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, EXT4_IGET_SPECIAL);
128 if (IS_ERR(inode_bl))
129 return PTR_ERR(inode_bl);
130 ei_bl = EXT4_I(inode_bl);
132 /* Protect orig inodes against a truncate and make sure,
133 * that only 1 swap_inode_boot_loader is running. */
134 lock_two_nondirectories(inode, inode_bl);
136 if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode) ||
137 IS_SWAPFILE(inode) || IS_ENCRYPTED(inode) ||
138 (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL) ||
139 ext4_has_inline_data(inode)) {
141 goto journal_err_out;
144 if (IS_RDONLY(inode) || IS_APPEND(inode) || IS_IMMUTABLE(inode) ||
145 !inode_owner_or_capable(mnt_userns, inode) ||
146 !capable(CAP_SYS_ADMIN)) {
148 goto journal_err_out;
151 filemap_invalidate_lock(inode->i_mapping);
152 err = filemap_write_and_wait(inode->i_mapping);
156 err = filemap_write_and_wait(inode_bl->i_mapping);
160 /* Wait for all existing dio workers */
161 inode_dio_wait(inode);
162 inode_dio_wait(inode_bl);
164 truncate_inode_pages(&inode->i_data, 0);
165 truncate_inode_pages(&inode_bl->i_data, 0);
167 handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
168 if (IS_ERR(handle)) {
172 ext4_fc_start_ineligible(sb, EXT4_FC_REASON_SWAP_BOOT);
174 /* Protect extent tree against block allocations via delalloc */
175 ext4_double_down_write_data_sem(inode, inode_bl);
177 if (inode_bl->i_nlink == 0) {
178 /* this inode has never been used as a BOOT_LOADER */
179 set_nlink(inode_bl, 1);
180 i_uid_write(inode_bl, 0);
181 i_gid_write(inode_bl, 0);
182 inode_bl->i_flags = 0;
184 inode_set_iversion(inode_bl, 1);
185 i_size_write(inode_bl, 0);
186 inode_bl->i_mode = S_IFREG;
187 if (ext4_has_feature_extents(sb)) {
188 ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
189 ext4_ext_tree_init(handle, inode_bl);
191 memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
194 err = dquot_initialize(inode);
198 size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
199 size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
200 diff = size - size_bl;
201 swap_inode_data(inode, inode_bl);
203 inode->i_ctime = inode_bl->i_ctime = current_time(inode);
205 inode->i_generation = prandom_u32();
206 inode_bl->i_generation = prandom_u32();
207 ext4_reset_inode_seed(inode);
208 ext4_reset_inode_seed(inode_bl);
210 ext4_discard_preallocations(inode, 0);
212 err = ext4_mark_inode_dirty(handle, inode);
214 /* No need to update quota information. */
215 ext4_warning(inode->i_sb,
216 "couldn't mark inode #%lu dirty (err %d)",
218 /* Revert all changes: */
219 swap_inode_data(inode, inode_bl);
220 ext4_mark_inode_dirty(handle, inode);
224 blocks = inode_bl->i_blocks;
225 bytes = inode_bl->i_bytes;
226 inode_bl->i_blocks = inode->i_blocks;
227 inode_bl->i_bytes = inode->i_bytes;
228 err = ext4_mark_inode_dirty(handle, inode_bl);
230 /* No need to update quota information. */
231 ext4_warning(inode_bl->i_sb,
232 "couldn't mark inode #%lu dirty (err %d)",
233 inode_bl->i_ino, err);
237 /* Bootloader inode should not be counted into quota information. */
239 dquot_free_space(inode, diff);
241 err = dquot_alloc_space(inode, -1 * diff);
245 /* Revert all changes: */
246 inode_bl->i_blocks = blocks;
247 inode_bl->i_bytes = bytes;
248 swap_inode_data(inode, inode_bl);
249 ext4_mark_inode_dirty(handle, inode);
250 ext4_mark_inode_dirty(handle, inode_bl);
254 ext4_journal_stop(handle);
255 ext4_fc_stop_ineligible(sb);
256 ext4_double_up_write_data_sem(inode, inode_bl);
259 filemap_invalidate_unlock(inode->i_mapping);
261 unlock_two_nondirectories(inode, inode_bl);
266 #ifdef CONFIG_FS_ENCRYPTION
267 static int uuid_is_zero(__u8 u[16])
271 for (i = 0; i < 16; i++)
279 * If immutable is set and we are not clearing it, we're not allowed to change
280 * anything else in the inode. Don't error out if we're only trying to set
281 * immutable on an immutable file.
