xref: /linux/fs/ext4/ioctl.c (revision 65989db7f88456273d0913d8d21f6097fa6aad19)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * linux/fs/ext4/ioctl.c
4  *
5  * Copyright (C) 1993, 1994, 1995
6  * Remy Card (card@masi.ibp.fr)
7  * Laboratoire MASI - Institut Blaise Pascal
8  * Universite Pierre et Marie Curie (Paris VI)
9  */
10 
11 #include <linux/fs.h>
12 #include <linux/capability.h>
13 #include <linux/time.h>
14 #include <linux/compat.h>
15 #include <linux/mount.h>
16 #include <linux/file.h>
17 #include <linux/quotaops.h>
18 #include <linux/random.h>
19 #include <linux/uaccess.h>
20 #include <linux/delay.h>
21 #include <linux/iversion.h>
22 #include <linux/fileattr.h>
23 #include <linux/uuid.h>
24 #include "ext4_jbd2.h"
25 #include "ext4.h"
26 #include <linux/fsmap.h>
27 #include "fsmap.h"
28 #include <trace/events/ext4.h>
29 
30 typedef void ext4_update_sb_callback(struct ext4_sb_info *sbi,
31 				     struct ext4_super_block *es,
32 				     const void *arg);
33 
34 /*
35  * Superblock modification callback function for changing file system
36  * label
37  */
ext4_sb_setlabel(struct ext4_sb_info * sbi,struct ext4_super_block * es,const void * arg)38 static void ext4_sb_setlabel(struct ext4_sb_info *sbi,
39 			     struct ext4_super_block *es, const void *arg)
40 {
41 	/* Sanity check, this should never happen */
42 	BUILD_BUG_ON(sizeof(es->s_volume_name) < EXT4_LABEL_MAX);
43 
44 	memcpy(es->s_volume_name, (char *)arg, EXT4_LABEL_MAX);
45 }
46 
47 /*
48  * Superblock modification callback function for changing file system
49  * UUID.
50  */
ext4_sb_setuuid(struct ext4_sb_info * sbi,struct ext4_super_block * es,const void * arg)51 static void ext4_sb_setuuid(struct ext4_sb_info *sbi,
52 			    struct ext4_super_block *es, const void *arg)
53 {
54 	memcpy(es->s_uuid, (__u8 *)arg, UUID_SIZE);
55 }
56 
57 static
ext4_update_primary_sb(struct super_block * sb,handle_t * handle,ext4_update_sb_callback func,const void * arg)58 int ext4_update_primary_sb(struct super_block *sb, handle_t *handle,
59 			   ext4_update_sb_callback func,
60 			   const void *arg)
61 {
62 	int err = 0;
63 	struct ext4_sb_info *sbi = EXT4_SB(sb);
64 	struct buffer_head *bh = sbi->s_sbh;
65 	struct ext4_super_block *es = sbi->s_es;
66 
67 	trace_ext4_update_sb(sb, bh->b_blocknr, 1);
68 
69 	BUFFER_TRACE(bh, "get_write_access");
70 	err = ext4_journal_get_write_access(handle, sb,
71 					    bh,
72 					    EXT4_JTR_NONE);
73 	if (err)
74 		goto out_err;
75 
76 	lock_buffer(bh);
77 	func(sbi, es, arg);
78 	ext4_superblock_csum_set(sb);
79 	unlock_buffer(bh);
80 
81 	if (buffer_write_io_error(bh) || !buffer_uptodate(bh)) {
82 		ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
83 			 "superblock detected");
84 		clear_buffer_write_io_error(bh);
85 		set_buffer_uptodate(bh);
86 	}
87 
88 	err = ext4_handle_dirty_metadata(handle, NULL, bh);
89 	if (err)
90 		goto out_err;
91 	err = sync_dirty_buffer(bh);
92 out_err:
93 	ext4_std_error(sb, err);
94 	return err;
95 }
96 
97 /*
98  * Update one backup superblock in the group 'grp' using the callback
99  * function 'func' and argument 'arg'. If the handle is NULL the
100  * modification is not journalled.
101  *
102  * Returns: 0 when no modification was done (no superblock in the group)
103  *	    1 when the modification was successful
104  *	   <0 on error
105  */
ext4_update_backup_sb(struct super_block * sb,handle_t * handle,ext4_group_t grp,ext4_update_sb_callback func,const void * arg)106 static int ext4_update_backup_sb(struct super_block *sb,
107 				 handle_t *handle, ext4_group_t grp,
108 				 ext4_update_sb_callback func, const void *arg)
109 {
110 	int err = 0;
111 	ext4_fsblk_t sb_block;
112 	struct buffer_head *bh;
113 	unsigned long offset = 0;
114 	struct ext4_super_block *es;
115 
116 	if (!ext4_bg_has_super(sb, grp))
117 		return 0;
118 
119 	/*
120 	 * For the group 0 there is always 1k padding, so we have
121 	 * either adjust offset, or sb_block depending on blocksize
122 	 */
123 	if (grp == 0) {
124 		sb_block = 1 * EXT4_MIN_BLOCK_SIZE;
125 		offset = do_div(sb_block, sb->s_blocksize);
126 	} else {
127 		sb_block = ext4_group_first_block_no(sb, grp);
128 		offset = 0;
129 	}
130 
131 	trace_ext4_update_sb(sb, sb_block, handle ? 1 : 0);
132 
133 	bh = ext4_sb_bread(sb, sb_block, 0);
134 	if (IS_ERR(bh))
135 		return PTR_ERR(bh);
136 
137 	if (handle) {
138 		BUFFER_TRACE(bh, "get_write_access");
139 		err = ext4_journal_get_write_access(handle, sb,
140 						    bh,
141 						    EXT4_JTR_NONE);
142 		if (err)
143 			goto out_bh;
144 	}
145 
146 	es = (struct ext4_super_block *) (bh->b_data + offset);
147 	lock_buffer(bh);
148 	if (ext4_has_feature_metadata_csum(sb) &&
149 	    es->s_checksum != ext4_superblock_csum(es)) {
150 		ext4_msg(sb, KERN_ERR, "Invalid checksum for backup "
151 		"superblock %llu", sb_block);
152 		unlock_buffer(bh);
153 		goto out_bh;
154 	}
155 	func(EXT4_SB(sb), es, arg);
156 	if (ext4_has_feature_metadata_csum(sb))
157 		es->s_checksum = ext4_superblock_csum(es);
158 	set_buffer_uptodate(bh);
159 	unlock_buffer(bh);
160 
161 	if (handle) {
162 		err = ext4_handle_dirty_metadata(handle, NULL, bh);
163 		if (err)
164 			goto out_bh;
165 	} else {
166 		BUFFER_TRACE(bh, "marking dirty");
167 		mark_buffer_dirty(bh);
168 	}
169 	err = sync_dirty_buffer(bh);
170 
171 out_bh:
172 	brelse(bh);
173 	ext4_std_error(sb, err);
174 	return (err) ? err : 1;
175 }
176 
177 /*
178  * Update primary and backup superblocks using the provided function
179  * func and argument arg.
180  *
181  * Only the primary superblock and at most two backup superblock
182  * modifications are journalled; the rest is modified without journal.
183  * This is safe because e2fsck will re-write them if there is a problem,
184  * and we're very unlikely to ever need more than two backups.
185  */
186 static
ext4_update_superblocks_fn(struct super_block * sb,ext4_update_sb_callback func,const void * arg)187 int ext4_update_superblocks_fn(struct super_block *sb,
188 			       ext4_update_sb_callback func,
189 			       const void *arg)
190 {
191 	handle_t *handle;
192 	ext4_group_t ngroups;
193 	unsigned int three = 1;
194 	unsigned int five = 5;
195 	unsigned int seven = 7;
196 	int err = 0, ret, i;
197 	ext4_group_t grp, primary_grp;
198 	struct ext4_sb_info *sbi = EXT4_SB(sb);
199 
200 	/*
201 	 * We can't update superblocks while the online resize is running
202 	 */
203 	if (test_and_set_bit_lock(EXT4_FLAGS_RESIZING,
204 				  &sbi->s_ext4_flags)) {
205 		ext4_msg(sb, KERN_ERR, "Can't modify superblock while"
206 			 "performing online resize");
207 		return -EBUSY;
208 	}
209 
210 	/*
211 	 * We're only going to update primary superblock and two
212 	 * backup superblocks in this transaction.
213 	 */
214 	handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 3);
215 	if (IS_ERR(handle)) {
216 		err = PTR_ERR(handle);
217 		goto out;
218 	}
219 
220 	/* Update primary superblock */
221 	err = ext4_update_primary_sb(sb, handle, func, arg);
222 	if (err) {
223 		ext4_msg(sb, KERN_ERR, "Failed to update primary "
224 			 "superblock");
225 		goto out_journal;
226 	}
227 
228 	primary_grp = ext4_get_group_number(sb, sbi->s_sbh->b_blocknr);
229 	ngroups = ext4_get_groups_count(sb);
230 
231 	/*
232 	 * Update backup superblocks. We have to start from group 0
233 	 * because it might not be where the primary superblock is
234 	 * if the fs is mounted with -o sb=<backup_sb_block>
235 	 */
236 	i = 0;
237 	grp = 0;
238 	while (grp < ngroups) {
239 		/* Skip primary superblock */
240 		if (grp == primary_grp)
241 			goto next_grp;
242 
243 		ret = ext4_update_backup_sb(sb, handle, grp, func, arg);
244 		if (ret < 0) {
245 			/* Ignore bad checksum; try to update next sb */
246 			if (ret == -EFSBADCRC)
247 				goto next_grp;
248 			err = ret;
249 			goto out_journal;
250 		}
251 
252 		i += ret;
253 		if (handle && i > 1) {
254 			/*
255 			 * We're only journalling primary superblock and
256 			 * two backup superblocks; the rest is not
257 			 * journalled.
258 			 */
259 			err = ext4_journal_stop(handle);
260 			if (err)
261 				goto out;
262 			handle = NULL;
263 		}
264 next_grp:
265 		grp = ext4_list_backups(sb, &three, &five, &seven);
266 	}
267 
268 out_journal:
269 	if (handle) {
270 		ret = ext4_journal_stop(handle);
271 		if (ret && !err)
272 			err = ret;
273 	}
274 out:
275 	clear_bit_unlock(EXT4_FLAGS_RESIZING, &sbi->s_ext4_flags);
276 	smp_mb__after_atomic();
277 	return err ? err : 0;
278 }
279 
280 /*
281  * Swap memory between @a and @b for @len bytes.
