xref: /openbmc/linux/fs/ext4/ioctl.c (revision c4d13222afd8a64bf11bc7ec68645496ee8b54b9)
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->i_ctime = inode_bl->i_ctime = current_time(inode);
453 	inode_inc_iversion(inode);
454 
455 	inode->i_generation = get_random_u32();
456 	inode_bl->i_generation = get_random_u32();
457 	ext4_reset_inode_seed(inode);
458 	ext4_reset_inode_seed(inode_bl);
459 
460 	ext4_discard_preallocations(inode, 0);
461 
462 	err = ext4_mark_inode_dirty(handle, inode);
463 	if (err < 0) {
464 		/* No need to update quota information. */
465 		ext4_warning(inode->i_sb,
466 			"couldn't mark inode #%lu dirty (err %d)",
467 			inode->i_ino, err);
468 		/* Revert all changes: */
469 		swap_inode_data(inode, inode_bl);
470 		ext4_mark_inode_dirty(handle, inode);
471 		goto err_out1;
472 	}
473 
474 	blocks = inode_bl->i_blocks;
475 	bytes = inode_bl->i_bytes;
476 	inode_bl->i_blocks = inode->i_blocks;
477 	inode_bl->i_bytes = inode->i_bytes;
478 	err = ext4_mark_inode_dirty(handle, inode_bl);
479 	if (err < 0) {
480 		/* No need to update quota information. */
481 		ext4_warning(inode_bl->i_sb,
482 			"couldn't mark inode #%lu dirty (err %d)",
483 			inode_bl->i_ino, err);
484 		goto revert;
485 	}
486 
487 	/* Bootloader inode should not be counted into quota information. */
488 	if (diff > 0)
489 		dquot_free_space(inode, diff);
490 	else
491 		err = dquot_alloc_space(inode, -1 * diff);
492 
493 	if (err < 0) {
494 revert:
495 		/* Revert all changes: */
496 		inode_bl->i_blocks = blocks;
497 		inode_bl->i_bytes = bytes;
498 		swap_inode_data(inode, inode_bl);
499 		ext4_mark_inode_dirty(handle, inode);
500 		ext4_mark_inode_dirty(handle, inode_bl);
501 	}
502 
503 err_out1:
504 	ext4_journal_stop(handle);
505 	ext4_double_up_write_data_sem(inode, inode_bl);
506 
507 err_out:
508 	filemap_invalidate_unlock(inode->i_mapping);
509 journal_err_out:
510 	unlock_two_nondirectories(inode, inode_bl);
511 	iput(inode_bl);
512 	return err;
513 }
514 
515 /*
516  * If immutable is set and we are not clearing it, we're not allowed to change
517  * anything else in the inode.  Don't error out if we're only trying to set
518  * immutable on an immutable file.
519  */
520 static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
521 				      unsigned int flags)
522 {
523 	struct ext4_inode_info *ei = EXT4_I(inode);
524 	unsigned int oldflags = ei->i_flags;
525 
526 	if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
527 		return 0;
528 
529 	if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
530 		return -EPERM;
531 	if (ext4_has_feature_project(inode->i_sb) &&
532 	    __kprojid_val(ei->i_projid) != new_projid)
533 		return -EPERM;
534 
535 	return 0;
536 }
537 
538 static void ext4_dax_dontcache(struct inode *inode, unsigned int flags)
539 {
540 	struct ext4_inode_info *ei = EXT4_I(inode);
541 
542 	if (S_ISDIR(inode->i_mode))
543 		return;
544 
545 	if (test_opt2(inode->i_sb, DAX_NEVER) ||
546 	    test_opt(inode->i_sb, DAX_ALWAYS))
547 		return;
548 
549 	if ((ei->i_flags ^ flags) & EXT4_DAX_FL)
550 		d_mark_dontcache(inode);
551 }
552 
553 static bool dax_compatible(struct inode *inode, unsigned int oldflags,
554 			   unsigned int flags)
555 {
556 	/* Allow the DAX flag to be changed on inline directories */
557 	if (S_ISDIR(inode->i_mode)) {
558 		flags &= ~EXT4_INLINE_DATA_FL;
559 		oldflags &= ~EXT4_INLINE_DATA_FL;
560 	}
561 
562 	if (flags & EXT4_DAX_FL) {
563 		if ((oldflags & EXT4_DAX_MUT_EXCL) ||
564 		     ext4_test_inode_state(inode,
565 					  EXT4_STATE_VERITY_IN_PROGRESS)) {
566 			return false;
567 		}
568 	}
569 
570 	if ((flags & EXT4_DAX_MUT_EXCL) && (oldflags & EXT4_DAX_FL))
571 			return false;
572 
573 	return true;
574 }
575 
576 static int ext4_ioctl_setflags(struct inode *inode,
577 			       unsigned int flags)
578 {
579 	struct ext4_inode_info *ei = EXT4_I(inode);
580 	handle_t *handle = NULL;
581 	int err = -EPERM, migrate = 0;
582 	struct ext4_iloc iloc;
583 	unsigned int oldflags, mask, i;
584 	struct super_block *sb = inode->i_sb;
585 
586 	/* Is it quota file? Do not allow user to mess with it */
587 	if (ext4_is_quota_file(inode))
588 		goto flags_out;
589 
590 	oldflags = ei->i_flags;
591 	/*
592 	 * The JOURNAL_DATA flag can only be changed by
593 	 * the relevant capability.
