xref: /openbmc/linux/fs/ext4/ioctl.c (revision d6e0cbb1)
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 /*
273  * If immutable is set and we are not clearing it, we're not allowed to change
274  * anything else in the inode.  Don't error out if we're only trying to set
275  * immutable on an immutable file.
276  */
277 static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
278 				      unsigned int flags)
279 {
280 	struct ext4_inode_info *ei = EXT4_I(inode);
281 	unsigned int oldflags = ei->i_flags;
282 
283 	if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
284 		return 0;
285 
286 	if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
287 		return -EPERM;
288 	if (ext4_has_feature_project(inode->i_sb) &&
289 	    __kprojid_val(ei->i_projid) != new_projid)
290 		return -EPERM;
291 
292 	return 0;
293 }
294 
295 static int ext4_ioctl_setflags(struct inode *inode,
296 			       unsigned int flags)
297 {
298 	struct ext4_inode_info *ei = EXT4_I(inode);
299 	handle_t *handle = NULL;
300 	int err = -EPERM, migrate = 0;
301 	struct ext4_iloc iloc;
302 	unsigned int oldflags, mask, i;
303 	unsigned int jflag;
304 	struct super_block *sb = inode->i_sb;
305 
306 	/* Is it quota file? Do not allow user to mess with it */
307 	if (ext4_is_quota_file(inode))
308 		goto flags_out;
309 
310 	oldflags = ei->i_flags;
311 
312 	/* The JOURNAL_DATA flag is modifiable only by root */
313 	jflag = flags & EXT4_JOURNAL_DATA_FL;
314 
315 	err = vfs_ioc_setflags_prepare(inode, oldflags, flags);
316 	if (err)
317 		goto flags_out;
318 
319 	/*
320 	 * The JOURNAL_DATA flag can only be changed by
321 	 * the relevant capability.
322 	 */
323 	if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
324 		if (!capable(CAP_SYS_RESOURCE))
325 			goto flags_out;
326 	}
327 	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
328 		migrate = 1;
329 
330 	if (flags & EXT4_EOFBLOCKS_FL) {
331 		/* we don't support adding EOFBLOCKS flag */
332 		if (!(oldflags & EXT4_EOFBLOCKS_FL)) {
333 			err = -EOPNOTSUPP;
334 			goto flags_out;
335 		}
336 	} else if (oldflags & EXT4_EOFBLOCKS_FL) {
337 		err = ext4_truncate(inode);
338 		if (err)
339 			goto flags_out;
340 	}
341 
342 	if ((flags ^ oldflags) & EXT4_CASEFOLD_FL) {
343 		if (!ext4_has_feature_casefold(sb)) {
344 			err = -EOPNOTSUPP;
345 			goto flags_out;
346 		}
347 
348 		if (!S_ISDIR(inode->i_mode)) {
349 			err = -ENOTDIR;
350 			goto flags_out;
351 		}
352 
353 		if (!ext4_empty_dir(inode)) {
354 			err = -ENOTEMPTY;
355 			goto flags_out;
356 		}
357 	}
358 
359 	/*
360 	 * Wait for all pending directio and then flush all the dirty pages
361 	 * for this file.  The flush marks all the pages readonly, so any
362 	 * subsequent attempt to write to the file (particularly mmap pages)
363 	 * will come through the filesystem and fail.
364 	 */
365 	if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
366 	    (flags & EXT4_IMMUTABLE_FL)) {
367 		inode_dio_wait(inode);
368 		err = filemap_write_and_wait(inode->i_mapping);
369 		if (err)
370 			goto flags_out;
371 	}
372 
373 	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
374 	if (IS_ERR(handle)) {
375 		err = PTR_ERR(handle);
376 		goto flags_out;
377 	}
378 	if (IS_SYNC(inode))
379 		ext4_handle_sync(handle);
380 	err = ext4_reserve_inode_write(handle, inode, &iloc);
381 	if (err)
382 		goto flags_err;
383 
384 	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
385 		if (!(mask & EXT4_FL_USER_MODIFIABLE))
386 			continue;
387 		/* These flags get special treatment later */
388 		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
389 			continue;
390 		if (mask & flags)
391 			ext4_set_inode_flag(inode, i);
392 		else
393 			ext4_clear_inode_flag(inode, i);
394 	}
395 
396 	ext4_set_inode_flags(inode);
397 	inode->i_ctime = current_time(inode);
398 
399 	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
400 flags_err:
401 	ext4_journal_stop(handle);
402 	if (err)
403 		goto flags_out;
404 
405 	if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
406 		/*
407 		 * Changes to the journaling mode can cause unsafe changes to
408 		 * S_DAX if we are using the DAX mount option.
