GNU Linux-libre 4.19.211-gnu1
[releases.git] / fs / ext4 / ioctl.c
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/uuid.h>
20 #include <linux/uaccess.h>
21 #include <linux/delay.h>
22 #include <linux/iversion.h>
23 #include "ext4_jbd2.h"
24 #include "ext4.h"
25 #include <linux/fsmap.h>
26 #include "fsmap.h"
27 #include <trace/events/ext4.h>
28
29 /**
30  * Swap memory between @a and @b for @len bytes.
31  *
32  * @a:          pointer to first memory area
33  * @b:          pointer to second memory area
34  * @len:        number of bytes to swap
35  *
36  */
37 static void memswap(void *a, void *b, size_t len)
38 {
39         unsigned char *ap, *bp;
40
41         ap = (unsigned char *)a;
42         bp = (unsigned char *)b;
43         while (len-- > 0) {
44                 swap(*ap, *bp);
45                 ap++;
46                 bp++;
47         }
48 }
49
50 /**
51  * Swap i_data and associated attributes between @inode1 and @inode2.
52  * This function is used for the primary swap between inode1 and inode2
53  * and also to revert this primary swap in case of errors.
54  *
55  * Therefore you have to make sure, that calling this method twice
56  * will revert all changes.
57  *
58  * @inode1:     pointer to first inode
59  * @inode2:     pointer to second inode
60  */
61 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
62 {
63         loff_t isize;
64         struct ext4_inode_info *ei1;
65         struct ext4_inode_info *ei2;
66         unsigned long tmp;
67
68         ei1 = EXT4_I(inode1);
69         ei2 = EXT4_I(inode2);
70
71         swap(inode1->i_version, inode2->i_version);
72         swap(inode1->i_atime, inode2->i_atime);
73         swap(inode1->i_mtime, inode2->i_mtime);
74
75         memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
76         tmp = ei1->i_flags & EXT4_FL_SHOULD_SWAP;
77         ei1->i_flags = (ei2->i_flags & EXT4_FL_SHOULD_SWAP) |
78                 (ei1->i_flags & ~EXT4_FL_SHOULD_SWAP);
79         ei2->i_flags = tmp | (ei2->i_flags & ~EXT4_FL_SHOULD_SWAP);
80         swap(ei1->i_disksize, ei2->i_disksize);
81         ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
82         ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
83
84         isize = i_size_read(inode1);
85         i_size_write(inode1, i_size_read(inode2));
86         i_size_write(inode2, isize);
87 }
88
89 static void reset_inode_seed(struct inode *inode)
90 {
91         struct ext4_inode_info *ei = EXT4_I(inode);
92         struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
93         __le32 inum = cpu_to_le32(inode->i_ino);
94         __le32 gen = cpu_to_le32(inode->i_generation);
95         __u32 csum;
96
97         if (!ext4_has_metadata_csum(inode->i_sb))
98                 return;
99
100         csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum, sizeof(inum));
101         ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen, sizeof(gen));
102 }
103
104 /**
105  * Swap the information from the given @inode and the inode
106  * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
107  * important fields of the inodes.
108  *
109  * @sb:         the super block of the filesystem
110  * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
111  *
112  */
113 static long swap_inode_boot_loader(struct super_block *sb,
114                                 struct inode *inode)
115 {
116         handle_t *handle;
117         int err;
118         struct inode *inode_bl;
119         struct ext4_inode_info *ei_bl;
120         qsize_t size, size_bl, diff;
121         blkcnt_t blocks;
122         unsigned short bytes;
123
124         inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO, EXT4_IGET_SPECIAL);
125         if (IS_ERR(inode_bl))
126                 return PTR_ERR(inode_bl);
127         ei_bl = EXT4_I(inode_bl);
128
129         /* Protect orig inodes against a truncate and make sure,
130          * that only 1 swap_inode_boot_loader is running. */
131         lock_two_nondirectories(inode, inode_bl);
132
133         if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode) ||
134             IS_SWAPFILE(inode) || IS_ENCRYPTED(inode) ||
135             (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL) ||
136             ext4_has_inline_data(inode)) {
137                 err = -EINVAL;
138                 goto journal_err_out;
139         }
140
141         if (IS_RDONLY(inode) || IS_APPEND(inode) || IS_IMMUTABLE(inode) ||
142             !inode_owner_or_capable(inode) || !capable(CAP_SYS_ADMIN)) {
143                 err = -EPERM;
144                 goto journal_err_out;
145         }
146
147         down_write(&EXT4_I(inode)->i_mmap_sem);
148         err = filemap_write_and_wait(inode->i_mapping);
149         if (err)
150                 goto err_out;
151
152         err = filemap_write_and_wait(inode_bl->i_mapping);
153         if (err)
154                 goto err_out;
155
156         /* Wait for all existing dio workers */
157         inode_dio_wait(inode);
158         inode_dio_wait(inode_bl);
159
160         truncate_inode_pages(&inode->i_data, 0);
161         truncate_inode_pages(&inode_bl->i_data, 0);
162
163         handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
164         if (IS_ERR(handle)) {
165                 err = -EINVAL;
166                 goto err_out;
167         }
168
169         /* Protect extent tree against block allocations via delalloc */
170         ext4_double_down_write_data_sem(inode, inode_bl);
171
172         if (inode_bl->i_nlink == 0) {
173                 /* this inode has never been used as a BOOT_LOADER */
174                 set_nlink(inode_bl, 1);
