 
1390 		err = mnt_want_write_file(filp);
1391 		if (err)
1392 			return err;
1393 		/*
1394 		 * inode_mutex prevent write and truncate on the file.
1395 		 * Read still goes through. We take i_data_sem in
1396 		 * ext4_ext_swap_inode_data before we switch the
1397 		 * inode format to prevent read.
1398 		 */
1399 		inode_lock((inode));
1400 		err = ext4_ext_migrate(inode);
1401 		inode_unlock((inode));
1402 		mnt_drop_write_file(filp);
1403 		return err;
1404 	}
1405 
1406 	case EXT4_IOC_ALLOC_DA_BLKS:
1407 	{
1408 		int err;
1409 		if (!inode_owner_or_capable(idmap, inode))
1410 			return -EACCES;
1411 
1412 		err = mnt_want_write_file(filp);
1413 		if (err)
1414 			return err;
1415 		err = ext4_alloc_da_blocks(inode);
1416 		mnt_drop_write_file(filp);
1417 		return err;
1418 	}
1419 
1420 	case EXT4_IOC_SWAP_BOOT:
1421 	{
1422 		int err;
1423 		if (!(filp->f_mode & FMODE_WRITE))
1424 			return -EBADF;
1425 		err = mnt_want_write_file(filp);
1426 		if (err)
1427 			return err;
1428 		err = swap_inode_boot_loader(sb, idmap, inode);
1429 		mnt_drop_write_file(filp);
1430 		return err;
1431 	}
1432 
1433 	case EXT4_IOC_RESIZE_FS: {
1434 		ext4_fsblk_t n_blocks_count;
1435 		int err = 0, err2 = 0;
1436 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
1437 
1438 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1439 				   sizeof(__u64))) {
1440 			return -EFAULT;
1441 		}
1442 
1443 		err = ext4_resize_begin(sb);
1444 		if (err)
1445 			return err;
1446 
1447 		err = mnt_want_write_file(filp);
1448 		if (err)
1449 			goto resizefs_out;
1450 
1451 		err = ext4_resize_fs(sb, n_blocks_count);
1452 		if (EXT4_SB(sb)->s_journal) {
1453 			ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE, NULL);
1454 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1455 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1456 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1457 		}
1458 		if (err == 0)
1459 			err = err2;
1460 		mnt_drop_write_file(filp);
1461 		if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1462 		    ext4_has_group_desc_csum(sb) &&
1463 		    test_opt(sb, INIT_INODE_TABLE))
1464 			err = ext4_register_li_request(sb, o_group);
1465 
1466 resizefs_out:
1467 		err2 = ext4_resize_end(sb, true);
1468 		if (err == 0)
1469 			err = err2;
1470 		return err;
1471 	}
1472 
1473 	case FITRIM:
1474 	{
1475 		struct fstrim_range range;
1476 		int ret = 0;
1477 
1478 		if (!capable(CAP_SYS_ADMIN))
1479 			return -EPERM;
1480 
1481 		if (!bdev_max_discard_sectors(sb->s_bdev))
1482 			return -EOPNOTSUPP;
1483 
1484 		/*
1485 		 * We haven't replayed the journal, so we cannot use our
1486 		 * block-bitmap-guided storage zapping commands.
