xref: /linux/fs/ext4/ioctl.c (revision e58e871becec2d3b04ed91c0c16fe8deac9c9dfa)
1 /*
2  * linux/fs/ext4/ioctl.c
3  *
4  * Copyright (C) 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  */
9 
10 #include <linux/fs.h>
11 #include <linux/capability.h>
12 #include <linux/time.h>
13 #include <linux/compat.h>
14 #include <linux/mount.h>
15 #include <linux/file.h>
16 #include <linux/quotaops.h>
17 #include <linux/uuid.h>
18 #include <linux/uaccess.h>
19 #include <linux/delay.h>
20 #include "ext4_jbd2.h"
21 #include "ext4.h"
22 #include <linux/fsmap.h>
23 #include "fsmap.h"
24 #include <trace/events/ext4.h>
25 
26 /**
27  * Swap memory between @a and @b for @len bytes.
28  *
29  * @a:          pointer to first memory area
30  * @b:          pointer to second memory area
31  * @len:        number of bytes to swap
32  *
33  */
34 static void memswap(void *a, void *b, size_t len)
35 {
36 	unsigned char *ap, *bp;
37 
38 	ap = (unsigned char *)a;
39 	bp = (unsigned char *)b;
40 	while (len-- > 0) {
41 		swap(*ap, *bp);
42 		ap++;
43 		bp++;
44 	}
45 }
46 
47 /**
48  * Swap i_data and associated attributes between @inode1 and @inode2.
49  * This function is used for the primary swap between inode1 and inode2
50  * and also to revert this primary swap in case of errors.
51  *
52  * Therefore you have to make sure, that calling this method twice
53  * will revert all changes.
54  *
55  * @inode1:     pointer to first inode
56  * @inode2:     pointer to second inode
57  */
58 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
59 {
60 	loff_t isize;
61 	struct ext4_inode_info *ei1;
62 	struct ext4_inode_info *ei2;
63 
64 	ei1 = EXT4_I(inode1);
65 	ei2 = EXT4_I(inode2);
66 
67 	memswap(&inode1->i_flags, &inode2->i_flags, sizeof(inode1->i_flags));
68 	memswap(&inode1->i_version, &inode2->i_version,
69 		  sizeof(inode1->i_version));
70 	memswap(&inode1->i_blocks, &inode2->i_blocks,
71 		  sizeof(inode1->i_blocks));
72 	memswap(&inode1->i_bytes, &inode2->i_bytes, sizeof(inode1->i_bytes));
73 	memswap(&inode1->i_atime, &inode2->i_atime, sizeof(inode1->i_atime));
74 	memswap(&inode1->i_mtime, &inode2->i_mtime, sizeof(inode1->i_mtime));
75 
76 	memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
77 	memswap(&ei1->i_flags, &ei2->i_flags, sizeof(ei1->i_flags));
78 	memswap(&ei1->i_disksize, &ei2->i_disksize, sizeof(ei1->i_disksize));
79 	ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
80 	ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
81 
82 	isize = i_size_read(inode1);
83 	i_size_write(inode1, i_size_read(inode2));
84 	i_size_write(inode2, isize);
85 }
86 
87 /**
88  * Swap the information from the given @inode and the inode
89  * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
90  * important fields of the inodes.
91  *
92  * @sb:         the super block of the filesystem
93  * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
94  *
95  */
96 static long swap_inode_boot_loader(struct super_block *sb,
97 				struct inode *inode)
98 {
99 	handle_t *handle;
100 	int err;
101 	struct inode *inode_bl;
102 	struct ext4_inode_info *ei_bl;
103 	struct ext4_sb_info *sbi = EXT4_SB(sb);
104 
105 	if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode))
106 		return -EINVAL;
107 
108 	if (!inode_owner_or_capable(inode) || !capable(CAP_SYS_ADMIN))
109 		return -EPERM;
110 
111 	inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO);
112 	if (IS_ERR(inode_bl))
113 		return PTR_ERR(inode_bl);
114 	ei_bl = EXT4_I(inode_bl);
115 
116 	filemap_flush(inode->i_mapping);
117 	filemap_flush(inode_bl->i_mapping);
118 
119 	/* Protect orig inodes against a truncate and make sure,