283 static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
286 struct ext4_inode_info *ei = EXT4_I(inode);
287 unsigned int oldflags = ei->i_flags;
289 if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
292 if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
294 if (ext4_has_feature_project(inode->i_sb) &&
295 __kprojid_val(ei->i_projid) != new_projid)
301 static void ext4_dax_dontcache(struct inode *inode, unsigned int flags)
303 struct ext4_inode_info *ei = EXT4_I(inode);
305 if (S_ISDIR(inode->i_mode))
308 if (test_opt2(inode->i_sb, DAX_NEVER) ||
309 test_opt(inode->i_sb, DAX_ALWAYS))
312 if ((ei->i_flags ^ flags) & EXT4_DAX_FL)
313 d_mark_dontcache(inode);
316 static bool dax_compatible(struct inode *inode, unsigned int oldflags,
319 /* Allow the DAX flag to be changed on inline directories */
320 if (S_ISDIR(inode->i_mode)) {
321 flags &= ~EXT4_INLINE_DATA_FL;
322 oldflags &= ~EXT4_INLINE_DATA_FL;
325 if (flags & EXT4_DAX_FL) {
326 if ((oldflags & EXT4_DAX_MUT_EXCL) ||
327 ext4_test_inode_state(inode,
328 EXT4_STATE_VERITY_IN_PROGRESS)) {
333 if ((flags & EXT4_DAX_MUT_EXCL) && (oldflags & EXT4_DAX_FL))
339 static int ext4_ioctl_setflags(struct inode *inode,
342 struct ext4_inode_info *ei = EXT4_I(inode);
343 handle_t *handle = NULL;
344 int err = -EPERM, migrate = 0;
345 struct ext4_iloc iloc;
346 unsigned int oldflags, mask, i;
347 struct super_block *sb = inode->i_sb;
349 /* Is it quota file? Do not allow user to mess with it */
350 if (ext4_is_quota_file(inode))
353 oldflags = ei->i_flags;
355 * The JOURNAL_DATA flag can only be changed by
356 * the relevant capability.
358 if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
359 if (!capable(CAP_SYS_RESOURCE))
363 if (!dax_compatible(inode, oldflags, flags)) {
368 if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
371 if ((flags ^ oldflags) & EXT4_CASEFOLD_FL) {
372 if (!ext4_has_feature_casefold(sb)) {
377 if (!S_ISDIR(inode->i_mode)) {
382 if (!ext4_empty_dir(inode)) {
389 * Wait for all pending directio and then flush all the dirty pages
390 * for this file. The flush marks all the pages readonly, so any
391 * subsequent attempt to write to the file (particularly mmap pages)
392 * will come through the filesystem and fail.
394 if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
395 (flags & EXT4_IMMUTABLE_FL)) {
396 inode_dio_wait(inode);
397 err = filemap_write_and_wait(inode->i_mapping);
402 handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
403 if (IS_ERR(handle)) {
404 err = PTR_ERR(handle);
408 ext4_handle_sync(handle);
409 err = ext4_reserve_inode_write(handle, inode, &iloc);
413 ext4_dax_dontcache(inode, flags);
415 for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
416 if (!(mask & EXT4_FL_USER_MODIFIABLE))
418 /* These flags get special treatment later */
419 if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
422 ext4_set_inode_flag(inode, i);
424 ext4_clear_inode_flag(inode, i);
427 ext4_set_inode_flags(inode, false);
429 inode->i_ctime = current_time(inode);
431 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
433 ext4_journal_stop(handle);
437 if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
439 * Changes to the journaling mode can cause unsafe changes to
440 * S_DAX if the inode is DAX
447 err = ext4_change_inode_journal_flag(inode,
448 flags & EXT4_JOURNAL_DATA_FL);
453 if (flags & EXT4_EXTENTS_FL)
454 err = ext4_ext_migrate(inode);
456 err = ext4_ind_migrate(inode);
464 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
466 struct super_block *sb = inode->i_sb;
467 struct ext4_inode_info *ei = EXT4_I(inode);
471 struct ext4_iloc iloc;
472 struct ext4_inode *raw_inode;
473 struct dquot *transfer_to[MAXQUOTAS] = { };
475 if (!ext4_has_feature_project(sb)) {
476 if (projid != EXT4_DEF_PROJID)
482 if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
485 kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
487 if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
491 /* Is it quota file? Do not allow user to mess with it */
492 if (ext4_is_quota_file(inode))
495 err = ext4_get_inode_loc(inode, &iloc);
499 raw_inode = ext4_raw_inode(&iloc);
500 if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
501 err = ext4_expand_extra_isize(inode,
502 EXT4_SB(sb)->s_want_extra_isize,
510 err = dquot_initialize(inode);
514 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
515 EXT4_QUOTA_INIT_BLOCKS(sb) +
516 EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
518 return PTR_ERR(handle);
520 err = ext4_reserve_inode_write(handle, inode, &iloc);
524 transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
525 if (!IS_ERR(transfer_to[PRJQUOTA])) {
527 /* __dquot_transfer() calls back ext4_get_inode_usage() which
528 * counts xattr inode references.
