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/uaccess.h>
  20 #include <linux/delay.h>
  21 #include <linux/iversion.h>
  22 #include <linux/fileattr.h>
  23 #include <linux/uuid.h>
  24 #include "ext4_jbd2.h"
  25 #include "ext4.h"
  26 #include <linux/fsmap.h>
  27 #include "fsmap.h"
  28 #include <trace/events/ext4.h>
  29
  30 typedef void ext4_update_sb_callback(struct ext4_super_block *es,
  31                                        const void *arg);
  32
  33 /*
  34  * Superblock modification callback function for changing file system
  35  * label
  36  */
  37 static void ext4_sb_setlabel(struct ext4_super_block *es, const void *arg)
  38 {
  39         /* Sanity check, this should never happen */
  40         BUILD_BUG_ON(sizeof(es->s_volume_name) < EXT4_LABEL_MAX);
  41
  42         memcpy(es->s_volume_name, (char *)arg, EXT4_LABEL_MAX);
  43 }
  44
  45 /*
  46  * Superblock modification callback function for changing file system
  47  * UUID.
  48  */
  49 static void ext4_sb_setuuid(struct ext4_super_block *es, const void *arg)
  50 {
  51         memcpy(es->s_uuid, (__u8 *)arg, UUID_SIZE);
  52 }
  53
  54 static
  55 int ext4_update_primary_sb(struct super_block *sb, handle_t *handle,
  56                            ext4_update_sb_callback func,
  57                            const void *arg)
  58 {
  59         int err = 0;
  60         struct ext4_sb_info *sbi = EXT4_SB(sb);
  61         struct buffer_head *bh = sbi->s_sbh;
  62         struct ext4_super_block *es = sbi->s_es;
  63
  64         trace_ext4_update_sb(sb, bh->b_blocknr, 1);
  65
  66         BUFFER_TRACE(bh, "get_write_access");
  67         err = ext4_journal_get_write_access(handle, sb,
  68                                             bh,
  69                                             EXT4_JTR_NONE);
  70         if (err)
  71                 goto out_err;
  72
  73         lock_buffer(bh);
  74         func(es, arg);
  75         ext4_superblock_csum_set(sb);
  76         unlock_buffer(bh);
  77
  78         if (buffer_write_io_error(bh) || !buffer_uptodate(bh)) {
  79                 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
  80                          "superblock detected");
  81                 clear_buffer_write_io_error(bh);
  82                 set_buffer_uptodate(bh);
  83         }
  84
  85         err = ext4_handle_dirty_metadata(handle, NULL, bh);
  86         if (err)
  87                 goto out_err;
  88         err = sync_dirty_buffer(bh);
  89 out_err:
  90         ext4_std_error(sb, err);
  91         return err;
  92 }
  93
  94 /*
  95  * Update one backup superblock in the group 'grp' using the callback
  96  * function 'func' and argument 'arg'. If the handle is NULL the
  97  * modification is not journalled.