282  *
283  * @a:          pointer to first memory area
284  * @b:          pointer to second memory area
285  * @len:        number of bytes to swap
286  *
287  */
memswap(void * a,void * b,size_t len)288 static void memswap(void *a, void *b, size_t len)
289 {
290 	unsigned char *ap, *bp;
291 
292 	ap = (unsigned char *)a;
293 	bp = (unsigned char *)b;
294 	while (len-- > 0) {
295 		swap(*ap, *bp);
296 		ap++;
297 		bp++;
298 	}
299 }
300 
301 /*
302  * Swap i_data and associated attributes between @inode1 and @inode2.
303  * This function is used for the primary swap between inode1 and inode2
304  * and also to revert this primary swap in case of errors.
305  *
306  * Therefore you have to make sure, that calling this method twice
307  * will revert all changes.
308  *
309  * @inode1:     pointer to first inode
310  * @inode2:     pointer to second inode
311  */
swap_inode_data(struct inode * inode1,struct inode * inode2)312 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
313 {
314 	loff_t isize;
315 	struct ext4_inode_info *ei1;
316 	struct ext4_inode_info *ei2;
317 	unsigned long tmp;
318 	struct timespec64 ts1, ts2;
319 
320 	ei1 = EXT4_I(inode1);
321 	ei2 = EXT4_I(inode2);
322 
323 	swap(inode1->i_version, inode2->i_version);
324 
325 	ts1 = inode_get_atime(inode1);
326 	ts2 = inode_get_atime(inode2);
327 	inode_set_atime_to_ts(inode1, ts2);
328 	inode_set_atime_to_ts(inode2, ts1);
329 
330 	ts1 = inode_get_mtime(inode1);
331 	ts2 = inode_get_mtime(inode2);
332 	inode_set_mtime_to_ts(inode1, ts2);
333 	inode_set_mtime_to_ts(inode2, ts1);
334 
335 	memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
336 	tmp = ei1->i_flags & EXT4_FL_SHOULD_SWAP;
337 	ei1->i_flags = (ei2->i_flags & EXT4_FL_SHOULD_SWAP) |
338 		(ei1->i_flags & ~EXT4_FL_SHOULD_SWAP);
339 	ei2->i_flags = tmp | (ei2->i_flags & ~EXT4_FL_SHOULD_SWAP);
340 	swap(ei1->i_disksize, ei2->i_disksize);
341 	ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
342 	ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
343 
344 	isize = i_size_read(inode1);
345 	i_size_write(inode1, i_size_read(inode2));
346 	i_size_write(inode2, isize);
347 }
348 
ext4_reset_inode_seed(struct inode * inode)349 void ext4_reset_inode_seed(struct inode *inode)
350 {
351 	struct ext4_inode_info *ei = EXT4_I(inode);
352 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
353 	__le32 inum = cpu_to_le32(inode->i_ino);
354 	__le32 gen = cpu_to_le32(inode->i_generation);
355 	__u32 csum;
356 
357 	if (!ext4_has_feature_metadata_csum(inode->i_sb))
358 		return;
359 
360 	csum = ext4_chksum(sbi->s_csum_seed, (__u8 *)&inum, sizeof(inum));
361 	ei->i_csum_seed = ext4_chksum(csum, (__u8 *)&gen, sizeof(gen));
362 }
363 
364 /*
365  * Swap the information from the given @inode and the inode
366  * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
367  * important fields of the inodes.
368  *
369  * @sb:         the super block of the filesystem
370  * @idmap:	idmap of the mount the inode was found from
371  * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
372  *
373  */
swap_inode_boot_loader(struct super_block * sb,struct mnt_idmap * idmap,struct inode * inode)374 static long swap_inode_boot_loader(struct super_block *sb,
375 				struct mnt_idmap *idmap,
376 				struct inode *inode)
377 {
378 	handle_t *handle;
379 	int err;
380 	struct inode *inode_bl;
381 	struct ext4_inode_info *ei_bl;
382 	qsize_t size, size_bl, diff;
383 	blkcnt_t blocks;
384 	unsigned short bytes;
385 
386 	inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO,
387 			EXT4_IGET_SPECIAL | EXT4_IGET_BAD);
388 	if (IS_ERR(inode_bl))
389 		return PTR_ERR(inode_bl);
390 	ei_bl = EXT4_I(inode_bl);
391 
392 	/* Protect orig inodes against a truncate and make sure,
393 	 * that only 1 swap_inode_boot_loader is running. */
394 	lock_two_nondirectories(inode, inode_bl);
395 
396 	if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode) ||
397 	    IS_SWAPFILE(inode) || IS_ENCRYPTED(inode) ||
398 	    (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL) ||
399 	    ext4_has_inline_data(inode)) {
400 		err = -EINVAL;
401 		goto journal_err_out;
402 	}
403 
404 	if (IS_RDONLY(inode) || IS_APPEND(inode) || IS_IMMUTABLE(inode) ||
405 	    !inode_owner_or_capable(idmap, inode) ||
406 	    !capable(CAP_SYS_ADMIN)) {
407 		err = -EPERM;
408 		goto journal_err_out;
409 	}
410 
411 	filemap_invalidate_lock(inode->i_mapping);
412 	err = filemap_write_and_wait(inode->i_mapping);
413 	if (err)
414 		goto err_out;
415 
416 	err = filemap_write_and_wait(inode_bl->i_mapping);
417 	if (err)
418 		goto err_out;
419 
420 	/* Wait for all existing dio workers */
421 	inode_dio_wait(inode);
422 	inode_dio_wait(inode_bl);
423 
424 	truncate_inode_pages(&inode->i_data, 0);
425 	truncate_inode_pages(&inode_bl->i_data, 0);
426 
427 	handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
428 	if (IS_ERR(handle)) {
429 		err = -EINVAL;
430 		goto err_out;
431 	}
432 	ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_SWAP_BOOT, handle);
433 
434 	/* Protect extent tree against block allocations via delalloc */
435 	ext4_double_down_write_data_sem(inode, inode_bl);
436 
437 	if (is_bad_inode(inode_bl) || !S_ISREG(inode_bl->i_mode)) {
438 		/* this inode has never been used as a BOOT_LOADER */
439 		set_nlink(inode_bl, 1);
440 		i_uid_write(inode_bl, 0);
441 		i_gid_write(inode_bl, 0);
442 		inode_bl->i_flags = 0;
443 		ei_bl->i_flags = 0;
444 		inode_set_iversion(inode_bl, 1);
445 		i_size_write(inode_bl, 0);
446 		EXT4_I(inode_bl)->i_disksize = inode_bl->i_size;
447 		inode_bl->i_mode = S_IFREG;
448 		if (ext4_has_feature_extents(sb)) {
449 			ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
450 			ext4_ext_tree_init(handle, inode_bl);
451 		} else
452 			memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
453 	}
454 
455 	err = dquot_initialize(inode);
456 	if (err)
457 		goto err_out1;
458 
459 	size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
460 	size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
461 	diff = size - size_bl;
462 	swap_inode_data(inode, inode_bl);
463 
464 	inode_set_ctime_current(inode);
465 	inode_set_ctime_current(inode_bl);
466 	inode_inc_iversion(inode);
467 
468 	inode->i_generation = get_random_u32();
469 	inode_bl->i_generation = get_random_u32();
470 	ext4_reset_inode_seed(inode);
471 	ext4_reset_inode_seed(inode_bl);
472 
473 	ext4_discard_preallocations(inode);
474 
475 	err = ext4_mark_inode_dirty(handle, inode);
476 	if (err < 0) {
477 		/* No need to update quota information. */
478 		ext4_warning(inode->i_sb,
479 			"couldn't mark inode #%lu dirty (err %d)",
480 			inode->i_ino, err);
481 		/* Revert all changes: */
482 		swap_inode_data(inode, inode_bl);
483 		ext4_mark_inode_dirty(handle, inode);
484 		goto err_out1;
485 	}
486 
487 	blocks = inode_bl->i_blocks;
488 	bytes = inode_bl->i_bytes;
489 	inode_bl->i_blocks = inode->i_blocks;
490 	inode_bl->i_bytes = inode->i_bytes;
491 	err = ext4_mark_inode_dirty(handle, inode_bl);
492 	if (err < 0) {
493 		/* No need to update quota information. */
494 		ext4_warning(inode_bl->i_sb,
495 			"couldn't mark inode #%lu dirty (err %d)",
496 			inode_bl->i_ino, err);
497 		goto revert;
498 	}
499 
500 	/* Bootloader inode should not be counted into quota information. */
501 	if (diff > 0)
502 		dquot_free_space(inode, diff);
503 	else
504 		err = dquot_alloc_space(inode, -1 * diff);
505 
506 	if (err < 0) {
507 revert:
508 		/* Revert all changes: */
509 		inode_bl->i_blocks = blocks;
510 		inode_bl->i_bytes = bytes;
511 		swap_inode_data(inode, inode_bl);
512 		ext4_mark_inode_dirty(handle, inode);
513 		ext4_mark_inode_dirty(handle, inode_bl);
514 	}
515 
516 err_out1:
517 	ext4_journal_stop(handle);
518 	ext4_double_up_write_data_sem(inode, inode_bl);
519 
520 err_out:
521 	filemap_invalidate_unlock(inode->i_mapping);
522 journal_err_out:
523 	unlock_two_nondirectories(inode, inode_bl);
524 	iput(inode_bl);
525 	return err;
526 }
527 
528 /*
529  * If immutable is set and we are not clearing it, we're not allowed to change
530  * anything else in the inode.  Don't error out if we're only trying to set
531  * immutable on an immutable file.