530 down_read(&EXT4_I(inode)->xattr_sem);
531 err = __dquot_transfer(inode, transfer_to);
532 up_read(&EXT4_I(inode)->xattr_sem);
533 dqput(transfer_to[PRJQUOTA]);
538 EXT4_I(inode)->i_projid = kprojid;
539 inode->i_ctime = current_time(inode);
541 rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
545 ext4_journal_stop(handle);
549 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
551 if (projid != EXT4_DEF_PROJID)
557 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
559 struct ext4_sb_info *sbi = EXT4_SB(sb);
562 if (!capable(CAP_SYS_ADMIN))
565 if (get_user(flags, (__u32 __user *)arg))
568 if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
571 if (ext4_forced_shutdown(sbi))
574 ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
575 trace_ext4_shutdown(sb, flags);
578 case EXT4_GOING_FLAGS_DEFAULT:
579 freeze_bdev(sb->s_bdev);
580 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
581 thaw_bdev(sb->s_bdev);
583 case EXT4_GOING_FLAGS_LOGFLUSH:
584 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
585 if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
586 (void) ext4_force_commit(sb);
587 jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
590 case EXT4_GOING_FLAGS_NOLOGFLUSH:
591 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
592 if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
593 jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
598 clear_opt(sb, DISCARD);
602 struct getfsmap_info {
603 struct super_block *gi_sb;
604 struct fsmap_head __user *gi_data;
609 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
611 struct getfsmap_info *info = priv;
614 trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
616 info->gi_last_flags = xfm->fmr_flags;
617 ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
618 if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
619 sizeof(struct fsmap)))
625 static int ext4_ioc_getfsmap(struct super_block *sb,
626 struct fsmap_head __user *arg)
628 struct getfsmap_info info = { NULL };
629 struct ext4_fsmap_head xhead = {0};
630 struct fsmap_head head;
631 bool aborted = false;
634 if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
636 if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
637 memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
638 sizeof(head.fmh_keys[0].fmr_reserved)) ||
639 memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
640 sizeof(head.fmh_keys[1].fmr_reserved)))
643 * ext4 doesn't report file extents at all, so the only valid
644 * file offsets are the magic ones (all zeroes or all ones).
