xref: /openbmc/linux/fs/ext4/ioctl.c (revision 7effbd18)
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 		inode_bl->i_mode = S_IFREG;
435 		if (ext4_has_feature_extents(sb)) {
436 			ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
437 			ext4_ext_tree_init(handle, inode_bl);
438 		} else
439 			memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
440 	}
441 
442 	err = dquot_initialize(inode);
443 	if (err)
444 		goto err_out1;
445 
446 	size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
447 	size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
448 	diff = size - size_bl;
449 	swap_inode_data(inode, inode_bl);
450 
451 	inode->i_ctime = inode_bl->i_ctime = current_time(inode);
452 	inode_inc_iversion(inode);
453 
454 	inode->i_generation = get_random_u32();
455 	inode_bl->i_generation = get_random_u32();
456 	ext4_reset_inode_seed(inode);
457 	ext4_reset_inode_seed(inode_bl);
458 
459 	ext4_discard_preallocations(inode, 0);
460 
461 	err = ext4_mark_inode_dirty(handle, inode);
462 	if (err < 0) {
463 		/* No need to update quota information. */
464 		ext4_warning(inode->i_sb,
465 			"couldn't mark inode #%lu dirty (err %d)",
466 			inode->i_ino, err);
467 		/* Revert all changes: */
468 		swap_inode_data(inode, inode_bl);
469 		ext4_mark_inode_dirty(handle, inode);
470 		goto err_out1;
471 	}
472 
473 	blocks = inode_bl->i_blocks;
474 	bytes = inode_bl->i_bytes;
475 	inode_bl->i_blocks = inode->i_blocks;
476 	inode_bl->i_bytes = inode->i_bytes;
477 	err = ext4_mark_inode_dirty(handle, inode_bl);
478 	if (err < 0) {
479 		/* No need to update quota information. */
480 		ext4_warning(inode_bl->i_sb,
481 			"couldn't mark inode #%lu dirty (err %d)",
482 			inode_bl->i_ino, err);
483 		goto revert;
484 	}
485 
486 	/* Bootloader inode should not be counted into quota information. */
487 	if (diff > 0)
488 		dquot_free_space(inode, diff);
489 	else
490 		err = dquot_alloc_space(inode, -1 * diff);
491 
492 	if (err < 0) {
493 revert:
494 		/* Revert all changes: */
495 		inode_bl->i_blocks = blocks;
496 		inode_bl->i_bytes = bytes;
497 		swap_inode_data(inode, inode_bl);
498 		ext4_mark_inode_dirty(handle, inode);
499 		ext4_mark_inode_dirty(handle, inode_bl);
500 	}
501 
502 err_out1:
503 	ext4_journal_stop(handle);
504 	ext4_double_up_write_data_sem(inode, inode_bl);
505 
506 err_out:
507 	filemap_invalidate_unlock(inode->i_mapping);
508 journal_err_out:
509 	unlock_two_nondirectories(inode, inode_bl);
510 	iput(inode_bl);
511 	return err;
512 }
513 
514 /*
515  * If immutable is set and we are not clearing it, we're not allowed to change
516  * anything else in the inode.  Don't error out if we're only trying to set
517  * immutable on an immutable file.
518  */
519 static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
520 				      unsigned int flags)
521 {
522 	struct ext4_inode_info *ei = EXT4_I(inode);
523 	unsigned int oldflags = ei->i_flags;
524 
525 	if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
526 		return 0;
527 
528 	if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
529 		return -EPERM;
530 	if (ext4_has_feature_project(inode->i_sb) &&
531 	    __kprojid_val(ei->i_projid) != new_projid)
532 		return -EPERM;
533 
534 	return 0;
535 }
536 
537 static void ext4_dax_dontcache(struct inode *inode, unsigned int flags)
538 {
539 	struct ext4_inode_info *ei = EXT4_I(inode);
540 
541 	if (S_ISDIR(inode->i_mode))
542 		return;
543 
544 	if (test_opt2(inode->i_sb, DAX_NEVER) ||
545 	    test_opt(inode->i_sb, DAX_ALWAYS))
546 		return;
547 
548 	if ((ei->i_flags ^ flags) & EXT4_DAX_FL)
549 		d_mark_dontcache(inode);
550 }
551 
552 static bool dax_compatible(struct inode *inode, unsigned int oldflags,
553 			   unsigned int flags)
554 {
555 	/* Allow the DAX flag to be changed on inline directories */
556 	if (S_ISDIR(inode->i_mode)) {
557 		flags &= ~EXT4_INLINE_DATA_FL;
558 		oldflags &= ~EXT4_INLINE_DATA_FL;
559 	}
560 
561 	if (flags & EXT4_DAX_FL) {
562 		if ((oldflags & EXT4_DAX_MUT_EXCL) ||
563 		     ext4_test_inode_state(inode,
564 					  EXT4_STATE_VERITY_IN_PROGRESS)) {
565 			return false;
566 		}
567 	}
568 
569 	if ((flags & EXT4_DAX_MUT_EXCL) && (oldflags & EXT4_DAX_FL))
570 			return false;
571 
572 	return true;
573 }
574 
575 static int ext4_ioctl_setflags(struct inode *inode,
576 			       unsigned int flags)
577 {
578 	struct ext4_inode_info *ei = EXT4_I(inode);
579 	handle_t *handle = NULL;
580 	int err = -EPERM, migrate = 0;
581 	struct ext4_iloc iloc;
582 	unsigned int oldflags, mask, i;
583 	struct super_block *sb = inode->i_sb;
584 
585 	/* Is it quota file? Do not allow user to mess with it */
586 	if (ext4_is_quota_file(inode))
587 		goto flags_out;
588 
589 	oldflags = ei->i_flags;
590 	/*
591 	 * The JOURNAL_DATA flag can only be changed by
592 	 * the relevant capability.