409 		 */
410 		if (test_opt(inode->i_sb, DAX)) {
411 			err = -EBUSY;
412 			goto flags_out;
413 		}
414 
415 		err = ext4_change_inode_journal_flag(inode, jflag);
416 		if (err)
417 			goto flags_out;
418 	}
419 	if (migrate) {
420 		if (flags & EXT4_EXTENTS_FL)
421 			err = ext4_ext_migrate(inode);
422 		else
423 			err = ext4_ind_migrate(inode);
424 	}
425 
426 flags_out:
427 	return err;
428 }
429 
430 #ifdef CONFIG_QUOTA
431 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
432 {
433 	struct inode *inode = file_inode(filp);
434 	struct super_block *sb = inode->i_sb;
435 	struct ext4_inode_info *ei = EXT4_I(inode);
436 	int err, rc;
437 	handle_t *handle;
438 	kprojid_t kprojid;
439 	struct ext4_iloc iloc;
440 	struct ext4_inode *raw_inode;
441 	struct dquot *transfer_to[MAXQUOTAS] = { };
442 
443 	if (!ext4_has_feature_project(sb)) {
444 		if (projid != EXT4_DEF_PROJID)
445 			return -EOPNOTSUPP;
446 		else
447 			return 0;
448 	}
449 
450 	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
451 		return -EOPNOTSUPP;
452 
453 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
454 
455 	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
456 		return 0;
457 
458 	err = -EPERM;
459 	/* Is it quota file? Do not allow user to mess with it */
460 	if (ext4_is_quota_file(inode))
461 		return err;
462 
463 	err = ext4_get_inode_loc(inode, &iloc);
464 	if (err)
465 		return err;
466 
467 	raw_inode = ext4_raw_inode(&iloc);
468 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
469 		err = ext4_expand_extra_isize(inode,
470 					      EXT4_SB(sb)->s_want_extra_isize,
471 					      &iloc);
472 		if (err)
473 			return err;
474 	} else {
475 		brelse(iloc.bh);
476 	}
477 
478 	err = dquot_initialize(inode);
479 	if (err)
480 		return err;
481 
482 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
483 		EXT4_QUOTA_INIT_BLOCKS(sb) +
484 		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
485 	if (IS_ERR(handle))
486 		return PTR_ERR(handle);
487 
488 	err = ext4_reserve_inode_write(handle, inode, &iloc);
489 	if (err)
490 		goto out_stop;
491 
492 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
493 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
494 
495 		/* __dquot_transfer() calls back ext4_get_inode_usage() which
496 		 * counts xattr inode references.