175                 i_uid_write(inode_bl, 0);
176                 i_gid_write(inode_bl, 0);
177                 inode_bl->i_flags = 0;
178                 ei_bl->i_flags = 0;
179                 inode_set_iversion(inode_bl, 1);
180                 i_size_write(inode_bl, 0);
181                 inode_bl->i_mode = S_IFREG;
182                 if (ext4_has_feature_extents(sb)) {
183                         ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
184                         ext4_ext_tree_init(handle, inode_bl);
185                 } else
186                         memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
187         }
188
189         err = dquot_initialize(inode);
190         if (err)
191                 goto err_out1;
192
193         size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
194         size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
195         diff = size - size_bl;
196         swap_inode_data(inode, inode_bl);
197
198         inode->i_ctime = inode_bl->i_ctime = current_time(inode);
199
200         inode->i_generation = prandom_u32();
201         inode_bl->i_generation = prandom_u32();
202         reset_inode_seed(inode);
203         reset_inode_seed(inode_bl);
204
205         ext4_discard_preallocations(inode);
206
207         err = ext4_mark_inode_dirty(handle, inode);
208         if (err < 0) {
209                 /* No need to update quota information. */
210                 ext4_warning(inode->i_sb,
211                         "couldn't mark inode #%lu dirty (err %d)",
212                         inode->i_ino, err);
213                 /* Revert all changes: */
214                 swap_inode_data(inode, inode_bl);
215                 ext4_mark_inode_dirty(handle, inode);
216                 goto err_out1;
217         }
218
219         blocks = inode_bl->i_blocks;
220         bytes = inode_bl->i_bytes;
221         inode_bl->i_blocks = inode->i_blocks;
222         inode_bl->i_bytes = inode->i_bytes;
223         err = ext4_mark_inode_dirty(handle, inode_bl);
224         if (err < 0) {
225                 /* No need to update quota information. */
226                 ext4_warning(inode_bl->i_sb,
227                         "couldn't mark inode #%lu dirty (err %d)",
228                         inode_bl->i_ino, err);
229                 goto revert;
230         }
231
232         /* Bootloader inode should not be counted into quota information. */
233         if (diff > 0)
234                 dquot_free_space(inode, diff);
235         else
236                 err = dquot_alloc_space(inode, -1 * diff);
237
238         if (err < 0) {
239 revert:
240                 /* Revert all changes: */
241                 inode_bl->i_blocks = blocks;
242                 inode_bl->i_bytes = bytes;
243                 swap_inode_data(inode, inode_bl);
244                 ext4_mark_inode_dirty(handle, inode);
245                 ext4_mark_inode_dirty(handle, inode_bl);
246         }
247
248 err_out1:
249         ext4_journal_stop(handle);
250         ext4_double_up_write_data_sem(inode, inode_bl);
251
252 err_out:
253         up_write(&EXT4_I(inode)->i_mmap_sem);
254 journal_err_out:
255         unlock_two_nondirectories(inode, inode_bl);
256         iput(inode_bl);
257         return err;
258 }
259
260 #ifdef CONFIG_EXT4_FS_ENCRYPTION
261 static int uuid_is_zero(__u8 u[16])
262 {
263         int     i;
264
265         for (i = 0; i < 16; i++)
266                 if (u[i])
267                         return 0;
268         return 1;
269 }
270 #endif
271
272 /*
273  * If immutable is set and we are not clearing it, we're not allowed to change
274  * anything else in the inode.  Don't error out if we're only trying to set
275  * immutable on an immutable file.
276  */
277 static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
278                                       unsigned int flags)
279 {
280         struct ext4_inode_info *ei = EXT4_I(inode);
281         unsigned int oldflags = ei->i_flags;
282
283         if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
284                 return 0;
285
286         if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
287                 return -EPERM;
288         if (ext4_has_feature_project(inode->i_sb) &&
289             __kprojid_val(ei->i_projid) != new_projid)
290                 return -EPERM;
291
292         return 0;
293 }
294
295 static int ext4_ioctl_setflags(struct inode *inode,
296                                unsigned int flags)
297 {
298         struct ext4_inode_info *ei = EXT4_I(inode);
299         handle_t *handle = NULL;
300         int err = -EPERM, migrate = 0;
301         struct ext4_iloc iloc;
302         unsigned int oldflags, mask, i;
303         unsigned int jflag;
304
305         /* Is it quota file? Do not allow user to mess with it */
306         if (ext4_is_quota_file(inode))
307                 goto flags_out;
308
309         oldflags = ei->i_flags;
310
311         /* The JOURNAL_DATA flag is modifiable only by root */
312         jflag = flags & EXT4_JOURNAL_DATA_FL;
313
314         /*
315          * The IMMUTABLE and APPEND_ONLY flags can only be changed by
316          * the relevant capability.
317          *
318          * This test looks nicer. Thanks to Pauline Middelink
319          */
320         if ((flags ^ oldflags) & (EXT4_APPEND_FL | EXT4_IMMUTABLE_FL)) {
321                 if (!capable(CAP_LINUX_IMMUTABLE))
322                         goto flags_out;
323         }
324
325         /*
326          * The JOURNAL_DATA flag can only be changed by
327          * the relevant capability.