593 	 */
594 	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
595 		if (!capable(CAP_SYS_RESOURCE))
596 			goto flags_out;
597 	}
598 
599 	if (!dax_compatible(inode, oldflags, flags)) {
600 		err = -EOPNOTSUPP;
601 		goto flags_out;
602 	}
603 
604 	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
605 		migrate = 1;
606 
607 	if ((flags ^ oldflags) & EXT4_CASEFOLD_FL) {
608 		if (!ext4_has_feature_casefold(sb)) {
609 			err = -EOPNOTSUPP;
610 			goto flags_out;
611 		}
612 
613 		if (!S_ISDIR(inode->i_mode)) {
614 			err = -ENOTDIR;
615 			goto flags_out;
616 		}
617 
618 		if (!ext4_empty_dir(inode)) {
619 			err = -ENOTEMPTY;
620 			goto flags_out;
621 		}
622 	}
623 
624 	/*
625 	 * Wait for all pending directio and then flush all the dirty pages
626 	 * for this file.  The flush marks all the pages readonly, so any
627 	 * subsequent attempt to write to the file (particularly mmap pages)
628 	 * will come through the filesystem and fail.
629 	 */
630 	if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
631 	    (flags & EXT4_IMMUTABLE_FL)) {
632 		inode_dio_wait(inode);
633 		err = filemap_write_and_wait(inode->i_mapping);
634 		if (err)
635 			goto flags_out;
636 	}
637 
638 	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
639 	if (IS_ERR(handle)) {
640 		err = PTR_ERR(handle);
641 		goto flags_out;
642 	}
643 	if (IS_SYNC(inode))
644 		ext4_handle_sync(handle);
645 	err = ext4_reserve_inode_write(handle, inode, &iloc);
646 	if (err)
647 		goto flags_err;
648 
649 	ext4_dax_dontcache(inode, flags);
650 
651 	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
652 		if (!(mask & EXT4_FL_USER_MODIFIABLE))
653 			continue;
654 		/* These flags get special treatment later */
655 		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
656 			continue;
657 		if (mask & flags)
658 			ext4_set_inode_flag(inode, i);
659 		else
660 			ext4_clear_inode_flag(inode, i);
661 	}
662 
663 	ext4_set_inode_flags(inode, false);
664 
665 	inode->i_ctime = current_time(inode);
666 	inode_inc_iversion(inode);
667 
668 	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
669 flags_err:
670 	ext4_journal_stop(handle);
671 	if (err)
672 		goto flags_out;
673 
674 	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
675 		/*
676 		 * Changes to the journaling mode can cause unsafe changes to
677 		 * S_DAX if the inode is DAX
678 		 */
679 		if (IS_DAX(inode)) {
680 			err = -EBUSY;
681 			goto flags_out;
682 		}
683 
684 		err = ext4_change_inode_journal_flag(inode,
685 						     flags & EXT4_JOURNAL_DATA_FL);
686 		if (err)
687 			goto flags_out;
688 	}
689 	if (migrate) {
690 		if (flags & EXT4_EXTENTS_FL)
691 			err = ext4_ext_migrate(inode);
692 		else
693 			err = ext4_ind_migrate(inode);
694 	}
695 
696 flags_out:
697 	return err;
698 }
699 
700 #ifdef CONFIG_QUOTA
701 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
702 {
703 	struct super_block *sb = inode->i_sb;
704 	struct ext4_inode_info *ei = EXT4_I(inode);
705 	int err, rc;
706 	handle_t *handle;
707 	kprojid_t kprojid;
708 	struct ext4_iloc iloc;
709 	struct ext4_inode *raw_inode;
710 	struct dquot *transfer_to[MAXQUOTAS] = { };
711 
712 	if (!ext4_has_feature_project(sb)) {
713 		if (projid != EXT4_DEF_PROJID)
714 			return -EOPNOTSUPP;
715 		else
716 			return 0;
717 	}
718 
719 	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
720 		return -EOPNOTSUPP;
721 
722 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
723 
724 	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
725 		return 0;
726 
727 	err = -EPERM;
728 	/* Is it quota file? Do not allow user to mess with it */
729 	if (ext4_is_quota_file(inode))
730 		return err;
731 
732 	err = dquot_initialize(inode);
733 	if (err)
734 		return err;
735 
736 	err = ext4_get_inode_loc(inode, &iloc);
737 	if (err)
738 		return err;
739 
740 	raw_inode = ext4_raw_inode(&iloc);
741 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
742 		err = ext4_expand_extra_isize(inode,
743 					      EXT4_SB(sb)->s_want_extra_isize,
744 					      &iloc);
745 		if (err)
746 			return err;
747 	} else {
748 		brelse(iloc.bh);
749 	}
750 
751 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
752 		EXT4_QUOTA_INIT_BLOCKS(sb) +
753 		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
754 	if (IS_ERR(handle))
755 		return PTR_ERR(handle);
756 
757 	err = ext4_reserve_inode_write(handle, inode, &iloc);
758 	if (err)
759 		goto out_stop;
760 
761 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
762 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
763 
764 		/* __dquot_transfer() calls back ext4_get_inode_usage() which
765 		 * counts xattr inode references.
