xref: /openbmc/linux/fs/ext4/ioctl.c (revision e145b35b)
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 	swap(inode1->i_flags, inode2->i_flags);
68 	swap(inode1->i_version, inode2->i_version);
69 	swap(inode1->i_blocks, inode2->i_blocks);
70 	swap(inode1->i_bytes, inode2->i_bytes);
71 	swap(inode1->i_atime, inode2->i_atime);
72 	swap(inode1->i_mtime, inode2->i_mtime);
73 
74 	memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
75 	swap(ei1->i_flags, ei2->i_flags);
76 	swap(ei1->i_disksize, ei2->i_disksize);
77 	ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
78 	ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
79 
80 	isize = i_size_read(inode1);
81 	i_size_write(inode1, i_size_read(inode2));
82 	i_size_write(inode2, isize);
83 }
84 
85 /**
86  * Swap the information from the given @inode and the inode
87  * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
88  * important fields of the inodes.
89  *
90  * @sb:         the super block of the filesystem
91  * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
92  *
93  */
94 static long swap_inode_boot_loader(struct super_block *sb,
95 				struct inode *inode)
96 {
97 	handle_t *handle;
98 	int err;
99 	struct inode *inode_bl;
100 	struct ext4_inode_info *ei_bl;
101 	struct ext4_sb_info *sbi = EXT4_SB(sb);
102 
103 	if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode))
104 		return -EINVAL;
105 
106 	if (!inode_owner_or_capable(inode) || !capable(CAP_SYS_ADMIN))
107 		return -EPERM;
108 
109 	inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO);
110 	if (IS_ERR(inode_bl))
111 		return PTR_ERR(inode_bl);
112 	ei_bl = EXT4_I(inode_bl);
113 
114 	filemap_flush(inode->i_mapping);
115 	filemap_flush(inode_bl->i_mapping);
116 
117 	/* Protect orig inodes against a truncate and make sure,
118 	 * that only 1 swap_inode_boot_loader is running. */
119 	lock_two_nondirectories(inode, inode_bl);
120 
121 	truncate_inode_pages(&inode->i_data, 0);
122 	truncate_inode_pages(&inode_bl->i_data, 0);
123 
124 	/* Wait for all existing dio workers */
125 	ext4_inode_block_unlocked_dio(inode);
126 	ext4_inode_block_unlocked_dio(inode_bl);
127 	inode_dio_wait(inode);
128 	inode_dio_wait(inode_bl);
129 
130 	handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
131 	if (IS_ERR(handle)) {
132 		err = -EINVAL;
133 		goto journal_err_out;
134 	}
135 
136 	/* Protect extent tree against block allocations via delalloc */
137 	ext4_double_down_write_data_sem(inode, inode_bl);
138 
139 	if (inode_bl->i_nlink == 0) {
140 		/* this inode has never been used as a BOOT_LOADER */
141 		set_nlink(inode_bl, 1);
142 		i_uid_write(inode_bl, 0);
143 		i_gid_write(inode_bl, 0);
144 		inode_bl->i_flags = 0;
145 		ei_bl->i_flags = 0;
146 		inode_bl->i_version = 1;
147 		i_size_write(inode_bl, 0);
148 		inode_bl->i_mode = S_IFREG;
149 		if (ext4_has_feature_extents(sb)) {
150 			ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
151 			ext4_ext_tree_init(handle, inode_bl);
152 		} else
153 			memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
154 	}
155 
156 	swap_inode_data(inode, inode_bl);
157 
158 	inode->i_ctime = inode_bl->i_ctime = current_time(inode);
159 
160 	spin_lock(&sbi->s_next_gen_lock);
161 	inode->i_generation = sbi->s_next_generation++;
162 	inode_bl->i_generation = sbi->s_next_generation++;
163 	spin_unlock(&sbi->s_next_gen_lock);
164 
165 	ext4_discard_preallocations(inode);
166 
167 	err = ext4_mark_inode_dirty(handle, inode);
168 	if (err < 0) {
169 		ext4_warning(inode->i_sb,
170 			"couldn't mark inode #%lu dirty (err %d)",
171 			inode->i_ino, err);
172 		/* Revert all changes: */
173 		swap_inode_data(inode, inode_bl);
174 	} else {
175 		err = ext4_mark_inode_dirty(handle, inode_bl);
176 		if (err < 0) {