328          */
329         if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
330                 if (!capable(CAP_SYS_RESOURCE))
331                         goto flags_out;
332         }
333         if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
334                 migrate = 1;
335
336         if (flags & EXT4_EOFBLOCKS_FL) {
337                 /* we don't support adding EOFBLOCKS flag */
338                 if (!(oldflags & EXT4_EOFBLOCKS_FL)) {
339                         err = -EOPNOTSUPP;
340                         goto flags_out;
341                 }
342         } else if (oldflags & EXT4_EOFBLOCKS_FL) {
343                 err = ext4_truncate(inode);
344                 if (err)
345                         goto flags_out;
346         }
347
348         /*
349          * Wait for all pending directio and then flush all the dirty pages
350          * for this file.  The flush marks all the pages readonly, so any
351          * subsequent attempt to write to the file (particularly mmap pages)
352          * will come through the filesystem and fail.
353          */
354         if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
355             (flags & EXT4_IMMUTABLE_FL)) {
356                 inode_dio_wait(inode);
357                 err = filemap_write_and_wait(inode->i_mapping);
358                 if (err)
359                         goto flags_out;
360         }
361
362         handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
363         if (IS_ERR(handle)) {
364                 err = PTR_ERR(handle);
365                 goto flags_out;
366         }
367         if (IS_SYNC(inode))
368                 ext4_handle_sync(handle);
369         err = ext4_reserve_inode_write(handle, inode, &iloc);
370         if (err)
371                 goto flags_err;
372
373         for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
374                 if (!(mask & EXT4_FL_USER_MODIFIABLE))
375                         continue;
376                 /* These flags get special treatment later */
377                 if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
378                         continue;
379                 if (mask & flags)
380                         ext4_set_inode_flag(inode, i);
381                 else
382                         ext4_clear_inode_flag(inode, i);
383         }
384
385         ext4_set_inode_flags(inode);
386         inode->i_ctime = current_time(inode);
387
388         err = ext4_mark_iloc_dirty(handle, inode, &iloc);
389 flags_err:
390         ext4_journal_stop(handle);
391         if (err)
392                 goto flags_out;
393
394         if ((jflag ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
395                 /*
396                  * Changes to the journaling mode can cause unsafe changes to
397                  * S_DAX if we are using the DAX mount option.
398                  */
399                 if (test_opt(inode->i_sb, DAX)) {
400                         err = -EBUSY;
401                         goto flags_out;
402                 }
403
404                 err = ext4_change_inode_journal_flag(inode, jflag);
405                 if (err)
406                         goto flags_out;
407         }
408         if (migrate) {
409                 if (flags & EXT4_EXTENTS_FL)
410                         err = ext4_ext_migrate(inode);
411                 else
412                         err = ext4_ind_migrate(inode);
413         }
414
415 flags_out:
416         return err;
417 }
418
419 #ifdef CONFIG_QUOTA
420 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
421 {
422         struct inode *inode = file_inode(filp);
423         struct super_block *sb = inode->i_sb;
424         struct ext4_inode_info *ei = EXT4_I(inode);
425         int err, rc;
426         handle_t *handle;
427         kprojid_t kprojid;
428         struct ext4_iloc iloc;
429         struct ext4_inode *raw_inode;
430         struct dquot *transfer_to[MAXQUOTAS] = { };
431
432         if (!ext4_has_feature_project(sb)) {
433                 if (projid != EXT4_DEF_PROJID)
434                         return -EOPNOTSUPP;
435                 else
436                         return 0;
437         }
438
439         if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
440                 return -EOPNOTSUPP;
441
442         kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
443
444         if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
445                 return 0;
446
447         err = -EPERM;
448         /* Is it quota file? Do not allow user to mess with it */
449         if (ext4_is_quota_file(inode))
450                 return err;
451
452         err = ext4_get_inode_loc(inode, &iloc);
453         if (err)
454                 return err;
455
456         raw_inode = ext4_raw_inode(&iloc);
457         if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
458                 err = ext4_expand_extra_isize(inode,
459                                               EXT4_SB(sb)->s_want_extra_isize,
460                                               &iloc);
461                 if (err)
462                         return err;
463         } else {
464                 brelse(iloc.bh);
465         }
466
467         err = dquot_initialize(inode);
468         if (err)
469                 return err;
470
471         handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
472                 EXT4_QUOTA_INIT_BLOCKS(sb) +
473                 EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
474         if (IS_ERR(handle))
475                 return PTR_ERR(handle);
476
477         err = ext4_reserve_inode_write(handle, inode, &iloc);
478         if (err)
479                 goto out_stop;
480
481         transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
482         if (!IS_ERR(transfer_to[PRJQUOTA])) {
483
484                 /* __dquot_transfer() calls back ext4_get_inode_usage() which
485                  * counts xattr inode references.