  98  *
  99  * Returns: 0 when no modification was done (no superblock in the group)
 100  *          1 when the modification was successful
 101  *         <0 on error
 102  */
 103 static int ext4_update_backup_sb(struct super_block *sb,
 104                                  handle_t *handle, ext4_group_t grp,
 105                                  ext4_update_sb_callback func, const void *arg)
 106 {
 107         int err = 0;
 108         ext4_fsblk_t sb_block;
 109         struct buffer_head *bh;
 110         unsigned long offset = 0;
 111         struct ext4_super_block *es;
 112
 113         if (!ext4_bg_has_super(sb, grp))
 114                 return 0;
 115
 116         /*
 117          * For the group 0 there is always 1k padding, so we have
 118          * either adjust offset, or sb_block depending on blocksize
 119          */
 120         if (grp == 0) {
 121                 sb_block = 1 * EXT4_MIN_BLOCK_SIZE;
 122                 offset = do_div(sb_block, sb->s_blocksize);
 123         } else {
 124                 sb_block = ext4_group_first_block_no(sb, grp);
 125                 offset = 0;
 126         }
 127
 128         trace_ext4_update_sb(sb, sb_block, handle ? 1 : 0);
 129
 130         bh = ext4_sb_bread(sb, sb_block, 0);
 131         if (IS_ERR(bh))
 132                 return PTR_ERR(bh);
 133
 134         if (handle) {
 135                 BUFFER_TRACE(bh, "get_write_access");
 136                 err = ext4_journal_get_write_access(handle, sb,
 137                                                     bh,
 138                                                     EXT4_JTR_NONE);
 139                 if (err)
 140                         goto out_bh;
 141         }
 142
 143         es = (struct ext4_super_block *) (bh->b_data + offset);
 144         lock_buffer(bh);
 145         if (ext4_has_metadata_csum(sb) &&
 146             es->s_checksum != ext4_superblock_csum(sb, es)) {
 147                 ext4_msg(sb, KERN_ERR, "Invalid checksum for backup "
 148                 "superblock %llu", sb_block);
 149                 unlock_buffer(bh);
 150                 goto out_bh;
 151         }
 152         func(es, arg);
 153         if (ext4_has_metadata_csum(sb))
 154                 es->s_checksum = ext4_superblock_csum(sb, es);
 155         set_buffer_uptodate(bh);
 156         unlock_buffer(bh);
 157
 158         if (err)
 159                 goto out_bh;
 160
 161         if (handle) {
 162                 err = ext4_handle_dirty_metadata(handle, NULL, bh);
 163                 if (err)
 164                         goto out_bh;
 165         } else {
 166                 BUFFER_TRACE(bh, "marking dirty");
 167                 mark_buffer_dirty(bh);
 168         }
 169         err = sync_dirty_buffer(bh);
 170
 171 out_bh:
 172         brelse(bh);
 173         ext4_std_error(sb, err);
 174         return (err) ? err : 1;
 175 }
 176
 177 /*
 178  * Update primary and backup superblocks using the provided function
 179  * func and argument arg.
 180  *
 181  * Only the primary superblock and at most two backup superblock
 182  * modifications are journalled; the rest is modified without journal.
 183  * This is safe because e2fsck will re-write them if there is a problem,
 184  * and we're very unlikely to ever need more than two backups.
 185  */
 186 static
 187 int ext4_update_superblocks_fn(struct super_block *sb,
 188                                ext4_update_sb_callback func,
 189                                const void *arg)
 190 {
 191         handle_t *handle;
 192         ext4_group_t ngroups;
 193         unsigned int three = 1;
 194         unsigned int five = 5;
 195         unsigned int seven = 7;
 196         int err = 0, ret, i;
 197         ext4_group_t grp, primary_grp;
 198         struct ext4_sb_info *sbi = EXT4_SB(sb);
 199
 200         /*
 201          * We can't update superblocks while the online resize is running
 202          */
 203         if (test_and_set_bit_lock(EXT4_FLAGS_RESIZING,
 204                                   &sbi->s_ext4_flags)) {
 205                 ext4_msg(sb, KERN_ERR, "Can't modify superblock while"
 206                          "performing online resize");
 207                 return -EBUSY;
 208         }
 209
 210         /*
 211          * We're only going to update primary superblock and two
 212          * backup superblocks in this transaction.
