xref: /openbmc/linux/fs/ext4/ioctl.c (revision 5b628549)
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/uuid.h>
20 #include <linux/uaccess.h>
21 #include <linux/delay.h>
22 #include <linux/iversion.h>
23 #include "ext4_jbd2.h"
24 #include "ext4.h"
25 #include <linux/fsmap.h>
26 #include "fsmap.h"
27 #include <trace/events/ext4.h>
28 
29 /**
30  * Swap memory between @a and @b for @len bytes.
31  *
32  * @a:          pointer to first memory area
33  * @b:          pointer to second memory area
34  * @len:        number of bytes to swap
35  *
36  */
37 static void memswap(void *a, void *b, size_t len)
38 {
39 	unsigned char *ap, *bp;
40 
41 	ap = (unsigned char *)a;
42 	bp = (unsigned char *)b;
43 	while (len-- > 0) {
44 		swap(*ap, *bp);
45 		ap++;
46 		bp++;
47 	}
48 }
49 
50 /**
51  * Swap i_data and associated attributes between @inode1 and @inode2.
52  * This function is used for the primary swap between inode1 and inode2
53  * and also to revert this primary swap in case of errors.
54  *
55  * Therefore you have to make sure, that calling this method twice
56  * will revert all changes.
57  *
58  * @inode1:     pointer to first inode
59  * @inode2:     pointer to second inode
60  */
61 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
62 {
63 	loff_t isize;
64 	struct ext4_inode_info *ei1;
65 	struct ext4_inode_info *ei2;
66 	unsigned long tmp;
67 
68 	ei1 = EXT4_I(inode1);
69 	ei2 = EXT4_I(inode2);
70 
71 	swap(inode1->i_version, inode2->i_version);
72 	swap(inode1->i_atime, inode2->i_atime);
73 	swap(inode1->i_mtime, inode2->i_mtime);
74 
75 	memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
76 	tmp = ei1->i_flags & EXT4_FL_SHOULD_SWAP;
77 	ei1->i_flags = (ei2->i_flags & EXT4_FL_SHOULD_SWAP) |
78 		(ei1->i_flags & ~EXT4_FL_SHOULD_SWAP);
79 	ei2->i_flags = tmp | (ei2->i_flags & ~EXT4_FL_SHOULD_SWAP);
80 	swap(ei1->i_disksize, ei2->i_disksize);
81 	ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
82 	ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
83 
84 	isize = i_size_read(inode1);
85 	i_size_write(inode1, i_size_read(inode2));
86 	i_size_write(inode2, isize);
87 }
88 
89 static void reset_inode_seed(struct inode *inode)
90 {
91 	struct ext4_inode_info *ei = EXT4_I(inode);
92 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
93 	__le32 inum = cpu_to_le32(inode->i_ino);
94 	__le32 gen = cpu_to_le32(inode->i_generation);
95 	__u32 csum;
96 
97 	if (!ext4_has_metadata_csum(inode->i_sb))
98 		return;
99 
100 	csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum, sizeof(inum));
101 	ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen, sizeof(gen));
102 }
103 
104 /**
105  * Swap the information from the given @inode and the inode
106  * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
107  * important fields of the inodes.
108  *
109  * @sb:         the super block of the filesystem
110  * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
111  *
112  */
113 static long swap_inode_boot_loader(struct super_block *sb,
114 				struct inode *inode)
115 {
116 	handle_t *handle;
117 	int err;
118 	struct inode *inode_bl;
119 	struct ext4_inode_info *ei_bl;
120 	qsize_t size, size_bl, diff;
121 	blkcnt_t blocks;
122 	unsigned short bytes;
123 
124 	inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO, EXT4_IGET_SPECIAL);
125 	if (IS_ERR(inode_bl))
126 		return PTR_ERR(inode_bl);
127 	ei_bl = EXT4_I(inode_bl);
128 
129 	/* Protect orig inodes against a truncate and make sure,
130 	 * that only 1 swap_inode_boot_loader is running. */
131 	lock_two_nondirectories(inode, inode_bl);
132 
133 	if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode) ||
134 	    IS_SWAPFILE(inode) || IS_ENCRYPTED(inode) ||
135 	    (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL) ||
136 	    ext4_has_inline_data(inode)) {
137 		err = -EINVAL;
138 		goto journal_err_out;
139 	}
140 
141 	if (IS_RDONLY(inode) || IS_APPEND(inode) || IS_IMMUTABLE(inode) ||
142 	    !inode_owner_or_capable(inode) || !capable(CAP_SYS_ADMIN)) {
143 		err = -EPERM;
144 		goto journal_err_out;
145 	}
146 
147 	down_write(&EXT4_I(inode)->i_mmap_sem);
148 	err = filemap_write_and_wait(inode->i_mapping);
