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
30typedef 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 */
37static 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 */
49static void ext4_sb_setuuid(struct ext4_super_block *es, const void *arg)
50{
51	memcpy(es->s_uuid, (__u8 *)arg, UUID_SIZE);
52}
53
54static
55int 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);
89out_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 */
103static 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
168out_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 */
183static
184int 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		}
261next_grp:
262		grp = ext4_list_backups(sb, &three, &five, &seven);
263	}
264
265out_journal:
266	if (handle) {
267		ret = ext4_journal_stop(handle);
268		if (ret && !err)
269			err = ret;
270	}
271out:
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 */
285static 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 */
309static 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	struct timespec64 ts1, ts2;
316
317	ei1 = EXT4_I(inode1);
318	ei2 = EXT4_I(inode2);
319
320	swap(inode1->i_version, inode2->i_version);
321
322	ts1 = inode_get_atime(inode1);
323	ts2 = inode_get_atime(inode2);
324	inode_set_atime_to_ts(inode1, ts2);
325	inode_set_atime_to_ts(inode2, ts1);
326
327	ts1 = inode_get_mtime(inode1);
328	ts2 = inode_get_mtime(inode2);
329	inode_set_mtime_to_ts(inode1, ts2);
330	inode_set_mtime_to_ts(inode2, ts1);
331
332	memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
333	tmp = ei1->i_flags & EXT4_FL_SHOULD_SWAP;
334	ei1->i_flags = (ei2->i_flags & EXT4_FL_SHOULD_SWAP) |
335		(ei1->i_flags & ~EXT4_FL_SHOULD_SWAP);
336	ei2->i_flags = tmp | (ei2->i_flags & ~EXT4_FL_SHOULD_SWAP);
337	swap(ei1->i_disksize, ei2->i_disksize);
338	ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
339	ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
340
341	isize = i_size_read(inode1);
342	i_size_write(inode1, i_size_read(inode2));
343	i_size_write(inode2, isize);
344}
345
346void ext4_reset_inode_seed(struct inode *inode)
347{
348	struct ext4_inode_info *ei = EXT4_I(inode);
349	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
350	__le32 inum = cpu_to_le32(inode->i_ino);
351	__le32 gen = cpu_to_le32(inode->i_generation);
352	__u32 csum;
353
354	if (!ext4_has_metadata_csum(inode->i_sb))
355		return;
356
357	csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum, sizeof(inum));
358	ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen, sizeof(gen));
359}
360
361/*
362 * Swap the information from the given @inode and the inode
363 * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
364 * important fields of the inodes.
365 *
366 * @sb:         the super block of the filesystem
367 * @idmap:	idmap of the mount the inode was found from
368 * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
369 *
370 */
371static long swap_inode_boot_loader(struct super_block *sb,
372				struct mnt_idmap *idmap,
373				struct inode *inode)
374{
375	handle_t *handle;
376	int err;
377	struct inode *inode_bl;
378	struct ext4_inode_info *ei_bl;
379	qsize_t size, size_bl, diff;
380	blkcnt_t blocks;
381	unsigned short bytes;
382
383	inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO,
384			EXT4_IGET_SPECIAL | EXT4_IGET_BAD);
385	if (IS_ERR(inode_bl))
386		return PTR_ERR(inode_bl);
387	ei_bl = EXT4_I(inode_bl);
388
389	/* Protect orig inodes against a truncate and make sure,
390	 * that only 1 swap_inode_boot_loader is running. */
391	lock_two_nondirectories(inode, inode_bl);
392
393	if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode) ||
394	    IS_SWAPFILE(inode) || IS_ENCRYPTED(inode) ||
395	    (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL) ||
396	    ext4_has_inline_data(inode)) {
397		err = -EINVAL;
398		goto journal_err_out;
399	}
400
401	if (IS_RDONLY(inode) || IS_APPEND(inode) || IS_IMMUTABLE(inode) ||
402	    !inode_owner_or_capable(idmap, inode) ||
403	    !capable(CAP_SYS_ADMIN)) {
404		err = -EPERM;
405		goto journal_err_out;
406	}
407
408	filemap_invalidate_lock(inode->i_mapping);
409	err = filemap_write_and_wait(inode->i_mapping);
410	if (err)
411		goto err_out;
412
413	err = filemap_write_and_wait(inode_bl->i_mapping);
414	if (err)
415		goto err_out;
416
417	/* Wait for all existing dio workers */
418	inode_dio_wait(inode);
419	inode_dio_wait(inode_bl);
420
421	truncate_inode_pages(&inode->i_data, 0);
422	truncate_inode_pages(&inode_bl->i_data, 0);
423
424	handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
425	if (IS_ERR(handle)) {
426		err = -EINVAL;
427		goto err_out;
428	}
429	ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_SWAP_BOOT, handle);
430
431	/* Protect extent tree against block allocations via delalloc */
432	ext4_double_down_write_data_sem(inode, inode_bl);
433
434	if (is_bad_inode(inode_bl) || !S_ISREG(inode_bl->i_mode)) {
435		/* this inode has never been used as a BOOT_LOADER */
436		set_nlink(inode_bl, 1);
437		i_uid_write(inode_bl, 0);
438		i_gid_write(inode_bl, 0);
439		inode_bl->i_flags = 0;
440		ei_bl->i_flags = 0;
441		inode_set_iversion(inode_bl, 1);
442		i_size_write(inode_bl, 0);
443		EXT4_I(inode_bl)->i_disksize = inode_bl->i_size;
444		inode_bl->i_mode = S_IFREG;
445		if (ext4_has_feature_extents(sb)) {
446			ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