 213          */
 214         handle = ext4_journal_start_sb(sb, EXT4_HT_MISC, 3);
 215         if (IS_ERR(handle)) {
 216                 err = PTR_ERR(handle);
 217                 goto out;
 218         }
 219
 220         /* Update primary superblock */
 221         err = ext4_update_primary_sb(sb, handle, func, arg);
 222         if (err) {
 223                 ext4_msg(sb, KERN_ERR, "Failed to update primary "
 224                          "superblock");
 225                 goto out_journal;
 226         }
 227
 228         primary_grp = ext4_get_group_number(sb, sbi->s_sbh->b_blocknr);
 229         ngroups = ext4_get_groups_count(sb);
 230
 231         /*
 232          * Update backup superblocks. We have to start from group 0
 233          * because it might not be where the primary superblock is
 234          * if the fs is mounted with -o sb=<backup_sb_block>
 235          */
 236         i = 0;
 237         grp = 0;
 238         while (grp < ngroups) {
 239                 /* Skip primary superblock */
 240                 if (grp == primary_grp)
 241                         goto next_grp;
 242
 243                 ret = ext4_update_backup_sb(sb, handle, grp, func, arg);
 244                 if (ret < 0) {
 245                         /* Ignore bad checksum; try to update next sb */
 246                         if (ret == -EFSBADCRC)
 247                                 goto next_grp;
 248                         err = ret;
 249                         goto out_journal;
 250                 }
 251
 252                 i += ret;
 253                 if (handle && i > 1) {
 254                         /*
 255                          * We're only journalling primary superblock and
 256                          * two backup superblocks; the rest is not
 257                          * journalled.
 258                          */
 259                         err = ext4_journal_stop(handle);
 260                         if (err)
 261                                 goto out;
 262                         handle = NULL;
 263                 }
 264 next_grp:
 265                 grp = ext4_list_backups(sb, &three, &five, &seven);
 266         }
 267
 268 out_journal:
 269         if (handle) {
 270                 ret = ext4_journal_stop(handle);
 271                 if (ret && !err)
 272                         err = ret;
 273         }
 274 out:
 275         clear_bit_unlock(EXT4_FLAGS_RESIZING, &sbi->s_ext4_flags);
 276         smp_mb__after_atomic();
 277         return err ? err : 0;
 278 }
 279
 280 /*
 281  * Swap memory between @a and @b for @len bytes.
 282  *
 283  * @a:          pointer to first memory area
 284  * @b:          pointer to second memory area
 285  * @len:        number of bytes to swap
 286  *
 287  */
 288 static void memswap(void *a, void *b, size_t len)
 289 {
 290         unsigned char *ap, *bp;
 291
 292         ap = (unsigned char *)a;
 293         bp = (unsigned char *)b;
 294         while (len-- > 0) {
 295                 swap(*ap, *bp);
 296                 ap++;
 297                 bp++;
 298         }
 299 }
 300
 301 /*
 302  * Swap i_data and associated attributes between @inode1 and @inode2.
 303  * This function is used for the primary swap between inode1 and inode2
 304  * and also to revert this primary swap in case of errors.
 305  *
 306  * Therefore you have to make sure, that calling this method twice
 307  * will revert all changes.
 308  *
 309  * @inode1:     pointer to first inode
 310  * @inode2:     pointer to second inode
 311  */
 312 static void swap_inode_data(struct inode *inode1, struct inode *inode2)
 313 {
 314         loff_t isize;
 315         struct ext4_inode_info *ei1;
 316         struct ext4_inode_info *ei2;
 317         unsigned long tmp;
 318
 319         ei1 = EXT4_I(inode1);
 320         ei2 = EXT4_I(inode2);
 321
 322         swap(inode1->i_version, inode2->i_version);
 323         swap(inode1->i_atime, inode2->i_atime);
 324         swap(inode1->i_mtime, inode2->i_mtime);
 325
 326         memswap(ei1->i_data, ei2->i_data, sizeof(ei1->i_data));
 327         tmp = ei1->i_flags & EXT4_FL_SHOULD_SWAP;
 328         ei1->i_flags = (ei2->i_flags & EXT4_FL_SHOULD_SWAP) |
 329                 (ei1->i_flags & ~EXT4_FL_SHOULD_SWAP);
 330         ei2->i_flags = tmp | (ei2->i_flags & ~EXT4_FL_SHOULD_SWAP);
 331         swap(ei1->i_disksize, ei2->i_disksize);
 332         ext4_es_remove_extent(inode1, 0, EXT_MAX_BLOCKS);
 333         ext4_es_remove_extent(inode2, 0, EXT_MAX_BLOCKS);
 334
 335         isize = i_size_read(inode1);
 336         i_size_write(inode1, i_size_read(inode2));
 337         i_size_write(inode2, isize);
 338 }
 339
 340 void ext4_reset_inode_seed(struct inode *inode)
 341 {
 342         struct ext4_inode_info *ei = EXT4_I(inode);
 343         struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
 344         __le32 inum = cpu_to_le32(inode->i_ino);
 345         __le32 gen = cpu_to_le32(inode->i_generation);
 346         __u32 csum;
 347
 348         if (!ext4_has_metadata_csum(inode->i_sb))
 349                 return;
 350
 351         csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&inum, sizeof(inum));
 352         ei->i_csum_seed = ext4_chksum(sbi, csum, (__u8 *)&gen, sizeof(gen));
 353 }
 354
 355 /*
 356  * Swap the information from the given @inode and the inode
 357  * EXT4_BOOT_LOADER_INO. It will basically swap i_data and all other
 358  * important fields of the inodes.