149 	if (err)
150 		goto err_out;
151 
152 	err = filemap_write_and_wait(inode_bl->i_mapping);
153 	if (err)
154 		goto err_out;
155 
156 	/* Wait for all existing dio workers */
157 	inode_dio_wait(inode);
158 	inode_dio_wait(inode_bl);
159 
160 	truncate_inode_pages(&inode->i_data, 0);
161 	truncate_inode_pages(&inode_bl->i_data, 0);
162 
163 	handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
164 	if (IS_ERR(handle)) {
165 		err = -EINVAL;
166 		goto err_out;
167 	}
168 
169 	/* Protect extent tree against block allocations via delalloc */
170 	ext4_double_down_write_data_sem(inode, inode_bl);
171 
172 	if (inode_bl->i_nlink == 0) {
173 		/* this inode has never been used as a BOOT_LOADER */
174 		set_nlink(inode_bl, 1);
175 		i_uid_write(inode_bl, 0);
176 		i_gid_write(inode_bl, 0);
177 		inode_bl->i_flags = 0;
178 		ei_bl->i_flags = 0;
179 		inode_set_iversion(inode_bl, 1);
180 		i_size_write(inode_bl, 0);
181 		inode_bl->i_mode = S_IFREG;
182 		if (ext4_has_feature_extents(sb)) {
183 			ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
184 			ext4_ext_tree_init(handle, inode_bl);
185 		} else
186 			memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
187 	}
188 
189 	err = dquot_initialize(inode);
190 	if (err)
191 		goto err_out1;
192 
193 	size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
194 	size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
195 	diff = size - size_bl;
196 	swap_inode_data(inode, inode_bl);
197 
198 	inode->i_ctime = inode_bl->i_ctime = current_time(inode);
199 
200 	inode->i_generation = prandom_u32();
201 	inode_bl->i_generation = prandom_u32();
202 	reset_inode_seed(inode);
203 	reset_inode_seed(inode_bl);
204 
205 	ext4_discard_preallocations(inode);
206 
207 	err = ext4_mark_inode_dirty(handle, inode);
208 	if (err < 0) {
209 		/* No need to update quota information. */
210 		ext4_warning(inode->i_sb,
211 			"couldn't mark inode #%lu dirty (err %d)",
212 			inode->i_ino, err);
213 		/* Revert all changes: */
214 		swap_inode_data(inode, inode_bl);
215 		ext4_mark_inode_dirty(handle, inode);
216 		goto err_out1;
217 	}
218 
219 	blocks = inode_bl->i_blocks;
220 	bytes = inode_bl->i_bytes;
221 	inode_bl->i_blocks = inode->i_blocks;
222 	inode_bl->i_bytes = inode->i_bytes;
223 	err = ext4_mark_inode_dirty(handle, inode_bl);
224 	if (err < 0) {
225 		/* No need to update quota information. */
226 		ext4_warning(inode_bl->i_sb,
227 			"couldn't mark inode #%lu dirty (err %d)",
228 			inode_bl->i_ino, err);
229 		goto revert;
230 	}
231 
232 	/* Bootloader inode should not be counted into quota information. */
233 	if (diff > 0)
234 		dquot_free_space(inode, diff);
235 	else
236 		err = dquot_alloc_space(inode, -1 * diff);
237 
238 	if (err < 0) {
239 revert:
240 		/* Revert all changes: */
241 		inode_bl->i_blocks = blocks;
242 		inode_bl->i_bytes = bytes;
243 		swap_inode_data(inode, inode_bl);
244 		ext4_mark_inode_dirty(handle, inode);
245 		ext4_mark_inode_dirty(handle, inode_bl);
246 	}
247 
248 err_out1:
249 	ext4_journal_stop(handle);
250 	ext4_double_up_write_data_sem(inode, inode_bl);
251 
252 err_out:
253 	up_write(&EXT4_I(inode)->i_mmap_sem);
254 journal_err_out:
255 	unlock_two_nondirectories(inode, inode_bl);
256 	iput(inode_bl);
257 	return err;
258 }
259 
260 #ifdef CONFIG_FS_ENCRYPTION
261 static int uuid_is_zero(__u8 u[16])
262 {
263 	int	i;
264 
265 	for (i = 0; i < 16; i++)
266 		if (u[i])
267 			return 0;
268 	return 1;
269 }
270 #endif
271 
272 static int ext4_ioctl_setflags(struct inode *inode,
273 			       unsigned int flags)
274 {
275 	struct ext4_inode_info *ei = EXT4_I(inode);
276 	handle_t *handle = NULL;
277 	int err = -EPERM, migrate = 0;
278 	struct ext4_iloc iloc;
279 	unsigned int oldflags, mask, i;
280 	unsigned int jflag;
281 
282 	/* Is it quota file? Do not allow user to mess with it */
283 	if (ext4_is_quota_file(inode))
284 		goto flags_out;
285 
286 	oldflags = ei->i_flags;
287 
288 	/* The JOURNAL_DATA flag is modifiable only by root */
289 	jflag = flags & EXT4_JOURNAL_DATA_FL;
290 
291 	/*
292 	 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
293 	 * the relevant capability.