497 		 */
498 		down_read(&EXT4_I(inode)->xattr_sem);
499 		err = __dquot_transfer(inode, transfer_to);
500 		up_read(&EXT4_I(inode)->xattr_sem);
501 		dqput(transfer_to[PRJQUOTA]);
502 		if (err)
503 			goto out_dirty;
504 	}
505 
506 	EXT4_I(inode)->i_projid = kprojid;
507 	inode->i_ctime = current_time(inode);
508 out_dirty:
509 	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
510 	if (!err)
511 		err = rc;
512 out_stop:
513 	ext4_journal_stop(handle);
514 	return err;
515 }
516 #else
517 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
518 {
519 	if (projid != EXT4_DEF_PROJID)
520 		return -EOPNOTSUPP;
521 	return 0;
522 }
523 #endif
524 
525 /* Transfer internal flags to xflags */
526 static inline __u32 ext4_iflags_to_xflags(unsigned long iflags)
527 {
528 	__u32 xflags = 0;
529 
530 	if (iflags & EXT4_SYNC_FL)
531 		xflags |= FS_XFLAG_SYNC;
532 	if (iflags & EXT4_IMMUTABLE_FL)
533 		xflags |= FS_XFLAG_IMMUTABLE;
534 	if (iflags & EXT4_APPEND_FL)
535 		xflags |= FS_XFLAG_APPEND;
536 	if (iflags & EXT4_NODUMP_FL)
537 		xflags |= FS_XFLAG_NODUMP;
538 	if (iflags & EXT4_NOATIME_FL)
539 		xflags |= FS_XFLAG_NOATIME;
540 	if (iflags & EXT4_PROJINHERIT_FL)
541 		xflags |= FS_XFLAG_PROJINHERIT;
542 	return xflags;
543 }
544 
545 #define EXT4_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
546 				  FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
547 				  FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
548 
549 /* Transfer xflags flags to internal */
550 static inline unsigned long ext4_xflags_to_iflags(__u32 xflags)
551 {
552 	unsigned long iflags = 0;
553 
554 	if (xflags & FS_XFLAG_SYNC)
555 		iflags |= EXT4_SYNC_FL;
556 	if (xflags & FS_XFLAG_IMMUTABLE)
557 		iflags |= EXT4_IMMUTABLE_FL;
558 	if (xflags & FS_XFLAG_APPEND)
559 		iflags |= EXT4_APPEND_FL;
560 	if (xflags & FS_XFLAG_NODUMP)
561 		iflags |= EXT4_NODUMP_FL;
562 	if (xflags & FS_XFLAG_NOATIME)
563 		iflags |= EXT4_NOATIME_FL;
564 	if (xflags & FS_XFLAG_PROJINHERIT)
565 		iflags |= EXT4_PROJINHERIT_FL;
566 
567 	return iflags;
568 }
569 
570 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
571 {
572 	struct ext4_sb_info *sbi = EXT4_SB(sb);
573 	__u32 flags;
574 
575 	if (!capable(CAP_SYS_ADMIN))
576 		return -EPERM;
577 
578 	if (get_user(flags, (__u32 __user *)arg))
579 		return -EFAULT;
580 
581 	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
582 		return -EINVAL;
583 
584 	if (ext4_forced_shutdown(sbi))
585 		return 0;
586 
587 	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
588 	trace_ext4_shutdown(sb, flags);
589 
590 	switch (flags) {
591 	case EXT4_GOING_FLAGS_DEFAULT:
592 		freeze_bdev(sb->s_bdev);
593 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
594 		thaw_bdev(sb->s_bdev, sb);
595 		break;
596 	case EXT4_GOING_FLAGS_LOGFLUSH:
597 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
598 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
599 			(void) ext4_force_commit(sb);
600 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
601 		}
602 		break;
603 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
604 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
605 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
606 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
607 		break;
608 	default:
609 		return -EINVAL;
610 	}
611 	clear_opt(sb, DISCARD);
612 	return 0;
613 }
614 
615 struct getfsmap_info {
616 	struct super_block	*gi_sb;
617 	struct fsmap_head __user *gi_data;
618 	unsigned int		gi_idx;
619 	__u32			gi_last_flags;
620 };
621 
622 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
623 {
624 	struct getfsmap_info *info = priv;
625 	struct fsmap fm;
626 
627 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
628 
629 	info->gi_last_flags = xfm->fmr_flags;
630 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
631 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
632 			sizeof(struct fsmap)))
633 		return -EFAULT;
634 
635 	return 0;
636 }
637 
638 static int ext4_ioc_getfsmap(struct super_block *sb,
639 			     struct fsmap_head __user *arg)
640 {
641 	struct getfsmap_info info = { NULL };
642 	struct ext4_fsmap_head xhead = {0};
643 	struct fsmap_head head;
644 	bool aborted = false;
645 	int error;
646 
647 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
648 		return -EFAULT;
649 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
650 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
651 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
652 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
653 		       sizeof(head.fmh_keys[1].fmr_reserved)))
654 		return -EINVAL;
655 	/*
656 	 * ext4 doesn't report file extents at all, so the only valid
657 	 * file offsets are the magic ones (all zeroes or all ones).