120 	 * that only 1 swap_inode_boot_loader is running. */
121 	lock_two_nondirectories(inode, inode_bl);
122 
123 	truncate_inode_pages(&inode->i_data, 0);
124 	truncate_inode_pages(&inode_bl->i_data, 0);
125 
126 	/* Wait for all existing dio workers */
127 	ext4_inode_block_unlocked_dio(inode);
128 	ext4_inode_block_unlocked_dio(inode_bl);
129 	inode_dio_wait(inode);
130 	inode_dio_wait(inode_bl);
131 
132 	handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
133 	if (IS_ERR(handle)) {
134 		err = -EINVAL;
135 		goto journal_err_out;
136 	}
137 
138 	/* Protect extent tree against block allocations via delalloc */
139 	ext4_double_down_write_data_sem(inode, inode_bl);
140 
141 	if (inode_bl->i_nlink == 0) {
142 		/* this inode has never been used as a BOOT_LOADER */
143 		set_nlink(inode_bl, 1);
144 		i_uid_write(inode_bl, 0);
145 		i_gid_write(inode_bl, 0);
146 		inode_bl->i_flags = 0;
147 		ei_bl->i_flags = 0;
148 		inode_bl->i_version = 1;
149 		i_size_write(inode_bl, 0);
150 		inode_bl->i_mode = S_IFREG;
151 		if (ext4_has_feature_extents(sb)) {
152 			ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
153 			ext4_ext_tree_init(handle, inode_bl);
154 		} else
155 			memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
156 	}
157 
158 	swap_inode_data(inode, inode_bl);
159 
160 	inode->i_ctime = inode_bl->i_ctime = current_time(inode);
161 
162 	spin_lock(&sbi->s_next_gen_lock);
163 	inode->i_generation = sbi->s_next_generation++;
164 	inode_bl->i_generation = sbi->s_next_generation++;
165 	spin_unlock(&sbi->s_next_gen_lock);
166 
167 	ext4_discard_preallocations(inode);
168 
169 	err = ext4_mark_inode_dirty(handle, inode);
170 	if (err < 0) {
171 		ext4_warning(inode->i_sb,
172 			"couldn't mark inode #%lu dirty (err %d)",
173 			inode->i_ino, err);
174 		/* Revert all changes: */
175 		swap_inode_data(inode, inode_bl);
176 	} else {
177 		err = ext4_mark_inode_dirty(handle, inode_bl);
178 		if (err < 0) {
179 			ext4_warning(inode_bl->i_sb,
180 				"couldn't mark inode #%lu dirty (err %d)",
181 				inode_bl->i_ino, err);
182 			/* Revert all changes: */
183 			swap_inode_data(inode, inode_bl);
184 			ext4_mark_inode_dirty(handle, inode);
185 		}
186 	}
187 	ext4_journal_stop(handle);
188 	ext4_double_up_write_data_sem(inode, inode_bl);
189 
190 journal_err_out:
191 	ext4_inode_resume_unlocked_dio(inode);
192 	ext4_inode_resume_unlocked_dio(inode_bl);
193 	unlock_two_nondirectories(inode, inode_bl);
194 	iput(inode_bl);
195 	return err;
196 }
197 
198 #ifdef CONFIG_EXT4_FS_ENCRYPTION
199 static int uuid_is_zero(__u8 u[16])
200 {
201 	int	i;
202 
203 	for (i = 0; i < 16; i++)
204 		if (u[i])
205 			return 0;
206 	return 1;
207 }
208 #endif
209 
210 static int ext4_ioctl_setflags(struct inode *inode,
211 			       unsigned int flags)
212 {
213 	struct ext4_inode_info *ei = EXT4_I(inode);
214 	handle_t *handle = NULL;
215 	int err = -EPERM, migrate = 0;
216 	struct ext4_iloc iloc;
217 	unsigned int oldflags, mask, i;
218 	unsigned int jflag;
219 
220 	/* Is it quota file? Do not allow user to mess with it */
221 	if (IS_NOQUOTA(inode))
222 		goto flags_out;
223 
224 	oldflags = ei->i_flags;
225 
226 	/* The JOURNAL_DATA flag is modifiable only by root */
227 	jflag = flags & EXT4_JOURNAL_DATA_FL;
228 
229 	/*
230 	 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
231 	 * the relevant capability.
232 	 *
233 	 * This test looks nicer. Thanks to Pauline Middelink
234 	 */
235 	if ((flags ^ oldflags) & (EXT4_APPEND_FL | EXT4_IMMUTABLE_FL)) {
236 		if (!capable(CAP_LINUX_IMMUTABLE))
237 			goto flags_out;
238 	}
239 
240 	/*
241 	 * The JOURNAL_DATA flag can only be changed by
242 	 * the relevant capability.