294 	 *
295 	 * This test looks nicer. Thanks to Pauline Middelink
296 	 */
297 	if ((flags ^ oldflags) & (EXT4_APPEND_FL | EXT4_IMMUTABLE_FL)) {
298 		if (!capable(CAP_LINUX_IMMUTABLE))
299 			goto flags_out;
300 	}
301 
302 	/*
303 	 * The JOURNAL_DATA flag can only be changed by
304 	 * the relevant capability.
305 	 */
306 	if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
307 		if (!capable(CAP_SYS_RESOURCE))
308 			goto flags_out;
309 	}
310 	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
311 		migrate = 1;
312 
313 	if (flags & EXT4_EOFBLOCKS_FL) {
314 		/* we don't support adding EOFBLOCKS flag */
315 		if (!(oldflags & EXT4_EOFBLOCKS_FL)) {
316 			err = -EOPNOTSUPP;
317 			goto flags_out;
318 		}
319 	} else if (oldflags & EXT4_EOFBLOCKS_FL) {
320 		err = ext4_truncate(inode);
321 		if (err)
322 			goto flags_out;
323 	}
324 
325 	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
326 	if (IS_ERR(handle)) {
327 		err = PTR_ERR(handle);
328 		goto flags_out;
329 	}
330 	if (IS_SYNC(inode))
331 		ext4_handle_sync(handle);
332 	err = ext4_reserve_inode_write(handle, inode, &iloc);
333 	if (err)
334 		goto flags_err;
335 
336 	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
337 		if (!(mask & EXT4_FL_USER_MODIFIABLE))
338 			continue;
339 		/* These flags get special treatment later */
340 		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
341 			continue;
342 		if (mask & flags)
343 			ext4_set_inode_flag(inode, i);
344 		else
345 			ext4_clear_inode_flag(inode, i);
346 	}
347 
348 	ext4_set_inode_flags(inode);
349 	inode->i_ctime = current_time(inode);
350 
351 	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
352 flags_err:
353 	ext4_journal_stop(handle);
354 	if (err)
355 		goto flags_out;
356 
357 	if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
358 		/*
359 		 * Changes to the journaling mode can cause unsafe changes to
360 		 * S_DAX if we are using the DAX mount option.
361 		 */
362 		if (test_opt(inode->i_sb, DAX)) {
363 			err = -EBUSY;
364 			goto flags_out;
365 		}
366 
367 		err = ext4_change_inode_journal_flag(inode, jflag);
368 		if (err)
369 			goto flags_out;
370 	}
371 	if (migrate) {
372 		if (flags & EXT4_EXTENTS_FL)
373 			err = ext4_ext_migrate(inode);
374 		else
375 			err = ext4_ind_migrate(inode);
376 	}
377 
378 flags_out:
379 	return err;
380 }
381 
382 #ifdef CONFIG_QUOTA
383 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
384 {
385 	struct inode *inode = file_inode(filp);
386 	struct super_block *sb = inode->i_sb;
387 	struct ext4_inode_info *ei = EXT4_I(inode);
388 	int err, rc;
389 	handle_t *handle;
390 	kprojid_t kprojid;
391 	struct ext4_iloc iloc;
392 	struct ext4_inode *raw_inode;
393 	struct dquot *transfer_to[MAXQUOTAS] = { };
394 
395 	if (!ext4_has_feature_project(sb)) {
396 		if (projid != EXT4_DEF_PROJID)
397 			return -EOPNOTSUPP;
398 		else
399 			return 0;
400 	}
401 
402 	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
403 		return -EOPNOTSUPP;
404 
405 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
406 
407 	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
408 		return 0;
409 
410 	err = -EPERM;
411 	/* Is it quota file? Do not allow user to mess with it */
412 	if (ext4_is_quota_file(inode))
413 		return err;
414 
415 	err = ext4_get_inode_loc(inode, &iloc);
416 	if (err)
417 		return err;
418 
419 	raw_inode = ext4_raw_inode(&iloc);
420 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
421 		err = ext4_expand_extra_isize(inode,
422 					      EXT4_SB(sb)->s_want_extra_isize,
423 					      &iloc);
424 		if (err)
425 			return err;
426 	} else {
427 		brelse(iloc.bh);
428 	}
429 
430 	err = dquot_initialize(inode);
431 	if (err)
432 		return err;
433 
434 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
435 		EXT4_QUOTA_INIT_BLOCKS(sb) +
436 		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
437 	if (IS_ERR(handle))
438 		return PTR_ERR(handle);
439 
440 	err = ext4_reserve_inode_write(handle, inode, &iloc);
441 	if (err)
442 		goto out_stop;
443 
444 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
445 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
446 
447 		/* __dquot_transfer() calls back ext4_get_inode_usage() which
448 		 * counts xattr inode references.