486                  */
487                 down_read(&EXT4_I(inode)->xattr_sem);
488                 err = __dquot_transfer(inode, transfer_to);
489                 up_read(&EXT4_I(inode)->xattr_sem);
490                 dqput(transfer_to[PRJQUOTA]);
491                 if (err)
492                         goto out_dirty;
493         }
494
495         EXT4_I(inode)->i_projid = kprojid;
496         inode->i_ctime = current_time(inode);
497 out_dirty:
498         rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
499         if (!err)
500                 err = rc;
501 out_stop:
502         ext4_journal_stop(handle);
503         return err;
504 }
505 #else
506 static int ext4_ioctl_setproject(struct file *filp, __u32 projid)
507 {
508         if (projid != EXT4_DEF_PROJID)
509                 return -EOPNOTSUPP;
510         return 0;
511 }
512 #endif
513
514 /* Transfer internal flags to xflags */
515 static inline __u32 ext4_iflags_to_xflags(unsigned long iflags)
516 {
517         __u32 xflags = 0;
518
519         if (iflags & EXT4_SYNC_FL)
520                 xflags |= FS_XFLAG_SYNC;
521         if (iflags & EXT4_IMMUTABLE_FL)
522                 xflags |= FS_XFLAG_IMMUTABLE;
523         if (iflags & EXT4_APPEND_FL)
524                 xflags |= FS_XFLAG_APPEND;
525         if (iflags & EXT4_NODUMP_FL)
526                 xflags |= FS_XFLAG_NODUMP;
527         if (iflags & EXT4_NOATIME_FL)
528                 xflags |= FS_XFLAG_NOATIME;
529         if (iflags & EXT4_PROJINHERIT_FL)
530                 xflags |= FS_XFLAG_PROJINHERIT;
531         return xflags;
532 }
533
534 #define EXT4_SUPPORTED_FS_XFLAGS (FS_XFLAG_SYNC | FS_XFLAG_IMMUTABLE | \
535                                   FS_XFLAG_APPEND | FS_XFLAG_NODUMP | \
536                                   FS_XFLAG_NOATIME | FS_XFLAG_PROJINHERIT)
537
538 /* Transfer xflags flags to internal */
539 static inline unsigned long ext4_xflags_to_iflags(__u32 xflags)
540 {
541         unsigned long iflags = 0;
542
543         if (xflags & FS_XFLAG_SYNC)
544                 iflags |= EXT4_SYNC_FL;
545         if (xflags & FS_XFLAG_IMMUTABLE)
546                 iflags |= EXT4_IMMUTABLE_FL;
547         if (xflags & FS_XFLAG_APPEND)
548                 iflags |= EXT4_APPEND_FL;
549         if (xflags & FS_XFLAG_NODUMP)
550                 iflags |= EXT4_NODUMP_FL;
551         if (xflags & FS_XFLAG_NOATIME)
552                 iflags |= EXT4_NOATIME_FL;
553         if (xflags & FS_XFLAG_PROJINHERIT)
554                 iflags |= EXT4_PROJINHERIT_FL;
555
556         return iflags;
557 }
558
559 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
560 {
561         struct ext4_sb_info *sbi = EXT4_SB(sb);
562         __u32 flags;
563
564         if (!capable(CAP_SYS_ADMIN))
565                 return -EPERM;
566
567         if (get_user(flags, (__u32 __user *)arg))
568                 return -EFAULT;
569
570         if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
571                 return -EINVAL;
572
573         if (ext4_forced_shutdown(sbi))
574                 return 0;
575
576         ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
577         trace_ext4_shutdown(sb, flags);
578
579         switch (flags) {
580         case EXT4_GOING_FLAGS_DEFAULT:
581                 freeze_bdev(sb->s_bdev);
582                 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
583                 thaw_bdev(sb->s_bdev, sb);
584                 break;
585         case EXT4_GOING_FLAGS_LOGFLUSH:
586                 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
587                 if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
588                         (void) ext4_force_commit(sb);
589                         jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
590                 }
591                 break;
592         case EXT4_GOING_FLAGS_NOLOGFLUSH:
593                 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
594                 if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
595                         jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
596                 break;
597         default:
598                 return -EINVAL;
599         }
600         clear_opt(sb, DISCARD);
601         return 0;
602 }
603
604 struct getfsmap_info {
605         struct super_block      *gi_sb;
606         struct fsmap_head __user *gi_data;
607         unsigned int            gi_idx;
608         __u32                   gi_last_flags;
609 };
610
611 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
612 {
613         struct getfsmap_info *info = priv;
614         struct fsmap fm;
615
616         trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
617
618         info->gi_last_flags = xfm->fmr_flags;
619         ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
620         if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
621                         sizeof(struct fsmap)))
622                 return -EFAULT;
623
624         return 0;
625 }
626
627 static int ext4_ioc_getfsmap(struct super_block *sb,
628                              struct fsmap_head __user *arg)
629 {
630         struct getfsmap_info info = {0};
631         struct ext4_fsmap_head xhead = {0};
632         struct fsmap_head head;
633         bool aborted = false;
634         int error;
635
636         if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
637                 return -EFAULT;
638         if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
639             memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
640                        sizeof(head.fmh_keys[0].fmr_reserved)) ||
641             memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
642                        sizeof(head.fmh_keys[1].fmr_reserved)))
643                 return -EINVAL;
644         /*
645          * ext4 doesn't report file extents at all, so the only valid
646          * file offsets are the magic ones (all zeroes or all ones).