243 	 */
244 	if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
245 		if (!capable(CAP_SYS_RESOURCE))
246 			goto flags_out;
247 	}
248 	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
249 		migrate = 1;
250 
251 	if (flags & EXT4_EOFBLOCKS_FL) {
252 		/* we don't support adding EOFBLOCKS flag */
253 		if (!(oldflags & EXT4_EOFBLOCKS_FL)) {
254 			err = -EOPNOTSUPP;
255 			goto flags_out;
256 		}
257 	} else if (oldflags & EXT4_EOFBLOCKS_FL) {
258 		err = ext4_truncate(inode);
259 		if (err)
260 			goto flags_out;
261 	}
262 
263 	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
264 	if (IS_ERR(handle)) {
265 		err = PTR_ERR(handle);
266 		goto flags_out;
267 	}
268 	if (IS_SYNC(inode))
269 		ext4_handle_sync(handle);
270 	err = ext4_reserve_inode_write(handle, inode, &iloc);
271 	if (err)
272 		goto flags_err;
273 
274 	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
275 		if (!(mask & EXT4_FL_USER_MODIFIABLE))
276 			continue;
277 		/* These flags get special treatment later */
278 		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
279 			continue;
280 		if (mask & flags)
281 			ext4_set_inode_flag(inode, i);
282 		else
283 			ext4_clear_inode_flag(inode, i);
284 	}
285 
286 	ext4_set_inode_flags(inode);
287 	inode->i_ctime = current_time(inode);
288 
289 	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
290 flags_err:
291 	ext4_journal_stop(handle);
292 	if (err)
293 		goto flags_out;
294 
295 	if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL))
296 		err = ext4_change_inode_journal_flag(inode, jflag);
297 	if (err)
298 		goto flags_out;
299 	if (migrate) {
300 		if (flags & EXT4_EXTENTS_FL)
301 			err = ext4_ext_migrate(inode);
302 		else
303 			err = ext4_ind_migrate(inode);
304 	}
305 
306 flags_out:
307 	return err;
308 }
309 
310 #ifdef CONFIG_QUOTA
311 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
312 {
313 	struct inode *inode = file_inode(filp);
314 	struct super_block *sb = inode->i_sb;
315 	struct ext4_inode_info *ei = EXT4_I(inode);
316 	int err, rc;
317 	handle_t *handle;
318 	kprojid_t kprojid;
319 	struct ext4_iloc iloc;
320 	struct ext4_inode *raw_inode;
321 	struct dquot *transfer_to[MAXQUOTAS] = { };
322 
323 	if (!ext4_has_feature_project(sb)) {
324 		if (projid != EXT4_DEF_PROJID)
325 			return -EOPNOTSUPP;
326 		else
327 			return 0;
328 	}
329 
330 	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
331 		return -EOPNOTSUPP;
332 
333 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
334 
335 	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
336 		return 0;
337 
338 	err = mnt_want_write_file(filp);
339 	if (err)
340 		return err;
341 
342 	err = -EPERM;
343 	inode_lock(inode);
344 	/* Is it quota file? Do not allow user to mess with it */
345 	if (IS_NOQUOTA(inode))
346 		goto out_unlock;
347 
348 	err = ext4_get_inode_loc(inode, &iloc);
349 	if (err)
350 		goto out_unlock;
351 
352 	raw_inode = ext4_raw_inode(&iloc);
353 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
354 		err = -EOVERFLOW;
355 		brelse(iloc.bh);
356 		goto out_unlock;
357 	}
358 	brelse(iloc.bh);
359 
360 	dquot_initialize(inode);
361 
362 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
363 		EXT4_QUOTA_INIT_BLOCKS(sb) +
364 		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
365 	if (IS_ERR(handle)) {
366 		err = PTR_ERR(handle);
367 		goto out_unlock;
368 	}
369 
370 	err = ext4_reserve_inode_write(handle, inode, &iloc);
371 	if (err)
372 		goto out_stop;
373 
374 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
375 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
376 		err = __dquot_transfer(inode, transfer_to);
377 		dqput(transfer_to[PRJQUOTA]);
378 		if (err)
379 			goto out_dirty;
380 	}
381 
382 	EXT4_I(inode)->i_projid = kprojid;
383 	inode->i_ctime = current_time(inode);
384 out_dirty:
385 	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
386 	if (!err)
387 		err = rc;
388 out_stop:
389 	ext4_journal_stop(handle);
390 out_unlock:
391 	inode_unlock(inode);
392 	mnt_drop_write_file(filp);
393 	return err;
394 }
395 #else
396 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
397 {