594 	 */
595 	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
596 		if (!capable(CAP_SYS_RESOURCE))
597 			goto flags_out;
598 	}
599 
600 	if (!dax_compatible(inode, oldflags, flags)) {
601 		err = -EOPNOTSUPP;
602 		goto flags_out;
603 	}
604 
605 	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
606 		migrate = 1;
607 
608 	if ((flags ^ oldflags) & EXT4_CASEFOLD_FL) {
609 		if (!ext4_has_feature_casefold(sb)) {
610 			err = -EOPNOTSUPP;
611 			goto flags_out;
612 		}
613 
614 		if (!S_ISDIR(inode->i_mode)) {
615 			err = -ENOTDIR;
616 			goto flags_out;
617 		}
618 
619 		if (!ext4_empty_dir(inode)) {
620 			err = -ENOTEMPTY;
621 			goto flags_out;
622 		}
623 	}
624 
625 	/*
626 	 * Wait for all pending directio and then flush all the dirty pages
627 	 * for this file.  The flush marks all the pages readonly, so any
628 	 * subsequent attempt to write to the file (particularly mmap pages)
629 	 * will come through the filesystem and fail.
630 	 */
631 	if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
632 	    (flags & EXT4_IMMUTABLE_FL)) {
633 		inode_dio_wait(inode);
634 		err = filemap_write_and_wait(inode->i_mapping);
635 		if (err)
636 			goto flags_out;
637 	}
638 
639 	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
640 	if (IS_ERR(handle)) {
641 		err = PTR_ERR(handle);
642 		goto flags_out;
643 	}
644 	if (IS_SYNC(inode))
645 		ext4_handle_sync(handle);
646 	err = ext4_reserve_inode_write(handle, inode, &iloc);
647 	if (err)
648 		goto flags_err;
649 
650 	ext4_dax_dontcache(inode, flags);
651 
652 	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
653 		if (!(mask & EXT4_FL_USER_MODIFIABLE))
654 			continue;
655 		/* These flags get special treatment later */
656 		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
657 			continue;
658 		if (mask & flags)
659 			ext4_set_inode_flag(inode, i);
660 		else
661 			ext4_clear_inode_flag(inode, i);
662 	}
663 
664 	ext4_set_inode_flags(inode, false);
665 
666 	inode->i_ctime = current_time(inode);
667 	inode_inc_iversion(inode);
668 
669 	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
670 flags_err:
671 	ext4_journal_stop(handle);
672 	if (err)
673 		goto flags_out;
674 
675 	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
676 		/*
677 		 * Changes to the journaling mode can cause unsafe changes to
678 		 * S_DAX if the inode is DAX
679 		 */
680 		if (IS_DAX(inode)) {
681 			err = -EBUSY;
682 			goto flags_out;
683 		}
684 
685 		err = ext4_change_inode_journal_flag(inode,
686 						     flags & EXT4_JOURNAL_DATA_FL);
687 		if (err)
688 			goto flags_out;
689 	}
690 	if (migrate) {
691 		if (flags & EXT4_EXTENTS_FL)
692 			err = ext4_ext_migrate(inode);
693 		else
694 			err = ext4_ind_migrate(inode);
695 	}
696 
697 flags_out:
698 	return err;
699 }
700 
701 #ifdef CONFIG_QUOTA
702 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
703 {
704 	struct super_block *sb = inode->i_sb;
705 	struct ext4_inode_info *ei = EXT4_I(inode);
706 	int err, rc;
707 	handle_t *handle;
708 	kprojid_t kprojid;
709 	struct ext4_iloc iloc;
710 	struct ext4_inode *raw_inode;
711 	struct dquot *transfer_to[MAXQUOTAS] = { };
712 
713 	if (!ext4_has_feature_project(sb)) {
714 		if (projid != EXT4_DEF_PROJID)
715 			return -EOPNOTSUPP;
716 		else
717 			return 0;
718 	}
719 
720 	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
721 		return -EOPNOTSUPP;
722 
723 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
724 
725 	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
726 		return 0;
727 
728 	err = -EPERM;
729 	/* Is it quota file? Do not allow user to mess with it */
730 	if (ext4_is_quota_file(inode))
731 		return err;
732 
733 	err = dquot_initialize(inode);
734 	if (err)
735 		return err;
736 
737 	err = ext4_get_inode_loc(inode, &iloc);
738 	if (err)
739 		return err;
740 
741 	raw_inode = ext4_raw_inode(&iloc);
742 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
743 		err = ext4_expand_extra_isize(inode,
744 					      EXT4_SB(sb)->s_want_extra_isize,
745 					      &iloc);
746 		if (err)
747 			return err;
748 	} else {
749 		brelse(iloc.bh);
750 	}
751 
752 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
753 		EXT4_QUOTA_INIT_BLOCKS(sb) +
754 		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
755 	if (IS_ERR(handle))
756 		return PTR_ERR(handle);
757 
758 	err = ext4_reserve_inode_write(handle, inode, &iloc);
759 	if (err)
760 		goto out_stop;
761 
762 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
763 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
764 
765 		/* __dquot_transfer() calls back ext4_get_inode_usage() which
766 		 * counts xattr inode references.
