xref: /openbmc/linux/fs/ext4/ioctl.c (revision 913447d0)
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 (err)
159 		goto out_bh;
160 
161 	if (handle) {
162 		err = ext4_handle_dirty_metadata(handle, NULL, bh);
163 		if (err)
164 			goto out_bh;
165 	} else {
166 		BUFFER_TRACE(bh, "marking dirty");
167 		mark_buffer_dirty(bh);
168 	}
169 	err = sync_dirty_buffer(bh);
170 
171 out_bh:
172 	brelse(bh);
173 	ext4_std_error(sb, err);
174 	return (err) ? err : 1;
175 }
176 
177 /*
178  * Update primary and backup superblocks using the provided function
179  * func and argument arg.
180  *
181  * Only the primary superblock and at most two backup superblock
182  * modifications are journalled; the rest is modified without journal.
183  * This is safe because e2fsck will re-write them if there is a problem,
184  * and we're very unlikely to ever need more than two backups.
185  */
186 static
187 int ext4_update_superblocks_fn(struct super_block *sb,
188 			       ext4_update_sb_callback func,
189 			       const void *arg)
190 {
191 	handle_t *handle;
192 	ext4_group_t ngroups;
193 	unsigned int three = 1;
194 	unsigned int five = 5;
195 	unsigned int seven = 7;
196 	int err = 0, ret, i;
197 	ext4_group_t grp, primary_grp;
198 	struct ext4_sb_info *sbi = EXT4_SB(sb);
199 
200 	/*
201 	 * We can't update superblocks while the online resize is running
202 	 */
203 	if (test_and_set_bit_lock(EXT4_FLAGS_RESIZING,
204 				  &sbi->s_ext4_flags)) {
205 		ext4_msg(sb, KERN_ERR, "Can't modify superblock while"
206 			 "performing online resize");
207 		return -EBUSY;
208 	}
209 
210 	/*
211 	 * We're only going to update primary superblock and two
212 	 * backup superblocks in this transaction.
213 	 */
214 	handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 3);
215 	if (IS_ERR(handle)) {
216 		err = PTR_ERR(handle);
217 		goto out;
218 	}
219 
220 	/* Update primary superblock */
221 	err = ext4_update_primary_sb(sb, handle, func, arg);
222 	if (err) {
223 		ext4_msg(sb, KERN_ERR, "Failed to update primary "
224 			 "superblock");
225 		goto out_journal;
226 	}
227 
228 	primary_grp = ext4_get_group_number(sb, sbi->s_sbh->b_blocknr);
229 	ngroups = ext4_get_groups_count(sb);
230 
231 	/*
232 	 * Update backup superblocks. We have to start from group 0
233 	 * because it might not be where the primary superblock is
234 	 * if the fs is mounted with -o sb=<backup_sb_block>
235 	 */
236 	i = 0;
237 	grp = 0;
238 	while (grp < ngroups) {
239 		/* Skip primary superblock */
240 		if (grp == primary_grp)
241 			goto next_grp;
242 
243 		ret = ext4_update_backup_sb(sb, handle, grp, func, arg);
244 		if (ret < 0) {
245 			/* Ignore bad checksum; try to update next sb */
246 			if (ret == -EFSBADCRC)
247 				goto next_grp;
248 			err = ret;
249 			goto out_journal;
250 		}
251 
252 		i += ret;
253 		if (handle && i > 1) {
254 			/*
255 			 * We're only journalling primary superblock and
256 			 * two backup superblocks; the rest is not
257 			 * journalled.
258 			 */
259 			err = ext4_journal_stop(handle);
260 			if (err)
261 				goto out;
262 			handle = NULL;
263 		}
264 next_grp:
265 		grp = ext4_list_backups(sb, &three, &five, &seven);
266 	}
267 
268 out_journal:
269 	if (handle) {
270 		ret = ext4_journal_stop(handle);
271 		if (ret && !err)
272 			err = ret;
273 	}
274 out:
275 	clear_bit_unlock(EXT4_FLAGS_RESIZING, &sbi->s_ext4_flags);
276 	smp_mb__after_atomic();
277 	return err ? err : 0;
278 }
279 
280 /*
281  * Swap memory between @a and @b for @len bytes.
282  *
283  * @a:          pointer to first memory area
284  * @b:          pointer to second memory area
285  * @len:        number of bytes to swap
286  *
287  */
288 static void memswap(void *a, void *b, size_t len)
289 {
290 	unsigned char *ap, *bp;
291 
292 	ap = (unsigned char *)a;
293 	bp = (unsigned char *)b;
294 	while (len-- > 0) {
295 		swap(*ap, *bp);
296 		ap++;
297 		bp++;
298 	}
299 }
300 
301 /*
302  * Swap i_data and associated attributes between @inode1 and @inode2.
303  * This function is used for the primary swap between inode1 and inode2
304  * and also to revert this primary swap in case of errors.
305  *
306  * Therefore you have to make sure, that calling this method twice
307  * will revert all changes.