658 	 */
659 	if (head.fmh_keys[0].fmr_offset ||
660 	    (head.fmh_keys[1].fmr_offset != 0 &&
661 	     head.fmh_keys[1].fmr_offset != -1ULL))
662 		return -EINVAL;
663 
664 	xhead.fmh_iflags = head.fmh_iflags;
665 	xhead.fmh_count = head.fmh_count;
666 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
667 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
668 
669 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
670 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
671 
672 	info.gi_sb = sb;
673 	info.gi_data = arg;
674 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
675 	if (error == EXT4_QUERY_RANGE_ABORT) {
676 		error = 0;
677 		aborted = true;
678 	} else if (error)
679 		return error;
680 
681 	/* If we didn't abort, set the "last" flag in the last fmx */
682 	if (!aborted && info.gi_idx) {
683 		info.gi_last_flags |= FMR_OF_LAST;
684 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
685 				 &info.gi_last_flags,
686 				 sizeof(info.gi_last_flags)))
687 			return -EFAULT;
688 	}
689 
690 	/* copy back header */
691 	head.fmh_entries = xhead.fmh_entries;
692 	head.fmh_oflags = xhead.fmh_oflags;
693 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
694 		return -EFAULT;
695 
696 	return 0;
697 }
698 
699 static long ext4_ioctl_group_add(struct file *file,
700 				 struct ext4_new_group_data *input)
701 {
702 	struct super_block *sb = file_inode(file)->i_sb;
703 	int err, err2=0;
704 
705 	err = ext4_resize_begin(sb);
706 	if (err)
707 		return err;
708 
709 	if (ext4_has_feature_bigalloc(sb)) {
710 		ext4_msg(sb, KERN_ERR,
711 			 "Online resizing not supported with bigalloc");
712 		err = -EOPNOTSUPP;
713 		goto group_add_out;
714 	}
715 
716 	err = mnt_want_write_file(file);
717 	if (err)
718 		goto group_add_out;
719 
720 	err = ext4_group_add(sb, input);
721 	if (EXT4_SB(sb)->s_journal) {
722 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
723 		err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
724 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
725 	}
726 	if (err == 0)
727 		err = err2;
728 	mnt_drop_write_file(file);
729 	if (!err && ext4_has_group_desc_csum(sb) &&
730 	    test_opt(sb, INIT_INODE_TABLE))
731 		err = ext4_register_li_request(sb, input->group);
732 group_add_out:
733 	ext4_resize_end(sb);
734 	return err;
735 }
736 
737 static void ext4_fill_fsxattr(struct inode *inode, struct fsxattr *fa)
738 {
739 	struct ext4_inode_info *ei = EXT4_I(inode);
740 
741 	simple_fill_fsxattr(fa, ext4_iflags_to_xflags(ei->i_flags &
742 						      EXT4_FL_USER_VISIBLE));
743 
744 	if (ext4_has_feature_project(inode->i_sb))
745 		fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
746 }
747 
748 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
749 {
750 	struct inode *inode = file_inode(filp);
751 	struct super_block *sb = inode->i_sb;
752 	struct ext4_inode_info *ei = EXT4_I(inode);
753 	unsigned int flags;
754 
755 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
756 
757 	switch (cmd) {
758 	case FS_IOC_GETFSMAP:
759 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
760 	case EXT4_IOC_GETFLAGS:
761 		flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
762 		if (S_ISREG(inode->i_mode))
763 			flags &= ~EXT4_PROJINHERIT_FL;
764 		return put_user(flags, (int __user *) arg);
765 	case EXT4_IOC_SETFLAGS: {
766 		int err;
767 
768 		if (!inode_owner_or_capable(inode))
769 			return -EACCES;
770 
771 		if (get_user(flags, (int __user *) arg))
772 			return -EFAULT;
773 
774 		if (flags & ~EXT4_FL_USER_VISIBLE)
775 			return -EOPNOTSUPP;
776 		/*
777 		 * chattr(1) grabs flags via GETFLAGS, modifies the result and
778 		 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
779 		 * more restrictive than just silently masking off visible but
780 		 * not settable flags as we always did.