532  */
ext4_ioctl_check_immutable(struct inode * inode,__u32 new_projid,unsigned int flags)533 static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
534 				      unsigned int flags)
535 {
536 	struct ext4_inode_info *ei = EXT4_I(inode);
537 	unsigned int oldflags = ei->i_flags;
538 
539 	if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
540 		return 0;
541 
542 	if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
543 		return -EPERM;
544 	if (ext4_has_feature_project(inode->i_sb) &&
545 	    __kprojid_val(ei->i_projid) != new_projid)
546 		return -EPERM;
547 
548 	return 0;
549 }
550 
ext4_dax_dontcache(struct inode * inode,unsigned int flags)551 static void ext4_dax_dontcache(struct inode *inode, unsigned int flags)
552 {
553 	struct ext4_inode_info *ei = EXT4_I(inode);
554 
555 	if (S_ISDIR(inode->i_mode))
556 		return;
557 
558 	if (test_opt2(inode->i_sb, DAX_NEVER) ||
559 	    test_opt(inode->i_sb, DAX_ALWAYS))
560 		return;
561 
562 	if ((ei->i_flags ^ flags) & EXT4_DAX_FL)
563 		d_mark_dontcache(inode);
564 }
565 
dax_compatible(struct inode * inode,unsigned int oldflags,unsigned int flags)566 static bool dax_compatible(struct inode *inode, unsigned int oldflags,
567 			   unsigned int flags)
568 {
569 	/* Allow the DAX flag to be changed on inline directories */
570 	if (S_ISDIR(inode->i_mode)) {
571 		flags &= ~EXT4_INLINE_DATA_FL;
572 		oldflags &= ~EXT4_INLINE_DATA_FL;
573 	}
574 
575 	if (flags & EXT4_DAX_FL) {
576 		if ((oldflags & EXT4_DAX_MUT_EXCL) ||
577 		     ext4_test_inode_state(inode,
578 					  EXT4_STATE_VERITY_IN_PROGRESS)) {
579 			return false;
580 		}
581 	}
582 
583 	if ((flags & EXT4_DAX_MUT_EXCL) && (oldflags & EXT4_DAX_FL))
584 			return false;
585 
586 	return true;
587 }
588 
ext4_ioctl_setflags(struct inode * inode,unsigned int flags)589 static int ext4_ioctl_setflags(struct inode *inode,
590 			       unsigned int flags)
591 {
592 	struct ext4_inode_info *ei = EXT4_I(inode);
593 	handle_t *handle = NULL;
594 	int err = -EPERM, migrate = 0;
595 	struct ext4_iloc iloc;
596 	unsigned int oldflags, mask, i;
597 	struct super_block *sb = inode->i_sb;
598 
599 	/* Is it quota file? Do not allow user to mess with it */
600 	if (ext4_is_quota_file(inode))
601 		goto flags_out;
602 
603 	oldflags = ei->i_flags;
604 	/*
605 	 * The JOURNAL_DATA flag can only be changed by
606 	 * the relevant capability.
607 	 */
608 	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
609 		if (!capable(CAP_SYS_RESOURCE))
610 			goto flags_out;
611 	}
612 
613 	if (!dax_compatible(inode, oldflags, flags)) {
614 		err = -EOPNOTSUPP;
615 		goto flags_out;
616 	}
617 
618 	if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
619 		migrate = 1;
620 
621 	if ((flags ^ oldflags) & EXT4_CASEFOLD_FL) {
622 		if (!ext4_has_feature_casefold(sb)) {
623 			err = -EOPNOTSUPP;
624 			goto flags_out;
625 		}
626 
627 		if (!S_ISDIR(inode->i_mode)) {
628 			err = -ENOTDIR;
629 			goto flags_out;
630 		}
631 
632 		if (!ext4_empty_dir(inode)) {
633 			err = -ENOTEMPTY;
634 			goto flags_out;
635 		}
636 	}
637 
638 	/*
639 	 * Wait for all pending directio and then flush all the dirty pages
640 	 * for this file.  The flush marks all the pages readonly, so any
641 	 * subsequent attempt to write to the file (particularly mmap pages)
642 	 * will come through the filesystem and fail.
643 	 */
644 	if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
645 	    (flags & EXT4_IMMUTABLE_FL)) {
646 		inode_dio_wait(inode);
647 		err = filemap_write_and_wait(inode->i_mapping);
648 		if (err)
649 			goto flags_out;
650 	}
651 
652 	handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
653 	if (IS_ERR(handle)) {
654 		err = PTR_ERR(handle);
655 		goto flags_out;
656 	}
657 	if (IS_SYNC(inode))
658 		ext4_handle_sync(handle);
659 	err = ext4_reserve_inode_write(handle, inode, &iloc);
660 	if (err)
661 		goto flags_err;
662 
663 	ext4_dax_dontcache(inode, flags);
664 
665 	for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
666 		if (!(mask & EXT4_FL_USER_MODIFIABLE))
667 			continue;
668 		/* These flags get special treatment later */
669 		if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
670 			continue;
671 		if (mask & flags)
672 			ext4_set_inode_flag(inode, i);
673 		else
674 			ext4_clear_inode_flag(inode, i);
675 	}
676 
677 	ext4_set_inode_flags(inode, false);
678 
679 	inode_set_ctime_current(inode);
680 	inode_inc_iversion(inode);
681 
682 	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
683 flags_err:
684 	ext4_journal_stop(handle);
685 	if (err)
686 		goto flags_out;
687 
688 	if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
689 		/*
690 		 * Changes to the journaling mode can cause unsafe changes to
691 		 * S_DAX if the inode is DAX
692 		 */
693 		if (IS_DAX(inode)) {
694 			err = -EBUSY;
695 			goto flags_out;
696 		}
697 
698 		err = ext4_change_inode_journal_flag(inode,
699 						     flags & EXT4_JOURNAL_DATA_FL);
700 		if (err)
701 			goto flags_out;
702 	}
703 	if (migrate) {
704 		if (flags & EXT4_EXTENTS_FL)
705 			err = ext4_ext_migrate(inode);
706 		else
707 			err = ext4_ind_migrate(inode);
708 	}
709 
710 flags_out:
711 	return err;
712 }
713 
714 #ifdef CONFIG_QUOTA
ext4_ioctl_setproject(struct inode * inode,__u32 projid)715 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
716 {
717 	struct super_block *sb = inode->i_sb;
718 	struct ext4_inode_info *ei = EXT4_I(inode);
719 	int err, rc;
720 	handle_t *handle;
721 	kprojid_t kprojid;
722 	struct ext4_iloc iloc;
723 	struct ext4_inode *raw_inode;
724 	struct dquot *transfer_to[MAXQUOTAS] = { };
725 
726 	if (!ext4_has_feature_project(sb)) {
727 		if (projid != EXT4_DEF_PROJID)
728 			return -EOPNOTSUPP;
729 		else
730 			return 0;
731 	}
732 
733 	if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
734 		return -EOPNOTSUPP;
735 
736 	kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
737 
738 	if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
739 		return 0;
740 
741 	err = -EPERM;
742 	/* Is it quota file? Do not allow user to mess with it */
743 	if (ext4_is_quota_file(inode))
744 		return err;
745 
746 	err = dquot_initialize(inode);
747 	if (err)
748 		return err;
749 
750 	err = ext4_get_inode_loc(inode, &iloc);
751 	if (err)
752 		return err;
753 
754 	raw_inode = ext4_raw_inode(&iloc);
755 	if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
756 		err = ext4_expand_extra_isize(inode,
757 					      EXT4_SB(sb)->s_want_extra_isize,
758 					      &iloc);
759 		if (err)
760 			return err;
761 	} else {
762 		brelse(iloc.bh);
763 	}
764 
765 	handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
766 		EXT4_QUOTA_INIT_BLOCKS(sb) +
767 		EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
768 	if (IS_ERR(handle))
769 		return PTR_ERR(handle);
770 
771 	err = ext4_reserve_inode_write(handle, inode, &iloc);
772 	if (err)
773 		goto out_stop;
774 
775 	transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
776 	if (!IS_ERR(transfer_to[PRJQUOTA])) {
777 
778 		/* __dquot_transfer() calls back ext4_get_inode_usage() which
779 		 * counts xattr inode references.
780 		 */
781 		down_read(&EXT4_I(inode)->xattr_sem);
782 		err = __dquot_transfer(inode, transfer_to);
783 		up_read(&EXT4_I(inode)->xattr_sem);
784 		dqput(transfer_to[PRJQUOTA]);
785 		if (err)
786 			goto out_dirty;
787 	}
788 
789 	EXT4_I(inode)->i_projid = kprojid;
790 	inode_set_ctime_current(inode);
791 	inode_inc_iversion(inode);
792 out_dirty:
793 	rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
794 	if (!err)
795 		err = rc;
796 out_stop:
797 	ext4_journal_stop(handle);
798 	return err;
799 }
800 #else
ext4_ioctl_setproject(struct inode * inode,__u32 projid)801 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
802 {
803 	if (projid != EXT4_DEF_PROJID)
804 		return -EOPNOTSUPP;
805 	return 0;
806 }
807 #endif
808 
ext4_force_shutdown(struct super_block * sb,u32 flags)809 int ext4_force_shutdown(struct super_block *sb, u32 flags)
810 {
811 	struct ext4_sb_info *sbi = EXT4_SB(sb);
812 	int ret;
813 
814 	if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
815 		return -EINVAL;
816 
817 	if (ext4_forced_shutdown(sb))
818 		return 0;
819 
820 	ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
821 	trace_ext4_shutdown(sb, flags);
822 
823 	switch (flags) {
824 	case EXT4_GOING_FLAGS_DEFAULT:
825 		ret = bdev_freeze(sb->s_bdev);
826 		if (ret)
827 			return ret;
828 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
829 		bdev_thaw(sb->s_bdev);
830 		break;
831 	case EXT4_GOING_FLAGS_LOGFLUSH:
832 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
833 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
834 			(void) ext4_force_commit(sb);
835 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
836 		}
837 		break;
838 	case EXT4_GOING_FLAGS_NOLOGFLUSH:
839 		set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
840 		if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
841 			jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
842 		break;
843 	default:
844 		return -EINVAL;
845 	}
846 	clear_opt(sb, DISCARD);
847 	return 0;
848 }
849 
ext4_ioctl_shutdown(struct super_block * sb,unsigned long arg)850 static int ext4_ioctl_shutdown(struct super_block *sb, unsigned long arg)
851 {
852 	u32 flags;
853 
854 	if (!capable(CAP_SYS_ADMIN))
855 		return -EPERM;
856 
857 	if (get_user(flags, (__u32 __user *)arg))
858 		return -EFAULT;
859 
860 	return ext4_force_shutdown(sb, flags);
861 }
862 
863 struct getfsmap_info {
864 	struct super_block	*gi_sb;
865 	struct fsmap_head __user *gi_data;
866 	unsigned int		gi_idx;
867 	__u32			gi_last_flags;
868 };
869 
ext4_getfsmap_format(struct ext4_fsmap * xfm,void * priv)870 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
871 {
872 	struct getfsmap_info *info = priv;
873 	struct fsmap fm;
874 
875 	trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
876 
877 	info->gi_last_flags = xfm->fmr_flags;
878 	ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
879 	if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
880 			sizeof(struct fsmap)))
881 		return -EFAULT;
882 
883 	return 0;
884 }
885 
ext4_ioc_getfsmap(struct super_block * sb,struct fsmap_head __user * arg)886 static int ext4_ioc_getfsmap(struct super_block *sb,
887 			     struct fsmap_head __user *arg)
888 {
889 	struct getfsmap_info info = { NULL };
890 	struct ext4_fsmap_head xhead = {0};
891 	struct fsmap_head head;
892 	bool aborted = false;
893 	int error;
894 
895 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
896 		return -EFAULT;
897 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
898 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
899 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
900 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
901 		       sizeof(head.fmh_keys[1].fmr_reserved)))
902 		return -EINVAL;
903 	/*
904 	 * ext4 doesn't report file extents at all, so the only valid
905 	 * file offsets are the magic ones (all zeroes or all ones).
