1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ext4/ioctl.c
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)
12 #include <linux/capability.h>
13 #include <linux/time.h>
14 #include <linux/compat.h>
15 #include <linux/mount.h>
16 #include <linux/file.h>
17 #include <linux/quotaops.h>
18 #include <linux/random.h>
19 #include <linux/uuid.h>
20 #include <linux/uaccess.h>
21 #include <linux/delay.h>
22 #include <linux/iversion.h>
23 #include <linux/fileattr.h>
24 #include "ext4_jbd2.h"
26 #include <linux/fsmap.h>
28 #include <trace/events/ext4.h>
31 * Swap memory between @a and @b for @len bytes.
33 * @a: pointer to first memory area
34 * @b: pointer to second memory area
35 * @len: number of bytes to swap
38 static void memswap(void *a, void *b, size_t len)
40 unsigned char *ap, *bp;
42 ap = (unsigned char *)a;
43 bp = (unsigned char *)b;
52 * Swap i_data and associated attributes between @inode1 and @inode2.
53 * This function is used for the primary swap between inode1 and inode2
54 * and also to revert this primary swap in case of errors.
56 * Therefore you have to make sure, that calling this method twice
57 * will revert all changes.
59 * @inode1: pointer to first inode
60 * @inode2: pointer to second inode
62 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
65 struct ext4_inode_info *ei1;
66 struct ext4_inode_info *ei2;
72 swap(inode1->i_version, inode2->i_version);
73 swap(inode1->i_atime, inode2->i_atime);
74 swap(inode1->i_mtime, inode2->i_mtime);
76 memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
77 tmp = ei1->i_flags & EXT4_FL_SHOULD_SWAP;
78 ei1->i_flags = (ei2->i_flags & EXT4_FL_SHOULD_SWAP) |
79 (ei1->i_flags & ~EXT4_FL_SHOULD_SWAP);
80 ei2->i_flags = tmp | (ei2->i_flags & ~EXT4_FL_SHOULD_SWAP);
81 swap(ei1->i_disksize, ei2->i_disksize);
82 ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
83 ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
85 isize = i_size_read(inode1);
86 i_size_write(inode1, i_size_read(inode2));
87 i_size_write(inode2, isize);
90 void ext4_reset_inode_seed(struct inode *inode)
92 struct ext4_inode_info *ei = EXT4_I(inode);
93 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
94 __le32 inum = cpu_to_le32(inode->i_ino);
95 __le32 gen = cpu_to_le32(inode->i_generation);
98 if (!ext4_has_metadata_csum(inode->i_sb))
101 csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum, sizeof(inum));
102 ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen, sizeof(gen));
106 * Swap the information from the given @inode and the inode
107 * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
108 * important fields of the inodes.
110 * @sb: the super block of the filesystem
111 * @mnt_userns: user namespace of the mount the inode was found from
112 * @inode: the inode to swap with EXT4_BOOT_LOADER_INO
115 static long swap_inode_boot_loader(struct super_block *sb,
116 struct user_namespace *mnt_userns,
121 struct inode *inode_bl;
122 struct ext4_inode_info *ei_bl;
123 qsize_t size, size_bl, diff;
125 unsigned short bytes;
127 inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO,
128 EXT4_IGET_SPECIAL | EXT4_IGET_BAD);
129 if (IS_ERR(inode_bl))
130 return PTR_ERR(inode_bl);
131 ei_bl = EXT4_I(inode_bl);
133 /* Protect orig inodes against a truncate and make sure,
134 * that only 1 swap_inode_boot_loader is running. */
135 lock_two_nondirectories(inode, inode_bl);
137 if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode) ||
138 IS_SWAPFILE(inode) || IS_ENCRYPTED(inode) ||