447			ext4_ext_tree_init(handle, inode_bl);
448		} else
449			memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
450	}
451
452	err = dquot_initialize(inode);
453	if (err)
454		goto err_out1;
455
456	size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
457	size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
458	diff = size - size_bl;
459	swap_inode_data(inode, inode_bl);
460
461	inode_set_ctime_current(inode);
462	inode_set_ctime_current(inode_bl);
463	inode_inc_iversion(inode);
464
465	inode->i_generation = get_random_u32();
466	inode_bl->i_generation = get_random_u32();
467	ext4_reset_inode_seed(inode);
468	ext4_reset_inode_seed(inode_bl);
469
470	ext4_discard_preallocations(inode);
471
472	err = ext4_mark_inode_dirty(handle, inode);
473	if (err < 0) {
474		/* No need to update quota information. */
475		ext4_warning(inode->i_sb,
476			"couldn't mark inode #%lu dirty (err %d)",
477			inode->i_ino, err);
478		/* Revert all changes: */
479		swap_inode_data(inode, inode_bl);
480		ext4_mark_inode_dirty(handle, inode);
481		goto err_out1;
482	}
483
484	blocks = inode_bl->i_blocks;
485	bytes = inode_bl->i_bytes;
486	inode_bl->i_blocks = inode->i_blocks;
487	inode_bl->i_bytes = inode->i_bytes;
488	err = ext4_mark_inode_dirty(handle, inode_bl);
489	if (err < 0) {
490		/* No need to update quota information. */
491		ext4_warning(inode_bl->i_sb,
492			"couldn't mark inode #%lu dirty (err %d)",
493			inode_bl->i_ino, err);
494		goto revert;
495	}
496
497	/* Bootloader inode should not be counted into quota information. */
498	if (diff > 0)
499		dquot_free_space(inode, diff);
500	else
501		err = dquot_alloc_space(inode, -1 * diff);
502
503	if (err < 0) {
504revert:
505		/* Revert all changes: */
506		inode_bl->i_blocks = blocks;
507		inode_bl->i_bytes = bytes;
508		swap_inode_data(inode, inode_bl);
509		ext4_mark_inode_dirty(handle, inode);
510		ext4_mark_inode_dirty(handle, inode_bl);
511	}
512
513err_out1:
514	ext4_journal_stop(handle);
515	ext4_double_up_write_data_sem(inode, inode_bl);
516
517err_out:
518	filemap_invalidate_unlock(inode->i_mapping);
519journal_err_out:
520	unlock_two_nondirectories(inode, inode_bl);
521	iput(inode_bl);
522	return err;
523}
524
525/*
526 * If immutable is set and we are not clearing it, we're not allowed to change
527 * anything else in the inode.  Don't error out if we're only trying to set
528 * immutable on an immutable file.
529 */
530static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
531				      unsigned int flags)
532{
533	struct ext4_inode_info *ei = EXT4_I(inode);
534	unsigned int oldflags = ei->i_flags;
535
536	if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
537		return 0;
538
539	if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
540		return -EPERM;
541	if (ext4_has_feature_project(inode->i_sb) &&
542	    __kprojid_val(ei->i_projid) != new_projid)
543		return -EPERM;
544
545	return 0;
546}
547
548static void ext4_dax_dontcache(struct inode *inode, unsigned int flags)
549{
550	struct ext4_inode_info *ei = EXT4_I(inode);
551
552	if (S_ISDIR(inode->i_mode))
553		return;
554
555	if (test_opt2(inode->i_sb, DAX_NEVER) ||
556	    test_opt(inode->i_sb, DAX_ALWAYS))
557		return;
558
559	if ((ei->i_flags ^ flags) & EXT4_DAX_FL)
560		d_mark_dontcache(inode);
561}
562
563static bool dax_compatible(struct inode *inode, unsigned int oldflags,
564			   unsigned int flags)
565{
566	/* Allow the DAX flag to be changed on inline directories */
567	if (S_ISDIR(inode->i_mode)) {
568		flags &= ~EXT4_INLINE_DATA_FL;
569		oldflags &= ~EXT4_INLINE_DATA_FL;
570	}
571
572	if (flags & EXT4_DAX_FL) {
573		if ((oldflags & EXT4_DAX_MUT_EXCL) ||
574		     ext4_test_inode_state(inode,
575					  EXT4_STATE_VERITY_IN_PROGRESS)) {
576			return false;
577		}
578	}
579
580	if ((flags & EXT4_DAX_MUT_EXCL) && (oldflags & EXT4_DAX_FL))
581			return false;
582
583	return true;
584}
585
586static int ext4_ioctl_setflags(struct inode *inode,
587			       unsigned int flags)
588{
589	struct ext4_inode_info *ei = EXT4_I(inode);
590	handle_t *handle = NULL;
591	int err = -EPERM, migrate = 0;
592	struct ext4_iloc iloc;
593	unsigned int oldflags, mask, i;
594	struct super_block *sb = inode->i_sb;
595
596	/* Is it quota file? Do not allow user to mess with it */
597	if (ext4_is_quota_file(inode))
598		goto flags_out;
599
600	oldflags = ei->i_flags;
601	/*
602	 * The JOURNAL_DATA flag can only be changed by
603	 * the relevant capability.
604	 */
605	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
606		if (!capable(CAP_SYS_RESOURCE))
607			goto flags_out;
608	}
609
610	if (!dax_compatible(inode, oldflags, flags)) {
611		err = -EOPNOTSUPP;
612		goto flags_out;
613	}
614
615	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
616		migrate = 1;
617
618	if ((flags ^ oldflags) & EXT4_CASEFOLD_FL) {
619		if (!ext4_has_feature_casefold(sb)) {
620			err = -EOPNOTSUPP;
621			goto flags_out;
622		}
623
624		if (!S_ISDIR(inode->i_mode)) {
625			err = -ENOTDIR;
626			goto flags_out;
627		}
628
629		if (!ext4_empty_dir(inode)) {
630			err = -ENOTEMPTY;
631			goto flags_out;
632		}
633	}
634
635	/*
636	 * Wait for all pending directio and then flush all the dirty pages
637	 * for this file.  The flush marks all the pages readonly, so any
638	 * subsequent attempt to write to the file (particularly mmap pages)
639	 * will come through the filesystem and fail.