781 		 */
782 		flags &= EXT4_FL_USER_MODIFIABLE;
783 		if (ext4_mask_flags(inode->i_mode, flags) != flags)
784 			return -EOPNOTSUPP;
785 
786 		err = mnt_want_write_file(filp);
787 		if (err)
788 			return err;
789 
790 		inode_lock(inode);
791 		err = ext4_ioctl_check_immutable(inode,
792 				from_kprojid(&init_user_ns, ei->i_projid),
793 				flags);
794 		if (!err)
795 			err = ext4_ioctl_setflags(inode, flags);
796 		inode_unlock(inode);
797 		mnt_drop_write_file(filp);
798 		return err;
799 	}
800 	case EXT4_IOC_GETVERSION:
801 	case EXT4_IOC_GETVERSION_OLD:
802 		return put_user(inode->i_generation, (int __user *) arg);
803 	case EXT4_IOC_SETVERSION:
804 	case EXT4_IOC_SETVERSION_OLD: {
805 		handle_t *handle;
806 		struct ext4_iloc iloc;
807 		__u32 generation;
808 		int err;
809 
810 		if (!inode_owner_or_capable(inode))
811 			return -EPERM;
812 
813 		if (ext4_has_metadata_csum(inode->i_sb)) {
814 			ext4_warning(sb, "Setting inode version is not "
815 				     "supported with metadata_csum enabled.");
816 			return -ENOTTY;
817 		}
818 
819 		err = mnt_want_write_file(filp);
820 		if (err)
821 			return err;
822 		if (get_user(generation, (int __user *) arg)) {
823 			err = -EFAULT;
824 			goto setversion_out;
825 		}
826 
827 		inode_lock(inode);
828 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
829 		if (IS_ERR(handle)) {
830 			err = PTR_ERR(handle);
831 			goto unlock_out;
832 		}
833 		err = ext4_reserve_inode_write(handle, inode, &iloc);
834 		if (err == 0) {
835 			inode->i_ctime = current_time(inode);
836 			inode->i_generation = generation;
837 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
838 		}
839 		ext4_journal_stop(handle);
840 
841 unlock_out:
842 		inode_unlock(inode);
843 setversion_out:
844 		mnt_drop_write_file(filp);
845 		return err;
846 	}
847 	case EXT4_IOC_GROUP_EXTEND: {
848 		ext4_fsblk_t n_blocks_count;
849 		int err, err2=0;
850 
851 		err = ext4_resize_begin(sb);
852 		if (err)
853 			return err;
854 
855 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
856 			err = -EFAULT;
857 			goto group_extend_out;
858 		}
859 
860 		if (ext4_has_feature_bigalloc(sb)) {
861 			ext4_msg(sb, KERN_ERR,
862 				 "Online resizing not supported with bigalloc");
863 			err = -EOPNOTSUPP;
864 			goto group_extend_out;
865 		}
866 
867 		err = mnt_want_write_file(filp);
868 		if (err)
869 			goto group_extend_out;
870 
871 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
872 		if (EXT4_SB(sb)->s_journal) {
873 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
874 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
875 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
876 		}
877 		if (err == 0)
878 			err = err2;
879 		mnt_drop_write_file(filp);
880 group_extend_out:
881 		ext4_resize_end(sb);
882 		return err;
883 	}
884 
885 	case EXT4_IOC_MOVE_EXT: {
886 		struct move_extent me;
887 		struct fd donor;
888 		int err;
889 
890 		if (!(filp->f_mode & FMODE_READ) ||
891 		    !(filp->f_mode & FMODE_WRITE))
892 			return -EBADF;
893 
894 		if (copy_from_user(&me,
895 			(struct move_extent __user *)arg, sizeof(me)))
896 			return -EFAULT;
897 		me.moved_len = 0;
898 
899 		donor = fdget(me.donor_fd);
900 		if (!donor.file)
901 			return -EBADF;
902 
903 		if (!(donor.file->f_mode & FMODE_WRITE)) {
904 			err = -EBADF;
905 			goto mext_out;
906 		}
907 
908 		if (ext4_has_feature_bigalloc(sb)) {
909 			ext4_msg(sb, KERN_ERR,
910 				 "Online defrag not supported with bigalloc");
911 			err = -EOPNOTSUPP;