139 (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL) ||
140 ext4_has_inline_data(inode)) {
142 goto journal_err_out;
145 if (IS_RDONLY(inode) || IS_APPEND(inode) || IS_IMMUTABLE(inode) ||
146 !inode_owner_or_capable(mnt_userns, inode) ||
147 !capable(CAP_SYS_ADMIN)) {
149 goto journal_err_out;
152 filemap_invalidate_lock(inode->i_mapping);
153 err = filemap_write_and_wait(inode->i_mapping);
157 err = filemap_write_and_wait(inode_bl->i_mapping);
161 /* Wait for all existing dio workers */
162 inode_dio_wait(inode);
163 inode_dio_wait(inode_bl);
165 truncate_inode_pages(&inode->i_data, 0);
166 truncate_inode_pages(&inode_bl->i_data, 0);
168 handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
169 if (IS_ERR(handle)) {
173 ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_SWAP_BOOT, handle);
175 /* Protect extent tree against block allocations via delalloc */
176 ext4_double_down_write_data_sem(inode, inode_bl);
178 if (is_bad_inode(inode_bl) || !S_ISREG(inode_bl->i_mode)) {
179 /* this inode has never been used as a BOOT_LOADER */
180 set_nlink(inode_bl, 1);
181 i_uid_write(inode_bl, 0);
182 i_gid_write(inode_bl, 0);
183 inode_bl->i_flags = 0;
185 inode_set_iversion(inode_bl, 1);
186 i_size_write(inode_bl, 0);
187 EXT4_I(inode_bl)->i_disksize = inode_bl->i_size;
188 inode_bl->i_mode = S_IFREG;
189 if (ext4_has_feature_extents(sb)) {
190 ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
191 ext4_ext_tree_init(handle, inode_bl);
193 memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
196 err = dquot_initialize(inode);
200 size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
201 size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
202 diff = size - size_bl;
203 swap_inode_data(inode, inode_bl);
205 inode->i_ctime = inode_bl->i_ctime = current_time(inode);
207 inode->i_generation = prandom_u32();
208 inode_bl->i_generation = prandom_u32();
209 ext4_reset_inode_seed(inode);
210 ext4_reset_inode_seed(inode_bl);
212 ext4_discard_preallocations(inode, 0);
214 err = ext4_mark_inode_dirty(handle, inode);
216 /* No need to update quota information. */
217 ext4_warning(inode->i_sb,
218 "couldn't mark inode #%lu dirty (err %d)",
220 /* Revert all changes: */
221 swap_inode_data(inode, inode_bl);
222 ext4_mark_inode_dirty(handle, inode);
226 blocks = inode_bl->i_blocks;
227 bytes = inode_bl->i_bytes;
228 inode_bl->i_blocks = inode->i_blocks;
229 inode_bl->i_bytes = inode->i_bytes;
230 err = ext4_mark_inode_dirty(handle, inode_bl);
232 /* No need to update quota information. */
233 ext4_warning(inode_bl->i_sb,
234 "couldn't mark inode #%lu dirty (err %d)",
235 inode_bl->i_ino, err);
239 /* Bootloader inode should not be counted into quota information. */
241 dquot_free_space(inode, diff);
243 err = dquot_alloc_space(inode, -1 * diff);
247 /* Revert all changes: */
248 inode_bl->i_blocks = blocks;
249 inode_bl->i_bytes = bytes;
250 swap_inode_data(inode, inode_bl);
251 ext4_mark_inode_dirty(handle, inode);
252 ext4_mark_inode_dirty(handle, inode_bl);
256 ext4_journal_stop(handle);
257 ext4_double_up_write_data_sem(inode, inode_bl);
260 filemap_invalidate_unlock(inode->i_mapping);
262 unlock_two_nondirectories(inode, inode_bl);
267 #ifdef CONFIG_FS_ENCRYPTION
268 static int uuid_is_zero(__u8 u[16])
272 for (i = 0; i < 16; i++)
280 * If immutable is set and we are not clearing it, we're not allowed to change
281 * anything else in the inode. Don't error out if we're only trying to set
282 * immutable on an immutable file.