640	 */
641	if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
642	    (flags & EXT4_IMMUTABLE_FL)) {
643		inode_dio_wait(inode);
644		err = filemap_write_and_wait(inode->i_mapping);
645		if (err)
646			goto flags_out;
647	}
648
649	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
650	if (IS_ERR(handle)) {
651		err = PTR_ERR(handle);
652		goto flags_out;
653	}
654	if (IS_SYNC(inode))
655		ext4_handle_sync(handle);
656	err = ext4_reserve_inode_write(handle, inode, &iloc);
657	if (err)
658		goto flags_err;
659
660	ext4_dax_dontcache(inode, flags);
661
662	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
663		if (!(mask & EXT4_FL_USER_MODIFIABLE))
664			continue;
665		/* These flags get special treatment later */
666		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
667			continue;
668		if (mask & flags)
669			ext4_set_inode_flag(inode, i);
670		else
671			ext4_clear_inode_flag(inode, i);
672	}
673
674	ext4_set_inode_flags(inode, false);
675
676	inode_set_ctime_current(inode);
677	inode_inc_iversion(inode);
678
679	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
680flags_err:
681	ext4_journal_stop(handle);
682	if (err)
683		goto flags_out;
684
685	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
686		/*
687		 * Changes to the journaling mode can cause unsafe changes to
688		 * S_DAX if the inode is DAX
689		 */
690		if (IS_DAX(inode)) {
691			err = -EBUSY;
692			goto flags_out;
693		}
694
695		err = ext4_change_inode_journal_flag(inode,
696						     flags & EXT4_JOURNAL_DATA_FL);
697		if (err)
698			goto flags_out;
699	}
700	if (migrate) {
701		if (flags & EXT4_EXTENTS_FL)
702			err = ext4_ext_migrate(inode);
703		else
704			err = ext4_ind_migrate(inode);
705	}
706
707flags_out:
708	return err;
709}
710
711#ifdef CONFIG_QUOTA
712static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
713{
714	struct super_block *sb = inode->i_sb;
715	struct ext4_inode_info *ei = EXT4_I(inode);
716	int err, rc;
717	handle_t *handle;
718	kprojid_t kprojid;
719	struct ext4_iloc iloc;
720	struct ext4_inode *raw_inode;
721	struct dquot *transfer_to[MAXQUOTAS] = { };
722
723	if (!ext4_has_feature_project(sb)) {
724		if (projid != EXT4_DEF_PROJID)
725			return -EOPNOTSUPP;
726		else
727			return 0;
728	}
729
730	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
731		return -EOPNOTSUPP;
732
733	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
734
735	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
736		return 0;
737
738	err = -EPERM;
739	/* Is it quota file? Do not allow user to mess with it */
740	if (ext4_is_quota_file(inode))
741		return err;
742
743	err = dquot_initialize(inode);
744	if (err)
745		return err;
746
747	err = ext4_get_inode_loc(inode, &iloc);
748	if (err)
749		return err;
750
751	raw_inode = ext4_raw_inode(&iloc);
752	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
753		err = ext4_expand_extra_isize(inode,
754					      EXT4_SB(sb)->s_want_extra_isize,
755					      &iloc);
756		if (err)
757			return err;
758	} else {
759		brelse(iloc.bh);
760	}
761
762	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
763		EXT4_QUOTA_INIT_BLOCKS(sb) +
764		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
765	if (IS_ERR(handle))
766		return PTR_ERR(handle);
767
768	err = ext4_reserve_inode_write(handle, inode, &iloc);
769	if (err)
770		goto out_stop;
771
772	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
773	if (!IS_ERR(transfer_to[PRJQUOTA])) {
774
775		/* __dquot_transfer() calls back ext4_get_inode_usage() which
776		 * counts xattr inode references.
777		 */
778		down_read(&EXT4_I(inode)->xattr_sem);
779		err = __dquot_transfer(inode, transfer_to);
780		up_read(&EXT4_I(inode)->xattr_sem);
781		dqput(transfer_to[PRJQUOTA]);
782		if (err)
783			goto out_dirty;
784	}
785
786	EXT4_I(inode)->i_projid = kprojid;
787	inode_set_ctime_current(inode);
788	inode_inc_iversion(inode);
789out_dirty:
790	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
791	if (!err)
792		err = rc;
793out_stop:
794	ext4_journal_stop(handle);
795	return err;
796}
797#else
798static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
799{
800	if (projid != EXT4_DEF_PROJID)
801		return -EOPNOTSUPP;
802	return 0;
803}
804#endif
805
806int ext4_force_shutdown(struct super_block *sb, u32 flags)
807{
808	struct ext4_sb_info *sbi = EXT4_SB(sb);
809	int ret;
810
811	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
812		return -EINVAL;
813
814	if (ext4_forced_shutdown(sb))
815		return 0;
816
817	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
818	trace_ext4_shutdown(sb, flags);
819
820	switch (flags) {
821	case EXT4_GOING_FLAGS_DEFAULT:
822		ret = bdev_freeze(sb->s_bdev);
823		if (ret)
824			return ret;
825		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
826		bdev_thaw(sb->s_bdev);
827		break;
828	case EXT4_GOING_FLAGS_LOGFLUSH:
829		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
830		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
831			(void) ext4_force_commit(sb);
832			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
833		}
834		break;
835	case EXT4_GOING_FLAGS_NOLOGFLUSH:
836		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
837		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
838			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
839		break;
840	default:
841		return -EINVAL;
842	}
843	clear_opt(sb, DISCARD);
844	return 0;
845}
846
847static int ext4_ioctl_shutdown(struct super_block *sb, unsigned long arg)
848{
849	u32 flags;
850
851	if (!capable(CAP_SYS_ADMIN))
852		return -EPERM;
853
854	if (get_user(flags, (__u32 __user *)arg))
855		return -EFAULT;
856
857	return ext4_force_shutdown(sb, flags);
858}
859
860struct getfsmap_info {
861	struct super_block	*gi_sb;
862	struct fsmap_head __user *gi_data;
863	unsigned int		gi_idx;
864	__u32			gi_last_flags;
865};
866
867static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
868{
869	struct getfsmap_info *info = priv;
870	struct fsmap fm;
871
872	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
873
874	info->gi_last_flags = xfm->fmr_flags;
875	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
876	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
877			sizeof(struct fsmap)))
878		return -EFAULT;
879
880	return 0;
881}
882
883static int ext4_ioc_getfsmap(struct super_block *sb,
884			     struct fsmap_head __user *arg)
885{
886	struct getfsmap_info info = { NULL };
887	struct ext4_fsmap_head xhead = {0};
888	struct fsmap_head head;
889	bool aborted = false;
890	int error;
891
892	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
893		return -EFAULT;
894	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
895	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
896		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
897	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
898		       sizeof(head.fmh_keys[1].fmr_reserved)))
899		return -EINVAL;
900	/*
901	 * ext4 doesn't report file extents at all, so the only valid
902	 * file offsets are the magic ones (all zeroes or all ones).