398 	if (projid != EXT4_DEF_PROJID)
399 		return -EOPNOTSUPP;
400 	return 0;
401 }
402 #endif
403 
404 /* Transfer internal flags to xflags */
405 static inline __u32 ext4_iflags_to_xflags(unsigned long iflags)
406 {
407 	__u32 xflags = 0;
408 
409 	if (iflags & EXT4_SYNC_FL)
410 		xflags |= FS_XFLAG_SYNC;
411 	if (iflags & EXT4_IMMUTABLE_FL)
412 		xflags |= FS_XFLAG_IMMUTABLE;
413 	if (iflags & EXT4_APPEND_FL)
414 		xflags |= FS_XFLAG_APPEND;
415 	if (iflags & EXT4_NODUMP_FL)
416 		xflags |= FS_XFLAG_NODUMP;
417 	if (iflags & EXT4_NOATIME_FL)
418 		xflags |= FS_XFLAG_NOATIME;
419 	if (iflags & EXT4_PROJINHERIT_FL)
420 		xflags |= FS_XFLAG_PROJINHERIT;
421 	return xflags;
422 }
423 
424 #define EXT4_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
425 				  FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
426 				  FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
427 
428 /* Transfer xflags flags to internal */
429 static inline unsigned long ext4_xflags_to_iflags(__u32 xflags)
430 {
431 	unsigned long iflags = 0;
432 
433 	if (xflags & FS_XFLAG_SYNC)
434 		iflags |= EXT4_SYNC_FL;
435 	if (xflags & FS_XFLAG_IMMUTABLE)
436 		iflags |= EXT4_IMMUTABLE_FL;
437 	if (xflags & FS_XFLAG_APPEND)
438 		iflags |= EXT4_APPEND_FL;
439 	if (xflags & FS_XFLAG_NODUMP)
440 		iflags |= EXT4_NODUMP_FL;
441 	if (xflags & FS_XFLAG_NOATIME)
442 		iflags |= EXT4_NOATIME_FL;
443 	if (xflags & FS_XFLAG_PROJINHERIT)
444 		iflags |= EXT4_PROJINHERIT_FL;
445 
446 	return iflags;
447 }
448 
449 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
450 {
451 	struct ext4_sb_info *sbi = EXT4_SB(sb);
452 	__u32 flags;
453 
454 	if (!capable(CAP_SYS_ADMIN))
455 		return -EPERM;
456 
457 	if (get_user(flags, (__u32 __user *)arg))
458 		return -EFAULT;
459 
460 	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
461 		return -EINVAL;
462 
463 	if (ext4_forced_shutdown(sbi))
464 		return 0;
465 
466 	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
467 
468 	switch (flags) {
469 	case EXT4_GOING_FLAGS_DEFAULT:
470 		freeze_bdev(sb->s_bdev);
471 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
472 		thaw_bdev(sb->s_bdev, sb);
473 		break;
474 	case EXT4_GOING_FLAGS_LOGFLUSH:
475 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
476 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
477 			(void) ext4_force_commit(sb);
478 			jbd2_journal_abort(sbi->s_journal, 0);
479 		}
480 		break;
481 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
482 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
483 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
484 			msleep(100);
485 			jbd2_journal_abort(sbi->s_journal, 0);
486 		}
487 		break;
488 	default:
489 		return -EINVAL;
490 	}
491 	clear_opt(sb, DISCARD);
492 	return 0;
493 }
494 
495 struct getfsmap_info {
496 	struct super_block	*gi_sb;
497 	struct fsmap_head __user *gi_data;
498 	unsigned int		gi_idx;
499 	__u32			gi_last_flags;
500 };
501 
502 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
503 {
504 	struct getfsmap_info *info = priv;
505 	struct fsmap fm;
506 
507 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
508 
509 	info->gi_last_flags = xfm->fmr_flags;
510 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
511 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
512 			sizeof(struct fsmap)))
513 		return -EFAULT;
514 
515 	return 0;
516 }
517 
518 static int ext4_ioc_getfsmap(struct super_block *sb,
519 			     struct fsmap_head __user *arg)
520 {
521 	struct getfsmap_info info = {0};
522 	struct ext4_fsmap_head xhead = {0};
523 	struct fsmap_head head;
524 	bool aborted = false;
525 	int error;
526 
527 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
528 		return -EFAULT;
529 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
530 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
531 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
532 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
533 		       sizeof(head.fmh_keys[1].fmr_reserved)))
534 		return -EINVAL;
535 	/*
536 	 * ext4 doesn't report file extents at all, so the only valid
537 	 * file offsets are the magic ones (all zeroes or all ones).