767 		 */
768 		down_read(&EXT4_I(inode)->xattr_sem);
769 		err = __dquot_transfer(inode, transfer_to);
770 		up_read(&EXT4_I(inode)->xattr_sem);
771 		dqput(transfer_to[PRJQUOTA]);
772 		if (err)
773 			goto out_dirty;
774 	}
775 
776 	EXT4_I(inode)->i_projid = kprojid;
777 	inode->i_ctime = current_time(inode);
778 	inode_inc_iversion(inode);
779 out_dirty:
780 	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
781 	if (!err)
782 		err = rc;
783 out_stop:
784 	ext4_journal_stop(handle);
785 	return err;
786 }
787 #else
788 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
789 {
790 	if (projid != EXT4_DEF_PROJID)
791 		return -EOPNOTSUPP;
792 	return 0;
793 }
794 #endif
795 
796 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
797 {
798 	struct ext4_sb_info *sbi = EXT4_SB(sb);
799 	__u32 flags;
800 	int ret;
801 
802 	if (!capable(CAP_SYS_ADMIN))
803 		return -EPERM;
804 
805 	if (get_user(flags, (__u32 __user *)arg))
806 		return -EFAULT;
807 
808 	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
809 		return -EINVAL;
810 
811 	if (ext4_forced_shutdown(sbi))
812 		return 0;
813 
814 	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
815 	trace_ext4_shutdown(sb, flags);
816 
817 	switch (flags) {
818 	case EXT4_GOING_FLAGS_DEFAULT:
819 		ret = freeze_bdev(sb->s_bdev);
820 		if (ret)
821 			return ret;
822 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
823 		thaw_bdev(sb->s_bdev);
824 		break;
825 	case EXT4_GOING_FLAGS_LOGFLUSH:
826 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
827 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
828 			(void) ext4_force_commit(sb);
829 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
830 		}
831 		break;
832 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
833 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
834 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
835 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
836 		break;
837 	default:
838 		return -EINVAL;
839 	}
840 	clear_opt(sb, DISCARD);
841 	return 0;
842 }
843 
844 struct getfsmap_info {
845 	struct super_block	*gi_sb;
846 	struct fsmap_head __user *gi_data;
847 	unsigned int		gi_idx;
848 	__u32			gi_last_flags;
849 };
850 
851 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
852 {
853 	struct getfsmap_info *info = priv;
854 	struct fsmap fm;
855 
856 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
857 
858 	info->gi_last_flags = xfm->fmr_flags;
859 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
860 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
861 			sizeof(struct fsmap)))
862 		return -EFAULT;
863 
864 	return 0;
865 }
866 
867 static int ext4_ioc_getfsmap(struct super_block *sb,
868 			     struct fsmap_head __user *arg)
869 {
870 	struct getfsmap_info info = { NULL };
871 	struct ext4_fsmap_head xhead = {0};
872 	struct fsmap_head head;
873 	bool aborted = false;
874 	int error;
875 
876 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
877 		return -EFAULT;
878 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
879 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
880 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
881 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
882 		       sizeof(head.fmh_keys[1].fmr_reserved)))
883 		return -EINVAL;
884 	/*
885 	 * ext4 doesn't report file extents at all, so the only valid
886 	 * file offsets are the magic ones (all zeroes or all ones).
887 	 */
888 	if (head.fmh_keys[0].fmr_offset ||
889 	    (head.fmh_keys[1].fmr_offset != 0 &&
890 	     head.fmh_keys[1].fmr_offset != -1ULL))
891 		return -EINVAL;
892 
893 	xhead.fmh_iflags = head.fmh_iflags;
894 	xhead.fmh_count = head.fmh_count;
895 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
896 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
897 
898 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
899 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
900 
901 	info.gi_sb = sb;
902 	info.gi_data = arg;
903 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
904 	if (error == EXT4_QUERY_RANGE_ABORT)
905 		aborted = true;
906 	else if (error)
907 		return error;
908 
909 	/* If we didn't abort, set the "last" flag in the last fmx */
910 	if (!aborted && info.gi_idx) {
911 		info.gi_last_flags |= FMR_OF_LAST;
912 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
913 				 &info.gi_last_flags,
914 				 sizeof(info.gi_last_flags)))
915 			return -EFAULT;
916 	}
917 
918 	/* copy back header */
919 	head.fmh_entries = xhead.fmh_entries;
920 	head.fmh_oflags = xhead.fmh_oflags;
921 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
922 		return -EFAULT;
923 
924 	return 0;
925 }
926 
927 static long ext4_ioctl_group_add(struct file *file,
928 				 struct ext4_new_group_data *input)
929 {
930 	struct super_block *sb = file_inode(file)->i_sb;
931 	int err, err2=0;
932 
933 	err = ext4_resize_begin(sb);
934 	if (err)
935 		return err;
936 
937 	if (ext4_has_feature_bigalloc(sb)) {
938 		ext4_msg(sb, KERN_ERR,
939 			 "Online resizing not supported with bigalloc");
940 		err = -EOPNOTSUPP;
941 		goto group_add_out;
942 	}
943 
944 	err = mnt_want_write_file(file);
945 	if (err)
946 		goto group_add_out;
947 
948 	err = ext4_group_add(sb, input);
949 	if (EXT4_SB(sb)->s_journal) {
950 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
951 		err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
952 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
953 	}
954 	if (err == 0)
955 		err = err2;
956 	mnt_drop_write_file(file);
957 	if (!err && ext4_has_group_desc_csum(sb) &&
958 	    test_opt(sb, INIT_INODE_TABLE))
959 		err = ext4_register_li_request(sb, input->group);
960 group_add_out:
961 	err2 = ext4_resize_end(sb, false);
962 	if (err == 0)
963 		err = err2;
964 	return err;
965 }
966 
967 int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa)
968 {
969 	struct inode *inode = d_inode(dentry);
970 	struct ext4_inode_info *ei = EXT4_I(inode);
971 	u32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
972 
973 	if (S_ISREG(inode->i_mode))
974 		flags &= ~FS_PROJINHERIT_FL;
975 
976 	fileattr_fill_flags(fa, flags);
977 	if (ext4_has_feature_project(inode->i_sb))
978 		fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
979 
980 	return 0;
981 }
982 
983 int ext4_fileattr_set(struct mnt_idmap *idmap,
984 		      struct dentry *dentry, struct fileattr *fa)
985 {
986 	struct inode *inode = d_inode(dentry);
987 	u32 flags = fa->flags;
988 	int err = -EOPNOTSUPP;
989 
990 	if (flags & ~EXT4_FL_USER_VISIBLE)
991 		goto out;
992 
993 	/*
994 	 * chattr(1) grabs flags via GETFLAGS, modifies the result and
995 	 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
996 	 * more restrictive than just silently masking off visible but
997 	 * not settable flags as we always did.