912 			goto mext_out;
913 		} else if (IS_DAX(inode)) {
914 			ext4_msg(sb, KERN_ERR,
915 				 "Online defrag not supported with DAX");
916 			err = -EOPNOTSUPP;
917 			goto mext_out;
918 		}
919 
920 		err = mnt_want_write_file(filp);
921 		if (err)
922 			goto mext_out;
923 
924 		err = ext4_move_extents(filp, donor.file, me.orig_start,
925 					me.donor_start, me.len, &me.moved_len);
926 		mnt_drop_write_file(filp);
927 
928 		if (copy_to_user((struct move_extent __user *)arg,
929 				 &me, sizeof(me)))
930 			err = -EFAULT;
931 mext_out:
932 		fdput(donor);
933 		return err;
934 	}
935 
936 	case EXT4_IOC_GROUP_ADD: {
937 		struct ext4_new_group_data input;
938 
939 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
940 				sizeof(input)))
941 			return -EFAULT;
942 
943 		return ext4_ioctl_group_add(filp, &input);
944 	}
945 
946 	case EXT4_IOC_MIGRATE:
947 	{
948 		int err;
949 		if (!inode_owner_or_capable(inode))
950 			return -EACCES;
951 
952 		err = mnt_want_write_file(filp);
953 		if (err)
954 			return err;
955 		/*
956 		 * inode_mutex prevent write and truncate on the file.
957 		 * Read still goes through. We take i_data_sem in
958 		 * ext4_ext_swap_inode_data before we switch the
959 		 * inode format to prevent read.
960 		 */
961 		inode_lock((inode));
962 		err = ext4_ext_migrate(inode);
963 		inode_unlock((inode));
964 		mnt_drop_write_file(filp);
965 		return err;
966 	}
967 
968 	case EXT4_IOC_ALLOC_DA_BLKS:
969 	{
970 		int err;
971 		if (!inode_owner_or_capable(inode))
972 			return -EACCES;
973 
974 		err = mnt_want_write_file(filp);
975 		if (err)
976 			return err;
977 		err = ext4_alloc_da_blocks(inode);
978 		mnt_drop_write_file(filp);
979 		return err;
980 	}
981 
982 	case EXT4_IOC_SWAP_BOOT:
983 	{
984 		int err;
985 		if (!(filp->f_mode & FMODE_WRITE))
986 			return -EBADF;
987 		err = mnt_want_write_file(filp);
988 		if (err)
989 			return err;
990 		err = swap_inode_boot_loader(sb, inode);
991 		mnt_drop_write_file(filp);
992 		return err;
993 	}
994 
995 	case EXT4_IOC_RESIZE_FS: {
996 		ext4_fsblk_t n_blocks_count;
997 		int err = 0, err2 = 0;
998 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
999 
1000 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1001 				   sizeof(__u64))) {
1002 			return -EFAULT;
1003 		}
1004 
1005 		err = ext4_resize_begin(sb);
1006 		if (err)
1007 			return err;
1008 
1009 		err = mnt_want_write_file(filp);
1010 		if (err)
1011 			goto resizefs_out;
1012 
1013 		err = ext4_resize_fs(sb, n_blocks_count);
1014 		if (EXT4_SB(sb)->s_journal) {
1015 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1016 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
1017 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1018 		}
1019 		if (err == 0)
1020 			err = err2;
1021 		mnt_drop_write_file(filp);
1022 		if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1023 		    ext4_has_group_desc_csum(sb) &&
1024 		    test_opt(sb, INIT_INODE_TABLE))
1025 			err = ext4_register_li_request(sb, o_group);
1026 
1027 resizefs_out:
1028 		ext4_resize_end(sb);
1029 		return err;
1030 	}
1031 
1032 	case FITRIM:
1033 	{
1034 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
1035 		struct fstrim_range range;
1036 		int ret = 0;
1037 
1038 		if (!capable(CAP_SYS_ADMIN))
1039 			return -EPERM;
1040 
1041 		if (!blk_queue_discard(q))
1042 			return -EOPNOTSUPP;
1043 
1044 		/*
1045 		 * We haven't replayed the journal, so we cannot use our
1046 		 * block-bitmap-guided storage zapping commands.