903	 */
904	if (head.fmh_keys[0].fmr_offset ||
905	    (head.fmh_keys[1].fmr_offset != 0 &&
906	     head.fmh_keys[1].fmr_offset != -1ULL))
907		return -EINVAL;
908
909	xhead.fmh_iflags = head.fmh_iflags;
910	xhead.fmh_count = head.fmh_count;
911	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
912	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
913
914	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
915	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
916
917	info.gi_sb = sb;
918	info.gi_data = arg;
919	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
920	if (error == EXT4_QUERY_RANGE_ABORT)
921		aborted = true;
922	else if (error)
923		return error;
924
925	/* If we didn't abort, set the "last" flag in the last fmx */
926	if (!aborted && info.gi_idx) {
927		info.gi_last_flags |= FMR_OF_LAST;
928		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
929				 &info.gi_last_flags,
930				 sizeof(info.gi_last_flags)))
931			return -EFAULT;
932	}
933
934	/* copy back header */
935	head.fmh_entries = xhead.fmh_entries;
936	head.fmh_oflags = xhead.fmh_oflags;
937	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
938		return -EFAULT;
939
940	return 0;
941}
942
943static long ext4_ioctl_group_add(struct file *file,
944				 struct ext4_new_group_data *input)
945{
946	struct super_block *sb = file_inode(file)->i_sb;
947	int err, err2=0;
948
949	err = ext4_resize_begin(sb);
950	if (err)
951		return err;
952
953	if (ext4_has_feature_bigalloc(sb)) {
954		ext4_msg(sb, KERN_ERR,
955			 "Online resizing not supported with bigalloc");
956		err = -EOPNOTSUPP;
957		goto group_add_out;
958	}
959
960	err = mnt_want_write_file(file);
961	if (err)
962		goto group_add_out;
963
964	err = ext4_group_add(sb, input);
965	if (EXT4_SB(sb)->s_journal) {
966		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
967		err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
968		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
969	}
970	if (err == 0)
971		err = err2;
972	mnt_drop_write_file(file);
973	if (!err && ext4_has_group_desc_csum(sb) &&
974	    test_opt(sb, INIT_INODE_TABLE))
975		err = ext4_register_li_request(sb, input->group);
976group_add_out:
977	err2 = ext4_resize_end(sb, false);
978	if (err == 0)
979		err = err2;
980	return err;
981}
982
983int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa)
984{
985	struct inode *inode = d_inode(dentry);
986	struct ext4_inode_info *ei = EXT4_I(inode);
987	u32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
988
989	if (S_ISREG(inode->i_mode))
990		flags &= ~FS_PROJINHERIT_FL;
991
992	fileattr_fill_flags(fa, flags);
993	if (ext4_has_feature_project(inode->i_sb))
994		fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
995
996	return 0;
997}
998
999int ext4_fileattr_set(struct mnt_idmap *idmap,
1000		      struct dentry *dentry, struct fileattr *fa)
1001{
1002	struct inode *inode = d_inode(dentry);
1003	u32 flags = fa->flags;
1004	int err = -EOPNOTSUPP;
1005
1006	if (flags & ~EXT4_FL_USER_VISIBLE)
1007		goto out;
1008
1009	/*
1010	 * chattr(1) grabs flags via GETFLAGS, modifies the result and
1011	 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
1012	 * more restrictive than just silently masking off visible but
1013	 * not settable flags as we always did.