177 			ext4_warning(inode_bl->i_sb,
178 				"couldn't mark inode #%lu dirty (err %d)",
179 				inode_bl->i_ino, err);
180 			/* Revert all changes: */
181 			swap_inode_data(inode, inode_bl);
182 			ext4_mark_inode_dirty(handle, inode);
183 		}
184 	}
185 	ext4_journal_stop(handle);
186 	ext4_double_up_write_data_sem(inode, inode_bl);
187 
188 journal_err_out:
189 	ext4_inode_resume_unlocked_dio(inode);
190 	ext4_inode_resume_unlocked_dio(inode_bl);
191 	unlock_two_nondirectories(inode, inode_bl);
192 	iput(inode_bl);
193 	return err;
194 }
195 
196 #ifdef CONFIG_EXT4_FS_ENCRYPTION
197 static int uuid_is_zero(__u8 u[16])
198 {
199 	int	i;
200 
201 	for (i = 0; i < 16; i++)
202 		if (u[i])
203 			return 0;
204 	return 1;
205 }
206 #endif
207 
208 static int ext4_ioctl_setflags(struct inode *inode,
209 			       unsigned int flags)
210 {
211 	struct ext4_inode_info *ei = EXT4_I(inode);
212 	handle_t *handle = NULL;
213 	int err = -EPERM, migrate = 0;
214 	struct ext4_iloc iloc;
215 	unsigned int oldflags, mask, i;
216 	unsigned int jflag;
217 
218 	/* Is it quota file? Do not allow user to mess with it */
219 	if (ext4_is_quota_file(inode))
220 		goto flags_out;
221 
222 	oldflags = ei->i_flags;
223 
224 	/* The JOURNAL_DATA flag is modifiable only by root */
225 	jflag = flags & EXT4_JOURNAL_DATA_FL;
226 
227 	/*
228 	 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
229 	 * the relevant capability.
230 	 *
231 	 * This test looks nicer. Thanks to Pauline Middelink
232 	 */
233 	if ((flags ^ oldflags) & (EXT4_APPEND_FL | EXT4_IMMUTABLE_FL)) {
234 		if (!capable(CAP_LINUX_IMMUTABLE))
235 			goto flags_out;
236 	}
237 
238 	/*
239 	 * The JOURNAL_DATA flag can only be changed by
240 	 * the relevant capability.
241 	 */
242 	if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
243 		if (!capable(CAP_SYS_RESOURCE))
244 			goto flags_out;
245 	}
246 	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
247 		migrate = 1;
248 
249 	if (flags & EXT4_EOFBLOCKS_FL) {
250 		/* we don't support adding EOFBLOCKS flag */
251 		if (!(oldflags & EXT4_EOFBLOCKS_FL)) {
252 			err = -EOPNOTSUPP;
253 			goto flags_out;
254 		}
255 	} else if (oldflags & EXT4_EOFBLOCKS_FL) {
256 		err = ext4_truncate(inode);
257 		if (err)
258 			goto flags_out;
259 	}
260 
261 	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
262 	if (IS_ERR(handle)) {
263 		err = PTR_ERR(handle);
264 		goto flags_out;
265 	}
266 	if (IS_SYNC(inode))
267 		ext4_handle_sync(handle);
268 	err = ext4_reserve_inode_write(handle, inode, &iloc);
269 	if (err)
270 		goto flags_err;
271 
272 	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
273 		if (!(mask & EXT4_FL_USER_MODIFIABLE))
274 			continue;
275 		/* These flags get special treatment later */
276 		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
277 			continue;
278 		if (mask & flags)
279 			ext4_set_inode_flag(inode, i);
280 		else
281 			ext4_clear_inode_flag(inode, i);
282 	}
283 
284 	ext4_set_inode_flags(inode);
285 	inode->i_ctime = current_time(inode);
286 
287 	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
288 flags_err:
289 	ext4_journal_stop(handle);
290 	if (err)
291 		goto flags_out;
292 
293 	if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL))
294 		err = ext4_change_inode_journal_flag(inode, jflag);
295 	if (err)
296 		goto flags_out;
297 	if (migrate) {
298 		if (flags & EXT4_EXTENTS_FL)
299 			err = ext4_ext_migrate(inode);
300 		else
301 			err = ext4_ind_migrate(inode);
302 	}
303 
304 flags_out:
305 	return err;
306 }
307 
308 #ifdef CONFIG_QUOTA
309 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
310 {
311 	struct inode *inode = file_inode(filp);
312 	struct super_block *sb = inode->i_sb;
313 	struct ext4_inode_info *ei = EXT4_I(inode);