1487 		 */
1488 		if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1489 			return -EROFS;
1490 
1491 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1492 		    sizeof(range)))
1493 			return -EFAULT;
1494 
1495 		ret = ext4_trim_fs(sb, &range);
1496 		if (ret < 0)
1497 			return ret;
1498 
1499 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
1500 		    sizeof(range)))
1501 			return -EFAULT;
1502 
1503 		return 0;
1504 	}
1505 	case EXT4_IOC_PRECACHE_EXTENTS:
1506 		return ext4_ext_precache(inode);
1507 
1508 	case FS_IOC_SET_ENCRYPTION_POLICY:
1509 		if (!ext4_has_feature_encrypt(sb))
1510 			return -EOPNOTSUPP;
1511 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1512 
1513 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1514 		return ext4_ioctl_get_encryption_pwsalt(filp, (void __user *)arg);
1515 
1516 	case FS_IOC_GET_ENCRYPTION_POLICY:
1517 		if (!ext4_has_feature_encrypt(sb))
1518 			return -EOPNOTSUPP;
1519 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1520 
1521 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1522 		if (!ext4_has_feature_encrypt(sb))
1523 			return -EOPNOTSUPP;
1524 		return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
1525 
1526 	case FS_IOC_ADD_ENCRYPTION_KEY:
1527 		if (!ext4_has_feature_encrypt(sb))
1528 			return -EOPNOTSUPP;
1529 		return fscrypt_ioctl_add_key(filp, (void __user *)arg);
1530 
1531 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1532 		if (!ext4_has_feature_encrypt(sb))
1533 			return -EOPNOTSUPP;
1534 		return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
1535 
1536 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1537 		if (!ext4_has_feature_encrypt(sb))
1538 			return -EOPNOTSUPP;
1539 		return fscrypt_ioctl_remove_key_all_users(filp,
1540 							  (void __user *)arg);
1541 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1542 		if (!ext4_has_feature_encrypt(sb))
1543 			return -EOPNOTSUPP;
1544 		return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
1545 
1546 	case FS_IOC_GET_ENCRYPTION_NONCE:
1547 		if (!ext4_has_feature_encrypt(sb))
1548 			return -EOPNOTSUPP;
1549 		return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
1550 
1551 	case EXT4_IOC_CLEAR_ES_CACHE:
1552 	{
1553 		if (!inode_owner_or_capable(idmap, inode))
1554 			return -EACCES;
1555 		ext4_clear_inode_es(inode);
1556 		return 0;
1557 	}
1558 
1559 	case EXT4_IOC_GETSTATE:
1560 	{
1561 		__u32	state = 0;
1562 
1563 		if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED))
1564 			state |= EXT4_STATE_FLAG_EXT_PRECACHED;
1565 		if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
1566 			state |= EXT4_STATE_FLAG_NEW;
1567 		if (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
1568 			state |= EXT4_STATE_FLAG_NEWENTRY;
1569 		if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE))
1570 			state |= EXT4_STATE_FLAG_DA_ALLOC_CLOSE;
1571 
1572 		return put_user(state, (__u32 __user *) arg);
1573 	}
1574 
1575 	case EXT4_IOC_GET_ES_CACHE:
1576 		return ext4_ioctl_get_es_cache(filp, arg);
1577 
1578 	case EXT4_IOC_SHUTDOWN:
1579 		return ext4_ioctl_shutdown(sb, arg);
1580 
1581 	case FS_IOC_ENABLE_VERITY:
1582 		if (!ext4_has_feature_verity(sb))
1583 			return -EOPNOTSUPP;
1584 		return fsverity_ioctl_enable(filp, (const void __user *)arg);
1585 
1586 	case FS_IOC_MEASURE_VERITY:
1587 		if (!ext4_has_feature_verity(sb))
1588 			return -EOPNOTSUPP;
1589 		return fsverity_ioctl_measure(filp, (void __user *)arg);
1590 
1591 	case FS_IOC_READ_VERITY_METADATA:
1592 		if (!ext4_has_feature_verity(sb))
1593 			return -EOPNOTSUPP;
1594 		return fsverity_ioctl_read_metadata(filp,
1595 						    (const void __user *)arg);
1596 
1597 	case EXT4_IOC_CHECKPOINT:
1598 		return ext4_ioctl_checkpoint(filp, arg);