647          */
648         if (head.fmh_keys[0].fmr_offset ||
649             (head.fmh_keys[1].fmr_offset != 0 &&
650              head.fmh_keys[1].fmr_offset != -1ULL))
651                 return -EINVAL;
652
653         xhead.fmh_iflags = head.fmh_iflags;
654         xhead.fmh_count = head.fmh_count;
655         ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
656         ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
657
658         trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
659         trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
660
661         info.gi_sb = sb;
662         info.gi_data = arg;
663         error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
664         if (error == EXT4_QUERY_RANGE_ABORT) {
665                 error = 0;
666                 aborted = true;
667         } else if (error)
668                 return error;
669
670         /* If we didn't abort, set the "last" flag in the last fmx */
671         if (!aborted && info.gi_idx) {
672                 info.gi_last_flags |= FMR_OF_LAST;
673                 if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
674                                  &info.gi_last_flags,
675                                  sizeof(info.gi_last_flags)))
676                         return -EFAULT;
677         }
678
679         /* copy back header */
680         head.fmh_entries = xhead.fmh_entries;
681         head.fmh_oflags = xhead.fmh_oflags;
682         if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
683                 return -EFAULT;
684
685         return 0;
686 }
687
688 static long ext4_ioctl_group_add(struct file *file,
689                                  struct ext4_new_group_data *input)
690 {
691         struct super_block *sb = file_inode(file)->i_sb;
692         int err, err2=0;
693
694         err = ext4_resize_begin(sb);
695         if (err)
696                 return err;
697
698         if (ext4_has_feature_bigalloc(sb)) {
699                 ext4_msg(sb, KERN_ERR,
700                          "Online resizing not supported with bigalloc");
701                 err = -EOPNOTSUPP;
702                 goto group_add_out;
703         }
704
705         err = mnt_want_write_file(file);
706         if (err)
707                 goto group_add_out;
708
709         err = ext4_group_add(sb, input);
710         if (EXT4_SB(sb)->s_journal) {
711                 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
712                 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
713                 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
714         }
715         if (err == 0)
716                 err = err2;
717         mnt_drop_write_file(file);
718         if (!err && ext4_has_group_desc_csum(sb) &&
719             test_opt(sb, INIT_INODE_TABLE))
720                 err = ext4_register_li_request(sb, input->group);
721 group_add_out:
722         ext4_resize_end(sb);
723         return err;
724 }
725
726 static int ext4_ioctl_check_project(struct inode *inode, struct fsxattr *fa)
727 {
728         /*
729          * Project Quota ID state is only allowed to change from within the init
730          * namespace. Enforce that restriction only if we are trying to change
731          * the quota ID state. Everything else is allowed in user namespaces.
732          */
733         if (current_user_ns() == &init_user_ns)
734                 return 0;
735
736         if (__kprojid_val(EXT4_I(inode)->i_projid) != fa->fsx_projid)
737                 return -EINVAL;
738
739         if (ext4_test_inode_flag(inode, EXT4_INODE_PROJINHERIT)) {
740                 if (!(fa->fsx_xflags & FS_XFLAG_PROJINHERIT))
741                         return -EINVAL;
742         } else {
743                 if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT)
744                         return -EINVAL;
745         }
746
747         return 0;
748 }
749
750 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
751 {
752         struct inode *inode = file_inode(filp);
753         struct super_block *sb = inode->i_sb;
754         struct ext4_inode_info *ei = EXT4_I(inode);
755         unsigned int flags;
756
757         ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
758
759         switch (cmd) {
760         case FS_IOC_GETFSMAP:
761                 return ext4_ioc_getfsmap(sb, (void __user *)arg);
762         case EXT4_IOC_GETFLAGS:
763                 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
764                 return put_user(flags, (int __user *) arg);
765         case EXT4_IOC_SETFLAGS: {
766                 int err;
767
768                 if (!inode_owner_or_capable(inode))
769                         return -EACCES;
770
771                 if (get_user(flags, (int __user *) arg))
772                         return -EFAULT;
773
774                 if (flags & ~EXT4_FL_USER_VISIBLE)
775                         return -EOPNOTSUPP;
776                 /*
777                  * chattr(1) grabs flags via GETFLAGS, modifies the result and
778                  * passes that to SETFLAGS. So we cannot easily make SETFLAGS
779                  * more restrictive than just silently masking off visible but
780                  * not settable flags as we always did.