284 static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
287 struct ext4_inode_info *ei = EXT4_I(inode);
288 unsigned int oldflags = ei->i_flags;
290 if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
293 if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
295 if (ext4_has_feature_project(inode->i_sb) &&
296 __kprojid_val(ei->i_projid) != new_projid)
302 static void ext4_dax_dontcache(struct inode *inode, unsigned int flags)
304 struct ext4_inode_info *ei = EXT4_I(inode);
306 if (S_ISDIR(inode->i_mode))
309 if (test_opt2(inode->i_sb, DAX_NEVER) ||
310 test_opt(inode->i_sb, DAX_ALWAYS))
313 if ((ei->i_flags ^ flags) & EXT4_DAX_FL)
314 d_mark_dontcache(inode);
317 static bool dax_compatible(struct inode *inode, unsigned int oldflags,
320 /* Allow the DAX flag to be changed on inline directories */
321 if (S_ISDIR(inode->i_mode)) {
322 flags &= ~EXT4_INLINE_DATA_FL;
323 oldflags &= ~EXT4_INLINE_DATA_FL;
326 if (flags & EXT4_DAX_FL) {
327 if ((oldflags & EXT4_DAX_MUT_EXCL) ||
328 ext4_test_inode_state(inode,
329 EXT4_STATE_VERITY_IN_PROGRESS)) {
334 if ((flags & EXT4_DAX_MUT_EXCL) && (oldflags & EXT4_DAX_FL))
340 static int ext4_ioctl_setflags(struct inode *inode,
343 struct ext4_inode_info *ei = EXT4_I(inode);
344 handle_t *handle = NULL;
345 int err = -EPERM, migrate = 0;
346 struct ext4_iloc iloc;
347 unsigned int oldflags, mask, i;
348 struct super_block *sb = inode->i_sb;
350 /* Is it quota file? Do not allow user to mess with it */
351 if (ext4_is_quota_file(inode))
354 oldflags = ei->i_flags;
356 * The JOURNAL_DATA flag can only be changed by
357 * the relevant capability.
359 if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
360 if (!capable(CAP_SYS_RESOURCE))
364 if (!dax_compatible(inode, oldflags, flags)) {
369 if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
372 if ((flags ^ oldflags) & EXT4_CASEFOLD_FL) {
373 if (!ext4_has_feature_casefold(sb)) {
378 if (!S_ISDIR(inode->i_mode)) {
383 if (!ext4_empty_dir(inode)) {
390 * Wait for all pending directio and then flush all the dirty pages
391 * for this file. The flush marks all the pages readonly, so any
392 * subsequent attempt to write to the file (particularly mmap pages)
393 * will come through the filesystem and fail.
395 if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
396 (flags & EXT4_IMMUTABLE_FL)) {
397 inode_dio_wait(inode);
398 err = filemap_write_and_wait(inode->i_mapping);
403 handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
404 if (IS_ERR(handle)) {
405 err = PTR_ERR(handle);
409 ext4_handle_sync(handle);
410 err = ext4_reserve_inode_write(handle, inode, &iloc);
414 ext4_dax_dontcache(inode, flags);
416 for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
417 if (!(mask & EXT4_FL_USER_MODIFIABLE))
419 /* These flags get special treatment later */
420 if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
423 ext4_set_inode_flag(inode, i);
425 ext4_clear_inode_flag(inode, i);
428 ext4_set_inode_flags(inode, false);
430 inode->i_ctime = current_time(inode);
432 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
434 ext4_journal_stop(handle);
438 if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
440 * Changes to the journaling mode can cause unsafe changes to
441 * S_DAX if the inode is DAX
448 err = ext4_change_inode_journal_flag(inode,
449 flags & EXT4_JOURNAL_DATA_FL);
454 if (flags & EXT4_EXTENTS_FL)
455 err = ext4_ext_migrate(inode);
457 err = ext4_ind_migrate(inode);
465 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
467 struct super_block *sb = inode->i_sb;
468 struct ext4_inode_info *ei = EXT4_I(inode);
472 struct ext4_iloc iloc;
473 struct ext4_inode *raw_inode;
474 struct dquot *transfer_to[MAXQUOTAS] = { };
476 if (!ext4_has_feature_project(sb)) {
477 if (projid != EXT4_DEF_PROJID)
483 if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
486 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
488 if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
492 /* Is it quota file? Do not allow user to mess with it */
493 if (ext4_is_quota_file(inode))
496 err = dquot_initialize(inode);
500 err = ext4_get_inode_loc(inode, &iloc);
504 raw_inode = ext4_raw_inode(&iloc);
505 if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
506 err = ext4_expand_extra_isize(inode,
507 EXT4_SB(sb)->s_want_extra_isize,
515 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
516 EXT4_QUOTA_INIT_BLOCKS(sb) +
517 EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
519 return PTR_ERR(handle);
521 err = ext4_reserve_inode_write(handle, inode, &iloc);
525 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
526 if (!IS_ERR(transfer_to[PRJQUOTA])) {
528 /* __dquot_transfer() calls back ext4_get_inode_usage() which
529 * counts xattr inode references.