449 		 */
450 		down_read(&EXT4_I(inode)->xattr_sem);
451 		err = __dquot_transfer(inode, transfer_to);
452 		up_read(&EXT4_I(inode)->xattr_sem);
453 		dqput(transfer_to[PRJQUOTA]);
454 		if (err)
455 			goto out_dirty;
456 	}
457 
458 	EXT4_I(inode)->i_projid = kprojid;
459 	inode->i_ctime = current_time(inode);
460 out_dirty:
461 	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
462 	if (!err)
463 		err = rc;
464 out_stop:
465 	ext4_journal_stop(handle);
466 	return err;
467 }
468 #else
469 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
470 {
471 	if (projid != EXT4_DEF_PROJID)
472 		return -EOPNOTSUPP;
473 	return 0;
474 }
475 #endif
476 
477 /* Transfer internal flags to xflags */
478 static inline __u32 ext4_iflags_to_xflags(unsigned long iflags)
479 {
480 	__u32 xflags = 0;
481 
482 	if (iflags & EXT4_SYNC_FL)
483 		xflags |= FS_XFLAG_SYNC;
484 	if (iflags & EXT4_IMMUTABLE_FL)
485 		xflags |= FS_XFLAG_IMMUTABLE;
486 	if (iflags & EXT4_APPEND_FL)
487 		xflags |= FS_XFLAG_APPEND;
488 	if (iflags & EXT4_NODUMP_FL)
489 		xflags |= FS_XFLAG_NODUMP;
490 	if (iflags & EXT4_NOATIME_FL)
491 		xflags |= FS_XFLAG_NOATIME;
492 	if (iflags & EXT4_PROJINHERIT_FL)
493 		xflags |= FS_XFLAG_PROJINHERIT;
494 	return xflags;
495 }
496 
497 #define EXT4_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
498 				  FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
499 				  FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
500 
501 /* Transfer xflags flags to internal */
502 static inline unsigned long ext4_xflags_to_iflags(__u32 xflags)
503 {
504 	unsigned long iflags = 0;
505 
506 	if (xflags & FS_XFLAG_SYNC)
507 		iflags |= EXT4_SYNC_FL;
508 	if (xflags & FS_XFLAG_IMMUTABLE)
509 		iflags |= EXT4_IMMUTABLE_FL;
510 	if (xflags & FS_XFLAG_APPEND)
511 		iflags |= EXT4_APPEND_FL;
512 	if (xflags & FS_XFLAG_NODUMP)
513 		iflags |= EXT4_NODUMP_FL;
514 	if (xflags & FS_XFLAG_NOATIME)
515 		iflags |= EXT4_NOATIME_FL;
516 	if (xflags & FS_XFLAG_PROJINHERIT)
517 		iflags |= EXT4_PROJINHERIT_FL;
518 
519 	return iflags;
520 }
521 
522 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
523 {
524 	struct ext4_sb_info *sbi = EXT4_SB(sb);
525 	__u32 flags;
526 
527 	if (!capable(CAP_SYS_ADMIN))
528 		return -EPERM;
529 
530 	if (get_user(flags, (__u32 __user *)arg))
531 		return -EFAULT;
532 
533 	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
534 		return -EINVAL;
535 
536 	if (ext4_forced_shutdown(sbi))
537 		return 0;
538 
539 	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
540 	trace_ext4_shutdown(sb, flags);
541 
542 	switch (flags) {
543 	case EXT4_GOING_FLAGS_DEFAULT:
544 		freeze_bdev(sb->s_bdev);
545 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
546 		thaw_bdev(sb->s_bdev, sb);
547 		break;
548 	case EXT4_GOING_FLAGS_LOGFLUSH:
549 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
550 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
551 			(void) ext4_force_commit(sb);
552 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
553 		}
554 		break;
555 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
556 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
557 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
558 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
559 		break;
560 	default:
561 		return -EINVAL;
562 	}
563 	clear_opt(sb, DISCARD);
564 	return 0;
565 }
566 
567 struct getfsmap_info {
568 	struct super_block	*gi_sb;
569 	struct fsmap_head __user *gi_data;
570 	unsigned int		gi_idx;
571 	__u32			gi_last_flags;
572 };
573 
574 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
575 {
576 	struct getfsmap_info *info = priv;
577 	struct fsmap fm;
578 
579 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
580 
581 	info->gi_last_flags = xfm->fmr_flags;
582 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
583 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
584 			sizeof(struct fsmap)))
585 		return -EFAULT;
586 
587 	return 0;
588 }
589 
590 static int ext4_ioc_getfsmap(struct super_block *sb,
591 			     struct fsmap_head __user *arg)
592 {
593 	struct getfsmap_info info = {0};
594 	struct ext4_fsmap_head xhead = {0};
595 	struct fsmap_head head;
596 	bool aborted = false;
597 	int error;
598 
599 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
600 		return -EFAULT;
601 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
602 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
603 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
604 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
605 		       sizeof(head.fmh_keys[1].fmr_reserved)))
606 		return -EINVAL;
607 	/*
608 	 * ext4 doesn't report file extents at all, so the only valid
609 	 * file offsets are the magic ones (all zeroes or all ones).