1014	 */
1015	flags &= EXT4_FL_USER_MODIFIABLE;
1016	if (ext4_mask_flags(inode->i_mode, flags) != flags)
1017		goto out;
1018	err = ext4_ioctl_check_immutable(inode, fa->fsx_projid, flags);
1019	if (err)
1020		goto out;
1021	err = ext4_ioctl_setflags(inode, flags);
1022	if (err)
1023		goto out;
1024	err = ext4_ioctl_setproject(inode, fa->fsx_projid);
1025out:
1026	return err;
1027}
1028
1029/* So that the fiemap access checks can't overflow on 32 bit machines. */
1030#define FIEMAP_MAX_EXTENTS	(UINT_MAX / sizeof(struct fiemap_extent))
1031
1032static int ext4_ioctl_get_es_cache(struct file *filp, unsigned long arg)
1033{
1034	struct fiemap fiemap;
1035	struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
1036	struct fiemap_extent_info fieinfo = { 0, };
1037	struct inode *inode = file_inode(filp);
1038	int error;
1039
1040	if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
1041		return -EFAULT;
1042
1043	if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
1044		return -EINVAL;
1045
1046	fieinfo.fi_flags = fiemap.fm_flags;
1047	fieinfo.fi_extents_max = fiemap.fm_extent_count;
1048	fieinfo.fi_extents_start = ufiemap->fm_extents;
1049
1050	error = ext4_get_es_cache(inode, &fieinfo, fiemap.fm_start,
1051			fiemap.fm_length);
1052	fiemap.fm_flags = fieinfo.fi_flags;
1053	fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
1054	if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
1055		error = -EFAULT;
1056
1057	return error;
1058}
1059
1060static int ext4_ioctl_checkpoint(struct file *filp, unsigned long arg)
1061{
1062	int err = 0;
1063	__u32 flags = 0;
1064	unsigned int flush_flags = 0;
1065	struct super_block *sb = file_inode(filp)->i_sb;
1066
1067	if (copy_from_user(&flags, (__u32 __user *)arg,
1068				sizeof(__u32)))
1069		return -EFAULT;
1070
1071	if (!capable(CAP_SYS_ADMIN))
1072		return -EPERM;
1073
1074	/* check for invalid bits set */
1075	if ((flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID) ||
1076				((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1077				(flags & JBD2_JOURNAL_FLUSH_ZEROOUT)))
1078		return -EINVAL;
1079
1080	if (!EXT4_SB(sb)->s_journal)
1081		return -ENODEV;
1082
1083	if ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1084	    !bdev_max_discard_sectors(EXT4_SB(sb)->s_journal->j_dev))
1085		return -EOPNOTSUPP;
1086
1087	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
1088		return 0;
1089
1090	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DISCARD)
1091		flush_flags |= JBD2_JOURNAL_FLUSH_DISCARD;
1092
1093	if (flags & EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT) {
1094		flush_flags |= JBD2_JOURNAL_FLUSH_ZEROOUT;
1095		pr_info_ratelimited("warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow");
1096	}
1097
1098	jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1099	err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, flush_flags);
1100	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1101
1102	return err;
1103}
1104
1105static int ext4_ioctl_setlabel(struct file *filp, const char __user *user_label)
1106{
1107	size_t len;
1108	int ret = 0;
1109	char new_label[EXT4_LABEL_MAX + 1];
1110	struct super_block *sb = file_inode(filp)->i_sb;
1111
1112	if (!capable(CAP_SYS_ADMIN))
1113		return -EPERM;
1114
1115	/*
1116	 * Copy the maximum length allowed for ext4 label with one more to
1117	 * find the required terminating null byte in order to test the
1118	 * label length. The on disk label doesn't need to be null terminated.
1119	 */
1120	if (copy_from_user(new_label, user_label, EXT4_LABEL_MAX + 1))
1121		return -EFAULT;
1122
1123	len = strnlen(new_label, EXT4_LABEL_MAX + 1);
1124	if (len > EXT4_LABEL_MAX)
1125		return -EINVAL;
1126
1127	/*
1128	 * Clear the buffer after the new label
1129	 */
1130	memset(new_label + len, 0, EXT4_LABEL_MAX - len);
1131
1132	ret = mnt_want_write_file(filp);
1133	if (ret)
1134		return ret;
1135
1136	ret = ext4_update_superblocks_fn(sb, ext4_sb_setlabel, new_label);
1137
1138	mnt_drop_write_file(filp);
1139	return ret;
1140}
1141
1142static int ext4_ioctl_getlabel(struct ext4_sb_info *sbi, char __user *user_label)
1143{
1144	char label[EXT4_LABEL_MAX + 1];
1145
1146	/*
1147	 * EXT4_LABEL_MAX must always be smaller than FSLABEL_MAX because
1148	 * FSLABEL_MAX must include terminating null byte, while s_volume_name
1149	 * does not have to.
1150	 */
1151	BUILD_BUG_ON(EXT4_LABEL_MAX >= FSLABEL_MAX);
1152
1153	memset(label, 0, sizeof(label));
1154	lock_buffer(sbi->s_sbh);
1155	strncpy(label, sbi->s_es->s_volume_name, EXT4_LABEL_MAX);
1156	unlock_buffer(sbi->s_sbh);
1157
1158	if (copy_to_user(user_label, label, sizeof(label)))
1159		return -EFAULT;
1160	return 0;
1161}
1162
1163static int ext4_ioctl_getuuid(struct ext4_sb_info *sbi,
1164			struct fsuuid __user *ufsuuid)
1165{
1166	struct fsuuid fsuuid;
1167	__u8 uuid[UUID_SIZE];
1168
1169	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1170		return -EFAULT;
1171
1172	if (fsuuid.fsu_len == 0) {
1173		fsuuid.fsu_len = UUID_SIZE;
1174		if (copy_to_user(&ufsuuid->fsu_len, &fsuuid.fsu_len,
1175					sizeof(fsuuid.fsu_len)))
1176			return -EFAULT;
1177		return 0;
1178	}
1179
1180	if (fsuuid.fsu_len < UUID_SIZE || fsuuid.fsu_flags != 0)
1181		return -EINVAL;
1182
1183	lock_buffer(sbi->s_sbh);
1184	memcpy(uuid, sbi->s_es->s_uuid, UUID_SIZE);
1185	unlock_buffer(sbi->s_sbh);
1186
1187	fsuuid.fsu_len = UUID_SIZE;
1188	if (copy_to_user(ufsuuid, &fsuuid, sizeof(fsuuid)) ||
1189	    copy_to_user(&ufsuuid->fsu_uuid[0], uuid, UUID_SIZE))
1190		return -EFAULT;
1191	return 0;
1192}
1193
1194static int ext4_ioctl_setuuid(struct file *filp,
1195			const struct fsuuid __user *ufsuuid)
1196{
1197	int ret = 0;
1198	struct super_block *sb = file_inode(filp)->i_sb;
1199	struct fsuuid fsuuid;
1200	__u8 uuid[UUID_SIZE];
1201
1202	if (!capable(CAP_SYS_ADMIN))
1203		return -EPERM;
1204
1205	/*
1206	 * If any checksums (group descriptors or metadata) are being used
1207	 * then the checksum seed feature is required to change the UUID.