531 down_read(&EXT4_I(inode)->xattr_sem);
532 err = __dquot_transfer(inode, transfer_to);
533 up_read(&EXT4_I(inode)->xattr_sem);
534 dqput(transfer_to[PRJQUOTA]);
539 EXT4_I(inode)->i_projid = kprojid;
540 inode->i_ctime = current_time(inode);
542 rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
546 ext4_journal_stop(handle);
550 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
552 if (projid != EXT4_DEF_PROJID)
558 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
560 struct ext4_sb_info *sbi = EXT4_SB(sb);
564 if (!capable(CAP_SYS_ADMIN))
567 if (get_user(flags, (__u32 __user *)arg))
570 if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
573 if (ext4_forced_shutdown(sbi))
576 ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
577 trace_ext4_shutdown(sb, flags);
580 case EXT4_GOING_FLAGS_DEFAULT:
581 ret = freeze_bdev(sb->s_bdev);
584 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
585 thaw_bdev(sb->s_bdev);
587 case EXT4_GOING_FLAGS_LOGFLUSH:
588 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
589 if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
590 (void) ext4_force_commit(sb);
591 jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
594 case EXT4_GOING_FLAGS_NOLOGFLUSH:
595 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
596 if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
597 jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
602 clear_opt(sb, DISCARD);
606 struct getfsmap_info {
607 struct super_block *gi_sb;
608 struct fsmap_head __user *gi_data;
613 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
615 struct getfsmap_info *info = priv;
618 trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
620 info->gi_last_flags = xfm->fmr_flags;
621 ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
622 if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
623 sizeof(struct fsmap)))
629 static int ext4_ioc_getfsmap(struct super_block *sb,
630 struct fsmap_head __user *arg)
632 struct getfsmap_info info = { NULL };
633 struct ext4_fsmap_head xhead = {0};
634 struct fsmap_head head;
635 bool aborted = false;
638 if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
640 if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
641 memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
642 sizeof(head.fmh_keys[0].fmr_reserved)) ||
643 memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
644 sizeof(head.fmh_keys[1].fmr_reserved)))
647 * ext4 doesn't report file extents at all, so the only valid
648 * file offsets are the magic ones (all zeroes or all ones).