610 	 */
611 	if (head.fmh_keys[0].fmr_offset ||
612 	    (head.fmh_keys[1].fmr_offset != 0 &&
613 	     head.fmh_keys[1].fmr_offset != -1ULL))
614 		return -EINVAL;
615 
616 	xhead.fmh_iflags = head.fmh_iflags;
617 	xhead.fmh_count = head.fmh_count;
618 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
619 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
620 
621 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
622 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
623 
624 	info.gi_sb = sb;
625 	info.gi_data = arg;
626 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
627 	if (error == EXT4_QUERY_RANGE_ABORT) {
628 		error = 0;
629 		aborted = true;
630 	} else if (error)
631 		return error;
632 
633 	/* If we didn't abort, set the "last" flag in the last fmx */
634 	if (!aborted && info.gi_idx) {
635 		info.gi_last_flags |= FMR_OF_LAST;
636 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
637 				 &info.gi_last_flags,
638 				 sizeof(info.gi_last_flags)))
639 			return -EFAULT;
640 	}
641 
642 	/* copy back header */
643 	head.fmh_entries = xhead.fmh_entries;
644 	head.fmh_oflags = xhead.fmh_oflags;
645 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
646 		return -EFAULT;
647 
648 	return 0;
649 }
650 
651 static long ext4_ioctl_group_add(struct file *file,
652 				 struct ext4_new_group_data *input)
653 {
654 	struct super_block *sb = file_inode(file)->i_sb;
655 	int err, err2=0;
656 
657 	err = ext4_resize_begin(sb);
658 	if (err)
659 		return err;
660 
661 	if (ext4_has_feature_bigalloc(sb)) {
662 		ext4_msg(sb, KERN_ERR,
663 			 "Online resizing not supported with bigalloc");
664 		err = -EOPNOTSUPP;
665 		goto group_add_out;
666 	}
667 
668 	err = mnt_want_write_file(file);
669 	if (err)
670 		goto group_add_out;
671 
672 	err = ext4_group_add(sb, input);
673 	if (EXT4_SB(sb)->s_journal) {
674 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
675 		err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
676 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
677 	}
678 	if (err == 0)
679 		err = err2;
680 	mnt_drop_write_file(file);
681 	if (!err && ext4_has_group_desc_csum(sb) &&
682 	    test_opt(sb, INIT_INODE_TABLE))
683 		err = ext4_register_li_request(sb, input->group);
684 group_add_out:
685 	ext4_resize_end(sb);
686 	return err;
687 }
688 
689 static int ext4_ioctl_check_project(struct inode *inode, struct fsxattr *fa)
690 {
691 	/*
692 	 * Project Quota ID state is only allowed to change from within the init
693 	 * namespace. Enforce that restriction only if we are trying to change
694 	 * the quota ID state. Everything else is allowed in user namespaces.
695 	 */
696 	if (current_user_ns() == &init_user_ns)
697 		return 0;
698 
699 	if (__kprojid_val(EXT4_I(inode)->i_projid) != fa->fsx_projid)
700 		return -EINVAL;
701 
702 	if (ext4_test_inode_flag(inode, EXT4_INODE_PROJINHERIT)) {
703 		if (!(fa->fsx_xflags & FS_XFLAG_PROJINHERIT))
704 			return -EINVAL;
705 	} else {
706 		if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT)
707 			return -EINVAL;
708 	}
709 
710 	return 0;
711 }
712 
713 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
714 {
715 	struct inode *inode = file_inode(filp);
716 	struct super_block *sb = inode->i_sb;
717 	struct ext4_inode_info *ei = EXT4_I(inode);
718 	unsigned int flags;
719 
720 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
721 
722 	switch (cmd) {
723 	case FS_IOC_GETFSMAP:
724 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
725 	case EXT4_IOC_GETFLAGS:
726 		flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
727 		return put_user(flags, (int __user *) arg);
728 	case EXT4_IOC_SETFLAGS: {
729 		int err;
730 
731 		if (!inode_owner_or_capable(inode))
732 			return -EACCES;
733 
734 		if (get_user(flags, (int __user *) arg))
735 			return -EFAULT;
736 
737 		if (flags & ~EXT4_FL_USER_VISIBLE)
738 			return -EOPNOTSUPP;
739 		/*
740 		 * chattr(1) grabs flags via GETFLAGS, modifies the result and
741 		 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
742 		 * more restrictive than just silently masking off visible but
743 		 * not settable flags as we always did.