1208	 */
1209	if (((ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb))
1210			&& !ext4_has_feature_csum_seed(sb))
1211		|| ext4_has_feature_stable_inodes(sb))
1212		return -EOPNOTSUPP;
1213
1214	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1215		return -EFAULT;
1216
1217	if (fsuuid.fsu_len != UUID_SIZE || fsuuid.fsu_flags != 0)
1218		return -EINVAL;
1219
1220	if (copy_from_user(uuid, &ufsuuid->fsu_uuid[0], UUID_SIZE))
1221		return -EFAULT;
1222
1223	ret = mnt_want_write_file(filp);
1224	if (ret)
1225		return ret;
1226
1227	ret = ext4_update_superblocks_fn(sb, ext4_sb_setuuid, &uuid);
1228	mnt_drop_write_file(filp);
1229
1230	return ret;
1231}
1232
1233static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1234{
1235	struct inode *inode = file_inode(filp);
1236	struct super_block *sb = inode->i_sb;
1237	struct mnt_idmap *idmap = file_mnt_idmap(filp);
1238
1239	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
1240
1241	switch (cmd) {
1242	case FS_IOC_GETFSMAP:
1243		return ext4_ioc_getfsmap(sb, (void __user *)arg);
1244	case EXT4_IOC_GETVERSION:
1245	case EXT4_IOC_GETVERSION_OLD:
1246		return put_user(inode->i_generation, (int __user *) arg);
1247	case EXT4_IOC_SETVERSION:
1248	case EXT4_IOC_SETVERSION_OLD: {
1249		handle_t *handle;
1250		struct ext4_iloc iloc;
1251		__u32 generation;
1252		int err;
1253
1254		if (!inode_owner_or_capable(idmap, inode))
1255			return -EPERM;
1256
1257		if (ext4_has_metadata_csum(inode->i_sb)) {
1258			ext4_warning(sb, "Setting inode version is not "
1259				     "supported with metadata_csum enabled.");
1260			return -ENOTTY;
1261		}
1262
1263		err = mnt_want_write_file(filp);
1264		if (err)
1265			return err;
1266		if (get_user(generation, (int __user *) arg)) {
1267			err = -EFAULT;
1268			goto setversion_out;
1269		}
1270
1271		inode_lock(inode);
1272		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
1273		if (IS_ERR(handle)) {
1274			err = PTR_ERR(handle);
1275			goto unlock_out;
1276		}
1277		err = ext4_reserve_inode_write(handle, inode, &iloc);
1278		if (err == 0) {
1279			inode_set_ctime_current(inode);
1280			inode_inc_iversion(inode);
1281			inode->i_generation = generation;
1282			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
1283		}
1284		ext4_journal_stop(handle);
1285
1286unlock_out:
1287		inode_unlock(inode);
1288setversion_out:
1289		mnt_drop_write_file(filp);
1290		return err;
1291	}
1292	case EXT4_IOC_GROUP_EXTEND: {
1293		ext4_fsblk_t n_blocks_count;
1294		int err, err2=0;
1295
1296		err = ext4_resize_begin(sb);
1297		if (err)
1298			return err;
1299
1300		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
1301			err = -EFAULT;
1302			goto group_extend_out;
1303		}
1304
1305		if (ext4_has_feature_bigalloc(sb)) {
1306			ext4_msg(sb, KERN_ERR,
1307				 "Online resizing not supported with bigalloc");
1308			err = -EOPNOTSUPP;
1309			goto group_extend_out;
1310		}
1311
1312		err = mnt_want_write_file(filp);
1313		if (err)
1314			goto group_extend_out;
1315
1316		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
1317		if (EXT4_SB(sb)->s_journal) {
1318			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1319			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1320			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1321		}
1322		if (err == 0)
1323			err = err2;
1324		mnt_drop_write_file(filp);
1325group_extend_out:
1326		err2 = ext4_resize_end(sb, false);
1327		if (err == 0)
1328			err = err2;
1329		return err;
1330	}
1331
1332	case EXT4_IOC_MOVE_EXT: {
1333		struct move_extent me;
1334		struct fd donor;
1335		int err;
1336
1337		if (!(filp->f_mode & FMODE_READ) ||
1338		    !(filp->f_mode & FMODE_WRITE))
1339			return -EBADF;
1340
1341		if (copy_from_user(&me,
1342			(struct move_extent __user *)arg, sizeof(me)))
1343			return -EFAULT;
1344		me.moved_len = 0;
1345
1346		donor = fdget(me.donor_fd);
1347		if (!donor.file)
1348			return -EBADF;
1349
1350		if (!(donor.file->f_mode & FMODE_WRITE)) {
1351			err = -EBADF;
1352			goto mext_out;
1353		}
1354
1355		if (ext4_has_feature_bigalloc(sb)) {
1356			ext4_msg(sb, KERN_ERR,
1357				 "Online defrag not supported with bigalloc");
1358			err = -EOPNOTSUPP;
1359			goto mext_out;
1360		} else if (IS_DAX(inode)) {
1361			ext4_msg(sb, KERN_ERR,
1362				 "Online defrag not supported with DAX");
1363			err = -EOPNOTSUPP;
1364			goto mext_out;
1365		}
1366
1367		err = mnt_want_write_file(filp);
1368		if (err)
1369			goto mext_out;
1370
1371		err = ext4_move_extents(filp, donor.file, me.orig_start,
1372					me.donor_start, me.len, &me.moved_len);
1373		mnt_drop_write_file(filp);
1374
1375		if (copy_to_user((struct move_extent __user *)arg,
1376				 &me, sizeof(me)))
1377			err = -EFAULT;
1378mext_out:
1379		fdput(donor);
1380		return err;
1381	}
1382
1383	case EXT4_IOC_GROUP_ADD: {
1384		struct ext4_new_group_data input;
1385
1386		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
1387				sizeof(input)))
1388			return -EFAULT;
1389
1390		return ext4_ioctl_group_add(filp, &input);
1391	}
1392
1393	case EXT4_IOC_MIGRATE:
1394	{
1395		int err;
1396		if (!inode_owner_or_capable(idmap, inode))
1397			return -EACCES;
1398
1399		err = mnt_want_write_file(filp);
1400		if (err)
1401			return err;
1402		/*
1403		 * inode_mutex prevent write and truncate on the file.