998 	 */
999 	flags &= EXT4_FL_USER_MODIFIABLE;
1000 	if (ext4_mask_flags(inode->i_mode, flags) != flags)
1001 		goto out;
1002 	err = ext4_ioctl_check_immutable(inode, fa->fsx_projid, flags);
1003 	if (err)
1004 		goto out;
1005 	err = ext4_ioctl_setflags(inode, flags);
1006 	if (err)
1007 		goto out;
1008 	err = ext4_ioctl_setproject(inode, fa->fsx_projid);
1009 out:
1010 	return err;
1011 }
1012 
1013 /* So that the fiemap access checks can't overflow on 32 bit machines. */
1014 #define FIEMAP_MAX_EXTENTS	(UINT_MAX / sizeof(struct fiemap_extent))
1015 
1016 static int ext4_ioctl_get_es_cache(struct file *filp, unsigned long arg)
1017 {
1018 	struct fiemap fiemap;
1019 	struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
1020 	struct fiemap_extent_info fieinfo = { 0, };
1021 	struct inode *inode = file_inode(filp);
1022 	int error;
1023 
1024 	if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
1025 		return -EFAULT;
1026 
1027 	if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
1028 		return -EINVAL;
1029 
1030 	fieinfo.fi_flags = fiemap.fm_flags;
1031 	fieinfo.fi_extents_max = fiemap.fm_extent_count;
1032 	fieinfo.fi_extents_start = ufiemap->fm_extents;
1033 
1034 	error = ext4_get_es_cache(inode, &fieinfo, fiemap.fm_start,
1035 			fiemap.fm_length);
1036 	fiemap.fm_flags = fieinfo.fi_flags;
1037 	fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
1038 	if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
1039 		error = -EFAULT;
1040 
1041 	return error;
1042 }
1043 
1044 static int ext4_ioctl_checkpoint(struct file *filp, unsigned long arg)
1045 {
1046 	int err = 0;
1047 	__u32 flags = 0;
1048 	unsigned int flush_flags = 0;
1049 	struct super_block *sb = file_inode(filp)->i_sb;
1050 
1051 	if (copy_from_user(&flags, (__u32 __user *)arg,
1052 				sizeof(__u32)))
1053 		return -EFAULT;
1054 
1055 	if (!capable(CAP_SYS_ADMIN))
1056 		return -EPERM;
1057 
1058 	/* check for invalid bits set */
1059 	if ((flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID) ||
1060 				((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1061 				(flags & JBD2_JOURNAL_FLUSH_ZEROOUT)))
1062 		return -EINVAL;
1063 
1064 	if (!EXT4_SB(sb)->s_journal)
1065 		return -ENODEV;
1066 
1067 	if ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1068 	    !bdev_max_discard_sectors(EXT4_SB(sb)->s_journal->j_dev))
1069 		return -EOPNOTSUPP;
1070 
1071 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
1072 		return 0;
1073 
1074 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DISCARD)
1075 		flush_flags |= JBD2_JOURNAL_FLUSH_DISCARD;
1076 
1077 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT) {
1078 		flush_flags |= JBD2_JOURNAL_FLUSH_ZEROOUT;
1079 		pr_info_ratelimited("warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow");
1080 	}
1081 
1082 	jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1083 	err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, flush_flags);
1084 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1085 
1086 	return err;
1087 }
1088 
1089 static int ext4_ioctl_setlabel(struct file *filp, const char __user *user_label)
1090 {
1091 	size_t len;
1092 	int ret = 0;
1093 	char new_label[EXT4_LABEL_MAX + 1];
1094 	struct super_block *sb = file_inode(filp)->i_sb;
1095 
1096 	if (!capable(CAP_SYS_ADMIN))
1097 		return -EPERM;
1098 
1099 	/*
1100 	 * Copy the maximum length allowed for ext4 label with one more to
1101 	 * find the required terminating null byte in order to test the
1102 	 * label length. The on disk label doesn't need to be null terminated.