308  *
309  * @inode1:     pointer to first inode
310  * @inode2:     pointer to second inode
311  */
312 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
313 {
314 	loff_t isize;
315 	struct ext4_inode_info *ei1;
316 	struct ext4_inode_info *ei2;
317 	unsigned long tmp;
318 
319 	ei1 = EXT4_I(inode1);
320 	ei2 = EXT4_I(inode2);
321 
322 	swap(inode1->i_version, inode2->i_version);
323 	swap(inode1->i_atime, inode2->i_atime);
324 	swap(inode1->i_mtime, inode2->i_mtime);
325 
326 	memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
327 	tmp = ei1->i_flags & EXT4_FL_SHOULD_SWAP;
328 	ei1->i_flags = (ei2->i_flags & EXT4_FL_SHOULD_SWAP) |
329 		(ei1->i_flags & ~EXT4_FL_SHOULD_SWAP);
330 	ei2->i_flags = tmp | (ei2->i_flags & ~EXT4_FL_SHOULD_SWAP);
331 	swap(ei1->i_disksize, ei2->i_disksize);
332 	ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
333 	ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
334 
335 	isize = i_size_read(inode1);
336 	i_size_write(inode1, i_size_read(inode2));
337 	i_size_write(inode2, isize);
338 }
339 
340 void ext4_reset_inode_seed(struct inode *inode)
341 {
342 	struct ext4_inode_info *ei = EXT4_I(inode);
343 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
344 	__le32 inum = cpu_to_le32(inode->i_ino);
345 	__le32 gen = cpu_to_le32(inode->i_generation);
346 	__u32 csum;
347 
348 	if (!ext4_has_metadata_csum(inode->i_sb))
349 		return;
350 
351 	csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum, sizeof(inum));
352 	ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen, sizeof(gen));
353 }
354 
355 /*
356  * Swap the information from the given @inode and the inode
357  * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
358  * important fields of the inodes.
359  *
360  * @sb:         the super block of the filesystem
361  * @mnt_userns:	user namespace of the mount the inode was found from
362  * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
363  *
364  */
365 static long swap_inode_boot_loader(struct super_block *sb,
366 				struct user_namespace *mnt_userns,
367 				struct inode *inode)
368 {
369 	handle_t *handle;
370 	int err;
371 	struct inode *inode_bl;
372 	struct ext4_inode_info *ei_bl;
373 	qsize_t size, size_bl, diff;
374 	blkcnt_t blocks;
375 	unsigned short bytes;
376 
377 	inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO, EXT4_IGET_SPECIAL);
378 	if (IS_ERR(inode_bl))
379 		return PTR_ERR(inode_bl);
380 	ei_bl = EXT4_I(inode_bl);
381 
382 	/* Protect orig inodes against a truncate and make sure,
383 	 * that only 1 swap_inode_boot_loader is running. */
384 	lock_two_nondirectories(inode, inode_bl);
385 
386 	if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode) ||
387 	    IS_SWAPFILE(inode) || IS_ENCRYPTED(inode) ||
388 	    (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL) ||
389 	    ext4_has_inline_data(inode)) {
390 		err = -EINVAL;
391 		goto journal_err_out;
392 	}
393 
394 	if (IS_RDONLY(inode) || IS_APPEND(inode) || IS_IMMUTABLE(inode) ||
395 	    !inode_owner_or_capable(mnt_userns, inode) ||
396 	    !capable(CAP_SYS_ADMIN)) {
397 		err = -EPERM;
398 		goto journal_err_out;
399 	}
400 
401 	filemap_invalidate_lock(inode->i_mapping);
402 	err = filemap_write_and_wait(inode->i_mapping);
403 	if (err)
404 		goto err_out;
405 
406 	err = filemap_write_and_wait(inode_bl->i_mapping);
407 	if (err)
408 		goto err_out;
409 
410 	/* Wait for all existing dio workers */
411 	inode_dio_wait(inode);
412 	inode_dio_wait(inode_bl);
413 
414 	truncate_inode_pages(&inode->i_data, 0);
415 	truncate_inode_pages(&inode_bl->i_data, 0);
416 
417 	handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
418 	if (IS_ERR(handle)) {
419 		err = -EINVAL;
420 		goto err_out;
421 	}
422 	ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_SWAP_BOOT, handle);
423 
424 	/* Protect extent tree against block allocations via delalloc */
425 	ext4_double_down_write_data_sem(inode, inode_bl);
426 
427 	if (inode_bl->i_nlink == 0) {
428 		/* this inode has never been used as a BOOT_LOADER */
429 		set_nlink(inode_bl, 1);
430 		i_uid_write(inode_bl, 0);
431 		i_gid_write(inode_bl, 0);
432 		inode_bl->i_flags = 0;
433 		ei_bl->i_flags = 0;
434 		inode_set_iversion(inode_bl, 1);
435 		i_size_write(inode_bl, 0);
436 		inode_bl->i_mode = S_IFREG;
437 		if (ext4_has_feature_extents(sb)) {
438 			ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
439 			ext4_ext_tree_init(handle, inode_bl);
440 		} else
441 			memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
442 	}
443 
444 	err = dquot_initialize(inode);
445 	if (err)
446 		goto err_out1;
447 
448 	size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
449 	size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
450 	diff = size - size_bl;
451 	swap_inode_data(inode, inode_bl);
452 
453 	inode->i_ctime = inode_bl->i_ctime = current_time(inode);
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->i_ctime = current_time(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 = ext4_get_inode_loc(inode, &iloc);
735 	if (err)
736 		return err;
737 
738 	raw_inode = ext4_raw_inode(&iloc);
739 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
740 		err = ext4_expand_extra_isize(inode,
741 					      EXT4_SB(sb)->s_want_extra_isize,
742 					      &iloc);
743 		if (err)
744 			return err;
745 	} else {
746 		brelse(iloc.bh);
747 	}
748 
749 	err = dquot_initialize(inode);
750 	if (err)
751 		return err;
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->i_ctime = current_time(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 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
798 {