766 		 */
767 		down_read(&EXT4_I(inode)->xattr_sem);
768 		err = __dquot_transfer(inode, transfer_to);
769 		up_read(&EXT4_I(inode)->xattr_sem);
770 		dqput(transfer_to[PRJQUOTA]);
771 		if (err)
772 			goto out_dirty;
773 	}
774 
775 	EXT4_I(inode)->i_projid = kprojid;
776 	inode->i_ctime = current_time(inode);
777 	inode_inc_iversion(inode);
778 out_dirty:
779 	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
780 	if (!err)
781 		err = rc;
782 out_stop:
783 	ext4_journal_stop(handle);
784 	return err;
785 }
786 #else
787 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
788 {
789 	if (projid != EXT4_DEF_PROJID)
790 		return -EOPNOTSUPP;
791 	return 0;
792 }
793 #endif
794 
795 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
796 {
797 	struct ext4_sb_info *sbi = EXT4_SB(sb);
798 	__u32 flags;
799 
800 	if (!capable(CAP_SYS_ADMIN))
801 		return -EPERM;
802 
803 	if (get_user(flags, (__u32 __user *)arg))
804 		return -EFAULT;
805 
806 	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
807 		return -EINVAL;
808 
809 	if (ext4_forced_shutdown(sbi))
810 		return 0;
811 
812 	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
813 	trace_ext4_shutdown(sb, flags);
814 
815 	switch (flags) {
816 	case EXT4_GOING_FLAGS_DEFAULT:
817 		freeze_bdev(sb->s_bdev);
818 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
819 		thaw_bdev(sb->s_bdev);
820 		break;
821 	case EXT4_GOING_FLAGS_LOGFLUSH:
822 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
823 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
824 			(void) ext4_force_commit(sb);
825 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
826 		}
827 		break;
828 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
829 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
830 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
831 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
832 		break;
833 	default:
834 		return -EINVAL;
835 	}
836 	clear_opt(sb, DISCARD);
837 	return 0;
838 }
839 
840 struct getfsmap_info {
841 	struct super_block	*gi_sb;
842 	struct fsmap_head __user *gi_data;
843 	unsigned int		gi_idx;
844 	__u32			gi_last_flags;
845 };
846 
847 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
848 {
849 	struct getfsmap_info *info = priv;
850 	struct fsmap fm;
851 
852 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
853 
854 	info->gi_last_flags = xfm->fmr_flags;
855 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
856 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
857 			sizeof(struct fsmap)))
858 		return -EFAULT;
859 
860 	return 0;
861 }
862 
863 static int ext4_ioc_getfsmap(struct super_block *sb,
864 			     struct fsmap_head __user *arg)
865 {
866 	struct getfsmap_info info = { NULL };
867 	struct ext4_fsmap_head xhead = {0};
868 	struct fsmap_head head;
869 	bool aborted = false;
870 	int error;
871 
872 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
873 		return -EFAULT;
874 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
875 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
876 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
877 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
878 		       sizeof(head.fmh_keys[1].fmr_reserved)))
879 		return -EINVAL;
880 	/*
881 	 * ext4 doesn't report file extents at all, so the only valid
882 	 * file offsets are the magic ones (all zeroes or all ones).
883 	 */
884 	if (head.fmh_keys[0].fmr_offset ||
885 	    (head.fmh_keys[1].fmr_offset != 0 &&
886 	     head.fmh_keys[1].fmr_offset != -1ULL))
887 		return -EINVAL;
888 
889 	xhead.fmh_iflags = head.fmh_iflags;
890 	xhead.fmh_count = head.fmh_count;
891 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
892 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
893 
894 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
895 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
896 
897 	info.gi_sb = sb;
898 	info.gi_data = arg;
899 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
900 	if (error == EXT4_QUERY_RANGE_ABORT)
901 		aborted = true;
902 	else if (error)
903 		return error;
904 
905 	/* If we didn't abort, set the "last" flag in the last fmx */
906 	if (!aborted && info.gi_idx) {
907 		info.gi_last_flags |= FMR_OF_LAST;
908 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
909 				 &info.gi_last_flags,
910 				 sizeof(info.gi_last_flags)))
911 			return -EFAULT;
912 	}
913 
914 	/* copy back header */
915 	head.fmh_entries = xhead.fmh_entries;
916 	head.fmh_oflags = xhead.fmh_oflags;
917 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
918 		return -EFAULT;
919 
920 	return 0;
921 }
922 
923 static long ext4_ioctl_group_add(struct file *file,
924 				 struct ext4_new_group_data *input)
925 {
926 	struct super_block *sb = file_inode(file)->i_sb;
927 	int err, err2=0;
928 
929 	err = ext4_resize_begin(sb);
930 	if (err)
931 		return err;
932 
933 	if (ext4_has_feature_bigalloc(sb)) {
934 		ext4_msg(sb, KERN_ERR,
935 			 "Online resizing not supported with bigalloc");
936 		err = -EOPNOTSUPP;
937 		goto group_add_out;
938 	}
939 
940 	err = mnt_want_write_file(file);
941 	if (err)
942 		goto group_add_out;
943 
944 	err = ext4_group_add(sb, input);
945 	if (EXT4_SB(sb)->s_journal) {
946 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
947 		err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
948 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
949 	}
950 	if (err == 0)
951 		err = err2;
952 	mnt_drop_write_file(file);
953 	if (!err && ext4_has_group_desc_csum(sb) &&
954 	    test_opt(sb, INIT_INODE_TABLE))
955 		err = ext4_register_li_request(sb, input->group);
956 group_add_out:
957 	err2 = ext4_resize_end(sb, false);
958 	if (err == 0)
959 		err = err2;
960 	return err;
961 }
962 
963 int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa)
964 {
965 	struct inode *inode = d_inode(dentry);
966 	struct ext4_inode_info *ei = EXT4_I(inode);
967 	u32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
968 
969 	if (S_ISREG(inode->i_mode))
970 		flags &= ~FS_PROJINHERIT_FL;
971 
972 	fileattr_fill_flags(fa, flags);
973 	if (ext4_has_feature_project(inode->i_sb))
974 		fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
975 
976 	return 0;
977 }
978 
979 int ext4_fileattr_set(struct mnt_idmap *idmap,
980 		      struct dentry *dentry, struct fileattr *fa)
981 {
982 	struct inode *inode = d_inode(dentry);
983 	u32 flags = fa->flags;
984 	int err = -EOPNOTSUPP;
985 
986 	if (flags & ~EXT4_FL_USER_VISIBLE)
987 		goto out;
988 
989 	/*
990 	 * chattr(1) grabs flags via GETFLAGS, modifies the result and
991 	 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
992 	 * more restrictive than just silently masking off visible but
993 	 * not settable flags as we always did.