1103 	 */
1104 	if (copy_from_user(new_label, user_label, EXT4_LABEL_MAX + 1))
1105 		return -EFAULT;
1106 
1107 	len = strnlen(new_label, EXT4_LABEL_MAX + 1);
1108 	if (len > EXT4_LABEL_MAX)
1109 		return -EINVAL;
1110 
1111 	/*
1112 	 * Clear the buffer after the new label
1113 	 */
1114 	memset(new_label + len, 0, EXT4_LABEL_MAX - len);
1115 
1116 	ret = mnt_want_write_file(filp);
1117 	if (ret)
1118 		return ret;
1119 
1120 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setlabel, new_label);
1121 
1122 	mnt_drop_write_file(filp);
1123 	return ret;
1124 }
1125 
1126 static int ext4_ioctl_getlabel(struct ext4_sb_info *sbi, char __user *user_label)
1127 {
1128 	char label[EXT4_LABEL_MAX + 1];
1129 
1130 	/*
1131 	 * EXT4_LABEL_MAX must always be smaller than FSLABEL_MAX because
1132 	 * FSLABEL_MAX must include terminating null byte, while s_volume_name
1133 	 * does not have to.
1134 	 */
1135 	BUILD_BUG_ON(EXT4_LABEL_MAX >= FSLABEL_MAX);
1136 
1137 	memset(label, 0, sizeof(label));
1138 	lock_buffer(sbi->s_sbh);
1139 	strncpy(label, sbi->s_es->s_volume_name, EXT4_LABEL_MAX);
1140 	unlock_buffer(sbi->s_sbh);
1141 
1142 	if (copy_to_user(user_label, label, sizeof(label)))
1143 		return -EFAULT;
1144 	return 0;
1145 }
1146 
1147 static int ext4_ioctl_getuuid(struct ext4_sb_info *sbi,
1148 			struct fsuuid __user *ufsuuid)
1149 {
1150 	struct fsuuid fsuuid;
1151 	__u8 uuid[UUID_SIZE];
1152 
1153 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1154 		return -EFAULT;
1155 
1156 	if (fsuuid.fsu_len == 0) {
1157 		fsuuid.fsu_len = UUID_SIZE;
1158 		if (copy_to_user(&ufsuuid->fsu_len, &fsuuid.fsu_len,
1159 					sizeof(fsuuid.fsu_len)))
1160 			return -EFAULT;
1161 		return 0;
1162 	}
1163 
1164 	if (fsuuid.fsu_len < UUID_SIZE || fsuuid.fsu_flags != 0)
1165 		return -EINVAL;
1166 
1167 	lock_buffer(sbi->s_sbh);
1168 	memcpy(uuid, sbi->s_es->s_uuid, UUID_SIZE);
1169 	unlock_buffer(sbi->s_sbh);
1170 
1171 	fsuuid.fsu_len = UUID_SIZE;
1172 	if (copy_to_user(ufsuuid, &fsuuid, sizeof(fsuuid)) ||
1173 	    copy_to_user(&ufsuuid->fsu_uuid[0], uuid, UUID_SIZE))
1174 		return -EFAULT;
1175 	return 0;
1176 }
1177 
1178 static int ext4_ioctl_setuuid(struct file *filp,
1179 			const struct fsuuid __user *ufsuuid)
1180 {
1181 	int ret = 0;
1182 	struct super_block *sb = file_inode(filp)->i_sb;
1183 	struct fsuuid fsuuid;
1184 	__u8 uuid[UUID_SIZE];
1185 
1186 	if (!capable(CAP_SYS_ADMIN))
1187 		return -EPERM;
1188 
1189 	/*
1190 	 * If any checksums (group descriptors or metadata) are being used
1191 	 * then the checksum seed feature is required to change the UUID.
1192 	 */
1193 	if (((ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb))
1194 			&& !ext4_has_feature_csum_seed(sb))
1195 		|| ext4_has_feature_stable_inodes(sb))
1196 		return -EOPNOTSUPP;
1197 
1198 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1199 		return -EFAULT;
1200 
1201 	if (fsuuid.fsu_len != UUID_SIZE || fsuuid.fsu_flags != 0)
1202 		return -EINVAL;
1203 
1204 	if (copy_from_user(uuid, &ufsuuid->fsu_uuid[0], UUID_SIZE))
1205 		return -EFAULT;
1206 
1207 	ret = mnt_want_write_file(filp);
1208 	if (ret)
1209 		return ret;
1210 
1211 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setuuid, &uuid);
1212 	mnt_drop_write_file(filp);
1213 
1214 	return ret;
1215 }
1216 
1217 static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1218 {
1219 	struct inode *inode = file_inode(filp);
1220 	struct super_block *sb = inode->i_sb;
1221 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
1222 
1223 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
1224 
1225 	switch (cmd) {
1226 	case FS_IOC_GETFSMAP:
1227 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
1228 	case EXT4_IOC_GETVERSION:
1229 	case EXT4_IOC_GETVERSION_OLD:
1230 		return put_user(inode->i_generation, (int __user *) arg);
1231 	case EXT4_IOC_SETVERSION:
1232 	case EXT4_IOC_SETVERSION_OLD: {
1233 		handle_t *handle;
1234 		struct ext4_iloc iloc;
1235 		__u32 generation;
1236 		int err;
1237 
1238 		if (!inode_owner_or_capable(idmap, inode))
1239 			return -EPERM;
1240 
1241 		if (ext4_has_metadata_csum(inode->i_sb)) {