314 	int err, rc;
315 	handle_t *handle;
316 	kprojid_t kprojid;
317 	struct ext4_iloc iloc;
318 	struct ext4_inode *raw_inode;
319 	struct dquot *transfer_to[MAXQUOTAS] = { };
320 
321 	if (!ext4_has_feature_project(sb)) {
322 		if (projid != EXT4_DEF_PROJID)
323 			return -EOPNOTSUPP;
324 		else
325 			return 0;
326 	}
327 
328 	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
329 		return -EOPNOTSUPP;
330 
331 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
332 
333 	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
334 		return 0;
335 
336 	err = mnt_want_write_file(filp);
337 	if (err)
338 		return err;
339 
340 	err = -EPERM;
341 	inode_lock(inode);
342 	/* Is it quota file? Do not allow user to mess with it */
343 	if (ext4_is_quota_file(inode))
344 		goto out_unlock;
345 
346 	err = ext4_get_inode_loc(inode, &iloc);
347 	if (err)
348 		goto out_unlock;
349 
350 	raw_inode = ext4_raw_inode(&iloc);
351 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
352 		err = ext4_expand_extra_isize(inode,
353 					      EXT4_SB(sb)->s_want_extra_isize,
354 					      &iloc);
355 		if (err)
356 			goto out_unlock;
357 	} else {
358 		brelse(iloc.bh);
359 	}
360 
361 	dquot_initialize(inode);
362 
363 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
364 		EXT4_QUOTA_INIT_BLOCKS(sb) +
365 		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
366 	if (IS_ERR(handle)) {
367 		err = PTR_ERR(handle);
368 		goto out_unlock;
369 	}
370 
371 	err = ext4_reserve_inode_write(handle, inode, &iloc);
372 	if (err)
373 		goto out_stop;
374 
375 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
376 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
377 
378 		/* __dquot_transfer() calls back ext4_get_inode_usage() which
379 		 * counts xattr inode references.
380 		 */
381 		down_read(&EXT4_I(inode)->xattr_sem);
382 		err = __dquot_transfer(inode, transfer_to);
383 		up_read(&EXT4_I(inode)->xattr_sem);
384 		dqput(transfer_to[PRJQUOTA]);
385 		if (err)
386 			goto out_dirty;
387 	}
388 
389 	EXT4_I(inode)->i_projid = kprojid;
390 	inode->i_ctime = current_time(inode);
391 out_dirty:
392 	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
393 	if (!err)
394 		err = rc;
395 out_stop:
396 	ext4_journal_stop(handle);
397 out_unlock:
398 	inode_unlock(inode);
399 	mnt_drop_write_file(filp);
400 	return err;
401 }
402 #else
403 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
404 {
405 	if (projid != EXT4_DEF_PROJID)
406 		return -EOPNOTSUPP;
407 	return 0;
408 }
409 #endif
410 
411 /* Transfer internal flags to xflags */
412 static inline __u32 ext4_iflags_to_xflags(unsigned long iflags)
413 {
414 	__u32 xflags = 0;
415 
416 	if (iflags & EXT4_SYNC_FL)
417 		xflags |= FS_XFLAG_SYNC;
418 	if (iflags & EXT4_IMMUTABLE_FL)
419 		xflags |= FS_XFLAG_IMMUTABLE;
420 	if (iflags & EXT4_APPEND_FL)
421 		xflags |= FS_XFLAG_APPEND;
422 	if (iflags & EXT4_NODUMP_FL)
423 		xflags |= FS_XFLAG_NODUMP;
424 	if (iflags & EXT4_NOATIME_FL)
425 		xflags |= FS_XFLAG_NOATIME;
426 	if (iflags & EXT4_PROJINHERIT_FL)
427 		xflags |= FS_XFLAG_PROJINHERIT;
428 	return xflags;
429 }
430 
431 #define EXT4_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
432 				  FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
433 				  FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
434 
435 /* Transfer xflags flags to internal */
436 static inline unsigned long ext4_xflags_to_iflags(__u32 xflags)
437 {
438 	unsigned long iflags = 0;
439 
440 	if (xflags & FS_XFLAG_SYNC)
441 		iflags |= EXT4_SYNC_FL;
442 	if (xflags & FS_XFLAG_IMMUTABLE)
443 		iflags |= EXT4_IMMUTABLE_FL;
444 	if (xflags & FS_XFLAG_APPEND)
445 		iflags |= EXT4_APPEND_FL;
446 	if (xflags & FS_XFLAG_NODUMP)
447 		iflags |= EXT4_NODUMP_FL;