538 	 */
539 	if (head.fmh_keys[0].fmr_offset ||
540 	    (head.fmh_keys[1].fmr_offset != 0 &&
541 	     head.fmh_keys[1].fmr_offset != -1ULL))
542 		return -EINVAL;
543 
544 	xhead.fmh_iflags = head.fmh_iflags;
545 	xhead.fmh_count = head.fmh_count;
546 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
547 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
548 
549 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
550 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
551 
552 	info.gi_sb = sb;
553 	info.gi_data = arg;
554 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
555 	if (error == EXT4_QUERY_RANGE_ABORT) {
556 		error = 0;
557 		aborted = true;
558 	} else if (error)
559 		return error;
560 
561 	/* If we didn't abort, set the "last" flag in the last fmx */
562 	if (!aborted && info.gi_idx) {
563 		info.gi_last_flags |= FMR_OF_LAST;
564 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
565 				 &info.gi_last_flags,
566 				 sizeof(info.gi_last_flags)))
567 			return -EFAULT;
568 	}
569 
570 	/* copy back header */
571 	head.fmh_entries = xhead.fmh_entries;
572 	head.fmh_oflags = xhead.fmh_oflags;
573 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
574 		return -EFAULT;
575 
576 	return 0;
577 }
578 
579 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
580 {
581 	struct inode *inode = file_inode(filp);
582 	struct super_block *sb = inode->i_sb;
583 	struct ext4_inode_info *ei = EXT4_I(inode);
584 	unsigned int flags;
585 
586 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
587 
588 	switch (cmd) {
589 	case FS_IOC_GETFSMAP:
590 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
591 	case EXT4_IOC_GETFLAGS:
592 		flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
593 		return put_user(flags, (int __user *) arg);
594 	case EXT4_IOC_SETFLAGS: {
595 		int err;
596 
597 		if (!inode_owner_or_capable(inode))
598 			return -EACCES;
599 
600 		if (get_user(flags, (int __user *) arg))
601 			return -EFAULT;
602 
603 		if (flags & ~EXT4_FL_USER_VISIBLE)
604 			return -EOPNOTSUPP;
605 		/*
606 		 * chattr(1) grabs flags via GETFLAGS, modifies the result and
607 		 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
608 		 * more restrictive than just silently masking off visible but
609 		 * not settable flags as we always did.
610 		 */
611 		flags &= EXT4_FL_USER_MODIFIABLE;
612 		if (ext4_mask_flags(inode->i_mode, flags) != flags)
613 			return -EOPNOTSUPP;
614 
615 		err = mnt_want_write_file(filp);
616 		if (err)
617 			return err;
618 
619 		inode_lock(inode);
620 		err = ext4_ioctl_setflags(inode, flags);
621 		inode_unlock(inode);
622 		mnt_drop_write_file(filp);
623 		return err;
624 	}
625 	case EXT4_IOC_GETVERSION:
626 	case EXT4_IOC_GETVERSION_OLD:
627 		return put_user(inode->i_generation, (int __user *) arg);
628 	case EXT4_IOC_SETVERSION:
629 	case EXT4_IOC_SETVERSION_OLD: {
630 		handle_t *handle;
631 		struct ext4_iloc iloc;
632 		__u32 generation;
633 		int err;
634 
635 		if (!inode_owner_or_capable(inode))
636 			return -EPERM;
637 
638 		if (ext4_has_metadata_csum(inode->i_sb)) {
639 			ext4_warning(sb, "Setting inode version is not "
640 				     "supported with metadata_csum enabled.");
641 			return -ENOTTY;
642 		}
643 
644 		err = mnt_want_write_file(filp);
645 		if (err)
646 			return err;
647 		if (get_user(generation, (int __user *) arg)) {
648 			err = -EFAULT;
649 			goto setversion_out;
650 		}
651 
652 		inode_lock(inode);
653 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
654 		if (IS_ERR(handle)) {
655 			err = PTR_ERR(handle);
656 			goto unlock_out;
657 		}
658 		err = ext4_reserve_inode_write(handle, inode, &iloc);
659 		if (err == 0) {
660 			inode->i_ctime = current_time(inode);
661 			inode->i_generation = generation;
662 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
663 		}
664 		ext4_journal_stop(handle);
665 
666 unlock_out:
667 		inode_unlock(inode);