1242 			ext4_warning(sb, "Setting inode version is not "
1243 				     "supported with metadata_csum enabled.");
1244 			return -ENOTTY;
1245 		}
1246 
1247 		err = mnt_want_write_file(filp);
1248 		if (err)
1249 			return err;
1250 		if (get_user(generation, (int __user *) arg)) {
1251 			err = -EFAULT;
1252 			goto setversion_out;
1253 		}
1254 
1255 		inode_lock(inode);
1256 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
1257 		if (IS_ERR(handle)) {
1258 			err = PTR_ERR(handle);
1259 			goto unlock_out;
1260 		}
1261 		err = ext4_reserve_inode_write(handle, inode, &iloc);
1262 		if (err == 0) {
1263 			inode->i_ctime = current_time(inode);
1264 			inode_inc_iversion(inode);
1265 			inode->i_generation = generation;
1266 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
1267 		}
1268 		ext4_journal_stop(handle);
1269 
1270 unlock_out:
1271 		inode_unlock(inode);
1272 setversion_out:
1273 		mnt_drop_write_file(filp);
1274 		return err;
1275 	}
1276 	case EXT4_IOC_GROUP_EXTEND: {
1277 		ext4_fsblk_t n_blocks_count;
1278 		int err, err2=0;
1279 
1280 		err = ext4_resize_begin(sb);
1281 		if (err)
1282 			return err;
1283 
1284 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
1285 			err = -EFAULT;
1286 			goto group_extend_out;
1287 		}
1288 
1289 		if (ext4_has_feature_bigalloc(sb)) {
1290 			ext4_msg(sb, KERN_ERR,
1291 				 "Online resizing not supported with bigalloc");
1292 			err = -EOPNOTSUPP;
1293 			goto group_extend_out;
1294 		}
1295 
1296 		err = mnt_want_write_file(filp);
1297 		if (err)
1298 			goto group_extend_out;
1299 
1300 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
1301 		if (EXT4_SB(sb)->s_journal) {
1302 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1303 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1304 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1305 		}
1306 		if (err == 0)
1307 			err = err2;
1308 		mnt_drop_write_file(filp);
1309 group_extend_out:
1310 		err2 = ext4_resize_end(sb, false);
1311 		if (err == 0)
1312 			err = err2;
1313 		return err;
1314 	}
1315 
1316 	case EXT4_IOC_MOVE_EXT: {
1317 		struct move_extent me;
1318 		struct fd donor;
1319 		int err;
1320 
1321 		if (!(filp->f_mode & FMODE_READ) ||
1322 		    !(filp->f_mode & FMODE_WRITE))
1323 			return -EBADF;
1324 
1325 		if (copy_from_user(&me,
1326 			(struct move_extent __user *)arg, sizeof(me)))
1327 			return -EFAULT;
1328 		me.moved_len = 0;
1329 
1330 		donor = fdget(me.donor_fd);
1331 		if (!donor.file)
1332 			return -EBADF;
1333 
1334 		if (!(donor.file->f_mode & FMODE_WRITE)) {
1335 			err = -EBADF;
1336 			goto mext_out;
1337 		}
1338 
1339 		if (ext4_has_feature_bigalloc(sb)) {
1340 			ext4_msg(sb, KERN_ERR,
1341 				 "Online defrag not supported with bigalloc");
1342 			err = -EOPNOTSUPP;
1343 			goto mext_out;
1344 		} else if (IS_DAX(inode)) {
1345 			ext4_msg(sb, KERN_ERR,
1346 				 "Online defrag not supported with DAX");
1347 			err = -EOPNOTSUPP;
1348 			goto mext_out;
1349 		}
1350 
1351 		err = mnt_want_write_file(filp);
1352 		if (err)
1353 			goto mext_out;
1354 
1355 		err = ext4_move_extents(filp, donor.file, me.orig_start,
1356 					me.donor_start, me.len, &me.moved_len);
1357 		mnt_drop_write_file(filp);
1358 
1359 		if (copy_to_user((struct move_extent __user *)arg,
1360 				 &me, sizeof(me)))
1361 			err = -EFAULT;
1362 mext_out:
1363 		fdput(donor);
1364 		return err;
1365 	}
1366 
1367 	case EXT4_IOC_GROUP_ADD: {
1368 		struct ext4_new_group_data input;
1369 
1370 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
1371 				sizeof(input)))
1372 			return -EFAULT;
1373 
1374 		return ext4_ioctl_group_add(filp, &input);
1375 	}
1376 
1377 	case EXT4_IOC_MIGRATE:
1378 	{
1379 		int err;
1380 		if (!inode_owner_or_capable(idmap, inode))
1381 			return -EACCES;
1382 
1383 		err = mnt_want_write_file(filp);
1384 		if (err)
1385 			return err;
1386 		/*
1387 		 * inode_mutex prevent write and truncate on the file.
1388 		 * Read still goes through. We take i_data_sem in
1389 		 * ext4_ext_swap_inode_data before we switch the
1390 		 * inode format to prevent read.