448 	if (xflags & FS_XFLAG_NOATIME)
449 		iflags |= EXT4_NOATIME_FL;
450 	if (xflags & FS_XFLAG_PROJINHERIT)
451 		iflags |= EXT4_PROJINHERIT_FL;
452 
453 	return iflags;
454 }
455 
456 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
457 {
458 	struct ext4_sb_info *sbi = EXT4_SB(sb);
459 	__u32 flags;
460 
461 	if (!capable(CAP_SYS_ADMIN))
462 		return -EPERM;
463 
464 	if (get_user(flags, (__u32 __user *)arg))
465 		return -EFAULT;
466 
467 	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
468 		return -EINVAL;
469 
470 	if (ext4_forced_shutdown(sbi))
471 		return 0;
472 
473 	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
474 
475 	switch (flags) {
476 	case EXT4_GOING_FLAGS_DEFAULT:
477 		freeze_bdev(sb->s_bdev);
478 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
479 		thaw_bdev(sb->s_bdev, sb);
480 		break;
481 	case EXT4_GOING_FLAGS_LOGFLUSH:
482 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
483 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
484 			(void) ext4_force_commit(sb);
485 			jbd2_journal_abort(sbi->s_journal, 0);
486 		}
487 		break;
488 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
489 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
490 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
491 			msleep(100);
492 			jbd2_journal_abort(sbi->s_journal, 0);
493 		}
494 		break;
495 	default:
496 		return -EINVAL;
497 	}
498 	clear_opt(sb, DISCARD);
499 	return 0;
500 }
501 
502 struct getfsmap_info {
503 	struct super_block	*gi_sb;
504 	struct fsmap_head __user *gi_data;
505 	unsigned int		gi_idx;
506 	__u32			gi_last_flags;
507 };
508 
509 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
510 {
511 	struct getfsmap_info *info = priv;
512 	struct fsmap fm;
513 
514 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
515 
516 	info->gi_last_flags = xfm->fmr_flags;
517 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
518 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
519 			sizeof(struct fsmap)))
520 		return -EFAULT;
521 
522 	return 0;
523 }
524 
525 static int ext4_ioc_getfsmap(struct super_block *sb,
526 			     struct fsmap_head __user *arg)
527 {
528 	struct getfsmap_info info = {0};
529 	struct ext4_fsmap_head xhead = {0};
530 	struct fsmap_head head;
531 	bool aborted = false;
532 	int error;
533 
534 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
535 		return -EFAULT;
536 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
537 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
538 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
539 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
540 		       sizeof(head.fmh_keys[1].fmr_reserved)))
541 		return -EINVAL;
542 	/*
543 	 * ext4 doesn't report file extents at all, so the only valid
544 	 * file offsets are the magic ones (all zeroes or all ones).
545 	 */
546 	if (head.fmh_keys[0].fmr_offset ||
547 	    (head.fmh_keys[1].fmr_offset != 0 &&
548 	     head.fmh_keys[1].fmr_offset != -1ULL))
549 		return -EINVAL;
550 
551 	xhead.fmh_iflags = head.fmh_iflags;
552 	xhead.fmh_count = head.fmh_count;
553 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
554 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
555 
556 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
557 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
558 
559 	info.gi_sb = sb;
560 	info.gi_data = arg;
561 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
562 	if (error == EXT4_QUERY_RANGE_ABORT) {
563 		error = 0;
564 		aborted = true;
565 	} else if (error)
566 		return error;
567 
568 	/* If we didn't abort, set the "last" flag in the last fmx */
569 	if (!aborted && info.gi_idx) {
570 		info.gi_last_flags |= FMR_OF_LAST;
571 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