781                  */
782                 flags &= EXT4_FL_USER_MODIFIABLE;
783                 if (ext4_mask_flags(inode->i_mode, flags) != flags)
784                         return -EOPNOTSUPP;
785
786                 err = mnt_want_write_file(filp);
787                 if (err)
788                         return err;
789
790                 inode_lock(inode);
791                 err = ext4_ioctl_check_immutable(inode,
792                                 from_kprojid(&init_user_ns, ei->i_projid),
793                                 flags);
794                 if (!err)
795                         err = ext4_ioctl_setflags(inode, flags);
796                 inode_unlock(inode);
797                 mnt_drop_write_file(filp);
798                 return err;
799         }
800         case EXT4_IOC_GETVERSION:
801         case EXT4_IOC_GETVERSION_OLD:
802                 return put_user(inode->i_generation, (int __user *) arg);
803         case EXT4_IOC_SETVERSION:
804         case EXT4_IOC_SETVERSION_OLD: {
805                 handle_t *handle;
806                 struct ext4_iloc iloc;
807                 __u32 generation;
808                 int err;
809
810                 if (!inode_owner_or_capable(inode))
811                         return -EPERM;
812
813                 if (ext4_has_metadata_csum(inode->i_sb)) {
814                         ext4_warning(sb, "Setting inode version is not "
815                                      "supported with metadata_csum enabled.");
816                         return -ENOTTY;
817                 }
818
819                 err = mnt_want_write_file(filp);
820                 if (err)
821                         return err;
822                 if (get_user(generation, (int __user *) arg)) {
823                         err = -EFAULT;
824                         goto setversion_out;
825                 }
826
827                 inode_lock(inode);
828                 handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
829                 if (IS_ERR(handle)) {
830                         err = PTR_ERR(handle);
831                         goto unlock_out;
832                 }
833                 err = ext4_reserve_inode_write(handle, inode, &iloc);
834                 if (err == 0) {
835                         inode->i_ctime = current_time(inode);
836                         inode->i_generation = generation;
837                         err = ext4_mark_iloc_dirty(handle, inode, &iloc);
838                 }
839                 ext4_journal_stop(handle);
840
841 unlock_out:
842                 inode_unlock(inode);
843 setversion_out:
844                 mnt_drop_write_file(filp);
845                 return err;
846         }
847         case EXT4_IOC_GROUP_EXTEND: {
848                 ext4_fsblk_t n_blocks_count;
849                 int err, err2=0;
850
851                 err = ext4_resize_begin(sb);
852                 if (err)
853                         return err;
854
855                 if (get_user(n_blocks_count, (__u32 __user *)arg)) {
856                         err = -EFAULT;
857                         goto group_extend_out;
858                 }
859
860                 if (ext4_has_feature_bigalloc(sb)) {
861                         ext4_msg(sb, KERN_ERR,
862                                  "Online resizing not supported with bigalloc");
863                         err = -EOPNOTSUPP;
864                         goto group_extend_out;
865                 }
866
867                 err = mnt_want_write_file(filp);
868                 if (err)
869                         goto group_extend_out;
870
871                 err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
872                 if (EXT4_SB(sb)->s_journal) {
873                         jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
874                         err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
875                         jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
876                 }
877                 if (err == 0)
878                         err = err2;
879                 mnt_drop_write_file(filp);
880 group_extend_out:
881                 ext4_resize_end(sb);
882                 return err;
883         }
884
885         case EXT4_IOC_MOVE_EXT: {
886                 struct move_extent me;
887                 struct fd donor;
888                 int err;
889
890                 if (!(filp->f_mode & FMODE_READ) ||
891                     !(filp->f_mode & FMODE_WRITE))
892                         return -EBADF;
893
894                 if (copy_from_user(&me,
895                         (struct move_extent __user *)arg, sizeof(me)))
896                         return -EFAULT;
897                 me.moved_len = 0;
898
899                 donor = fdget(me.donor_fd);
900                 if (!donor.file)
901                         return -EBADF;
902
903                 if (!(donor.file->f_mode & FMODE_WRITE)) {
904                         err = -EBADF;
905                         goto mext_out;
906                 }
907
908                 if (ext4_has_feature_bigalloc(sb)) {
909                         ext4_msg(sb, KERN_ERR,
910                                  "Online defrag not supported with bigalloc");
911                         err = -EOPNOTSUPP;
912                         goto mext_out;
913                 } else if (IS_DAX(inode)) {
914                         ext4_msg(sb, KERN_ERR,
915                                  "Online defrag not supported with DAX");
916                         err = -EOPNOTSUPP;
917                         goto mext_out;
918                 }
919
920                 err = mnt_want_write_file(filp);
921                 if (err)
922                         goto mext_out;
923
924                 err = ext4_move_extents(filp, donor.file, me.orig_start,
925                                         me.donor_start, me.len, &me.moved_len);
926                 mnt_drop_write_file(filp);
927
928                 if (copy_to_user((struct move_extent __user *)arg,
929                                  &me, sizeof(me)))
930                         err = -EFAULT;
931 mext_out:
932                 fdput(donor);
933                 return err;
934         }
935
936         case EXT4_IOC_GROUP_ADD: {
937                 struct ext4_new_group_data input;
938
939                 if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
940                                 sizeof(input)))
941                         return -EFAULT;
942
943                 return ext4_ioctl_group_add(filp, &input);
944         }
945
946         case EXT4_IOC_MIGRATE:
947         {
948                 int err;
949                 if (!inode_owner_or_capable(inode))
950                         return -EACCES;
951
952                 err = mnt_want_write_file(filp);
953                 if (err)
954                         return err;
955                 /*
956                  * inode_mutex prevent write and truncate on the file.
957                  * Read still goes through. We take i_data_sem in
958                  * ext4_ext_swap_inode_data before we switch the
959                  * inode format to prevent read.