668 setversion_out:
669 		mnt_drop_write_file(filp);
670 		return err;
671 	}
672 	case EXT4_IOC_GROUP_EXTEND: {
673 		ext4_fsblk_t n_blocks_count;
674 		int err, err2=0;
675 
676 		err = ext4_resize_begin(sb);
677 		if (err)
678 			return err;
679 
680 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
681 			err = -EFAULT;
682 			goto group_extend_out;
683 		}
684 
685 		if (ext4_has_feature_bigalloc(sb)) {
686 			ext4_msg(sb, KERN_ERR,
687 				 "Online resizing not supported with bigalloc");
688 			err = -EOPNOTSUPP;
689 			goto group_extend_out;
690 		}
691 
692 		err = mnt_want_write_file(filp);
693 		if (err)
694 			goto group_extend_out;
695 
696 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
697 		if (EXT4_SB(sb)->s_journal) {
698 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
699 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
700 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
701 		}
702 		if (err == 0)
703 			err = err2;
704 		mnt_drop_write_file(filp);
705 group_extend_out:
706 		ext4_resize_end(sb);
707 		return err;
708 	}
709 
710 	case EXT4_IOC_MOVE_EXT: {
711 		struct move_extent me;
712 		struct fd donor;
713 		int err;
714 
715 		if (!(filp->f_mode & FMODE_READ) ||
716 		    !(filp->f_mode & FMODE_WRITE))
717 			return -EBADF;
718 
719 		if (copy_from_user(&me,
720 			(struct move_extent __user *)arg, sizeof(me)))
721 			return -EFAULT;
722 		me.moved_len = 0;
723 
724 		donor = fdget(me.donor_fd);
725 		if (!donor.file)
726 			return -EBADF;
727 
728 		if (!(donor.file->f_mode & FMODE_WRITE)) {
729 			err = -EBADF;
730 			goto mext_out;
731 		}
732 
733 		if (ext4_has_feature_bigalloc(sb)) {
734 			ext4_msg(sb, KERN_ERR,
735 				 "Online defrag not supported with bigalloc");
736 			err = -EOPNOTSUPP;
737 			goto mext_out;
738 		} else if (IS_DAX(inode)) {
739 			ext4_msg(sb, KERN_ERR,
740 				 "Online defrag not supported with DAX");
741 			err = -EOPNOTSUPP;
742 			goto mext_out;
743 		}
744 
745 		err = mnt_want_write_file(filp);
746 		if (err)
747 			goto mext_out;
748 
749 		err = ext4_move_extents(filp, donor.file, me.orig_start,
750 					me.donor_start, me.len, &me.moved_len);
751 		mnt_drop_write_file(filp);
752 
753 		if (copy_to_user((struct move_extent __user *)arg,
754 				 &me, sizeof(me)))
755 			err = -EFAULT;
756 mext_out:
757 		fdput(donor);
758 		return err;
759 	}
760 
761 	case EXT4_IOC_GROUP_ADD: {
762 		struct ext4_new_group_data input;
763 		int err, err2=0;
764 
765 		err = ext4_resize_begin(sb);
766 		if (err)
767 			return err;
768 
769 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
770 				sizeof(input))) {
771 			err = -EFAULT;
772 			goto group_add_out;
773 		}
774 
775 		if (ext4_has_feature_bigalloc(sb)) {
776 			ext4_msg(sb, KERN_ERR,
777 				 "Online resizing not supported with bigalloc");
778 			err = -EOPNOTSUPP;
779 			goto group_add_out;
780 		}
781 
782 		err = mnt_want_write_file(filp);
783 		if (err)
784 			goto group_add_out;
785 
786 		err = ext4_group_add(sb, &input);
787 		if (EXT4_SB(sb)->s_journal) {
788 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
789 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
790 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
791 		}
792 		if (err == 0)
793 			err = err2;
794 		mnt_drop_write_file(filp);
795 		if (!err && ext4_has_group_desc_csum(sb) &&
796 		    test_opt(sb, INIT_INODE_TABLE))
797 			err = ext4_register_li_request(sb, input.group);
798 group_add_out:
799 		ext4_resize_end(sb);
800 		return err;
801 	}
802 
803 	case EXT4_IOC_MIGRATE:
804 	{
805 		int err;
806 		if (!inode_owner_or_capable(inode))
807 			return -EACCES;
808 
809 		err = mnt_want_write_file(filp);
810 		if (err)
811 			return err;
812 		/*
813 		 * inode_mutex prevent write and truncate on the file.
814 		 * Read still goes through. We take i_data_sem in
815 		 * ext4_ext_swap_inode_data before we switch the
816 		 * inode format to prevent read.