650 if (head.fmh_keys[0].fmr_offset ||
651 (head.fmh_keys[1].fmr_offset != 0 &&
652 head.fmh_keys[1].fmr_offset != -1ULL))
655 xhead.fmh_iflags = head.fmh_iflags;
656 xhead.fmh_count = head.fmh_count;
657 ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
658 ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
660 trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
661 trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
665 error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
666 if (error == EXT4_QUERY_RANGE_ABORT)
671 /* If we didn't abort, set the "last" flag in the last fmx */
672 if (!aborted && info.gi_idx) {
673 info.gi_last_flags |= FMR_OF_LAST;
674 if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
676 sizeof(info.gi_last_flags)))
680 /* copy back header */
681 head.fmh_entries = xhead.fmh_entries;
682 head.fmh_oflags = xhead.fmh_oflags;
683 if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
689 static long ext4_ioctl_group_add(struct file *file,
690 struct ext4_new_group_data *input)
692 struct super_block *sb = file_inode(file)->i_sb;
695 err = ext4_resize_begin(sb);
699 if (ext4_has_feature_bigalloc(sb)) {
700 ext4_msg(sb, KERN_ERR,
701 "Online resizing not supported with bigalloc");
706 err = mnt_want_write_file(file);
710 err = ext4_group_add(sb, input);
711 if (EXT4_SB(sb)->s_journal) {
712 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
713 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
714 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
718 mnt_drop_write_file(file);
719 if (!err && ext4_has_group_desc_csum(sb) &&
720 test_opt(sb, INIT_INODE_TABLE))
721 err = ext4_register_li_request(sb, input->group);
727 int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa)
729 struct inode *inode = d_inode(dentry);
730 struct ext4_inode_info *ei = EXT4_I(inode);
731 u32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
733 if (S_ISREG(inode->i_mode))
734 flags &= ~FS_PROJINHERIT_FL;
736 fileattr_fill_flags(fa, flags);
737 if (ext4_has_feature_project(inode->i_sb))
738 fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
743 int ext4_fileattr_set(struct user_namespace *mnt_userns,
744 struct dentry *dentry, struct fileattr *fa)
746 struct inode *inode = d_inode(dentry);
747 u32 flags = fa->flags;
748 int err = -EOPNOTSUPP;
750 if (flags & ~EXT4_FL_USER_VISIBLE)
754 * chattr(1) grabs flags via GETFLAGS, modifies the result and
755 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
756 * more restrictive than just silently masking off visible but
757 * not settable flags as we always did.
759 flags &= EXT4_FL_USER_MODIFIABLE;
760 if (ext4_mask_flags(inode->i_mode, flags) != flags)
762 err = ext4_ioctl_check_immutable(inode, fa->fsx_projid, flags);
765 err = ext4_ioctl_setflags(inode, flags);
768 err = ext4_ioctl_setproject(inode, fa->fsx_projid);
773 /* So that the fiemap access checks can't overflow on 32 bit machines. */
774 #define FIEMAP_MAX_EXTENTS (UINT_MAX / sizeof(struct fiemap_extent))
776 static int ext4_ioctl_get_es_cache(struct file *filp, unsigned long arg)
778 struct fiemap fiemap;
779 struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
780 struct fiemap_extent_info fieinfo = { 0, };
781 struct inode *inode = file_inode(filp);
784 if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
787 if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
790 fieinfo.fi_flags = fiemap.fm_flags;
791 fieinfo.fi_extents_max = fiemap.fm_extent_count;
792 fieinfo.fi_extents_start = ufiemap->fm_extents;
794 error = ext4_get_es_cache(inode, &fieinfo, fiemap.fm_start,
796 fiemap.fm_flags = fieinfo.fi_flags;
797 fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
798 if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
804 static int ext4_ioctl_checkpoint(struct file *filp, unsigned long arg)
808 unsigned int flush_flags = 0;
809 struct super_block *sb = file_inode(filp)->i_sb;
810 struct request_queue *q;
812 if (copy_from_user(&flags, (__u32 __user *)arg,
816 if (!capable(CAP_SYS_ADMIN))
819 /* check for invalid bits set */