906 	 */
907 	if (head.fmh_keys[0].fmr_offset ||
908 	    (head.fmh_keys[1].fmr_offset != 0 &&
909 	     head.fmh_keys[1].fmr_offset != -1ULL))
910 		return -EINVAL;
911 
912 	xhead.fmh_iflags = head.fmh_iflags;
913 	xhead.fmh_count = head.fmh_count;
914 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
915 	ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
916 
917 	trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
918 	trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
919 
920 	info.gi_sb = sb;
921 	info.gi_data = arg;
922 	error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
923 	if (error == EXT4_QUERY_RANGE_ABORT)
924 		aborted = true;
925 	else if (error)
926 		return error;
927 
928 	/* If we didn't abort, set the "last" flag in the last fmx */
929 	if (!aborted && info.gi_idx) {
930 		info.gi_last_flags |= FMR_OF_LAST;
931 		if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
932 				 &info.gi_last_flags,
933 				 sizeof(info.gi_last_flags)))
934 			return -EFAULT;
935 	}
936 
937 	/* copy back header */
938 	head.fmh_entries = xhead.fmh_entries;
939 	head.fmh_oflags = xhead.fmh_oflags;
940 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
941 		return -EFAULT;
942 
943 	return 0;
944 }
945 
ext4_ioctl_group_add(struct file * file,struct ext4_new_group_data * input)946 static long ext4_ioctl_group_add(struct file *file,
947 				 struct ext4_new_group_data *input)
948 {
949 	struct super_block *sb = file_inode(file)->i_sb;
950 	int err, err2=0;
951 
952 	err = ext4_resize_begin(sb);
953 	if (err)
954 		return err;
955 
956 	if (ext4_has_feature_bigalloc(sb)) {
957 		ext4_msg(sb, KERN_ERR,
958 			 "Online resizing not supported with bigalloc");
959 		err = -EOPNOTSUPP;
960 		goto group_add_out;
961 	}
962 
963 	err = mnt_want_write_file(file);
964 	if (err)
965 		goto group_add_out;
966 
967 	err = ext4_group_add(sb, input);
968 	if (EXT4_SB(sb)->s_journal) {
969 		jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
970 		err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
971 		jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
972 	}
973 	if (err == 0)
974 		err = err2;
975 	mnt_drop_write_file(file);
976 	if (!err && ext4_has_group_desc_csum(sb) &&
977 	    test_opt(sb, INIT_INODE_TABLE))
978 		err = ext4_register_li_request(sb, input->group);
979 group_add_out:
980 	err2 = ext4_resize_end(sb, false);
981 	if (err == 0)
982 		err = err2;
983 	return err;
984 }
985 
ext4_fileattr_get(struct dentry * dentry,struct file_kattr * fa)986 int ext4_fileattr_get(struct dentry *dentry, struct file_kattr *fa)
987 {
988 	struct inode *inode = d_inode(dentry);
989 	struct ext4_inode_info *ei = EXT4_I(inode);
990 	u32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
991 
992 	if (S_ISREG(inode->i_mode))
993 		flags &= ~FS_PROJINHERIT_FL;
994 
995 	fileattr_fill_flags(fa, flags);
996 	if (ext4_has_feature_project(inode->i_sb))
997 		fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
998 
999 	return 0;
1000 }
1001 
ext4_fileattr_set(struct mnt_idmap * idmap,struct dentry * dentry,struct file_kattr * fa)1002 int ext4_fileattr_set(struct mnt_idmap *idmap,
1003 		      struct dentry *dentry, struct file_kattr *fa)
1004 {
1005 	struct inode *inode = d_inode(dentry);
1006 	u32 flags = fa->flags;
1007 	int err = -EOPNOTSUPP;
1008 
1009 	if (flags & ~EXT4_FL_USER_VISIBLE)
1010 		goto out;
1011 
1012 	/*
1013 	 * chattr(1) grabs flags via GETFLAGS, modifies the result and
1014 	 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
1015 	 * more restrictive than just silently masking off visible but
1016 	 * not settable flags as we always did.
1017 	 */
1018 	flags &= EXT4_FL_USER_MODIFIABLE;
1019 	if (ext4_mask_flags(inode->i_mode, flags) != flags)
1020 		goto out;
1021 	err = ext4_ioctl_check_immutable(inode, fa->fsx_projid, flags);
1022 	if (err)
1023 		goto out;
1024 	err = ext4_ioctl_setflags(inode, flags);
1025 	if (err)
1026 		goto out;
1027 	err = ext4_ioctl_setproject(inode, fa->fsx_projid);
1028 out:
1029 	return err;
1030 }
1031 
1032 /* So that the fiemap access checks can't overflow on 32 bit machines. */
1033 #define FIEMAP_MAX_EXTENTS	(UINT_MAX / sizeof(struct fiemap_extent))
1034 
ext4_ioctl_get_es_cache(struct file * filp,unsigned long arg)1035 static int ext4_ioctl_get_es_cache(struct file *filp, unsigned long arg)
1036 {
1037 	struct fiemap fiemap;
1038 	struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
1039 	struct fiemap_extent_info fieinfo = { 0, };
1040 	struct inode *inode = file_inode(filp);
1041 	int error;
1042 
1043 	if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
1044 		return -EFAULT;
1045 
1046 	if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
1047 		return -EINVAL;
1048 
1049 	fieinfo.fi_flags = fiemap.fm_flags;
1050 	fieinfo.fi_extents_max = fiemap.fm_extent_count;
1051 	fieinfo.fi_extents_start = ufiemap->fm_extents;
1052 
1053 	error = ext4_get_es_cache(inode, &fieinfo, fiemap.fm_start,
1054 			fiemap.fm_length);
1055 	fiemap.fm_flags = fieinfo.fi_flags;
1056 	fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
1057 	if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
1058 		error = -EFAULT;
1059 
1060 	return error;
1061 }
1062 
ext4_ioctl_checkpoint(struct file * filp,unsigned long arg)1063 static int ext4_ioctl_checkpoint(struct file *filp, unsigned long arg)
1064 {
1065 	int err = 0;
1066 	__u32 flags = 0;
1067 	unsigned int flush_flags = 0;
1068 	struct super_block *sb = file_inode(filp)->i_sb;
1069 
1070 	if (copy_from_user(&flags, (__u32 __user *)arg,
1071 				sizeof(__u32)))
1072 		return -EFAULT;
1073 
1074 	if (!capable(CAP_SYS_ADMIN))
1075 		return -EPERM;
1076 
1077 	/* check for invalid bits set */
1078 	if ((flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID) ||
1079 				((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1080 				(flags & JBD2_JOURNAL_FLUSH_ZEROOUT)))
1081 		return -EINVAL;
1082 
1083 	if (!EXT4_SB(sb)->s_journal)
1084 		return -ENODEV;
1085 
1086 	if ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1087 	    !bdev_max_discard_sectors(EXT4_SB(sb)->s_journal->j_dev))
1088 		return -EOPNOTSUPP;
1089 
1090 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
1091 		return 0;
1092 
1093 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_DISCARD)
1094 		flush_flags |= JBD2_JOURNAL_FLUSH_DISCARD;
1095 
1096 	if (flags & EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT) {
1097 		flush_flags |= JBD2_JOURNAL_FLUSH_ZEROOUT;
1098 		pr_info_ratelimited("warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow");
1099 	}
1100 
1101 	jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1102 	err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, flush_flags);
1103 	jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1104 
1105 	return err;
1106 }
1107 
ext4_ioctl_setlabel(struct file * filp,const char __user * user_label)1108 static int ext4_ioctl_setlabel(struct file *filp, const char __user *user_label)
1109 {
1110 	size_t len;
1111 	int ret = 0;
1112 	char new_label[EXT4_LABEL_MAX + 1];
1113 	struct super_block *sb = file_inode(filp)->i_sb;
1114 
1115 	if (!capable(CAP_SYS_ADMIN))
1116 		return -EPERM;
1117 
1118 	/*
1119 	 * Copy the maximum length allowed for ext4 label with one more to
1120 	 * find the required terminating null byte in order to test the
1121 	 * label length. The on disk label doesn't need to be null terminated.
1122 	 */
1123 	if (copy_from_user(new_label, user_label, EXT4_LABEL_MAX + 1))
1124 		return -EFAULT;
1125 
1126 	len = strnlen(new_label, EXT4_LABEL_MAX + 1);
1127 	if (len > EXT4_LABEL_MAX)
1128 		return -EINVAL;
1129 
1130 	/*
1131 	 * Clear the buffer after the new label
1132 	 */
1133 	memset(new_label + len, 0, EXT4_LABEL_MAX - len);
1134 
1135 	ret = mnt_want_write_file(filp);
1136 	if (ret)
1137 		return ret;
1138 
1139 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setlabel, new_label);
1140 
1141 	mnt_drop_write_file(filp);
1142 	return ret;
1143 }
1144 
ext4_ioctl_getlabel(struct ext4_sb_info * sbi,char __user * user_label)1145 static int ext4_ioctl_getlabel(struct ext4_sb_info *sbi, char __user *user_label)
1146 {
1147 	char label[EXT4_LABEL_MAX + 1];
1148 
1149 	/*
1150 	 * EXT4_LABEL_MAX must always be smaller than FSLABEL_MAX because
1151 	 * FSLABEL_MAX must include terminating null byte, while s_volume_name
1152 	 * does not have to.