1391 		 */
1392 		inode_lock((inode));
1393 		err = ext4_ext_migrate(inode);
1394 		inode_unlock((inode));
1395 		mnt_drop_write_file(filp);
1396 		return err;
1397 	}
1398 
1399 	case EXT4_IOC_ALLOC_DA_BLKS:
1400 	{
1401 		int err;
1402 		if (!inode_owner_or_capable(idmap, inode))
1403 			return -EACCES;
1404 
1405 		err = mnt_want_write_file(filp);
1406 		if (err)
1407 			return err;
1408 		err = ext4_alloc_da_blocks(inode);
1409 		mnt_drop_write_file(filp);
1410 		return err;
1411 	}
1412 
1413 	case EXT4_IOC_SWAP_BOOT:
1414 	{
1415 		int err;
1416 		if (!(filp->f_mode & FMODE_WRITE))
1417 			return -EBADF;
1418 		err = mnt_want_write_file(filp);
1419 		if (err)
1420 			return err;
1421 		err = swap_inode_boot_loader(sb, idmap, inode);
1422 		mnt_drop_write_file(filp);
1423 		return err;
1424 	}
1425 
1426 	case EXT4_IOC_RESIZE_FS: {
1427 		ext4_fsblk_t n_blocks_count;
1428 		int err = 0, err2 = 0;
1429 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
1430 
1431 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1432 				   sizeof(__u64))) {
1433 			return -EFAULT;
1434 		}
1435 
1436 		err = ext4_resize_begin(sb);
1437 		if (err)
1438 			return err;
1439 
1440 		err = mnt_want_write_file(filp);
1441 		if (err)
1442 			goto resizefs_out;
1443 
1444 		err = ext4_resize_fs(sb, n_blocks_count);
1445 		if (EXT4_SB(sb)->s_journal) {
1446 			ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE, NULL);
1447 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1448 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1449 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1450 		}
1451 		if (err == 0)
1452 			err = err2;
1453 		mnt_drop_write_file(filp);
1454 		if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1455 		    ext4_has_group_desc_csum(sb) &&
1456 		    test_opt(sb, INIT_INODE_TABLE))
1457 			err = ext4_register_li_request(sb, o_group);
1458 
1459 resizefs_out:
1460 		err2 = ext4_resize_end(sb, true);
1461 		if (err == 0)
1462 			err = err2;
1463 		return err;
1464 	}
1465 
1466 	case FITRIM:
1467 	{
1468 		struct fstrim_range range;
1469 		int ret = 0;
1470 
1471 		if (!capable(CAP_SYS_ADMIN))
1472 			return -EPERM;
1473 
1474 		if (!bdev_max_discard_sectors(sb->s_bdev))
1475 			return -EOPNOTSUPP;
1476 
1477 		/*
1478 		 * We haven't replayed the journal, so we cannot use our
1479 		 * block-bitmap-guided storage zapping commands.
1480 		 */
1481 		if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1482 			return -EROFS;
1483 
1484 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1485 		    sizeof(range)))
1486 			return -EFAULT;
1487 
1488 		ret = ext4_trim_fs(sb, &range);
1489 		if (ret < 0)
1490 			return ret;
1491 
1492 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
1493 		    sizeof(range)))
1494 			return -EFAULT;
1495 
1496 		return 0;
1497 	}
1498 	case EXT4_IOC_PRECACHE_EXTENTS:
1499 		return ext4_ext_precache(inode);
1500 
1501 	case FS_IOC_SET_ENCRYPTION_POLICY:
1502 		if (!ext4_has_feature_encrypt(sb))
1503 			return -EOPNOTSUPP;
1504 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1505 
1506 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1507 		return ext4_ioctl_get_encryption_pwsalt(filp, (void __user *)arg);
1508 
1509 	case FS_IOC_GET_ENCRYPTION_POLICY:
1510 		if (!ext4_has_feature_encrypt(sb))
1511 			return -EOPNOTSUPP;
1512 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1513 
1514 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1515 		if (!ext4_has_feature_encrypt(sb))
1516 			return -EOPNOTSUPP;
1517 		return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
1518 
1519 	case FS_IOC_ADD_ENCRYPTION_KEY:
1520 		if (!ext4_has_feature_encrypt(sb))
1521 			return -EOPNOTSUPP;
1522 		return fscrypt_ioctl_add_key(filp, (void __user *)arg);
1523 
1524 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1525 		if (!ext4_has_feature_encrypt(sb))
1526 			return -EOPNOTSUPP;
1527 		return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
1528 
1529 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1530 		if (!ext4_has_feature_encrypt(sb))
1531 			return -EOPNOTSUPP;
1532 		return fscrypt_ioctl_remove_key_all_users(filp,
1533 							  (void __user *)arg);
1534 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1535 		if (!ext4_has_feature_encrypt(sb))
1536 			return -EOPNOTSUPP;
1537 		return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
1538 
1539 	case FS_IOC_GET_ENCRYPTION_NONCE:
1540 		if (!ext4_has_feature_encrypt(sb))
1541 			return -EOPNOTSUPP;
1542 		return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
1543 
1544 	case EXT4_IOC_CLEAR_ES_CACHE:
1545 	{
1546 		if (!inode_owner_or_capable(idmap, inode))
1547 			return -EACCES;
1548 		ext4_clear_inode_es(inode);
1549 		return 0;
1550 	}
1551 
1552 	case EXT4_IOC_GETSTATE:
1553 	{
1554 		__u32	state = 0;
1555 
1556 		if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED))
1557 			state |= EXT4_STATE_FLAG_EXT_PRECACHED;