820 if ((flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID) ||
821 ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
822 (flags & JBD2_JOURNAL_FLUSH_ZEROOUT)))
825 if (!EXT4_SB(sb)->s_journal)
828 if (flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID)
831 q = bdev_get_queue(EXT4_SB(sb)->s_journal->j_dev);
834 if ((flags & JBD2_JOURNAL_FLUSH_DISCARD) && !blk_queue_discard(q))
837 if (flags & EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
840 if (flags & EXT4_IOC_CHECKPOINT_FLAG_DISCARD)
841 flush_flags |= JBD2_JOURNAL_FLUSH_DISCARD;
843 if (flags & EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT) {
844 flush_flags |= JBD2_JOURNAL_FLUSH_ZEROOUT;
845 pr_info_ratelimited("warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow");
848 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
849 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, flush_flags);
850 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
855 static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
857 struct inode *inode = file_inode(filp);
858 struct super_block *sb = inode->i_sb;
859 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
861 ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
864 case FS_IOC_GETFSMAP:
865 return ext4_ioc_getfsmap(sb, (void __user *)arg);
866 case EXT4_IOC_GETVERSION:
867 case EXT4_IOC_GETVERSION_OLD:
868 return put_user(inode->i_generation, (int __user *) arg);
869 case EXT4_IOC_SETVERSION:
870 case EXT4_IOC_SETVERSION_OLD: {
872 struct ext4_iloc iloc;
876 if (!inode_owner_or_capable(mnt_userns, inode))
879 if (ext4_has_metadata_csum(inode->i_sb)) {
880 ext4_warning(sb, "Setting inode version is not "
881 "supported with metadata_csum enabled.");
885 err = mnt_want_write_file(filp);
888 if (get_user(generation, (int __user *) arg)) {
894 handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
895 if (IS_ERR(handle)) {
896 err = PTR_ERR(handle);
899 err = ext4_reserve_inode_write(handle, inode, &iloc);
901 inode->i_ctime = current_time(inode);
902 inode->i_generation = generation;
903 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
905 ext4_journal_stop(handle);
910 mnt_drop_write_file(filp);
913 case EXT4_IOC_GROUP_EXTEND: {
914 ext4_fsblk_t n_blocks_count;
917 err = ext4_resize_begin(sb);
921 if (get_user(n_blocks_count, (__u32 __user *)arg)) {
923 goto group_extend_out;
926 if (ext4_has_feature_bigalloc(sb)) {
927 ext4_msg(sb, KERN_ERR,
928 "Online resizing not supported with bigalloc");
930 goto group_extend_out;
933 err = mnt_want_write_file(filp);
935 goto group_extend_out;
937 err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
938 if (EXT4_SB(sb)->s_journal) {
939 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
940 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
941 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
945 mnt_drop_write_file(filp);
951 case EXT4_IOC_MOVE_EXT: {
952 struct move_extent me;
956 if (!(filp->f_mode & FMODE_READ) ||
957 !(filp->f_mode & FMODE_WRITE))
960 if (copy_from_user(&me,
961 (struct move_extent __user *)arg, sizeof(me)))
965 donor = fdget(me.donor_fd);
969 if (!(donor.file->f_mode & FMODE_WRITE)) {
974 if (ext4_has_feature_bigalloc(sb)) {
975 ext4_msg(sb, KERN_ERR,
976 "Online defrag not supported with bigalloc");
979 } else if (IS_DAX(inode)) {
980 ext4_msg(sb, KERN_ERR,
981 "Online defrag not supported with DAX");
986 err = mnt_want_write_file(filp);
990 err = ext4_move_extents(filp, donor.file, me.orig_start,
991 me.donor_start, me.len, &me.moved_len);
992 mnt_drop_write_file(filp);
994 if (copy_to_user((struct move_extent __user *)arg,
1002 case EXT4_IOC_GROUP_ADD: {
1003 struct ext4_new_group_data input;
1005 if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
1009 return ext4_ioctl_group_add(filp, &input);
1012 case EXT4_IOC_MIGRATE:
1015 if (!inode_owner_or_capable(mnt_userns, inode))
1018 err = mnt_want_write_file(filp);
1022 * inode_mutex prevent write and truncate on the file.
1023 * Read still goes through. We take i_data_sem in
1024 * ext4_ext_swap_inode_data before we switch the
1025 * inode format to prevent read.