1153 	 */
1154 	BUILD_BUG_ON(EXT4_LABEL_MAX >= FSLABEL_MAX);
1155 
1156 	lock_buffer(sbi->s_sbh);
1157 	memtostr_pad(label, sbi->s_es->s_volume_name);
1158 	unlock_buffer(sbi->s_sbh);
1159 
1160 	if (copy_to_user(user_label, label, sizeof(label)))
1161 		return -EFAULT;
1162 	return 0;
1163 }
1164 
ext4_ioctl_getuuid(struct ext4_sb_info * sbi,struct fsuuid __user * ufsuuid)1165 static int ext4_ioctl_getuuid(struct ext4_sb_info *sbi,
1166 			struct fsuuid __user *ufsuuid)
1167 {
1168 	struct fsuuid fsuuid;
1169 	__u8 uuid[UUID_SIZE];
1170 
1171 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1172 		return -EFAULT;
1173 
1174 	if (fsuuid.fsu_len == 0) {
1175 		fsuuid.fsu_len = UUID_SIZE;
1176 		if (copy_to_user(&ufsuuid->fsu_len, &fsuuid.fsu_len,
1177 					sizeof(fsuuid.fsu_len)))
1178 			return -EFAULT;
1179 		return 0;
1180 	}
1181 
1182 	if (fsuuid.fsu_len < UUID_SIZE || fsuuid.fsu_flags != 0)
1183 		return -EINVAL;
1184 
1185 	lock_buffer(sbi->s_sbh);
1186 	memcpy(uuid, sbi->s_es->s_uuid, UUID_SIZE);
1187 	unlock_buffer(sbi->s_sbh);
1188 
1189 	fsuuid.fsu_len = UUID_SIZE;
1190 	if (copy_to_user(ufsuuid, &fsuuid, sizeof(fsuuid)) ||
1191 	    copy_to_user(&ufsuuid->fsu_uuid[0], uuid, UUID_SIZE))
1192 		return -EFAULT;
1193 	return 0;
1194 }
1195 
ext4_ioctl_setuuid(struct file * filp,const struct fsuuid __user * ufsuuid)1196 static int ext4_ioctl_setuuid(struct file *filp,
1197 			const struct fsuuid __user *ufsuuid)
1198 {
1199 	int ret = 0;
1200 	struct super_block *sb = file_inode(filp)->i_sb;
1201 	struct fsuuid fsuuid;
1202 	__u8 uuid[UUID_SIZE];
1203 
1204 	if (!capable(CAP_SYS_ADMIN))
1205 		return -EPERM;
1206 
1207 	/*
1208 	 * If any checksums (group descriptors or metadata) are being used
1209 	 * then the checksum seed feature is required to change the UUID.
1210 	 */
1211 	if (((ext4_has_feature_gdt_csum(sb) ||
1212 	      ext4_has_feature_metadata_csum(sb))
1213 			&& !ext4_has_feature_csum_seed(sb))
1214 		|| ext4_has_feature_stable_inodes(sb))
1215 		return -EOPNOTSUPP;
1216 
1217 	if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1218 		return -EFAULT;
1219 
1220 	if (fsuuid.fsu_len != UUID_SIZE || fsuuid.fsu_flags != 0)
1221 		return -EINVAL;
1222 
1223 	if (copy_from_user(uuid, &ufsuuid->fsu_uuid[0], UUID_SIZE))
1224 		return -EFAULT;
1225 
1226 	ret = mnt_want_write_file(filp);
1227 	if (ret)
1228 		return ret;
1229 
1230 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setuuid, &uuid);
1231 	mnt_drop_write_file(filp);
1232 
1233 	return ret;
1234 }
1235 
1236 
1237 #define TUNE_OPS_SUPPORTED (EXT4_TUNE_FL_ERRORS_BEHAVIOR |    \
1238 	EXT4_TUNE_FL_MNT_COUNT | EXT4_TUNE_FL_MAX_MNT_COUNT | \
1239 	EXT4_TUNE_FL_CHECKINTRVAL | EXT4_TUNE_FL_LAST_CHECK_TIME | \
1240 	EXT4_TUNE_FL_RESERVED_BLOCKS | EXT4_TUNE_FL_RESERVED_UID | \
1241 	EXT4_TUNE_FL_RESERVED_GID | EXT4_TUNE_FL_DEFAULT_MNT_OPTS | \
1242 	EXT4_TUNE_FL_DEF_HASH_ALG | EXT4_TUNE_FL_RAID_STRIDE | \
1243 	EXT4_TUNE_FL_RAID_STRIPE_WIDTH | EXT4_TUNE_FL_MOUNT_OPTS | \
1244 	EXT4_TUNE_FL_FEATURES | EXT4_TUNE_FL_EDIT_FEATURES | \
1245 	EXT4_TUNE_FL_FORCE_FSCK | EXT4_TUNE_FL_ENCODING | \
1246 	EXT4_TUNE_FL_ENCODING_FLAGS)
1247 
1248 #define EXT4_TUNE_SET_COMPAT_SUPP \
1249 		(EXT4_FEATURE_COMPAT_DIR_INDEX |	\
1250 		 EXT4_FEATURE_COMPAT_STABLE_INODES)
1251 #define EXT4_TUNE_SET_INCOMPAT_SUPP \
1252 		(EXT4_FEATURE_INCOMPAT_EXTENTS |	\
1253 		 EXT4_FEATURE_INCOMPAT_EA_INODE |	\
1254 		 EXT4_FEATURE_INCOMPAT_ENCRYPT |	\
1255 		 EXT4_FEATURE_INCOMPAT_CSUM_SEED |	\
1256 		 EXT4_FEATURE_INCOMPAT_LARGEDIR |	\
1257 		 EXT4_FEATURE_INCOMPAT_CASEFOLD)
1258 #define EXT4_TUNE_SET_RO_COMPAT_SUPP \
1259 		(EXT4_FEATURE_RO_COMPAT_LARGE_FILE |	\
1260 		 EXT4_FEATURE_RO_COMPAT_DIR_NLINK |	\
1261 		 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE |	\
1262 		 EXT4_FEATURE_RO_COMPAT_PROJECT |	\
1263 		 EXT4_FEATURE_RO_COMPAT_VERITY)
1264 
1265 #define EXT4_TUNE_CLEAR_COMPAT_SUPP (0)
1266 #define EXT4_TUNE_CLEAR_INCOMPAT_SUPP (0)
1267 #define EXT4_TUNE_CLEAR_RO_COMPAT_SUPP (0)
1268 
1269 #define SB_ENC_SUPP_MASK (SB_ENC_STRICT_MODE_FL |	\
1270 			  SB_ENC_NO_COMPAT_FALLBACK_FL)
1271 
ext4_ioctl_get_tune_sb(struct ext4_sb_info * sbi,struct ext4_tune_sb_params __user * params)1272 static int ext4_ioctl_get_tune_sb(struct ext4_sb_info *sbi,
1273 				  struct ext4_tune_sb_params __user *params)
1274 {
1275 	struct ext4_tune_sb_params ret;
1276 	struct ext4_super_block *es = sbi->s_es;
1277 
1278 	memset(&ret, 0, sizeof(ret));
1279 	ret.set_flags = TUNE_OPS_SUPPORTED;
1280 	ret.errors_behavior = le16_to_cpu(es->s_errors);
1281 	ret.mnt_count = le16_to_cpu(es->s_mnt_count);
1282 	ret.max_mnt_count = le16_to_cpu(es->s_max_mnt_count);
1283 	ret.checkinterval = le32_to_cpu(es->s_checkinterval);
1284 	ret.last_check_time = le32_to_cpu(es->s_lastcheck);
1285 	ret.reserved_blocks = ext4_r_blocks_count(es);
1286 	ret.blocks_count = ext4_blocks_count(es);
1287 	ret.reserved_uid = ext4_get_resuid(es);
1288 	ret.reserved_gid = ext4_get_resgid(es);
1289 	ret.default_mnt_opts = le32_to_cpu(es->s_default_mount_opts);
1290 	ret.def_hash_alg = es->s_def_hash_version;
1291 	ret.raid_stride = le16_to_cpu(es->s_raid_stride);
1292 	ret.raid_stripe_width = le32_to_cpu(es->s_raid_stripe_width);
1293 	ret.encoding = le16_to_cpu(es->s_encoding);
1294 	ret.encoding_flags = le16_to_cpu(es->s_encoding_flags);
1295 	strscpy_pad(ret.mount_opts, es->s_mount_opts);
1296 	ret.feature_compat = le32_to_cpu(es->s_feature_compat);
1297 	ret.feature_incompat = le32_to_cpu(es->s_feature_incompat);
1298 	ret.feature_ro_compat = le32_to_cpu(es->s_feature_ro_compat);
1299 	ret.set_feature_compat_mask = EXT4_TUNE_SET_COMPAT_SUPP;
1300 	ret.set_feature_incompat_mask = EXT4_TUNE_SET_INCOMPAT_SUPP;
1301 	ret.set_feature_ro_compat_mask = EXT4_TUNE_SET_RO_COMPAT_SUPP;
1302 	ret.clear_feature_compat_mask = EXT4_TUNE_CLEAR_COMPAT_SUPP;
1303 	ret.clear_feature_incompat_mask = EXT4_TUNE_CLEAR_INCOMPAT_SUPP;
1304 	ret.clear_feature_ro_compat_mask = EXT4_TUNE_CLEAR_RO_COMPAT_SUPP;
1305 	if (copy_to_user(params, &ret, sizeof(ret)))
1306 		return -EFAULT;
1307 	return 0;
1308 }
1309 
ext4_sb_setparams(struct ext4_sb_info * sbi,struct ext4_super_block * es,const void * arg)1310 static void ext4_sb_setparams(struct ext4_sb_info *sbi,