646 if (head.fmh_keys[0].fmr_offset ||
647 (head.fmh_keys[1].fmr_offset != 0 &&
648 head.fmh_keys[1].fmr_offset != -1ULL))
651 xhead.fmh_iflags = head.fmh_iflags;
652 xhead.fmh_count = head.fmh_count;
653 ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
654 ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
656 trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
657 trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
661 error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
662 if (error == EXT4_QUERY_RANGE_ABORT)
667 /* If we didn't abort, set the "last" flag in the last fmx */
668 if (!aborted && info.gi_idx) {
669 info.gi_last_flags |= FMR_OF_LAST;
670 if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
672 sizeof(info.gi_last_flags)))
676 /* copy back header */
677 head.fmh_entries = xhead.fmh_entries;
678 head.fmh_oflags = xhead.fmh_oflags;
679 if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
685 static long ext4_ioctl_group_add(struct file *file,
686 struct ext4_new_group_data *input)
688 struct super_block *sb = file_inode(file)->i_sb;
691 err = ext4_resize_begin(sb);
695 if (ext4_has_feature_bigalloc(sb)) {
696 ext4_msg(sb, KERN_ERR,
697 "Online resizing not supported with bigalloc");
702 err = mnt_want_write_file(file);
706 err = ext4_group_add(sb, input);
707 if (EXT4_SB(sb)->s_journal) {
708 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
709 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
710 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
714 mnt_drop_write_file(file);
715 if (!err && ext4_has_group_desc_csum(sb) &&
716 test_opt(sb, INIT_INODE_TABLE))
717 err = ext4_register_li_request(sb, input->group);
723 int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa)
725 struct inode *inode = d_inode(dentry);
726 struct ext4_inode_info *ei = EXT4_I(inode);
727 u32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
729 if (S_ISREG(inode->i_mode))
730 flags &= ~FS_PROJINHERIT_FL;
732 fileattr_fill_flags(fa, flags);
733 if (ext4_has_feature_project(inode->i_sb))
734 fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
739 int ext4_fileattr_set(struct user_namespace *mnt_userns,
740 struct dentry *dentry, struct fileattr *fa)
742 struct inode *inode = d_inode(dentry);
743 u32 flags = fa->flags;
744 int err = -EOPNOTSUPP;
746 ext4_fc_start_update(inode);
747 if (flags & ~EXT4_FL_USER_VISIBLE)
751 * chattr(1) grabs flags via GETFLAGS, modifies the result and
752 * passes that to SETFLAGS. So we cannot easily make SETFLAGS
753 * more restrictive than just silently masking off visible but
754 * not settable flags as we always did.
756 flags &= EXT4_FL_USER_MODIFIABLE;
757 if (ext4_mask_flags(inode->i_mode, flags) != flags)
759 err = ext4_ioctl_check_immutable(inode, fa->fsx_projid, flags);
762 err = ext4_ioctl_setflags(inode, flags);
765 err = ext4_ioctl_setproject(inode, fa->fsx_projid);
767 ext4_fc_stop_update(inode);
771 /* So that the fiemap access checks can't overflow on 32 bit machines. */
772 #define FIEMAP_MAX_EXTENTS (UINT_MAX / sizeof(struct fiemap_extent))
774 static int ext4_ioctl_get_es_cache(struct file *filp, unsigned long arg)
776 struct fiemap fiemap;
777 struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
778 struct fiemap_extent_info fieinfo = { 0, };
779 struct inode *inode = file_inode(filp);
782 if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
785 if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
788 fieinfo.fi_flags = fiemap.fm_flags;
789 fieinfo.fi_extents_max = fiemap.fm_extent_count;
790 fieinfo.fi_extents_start = ufiemap->fm_extents;
792 error = ext4_get_es_cache(inode, &fieinfo, fiemap.fm_start,
794 fiemap.fm_flags = fieinfo.fi_flags;
795 fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
796 if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
802 static int ext4_ioctl_checkpoint(struct file *filp, unsigned long arg)
806 unsigned int flush_flags = 0;
807 struct super_block *sb = file_inode(filp)->i_sb;
808 struct request_queue *q;
810 if (copy_from_user(&flags, (__u32 __user *)arg,
814 if (!capable(CAP_SYS_ADMIN))
817 /* check for invalid bits set */