1027 inode_lock((inode));
1028 err = ext4_ext_migrate(inode);
1029 inode_unlock((inode));
1030 mnt_drop_write_file(filp);
1034 case EXT4_IOC_ALLOC_DA_BLKS:
1037 if (!inode_owner_or_capable(mnt_userns, inode))
1040 err = mnt_want_write_file(filp);
1043 err = ext4_alloc_da_blocks(inode);
1044 mnt_drop_write_file(filp);
1048 case EXT4_IOC_SWAP_BOOT:
1051 if (!(filp->f_mode & FMODE_WRITE))
1053 err = mnt_want_write_file(filp);
1056 err = swap_inode_boot_loader(sb, mnt_userns, inode);
1057 mnt_drop_write_file(filp);
1061 case EXT4_IOC_RESIZE_FS: {
1062 ext4_fsblk_t n_blocks_count;
1063 int err = 0, err2 = 0;
1064 ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
1066 if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1071 err = ext4_resize_begin(sb);
1075 err = mnt_want_write_file(filp);
1079 err = ext4_resize_fs(sb, n_blocks_count);
1080 if (EXT4_SB(sb)->s_journal) {
1081 ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE, NULL);
1082 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1083 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1084 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1088 mnt_drop_write_file(filp);
1089 if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1090 ext4_has_group_desc_csum(sb) &&
1091 test_opt(sb, INIT_INODE_TABLE))
1092 err = ext4_register_li_request(sb, o_group);
1095 ext4_resize_end(sb);
1101 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1102 struct fstrim_range range;
1105 if (!capable(CAP_SYS_ADMIN))
1108 if (!blk_queue_discard(q))
1112 * We haven't replayed the journal, so we cannot use our
1113 * block-bitmap-guided storage zapping commands.
1115 if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1118 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1122 ret = ext4_trim_fs(sb, &range);
1126 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1132 case EXT4_IOC_PRECACHE_EXTENTS:
1133 return ext4_ext_precache(inode);
1135 case FS_IOC_SET_ENCRYPTION_POLICY:
1136 if (!ext4_has_feature_encrypt(sb))
1138 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1140 case FS_IOC_GET_ENCRYPTION_PWSALT: {
1141 #ifdef CONFIG_FS_ENCRYPTION
1143 struct ext4_sb_info *sbi = EXT4_SB(sb);
1146 if (!ext4_has_feature_encrypt(sb))
1148 if (uuid_is_zero(sbi->s_es->s_encrypt_pw_salt)) {
1149 err = mnt_want_write_file(filp);
1152 handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
1153 if (IS_ERR(handle)) {
1154 err = PTR_ERR(handle);
1155 goto pwsalt_err_exit;
1157 err = ext4_journal_get_write_access(handle, sb,
1161 goto pwsalt_err_journal;
1162 lock_buffer(sbi->s_sbh);
1163 generate_random_uuid(sbi->s_es->s_encrypt_pw_salt);
1164 ext4_superblock_csum_set(sb);
1165 unlock_buffer(sbi->s_sbh);
1166 err = ext4_handle_dirty_metadata(handle, NULL,
1169 err2 = ext4_journal_stop(handle);
1173 mnt_drop_write_file(filp);
1177 if (copy_to_user((void __user *) arg,
1178 sbi->s_es->s_encrypt_pw_salt, 16))
1185 case FS_IOC_GET_ENCRYPTION_POLICY:
1186 if (!ext4_has_feature_encrypt(sb))
1188 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1190 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1191 if (!ext4_has_feature_encrypt(sb))
1193 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
1195 case FS_IOC_ADD_ENCRYPTION_KEY:
1196 if (!ext4_has_feature_encrypt(sb))
1198 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
1200 case FS_IOC_REMOVE_ENCRYPTION_KEY:
1201 if (!ext4_has_feature_encrypt(sb))
1203 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
1205 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1206 if (!ext4_has_feature_encrypt(sb))
1208 return fscrypt_ioctl_remove_key_all_users(filp,
1209 (void __user *)arg);
1210 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1211 if (!ext4_has_feature_encrypt(sb))