799 	struct ext4_sb_info *sbi = EXT4_SB(sb);
800 	__u32 flags;
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 		freeze_bdev(sb->s_bdev);
820 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
821 		thaw_bdev(sb->s_bdev);
822 		break;
823 	case EXT4_GOING_FLAGS_LOGFLUSH:
824 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
825 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
826 			(void) ext4_force_commit(sb);
827 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
828 		}
829 		break;
830 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
831 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
832 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
833 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
834 		break;
835 	default:
836 		return -EINVAL;
837 	}
838 	clear_opt(sb, DISCARD);
839 	return 0;
840 }
841 
842 struct getfsmap_info {
843 	struct super_block	*gi_sb;
844 	struct fsmap_head __user *gi_data;
845 	unsigned int		gi_idx;
846 	__u32			gi_last_flags;
847 };
848 
849 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
850 {
851 	struct getfsmap_info *info = priv;
852 	struct fsmap fm;
853 
854 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
855 
856 	info->gi_last_flags = xfm->fmr_flags;
857 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
858 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
859 			sizeof(struct fsmap)))
860 		return -EFAULT;
861 
862 	return 0;
863 }
864 
865 static int ext4_ioc_getfsmap(struct super_block *sb,
866 			     struct fsmap_head __user *arg)
867 {
868 	struct getfsmap_info info = { NULL };
869 	struct ext4_fsmap_head xhead = {0};
870 	struct fsmap_head head;
871 	bool aborted = false;
872 	int error;
873 
874 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
875 		return -EFAULT;
876 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
877 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
878 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
879 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
880 		       sizeof(head.fmh_keys[1].fmr_reserved)))
881 		return -EINVAL;
882 	/*
883 	 * ext4 doesn't report file extents at all, so the only valid
884 	 * file offsets are the magic ones (all zeroes or all ones).
885 	 */
886 	if (head.fmh_keys[0].fmr_offset ||
887 	    (head.fmh_keys[1].fmr_offset != 0 &&
888 	     head.fmh_keys[1].fmr_offset != -1ULL))
889 		return -EINVAL;
890 
891 	xhead.fmh_iflags = head.fmh_iflags;
892 	xhead.fmh_count = head.fmh_count;
893 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
894 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
895 
896 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
897 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
898 
899 	info.gi_sb = sb;
900 	info.gi_data = arg;
901 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
902 	if (error == EXT4_QUERY_RANGE_ABORT)
903 		aborted = true;
904 	else if (error)
905 		return error;
906 
907 	/* If we didn't abort, set the "last" flag in the last fmx */
908 	if (!aborted && info.gi_idx) {
909 		info.gi_last_flags |= FMR_OF_LAST;
910 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
911 				 &info.gi_last_flags,
912 				 sizeof(info.gi_last_flags)))
913 			return -EFAULT;
914 	}
915 
916 	/* copy back header */
917 	head.fmh_entries = xhead.fmh_entries;
918 	head.fmh_oflags = xhead.fmh_oflags;
919 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
920 		return -EFAULT;
921 
922 	return 0;
923 }
924 
925 static long ext4_ioctl_group_add(struct file *file,
926 				 struct ext4_new_group_data *input)
927 {
928 	struct super_block *sb = file_inode(file)->i_sb;
929 	int err, err2=0;
930 
931 	err = ext4_resize_begin(sb);
932 	if (err)
933 		return err;
934 
935 	if (ext4_has_feature_bigalloc(sb)) {
936 		ext4_msg(sb, KERN_ERR,
937 			 "Online resizing not supported with bigalloc");
938 		err = -EOPNOTSUPP;
939 		goto group_add_out;
940 	}
941 
942 	err = mnt_want_write_file(file);
943 	if (err)
944 		goto group_add_out;
945 
946 	err = ext4_group_add(sb, input);
947 	if (EXT4_SB(sb)->s_journal) {
948 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
949 		err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
950 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
951 	}
952 	if (err == 0)
953 		err = err2;
954 	mnt_drop_write_file(file);
955 	if (!err && ext4_has_group_desc_csum(sb) &&
956 	    test_opt(sb, INIT_INODE_TABLE))
957 		err = ext4_register_li_request(sb, input->group);
958 group_add_out:
959 	err2 = ext4_resize_end(sb, false);
960 	if (err == 0)
961 		err = err2;
962 	return err;
963 }
964 
965 int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa)
966 {
967 	struct inode *inode = d_inode(dentry);
968 	struct ext4_inode_info *ei = EXT4_I(inode);
969 	u32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
970 
971 	if (S_ISREG(inode->i_mode))
972 		flags &= ~FS_PROJINHERIT_FL;
973 
974 	fileattr_fill_flags(fa, flags);
975 	if (ext4_has_feature_project(inode->i_sb))
976 		fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
977 
978 	return 0;
979 }
980 
981 int ext4_fileattr_set(struct user_namespace *mnt_userns,
982 		      struct dentry *dentry, struct fileattr *fa)
983 {
984 	struct inode *inode = d_inode(dentry);
985 	u32 flags = fa->flags;
986 	int err = -EOPNOTSUPP;
987 
988 	if (flags & ~EXT4_FL_USER_VISIBLE)
989 		goto out;
990 
991 	/*
992 	 * chattr(1) grabs flags via GETFLAGS, modifies the result and
993 	 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
994 	 * more restrictive than just silently masking off visible but
995 	 * not settable flags as we always did.