 359  *
 360  * @sb:         the super block of the filesystem
 361  * @mnt_userns: user namespace of the mount the inode was found from
 362  * @inode:      the inode to swap with EXT4_BOOT_LOADER_INO
 363  *
 364  */
 365 static long swap_inode_boot_loader(struct super_block *sb,
 366                                 struct user_namespace *mnt_userns,
 367                                 struct inode *inode)
 368 {
 369         handle_t *handle;
 370         int err;
 371         struct inode *inode_bl;
 372         struct ext4_inode_info *ei_bl;
 373         qsize_t size, size_bl, diff;
 374         blkcnt_t blocks;
 375         unsigned short bytes;
 376
 377         inode_bl = ext4_iget(sb, EXT4_BOOT_LOADER_INO,
 378                         EXT4_IGET_SPECIAL | EXT4_IGET_BAD);
 379         if (IS_ERR(inode_bl))
 380                 return PTR_ERR(inode_bl);
 381         ei_bl = EXT4_I(inode_bl);
 382
 383         /* Protect orig inodes against a truncate and make sure,
 384          * that only 1 swap_inode_boot_loader is running. */
 385         lock_two_nondirectories(inode, inode_bl);
 386
 387         if (inode->i_nlink != 1 || !S_ISREG(inode->i_mode) ||
 388             IS_SWAPFILE(inode) || IS_ENCRYPTED(inode) ||
 389             (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL) ||
 390             ext4_has_inline_data(inode)) {
 391                 err = -EINVAL;
 392                 goto journal_err_out;
 393         }
 394
 395         if (IS_RDONLY(inode) || IS_APPEND(inode) || IS_IMMUTABLE(inode) ||
 396             !inode_owner_or_capable(mnt_userns, inode) ||
 397             !capable(CAP_SYS_ADMIN)) {
 398                 err = -EPERM;
 399                 goto journal_err_out;
 400         }
 401
 402         filemap_invalidate_lock(inode->i_mapping);
 403         err = filemap_write_and_wait(inode->i_mapping);
 404         if (err)
 405                 goto err_out;
 406
 407         err = filemap_write_and_wait(inode_bl->i_mapping);
 408         if (err)
 409                 goto err_out;
 410
 411         /* Wait for all existing dio workers */
 412         inode_dio_wait(inode);
 413         inode_dio_wait(inode_bl);
 414
 415         truncate_inode_pages(&inode->i_data, 0);
 416         truncate_inode_pages(&inode_bl->i_data, 0);
 417
 418         handle = ext4_journal_start(inode_bl, EXT4_HT_MOVE_EXTENTS, 2);
 419         if (IS_ERR(handle)) {
 420                 err = -EINVAL;
 421                 goto err_out;
 422         }
 423         ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_SWAP_BOOT, handle);
 424
 425         /* Protect extent tree against block allocations via delalloc */
 426         ext4_double_down_write_data_sem(inode, inode_bl);
 427
 428         if (is_bad_inode(inode_bl) || !S_ISREG(inode_bl->i_mode)) {
 429                 /* this inode has never been used as a BOOT_LOADER */
 430                 set_nlink(inode_bl, 1);
 431                 i_uid_write(inode_bl, 0);
 432                 i_gid_write(inode_bl, 0);
 433                 inode_bl->i_flags = 0;
 434                 ei_bl->i_flags = 0;
 435                 inode_set_iversion(inode_bl, 1);
 436                 i_size_write(inode_bl, 0);
 437                 EXT4_I(inode_bl)->i_disksize = inode_bl->i_size;
 438                 inode_bl->i_mode = S_IFREG;