572 				 &info.gi_last_flags,
573 				 sizeof(info.gi_last_flags)))
574 			return -EFAULT;
575 	}
576 
577 	/* copy back header */
578 	head.fmh_entries = xhead.fmh_entries;
579 	head.fmh_oflags = xhead.fmh_oflags;
580 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
581 		return -EFAULT;
582 
583 	return 0;
584 }
585 
586 static long ext4_ioctl_group_add(struct file *file,
587 				 struct ext4_new_group_data *input)
588 {
589 	struct super_block *sb = file_inode(file)->i_sb;
590 	int err, err2=0;
591 
592 	err = ext4_resize_begin(sb);
593 	if (err)
594 		return err;
595 
596 	if (ext4_has_feature_bigalloc(sb)) {
597 		ext4_msg(sb, KERN_ERR,
598 			 "Online resizing not supported with bigalloc");
599 		err = -EOPNOTSUPP;
600 		goto group_add_out;
601 	}
602 
603 	err = mnt_want_write_file(file);
604 	if (err)
605 		goto group_add_out;
606 
607 	err = ext4_group_add(sb, input);
608 	if (EXT4_SB(sb)->s_journal) {
609 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
610 		err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
611 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
612 	}
613 	if (err == 0)
614 		err = err2;
615 	mnt_drop_write_file(file);
616 	if (!err && ext4_has_group_desc_csum(sb) &&
617 	    test_opt(sb, INIT_INODE_TABLE))
618 		err = ext4_register_li_request(sb, input->group);
619 group_add_out:
620 	ext4_resize_end(sb);
621 	return err;
622 }
623 
624 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
625 {
626 	struct inode *inode = file_inode(filp);
627 	struct super_block *sb = inode->i_sb;
628 	struct ext4_inode_info *ei = EXT4_I(inode);
629 	unsigned int flags;
630 
631 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
632 
633 	switch (cmd) {
634 	case FS_IOC_GETFSMAP:
635 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
636 	case EXT4_IOC_GETFLAGS:
637 		flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
638 		return put_user(flags, (int __user *) arg);
639 	case EXT4_IOC_SETFLAGS: {
640 		int err;
641 
642 		if (!inode_owner_or_capable(inode))
643 			return -EACCES;
644 
645 		if (get_user(flags, (int __user *) arg))
646 			return -EFAULT;
647 
648 		if (flags & ~EXT4_FL_USER_VISIBLE)
649 			return -EOPNOTSUPP;
650 		/*
651 		 * chattr(1) grabs flags via GETFLAGS, modifies the result and
652 		 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
653 		 * more restrictive than just silently masking off visible but
654 		 * not settable flags as we always did.
655 		 */
656 		flags &= EXT4_FL_USER_MODIFIABLE;
657 		if (ext4_mask_flags(inode->i_mode, flags) != flags)
658 			return -EOPNOTSUPP;
659 
660 		err = mnt_want_write_file(filp);
661 		if (err)
662 			return err;
663 
664 		inode_lock(inode);
665 		err = ext4_ioctl_setflags(inode, flags);
666 		inode_unlock(inode);
667 		mnt_drop_write_file(filp);
668 		return err;
669 	}
670 	case EXT4_IOC_GETVERSION:
671 	case EXT4_IOC_GETVERSION_OLD:
672 		return put_user(inode->i_generation, (int __user *) arg);
673 	case EXT4_IOC_SETVERSION:
674 	case EXT4_IOC_SETVERSION_OLD: {
675 		handle_t *handle;
676 		struct ext4_iloc iloc;
677 		__u32 generation;
678 		int err;
679 
680 		if (!inode_owner_or_capable(inode))
681 			return -EPERM;
682 
683 		if (ext4_has_metadata_csum(inode->i_sb)) {
684 			ext4_warning(sb, "Setting inode version is not "
685 				     "supported with metadata_csum enabled.");
686 			return -ENOTTY;
687 		}
688 
689 		err = mnt_want_write_file(filp);
690 		if (err)
691 			return err;
692 		if (get_user(generation, (int __user *) arg)) {
693 			err = -EFAULT;
694 			goto setversion_out;
695 		}
696 
697 		inode_lock(inode);
698 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
699 		if (IS_ERR(handle)) {
700 			err = PTR_ERR(handle);
701 			goto unlock_out;