744 		 */
745 		flags &= EXT4_FL_USER_MODIFIABLE;
746 		if (ext4_mask_flags(inode->i_mode, flags) != flags)
747 			return -EOPNOTSUPP;
748 
749 		err = mnt_want_write_file(filp);
750 		if (err)
751 			return err;
752 
753 		inode_lock(inode);
754 		err = ext4_ioctl_setflags(inode, flags);
755 		inode_unlock(inode);
756 		mnt_drop_write_file(filp);
757 		return err;
758 	}
759 	case EXT4_IOC_GETVERSION:
760 	case EXT4_IOC_GETVERSION_OLD:
761 		return put_user(inode->i_generation, (int __user *) arg);
762 	case EXT4_IOC_SETVERSION:
763 	case EXT4_IOC_SETVERSION_OLD: {
764 		handle_t *handle;
765 		struct ext4_iloc iloc;
766 		__u32 generation;
767 		int err;
768 
769 		if (!inode_owner_or_capable(inode))
770 			return -EPERM;
771 
772 		if (ext4_has_metadata_csum(inode->i_sb)) {
773 			ext4_warning(sb, "Setting inode version is not "
774 				     "supported with metadata_csum enabled.");
775 			return -ENOTTY;
776 		}
777 
778 		err = mnt_want_write_file(filp);
779 		if (err)
780 			return err;
781 		if (get_user(generation, (int __user *) arg)) {
782 			err = -EFAULT;
783 			goto setversion_out;
784 		}
785 
786 		inode_lock(inode);
787 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
788 		if (IS_ERR(handle)) {
789 			err = PTR_ERR(handle);
790 			goto unlock_out;
791 		}
792 		err = ext4_reserve_inode_write(handle, inode, &iloc);
793 		if (err == 0) {
794 			inode->i_ctime = current_time(inode);
795 			inode->i_generation = generation;
796 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
797 		}
798 		ext4_journal_stop(handle);
799 
800 unlock_out:
801 		inode_unlock(inode);
802 setversion_out:
803 		mnt_drop_write_file(filp);
804 		return err;
805 	}
806 	case EXT4_IOC_GROUP_EXTEND: {
807 		ext4_fsblk_t n_blocks_count;
808 		int err, err2=0;
809 
810 		err = ext4_resize_begin(sb);
811 		if (err)
812 			return err;
813 
814 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
815 			err = -EFAULT;
816 			goto group_extend_out;
817 		}
818 
819 		if (ext4_has_feature_bigalloc(sb)) {
820 			ext4_msg(sb, KERN_ERR,
821 				 "Online resizing not supported with bigalloc");
822 			err = -EOPNOTSUPP;
823 			goto group_extend_out;
824 		}
825 
826 		err = mnt_want_write_file(filp);
827 		if (err)
828 			goto group_extend_out;
829 
830 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
831 		if (EXT4_SB(sb)->s_journal) {
832 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
833 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
834 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
835 		}
836 		if (err == 0)
837 			err = err2;
838 		mnt_drop_write_file(filp);
839 group_extend_out:
840 		ext4_resize_end(sb);
841 		return err;
842 	}
843 
844 	case EXT4_IOC_MOVE_EXT: {
845 		struct move_extent me;
846 		struct fd donor;
847 		int err;
848 
849 		if (!(filp->f_mode & FMODE_READ) ||
850 		    !(filp->f_mode & FMODE_WRITE))
851 			return -EBADF;
852 
853 		if (copy_from_user(&me,
854 			(struct move_extent __user *)arg, sizeof(me)))
855 			return -EFAULT;
856 		me.moved_len = 0;
857 
858 		donor = fdget(me.donor_fd);
859 		if (!donor.file)
860 			return -EBADF;
861 
862 		if (!(donor.file->f_mode & FMODE_WRITE)) {
863 			err = -EBADF;
864 			goto mext_out;
865 		}
866 
867 		if (ext4_has_feature_bigalloc(sb)) {
868 			ext4_msg(sb, KERN_ERR,
869 				 "Online defrag not supported with bigalloc");
870 			err = -EOPNOTSUPP;
871 			goto mext_out;
872 		} else if (IS_DAX(inode)) {
873 			ext4_msg(sb, KERN_ERR,
874 				 "Online defrag not supported with DAX");
875 			err = -EOPNOTSUPP;
876 			goto mext_out;
877 		}
878 
879 		err = mnt_want_write_file(filp);
880 		if (err)
881 			goto mext_out;
882 
883 		err = ext4_move_extents(filp, donor.file, me.orig_start,
884 					me.donor_start, me.len, &me.moved_len);
885 		mnt_drop_write_file(filp);
886 
887 		if (copy_to_user((struct move_extent __user *)arg,
888 				 &me, sizeof(me)))
889 			err = -EFAULT;
890 mext_out:
891 		fdput(donor);
892 		return err;
893 	}
894 
895 	case EXT4_IOC_GROUP_ADD: {
896 		struct ext4_new_group_data input;
897 
898 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
899 				sizeof(input)))
900 			return -EFAULT;
901 
902 		return ext4_ioctl_group_add(filp, &input);
903 	}
904 
905 	case EXT4_IOC_MIGRATE:
906 	{
907 		int err;
908 		if (!inode_owner_or_capable(inode))
909 			return -EACCES;
910 
911 		err = mnt_want_write_file(filp);
912 		if (err)
913 			return err;
914 		/*
915 		 * inode_mutex prevent write and truncate on the file.
916 		 * Read still goes through. We take i_data_sem in
917 		 * ext4_ext_swap_inode_data before we switch the
918 		 * inode format to prevent read.