996 	 */
997 	flags &= EXT4_FL_USER_MODIFIABLE;
998 	if (ext4_mask_flags(inode->i_mode, flags) != flags)
999 		goto out;
1000 	err = ext4_ioctl_check_immutable(inode, fa->fsx_projid, flags);
1001 	if (err)
1002 		goto out;
1003 	err = ext4_ioctl_setflags(inode, flags);
1004 	if (err)
1005 		goto out;
1006 	err = ext4_ioctl_setproject(inode, fa->fsx_projid);
1007 out:
1008 	return err;
1009 }
1010 
1011 /* So that the fiemap access checks can't overflow on 32 bit machines. */
1012 #define FIEMAP_MAX_EXTENTS	(UINT_MAX / sizeof(struct fiemap_extent))
1013 
1014 static int ext4_ioctl_get_es_cache(struct file *filp, unsigned long arg)
1015 {
1016 	struct fiemap fiemap;
1017 	struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
1018 	struct fiemap_extent_info fieinfo = { 0, };
1019 	struct inode *inode = file_inode(filp);
1020 	int error;
1021 
1022 	if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
1023 		return -EFAULT;
1024 
1025 	if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
1026 		return -EINVAL;
1027 
1028 	fieinfo.fi_flags = fiemap.fm_flags;
1029 	fieinfo.fi_extents_max = fiemap.fm_extent_count;
1030 	fieinfo.fi_extents_start = ufiemap->fm_extents;
1031 
1032 	error = ext4_get_es_cache(inode, &fieinfo, fiemap.fm_start,
1033 			fiemap.fm_length);
1034 	fiemap.fm_flags = fieinfo.fi_flags;
1035 	fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
1036 	if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
1037 		error = -EFAULT;
1038 
1039 	return error;
1040 }
1041 
1042 static int ext4_ioctl_checkpoint(struct file *filp, unsigned long arg)
1043 {
1044 	int err = 0;
1045 	__u32 flags = 0;
1046 	unsigned int flush_flags = 0;
1047 	struct super_block *sb = file_inode(filp)->i_sb;
1048 
1049 	if (copy_from_user(&flags, (__u32 __user *)arg,
1050 				sizeof(__u32)))
1051 		return -EFAULT;
1052 
1053 	if (!capable(CAP_SYS_ADMIN))
1054 		return -EPERM;
1055 
1056 	/* check for invalid bits set */
1057 	if ((flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID) ||
1058 				((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1059 				(flags & JBD2_JOURNAL_FLUSH_ZEROOUT)))
1060 		return -EINVAL;
1061 
1062 	if (!EXT4_SB(sb)->s_journal)
1063 		return -ENODEV;
1064 
1065 	if ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1066 	    !bdev_max_discard_sectors(EXT4_SB(sb)->s_journal->j_dev))
1067 		return -EOPNOTSUPP;
1068 
1069 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
1070 		return 0;
1071 
1072 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DISCARD)
1073 		flush_flags |= JBD2_JOURNAL_FLUSH_DISCARD;
1074 
1075 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT) {
1076 		flush_flags |= JBD2_JOURNAL_FLUSH_ZEROOUT;
1077 		pr_info_ratelimited("warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow");
1078 	}
1079 
1080 	jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1081 	err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, flush_flags);
1082 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1083 
1084 	return err;
1085 }
1086 
1087 static int ext4_ioctl_setlabel(struct file *filp, const char __user *user_label)
1088 {
1089 	size_t len;
1090 	int ret = 0;
1091 	char new_label[EXT4_LABEL_MAX + 1];
1092 	struct super_block *sb = file_inode(filp)->i_sb;
1093 
1094 	if (!capable(CAP_SYS_ADMIN))
1095 		return -EPERM;
1096 
1097 	/*
1098 	 * Copy the maximum length allowed for ext4 label with one more to
1099 	 * find the required terminating null byte in order to test the
1100 	 * label length. The on disk label doesn't need to be null terminated.
1101 	 */
1102 	if (copy_from_user(new_label, user_label, EXT4_LABEL_MAX + 1))
1103 		return -EFAULT;
1104 
1105 	len = strnlen(new_label, EXT4_LABEL_MAX + 1);
1106 	if (len > EXT4_LABEL_MAX)
1107 		return -EINVAL;
1108 
1109 	/*
1110 	 * Clear the buffer after the new label
1111 	 */
1112 	memset(new_label + len, 0, EXT4_LABEL_MAX - len);
1113 
1114 	ret = mnt_want_write_file(filp);
1115 	if (ret)
1116 		return ret;
1117 
1118 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setlabel, new_label);
1119 
1120 	mnt_drop_write_file(filp);
1121 	return ret;
1122 }
1123 
1124 static int ext4_ioctl_getlabel(struct ext4_sb_info *sbi, char __user *user_label)
1125 {
1126 	char label[EXT4_LABEL_MAX + 1];
1127 
1128 	/*
1129 	 * EXT4_LABEL_MAX must always be smaller than FSLABEL_MAX because
1130 	 * FSLABEL_MAX must include terminating null byte, while s_volume_name
1131 	 * does not have to.