1047 		 */
1048 		if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1049 			return -EROFS;
1050 
1051 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1052 		    sizeof(range)))
1053 			return -EFAULT;
1054 
1055 		range.minlen = max((unsigned int)range.minlen,
1056 				   q->limits.discard_granularity);
1057 		ret = ext4_trim_fs(sb, &range);
1058 		if (ret < 0)
1059 			return ret;
1060 
1061 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
1062 		    sizeof(range)))
1063 			return -EFAULT;
1064 
1065 		return 0;
1066 	}
1067 	case EXT4_IOC_PRECACHE_EXTENTS:
1068 		return ext4_ext_precache(inode);
1069 
1070 	case EXT4_IOC_SET_ENCRYPTION_POLICY:
1071 		if (!ext4_has_feature_encrypt(sb))
1072 			return -EOPNOTSUPP;
1073 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1074 
1075 	case EXT4_IOC_GET_ENCRYPTION_PWSALT: {
1076 #ifdef CONFIG_FS_ENCRYPTION
1077 		int err, err2;
1078 		struct ext4_sb_info *sbi = EXT4_SB(sb);
1079 		handle_t *handle;
1080 
1081 		if (!ext4_has_feature_encrypt(sb))
1082 			return -EOPNOTSUPP;
1083 		if (uuid_is_zero(sbi->s_es->s_encrypt_pw_salt)) {
1084 			err = mnt_want_write_file(filp);
1085 			if (err)
1086 				return err;
1087 			handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
1088 			if (IS_ERR(handle)) {
1089 				err = PTR_ERR(handle);
1090 				goto pwsalt_err_exit;
1091 			}
1092 			err = ext4_journal_get_write_access(handle, sbi->s_sbh);
1093 			if (err)
1094 				goto pwsalt_err_journal;
1095 			generate_random_uuid(sbi->s_es->s_encrypt_pw_salt);
1096 			err = ext4_handle_dirty_metadata(handle, NULL,
1097 							 sbi->s_sbh);
1098 		pwsalt_err_journal:
1099 			err2 = ext4_journal_stop(handle);
1100 			if (err2 && !err)
1101 				err = err2;
1102 		pwsalt_err_exit:
1103 			mnt_drop_write_file(filp);
1104 			if (err)
1105 				return err;
1106 		}
1107 		if (copy_to_user((void __user *) arg,
1108 				 sbi->s_es->s_encrypt_pw_salt, 16))
1109 			return -EFAULT;
1110 		return 0;
1111 #else
1112 		return -EOPNOTSUPP;
1113 #endif
1114 	}
1115 	case EXT4_IOC_GET_ENCRYPTION_POLICY:
1116 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1117 
1118 	case EXT4_IOC_FSGETXATTR:
1119 	{
1120 		struct fsxattr fa;
1121 
1122 		ext4_fill_fsxattr(inode, &fa);
1123 
1124 		if (copy_to_user((struct fsxattr __user *)arg,
1125 				 &fa, sizeof(fa)))
1126 			return -EFAULT;
1127 		return 0;
1128 	}
1129 	case EXT4_IOC_FSSETXATTR:
1130 	{
1131 		struct fsxattr fa, old_fa;
1132 		int err;
1133 
1134 		if (copy_from_user(&fa, (struct fsxattr __user *)arg,
1135 				   sizeof(fa)))
1136 			return -EFAULT;
1137 
1138 		/* Make sure caller has proper permission */
1139 		if (!inode_owner_or_capable(inode))
1140 			return -EACCES;
1141 
1142 		if (fa.fsx_xflags & ~EXT4_SUPPORTED_FS_XFLAGS)
1143 			return -EOPNOTSUPP;
1144 
1145 		flags = ext4_xflags_to_iflags(fa.fsx_xflags);
1146 		if (ext4_mask_flags(inode->i_mode, flags) != flags)