702 		}
703 		err = ext4_reserve_inode_write(handle, inode, &iloc);
704 		if (err == 0) {
705 			inode->i_ctime = current_time(inode);
706 			inode->i_generation = generation;
707 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
708 		}
709 		ext4_journal_stop(handle);
710 
711 unlock_out:
712 		inode_unlock(inode);
713 setversion_out:
714 		mnt_drop_write_file(filp);
715 		return err;
716 	}
717 	case EXT4_IOC_GROUP_EXTEND: {
718 		ext4_fsblk_t n_blocks_count;
719 		int err, err2=0;
720 
721 		err = ext4_resize_begin(sb);
722 		if (err)
723 			return err;
724 
725 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
726 			err = -EFAULT;
727 			goto group_extend_out;
728 		}
729 
730 		if (ext4_has_feature_bigalloc(sb)) {
731 			ext4_msg(sb, KERN_ERR,
732 				 "Online resizing not supported with bigalloc");
733 			err = -EOPNOTSUPP;
734 			goto group_extend_out;
735 		}
736 
737 		err = mnt_want_write_file(filp);
738 		if (err)
739 			goto group_extend_out;
740 
741 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
742 		if (EXT4_SB(sb)->s_journal) {
743 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
744 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
745 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
746 		}
747 		if (err == 0)
748 			err = err2;
749 		mnt_drop_write_file(filp);
750 group_extend_out:
751 		ext4_resize_end(sb);
752 		return err;
753 	}
754 
755 	case EXT4_IOC_MOVE_EXT: {
756 		struct move_extent me;
757 		struct fd donor;
758 		int err;
759 
760 		if (!(filp->f_mode & FMODE_READ) ||
761 		    !(filp->f_mode & FMODE_WRITE))
762 			return -EBADF;
763 
764 		if (copy_from_user(&me,
765 			(struct move_extent __user *)arg, sizeof(me)))
766 			return -EFAULT;
767 		me.moved_len = 0;
768 
769 		donor = fdget(me.donor_fd);
770 		if (!donor.file)
771 			return -EBADF;
772 
773 		if (!(donor.file->f_mode & FMODE_WRITE)) {
774 			err = -EBADF;
775 			goto mext_out;
776 		}
777 
778 		if (ext4_has_feature_bigalloc(sb)) {
779 			ext4_msg(sb, KERN_ERR,
780 				 "Online defrag not supported with bigalloc");
781 			err = -EOPNOTSUPP;
782 			goto mext_out;
783 		} else if (IS_DAX(inode)) {
784 			ext4_msg(sb, KERN_ERR,
785 				 "Online defrag not supported with DAX");
786 			err = -EOPNOTSUPP;
787 			goto mext_out;
788 		}
789 
790 		err = mnt_want_write_file(filp);
791 		if (err)
792 			goto mext_out;
793 
794 		err = ext4_move_extents(filp, donor.file, me.orig_start,
795 					me.donor_start, me.len, &me.moved_len);
796 		mnt_drop_write_file(filp);
797 
798 		if (copy_to_user((struct move_extent __user *)arg,
799 				 &me, sizeof(me)))
800 			err = -EFAULT;
801 mext_out:
802 		fdput(donor);
803 		return err;
804 	}
805 
806 	case EXT4_IOC_GROUP_ADD: {
807 		struct ext4_new_group_data input;
808 
809 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
810 				sizeof(input)))
811 			return -EFAULT;
812 
813 		return ext4_ioctl_group_add(filp, &input);
814 	}
815 
816 	case EXT4_IOC_MIGRATE:
817 	{
818 		int err;
819 		if (!inode_owner_or_capable(inode))
820 			return -EACCES;
821 
822 		err = mnt_want_write_file(filp);
823 		if (err)
824 			return err;
825 		/*
826 		 * inode_mutex prevent write and truncate on the file.
827 		 * Read still goes through. We take i_data_sem in
828 		 * ext4_ext_swap_inode_data before we switch the
829 		 * inode format to prevent read.
830 		 */
831 		inode_lock((inode));
832 		err = ext4_ext_migrate(inode);
833 		inode_unlock((inode));
834 		mnt_drop_write_file(filp);
835 		return err;
836 	}
837 
838 	case EXT4_IOC_ALLOC_DA_BLKS:
839 	{
840 		int err;
841 		if (!inode_owner_or_capable(inode))
842 			return -EACCES;
843 
844 		err = mnt_want_write_file(filp);
845 		if (err)
846 			return err;
847 		err = ext4_alloc_da_blocks(inode);