1132 	 */
1133 	BUILD_BUG_ON(EXT4_LABEL_MAX >= FSLABEL_MAX);
1134 
1135 	memset(label, 0, sizeof(label));
1136 	lock_buffer(sbi->s_sbh);
1137 	strncpy(label, sbi->s_es->s_volume_name, EXT4_LABEL_MAX);
1138 	unlock_buffer(sbi->s_sbh);
1139 
1140 	if (copy_to_user(user_label, label, sizeof(label)))
1141 		return -EFAULT;
1142 	return 0;
1143 }
1144 
1145 static int ext4_ioctl_getuuid(struct ext4_sb_info *sbi,
1146 			struct fsuuid __user *ufsuuid)
1147 {
1148 	struct fsuuid fsuuid;
1149 	__u8 uuid[UUID_SIZE];
1150 
1151 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1152 		return -EFAULT;
1153 
1154 	if (fsuuid.fsu_len == 0) {
1155 		fsuuid.fsu_len = UUID_SIZE;
1156 		if (copy_to_user(ufsuuid, &fsuuid, sizeof(fsuuid.fsu_len)))
1157 			return -EFAULT;
1158 		return -EINVAL;
1159 	}
1160 
1161 	if (fsuuid.fsu_len != UUID_SIZE || fsuuid.fsu_flags != 0)
1162 		return -EINVAL;
1163 
1164 	lock_buffer(sbi->s_sbh);
1165 	memcpy(uuid, sbi->s_es->s_uuid, UUID_SIZE);
1166 	unlock_buffer(sbi->s_sbh);
1167 
1168 	if (copy_to_user(&ufsuuid->fsu_uuid[0], uuid, UUID_SIZE))
1169 		return -EFAULT;
1170 	return 0;
1171 }
1172 
1173 static int ext4_ioctl_setuuid(struct file *filp,
1174 			const struct fsuuid __user *ufsuuid)
1175 {
1176 	int ret = 0;
1177 	struct super_block *sb = file_inode(filp)->i_sb;
1178 	struct fsuuid fsuuid;
1179 	__u8 uuid[UUID_SIZE];
1180 
1181 	if (!capable(CAP_SYS_ADMIN))
1182 		return -EPERM;
1183 
1184 	/*
1185 	 * If any checksums (group descriptors or metadata) are being used
1186 	 * then the checksum seed feature is required to change the UUID.
1187 	 */
1188 	if (((ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb))
1189 			&& !ext4_has_feature_csum_seed(sb))
1190 		|| ext4_has_feature_stable_inodes(sb))
1191 		return -EOPNOTSUPP;
1192 
1193 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1194 		return -EFAULT;
1195 
1196 	if (fsuuid.fsu_len != UUID_SIZE || fsuuid.fsu_flags != 0)
1197 		return -EINVAL;
1198 
1199 	if (copy_from_user(uuid, &ufsuuid->fsu_uuid[0], UUID_SIZE))
1200 		return -EFAULT;
1201 
1202 	ret = mnt_want_write_file(filp);
1203 	if (ret)
1204 		return ret;
1205 
1206 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setuuid, &uuid);
1207 	mnt_drop_write_file(filp);
1208 
1209 	return ret;
1210 }
1211 
1212 static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1213 {
1214 	struct inode *inode = file_inode(filp);
1215 	struct super_block *sb = inode->i_sb;
1216 	struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
1217 
1218 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
1219 
1220 	switch (cmd) {
1221 	case FS_IOC_GETFSMAP:
1222 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
1223 	case EXT4_IOC_GETVERSION:
1224 	case EXT4_IOC_GETVERSION_OLD:
1225 		return put_user(inode->i_generation, (int __user *) arg);
1226 	case EXT4_IOC_SETVERSION:
1227 	case EXT4_IOC_SETVERSION_OLD: {
1228 		handle_t *handle;
1229 		struct ext4_iloc iloc;
1230 		__u32 generation;
1231 		int err;
1232 
1233 		if (!inode_owner_or_capable(mnt_userns, inode))
1234 			return -EPERM;
1235 
1236 		if (ext4_has_metadata_csum(inode->i_sb)) {
1237 			ext4_warning(sb, "Setting inode version is not "
1238 				     "supported with metadata_csum enabled.");
1239 			return -ENOTTY;
1240 		}
1241 
1242 		err = mnt_want_write_file(filp);
1243 		if (err)
1244 			return err;
1245 		if (get_user(generation, (int __user *) arg)) {
1246 			err = -EFAULT;
1247 			goto setversion_out;
1248 		}
1249 
1250 		inode_lock(inode);
1251 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
1252 		if (IS_ERR(handle)) {
1253 			err = PTR_ERR(handle);