1311 			      struct ext4_super_block *es, const void *arg)
1312 {
1313 	const struct ext4_tune_sb_params *params = arg;
1314 
1315 	if (params->set_flags & EXT4_TUNE_FL_ERRORS_BEHAVIOR)
1316 		es->s_errors = cpu_to_le16(params->errors_behavior);
1317 	if (params->set_flags & EXT4_TUNE_FL_MNT_COUNT)
1318 		es->s_mnt_count = cpu_to_le16(params->mnt_count);
1319 	if (params->set_flags & EXT4_TUNE_FL_MAX_MNT_COUNT)
1320 		es->s_max_mnt_count = cpu_to_le16(params->max_mnt_count);
1321 	if (params->set_flags & EXT4_TUNE_FL_CHECKINTRVAL)
1322 		es->s_checkinterval = cpu_to_le32(params->checkinterval);
1323 	if (params->set_flags & EXT4_TUNE_FL_LAST_CHECK_TIME)
1324 		es->s_lastcheck = cpu_to_le32(params->last_check_time);
1325 	if (params->set_flags & EXT4_TUNE_FL_RESERVED_BLOCKS) {
1326 		ext4_fsblk_t blk = params->reserved_blocks;
1327 
1328 		es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1329 		es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1330 	}
1331 	if (params->set_flags & EXT4_TUNE_FL_RESERVED_UID) {
1332 		int uid = params->reserved_uid;
1333 
1334 		es->s_def_resuid = cpu_to_le16(uid & 0xFFFF);
1335 		es->s_def_resuid_hi = cpu_to_le16(uid >> 16);
1336 	}
1337 	if (params->set_flags & EXT4_TUNE_FL_RESERVED_GID) {
1338 		int gid = params->reserved_gid;
1339 
1340 		es->s_def_resgid = cpu_to_le16(gid & 0xFFFF);
1341 		es->s_def_resgid_hi = cpu_to_le16(gid >> 16);
1342 	}
1343 	if (params->set_flags & EXT4_TUNE_FL_DEFAULT_MNT_OPTS)
1344 		es->s_default_mount_opts = cpu_to_le32(params->default_mnt_opts);
1345 	if (params->set_flags & EXT4_TUNE_FL_DEF_HASH_ALG)
1346 		es->s_def_hash_version = params->def_hash_alg;
1347 	if (params->set_flags & EXT4_TUNE_FL_RAID_STRIDE)
1348 		es->s_raid_stride = cpu_to_le16(params->raid_stride);
1349 	if (params->set_flags & EXT4_TUNE_FL_RAID_STRIPE_WIDTH)
1350 		es->s_raid_stripe_width =
1351 			cpu_to_le32(params->raid_stripe_width);
1352 	if (params->set_flags & EXT4_TUNE_FL_ENCODING)
1353 		es->s_encoding = cpu_to_le16(params->encoding);
1354 	if (params->set_flags & EXT4_TUNE_FL_ENCODING_FLAGS)
1355 		es->s_encoding_flags = cpu_to_le16(params->encoding_flags);
1356 	strscpy_pad(es->s_mount_opts, params->mount_opts);
1357 	if (params->set_flags & EXT4_TUNE_FL_EDIT_FEATURES) {
1358 		es->s_feature_compat |=
1359 			cpu_to_le32(params->set_feature_compat_mask);
1360 		es->s_feature_incompat |=
1361 			cpu_to_le32(params->set_feature_incompat_mask);
1362 		es->s_feature_ro_compat |=
1363 			cpu_to_le32(params->set_feature_ro_compat_mask);
1364 		es->s_feature_compat &=
1365 			~cpu_to_le32(params->clear_feature_compat_mask);
1366 		es->s_feature_incompat &=
1367 			~cpu_to_le32(params->clear_feature_incompat_mask);
1368 		es->s_feature_ro_compat &=
1369 			~cpu_to_le32(params->clear_feature_ro_compat_mask);
1370 		if (params->set_feature_compat_mask &
1371 		    EXT4_FEATURE_COMPAT_DIR_INDEX)
1372 			es->s_def_hash_version = sbi->s_def_hash_version;
1373 		if (params->set_feature_incompat_mask &
1374 		    EXT4_FEATURE_INCOMPAT_CSUM_SEED)
1375 			es->s_checksum_seed = cpu_to_le32(sbi->s_csum_seed);
1376 	}
1377 	if (params->set_flags & EXT4_TUNE_FL_FORCE_FSCK)
1378 		es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
1379 }
1380 
ext4_ioctl_set_tune_sb(struct file * filp,struct ext4_tune_sb_params __user * in)1381 static int ext4_ioctl_set_tune_sb(struct file *filp,
1382 				  struct ext4_tune_sb_params __user *in)
1383 {
1384 	struct ext4_tune_sb_params params;
1385 	struct super_block *sb = file_inode(filp)->i_sb;
1386 	struct ext4_sb_info *sbi = EXT4_SB(sb);
1387 	struct ext4_super_block *es = sbi->s_es;
1388 	int enabling_casefold = 0;
1389 	int ret;
1390 
1391 	if (!capable(CAP_SYS_ADMIN))
1392 		return -EPERM;
1393 
1394 	if (copy_from_user(&params, in, sizeof(params)))
1395 		return -EFAULT;
1396 
1397 	if ((params.set_flags & ~TUNE_OPS_SUPPORTED) != 0)
1398 		return -EOPNOTSUPP;
1399 
1400 	if ((params.set_flags & EXT4_TUNE_FL_ERRORS_BEHAVIOR) &&
1401 	    (params.errors_behavior > EXT4_ERRORS_PANIC))
1402 		return -EINVAL;
1403 
1404 	if ((params.set_flags & EXT4_TUNE_FL_RESERVED_BLOCKS) &&
1405 	    (params.reserved_blocks > ext4_blocks_count(sbi->s_es) / 2))
1406 		return -EINVAL;
1407 	if ((params.set_flags & EXT4_TUNE_FL_DEF_HASH_ALG) &&
1408 	    ((params.def_hash_alg > DX_HASH_LAST) ||
1409 	     (params.def_hash_alg == DX_HASH_SIPHASH)))
1410 		return -EINVAL;
1411 	if ((params.set_flags & EXT4_TUNE_FL_FEATURES) &&
1412 	    (params.set_flags & EXT4_TUNE_FL_EDIT_FEATURES))
1413 		return -EINVAL;
1414 
1415 	if (params.set_flags & EXT4_TUNE_FL_FEATURES) {
1416 		params.set_feature_compat_mask =
1417 			params.feature_compat &
1418 			~le32_to_cpu(es->s_feature_compat);
1419 		params.set_feature_incompat_mask =
1420 			params.feature_incompat &
1421 			~le32_to_cpu(es->s_feature_incompat);
1422 		params.set_feature_ro_compat_mask =
1423 			params.feature_ro_compat &
1424 			~le32_to_cpu(es->s_feature_ro_compat);
1425 		params.clear_feature_compat_mask =
1426 			~params.feature_compat &
1427 			le32_to_cpu(es->s_feature_compat);
1428 		params.clear_feature_incompat_mask =
1429 			~params.feature_incompat &
1430 			le32_to_cpu(es->s_feature_incompat);
1431 		params.clear_feature_ro_compat_mask =
1432 			~params.feature_ro_compat &
1433 			le32_to_cpu(es->s_feature_ro_compat);
1434 		params.set_flags |= EXT4_TUNE_FL_EDIT_FEATURES;
1435 	}
1436 	if (params.set_flags & EXT4_TUNE_FL_EDIT_FEATURES) {
1437 		if ((params.set_feature_compat_mask &
1438 		     ~EXT4_TUNE_SET_COMPAT_SUPP) ||
1439 		    (params.set_feature_incompat_mask &
1440 		     ~EXT4_TUNE_SET_INCOMPAT_SUPP) ||
1441 		    (params.set_feature_ro_compat_mask &
1442 		     ~EXT4_TUNE_SET_RO_COMPAT_SUPP) ||
1443 		    (params.clear_feature_compat_mask &
1444 		     ~EXT4_TUNE_CLEAR_COMPAT_SUPP) ||
1445 		    (params.clear_feature_incompat_mask &
1446 		     ~EXT4_TUNE_CLEAR_INCOMPAT_SUPP) ||
1447 		    (params.clear_feature_ro_compat_mask &
1448 		     ~EXT4_TUNE_CLEAR_RO_COMPAT_SUPP))
1449 			return -EOPNOTSUPP;
1450 
1451 		/*
1452 		 * Filter out the features that are already set from
1453 		 * the set_mask.