1404		 * Read still goes through. We take i_data_sem in
1405		 * ext4_ext_swap_inode_data before we switch the
1406		 * inode format to prevent read.
1407		 */
1408		inode_lock((inode));
1409		err = ext4_ext_migrate(inode);
1410		inode_unlock((inode));
1411		mnt_drop_write_file(filp);
1412		return err;
1413	}
1414
1415	case EXT4_IOC_ALLOC_DA_BLKS:
1416	{
1417		int err;
1418		if (!inode_owner_or_capable(idmap, inode))
1419			return -EACCES;
1420
1421		err = mnt_want_write_file(filp);
1422		if (err)
1423			return err;
1424		err = ext4_alloc_da_blocks(inode);
1425		mnt_drop_write_file(filp);
1426		return err;
1427	}
1428
1429	case EXT4_IOC_SWAP_BOOT:
1430	{
1431		int err;
1432		if (!(filp->f_mode & FMODE_WRITE))
1433			return -EBADF;
1434		err = mnt_want_write_file(filp);
1435		if (err)
1436			return err;
1437		err = swap_inode_boot_loader(sb, idmap, inode);
1438		mnt_drop_write_file(filp);
1439		return err;
1440	}
1441
1442	case EXT4_IOC_RESIZE_FS: {
1443		ext4_fsblk_t n_blocks_count;
1444		int err = 0, err2 = 0;
1445		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
1446
1447		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1448				   sizeof(__u64))) {
1449			return -EFAULT;
1450		}
1451
1452		err = ext4_resize_begin(sb);
1453		if (err)
1454			return err;
1455
1456		err = mnt_want_write_file(filp);
1457		if (err)
1458			goto resizefs_out;
1459
1460		err = ext4_resize_fs(sb, n_blocks_count);
1461		if (EXT4_SB(sb)->s_journal) {
1462			ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE, NULL);
1463			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1464			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1465			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1466		}
1467		if (err == 0)
1468			err = err2;
1469		mnt_drop_write_file(filp);
1470		if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1471		    ext4_has_group_desc_csum(sb) &&
1472		    test_opt(sb, INIT_INODE_TABLE))
1473			err = ext4_register_li_request(sb, o_group);
1474
1475resizefs_out:
1476		err2 = ext4_resize_end(sb, true);
1477		if (err == 0)
1478			err = err2;
1479		return err;
1480	}
1481
1482	case FITRIM:
1483	{
1484		struct fstrim_range range;
1485		int ret = 0;
1486
1487		if (!capable(CAP_SYS_ADMIN))
1488			return -EPERM;
1489
1490		if (!bdev_max_discard_sectors(sb->s_bdev))
1491			return -EOPNOTSUPP;
1492
1493		/*
1494		 * We haven't replayed the journal, so we cannot use our
1495		 * block-bitmap-guided storage zapping commands.
1496		 */
1497		if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1498			return -EROFS;
1499
1500		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1501		    sizeof(range)))
1502			return -EFAULT;
1503
1504		ret = ext4_trim_fs(sb, &range);
1505		if (ret < 0)
1506			return ret;
1507
1508		if (copy_to_user((struct fstrim_range __user *)arg, &range,
1509		    sizeof(range)))
1510			return -EFAULT;
1511
1512		return 0;
1513	}
1514	case EXT4_IOC_PRECACHE_EXTENTS:
1515		return ext4_ext_precache(inode);
1516
1517	case FS_IOC_SET_ENCRYPTION_POLICY:
1518		if (!ext4_has_feature_encrypt(sb))
1519			return -EOPNOTSUPP;
1520		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1521
1522	case FS_IOC_GET_ENCRYPTION_PWSALT:
1523		return ext4_ioctl_get_encryption_pwsalt(filp, (void __user *)arg);
1524
1525	case FS_IOC_GET_ENCRYPTION_POLICY:
1526		if (!ext4_has_feature_encrypt(sb))
1527			return -EOPNOTSUPP;
1528		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1529
1530	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1531		if (!ext4_has_feature_encrypt(sb))
1532			return -EOPNOTSUPP;
1533		return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
1534
1535	case FS_IOC_ADD_ENCRYPTION_KEY:
1536		if (!ext4_has_feature_encrypt(sb))
1537			return -EOPNOTSUPP;
1538		return fscrypt_ioctl_add_key(filp, (void __user *)arg);
1539
1540	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1541		if (!ext4_has_feature_encrypt(sb))
1542			return -EOPNOTSUPP;
1543		return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
1544
1545	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1546		if (!ext4_has_feature_encrypt(sb))
1547			return -EOPNOTSUPP;
1548		return fscrypt_ioctl_remove_key_all_users(filp,
1549							  (void __user *)arg);
1550	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1551		if (!ext4_has_feature_encrypt(sb))
1552			return -EOPNOTSUPP;
1553		return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
1554
1555	case FS_IOC_GET_ENCRYPTION_NONCE:
1556		if (!ext4_has_feature_encrypt(sb))
1557			return -EOPNOTSUPP;
1558		return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
1559
1560	case EXT4_IOC_CLEAR_ES_CACHE:
1561	{
1562		if (!inode_owner_or_capable(idmap, inode))
1563			return -EACCES;
1564		ext4_clear_inode_es(inode);
1565		return 0;
1566	}
1567
1568	case EXT4_IOC_GETSTATE:
1569	{
1570		__u32	state = 0;
1571
1572		if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED))