817 		 */
818 		inode_lock((inode));
819 		err = ext4_ext_migrate(inode);
820 		inode_unlock((inode));
821 		mnt_drop_write_file(filp);
822 		return err;
823 	}
824 
825 	case EXT4_IOC_ALLOC_DA_BLKS:
826 	{
827 		int err;
828 		if (!inode_owner_or_capable(inode))
829 			return -EACCES;
830 
831 		err = mnt_want_write_file(filp);
832 		if (err)
833 			return err;
834 		err = ext4_alloc_da_blocks(inode);
835 		mnt_drop_write_file(filp);
836 		return err;
837 	}
838 
839 	case EXT4_IOC_SWAP_BOOT:
840 	{
841 		int err;
842 		if (!(filp->f_mode & FMODE_WRITE))
843 			return -EBADF;
844 		err = mnt_want_write_file(filp);
845 		if (err)
846 			return err;
847 		err = swap_inode_boot_loader(sb, inode);
848 		mnt_drop_write_file(filp);
849 		return err;
850 	}
851 
852 	case EXT4_IOC_RESIZE_FS: {
853 		ext4_fsblk_t n_blocks_count;
854 		int err = 0, err2 = 0;
855 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
856 
857 		if (ext4_has_feature_bigalloc(sb)) {
858 			ext4_msg(sb, KERN_ERR,
859 				 "Online resizing not (yet) supported with bigalloc");
860 			return -EOPNOTSUPP;
861 		}
862 
863 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
864 				   sizeof(__u64))) {
865 			return -EFAULT;
866 		}
867 
868 		err = ext4_resize_begin(sb);
869 		if (err)
870 			return err;
871 
872 		err = mnt_want_write_file(filp);
873 		if (err)
874 			goto resizefs_out;
875 
876 		err = ext4_resize_fs(sb, n_blocks_count);
877 		if (EXT4_SB(sb)->s_journal) {
878 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
879 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
880 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
881 		}
882 		if (err == 0)
883 			err = err2;
884 		mnt_drop_write_file(filp);
885 		if (!err && (o_group > EXT4_SB(sb)->s_groups_count) &&
886 		    ext4_has_group_desc_csum(sb) &&
887 		    test_opt(sb, INIT_INODE_TABLE))
888 			err = ext4_register_li_request(sb, o_group);
889 
890 resizefs_out:
891 		ext4_resize_end(sb);
892 		return err;
893 	}
894 
895 	case FITRIM:
896 	{
897 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
898 		struct fstrim_range range;
899 		int ret = 0;
900 
901 		if (!capable(CAP_SYS_ADMIN))
902 			return -EPERM;
903 
904 		if (!blk_queue_discard(q))
905 			return -EOPNOTSUPP;
906 
907 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
908 		    sizeof(range)))
909 			return -EFAULT;
910 
911 		range.minlen = max((unsigned int)range.minlen,
912 				   q->limits.discard_granularity);
913 		ret = ext4_trim_fs(sb, &range);
914 		if (ret < 0)
915 			return ret;
916 
917 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
918 		    sizeof(range)))
919 			return -EFAULT;
920 
921 		return 0;
922 	}
923 	case EXT4_IOC_PRECACHE_EXTENTS:
924 		return ext4_ext_precache(inode);
925 
926 	case EXT4_IOC_SET_ENCRYPTION_POLICY:
927 		if (!ext4_has_feature_encrypt(sb))
928 			return -EOPNOTSUPP;
929 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
930 
931 	case EXT4_IOC_GET_ENCRYPTION_PWSALT: {
932 #ifdef CONFIG_EXT4_FS_ENCRYPTION
933 		int err, err2;
934 		struct ext4_sb_info *sbi = EXT4_SB(sb);
935 		handle_t *handle;
936 
937 		if (!ext4_has_feature_encrypt(sb))
938 			return -EOPNOTSUPP;
939 		if (uuid_is_zero(sbi->s_es->s_encrypt_pw_salt)) {
940 			err = mnt_want_write_file(filp);
941 			if (err)
942 				return err;
943 			handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
944 			if (IS_ERR(handle)) {
945 				err = PTR_ERR(handle);
946 				goto pwsalt_err_exit;
947 			}
948 			err = ext4_journal_get_write_access(handle, sbi->s_sbh);
949 			if (err)
950 				goto pwsalt_err_journal;
951 			generate_random_uuid(sbi->s_es->s_encrypt_pw_salt);
952 			err = ext4_handle_dirty_metadata(handle, NULL,
953 							 sbi->s_sbh);
954 		pwsalt_err_journal:
955 			err2 = ext4_journal_stop(handle);
956 			if (err2 && !err)
957 				err = err2;
958 		pwsalt_err_exit:
959 			mnt_drop_write_file(filp);
960 			if (err)
961 				return err;
962 		}
963 		if (copy_to_user((void __user *) arg,
964 				 sbi->s_es->s_encrypt_pw_salt, 16))