1254 			goto unlock_out;
1255 		}
1256 		err = ext4_reserve_inode_write(handle, inode, &iloc);
1257 		if (err == 0) {
1258 			inode->i_ctime = current_time(inode);
1259 			inode_inc_iversion(inode);
1260 			inode->i_generation = generation;
1261 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
1262 		}
1263 		ext4_journal_stop(handle);
1264 
1265 unlock_out:
1266 		inode_unlock(inode);
1267 setversion_out:
1268 		mnt_drop_write_file(filp);
1269 		return err;
1270 	}
1271 	case EXT4_IOC_GROUP_EXTEND: {
1272 		ext4_fsblk_t n_blocks_count;
1273 		int err, err2=0;
1274 
1275 		err = ext4_resize_begin(sb);
1276 		if (err)
1277 			return err;
1278 
1279 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
1280 			err = -EFAULT;
1281 			goto group_extend_out;
1282 		}
1283 
1284 		if (ext4_has_feature_bigalloc(sb)) {
1285 			ext4_msg(sb, KERN_ERR,
1286 				 "Online resizing not supported with bigalloc");
1287 			err = -EOPNOTSUPP;
1288 			goto group_extend_out;
1289 		}
1290 
1291 		err = mnt_want_write_file(filp);
1292 		if (err)
1293 			goto group_extend_out;
1294 
1295 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
1296 		if (EXT4_SB(sb)->s_journal) {
1297 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1298 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1299 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1300 		}
1301 		if (err == 0)
1302 			err = err2;
1303 		mnt_drop_write_file(filp);
1304 group_extend_out:
1305 		err2 = ext4_resize_end(sb, false);
1306 		if (err == 0)
1307 			err = err2;
1308 		return err;
1309 	}
1310 
1311 	case EXT4_IOC_MOVE_EXT: {
1312 		struct move_extent me;
1313 		struct fd donor;
1314 		int err;
1315 
1316 		if (!(filp->f_mode & FMODE_READ) ||
1317 		    !(filp->f_mode & FMODE_WRITE))
1318 			return -EBADF;
1319 
1320 		if (copy_from_user(&me,
1321 			(struct move_extent __user *)arg, sizeof(me)))
1322 			return -EFAULT;
1323 		me.moved_len = 0;
1324 
1325 		donor = fdget(me.donor_fd);
1326 		if (!donor.file)
1327 			return -EBADF;
1328 
1329 		if (!(donor.file->f_mode & FMODE_WRITE)) {
1330 			err = -EBADF;
1331 			goto mext_out;
1332 		}
1333 
1334 		if (ext4_has_feature_bigalloc(sb)) {
1335 			ext4_msg(sb, KERN_ERR,
1336 				 "Online defrag not supported with bigalloc");
1337 			err = -EOPNOTSUPP;
1338 			goto mext_out;
1339 		} else if (IS_DAX(inode)) {
1340 			ext4_msg(sb, KERN_ERR,
1341 				 "Online defrag not supported with DAX");
1342 			err = -EOPNOTSUPP;
1343 			goto mext_out;
1344 		}
1345 
1346 		err = mnt_want_write_file(filp);
1347 		if (err)
1348 			goto mext_out;
1349 
1350 		err = ext4_move_extents(filp, donor.file, me.orig_start,
1351 					me.donor_start, me.len, &me.moved_len);
1352 		mnt_drop_write_file(filp);
1353 
1354 		if (copy_to_user((struct move_extent __user *)arg,
1355 				 &me, sizeof(me)))
1356 			err = -EFAULT;
1357 mext_out:
1358 		fdput(donor);
1359 		return err;
1360 	}
1361 
1362 	case EXT4_IOC_GROUP_ADD: {
1363 		struct ext4_new_group_data input;
1364 
1365 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
1366 				sizeof(input)))
1367 			return -EFAULT;
1368 
1369 		return ext4_ioctl_group_add(filp, &input);
1370 	}
1371 
1372 	case EXT4_IOC_MIGRATE:
1373 	{
1374 		int err;
1375 		if (!inode_owner_or_capable(mnt_userns, inode))
1376 			return -EACCES;
1377 
1378 		err = mnt_want_write_file(filp);
1379 		if (err)
1380 			return err;
1381 		/*
1382 		 * inode_mutex prevent write and truncate on the file.
1383 		 * Read still goes through. We take i_data_sem in
1384 		 * ext4_ext_swap_inode_data before we switch the
1385 		 * inode format to prevent read.