 439                 if (ext4_has_feature_extents(sb)) {
 440                         ext4_set_inode_flag(inode_bl, EXT4_INODE_EXTENTS);
 441                         ext4_ext_tree_init(handle, inode_bl);
 442                 } else
 443                         memset(ei_bl->i_data, 0, sizeof(ei_bl->i_data));
 444         }
 445
 446         err = dquot_initialize(inode);
 447         if (err)
 448                 goto err_out1;
 449
 450         size = (qsize_t)(inode->i_blocks) * (1 << 9) + inode->i_bytes;
 451         size_bl = (qsize_t)(inode_bl->i_blocks) * (1 << 9) + inode_bl->i_bytes;
 452         diff = size - size_bl;
 453         swap_inode_data(inode, inode_bl);
 454
 455         inode->i_ctime = inode_bl->i_ctime = current_time(inode);
 456         inode_inc_iversion(inode);
 457
 458         inode->i_generation = get_random_u32();
 459         inode_bl->i_generation = get_random_u32();
 460         ext4_reset_inode_seed(inode);
 461         ext4_reset_inode_seed(inode_bl);
 462
 463         ext4_discard_preallocations(inode, 0);
 464
 465         err = ext4_mark_inode_dirty(handle, inode);
 466         if (err < 0) {
 467                 /* No need to update quota information. */
 468                 ext4_warning(inode->i_sb,
 469                         "couldn't mark inode #%lu dirty (err %d)",
 470                         inode->i_ino, err);
 471                 /* Revert all changes: */
 472                 swap_inode_data(inode, inode_bl);
 473                 ext4_mark_inode_dirty(handle, inode);
 474                 goto err_out1;
 475         }
 476
 477         blocks = inode_bl->i_blocks;
 478         bytes = inode_bl->i_bytes;
 479         inode_bl->i_blocks = inode->i_blocks;
 480         inode_bl->i_bytes = inode->i_bytes;
 481         err = ext4_mark_inode_dirty(handle, inode_bl);
 482         if (err < 0) {
 483                 /* No need to update quota information. */
 484                 ext4_warning(inode_bl->i_sb,
 485                         "couldn't mark inode #%lu dirty (err %d)",
 486                         inode_bl->i_ino, err);
 487                 goto revert;
 488         }
 489
 490         /* Bootloader inode should not be counted into quota information. */
 491         if (diff > 0)
 492                 dquot_free_space(inode, diff);
 493         else
 494                 err = dquot_alloc_space(inode, -1 * diff);
 495
 496         if (err < 0) {
 497 revert:
 498                 /* Revert all changes: */
 499                 inode_bl->i_blocks = blocks;
 500                 inode_bl->i_bytes = bytes;
 501                 swap_inode_data(inode, inode_bl);
 502                 ext4_mark_inode_dirty(handle, inode);
 503                 ext4_mark_inode_dirty(handle, inode_bl);
 504         }
 505
 506 err_out1:
 507         ext4_journal_stop(handle);
 508         ext4_double_up_write_data_sem(inode, inode_bl);
 509
 510 err_out:
 511         filemap_invalidate_unlock(inode->i_mapping);
 512 journal_err_out:
 513         unlock_two_nondirectories(inode, inode_bl);
 514         iput(inode_bl);
 515         return err;
 516 }
 517
 518 /*
 519  * If immutable is set and we are not clearing it, we're not allowed to change
 520  * anything else in the inode.  Don't error out if we're only trying to set
 521  * immutable on an immutable file.