1558 		if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
1559 			state |= EXT4_STATE_FLAG_NEW;
1560 		if (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
1561 			state |= EXT4_STATE_FLAG_NEWENTRY;
1562 		if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE))
1563 			state |= EXT4_STATE_FLAG_DA_ALLOC_CLOSE;
1564 
1565 		return put_user(state, (__u32 __user *) arg);
1566 	}
1567 
1568 	case EXT4_IOC_GET_ES_CACHE:
1569 		return ext4_ioctl_get_es_cache(filp, arg);
1570 
1571 	case EXT4_IOC_SHUTDOWN:
1572 		return ext4_shutdown(sb, arg);
1573 
1574 	case FS_IOC_ENABLE_VERITY:
1575 		if (!ext4_has_feature_verity(sb))
1576 			return -EOPNOTSUPP;
1577 		return fsverity_ioctl_enable(filp, (const void __user *)arg);
1578 
1579 	case FS_IOC_MEASURE_VERITY:
1580 		if (!ext4_has_feature_verity(sb))
1581 			return -EOPNOTSUPP;
1582 		return fsverity_ioctl_measure(filp, (void __user *)arg);
1583 
1584 	case FS_IOC_READ_VERITY_METADATA:
1585 		if (!ext4_has_feature_verity(sb))
1586 			return -EOPNOTSUPP;
1587 		return fsverity_ioctl_read_metadata(filp,
1588 						    (const void __user *)arg);
1589 
1590 	case EXT4_IOC_CHECKPOINT:
1591 		return ext4_ioctl_checkpoint(filp, arg);
1592 
1593 	case FS_IOC_GETFSLABEL:
1594 		return ext4_ioctl_getlabel(EXT4_SB(sb), (void __user *)arg);
1595 
1596 	case FS_IOC_SETFSLABEL:
1597 		return ext4_ioctl_setlabel(filp,
1598 					   (const void __user *)arg);
1599 
1600 	case EXT4_IOC_GETFSUUID:
1601 		return ext4_ioctl_getuuid(EXT4_SB(sb), (void __user *)arg);
1602 	case EXT4_IOC_SETFSUUID:
1603 		return ext4_ioctl_setuuid(filp, (const void __user *)arg);
1604 	default:
1605 		return -ENOTTY;
1606 	}
1607 }
1608 
1609 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1610 {
1611 	return __ext4_ioctl(filp, cmd, arg);
1612 }
1613 
1614 #ifdef CONFIG_COMPAT
1615 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1616 {
1617 	/* These are just misnamed, they actually get/put from/to user an int */
1618 	switch (cmd) {
1619 	case EXT4_IOC32_GETVERSION:
1620 		cmd = EXT4_IOC_GETVERSION;
1621 		break;
1622 	case EXT4_IOC32_SETVERSION:
1623 		cmd = EXT4_IOC_SETVERSION;
1624 		break;
1625 	case EXT4_IOC32_GROUP_EXTEND:
1626 		cmd = EXT4_IOC_GROUP_EXTEND;
1627 		break;
1628 	case EXT4_IOC32_GETVERSION_OLD:
1629 		cmd = EXT4_IOC_GETVERSION_OLD;
1630 		break;
1631 	case EXT4_IOC32_SETVERSION_OLD:
1632 		cmd = EXT4_IOC_SETVERSION_OLD;
1633 		break;
1634 	case EXT4_IOC32_GETRSVSZ:
1635 		cmd = EXT4_IOC_GETRSVSZ;
1636 		break;
1637 	case EXT4_IOC32_SETRSVSZ:
1638 		cmd = EXT4_IOC_SETRSVSZ;
1639 		break;
1640 	case EXT4_IOC32_GROUP_ADD: {
1641 		struct compat_ext4_new_group_input __user *uinput;
1642 		struct ext4_new_group_data input;
1643 		int err;
1644 
1645 		uinput = compat_ptr(arg);
1646 		err = get_user(input.group, &uinput->group);
1647 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1648 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1649 		err |= get_user(input.inode_table, &uinput->inode_table);
1650 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1651 		err |= get_user(input.reserved_blocks,
1652 				&uinput->reserved_blocks);
1653 		if (err)
1654 			return -EFAULT;
1655 		return ext4_ioctl_group_add(file, &input);
1656 	}
1657 	case EXT4_IOC_MOVE_EXT:
1658 	case EXT4_IOC_RESIZE_FS:
1659 	case FITRIM:
1660 	case EXT4_IOC_PRECACHE_EXTENTS:
1661 	case FS_IOC_SET_ENCRYPTION_POLICY:
1662 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1663 	case FS_IOC_GET_ENCRYPTION_POLICY:
1664 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1665 	case FS_IOC_ADD_ENCRYPTION_KEY:
1666 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1667 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1668 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1669 	case FS_IOC_GET_ENCRYPTION_NONCE:
1670 	case EXT4_IOC_SHUTDOWN:
1671 	case FS_IOC_GETFSMAP:
1672 	case FS_IOC_ENABLE_VERITY:
1673 	case FS_IOC_MEASURE_VERITY:
1674 	case FS_IOC_READ_VERITY_METADATA:
1675 	case EXT4_IOC_CLEAR_ES_CACHE:
1676 	case EXT4_IOC_GETSTATE:
1677 	case EXT4_IOC_GET_ES_CACHE:
1678 	case EXT4_IOC_CHECKPOINT:
1679 	case FS_IOC_GETFSLABEL:
1680 	case FS_IOC_SETFSLABEL:
1681 	case EXT4_IOC_GETFSUUID:
1682 	case EXT4_IOC_SETFSUUID:
1683 		break;
1684 	default:
1685 		return -ENOIOCTLCMD;
1686 	}
1687 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1688 }
1689 #endif
1690 
1691 static void set_overhead(struct ext4_super_block *es, const void *arg)
1692 {
1693 	es->s_overhead_clusters = cpu_to_le32(*((unsigned long *) arg));
1694 }
1695 
1696 int ext4_update_overhead(struct super_block *sb, bool force)
1697 {
1698 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1699 
1700 	if (sb_rdonly(sb))
1701 		return 0;
1702 	if (!force &&
1703 	    (sbi->s_overhead == 0 ||
1704 	     sbi->s_overhead == le32_to_cpu(sbi->s_es->s_overhead_clusters)))
1705 		return 0;
1706 	return ext4_update_superblocks_fn(sb, set_overhead, &sbi->s_overhead);
1707 }
1708