1147 			return -EOPNOTSUPP;
1148 
1149 		err = mnt_want_write_file(filp);
1150 		if (err)
1151 			return err;
1152 
1153 		inode_lock(inode);
1154 		ext4_fill_fsxattr(inode, &old_fa);
1155 		err = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
1156 		if (err)
1157 			goto out;
1158 		flags = (ei->i_flags & ~EXT4_FL_XFLAG_VISIBLE) |
1159 			 (flags & EXT4_FL_XFLAG_VISIBLE);
1160 		err = ext4_ioctl_check_immutable(inode, fa.fsx_projid, flags);
1161 		if (err)
1162 			goto out;
1163 		err = ext4_ioctl_setflags(inode, flags);
1164 		if (err)
1165 			goto out;
1166 		err = ext4_ioctl_setproject(filp, fa.fsx_projid);
1167 out:
1168 		inode_unlock(inode);
1169 		mnt_drop_write_file(filp);
1170 		return err;
1171 	}
1172 	case EXT4_IOC_SHUTDOWN:
1173 		return ext4_shutdown(sb, arg);
1174 	default:
1175 		return -ENOTTY;
1176 	}
1177 }
1178 
1179 #ifdef CONFIG_COMPAT
1180 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1181 {
1182 	/* These are just misnamed, they actually get/put from/to user an int */
1183 	switch (cmd) {
1184 	case EXT4_IOC32_GETFLAGS:
1185 		cmd = EXT4_IOC_GETFLAGS;
1186 		break;
1187 	case EXT4_IOC32_SETFLAGS:
1188 		cmd = EXT4_IOC_SETFLAGS;
1189 		break;
1190 	case EXT4_IOC32_GETVERSION:
1191 		cmd = EXT4_IOC_GETVERSION;
1192 		break;
1193 	case EXT4_IOC32_SETVERSION:
1194 		cmd = EXT4_IOC_SETVERSION;
1195 		break;
1196 	case EXT4_IOC32_GROUP_EXTEND:
1197 		cmd = EXT4_IOC_GROUP_EXTEND;
1198 		break;
1199 	case EXT4_IOC32_GETVERSION_OLD:
1200 		cmd = EXT4_IOC_GETVERSION_OLD;
1201 		break;
1202 	case EXT4_IOC32_SETVERSION_OLD:
1203 		cmd = EXT4_IOC_SETVERSION_OLD;
1204 		break;
1205 	case EXT4_IOC32_GETRSVSZ:
1206 		cmd = EXT4_IOC_GETRSVSZ;
1207 		break;
1208 	case EXT4_IOC32_SETRSVSZ:
1209 		cmd = EXT4_IOC_SETRSVSZ;
1210 		break;
1211 	case EXT4_IOC32_GROUP_ADD: {
1212 		struct compat_ext4_new_group_input __user *uinput;
1213 		struct ext4_new_group_data input;
1214 		int err;
1215 
1216 		uinput = compat_ptr(arg);
1217 		err = get_user(input.group, &uinput->group);
1218 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1219 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1220 		err |= get_user(input.inode_table, &uinput->inode_table);
1221 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1222 		err |= get_user(input.reserved_blocks,
1223 				&uinput->reserved_blocks);
1224 		if (err)
1225 			return -EFAULT;
1226 		return ext4_ioctl_group_add(file, &input);
1227 	}
1228 	case EXT4_IOC_MOVE_EXT:
1229 	case EXT4_IOC_RESIZE_FS:
1230 	case EXT4_IOC_PRECACHE_EXTENTS:
1231 	case EXT4_IOC_SET_ENCRYPTION_POLICY:
1232 	case EXT4_IOC_GET_ENCRYPTION_PWSALT:
1233 	case EXT4_IOC_GET_ENCRYPTION_POLICY:
1234 	case EXT4_IOC_SHUTDOWN:
1235 	case FS_IOC_GETFSMAP:
1236 		break;
1237 	default:
1238 		return -ENOIOCTLCMD;
1239 	}
1240 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1241 }
1242 #endif
1243