919 		 */
920 		inode_lock((inode));
921 		err = ext4_ext_migrate(inode);
922 		inode_unlock((inode));
923 		mnt_drop_write_file(filp);
924 		return err;
925 	}
926 
927 	case EXT4_IOC_ALLOC_DA_BLKS:
928 	{
929 		int err;
930 		if (!inode_owner_or_capable(inode))
931 			return -EACCES;
932 
933 		err = mnt_want_write_file(filp);
934 		if (err)
935 			return err;
936 		err = ext4_alloc_da_blocks(inode);
937 		mnt_drop_write_file(filp);
938 		return err;
939 	}
940 
941 	case EXT4_IOC_SWAP_BOOT:
942 	{
943 		int err;
944 		if (!(filp->f_mode & FMODE_WRITE))
945 			return -EBADF;
946 		err = mnt_want_write_file(filp);
947 		if (err)
948 			return err;
949 		err = swap_inode_boot_loader(sb, inode);
950 		mnt_drop_write_file(filp);
951 		return err;
952 	}
953 
954 	case EXT4_IOC_RESIZE_FS: {
955 		ext4_fsblk_t n_blocks_count;
956 		int err = 0, err2 = 0;
957 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
958 
959 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
960 				   sizeof(__u64))) {
961 			return -EFAULT;
962 		}
963 
964 		err = ext4_resize_begin(sb);
965 		if (err)
966 			return err;
967 
968 		err = mnt_want_write_file(filp);
969 		if (err)
970 			goto resizefs_out;
971 
972 		err = ext4_resize_fs(sb, n_blocks_count);
973 		if (EXT4_SB(sb)->s_journal) {
974 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
975 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
976 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
977 		}
978 		if (err == 0)
979 			err = err2;
980 		mnt_drop_write_file(filp);
981 		if (!err && (o_group > EXT4_SB(sb)->s_groups_count) &&
982 		    ext4_has_group_desc_csum(sb) &&
983 		    test_opt(sb, INIT_INODE_TABLE))
984 			err = ext4_register_li_request(sb, o_group);
985 
986 resizefs_out:
987 		ext4_resize_end(sb);
988 		return err;
989 	}
990 
991 	case FITRIM:
992 	{
993 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
994 		struct fstrim_range range;
995 		int ret = 0;
996 
997 		if (!capable(CAP_SYS_ADMIN))
998 			return -EPERM;
999 
1000 		if (!blk_queue_discard(q))
1001 			return -EOPNOTSUPP;
1002 
1003 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1004 		    sizeof(range)))
1005 			return -EFAULT;
1006 
1007 		range.minlen = max((unsigned int)range.minlen,
1008 				   q->limits.discard_granularity);
1009 		ret = ext4_trim_fs(sb, &range);
1010 		if (ret < 0)
1011 			return ret;
1012 
1013 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
1014 		    sizeof(range)))
1015 			return -EFAULT;
1016 
1017 		return 0;
1018 	}
1019 	case EXT4_IOC_PRECACHE_EXTENTS:
1020 		return ext4_ext_precache(inode);
1021 
1022 	case EXT4_IOC_SET_ENCRYPTION_POLICY:
1023 		if (!ext4_has_feature_encrypt(sb))
1024 			return -EOPNOTSUPP;
1025 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1026 
1027 	case EXT4_IOC_GET_ENCRYPTION_PWSALT: {
1028 #ifdef CONFIG_FS_ENCRYPTION
1029 		int err, err2;
1030 		struct ext4_sb_info *sbi = EXT4_SB(sb);
1031 		handle_t *handle;
1032 
1033 		if (!ext4_has_feature_encrypt(sb))
1034 			return -EOPNOTSUPP;
1035 		if (uuid_is_zero(sbi->s_es->s_encrypt_pw_salt)) {
1036 			err = mnt_want_write_file(filp);
1037 			if (err)
1038 				return err;
1039 			handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
1040 			if (IS_ERR(handle)) {
1041 				err = PTR_ERR(handle);
1042 				goto pwsalt_err_exit;
1043 			}
1044 			err = ext4_journal_get_write_access(handle, sbi->s_sbh);
1045 			if (err)
1046 				goto pwsalt_err_journal;
1047 			generate_random_uuid(sbi->s_es->s_encrypt_pw_salt);
1048 			err = ext4_handle_dirty_metadata(handle, NULL,
1049 							 sbi->s_sbh);
1050 		pwsalt_err_journal:
1051 			err2 = ext4_journal_stop(handle);
1052 			if (err2 && !err)
1053 				err = err2;
1054 		pwsalt_err_exit:
1055 			mnt_drop_write_file(filp);
1056 			if (err)
1057 				return err;
1058 		}
1059 		if (copy_to_user((void __user *) arg,
1060 				 sbi->s_es->s_encrypt_pw_salt, 16))
1061 			return -EFAULT;