994 	 */
995 	flags &= EXT4_FL_USER_MODIFIABLE;
996 	if (ext4_mask_flags(inode->i_mode, flags) != flags)
997 		goto out;
998 	err = ext4_ioctl_check_immutable(inode, fa->fsx_projid, flags);
999 	if (err)
1000 		goto out;
1001 	err = ext4_ioctl_setflags(inode, flags);
1002 	if (err)
1003 		goto out;
1004 	err = ext4_ioctl_setproject(inode, fa->fsx_projid);
1005 out:
1006 	return err;
1007 }
1008 
1009 /* So that the fiemap access checks can't overflow on 32 bit machines. */
1010 #define FIEMAP_MAX_EXTENTS	(UINT_MAX / sizeof(struct fiemap_extent))
1011 
1012 static int ext4_ioctl_get_es_cache(struct file *filp, unsigned long arg)
1013 {
1014 	struct fiemap fiemap;
1015 	struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
1016 	struct fiemap_extent_info fieinfo = { 0, };
1017 	struct inode *inode = file_inode(filp);
1018 	int error;
1019 
1020 	if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
1021 		return -EFAULT;
1022 
1023 	if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
1024 		return -EINVAL;
1025 
1026 	fieinfo.fi_flags = fiemap.fm_flags;
1027 	fieinfo.fi_extents_max = fiemap.fm_extent_count;
1028 	fieinfo.fi_extents_start = ufiemap->fm_extents;
1029 
1030 	error = ext4_get_es_cache(inode, &fieinfo, fiemap.fm_start,
1031 			fiemap.fm_length);
1032 	fiemap.fm_flags = fieinfo.fi_flags;
1033 	fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
1034 	if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
1035 		error = -EFAULT;
1036 
1037 	return error;
1038 }
1039 
1040 static int ext4_ioctl_checkpoint(struct file *filp, unsigned long arg)
1041 {
1042 	int err = 0;
1043 	__u32 flags = 0;
1044 	unsigned int flush_flags = 0;
1045 	struct super_block *sb = file_inode(filp)->i_sb;
1046 
1047 	if (copy_from_user(&flags, (__u32 __user *)arg,
1048 				sizeof(__u32)))
1049 		return -EFAULT;
1050 
1051 	if (!capable(CAP_SYS_ADMIN))
1052 		return -EPERM;
1053 
1054 	/* check for invalid bits set */
1055 	if ((flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID) ||
1056 				((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1057 				(flags & JBD2_JOURNAL_FLUSH_ZEROOUT)))
1058 		return -EINVAL;
1059 
1060 	if (!EXT4_SB(sb)->s_journal)
1061 		return -ENODEV;
1062 
1063 	if ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1064 	    !bdev_max_discard_sectors(EXT4_SB(sb)->s_journal->j_dev))
1065 		return -EOPNOTSUPP;
1066 
1067 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
1068 		return 0;
1069 
1070 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DISCARD)
1071 		flush_flags |= JBD2_JOURNAL_FLUSH_DISCARD;
1072 
1073 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT) {
1074 		flush_flags |= JBD2_JOURNAL_FLUSH_ZEROOUT;
1075 		pr_info_ratelimited("warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow");
1076 	}
1077 
1078 	jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1079 	err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, flush_flags);
1080 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1081 
1082 	return err;
1083 }
1084 
1085 static int ext4_ioctl_setlabel(struct file *filp, const char __user *user_label)
1086 {
1087 	size_t len;
1088 	int ret = 0;
1089 	char new_label[EXT4_LABEL_MAX + 1];
1090 	struct super_block *sb = file_inode(filp)->i_sb;
1091 
1092 	if (!capable(CAP_SYS_ADMIN))
1093 		return -EPERM;
1094 
1095 	/*
1096 	 * Copy the maximum length allowed for ext4 label with one more to
1097 	 * find the required terminating null byte in order to test the
1098 	 * label length. The on disk label doesn't need to be null terminated.