1386 		 */
1387 		inode_lock((inode));
1388 		err = ext4_ext_migrate(inode);
1389 		inode_unlock((inode));
1390 		mnt_drop_write_file(filp);
1391 		return err;
1392 	}
1393 
1394 	case EXT4_IOC_ALLOC_DA_BLKS:
1395 	{
1396 		int err;
1397 		if (!inode_owner_or_capable(mnt_userns, inode))
1398 			return -EACCES;
1399 
1400 		err = mnt_want_write_file(filp);
1401 		if (err)
1402 			return err;
1403 		err = ext4_alloc_da_blocks(inode);
1404 		mnt_drop_write_file(filp);
1405 		return err;
1406 	}
1407 
1408 	case EXT4_IOC_SWAP_BOOT:
1409 	{
1410 		int err;
1411 		if (!(filp->f_mode & FMODE_WRITE))
1412 			return -EBADF;
1413 		err = mnt_want_write_file(filp);
1414 		if (err)
1415 			return err;
1416 		err = swap_inode_boot_loader(sb, mnt_userns, inode);
1417 		mnt_drop_write_file(filp);
1418 		return err;
1419 	}
1420 
1421 	case EXT4_IOC_RESIZE_FS: {
1422 		ext4_fsblk_t n_blocks_count;
1423 		int err = 0, err2 = 0;
1424 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
1425 
1426 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1427 				   sizeof(__u64))) {
1428 			return -EFAULT;
1429 		}
1430 
1431 		err = ext4_resize_begin(sb);
1432 		if (err)
1433 			return err;
1434 
1435 		err = mnt_want_write_file(filp);
1436 		if (err)
1437 			goto resizefs_out;
1438 
1439 		err = ext4_resize_fs(sb, n_blocks_count);
1440 		if (EXT4_SB(sb)->s_journal) {
1441 			ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE, NULL);
1442 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1443 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1444 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1445 		}
1446 		if (err == 0)
1447 			err = err2;
1448 		mnt_drop_write_file(filp);
1449 		if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1450 		    ext4_has_group_desc_csum(sb) &&
1451 		    test_opt(sb, INIT_INODE_TABLE))
1452 			err = ext4_register_li_request(sb, o_group);
1453 
1454 resizefs_out:
1455 		err2 = ext4_resize_end(sb, true);
1456 		if (err == 0)
1457 			err = err2;
1458 		return err;
1459 	}
1460 
1461 	case FITRIM:
1462 	{
1463 		struct fstrim_range range;
1464 		int ret = 0;
1465 
1466 		if (!capable(CAP_SYS_ADMIN))
1467 			return -EPERM;
1468 
1469 		if (!bdev_max_discard_sectors(sb->s_bdev))
1470 			return -EOPNOTSUPP;
1471 
1472 		/*
1473 		 * We haven't replayed the journal, so we cannot use our
1474 		 * block-bitmap-guided storage zapping commands.
1475 		 */
1476 		if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1477 			return -EROFS;
1478 
1479 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1480 		    sizeof(range)))
1481 			return -EFAULT;
1482 
1483 		ret = ext4_trim_fs(sb, &range);
1484 		if (ret < 0)
1485 			return ret;
1486 
1487 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
1488 		    sizeof(range)))
1489 			return -EFAULT;
1490 
1491 		return 0;
1492 	}
1493 	case EXT4_IOC_PRECACHE_EXTENTS:
1494 		return ext4_ext_precache(inode);
1495 
1496 	case FS_IOC_SET_ENCRYPTION_POLICY:
1497 		if (!ext4_has_feature_encrypt(sb))
1498 			return -EOPNOTSUPP;
1499 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1500 
1501 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1502 		return ext4_ioctl_get_encryption_pwsalt(filp, (void __user *)arg);
1503 
1504 	case FS_IOC_GET_ENCRYPTION_POLICY:
1505 		if (!ext4_has_feature_encrypt(sb))
1506 			return -EOPNOTSUPP;
1507 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1508 
1509 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1510 		if (!ext4_has_feature_encrypt(sb))
1511 			return -EOPNOTSUPP;
1512 		return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
1513 
1514 	case FS_IOC_ADD_ENCRYPTION_KEY:
1515 		if (!ext4_has_feature_encrypt(sb))
1516 			return -EOPNOTSUPP;
1517 		return fscrypt_ioctl_add_key(filp, (void __user *)arg);
1518 
1519 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1520 		if (!ext4_has_feature_encrypt(sb))
1521 			return -EOPNOTSUPP;
1522 		return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
1523 
1524 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1525 		if (!ext4_has_feature_encrypt(sb))
1526 			return -EOPNOTSUPP;
1527 		return fscrypt_ioctl_remove_key_all_users(filp,
1528 							  (void __user *)arg);
1529 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1530 		if (!ext4_has_feature_encrypt(sb))
1531 			return -EOPNOTSUPP;
1532 		return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
1533 
1534 	case FS_IOC_GET_ENCRYPTION_NONCE:
1535 		if (!ext4_has_feature_encrypt(sb))
1536 			return -EOPNOTSUPP;
1537 		return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
1538 
1539 	case EXT4_IOC_CLEAR_ES_CACHE:
1540 	{
1541 		if (!inode_owner_or_capable(mnt_userns, inode))
1542 			return -EACCES;
1543 		ext4_clear_inode_es(inode);
1544 		return 0;
1545 	}
1546 
1547 	case EXT4_IOC_GETSTATE:
1548 	{
1549 		__u32	state = 0;
1550 
1551 		if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED))
1552 			state |= EXT4_STATE_FLAG_EXT_PRECACHED;