1454 		 */
1455 		params.set_feature_compat_mask &=
1456 			~le32_to_cpu(es->s_feature_compat);
1457 		params.set_feature_incompat_mask &=
1458 			~le32_to_cpu(es->s_feature_incompat);
1459 		params.set_feature_ro_compat_mask &=
1460 			~le32_to_cpu(es->s_feature_ro_compat);
1461 		if ((params.set_feature_incompat_mask &
1462 		     EXT4_FEATURE_INCOMPAT_CASEFOLD)) {
1463 			enabling_casefold = 1;
1464 			if (!(params.set_flags & EXT4_TUNE_FL_ENCODING)) {
1465 				params.encoding = EXT4_ENC_UTF8_12_1;
1466 				params.set_flags |= EXT4_TUNE_FL_ENCODING;
1467 			}
1468 			if (!(params.set_flags & EXT4_TUNE_FL_ENCODING_FLAGS)) {
1469 				params.encoding_flags = 0;
1470 				params.set_flags |= EXT4_TUNE_FL_ENCODING_FLAGS;
1471 			}
1472 		}
1473 		if ((params.set_feature_compat_mask &
1474 		     EXT4_FEATURE_COMPAT_DIR_INDEX)) {
1475 			uuid_t	uu;
1476 
1477 			memcpy(&uu, sbi->s_hash_seed, UUID_SIZE);
1478 			if (uuid_is_null(&uu))
1479 				generate_random_uuid((char *)
1480 						     &sbi->s_hash_seed);
1481 			if (params.set_flags & EXT4_TUNE_FL_DEF_HASH_ALG)
1482 				sbi->s_def_hash_version = params.def_hash_alg;
1483 			else if (sbi->s_def_hash_version == 0)
1484 				sbi->s_def_hash_version = DX_HASH_HALF_MD4;
1485 			if (!(es->s_flags &
1486 			      cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH)) &&
1487 			    !(es->s_flags &
1488 			      cpu_to_le32(EXT2_FLAGS_SIGNED_HASH))) {
1489 #ifdef __CHAR_UNSIGNED__
1490 				sbi->s_hash_unsigned = 3;
1491 #else
1492 				sbi->s_hash_unsigned = 0;
1493 #endif
1494 			}
1495 		}
1496 	}
1497 	if (params.set_flags & EXT4_TUNE_FL_ENCODING) {
1498 		if (!enabling_casefold)
1499 			return -EINVAL;
1500 		if (params.encoding == 0)
1501 			params.encoding = EXT4_ENC_UTF8_12_1;
1502 		else if (params.encoding != EXT4_ENC_UTF8_12_1)
1503 			return -EINVAL;
1504 	}
1505 	if (params.set_flags & EXT4_TUNE_FL_ENCODING_FLAGS) {
1506 		if (!enabling_casefold)
1507 			return -EINVAL;
1508 		if (params.encoding_flags & ~SB_ENC_SUPP_MASK)
1509 			return -EINVAL;
1510 	}
1511 
1512 	ret = mnt_want_write_file(filp);
1513 	if (ret)
1514 		return ret;
1515 
1516 	ret = ext4_update_superblocks_fn(sb, ext4_sb_setparams, &params);
1517 	mnt_drop_write_file(filp);
1518 
1519 	if (params.set_flags & EXT4_TUNE_FL_DEF_HASH_ALG)
1520 		sbi->s_def_hash_version = params.def_hash_alg;
1521 
1522 	return ret;
1523 }
1524 
__ext4_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)1525 static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1526 {
1527 	struct inode *inode = file_inode(filp);
1528 	struct super_block *sb = inode->i_sb;
1529 	struct mnt_idmap *idmap = file_mnt_idmap(filp);
1530 
1531 	ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
1532 
1533 	switch (cmd) {
1534 	case FS_IOC_GETFSMAP:
1535 		return ext4_ioc_getfsmap(sb, (void __user *)arg);
1536 	case EXT4_IOC_GETVERSION:
1537 	case EXT4_IOC_GETVERSION_OLD:
1538 		return put_user(inode->i_generation, (int __user *) arg);
1539 	case EXT4_IOC_SETVERSION:
1540 	case EXT4_IOC_SETVERSION_OLD: {
1541 		handle_t *handle;
1542 		struct ext4_iloc iloc;
1543 		__u32 generation;
1544 		int err;
1545 
1546 		if (!inode_owner_or_capable(idmap, inode))
1547 			return -EPERM;
1548 
1549 		if (ext4_has_feature_metadata_csum(inode->i_sb)) {
1550 			ext4_warning(sb, "Setting inode version is not "
1551 				     "supported with metadata_csum enabled.");
1552 			return -ENOTTY;
1553 		}
1554 
1555 		err = mnt_want_write_file(filp);
1556 		if (err)
1557 			return err;
1558 		if (get_user(generation, (int __user *) arg)) {
1559 			err = -EFAULT;
1560 			goto setversion_out;
1561 		}
1562 
1563 		inode_lock(inode);
1564 		handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
1565 		if (IS_ERR(handle)) {
1566 			err = PTR_ERR(handle);
1567 			goto unlock_out;
1568 		}
1569 		err = ext4_reserve_inode_write(handle, inode, &iloc);
1570 		if (err == 0) {
1571 			inode_set_ctime_current(inode);
1572 			inode_inc_iversion(inode);
1573 			inode->i_generation = generation;
1574 			err = ext4_mark_iloc_dirty(handle, inode, &iloc);
1575 		}
1576 		ext4_journal_stop(handle);
1577 
1578 unlock_out:
1579 		inode_unlock(inode);
1580 setversion_out:
1581 		mnt_drop_write_file(filp);
1582 		return err;
1583 	}
1584 	case EXT4_IOC_GROUP_EXTEND: {
1585 		ext4_fsblk_t n_blocks_count;
1586 		int err, err2=0;
1587 
1588 		err = ext4_resize_begin(sb);
1589 		if (err)
1590 			return err;
1591 
1592 		if (get_user(n_blocks_count, (__u32 __user *)arg)) {
1593 			err = -EFAULT;
1594 			goto group_extend_out;
1595 		}
1596 
1597 		if (ext4_has_feature_bigalloc(sb)) {
1598 			ext4_msg(sb, KERN_ERR,
1599 				 "Online resizing not supported with bigalloc");
1600 			err = -EOPNOTSUPP;
1601 			goto group_extend_out;
1602 		}
1603 
1604 		err = mnt_want_write_file(filp);
1605 		if (err)
1606 			goto group_extend_out;
1607 
1608 		err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
1609 		if (EXT4_SB(sb)->s_journal) {
1610 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1611 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1612 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1613 		}
1614 		if (err == 0)
1615 			err = err2;
1616 		mnt_drop_write_file(filp);
1617 group_extend_out:
1618 		err2 = ext4_resize_end(sb, false);
1619 		if (err == 0)
1620 			err = err2;
1621 		return err;
1622 	}
1623 
1624 	case EXT4_IOC_MOVE_EXT: {
1625 		struct move_extent me;
1626 		int err;
1627 
1628 		if (!(filp->f_mode & FMODE_READ) ||
1629 		    !(filp->f_mode & FMODE_WRITE))
1630 			return -EBADF;
1631 
1632 		if (copy_from_user(&me,
1633 			(struct move_extent __user *)arg, sizeof(me)))
1634 			return -EFAULT;
1635 		me.moved_len = 0;
1636 
1637 		CLASS(fd, donor)(me.donor_fd);
1638 		if (fd_empty(donor))
1639 			return -EBADF;
1640 
1641 		if (!(fd_file(donor)->f_mode & FMODE_WRITE))
1642 			return -EBADF;
1643 
1644 		if (ext4_has_feature_bigalloc(sb)) {
1645 			ext4_msg(sb, KERN_ERR,
1646 				 "Online defrag not supported with bigalloc");
1647 			return -EOPNOTSUPP;
1648 		} else if (IS_DAX(inode)) {
1649 			ext4_msg(sb, KERN_ERR,
1650 				 "Online defrag not supported with DAX");
1651 			return -EOPNOTSUPP;
1652 		}
1653 
1654 		err = mnt_want_write_file(filp);
1655 		if (err)
1656 			return err;
1657 
1658 		err = ext4_move_extents(filp, fd_file(donor), me.orig_start,
1659 					me.donor_start, me.len, &me.moved_len);
1660 		mnt_drop_write_file(filp);
1661 
1662 		if (copy_to_user((struct move_extent __user *)arg,
1663 				 &me, sizeof(me)))
1664 			err = -EFAULT;
1665 		return err;
1666 	}
1667 
1668 	case EXT4_IOC_GROUP_ADD: {
1669 		struct ext4_new_group_data input;
1670 
1671 		if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
1672 				sizeof(input)))
1673 			return -EFAULT;
1674 
1675 		return ext4_ioctl_group_add(filp, &input);
1676 	}
1677 
1678 	case EXT4_IOC_MIGRATE:
1679 	{
1680 		int err;
1681 		if (!inode_owner_or_capable(idmap, inode))
1682 			return -EACCES;
1683 
1684 		err = mnt_want_write_file(filp);
1685 		if (err)
1686 			return err;
1687 		/*
1688 		 * inode_mutex prevent write and truncate on the file.
1689 		 * Read still goes through. We take i_data_sem in
1690 		 * ext4_ext_swap_inode_data before we switch the
1691 		 * inode format to prevent read.
1692 		 */
1693 		inode_lock((inode));
1694 		err = ext4_ext_migrate(inode);
1695 		inode_unlock((inode));
1696 		mnt_drop_write_file(filp);
1697 		return err;
1698 	}
1699 
1700 	case EXT4_IOC_ALLOC_DA_BLKS:
1701 	{
1702 		int err;
1703 		if (!inode_owner_or_capable(idmap, inode))
1704 			return -EACCES;
1705 
1706 		err = mnt_want_write_file(filp);
1707 		if (err)
1708 			return err;
1709 		err = ext4_alloc_da_blocks(inode);
1710 		mnt_drop_write_file(filp);
1711 		return err;
1712 	}
1713 
1714 	case EXT4_IOC_SWAP_BOOT:
1715 	{
1716 		int err;
1717 		if (!(filp->f_mode & FMODE_WRITE))
1718 			return -EBADF;
1719 		err = mnt_want_write_file(filp);
1720 		if (err)
1721 			return err;
1722 		err = swap_inode_boot_loader(sb, idmap, inode);
1723 		mnt_drop_write_file(filp);
1724 		return err;
1725 	}
1726 
1727 	case EXT4_IOC_RESIZE_FS: {
1728 		ext4_fsblk_t n_blocks_count;
1729 		int err = 0, err2 = 0;
1730 		ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
1731 
1732 		if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1733 				   sizeof(__u64))) {
1734 			return -EFAULT;
1735 		}
1736 
1737 		err = ext4_resize_begin(sb);
1738 		if (err)
1739 			return err;
1740 
1741 		err = mnt_want_write_file(filp);
1742 		if (err)
1743 			goto resizefs_out;
1744 
1745 		err = ext4_resize_fs(sb, n_blocks_count);
1746 		if (EXT4_SB(sb)->s_journal) {
1747 			ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE, NULL);
1748 			jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1749 			err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1750 			jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1751 		}
1752 		if (err == 0)
1753 			err = err2;
1754 		mnt_drop_write_file(filp);
1755 		if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1756 		    ext4_has_group_desc_csum(sb) &&
1757 		    test_opt(sb, INIT_INODE_TABLE))
1758 			err = ext4_register_li_request(sb, o_group);
1759 
1760 resizefs_out:
1761 		err2 = ext4_resize_end(sb, true);
1762 		if (err == 0)
1763 			err = err2;
1764 		return err;
1765 	}
1766 
1767 	case FITRIM:
1768 	{
1769 		struct fstrim_range range;
1770 		int ret = 0;
1771 
1772 		if (!capable(CAP_SYS_ADMIN))
1773 			return -EPERM;
1774 
1775 		if (!bdev_max_discard_sectors(sb->s_bdev))
1776 			return -EOPNOTSUPP;
1777 
1778 		/*
1779 		 * We haven't replayed the journal, so we cannot use our
1780 		 * block-bitmap-guided storage zapping commands.