1099 	 */
1100 	if (copy_from_user(new_label, user_label, EXT4_LABEL_MAX + 1))
1101 		return -EFAULT;
1102 
1103 	len = strnlen(new_label, EXT4_LABEL_MAX + 1);
1104 	if (len > EXT4_LABEL_MAX)
1105 		return -EINVAL;
1106 
1107 	/*
1108 	 * Clear the buffer after the new label
1109 	 */
1110 	memset(new_label + len, 0, EXT4_LABEL_MAX - len);
1111 
1112 	ret = mnt_want_write_file(filp);
1113 	if (ret)
1114 		return ret;
1115 
1116 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setlabel, new_label);
1117 
1118 	mnt_drop_write_file(filp);
1119 	return ret;
1120 }
1121 
1122 static int ext4_ioctl_getlabel(struct ext4_sb_info *sbi, char __user *user_label)
1123 {
1124 	char label[EXT4_LABEL_MAX + 1];
1125 
1126 	/*
1127 	 * EXT4_LABEL_MAX must always be smaller than FSLABEL_MAX because
1128 	 * FSLABEL_MAX must include terminating null byte, while s_volume_name
1129 	 * does not have to.
1130 	 */
1131 	BUILD_BUG_ON(EXT4_LABEL_MAX >= FSLABEL_MAX);
1132 
1133 	memset(label, 0, sizeof(label));
1134 	lock_buffer(sbi->s_sbh);
1135 	strncpy(label, sbi->s_es->s_volume_name, EXT4_LABEL_MAX);
1136 	unlock_buffer(sbi->s_sbh);
1137 
1138 	if (copy_to_user(user_label, label, sizeof(label)))
1139 		return -EFAULT;
1140 	return 0;
1141 }
1142 
1143 static int ext4_ioctl_getuuid(struct ext4_sb_info *sbi,
1144 			struct fsuuid __user *ufsuuid)
1145 {
1146 	struct fsuuid fsuuid;
1147 	__u8 uuid[UUID_SIZE];
1148 
1149 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1150 		return -EFAULT;
1151 
1152 	if (fsuuid.fsu_len == 0) {
1153 		fsuuid.fsu_len = UUID_SIZE;
1154 		if (copy_to_user(&ufsuuid->fsu_len, &fsuuid.fsu_len,
1155 					sizeof(fsuuid.fsu_len)))
1156 			return -EFAULT;
1157 		return 0;
1158 	}
1159 
1160 	if (fsuuid.fsu_len < UUID_SIZE || fsuuid.fsu_flags != 0)
1161 		return -EINVAL;
1162 
1163 	lock_buffer(sbi->s_sbh);
1164 	memcpy(uuid, sbi->s_es->s_uuid, UUID_SIZE);
1165 	unlock_buffer(sbi->s_sbh);
1166 
1167 	fsuuid.fsu_len = UUID_SIZE;
1168 	if (copy_to_user(ufsuuid, &fsuuid, sizeof(fsuuid)) ||
1169 	    copy_to_user(&ufsuuid->fsu_uuid[0], uuid, UUID_SIZE))
1170 		return -EFAULT;
1171 	return 0;
1172 }
1173 
1174 static int ext4_ioctl_setuuid(struct file *filp,
1175 			const struct fsuuid __user *ufsuuid)
1176 {
1177 	int ret = 0;
1178 	struct super_block *sb = file_inode(filp)->i_sb;
1179 	struct fsuuid fsuuid;
1180 	__u8 uuid[UUID_SIZE];
1181 
1182 	if (!capable(CAP_SYS_ADMIN))
1183 		return -EPERM;
1184 
1185 	/*
1186 	 * If any checksums (group descriptors or metadata) are being used
1187 	 * then the checksum seed feature is required to change the UUID.
1188 	 */
1189 	if (((ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb))
1190 			&& !ext4_has_feature_csum_seed(sb))
1191 		|| ext4_has_feature_stable_inodes(sb))
1192 		return -EOPNOTSUPP;
1193 
1194 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1195 		return -EFAULT;
1196 
1197 	if (fsuuid.fsu_len != UUID_SIZE || fsuuid.fsu_flags != 0)
1198 		return -EINVAL;
1199 
1200 	if (copy_from_user(uuid, &ufsuuid->fsu_uuid[0], UUID_SIZE))
1201 		return -EFAULT;
1202 
1203 	ret = mnt_want_write_file(filp);
1204 	if (ret)
1205 		return ret;
1206 
1207 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setuuid, &uuid);
1208 	mnt_drop_write_file(filp);
1209 
1210 	return ret;
1211 }
1212 
1213 static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1214 {
1215 	struct inode *inode = file_inode(filp);
1216 	struct super_block *sb = inode->i_sb;
1217 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
1218 
1219 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
1220 
1221 	switch (cmd) {
1222 	case FS_IOC_GETFSMAP:
1223 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
1224 	case EXT4_IOC_GETVERSION:
1225 	case EXT4_IOC_GETVERSION_OLD:
1226 		return put_user(inode->i_generation, (int __user *) arg);
1227 	case EXT4_IOC_SETVERSION:
1228 	case EXT4_IOC_SETVERSION_OLD: {
1229 		handle_t *handle;
1230 		struct ext4_iloc iloc;
1231 		__u32 generation;
1232 		int err;
1233 
1234 		if (!inode_owner_or_capable(idmap, inode))
1235 			return -EPERM;
1236 
1237 		if (ext4_has_metadata_csum(inode->i_sb)) {