818 if ((flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID) ||
819 ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
820 (flags & JBD2_JOURNAL_FLUSH_ZEROOUT)))
823 if (!EXT4_SB(sb)->s_journal)
826 if (flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID)
829 q = bdev_get_queue(EXT4_SB(sb)->s_journal->j_dev);
832 if ((flags & JBD2_JOURNAL_FLUSH_DISCARD) && !blk_queue_discard(q))
835 if (flags & EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
838 if (flags & EXT4_IOC_CHECKPOINT_FLAG_DISCARD)
839 flush_flags |= JBD2_JOURNAL_FLUSH_DISCARD;
841 if (flags & EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT) {
842 flush_flags |= JBD2_JOURNAL_FLUSH_ZEROOUT;
843 pr_info_ratelimited("warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow");
846 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
847 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, flush_flags);
848 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
853 static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
855 struct inode *inode = file_inode(filp);
856 struct super_block *sb = inode->i_sb;
857 struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
859 ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
862 case FS_IOC_GETFSMAP:
863 return ext4_ioc_getfsmap(sb, (void __user *)arg);
864 case EXT4_IOC_GETVERSION:
865 case EXT4_IOC_GETVERSION_OLD:
866 return put_user(inode->i_generation, (int __user *) arg);
867 case EXT4_IOC_SETVERSION:
868 case EXT4_IOC_SETVERSION_OLD: {
870 struct ext4_iloc iloc;
874 if (!inode_owner_or_capable(mnt_userns, inode))
877 if (ext4_has_metadata_csum(inode->i_sb)) {
878 ext4_warning(sb, "Setting inode version is not "
879 "supported with metadata_csum enabled.");
883 err = mnt_want_write_file(filp);
886 if (get_user(generation, (int __user *) arg)) {
892 handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
893 if (IS_ERR(handle)) {
894 err = PTR_ERR(handle);
897 err = ext4_reserve_inode_write(handle, inode, &iloc);
899 inode->i_ctime = current_time(inode);
900 inode->i_generation = generation;
901 err = ext4_mark_iloc_dirty(handle, inode, &iloc);
903 ext4_journal_stop(handle);
908 mnt_drop_write_file(filp);
911 case EXT4_IOC_GROUP_EXTEND: {
912 ext4_fsblk_t n_blocks_count;
915 err = ext4_resize_begin(sb);
919 if (get_user(n_blocks_count, (__u32 __user *)arg)) {
921 goto group_extend_out;
924 if (ext4_has_feature_bigalloc(sb)) {
925 ext4_msg(sb, KERN_ERR,
926 "Online resizing not supported with bigalloc");
928 goto group_extend_out;
931 err = mnt_want_write_file(filp);
933 goto group_extend_out;
935 err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
936 if (EXT4_SB(sb)->s_journal) {
937 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
938 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
939 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
943 mnt_drop_write_file(filp);
949 case EXT4_IOC_MOVE_EXT: {
950 struct move_extent me;
954 if (!(filp->f_mode & FMODE_READ) ||
955 !(filp->f_mode & FMODE_WRITE))
958 if (copy_from_user(&me,
959 (struct move_extent __user *)arg, sizeof(me)))
963 donor = fdget(me.donor_fd);
967 if (!(donor.file->f_mode & FMODE_WRITE)) {
972 if (ext4_has_feature_bigalloc(sb)) {
973 ext4_msg(sb, KERN_ERR,
974 "Online defrag not supported with bigalloc");
977 } else if (IS_DAX(inode)) {
978 ext4_msg(sb, KERN_ERR,
979 "Online defrag not supported with DAX");
984 err = mnt_want_write_file(filp);
988 err = ext4_move_extents(filp, donor.file, me.orig_start,
989 me.donor_start, me.len, &me.moved_len);
990 mnt_drop_write_file(filp);
992 if (copy_to_user((struct move_extent __user *)arg,
1000 case EXT4_IOC_GROUP_ADD: {
1001 struct ext4_new_group_data input;
1003 if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
1007 return ext4_ioctl_group_add(filp, &input);
1010 case EXT4_IOC_MIGRATE:
1013 if (!inode_owner_or_capable(mnt_userns, inode))
1016 err = mnt_want_write_file(filp);
1020 * inode_mutex prevent write and truncate on the file.
1021 * Read still goes through. We take i_data_sem in
1022 * ext4_ext_swap_inode_data before we switch the
1023 * inode format to prevent read.