1213 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
1215 case FS_IOC_GET_ENCRYPTION_NONCE:
1216 if (!ext4_has_feature_encrypt(sb))
1218 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
1220 case EXT4_IOC_CLEAR_ES_CACHE:
1222 if (!inode_owner_or_capable(mnt_userns, inode))
1224 ext4_clear_inode_es(inode);
1228 case EXT4_IOC_GETSTATE:
1232 if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED))
1233 state |= EXT4_STATE_FLAG_EXT_PRECACHED;
1234 if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
1235 state |= EXT4_STATE_FLAG_NEW;
1236 if (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
1237 state |= EXT4_STATE_FLAG_NEWENTRY;
1238 if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE))
1239 state |= EXT4_STATE_FLAG_DA_ALLOC_CLOSE;
1241 return put_user(state, (__u32 __user *) arg);
1244 case EXT4_IOC_GET_ES_CACHE:
1245 return ext4_ioctl_get_es_cache(filp, arg);
1247 case EXT4_IOC_SHUTDOWN:
1248 return ext4_shutdown(sb, arg);
1250 case FS_IOC_ENABLE_VERITY:
1251 if (!ext4_has_feature_verity(sb))
1253 return fsverity_ioctl_enable(filp, (const void __user *)arg);
1255 case FS_IOC_MEASURE_VERITY:
1256 if (!ext4_has_feature_verity(sb))
1258 return fsverity_ioctl_measure(filp, (void __user *)arg);
1260 case FS_IOC_READ_VERITY_METADATA:
1261 if (!ext4_has_feature_verity(sb))
1263 return fsverity_ioctl_read_metadata(filp,
1264 (const void __user *)arg);
1266 case EXT4_IOC_CHECKPOINT:
1267 return ext4_ioctl_checkpoint(filp, arg);
1274 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1276 return __ext4_ioctl(filp, cmd, arg);
1279 #ifdef CONFIG_COMPAT
1280 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1282 /* These are just misnamed, they actually get/put from/to user an int */
1284 case EXT4_IOC32_GETVERSION:
1285 cmd = EXT4_IOC_GETVERSION;
1287 case EXT4_IOC32_SETVERSION:
1288 cmd = EXT4_IOC_SETVERSION;
1290 case EXT4_IOC32_GROUP_EXTEND:
1291 cmd = EXT4_IOC_GROUP_EXTEND;
1293 case EXT4_IOC32_GETVERSION_OLD:
1294 cmd = EXT4_IOC_GETVERSION_OLD;
1296 case EXT4_IOC32_SETVERSION_OLD:
1297 cmd = EXT4_IOC_SETVERSION_OLD;
1299 case EXT4_IOC32_GETRSVSZ:
1300 cmd = EXT4_IOC_GETRSVSZ;
1302 case EXT4_IOC32_SETRSVSZ:
1303 cmd = EXT4_IOC_SETRSVSZ;
1305 case EXT4_IOC32_GROUP_ADD: {
1306 struct compat_ext4_new_group_input __user *uinput;
1307 struct ext4_new_group_data input;
1310 uinput = compat_ptr(arg);
1311 err = get_user(input.group, &uinput->group);
1312 err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1313 err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1314 err |= get_user(input.inode_table, &uinput->inode_table);
1315 err |= get_user(input.blocks_count, &uinput->blocks_count);
1316 err |= get_user(input.reserved_blocks,
1317 &uinput->reserved_blocks);
1320 return ext4_ioctl_group_add(file, &input);
1322 case EXT4_IOC_MOVE_EXT:
1323 case EXT4_IOC_RESIZE_FS:
1325 case EXT4_IOC_PRECACHE_EXTENTS:
1326 case FS_IOC_SET_ENCRYPTION_POLICY:
1327 case FS_IOC_GET_ENCRYPTION_PWSALT:
1328 case FS_IOC_GET_ENCRYPTION_POLICY:
1329 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1330 case FS_IOC_ADD_ENCRYPTION_KEY:
1331 case FS_IOC_REMOVE_ENCRYPTION_KEY:
1332 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1333 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1334 case FS_IOC_GET_ENCRYPTION_NONCE:
1335 case EXT4_IOC_SHUTDOWN:
1336 case FS_IOC_GETFSMAP:
1337 case FS_IOC_ENABLE_VERITY:
1338 case FS_IOC_MEASURE_VERITY:
1339 case FS_IOC_READ_VERITY_METADATA:
1340 case EXT4_IOC_CLEAR_ES_CACHE:
1341 case EXT4_IOC_GETSTATE:
1342 case EXT4_IOC_GET_ES_CACHE:
1343 case EXT4_IOC_CHECKPOINT:
1346 return -ENOIOCTLCMD;
1348 return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));