965 			return -EFAULT;
966 		return 0;
967 #else
968 		return -EOPNOTSUPP;
969 #endif
970 	}
971 	case EXT4_IOC_GET_ENCRYPTION_POLICY:
972 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
973 
974 	case EXT4_IOC_FSGETXATTR:
975 	{
976 		struct fsxattr fa;
977 
978 		memset(&fa, 0, sizeof(struct fsxattr));
979 		fa.fsx_xflags = ext4_iflags_to_xflags(ei->i_flags & EXT4_FL_USER_VISIBLE);
980 
981 		if (ext4_has_feature_project(inode->i_sb)) {
982 			fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
983 				EXT4_I(inode)->i_projid);
984 		}
985 
986 		if (copy_to_user((struct fsxattr __user *)arg,
987 				 &fa, sizeof(fa)))
988 			return -EFAULT;
989 		return 0;
990 	}
991 	case EXT4_IOC_FSSETXATTR:
992 	{
993 		struct fsxattr fa;
994 		int err;
995 
996 		if (copy_from_user(&fa, (struct fsxattr __user *)arg,
997 				   sizeof(fa)))
998 			return -EFAULT;
999 
1000 		/* Make sure caller has proper permission */
1001 		if (!inode_owner_or_capable(inode))
1002 			return -EACCES;
1003 
1004 		if (fa.fsx_xflags & ~EXT4_SUPPORTED_FS_XFLAGS)
1005 			return -EOPNOTSUPP;
1006 
1007 		flags = ext4_xflags_to_iflags(fa.fsx_xflags);
1008 		if (ext4_mask_flags(inode->i_mode, flags) != flags)
1009 			return -EOPNOTSUPP;
1010 
1011 		err = mnt_want_write_file(filp);
1012 		if (err)
1013 			return err;
1014 
1015 		inode_lock(inode);
1016 		flags = (ei->i_flags & ~EXT4_FL_XFLAG_VISIBLE) |
1017 			 (flags & EXT4_FL_XFLAG_VISIBLE);
1018 		err = ext4_ioctl_setflags(inode, flags);
1019 		inode_unlock(inode);
1020 		mnt_drop_write_file(filp);
1021 		if (err)
1022 			return err;
1023 
1024 		err = ext4_ioctl_setproject(filp, fa.fsx_projid);
1025 		if (err)
1026 			return err;
1027 
1028 		return 0;
1029 	}
1030 	case EXT4_IOC_SHUTDOWN:
1031 		return ext4_shutdown(sb, arg);
1032 	default:
1033 		return -ENOTTY;
1034 	}
1035 }
1036 
1037 #ifdef CONFIG_COMPAT
1038 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1039 {
1040 	/* These are just misnamed, they actually get/put from/to user an int */
1041 	switch (cmd) {
1042 	case EXT4_IOC32_GETFLAGS:
1043 		cmd = EXT4_IOC_GETFLAGS;
1044 		break;
1045 	case EXT4_IOC32_SETFLAGS:
1046 		cmd = EXT4_IOC_SETFLAGS;
1047 		break;
1048 	case EXT4_IOC32_GETVERSION:
1049 		cmd = EXT4_IOC_GETVERSION;
1050 		break;
1051 	case EXT4_IOC32_SETVERSION:
1052 		cmd = EXT4_IOC_SETVERSION;
1053 		break;
1054 	case EXT4_IOC32_GROUP_EXTEND:
1055 		cmd = EXT4_IOC_GROUP_EXTEND;
1056 		break;
1057 	case EXT4_IOC32_GETVERSION_OLD:
1058 		cmd = EXT4_IOC_GETVERSION_OLD;
1059 		break;
1060 	case EXT4_IOC32_SETVERSION_OLD:
1061 		cmd = EXT4_IOC_SETVERSION_OLD;
1062 		break;
1063 	case EXT4_IOC32_GETRSVSZ:
1064 		cmd = EXT4_IOC_GETRSVSZ;
1065 		break;
1066 	case EXT4_IOC32_SETRSVSZ:
1067 		cmd = EXT4_IOC_SETRSVSZ;
1068 		break;
1069 	case EXT4_IOC32_GROUP_ADD: {
1070 		struct compat_ext4_new_group_input __user *uinput;
1071 		struct ext4_new_group_input input;
1072 		mm_segment_t old_fs;
1073 		int err;
1074 
1075 		uinput = compat_ptr(arg);
1076 		err = get_user(input.group, &uinput->group);
1077 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1078 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1079 		err |= get_user(input.inode_table, &uinput->inode_table);
1080 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1081 		err |= get_user(input.reserved_blocks,
1082 				&uinput->reserved_blocks);
1083 		if (err)
1084 			return -EFAULT;
1085 		old_fs = get_fs();
1086 		set_fs(KERNEL_DS);
1087 		err = ext4_ioctl(file, EXT4_IOC_GROUP_ADD,
1088 				 (unsigned long) &input);
1089 		set_fs(old_fs);
1090 		return err;
1091 	}
1092 	case EXT4_IOC_MOVE_EXT:
1093 	case EXT4_IOC_RESIZE_FS:
1094 	case EXT4_IOC_PRECACHE_EXTENTS:
1095 	case EXT4_IOC_SET_ENCRYPTION_POLICY:
1096 	case EXT4_IOC_GET_ENCRYPTION_PWSALT:
1097 	case EXT4_IOC_GET_ENCRYPTION_POLICY:
1098 	case EXT4_IOC_SHUTDOWN:
1099 	case FS_IOC_GETFSMAP:
1100 		break;
1101 	default:
1102 		return -ENOIOCTLCMD;
1103 	}
1104 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1105 }
1106 #endif
1107