1781 		 */
1782 		if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1783 			return -EROFS;
1784 
1785 		if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1786 		    sizeof(range)))
1787 			return -EFAULT;
1788 
1789 		ret = ext4_trim_fs(sb, &range);
1790 		if (ret < 0)
1791 			return ret;
1792 
1793 		if (copy_to_user((struct fstrim_range __user *)arg, &range,
1794 		    sizeof(range)))
1795 			return -EFAULT;
1796 
1797 		return 0;
1798 	}
1799 	case EXT4_IOC_PRECACHE_EXTENTS:
1800 	{
1801 		int ret;
1802 
1803 		inode_lock_shared(inode);
1804 		ret = ext4_ext_precache(inode);
1805 		inode_unlock_shared(inode);
1806 		return ret;
1807 	}
1808 	case FS_IOC_SET_ENCRYPTION_POLICY:
1809 		if (!ext4_has_feature_encrypt(sb))
1810 			return -EOPNOTSUPP;
1811 		return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1812 
1813 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1814 		return ext4_ioctl_get_encryption_pwsalt(filp, (void __user *)arg);
1815 
1816 	case FS_IOC_GET_ENCRYPTION_POLICY:
1817 		if (!ext4_has_feature_encrypt(sb))
1818 			return -EOPNOTSUPP;
1819 		return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1820 
1821 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1822 		if (!ext4_has_feature_encrypt(sb))
1823 			return -EOPNOTSUPP;
1824 		return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
1825 
1826 	case FS_IOC_ADD_ENCRYPTION_KEY:
1827 		if (!ext4_has_feature_encrypt(sb))
1828 			return -EOPNOTSUPP;
1829 		return fscrypt_ioctl_add_key(filp, (void __user *)arg);
1830 
1831 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1832 		if (!ext4_has_feature_encrypt(sb))
1833 			return -EOPNOTSUPP;
1834 		return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
1835 
1836 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1837 		if (!ext4_has_feature_encrypt(sb))
1838 			return -EOPNOTSUPP;
1839 		return fscrypt_ioctl_remove_key_all_users(filp,
1840 							  (void __user *)arg);
1841 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1842 		if (!ext4_has_feature_encrypt(sb))
1843 			return -EOPNOTSUPP;
1844 		return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
1845 
1846 	case FS_IOC_GET_ENCRYPTION_NONCE:
1847 		if (!ext4_has_feature_encrypt(sb))
1848 			return -EOPNOTSUPP;
1849 		return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
1850 
1851 	case EXT4_IOC_CLEAR_ES_CACHE:
1852 	{
1853 		if (!inode_owner_or_capable(idmap, inode))
1854 			return -EACCES;
1855 		ext4_clear_inode_es(inode);
1856 		return 0;
1857 	}
1858 
1859 	case EXT4_IOC_GETSTATE:
1860 	{
1861 		__u32	state = 0;
1862 
1863 		if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED))
1864 			state |= EXT4_STATE_FLAG_EXT_PRECACHED;
1865 		if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
1866 			state |= EXT4_STATE_FLAG_NEW;
1867 		if (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
1868 			state |= EXT4_STATE_FLAG_NEWENTRY;
1869 		if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE))
1870 			state |= EXT4_STATE_FLAG_DA_ALLOC_CLOSE;
1871 
1872 		return put_user(state, (__u32 __user *) arg);
1873 	}
1874 
1875 	case EXT4_IOC_GET_ES_CACHE:
1876 		return ext4_ioctl_get_es_cache(filp, arg);
1877 
1878 	case EXT4_IOC_SHUTDOWN:
1879 		return ext4_ioctl_shutdown(sb, arg);
1880 
1881 	case FS_IOC_ENABLE_VERITY:
1882 		if (!ext4_has_feature_verity(sb))
1883 			return -EOPNOTSUPP;
1884 		return fsverity_ioctl_enable(filp, (const void __user *)arg);
1885 
1886 	case FS_IOC_MEASURE_VERITY:
1887 		if (!ext4_has_feature_verity(sb))
1888 			return -EOPNOTSUPP;
1889 		return fsverity_ioctl_measure(filp, (void __user *)arg);
1890 
1891 	case FS_IOC_READ_VERITY_METADATA:
1892 		if (!ext4_has_feature_verity(sb))
1893 			return -EOPNOTSUPP;
1894 		return fsverity_ioctl_read_metadata(filp,
1895 						    (const void __user *)arg);
1896 
1897 	case EXT4_IOC_CHECKPOINT:
1898 		return ext4_ioctl_checkpoint(filp, arg);
1899 
1900 	case FS_IOC_GETFSLABEL:
1901 		return ext4_ioctl_getlabel(EXT4_SB(sb), (void __user *)arg);
1902 
1903 	case FS_IOC_SETFSLABEL:
1904 		return ext4_ioctl_setlabel(filp,
1905 					   (const void __user *)arg);
1906 
1907 	case EXT4_IOC_GETFSUUID:
1908 		return ext4_ioctl_getuuid(EXT4_SB(sb), (void __user *)arg);
1909 	case EXT4_IOC_SETFSUUID:
1910 		return ext4_ioctl_setuuid(filp, (const void __user *)arg);
1911 	case EXT4_IOC_GET_TUNE_SB_PARAM:
1912 		return ext4_ioctl_get_tune_sb(EXT4_SB(sb),
1913 					      (void __user *)arg);
1914 	case EXT4_IOC_SET_TUNE_SB_PARAM:
1915 		return ext4_ioctl_set_tune_sb(filp, (void __user *)arg);
1916 	default:
1917 		return -ENOTTY;
1918 	}
1919 }
1920 
ext4_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)1921 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1922 {
1923 	return __ext4_ioctl(filp, cmd, arg);
1924 }
1925 
1926 #ifdef CONFIG_COMPAT
ext4_compat_ioctl(struct file * file,unsigned int cmd,unsigned long arg)1927 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1928 {
1929 	/* These are just misnamed, they actually get/put from/to user an int */
1930 	switch (cmd) {
1931 	case EXT4_IOC32_GETVERSION:
1932 		cmd = EXT4_IOC_GETVERSION;
1933 		break;
1934 	case EXT4_IOC32_SETVERSION:
1935 		cmd = EXT4_IOC_SETVERSION;
1936 		break;
1937 	case EXT4_IOC32_GROUP_EXTEND:
1938 		cmd = EXT4_IOC_GROUP_EXTEND;
1939 		break;
1940 	case EXT4_IOC32_GETVERSION_OLD:
1941 		cmd = EXT4_IOC_GETVERSION_OLD;
1942 		break;
1943 	case EXT4_IOC32_SETVERSION_OLD:
1944 		cmd = EXT4_IOC_SETVERSION_OLD;
1945 		break;
1946 	case EXT4_IOC32_GETRSVSZ:
1947 		cmd = EXT4_IOC_GETRSVSZ;
1948 		break;
1949 	case EXT4_IOC32_SETRSVSZ:
1950 		cmd = EXT4_IOC_SETRSVSZ;
1951 		break;
1952 	case EXT4_IOC32_GROUP_ADD: {
1953 		struct compat_ext4_new_group_input __user *uinput;
1954 		struct ext4_new_group_data input;
1955 		int err;
1956 
1957 		uinput = compat_ptr(arg);
1958 		err = get_user(input.group, &uinput->group);
1959 		err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1960 		err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1961 		err |= get_user(input.inode_table, &uinput->inode_table);
1962 		err |= get_user(input.blocks_count, &uinput->blocks_count);
1963 		err |= get_user(input.reserved_blocks,
1964 				&uinput->reserved_blocks);
1965 		if (err)
1966 			return -EFAULT;
1967 		return ext4_ioctl_group_add(file, &input);
1968 	}
1969 	case EXT4_IOC_MOVE_EXT:
1970 	case EXT4_IOC_RESIZE_FS:
1971 	case FITRIM:
1972 	case EXT4_IOC_PRECACHE_EXTENTS:
1973 	case FS_IOC_SET_ENCRYPTION_POLICY:
1974 	case FS_IOC_GET_ENCRYPTION_PWSALT:
1975 	case FS_IOC_GET_ENCRYPTION_POLICY:
1976 	case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1977 	case FS_IOC_ADD_ENCRYPTION_KEY:
1978 	case FS_IOC_REMOVE_ENCRYPTION_KEY:
1979 	case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1980 	case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1981 	case FS_IOC_GET_ENCRYPTION_NONCE:
1982 	case EXT4_IOC_SHUTDOWN:
1983 	case FS_IOC_GETFSMAP:
1984 	case FS_IOC_ENABLE_VERITY:
1985 	case FS_IOC_MEASURE_VERITY:
1986 	case FS_IOC_READ_VERITY_METADATA:
1987 	case EXT4_IOC_CLEAR_ES_CACHE:
1988 	case EXT4_IOC_GETSTATE:
1989 	case EXT4_IOC_GET_ES_CACHE:
1990 	case EXT4_IOC_CHECKPOINT:
1991 	case FS_IOC_GETFSLABEL:
1992 	case FS_IOC_SETFSLABEL:
1993 	case EXT4_IOC_GETFSUUID:
1994 	case EXT4_IOC_SETFSUUID:
1995 		break;
1996 	default:
1997 		return -ENOIOCTLCMD;
1998 	}
1999 	return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
2000 }
2001 #endif
2002 
set_overhead(struct ext4_sb_info * sbi,struct ext4_super_block * es,const void * arg)2003 static void set_overhead(struct ext4_sb_info *sbi,
2004 			 struct ext4_super_block *es, const void *arg)
2005 {
2006 	es->s_overhead_clusters = cpu_to_le32(*((unsigned long *) arg));
2007 }
2008 
ext4_update_overhead(struct super_block * sb,bool force)2009 int ext4_update_overhead(struct super_block *sb, bool force)
2010 {
2011 	struct ext4_sb_info *sbi = EXT4_SB(sb);
2012 
2013 	if (ext4_emergency_state(sb) || sb_rdonly(sb))
2014 		return 0;
2015 	if (!force &&
2016 	    (sbi->s_overhead == 0 ||
2017 	     sbi->s_overhead == le32_to_cpu(sbi->s_es->s_overhead_clusters)))
2018 		return 0;
2019 	return ext4_update_superblocks_fn(sb, set_overhead, &sbi->s_overhead);
2020 }
2021