960                  */
961                 inode_lock((inode));
962                 err = ext4_ext_migrate(inode);
963                 inode_unlock((inode));
964                 mnt_drop_write_file(filp);
965                 return err;
966         }
967
968         case EXT4_IOC_ALLOC_DA_BLKS:
969         {
970                 int err;
971                 if (!inode_owner_or_capable(inode))
972                         return -EACCES;
973
974                 err = mnt_want_write_file(filp);
975                 if (err)
976                         return err;
977                 err = ext4_alloc_da_blocks(inode);
978                 mnt_drop_write_file(filp);
979                 return err;
980         }
981
982         case EXT4_IOC_SWAP_BOOT:
983         {
984                 int err;
985                 if (!(filp->f_mode & FMODE_WRITE))
986                         return -EBADF;
987                 err = mnt_want_write_file(filp);
988                 if (err)
989                         return err;
990                 err = swap_inode_boot_loader(sb, inode);
991                 mnt_drop_write_file(filp);
992                 return err;
993         }
994
995         case EXT4_IOC_RESIZE_FS: {
996                 ext4_fsblk_t n_blocks_count;
997                 int err = 0, err2 = 0;
998                 ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
999
1000                 if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1001                                    sizeof(__u64))) {
1002                         return -EFAULT;
1003                 }
1004
1005                 err = ext4_resize_begin(sb);
1006                 if (err)
1007                         return err;
1008
1009                 err = mnt_want_write_file(filp);
1010                 if (err)
1011                         goto resizefs_out;
1012
1013                 err = ext4_resize_fs(sb, n_blocks_count);
1014                 if (EXT4_SB(sb)->s_journal) {
1015                         jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1016                         err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
1017                         jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1018                 }
1019                 if (err == 0)
1020                         err = err2;
1021                 mnt_drop_write_file(filp);
1022                 if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1023                     ext4_has_group_desc_csum(sb) &&
1024                     test_opt(sb, INIT_INODE_TABLE))
1025                         err = ext4_register_li_request(sb, o_group);
1026
1027 resizefs_out:
1028                 ext4_resize_end(sb);
1029                 return err;
1030         }
1031
1032         case FITRIM:
1033         {
1034                 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1035                 struct fstrim_range range;
1036                 int ret = 0;
1037
1038                 if (!capable(CAP_SYS_ADMIN))
1039                         return -EPERM;
1040
1041                 if (!blk_queue_discard(q))
1042                         return -EOPNOTSUPP;
1043
1044                 /*
1045                  * We haven't replayed the journal, so we cannot use our
1046                  * block-bitmap-guided storage zapping commands.
1047                  */
1048                 if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1049                         return -EROFS;
1050
1051                 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1052                     sizeof(range)))
1053                         return -EFAULT;
1054
1055                 range.minlen = max((unsigned int)range.minlen,
1056                                    q->limits.discard_granularity);
1057                 ret = ext4_trim_fs(sb, &range);
1058                 if (ret < 0)
1059                         return ret;
1060
1061                 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1062                     sizeof(range)))
1063                         return -EFAULT;
1064
1065                 return 0;
1066         }
1067         case EXT4_IOC_PRECACHE_EXTENTS:
1068                 return ext4_ext_precache(inode);
1069
1070         case EXT4_IOC_SET_ENCRYPTION_POLICY:
1071                 if (!ext4_has_feature_encrypt(sb))
1072                         return -EOPNOTSUPP;
1073                 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1074
1075         case EXT4_IOC_GET_ENCRYPTION_PWSALT: {
1076 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1077                 int err, err2;
1078                 struct ext4_sb_info *sbi = EXT4_SB(sb);
1079                 handle_t *handle;
1080
1081                 if (!ext4_has_feature_encrypt(sb))
1082                         return -EOPNOTSUPP;
1083                 if (uuid_is_zero(sbi->s_es->s_encrypt_pw_salt)) {
1084                         err = mnt_want_write_file(filp);
1085                         if (err)
1086                                 return err;
1087                         handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 1);
1088                         if (IS_ERR(handle)) {
1089                                 err = PTR_ERR(handle);
1090                                 goto pwsalt_err_exit;
1091                         }
1092                         err = ext4_journal_get_write_access(handle, sbi->s_sbh);
1093                         if (err)
1094                                 goto pwsalt_err_journal;
1095                         lock_buffer(sbi->s_sbh);
1096                         generate_random_uuid(sbi->s_es->s_encrypt_pw_salt);
1097                         ext4_superblock_csum_set(sb);
1098                         unlock_buffer(sbi->s_sbh);
1099                         err = ext4_handle_dirty_metadata(handle, NULL,
1100                                                          sbi->s_sbh);
1101                 pwsalt_err_journal:
1102                         err2 = ext4_journal_stop(handle);