1025 inode_lock((inode));
1026 err = ext4_ext_migrate(inode);
1027 inode_unlock((inode));
1028 mnt_drop_write_file(filp);
1032 case EXT4_IOC_ALLOC_DA_BLKS:
1035 if (!inode_owner_or_capable(mnt_userns, inode))
1038 err = mnt_want_write_file(filp);
1041 err = ext4_alloc_da_blocks(inode);
1042 mnt_drop_write_file(filp);
1046 case EXT4_IOC_SWAP_BOOT:
1049 if (!(filp->f_mode & FMODE_WRITE))
1051 err = mnt_want_write_file(filp);
1054 err = swap_inode_boot_loader(sb, mnt_userns, inode);
1055 mnt_drop_write_file(filp);
1059 case EXT4_IOC_RESIZE_FS: {
1060 ext4_fsblk_t n_blocks_count;
1061 int err = 0, err2 = 0;
1062 ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
1064 if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1069 err = ext4_resize_begin(sb);
1073 err = mnt_want_write_file(filp);
1077 err = ext4_resize_fs(sb, n_blocks_count);
1078 if (EXT4_SB(sb)->s_journal) {
1079 ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE);
1080 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1081 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1082 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1086 mnt_drop_write_file(filp);
1087 if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1088 ext4_has_group_desc_csum(sb) &&
1089 test_opt(sb, INIT_INODE_TABLE))
1090 err = ext4_register_li_request(sb, o_group);
1093 ext4_resize_end(sb);
1099 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1100 struct fstrim_range range;
1103 if (!capable(CAP_SYS_ADMIN))
1106 if (!blk_queue_discard(q))
1110 * We haven't replayed the journal, so we cannot use our
1111 * block-bitmap-guided storage zapping commands.
1113 if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1116 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1120 range.minlen = max((unsigned int)range.minlen,
1121 q->limits.discard_granularity);
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)
1278 ext4_fc_start_update(file_inode(filp));
1279 ret = __ext4_ioctl(filp, cmd, arg);
1280 ext4_fc_stop_update(file_inode(filp));
1285 #ifdef CONFIG_COMPAT
1286 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1288 /* These are just misnamed, they actually get/put from/to user an int */
1290 case EXT4_IOC32_GETVERSION:
1291 cmd = EXT4_IOC_GETVERSION;
1293 case EXT4_IOC32_SETVERSION:
1294 cmd = EXT4_IOC_SETVERSION;
1296 case EXT4_IOC32_GROUP_EXTEND:
1297 cmd = EXT4_IOC_GROUP_EXTEND;
1299 case EXT4_IOC32_GETVERSION_OLD:
1300 cmd = EXT4_IOC_GETVERSION_OLD;
1302 case EXT4_IOC32_SETVERSION_OLD:
1303 cmd = EXT4_IOC_SETVERSION_OLD;
1305 case EXT4_IOC32_GETRSVSZ:
1306 cmd = EXT4_IOC_GETRSVSZ;
1308 case EXT4_IOC32_SETRSVSZ:
1309 cmd = EXT4_IOC_SETRSVSZ;
1311 case EXT4_IOC32_GROUP_ADD: {
1312 struct compat_ext4_new_group_input __user *uinput;
1313 struct ext4_new_group_data input;
1316 uinput = compat_ptr(arg);
1317 err = get_user(input.group, &uinput->group);
1318 err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1319 err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1320 err |= get_user(input.inode_table, &uinput->inode_table);
1321 err |= get_user(input.blocks_count, &uinput->blocks_count);
1322 err |= get_user(input.reserved_blocks,
1323 &uinput->reserved_blocks);
1326 return ext4_ioctl_group_add(file, &input);
1328 case EXT4_IOC_MOVE_EXT:
1329 case EXT4_IOC_RESIZE_FS:
1331 case EXT4_IOC_PRECACHE_EXTENTS:
1332 case FS_IOC_SET_ENCRYPTION_POLICY:
1333 case FS_IOC_GET_ENCRYPTION_PWSALT:
1334 case FS_IOC_GET_ENCRYPTION_POLICY:
1335 case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1336 case FS_IOC_ADD_ENCRYPTION_KEY:
1337 case FS_IOC_REMOVE_ENCRYPTION_KEY:
1338 case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1339 case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1340 case FS_IOC_GET_ENCRYPTION_NONCE:
1341 case EXT4_IOC_SHUTDOWN:
1342 case FS_IOC_GETFSMAP:
1343 case FS_IOC_ENABLE_VERITY:
1344 case FS_IOC_MEASURE_VERITY:
1345 case FS_IOC_READ_VERITY_METADATA:
1346 case EXT4_IOC_CLEAR_ES_CACHE:
1347 case EXT4_IOC_GETSTATE:
1348 case EXT4_IOC_GET_ES_CACHE:
1349 case EXT4_IOC_CHECKPOINT:
1352 return -ENOIOCTLCMD;
1354 return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));