1599 
1600 	case FS_IOC_GETFSLABEL:
1601 		return ext4_ioctl_getlabel(EXT4_SB(sb), (void __user *)arg);
1602 
1603 	case FS_IOC_SETFSLABEL:
1604 		return ext4_ioctl_setlabel(filp,
1605 					   (const void __user *)arg);
1606 
1607 	case EXT4_IOC_GETFSUUID:
1608 		return ext4_ioctl_getuuid(EXT4_SB(sb), (void __user *)arg);
1609 	case EXT4_IOC_SETFSUUID:
1610 		return ext4_ioctl_setuuid(filp, (const void __user *)arg);
1611 	default:
1612 		return -ENOTTY;
1613 	}
1614 }
1615 
1616 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1617 {
1618 	return __ext4_ioctl(filp, cmd, arg);
1619 }
1620 
1621 #ifdef CONFIG_COMPAT
1622 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1623 {
1624 	/* These are just misnamed, they actually get/put from/to user an int */
1625 	switch (cmd) {
1626 	case EXT4_IOC32_GETVERSION:
1627 		cmd = EXT4_IOC_GETVERSION;
1628 		break;
1629 	case EXT4_IOC32_SETVERSION:
1630 		cmd = EXT4_IOC_SETVERSION;
1631 		break;
1632 	case EXT4_IOC32_GROUP_EXTEND:
1633 		cmd = EXT4_IOC_GROUP_EXTEND;
1634 		break;
1635 	case EXT4_IOC32_GETVERSION_OLD:
1636 		cmd = EXT4_IOC_GETVERSION_OLD;
1637 		break;
1638 	case EXT4_IOC32_SETVERSION_OLD:
1639 		cmd = EXT4_IOC_SETVERSION_OLD;
1640 		break;
1641 	case EXT4_IOC32_GETRSVSZ:
1642 		cmd = EXT4_IOC_GETRSVSZ;
1643 		break;
1644 	case EXT4_IOC32_SETRSVSZ:
1645 		cmd = EXT4_IOC_SETRSVSZ;
1646 		break;
1647 	case EXT4_IOC32_GROUP_ADD: {
1648 		struct compat_ext4_new_group_input __user *uinput;
1649 		struct ext4_new_group_data input;
1650 		int err;
1651 
1652 		uinput = compat_ptr(arg);
1653 		err = get_user(input.group, &uinput->group);
1654 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1655 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1656 		err |= get_user(input.inode_table, &uinput->inode_table);
1657 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1658 		err |= get_user(input.reserved_blocks,
1659 				&uinput->reserved_blocks);
1660 		if (err)
1661 			return -EFAULT;
1662 		return ext4_ioctl_group_add(file, &input);
1663 	}
1664 	case EXT4_IOC_MOVE_EXT:
1665 	case EXT4_IOC_RESIZE_FS:
1666 	case FITRIM:
1667 	case EXT4_IOC_PRECACHE_EXTENTS:
1668 	case FS_IOC_SET_ENCRYPTION_POLICY:
1669 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1670 	case FS_IOC_GET_ENCRYPTION_POLICY:
1671 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1672 	case FS_IOC_ADD_ENCRYPTION_KEY:
1673 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1674 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1675 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1676 	case FS_IOC_GET_ENCRYPTION_NONCE:
1677 	case EXT4_IOC_SHUTDOWN:
1678 	case FS_IOC_GETFSMAP:
1679 	case FS_IOC_ENABLE_VERITY:
1680 	case FS_IOC_MEASURE_VERITY:
1681 	case FS_IOC_READ_VERITY_METADATA:
1682 	case EXT4_IOC_CLEAR_ES_CACHE:
1683 	case EXT4_IOC_GETSTATE:
1684 	case EXT4_IOC_GET_ES_CACHE:
1685 	case EXT4_IOC_CHECKPOINT:
1686 	case FS_IOC_GETFSLABEL:
1687 	case FS_IOC_SETFSLABEL:
1688 	case EXT4_IOC_GETFSUUID:
1689 	case EXT4_IOC_SETFSUUID:
1690 		break;
1691 	default:
1692 		return -ENOIOCTLCMD;
1693 	}
1694 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1695 }
1696 #endif
1697 
1698 static void set_overhead(struct ext4_super_block *es, const void *arg)
1699 {
1700 	es->s_overhead_clusters = cpu_to_le32(*((unsigned long *) arg));
1701 }
1702 
1703 int ext4_update_overhead(struct super_block *sb, bool force)
1704 {
1705 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1706 
1707 	if (sb_rdonly(sb))
1708 		return 0;
1709 	if (!force &&
1710 	    (sbi->s_overhead == 0 ||
1711 	     sbi->s_overhead == le32_to_cpu(sbi->s_es->s_overhead_clusters)))
1712 		return 0;
1713 	return ext4_update_superblocks_fn(sb, set_overhead, &sbi->s_overhead);
1714 }
1715