1103                         if (err2 && !err)
1104                                 err = err2;
1105                 pwsalt_err_exit:
1106                         mnt_drop_write_file(filp);
1107                         if (err)
1108                                 return err;
1109                 }
1110                 if (copy_to_user((void __user *) arg,
1111                                  sbi->s_es->s_encrypt_pw_salt, 16))
1112                         return -EFAULT;
1113                 return 0;
1114 #else
1115                 return -EOPNOTSUPP;
1116 #endif
1117         }
1118         case EXT4_IOC_GET_ENCRYPTION_POLICY:
1119                 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1120
1121         case EXT4_IOC_FSGETXATTR:
1122         {
1123                 struct fsxattr fa;
1124
1125                 memset(&fa, 0, sizeof(struct fsxattr));
1126                 fa.fsx_xflags = ext4_iflags_to_xflags(ei->i_flags & EXT4_FL_USER_VISIBLE);
1127
1128                 if (ext4_has_feature_project(inode->i_sb)) {
1129                         fa.fsx_projid = (__u32)from_kprojid(&init_user_ns,
1130                                 EXT4_I(inode)->i_projid);
1131                 }
1132
1133                 if (copy_to_user((struct fsxattr __user *)arg,
1134                                  &fa, sizeof(fa)))
1135                         return -EFAULT;
1136                 return 0;
1137         }
1138         case EXT4_IOC_FSSETXATTR:
1139         {
1140                 struct fsxattr fa;
1141                 int err;
1142
1143                 if (copy_from_user(&fa, (struct fsxattr __user *)arg,
1144                                    sizeof(fa)))
1145                         return -EFAULT;
1146
1147                 /* Make sure caller has proper permission */
1148                 if (!inode_owner_or_capable(inode))
1149                         return -EACCES;
1150
1151                 if (fa.fsx_xflags & ~EXT4_SUPPORTED_FS_XFLAGS)
1152                         return -EOPNOTSUPP;
1153
1154                 flags = ext4_xflags_to_iflags(fa.fsx_xflags);
1155                 if (ext4_mask_flags(inode->i_mode, flags) != flags)
1156                         return -EOPNOTSUPP;
1157
1158                 err = mnt_want_write_file(filp);
1159                 if (err)
1160                         return err;
1161
1162                 inode_lock(inode);
1163                 err = ext4_ioctl_check_project(inode, &fa);
1164                 if (err)
1165                         goto out;
1166                 flags = (ei->i_flags & ~EXT4_FL_XFLAG_VISIBLE) |
1167                          (flags & EXT4_FL_XFLAG_VISIBLE);
1168                 err = ext4_ioctl_check_immutable(inode, fa.fsx_projid, flags);
1169                 if (err)
1170                         goto out;
1171                 err = ext4_ioctl_setflags(inode, flags);
1172                 if (err)
1173                         goto out;
1174                 err = ext4_ioctl_setproject(filp, fa.fsx_projid);
1175 out:
1176                 inode_unlock(inode);
1177                 mnt_drop_write_file(filp);
1178                 return err;
1179         }
1180         case EXT4_IOC_SHUTDOWN:
1181                 return ext4_shutdown(sb, arg);
1182         default:
1183                 return -ENOTTY;
1184         }
1185 }
1186
1187 #ifdef CONFIG_COMPAT
1188 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1189 {
1190         /* These are just misnamed, they actually get/put from/to user an int */
1191         switch (cmd) {
1192         case EXT4_IOC32_GETFLAGS:
1193                 cmd = EXT4_IOC_GETFLAGS;
1194                 break;
1195         case EXT4_IOC32_SETFLAGS:
1196                 cmd = EXT4_IOC_SETFLAGS;
1197                 break;
1198         case EXT4_IOC32_GETVERSION:
1199                 cmd = EXT4_IOC_GETVERSION;
1200                 break;
1201         case EXT4_IOC32_SETVERSION:
1202                 cmd = EXT4_IOC_SETVERSION;
1203                 break;
1204         case EXT4_IOC32_GROUP_EXTEND:
1205                 cmd = EXT4_IOC_GROUP_EXTEND;
1206                 break;
1207         case EXT4_IOC32_GETVERSION_OLD:
1208                 cmd = EXT4_IOC_GETVERSION_OLD;
1209                 break;
1210         case EXT4_IOC32_SETVERSION_OLD:
1211                 cmd = EXT4_IOC_SETVERSION_OLD;
1212                 break;
1213         case EXT4_IOC32_GETRSVSZ:
1214                 cmd = EXT4_IOC_GETRSVSZ;
1215                 break;
1216         case EXT4_IOC32_SETRSVSZ:
1217                 cmd = EXT4_IOC_SETRSVSZ;
1218                 break;
1219         case EXT4_IOC32_GROUP_ADD: {
1220                 struct compat_ext4_new_group_input __user *uinput;
1221                 struct ext4_new_group_data input;
1222                 int err;
1223
1224                 uinput = compat_ptr(arg);
1225                 err = get_user(input.group, &uinput->group);
1226                 err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1227                 err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1228                 err |= get_user(input.inode_table, &uinput->inode_table);
1229                 err |= get_user(input.blocks_count, &uinput->blocks_count);
1230                 err |= get_user(input.reserved_blocks,
1231                                 &uinput->reserved_blocks);
1232                 if (err)
1233                         return -EFAULT;
1234                 return ext4_ioctl_group_add(file, &input);
1235         }
1236         case EXT4_IOC_MOVE_EXT:
1237         case EXT4_IOC_RESIZE_FS:
1238         case EXT4_IOC_PRECACHE_EXTENTS:
1239         case EXT4_IOC_SET_ENCRYPTION_POLICY:
1240         case EXT4_IOC_GET_ENCRYPTION_PWSALT:
1241         case EXT4_IOC_GET_ENCRYPTION_POLICY:
1242         case EXT4_IOC_SHUTDOWN:
1243         case FS_IOC_GETFSMAP:
1244                 break;
1245         default:
1246                 return -ENOIOCTLCMD;
1247         }
1248         return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1249 }
1250 #endif