1238 			ext4_warning(sb, "Setting inode version is not "
1239 				     "supported with metadata_csum enabled.");
1240 			return -ENOTTY;
1241 		}
1242 
1243 		err = mnt_want_write_file(filp);
1244 		if (err)
1245 			return err;
1246 		if (get_user(generation, (int __user *) arg)) {
1247 			err = -EFAULT;
1248 			goto setversion_out;
1249 		}
1250 
1251 		inode_lock(inode);
1252 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
1253 		if (IS_ERR(handle)) {
1254 			err = PTR_ERR(handle);
1255 			goto unlock_out;
1256 		}
1257 		err = ext4_reserve_inode_write(handle, inode, &iloc);
1258 		if (err == 0) {
1259 			inode->i_ctime = current_time(inode);
1260 			inode_inc_iversion(inode);
1261 			inode->i_generation = generation;
1262 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
1263 		}
1264 		ext4_journal_stop(handle);
1265 
1266 unlock_out:
1267 		inode_unlock(inode);
1268 setversion_out:
1269 		mnt_drop_write_file(filp);
1270 		return err;
1271 	}
1272 	case EXT4_IOC_GROUP_EXTEND: {
1273 		ext4_fsblk_t n_blocks_count;
1274 		int err, err2=0;
1275 
1276 		err = ext4_resize_begin(sb);
1277 		if (err)
1278 			return err;
1279 
1280 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
1281 			err = -EFAULT;
1282 			goto group_extend_out;
1283 		}
1284 
1285 		if (ext4_has_feature_bigalloc(sb)) {
1286 			ext4_msg(sb, KERN_ERR,
1287 				 "Online resizing not supported with bigalloc");
1288 			err = -EOPNOTSUPP;
1289 			goto group_extend_out;
1290 		}
1291 
1292 		err = mnt_want_write_file(filp);
1293 		if (err)
1294 			goto group_extend_out;
1295 
1296 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
1297 		if (EXT4_SB(sb)->s_journal) {
1298 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1299 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1300 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1301 		}
1302 		if (err == 0)
1303 			err = err2;
1304 		mnt_drop_write_file(filp);
1305 group_extend_out:
1306 		err2 = ext4_resize_end(sb, false);
1307 		if (err == 0)
1308 			err = err2;
1309 		return err;
1310 	}
1311 
1312 	case EXT4_IOC_MOVE_EXT: {
1313 		struct move_extent me;
1314 		struct fd donor;
1315 		int err;
1316 
1317 		if (!(filp->f_mode & FMODE_READ) ||
1318 		    !(filp->f_mode & FMODE_WRITE))
1319 			return -EBADF;
1320 
1321 		if (copy_from_user(&me,
1322 			(struct move_extent __user *)arg, sizeof(me)))
1323 			return -EFAULT;
1324 		me.moved_len = 0;
1325 
1326 		donor = fdget(me.donor_fd);
1327 		if (!donor.file)
1328 			return -EBADF;
1329 
1330 		if (!(donor.file->f_mode & FMODE_WRITE)) {
1331 			err = -EBADF;
1332 			goto mext_out;
1333 		}
1334 
1335 		if (ext4_has_feature_bigalloc(sb)) {
1336 			ext4_msg(sb, KERN_ERR,
1337 				 "Online defrag not supported with bigalloc");
1338 			err = -EOPNOTSUPP;
1339 			goto mext_out;
1340 		} else if (IS_DAX(inode)) {
1341 			ext4_msg(sb, KERN_ERR,
1342 				 "Online defrag not supported with DAX");
1343 			err = -EOPNOTSUPP;
1344 			goto mext_out;
1345 		}
1346 
1347 		err = mnt_want_write_file(filp);
1348 		if (err)
1349 			goto mext_out;
1350 
1351 		err = ext4_move_extents(filp, donor.file, me.orig_start,
1352 					me.donor_start, me.len, &me.moved_len);
1353 		mnt_drop_write_file(filp);
1354 
1355 		if (copy_to_user((struct move_extent __user *)arg,
1356 				 &me, sizeof(me)))
1357 			err = -EFAULT;
1358 mext_out:
1359 		fdput(donor);
1360 		return err;
1361 	}
1362 
1363 	case EXT4_IOC_GROUP_ADD: {
1364 		struct ext4_new_group_data input;
1365 
1366 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
1367 				sizeof(input)))
1368 			return -EFAULT;
1369 
1370 		return ext4_ioctl_group_add(filp, &input);
1371 	}
1372 
1373 	case EXT4_IOC_MIGRATE:
1374 	{
1375 		int err;
1376 		if (!inode_owner_or_capable(idmap, inode))
1377 			return -EACCES;
1378 
1379 		err = mnt_want_write_file(filp);
1380 		if (err)
1381 			return err;
1382 		/*
1383 		 * inode_mutex prevent write and truncate on the file.
1384 		 * Read still goes through. We take i_data_sem in
1385 		 * ext4_ext_swap_inode_data before we switch the
1386 		 * inode format to prevent read.