1062 		return 0;
1063 #else
1064 		return -EOPNOTSUPP;
1065 #endif
1066 	}
1067 	case EXT4_IOC_GET_ENCRYPTION_POLICY:
1068 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1069 
1070 	case EXT4_IOC_FSGETXATTR:
1071 	{
1072 		struct fsxattr fa;
1073 
1074 		memset(&fa, 0, sizeof(struct fsxattr));
1075 		fa.fsx_xflags = ext4_iflags_to_xflags(ei->i_flags & EXT4_FL_USER_VISIBLE);
1076 
1077 		if (ext4_has_feature_project(inode->i_sb)) {
1078 			fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
1079 				EXT4_I(inode)->i_projid);
1080 		}
1081 
1082 		if (copy_to_user((struct fsxattr __user *)arg,
1083 				 &fa, sizeof(fa)))
1084 			return -EFAULT;
1085 		return 0;
1086 	}
1087 	case EXT4_IOC_FSSETXATTR:
1088 	{
1089 		struct fsxattr fa;
1090 		int err;
1091 
1092 		if (copy_from_user(&fa, (struct fsxattr __user *)arg,
1093 				   sizeof(fa)))
1094 			return -EFAULT;
1095 
1096 		/* Make sure caller has proper permission */
1097 		if (!inode_owner_or_capable(inode))
1098 			return -EACCES;
1099 
1100 		if (fa.fsx_xflags & ~EXT4_SUPPORTED_FS_XFLAGS)
1101 			return -EOPNOTSUPP;
1102 
1103 		flags = ext4_xflags_to_iflags(fa.fsx_xflags);
1104 		if (ext4_mask_flags(inode->i_mode, flags) != flags)
1105 			return -EOPNOTSUPP;
1106 
1107 		err = mnt_want_write_file(filp);
1108 		if (err)
1109 			return err;
1110 
1111 		inode_lock(inode);
1112 		err = ext4_ioctl_check_project(inode, &fa);
1113 		if (err)
1114 			goto out;
1115 		flags = (ei->i_flags & ~EXT4_FL_XFLAG_VISIBLE) |
1116 			 (flags & EXT4_FL_XFLAG_VISIBLE);
1117 		err = ext4_ioctl_setflags(inode, flags);
1118 		if (err)
1119 			goto out;
1120 		err = ext4_ioctl_setproject(filp, fa.fsx_projid);
1121 out:
1122 		inode_unlock(inode);
1123 		mnt_drop_write_file(filp);
1124 		return err;
1125 	}
1126 	case EXT4_IOC_SHUTDOWN:
1127 		return ext4_shutdown(sb, arg);
1128 	default:
1129 		return -ENOTTY;
1130 	}
1131 }
1132 
1133 #ifdef CONFIG_COMPAT
1134 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1135 {
1136 	/* These are just misnamed, they actually get/put from/to user an int */
1137 	switch (cmd) {
1138 	case EXT4_IOC32_GETFLAGS:
1139 		cmd = EXT4_IOC_GETFLAGS;
1140 		break;
1141 	case EXT4_IOC32_SETFLAGS:
1142 		cmd = EXT4_IOC_SETFLAGS;
1143 		break;
1144 	case EXT4_IOC32_GETVERSION:
1145 		cmd = EXT4_IOC_GETVERSION;
1146 		break;
1147 	case EXT4_IOC32_SETVERSION:
1148 		cmd = EXT4_IOC_SETVERSION;
1149 		break;
1150 	case EXT4_IOC32_GROUP_EXTEND:
1151 		cmd = EXT4_IOC_GROUP_EXTEND;
1152 		break;
1153 	case EXT4_IOC32_GETVERSION_OLD:
1154 		cmd = EXT4_IOC_GETVERSION_OLD;
1155 		break;
1156 	case EXT4_IOC32_SETVERSION_OLD:
1157 		cmd = EXT4_IOC_SETVERSION_OLD;
1158 		break;
1159 	case EXT4_IOC32_GETRSVSZ:
1160 		cmd = EXT4_IOC_GETRSVSZ;
1161 		break;
1162 	case EXT4_IOC32_SETRSVSZ:
1163 		cmd = EXT4_IOC_SETRSVSZ;
1164 		break;
1165 	case EXT4_IOC32_GROUP_ADD: {
1166 		struct compat_ext4_new_group_input __user *uinput;
1167 		struct ext4_new_group_data input;
1168 		int err;
1169 
1170 		uinput = compat_ptr(arg);
1171 		err = get_user(input.group, &uinput->group);
1172 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1173 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1174 		err |= get_user(input.inode_table, &uinput->inode_table);
1175 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1176 		err |= get_user(input.reserved_blocks,
1177 				&uinput->reserved_blocks);
1178 		if (err)
1179 			return -EFAULT;
1180 		return ext4_ioctl_group_add(file, &input);
1181 	}
1182 	case EXT4_IOC_MOVE_EXT:
1183 	case EXT4_IOC_RESIZE_FS:
1184 	case EXT4_IOC_PRECACHE_EXTENTS:
1185 	case EXT4_IOC_SET_ENCRYPTION_POLICY:
1186 	case EXT4_IOC_GET_ENCRYPTION_PWSALT:
1187 	case EXT4_IOC_GET_ENCRYPTION_POLICY:
1188 	case EXT4_IOC_SHUTDOWN:
1189 	case FS_IOC_GETFSMAP:
1190 		break;
1191 	default:
1192 		return -ENOIOCTLCMD;
1193 	}
1194 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1195 }
1196 #endif
1197