848 		mnt_drop_write_file(filp);
849 		return err;
850 	}
851 
852 	case EXT4_IOC_SWAP_BOOT:
853 	{
854 		int err;
855 		if (!(filp->f_mode & FMODE_WRITE))
856 			return -EBADF;
857 		err = mnt_want_write_file(filp);
858 		if (err)
859 			return err;
860 		err = swap_inode_boot_loader(sb, inode);
861 		mnt_drop_write_file(filp);
862 		return err;
863 	}
864 
865 	case EXT4_IOC_RESIZE_FS: {
866 		ext4_fsblk_t n_blocks_count;
867 		int err = 0, err2 = 0;
868 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
869 
870 		if (ext4_has_feature_bigalloc(sb)) {
871 			ext4_msg(sb, KERN_ERR,
872 				 "Online resizing not (yet) supported with bigalloc");
873 			return -EOPNOTSUPP;
874 		}
875 
876 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
877 				   sizeof(__u64))) {
878 			return -EFAULT;
879 		}
880 
881 		err = ext4_resize_begin(sb);
882 		if (err)
883 			return err;
884 
885 		err = mnt_want_write_file(filp);
886 		if (err)
887 			goto resizefs_out;
888 
889 		err = ext4_resize_fs(sb, n_blocks_count);
890 		if (EXT4_SB(sb)->s_journal) {
891 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
892 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
893 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
894 		}
895 		if (err == 0)
896 			err = err2;
897 		mnt_drop_write_file(filp);
898 		if (!err && (o_group > EXT4_SB(sb)->s_groups_count) &&
899 		    ext4_has_group_desc_csum(sb) &&
900 		    test_opt(sb, INIT_INODE_TABLE))
901 			err = ext4_register_li_request(sb, o_group);
902 
903 resizefs_out:
904 		ext4_resize_end(sb);
905 		return err;
906 	}
907 
908 	case FITRIM:
909 	{
910 		struct request_queue *q = bdev_get_queue(sb->s_bdev);
911 		struct fstrim_range range;
912 		int ret = 0;
913 
914 		if (!capable(CAP_SYS_ADMIN))
915 			return -EPERM;
916 
917 		if (!blk_queue_discard(q))
918 			return -EOPNOTSUPP;
919 
920 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
921 		    sizeof(range)))
922 			return -EFAULT;
923 
924 		range.minlen = max((unsigned int)range.minlen,
925 				   q->limits.discard_granularity);
926 		ret = ext4_trim_fs(sb, &range);
927 		if (ret < 0)
928 			return ret;
929 
930 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
931 		    sizeof(range)))
932 			return -EFAULT;
933 
934 		return 0;
935 	}
936 	case EXT4_IOC_PRECACHE_EXTENTS:
937 		return ext4_ext_precache(inode);
938 
939 	case EXT4_IOC_SET_ENCRYPTION_POLICY:
940 		if (!ext4_has_feature_encrypt(sb))
941 			return -EOPNOTSUPP;
942 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
943 
944 	case EXT4_IOC_GET_ENCRYPTION_PWSALT: {
945 #ifdef CONFIG_EXT4_FS_ENCRYPTION
946 		int err, err2;
947 		struct ext4_sb_info *sbi = EXT4_SB(sb);
948 		handle_t *handle;
949 
950 		if (!ext4_has_feature_encrypt(sb))
951 			return -EOPNOTSUPP;
952 		if (uuid_is_zero(sbi->s_es->s_encrypt_pw_salt)) {
953 			err = mnt_want_write_file(filp);
954 			if (err)
955 				return err;
956 			handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
957 			if (IS_ERR(handle)) {
958 				err = PTR_ERR(handle);
959 				goto pwsalt_err_exit;
960 			}
961 			err = ext4_journal_get_write_access(handle, sbi->s_sbh);
962 			if (err)
963 				goto pwsalt_err_journal;
964 			generate_random_uuid(sbi->s_es->s_encrypt_pw_salt);
965 			err = ext4_handle_dirty_metadata(handle, NULL,
966 							 sbi->s_sbh);
967 		pwsalt_err_journal:
968 			err2 = ext4_journal_stop(handle);
969 			if (err2 && !err)
970 				err = err2;
971 		pwsalt_err_exit:
972 			mnt_drop_write_file(filp);
973 			if (err)
974 				return err;
975 		}
976 		if (copy_to_user((void __user *) arg,
977 				 sbi->s_es->s_encrypt_pw_salt, 16))
978 			return -EFAULT;
979 		return 0;
980 #else
981 		return -EOPNOTSUPP;
982 #endif
983 	}
984 	case EXT4_IOC_GET_ENCRYPTION_POLICY:
985 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
986 
987 	case EXT4_IOC_FSGETXATTR:
988 	{
989 		struct fsxattr fa;
990 
991 		memset(&fa, 0, sizeof(struct fsxattr));
992 		fa.fsx_xflags = ext4_iflags_to_xflags(ei->i_flags & EXT4_FL_USER_VISIBLE);
993 
994 		if (ext4_has_feature_project(inode->i_sb)) {
995 			fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
996 				EXT4_I(inode)->i_projid);
997 		}
998 
999 		if (copy_to_user((struct fsxattr __user *)arg,
1000 				 &fa, sizeof(fa)))
1001 			return -EFAULT;
1002 		return 0;
1003 	}
1004 	case EXT4_IOC_FSSETXATTR:
1005 	{
1006 		struct fsxattr fa;
1007 		int err;
1008 
1009 		if (copy_from_user(&fa, (struct fsxattr __user *)arg,
1010 				   sizeof(fa)))
1011 			return -EFAULT;
1012 
1013 		/* Make sure caller has proper permission */
1014 		if (!inode_owner_or_capable(inode))
1015 			return -EACCES;
1016 
1017 		if (fa.fsx_xflags & ~EXT4_SUPPORTED_FS_XFLAGS)
1018 			return -EOPNOTSUPP;
1019 
1020 		flags = ext4_xflags_to_iflags(fa.fsx_xflags);
1021 		if (ext4_mask_flags(inode->i_mode, flags) != flags)
1022 			return -EOPNOTSUPP;
1023 
1024 		err = mnt_want_write_file(filp);
1025 		if (err)
1026 			return err;
1027 
1028 		inode_lock(inode);
1029 		flags = (ei->i_flags & ~EXT4_FL_XFLAG_VISIBLE) |
1030 			 (flags & EXT4_FL_XFLAG_VISIBLE);
1031 		err = ext4_ioctl_setflags(inode, flags);
1032 		inode_unlock(inode);
1033 		mnt_drop_write_file(filp);
1034 		if (err)
1035 			return err;
1036 
1037 		err = ext4_ioctl_setproject(filp, fa.fsx_projid);
1038 		if (err)
1039 			return err;
1040 
1041 		return 0;
1042 	}
1043 	case EXT4_IOC_SHUTDOWN:
1044 		return ext4_shutdown(sb, arg);
1045 	default:
1046 		return -ENOTTY;
1047 	}
1048 }
1049 
1050 #ifdef CONFIG_COMPAT
1051 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1052 {
1053 	/* These are just misnamed, they actually get/put from/to user an int */
1054 	switch (cmd) {
1055 	case EXT4_IOC32_GETFLAGS:
1056 		cmd = EXT4_IOC_GETFLAGS;
1057 		break;
1058 	case EXT4_IOC32_SETFLAGS:
1059 		cmd = EXT4_IOC_SETFLAGS;
1060 		break;
1061 	case EXT4_IOC32_GETVERSION:
1062 		cmd = EXT4_IOC_GETVERSION;
1063 		break;
1064 	case EXT4_IOC32_SETVERSION:
1065 		cmd = EXT4_IOC_SETVERSION;
1066 		break;
1067 	case EXT4_IOC32_GROUP_EXTEND:
1068 		cmd = EXT4_IOC_GROUP_EXTEND;
1069 		break;
1070 	case EXT4_IOC32_GETVERSION_OLD:
1071 		cmd = EXT4_IOC_GETVERSION_OLD;
1072 		break;
1073 	case EXT4_IOC32_SETVERSION_OLD:
1074 		cmd = EXT4_IOC_SETVERSION_OLD;
1075 		break;
1076 	case EXT4_IOC32_GETRSVSZ:
1077 		cmd = EXT4_IOC_GETRSVSZ;
1078 		break;
1079 	case EXT4_IOC32_SETRSVSZ:
1080 		cmd = EXT4_IOC_SETRSVSZ;
1081 		break;
1082 	case EXT4_IOC32_GROUP_ADD: {
1083 		struct compat_ext4_new_group_input __user *uinput;
1084 		struct ext4_new_group_data input;
1085 		int err;
1086 
1087 		uinput = compat_ptr(arg);
1088 		err = get_user(input.group, &uinput->group);
1089 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1090 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1091 		err |= get_user(input.inode_table, &uinput->inode_table);
1092 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1093 		err |= get_user(input.reserved_blocks,
1094 				&uinput->reserved_blocks);
1095 		if (err)
1096 			return -EFAULT;
1097 		return ext4_ioctl_group_add(file, &input);
1098 	}
1099 	case EXT4_IOC_MOVE_EXT:
1100 	case EXT4_IOC_RESIZE_FS:
1101 	case EXT4_IOC_PRECACHE_EXTENTS:
1102 	case EXT4_IOC_SET_ENCRYPTION_POLICY:
1103 	case EXT4_IOC_GET_ENCRYPTION_PWSALT:
1104 	case EXT4_IOC_GET_ENCRYPTION_POLICY:
1105 	case EXT4_IOC_SHUTDOWN:
1106 	case FS_IOC_GETFSMAP:
1107 		break;
1108 	default:
1109 		return -ENOIOCTLCMD;
1110 	}
1111 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1112 }
1113 #endif
1114