1387 		 */
1388 		inode_lock((inode));
1389 		err = ext4_ext_migrate(inode);
1390 		inode_unlock((inode));
1391 		mnt_drop_write_file(filp);
1392 		return err;
1393 	}
1394 
1395 	case EXT4_IOC_ALLOC_DA_BLKS:
1396 	{
1397 		int err;
1398 		if (!inode_owner_or_capable(idmap, inode))
1399 			return -EACCES;
1400 
1401 		err = mnt_want_write_file(filp);
1402 		if (err)
1403 			return err;
1404 		err = ext4_alloc_da_blocks(inode);
1405 		mnt_drop_write_file(filp);
1406 		return err;
1407 	}
1408 
1409 	case EXT4_IOC_SWAP_BOOT:
1410 	{
1411 		int err;
1412 		if (!(filp->f_mode & FMODE_WRITE))
1413 			return -EBADF;
1414 		err = mnt_want_write_file(filp);
1415 		if (err)
1416 			return err;
1417 		err = swap_inode_boot_loader(sb, idmap, inode);
1418 		mnt_drop_write_file(filp);
1419 		return err;
1420 	}
1421 
1422 	case EXT4_IOC_RESIZE_FS: {
1423 		ext4_fsblk_t n_blocks_count;
1424 		int err = 0, err2 = 0;
1425 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
1426 
1427 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1428 				   sizeof(__u64))) {
1429 			return -EFAULT;
1430 		}
1431 
1432 		err = ext4_resize_begin(sb);
1433 		if (err)
1434 			return err;
1435 
1436 		err = mnt_want_write_file(filp);
1437 		if (err)
1438 			goto resizefs_out;
1439 
1440 		err = ext4_resize_fs(sb, n_blocks_count);
1441 		if (EXT4_SB(sb)->s_journal) {
1442 			ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE, NULL);
1443 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1444 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1445 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1446 		}
1447 		if (err == 0)
1448 			err = err2;
1449 		mnt_drop_write_file(filp);
1450 		if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1451 		    ext4_has_group_desc_csum(sb) &&
1452 		    test_opt(sb, INIT_INODE_TABLE))
1453 			err = ext4_register_li_request(sb, o_group);
1454 
1455 resizefs_out:
1456 		err2 = ext4_resize_end(sb, true);
1457 		if (err == 0)
1458 			err = err2;
1459 		return err;
1460 	}
1461 
1462 	case FITRIM:
1463 	{
1464 		struct fstrim_range range;
1465 		int ret = 0;
1466 
1467 		if (!capable(CAP_SYS_ADMIN))
1468 			return -EPERM;
1469 
1470 		if (!bdev_max_discard_sectors(sb->s_bdev))
1471 			return -EOPNOTSUPP;
1472 
1473 		/*
1474 		 * We haven't replayed the journal, so we cannot use our
1475 		 * block-bitmap-guided storage zapping commands.
1476 		 */
1477 		if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1478 			return -EROFS;
1479 
1480 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1481 		    sizeof(range)))
1482 			return -EFAULT;
1483 
1484 		ret = ext4_trim_fs(sb, &range);
1485 		if (ret < 0)
1486 			return ret;
1487 
1488 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
1489 		    sizeof(range)))
1490 			return -EFAULT;
1491 
1492 		return 0;
1493 	}
1494 	case EXT4_IOC_PRECACHE_EXTENTS:
1495 		return ext4_ext_precache(inode);
1496 
1497 	case FS_IOC_SET_ENCRYPTION_POLICY:
1498 		if (!ext4_has_feature_encrypt(sb))
1499 			return -EOPNOTSUPP;
1500 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1501 
1502 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1503 		return ext4_ioctl_get_encryption_pwsalt(filp, (void __user *)arg);
1504 
1505 	case FS_IOC_GET_ENCRYPTION_POLICY:
1506 		if (!ext4_has_feature_encrypt(sb))
1507 			return -EOPNOTSUPP;
1508 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1509 
1510 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1511 		if (!ext4_has_feature_encrypt(sb))
1512 			return -EOPNOTSUPP;
1513 		return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
1514 
1515 	case FS_IOC_ADD_ENCRYPTION_KEY:
1516 		if (!ext4_has_feature_encrypt(sb))
1517 			return -EOPNOTSUPP;
1518 		return fscrypt_ioctl_add_key(filp, (void __user *)arg);
1519 
1520 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1521 		if (!ext4_has_feature_encrypt(sb))
1522 			return -EOPNOTSUPP;
1523 		return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
1524 
1525 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1526 		if (!ext4_has_feature_encrypt(sb))
1527 			return -EOPNOTSUPP;
1528 		return fscrypt_ioctl_remove_key_all_users(filp,
1529 							  (void __user *)arg);
1530 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1531 		if (!ext4_has_feature_encrypt(sb))
1532 			return -EOPNOTSUPP;
1533 		return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
1534 
1535 	case FS_IOC_GET_ENCRYPTION_NONCE:
1536 		if (!ext4_has_feature_encrypt(sb))
1537 			return -EOPNOTSUPP;
1538 		return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
1539 
1540 	case EXT4_IOC_CLEAR_ES_CACHE:
1541 	{
1542 		if (!inode_owner_or_capable(idmap, inode))
1543 			return -EACCES;
1544 		ext4_clear_inode_es(inode);
1545 		return 0;
1546 	}
1547 
1548 	case EXT4_IOC_GETSTATE:
1549 	{
1550 		__u32	state = 0;
1551 
1552 		if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED))
1553 			state |= EXT4_STATE_FLAG_EXT_PRECACHED;