 522  */
 523 static int ext4_ioctl_check_immutable(struct inode *inode, __u32 new_projid,
 524                                       unsigned int flags)
 525 {
 526         struct ext4_inode_info *ei = EXT4_I(inode);
 527         unsigned int oldflags = ei->i_flags;
 528
 529         if (!(oldflags & EXT4_IMMUTABLE_FL) || !(flags & EXT4_IMMUTABLE_FL))
 530                 return 0;
 531
 532         if ((oldflags & ~EXT4_IMMUTABLE_FL) != (flags & ~EXT4_IMMUTABLE_FL))
 533                 return -EPERM;
 534         if (ext4_has_feature_project(inode->i_sb) &&
 535             __kprojid_val(ei->i_projid) != new_projid)
 536                 return -EPERM;
 537
 538         return 0;
 539 }
 540
 541 static void ext4_dax_dontcache(struct inode *inode, unsigned int flags)
 542 {
 543         struct ext4_inode_info *ei = EXT4_I(inode);
 544
 545         if (S_ISDIR(inode->i_mode))
 546                 return;
 547
 548         if (test_opt2(inode->i_sb, DAX_NEVER) ||
 549             test_opt(inode->i_sb, DAX_ALWAYS))
 550                 return;
 551
 552         if ((ei->i_flags ^ flags) & EXT4_DAX_FL)
 553                 d_mark_dontcache(inode);
 554 }
 555
 556 static bool dax_compatible(struct inode *inode, unsigned int oldflags,
 557                            unsigned int flags)
 558 {
 559         /* Allow the DAX flag to be changed on inline directories */
 560         if (S_ISDIR(inode->i_mode)) {
 561                 flags &= ~EXT4_INLINE_DATA_FL;
 562                 oldflags &= ~EXT4_INLINE_DATA_FL;
 563         }
 564
 565         if (flags & EXT4_DAX_FL) {
 566                 if ((oldflags & EXT4_DAX_MUT_EXCL) ||
 567                      ext4_test_inode_state(inode,
 568                                           EXT4_STATE_VERITY_IN_PROGRESS)) {
 569                         return false;
 570                 }
 571         }
 572
 573         if ((flags & EXT4_DAX_MUT_EXCL) && (oldflags & EXT4_DAX_FL))
 574                         return false;
 575
 576         return true;
 577 }
 578
 579 static int ext4_ioctl_setflags(struct inode *inode,
 580                                unsigned int flags)
 581 {
 582         struct ext4_inode_info *ei = EXT4_I(inode);
 583         handle_t *handle = NULL;
 584         int err = -EPERM, migrate = 0;
 585         struct ext4_iloc iloc;
 586         unsigned int oldflags, mask, i;
 587         struct super_block *sb = inode->i_sb;
 588
 589         /* Is it quota file? Do not allow user to mess with it */
 590         if (ext4_is_quota_file(inode))
 591                 goto flags_out;
 592
 593         oldflags = ei->i_flags;
 594         /*
 595          * The JOURNAL_DATA flag can only be changed by
 596          * the relevant capability.
 597          */
 598         if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
 599                 if (!capable(CAP_SYS_RESOURCE))
 600                         goto flags_out;
 601         }
 602
 603         if (!dax_compatible(inode, oldflags, flags)) {
 604                 err = -EOPNOTSUPP;
 605                 goto flags_out;
 606         }
 607
 608         if ((flags ^ oldflags) & EXT4_EXTENTS_FL)
 609                 migrate = 1;
 610
 611         if ((flags ^ oldflags) & EXT4_CASEFOLD_FL) {
 612                 if (!ext4_has_feature_casefold(sb)) {
 613                         err = -EOPNOTSUPP;
 614                         goto flags_out;
 615                 }
 616
 617                 if (!S_ISDIR(inode->i_mode)) {
 618                         err = -ENOTDIR;
 619                         goto flags_out;
 620                 }
 621
 622                 if (!ext4_empty_dir(inode)) {
 623                         err = -ENOTEMPTY;
 624                         goto flags_out;
 625                 }
 626         }
 627
 628         /*
 629          * Wait for all pending directio and then flush all the dirty pages
 630          * for this file.  The flush marks all the pages readonly, so any
 631          * subsequent attempt to write to the file (particularly mmap pages)
 632          * will come through the filesystem and fail.