1553 		if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
1554 			state |= EXT4_STATE_FLAG_NEW;
1555 		if (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
1556 			state |= EXT4_STATE_FLAG_NEWENTRY;
1557 		if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE))
1558 			state |= EXT4_STATE_FLAG_DA_ALLOC_CLOSE;
1559 
1560 		return put_user(state, (__u32 __user *) arg);
1561 	}
1562 
1563 	case EXT4_IOC_GET_ES_CACHE:
1564 		return ext4_ioctl_get_es_cache(filp, arg);
1565 
1566 	case EXT4_IOC_SHUTDOWN:
1567 		return ext4_shutdown(sb, arg);
1568 
1569 	case FS_IOC_ENABLE_VERITY:
1570 		if (!ext4_has_feature_verity(sb))
1571 			return -EOPNOTSUPP;
1572 		return fsverity_ioctl_enable(filp, (const void __user *)arg);
1573 
1574 	case FS_IOC_MEASURE_VERITY:
1575 		if (!ext4_has_feature_verity(sb))
1576 			return -EOPNOTSUPP;
1577 		return fsverity_ioctl_measure(filp, (void __user *)arg);
1578 
1579 	case FS_IOC_READ_VERITY_METADATA:
1580 		if (!ext4_has_feature_verity(sb))
1581 			return -EOPNOTSUPP;
1582 		return fsverity_ioctl_read_metadata(filp,
1583 						    (const void __user *)arg);
1584 
1585 	case EXT4_IOC_CHECKPOINT:
1586 		return ext4_ioctl_checkpoint(filp, arg);
1587 
1588 	case FS_IOC_GETFSLABEL:
1589 		return ext4_ioctl_getlabel(EXT4_SB(sb), (void __user *)arg);
1590 
1591 	case FS_IOC_SETFSLABEL:
1592 		return ext4_ioctl_setlabel(filp,
1593 					   (const void __user *)arg);
1594 
1595 	case EXT4_IOC_GETFSUUID:
1596 		return ext4_ioctl_getuuid(EXT4_SB(sb), (void __user *)arg);
1597 	case EXT4_IOC_SETFSUUID:
1598 		return ext4_ioctl_setuuid(filp, (const void __user *)arg);
1599 	default:
1600 		return -ENOTTY;
1601 	}
1602 }
1603 
1604 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1605 {
1606 	return __ext4_ioctl(filp, cmd, arg);
1607 }
1608 
1609 #ifdef CONFIG_COMPAT
1610 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1611 {
1612 	/* These are just misnamed, they actually get/put from/to user an int */
1613 	switch (cmd) {
1614 	case EXT4_IOC32_GETVERSION:
1615 		cmd = EXT4_IOC_GETVERSION;
1616 		break;
1617 	case EXT4_IOC32_SETVERSION:
1618 		cmd = EXT4_IOC_SETVERSION;
1619 		break;
1620 	case EXT4_IOC32_GROUP_EXTEND:
1621 		cmd = EXT4_IOC_GROUP_EXTEND;
1622 		break;
1623 	case EXT4_IOC32_GETVERSION_OLD:
1624 		cmd = EXT4_IOC_GETVERSION_OLD;
1625 		break;
1626 	case EXT4_IOC32_SETVERSION_OLD:
1627 		cmd = EXT4_IOC_SETVERSION_OLD;
1628 		break;
1629 	case EXT4_IOC32_GETRSVSZ:
1630 		cmd = EXT4_IOC_GETRSVSZ;
1631 		break;
1632 	case EXT4_IOC32_SETRSVSZ:
1633 		cmd = EXT4_IOC_SETRSVSZ;
1634 		break;
1635 	case EXT4_IOC32_GROUP_ADD: {
1636 		struct compat_ext4_new_group_input __user *uinput;
1637 		struct ext4_new_group_data input;
1638 		int err;
1639 
1640 		uinput = compat_ptr(arg);
1641 		err = get_user(input.group, &uinput->group);
1642 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1643 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1644 		err |= get_user(input.inode_table, &uinput->inode_table);
1645 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1646 		err |= get_user(input.reserved_blocks,
1647 				&uinput->reserved_blocks);
1648 		if (err)
1649 			return -EFAULT;
1650 		return ext4_ioctl_group_add(file, &input);
1651 	}
1652 	case EXT4_IOC_MOVE_EXT:
1653 	case EXT4_IOC_RESIZE_FS:
1654 	case FITRIM:
1655 	case EXT4_IOC_PRECACHE_EXTENTS:
1656 	case FS_IOC_SET_ENCRYPTION_POLICY:
1657 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1658 	case FS_IOC_GET_ENCRYPTION_POLICY:
1659 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1660 	case FS_IOC_ADD_ENCRYPTION_KEY:
1661 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1662 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1663 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1664 	case FS_IOC_GET_ENCRYPTION_NONCE:
1665 	case EXT4_IOC_SHUTDOWN:
1666 	case FS_IOC_GETFSMAP:
1667 	case FS_IOC_ENABLE_VERITY:
1668 	case FS_IOC_MEASURE_VERITY:
1669 	case FS_IOC_READ_VERITY_METADATA:
1670 	case EXT4_IOC_CLEAR_ES_CACHE:
1671 	case EXT4_IOC_GETSTATE:
1672 	case EXT4_IOC_GET_ES_CACHE:
1673 	case EXT4_IOC_CHECKPOINT:
1674 	case FS_IOC_GETFSLABEL:
1675 	case FS_IOC_SETFSLABEL:
1676 	case EXT4_IOC_GETFSUUID:
1677 	case EXT4_IOC_SETFSUUID:
1678 		break;
1679 	default:
1680 		return -ENOIOCTLCMD;
1681 	}
1682 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1683 }
1684 #endif
1685 
1686 static void set_overhead(struct ext4_super_block *es, const void *arg)
1687 {
1688 	es->s_overhead_clusters = cpu_to_le32(*((unsigned long *) arg));
1689 }
1690 
1691 int ext4_update_overhead(struct super_block *sb, bool force)
1692 {
1693 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1694 
1695 	if (sb_rdonly(sb))
1696 		return 0;
1697 	if (!force &&
1698 	    (sbi->s_overhead == 0 ||
1699 	     sbi->s_overhead == le32_to_cpu(sbi->s_es->s_overhead_clusters)))
1700 		return 0;
1701 	return ext4_update_superblocks_fn(sb, set_overhead, &sbi->s_overhead);
1702 }
1703