1554 		if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
1555 			state |= EXT4_STATE_FLAG_NEW;
1556 		if (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
1557 			state |= EXT4_STATE_FLAG_NEWENTRY;
1558 		if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE))
1559 			state |= EXT4_STATE_FLAG_DA_ALLOC_CLOSE;
1560 
1561 		return put_user(state, (__u32 __user *) arg);
1562 	}
1563 
1564 	case EXT4_IOC_GET_ES_CACHE:
1565 		return ext4_ioctl_get_es_cache(filp, arg);
1566 
1567 	case EXT4_IOC_SHUTDOWN:
1568 		return ext4_shutdown(sb, arg);
1569 
1570 	case FS_IOC_ENABLE_VERITY:
1571 		if (!ext4_has_feature_verity(sb))
1572 			return -EOPNOTSUPP;
1573 		return fsverity_ioctl_enable(filp, (const void __user *)arg);
1574 
1575 	case FS_IOC_MEASURE_VERITY:
1576 		if (!ext4_has_feature_verity(sb))
1577 			return -EOPNOTSUPP;
1578 		return fsverity_ioctl_measure(filp, (void __user *)arg);
1579 
1580 	case FS_IOC_READ_VERITY_METADATA:
1581 		if (!ext4_has_feature_verity(sb))
1582 			return -EOPNOTSUPP;
1583 		return fsverity_ioctl_read_metadata(filp,
1584 						    (const void __user *)arg);
1585 
1586 	case EXT4_IOC_CHECKPOINT:
1587 		return ext4_ioctl_checkpoint(filp, arg);
1588 
1589 	case FS_IOC_GETFSLABEL:
1590 		return ext4_ioctl_getlabel(EXT4_SB(sb), (void __user *)arg);
1591 
1592 	case FS_IOC_SETFSLABEL:
1593 		return ext4_ioctl_setlabel(filp,
1594 					   (const void __user *)arg);
1595 
1596 	case EXT4_IOC_GETFSUUID:
1597 		return ext4_ioctl_getuuid(EXT4_SB(sb), (void __user *)arg);
1598 	case EXT4_IOC_SETFSUUID:
1599 		return ext4_ioctl_setuuid(filp, (const void __user *)arg);
1600 	default:
1601 		return -ENOTTY;
1602 	}
1603 }
1604 
1605 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1606 {
1607 	return __ext4_ioctl(filp, cmd, arg);
1608 }
1609 
1610 #ifdef CONFIG_COMPAT
1611 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1612 {
1613 	/* These are just misnamed, they actually get/put from/to user an int */
1614 	switch (cmd) {
1615 	case EXT4_IOC32_GETVERSION:
1616 		cmd = EXT4_IOC_GETVERSION;
1617 		break;
1618 	case EXT4_IOC32_SETVERSION:
1619 		cmd = EXT4_IOC_SETVERSION;
1620 		break;
1621 	case EXT4_IOC32_GROUP_EXTEND:
1622 		cmd = EXT4_IOC_GROUP_EXTEND;
1623 		break;
1624 	case EXT4_IOC32_GETVERSION_OLD:
1625 		cmd = EXT4_IOC_GETVERSION_OLD;
1626 		break;
1627 	case EXT4_IOC32_SETVERSION_OLD:
1628 		cmd = EXT4_IOC_SETVERSION_OLD;
1629 		break;
1630 	case EXT4_IOC32_GETRSVSZ:
1631 		cmd = EXT4_IOC_GETRSVSZ;
1632 		break;
1633 	case EXT4_IOC32_SETRSVSZ:
1634 		cmd = EXT4_IOC_SETRSVSZ;
1635 		break;
1636 	case EXT4_IOC32_GROUP_ADD: {
1637 		struct compat_ext4_new_group_input __user *uinput;
1638 		struct ext4_new_group_data input;
1639 		int err;
1640 
1641 		uinput = compat_ptr(arg);
1642 		err = get_user(input.group, &uinput->group);
1643 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1644 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1645 		err |= get_user(input.inode_table, &uinput->inode_table);
1646 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1647 		err |= get_user(input.reserved_blocks,
1648 				&uinput->reserved_blocks);
1649 		if (err)
1650 			return -EFAULT;
1651 		return ext4_ioctl_group_add(file, &input);
1652 	}
1653 	case EXT4_IOC_MOVE_EXT:
1654 	case EXT4_IOC_RESIZE_FS:
1655 	case FITRIM:
1656 	case EXT4_IOC_PRECACHE_EXTENTS:
1657 	case FS_IOC_SET_ENCRYPTION_POLICY:
1658 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1659 	case FS_IOC_GET_ENCRYPTION_POLICY:
1660 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1661 	case FS_IOC_ADD_ENCRYPTION_KEY:
1662 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1663 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1664 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1665 	case FS_IOC_GET_ENCRYPTION_NONCE:
1666 	case EXT4_IOC_SHUTDOWN:
1667 	case FS_IOC_GETFSMAP:
1668 	case FS_IOC_ENABLE_VERITY:
1669 	case FS_IOC_MEASURE_VERITY:
1670 	case FS_IOC_READ_VERITY_METADATA:
1671 	case EXT4_IOC_CLEAR_ES_CACHE:
1672 	case EXT4_IOC_GETSTATE:
1673 	case EXT4_IOC_GET_ES_CACHE:
1674 	case EXT4_IOC_CHECKPOINT:
1675 	case FS_IOC_GETFSLABEL:
1676 	case FS_IOC_SETFSLABEL:
1677 	case EXT4_IOC_GETFSUUID:
1678 	case EXT4_IOC_SETFSUUID:
1679 		break;
1680 	default:
1681 		return -ENOIOCTLCMD;
1682 	}
1683 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1684 }
1685 #endif
1686 
1687 static void set_overhead(struct ext4_super_block *es, const void *arg)
1688 {
1689 	es->s_overhead_clusters = cpu_to_le32(*((unsigned long *) arg));
1690 }
1691 
1692 int ext4_update_overhead(struct super_block *sb, bool force)
1693 {
1694 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1695 
1696 	if (sb_rdonly(sb))
1697 		return 0;
1698 	if (!force &&
1699 	    (sbi->s_overhead == 0 ||
1700 	     sbi->s_overhead == le32_to_cpu(sbi->s_es->s_overhead_clusters)))
1701 		return 0;
1702 	return ext4_update_superblocks_fn(sb, set_overhead, &sbi->s_overhead);
1703 }
1704