1573			state |= EXT4_STATE_FLAG_EXT_PRECACHED;
1574		if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
1575			state |= EXT4_STATE_FLAG_NEW;
1576		if (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
1577			state |= EXT4_STATE_FLAG_NEWENTRY;
1578		if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE))
1579			state |= EXT4_STATE_FLAG_DA_ALLOC_CLOSE;
1580
1581		return put_user(state, (__u32 __user *) arg);
1582	}
1583
1584	case EXT4_IOC_GET_ES_CACHE:
1585		return ext4_ioctl_get_es_cache(filp, arg);
1586
1587	case EXT4_IOC_SHUTDOWN:
1588		return ext4_ioctl_shutdown(sb, arg);
1589
1590	case FS_IOC_ENABLE_VERITY:
1591		if (!ext4_has_feature_verity(sb))
1592			return -EOPNOTSUPP;
1593		return fsverity_ioctl_enable(filp, (const void __user *)arg);
1594
1595	case FS_IOC_MEASURE_VERITY:
1596		if (!ext4_has_feature_verity(sb))
1597			return -EOPNOTSUPP;
1598		return fsverity_ioctl_measure(filp, (void __user *)arg);
1599
1600	case FS_IOC_READ_VERITY_METADATA:
1601		if (!ext4_has_feature_verity(sb))
1602			return -EOPNOTSUPP;
1603		return fsverity_ioctl_read_metadata(filp,
1604						    (const void __user *)arg);
1605
1606	case EXT4_IOC_CHECKPOINT:
1607		return ext4_ioctl_checkpoint(filp, arg);
1608
1609	case FS_IOC_GETFSLABEL:
1610		return ext4_ioctl_getlabel(EXT4_SB(sb), (void __user *)arg);
1611
1612	case FS_IOC_SETFSLABEL:
1613		return ext4_ioctl_setlabel(filp,
1614					   (const void __user *)arg);
1615
1616	case EXT4_IOC_GETFSUUID:
1617		return ext4_ioctl_getuuid(EXT4_SB(sb), (void __user *)arg);
1618	case EXT4_IOC_SETFSUUID:
1619		return ext4_ioctl_setuuid(filp, (const void __user *)arg);
1620	default:
1621		return -ENOTTY;
1622	}
1623}
1624
1625long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1626{
1627	return __ext4_ioctl(filp, cmd, arg);
1628}
1629
1630#ifdef CONFIG_COMPAT
1631long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1632{
1633	/* These are just misnamed, they actually get/put from/to user an int */
1634	switch (cmd) {
1635	case EXT4_IOC32_GETVERSION:
1636		cmd = EXT4_IOC_GETVERSION;
1637		break;
1638	case EXT4_IOC32_SETVERSION:
1639		cmd = EXT4_IOC_SETVERSION;
1640		break;
1641	case EXT4_IOC32_GROUP_EXTEND:
1642		cmd = EXT4_IOC_GROUP_EXTEND;
1643		break;
1644	case EXT4_IOC32_GETVERSION_OLD:
1645		cmd = EXT4_IOC_GETVERSION_OLD;
1646		break;
1647	case EXT4_IOC32_SETVERSION_OLD:
1648		cmd = EXT4_IOC_SETVERSION_OLD;
1649		break;
1650	case EXT4_IOC32_GETRSVSZ:
1651		cmd = EXT4_IOC_GETRSVSZ;
1652		break;
1653	case EXT4_IOC32_SETRSVSZ:
1654		cmd = EXT4_IOC_SETRSVSZ;
1655		break;
1656	case EXT4_IOC32_GROUP_ADD: {
1657		struct compat_ext4_new_group_input __user *uinput;
1658		struct ext4_new_group_data input;
1659		int err;
1660
1661		uinput = compat_ptr(arg);
1662		err = get_user(input.group, &uinput->group);
1663		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1664		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1665		err |= get_user(input.inode_table, &uinput->inode_table);
1666		err |= get_user(input.blocks_count, &uinput->blocks_count);
1667		err |= get_user(input.reserved_blocks,
1668				&uinput->reserved_blocks);
1669		if (err)
1670			return -EFAULT;
1671		return ext4_ioctl_group_add(file, &input);
1672	}
1673	case EXT4_IOC_MOVE_EXT:
1674	case EXT4_IOC_RESIZE_FS:
1675	case FITRIM:
1676	case EXT4_IOC_PRECACHE_EXTENTS:
1677	case FS_IOC_SET_ENCRYPTION_POLICY:
1678	case FS_IOC_GET_ENCRYPTION_PWSALT:
1679	case FS_IOC_GET_ENCRYPTION_POLICY:
1680	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1681	case FS_IOC_ADD_ENCRYPTION_KEY:
1682	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1683	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1684	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1685	case FS_IOC_GET_ENCRYPTION_NONCE:
1686	case EXT4_IOC_SHUTDOWN:
1687	case FS_IOC_GETFSMAP:
1688	case FS_IOC_ENABLE_VERITY:
1689	case FS_IOC_MEASURE_VERITY:
1690	case FS_IOC_READ_VERITY_METADATA:
1691	case EXT4_IOC_CLEAR_ES_CACHE:
1692	case EXT4_IOC_GETSTATE:
1693	case EXT4_IOC_GET_ES_CACHE:
1694	case EXT4_IOC_CHECKPOINT:
1695	case FS_IOC_GETFSLABEL:
1696	case FS_IOC_SETFSLABEL:
1697	case EXT4_IOC_GETFSUUID:
1698	case EXT4_IOC_SETFSUUID:
1699		break;
1700	default:
1701		return -ENOIOCTLCMD;
1702	}
1703	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1704}
1705#endif
1706
1707static void set_overhead(struct ext4_super_block *es, const void *arg)
1708{
1709	es->s_overhead_clusters = cpu_to_le32(*((unsigned long *) arg));
1710}
1711
1712int ext4_update_overhead(struct super_block *sb, bool force)
1713{
1714	struct ext4_sb_info *sbi = EXT4_SB(sb);
1715
1716	if (sb_rdonly(sb))
1717		return 0;
1718	if (!force &&
1719	    (sbi->s_overhead == 0 ||
1720	     sbi->s_overhead == le32_to_cpu(sbi->s_es->s_overhead_clusters)))
1721		return 0;
1722	return ext4_update_superblocks_fn(sb, set_overhead, &sbi->s_overhead);
1723}
1724