 633          */
 634         if (S_ISREG(inode->i_mode) && !IS_IMMUTABLE(inode) &&
 635             (flags & EXT4_IMMUTABLE_FL)) {
 636                 inode_dio_wait(inode);
 637                 err = filemap_write_and_wait(inode->i_mapping);
 638                 if (err)
 639                         goto flags_out;
 640         }
 641
 642         handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
 643         if (IS_ERR(handle)) {
 644                 err = PTR_ERR(handle);
 645                 goto flags_out;
 646         }
 647         if (IS_SYNC(inode))
 648                 ext4_handle_sync(handle);
 649         err = ext4_reserve_inode_write(handle, inode, &iloc);
 650         if (err)
 651                 goto flags_err;
 652
 653         ext4_dax_dontcache(inode, flags);
 654
 655         for (i = 0, mask = 1; i < 32; i++, mask <<= 1) {
 656                 if (!(mask & EXT4_FL_USER_MODIFIABLE))
 657                         continue;
 658                 /* These flags get special treatment later */
 659                 if (mask == EXT4_JOURNAL_DATA_FL || mask == EXT4_EXTENTS_FL)
 660                         continue;
 661                 if (mask & flags)
 662                         ext4_set_inode_flag(inode, i);
 663                 else
 664                         ext4_clear_inode_flag(inode, i);
 665         }
 666
 667         ext4_set_inode_flags(inode, false);
 668
 669         inode->i_ctime = current_time(inode);
 670         inode_inc_iversion(inode);
 671
 672         err = ext4_mark_iloc_dirty(handle, inode, &iloc);
 673 flags_err:
 674         ext4_journal_stop(handle);
 675         if (err)
 676                 goto flags_out;
 677
 678         if ((flags ^ oldflags) & (EXT4_JOURNAL_DATA_FL)) {
 679                 /*
 680                  * Changes to the journaling mode can cause unsafe changes to
 681                  * S_DAX if the inode is DAX
 682                  */
 683                 if (IS_DAX(inode)) {
 684                         err = -EBUSY;
 685                         goto flags_out;
 686                 }
 687
 688                 err = ext4_change_inode_journal_flag(inode,
 689                                                      flags & EXT4_JOURNAL_DATA_FL);
 690                 if (err)
 691                         goto flags_out;
 692         }
 693         if (migrate) {
 694                 if (flags & EXT4_EXTENTS_FL)
 695                         err = ext4_ext_migrate(inode);
 696                 else
 697                         err = ext4_ind_migrate(inode);
 698         }
 699
 700 flags_out:
 701         return err;
 702 }
 703
 704 #ifdef CONFIG_QUOTA
 705 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
 706 {
 707         struct super_block *sb = inode->i_sb;
 708         struct ext4_inode_info *ei = EXT4_I(inode);
 709         int err, rc;
 710         handle_t *handle;
 711         kprojid_t kprojid;
 712         struct ext4_iloc iloc;
 713         struct ext4_inode *raw_inode;
 714         struct dquot *transfer_to[MAXQUOTAS] = { };
 715
 716         if (!ext4_has_feature_project(sb)) {
 717                 if (projid != EXT4_DEF_PROJID)
 718                         return -EOPNOTSUPP;
 719                 else
 720                         return 0;
 721         }
 722
 723         if (EXT4_INODE_SIZE(sb) <= EXT4_GOOD_OLD_INODE_SIZE)
 724                 return -EOPNOTSUPP;
 725
 726         kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
 727
 728         if (projid_eq(kprojid, EXT4_I(inode)->i_projid))
 729                 return 0;
 730
 731         err = -EPERM;
 732         /* Is it quota file? Do not allow user to mess with it */
 733         if (ext4_is_quota_file(inode))
 734                 return err;
 735
 736         err = dquot_initialize(inode);
 737         if (err)
 738                 return err;
 739
 740         err = ext4_get_inode_loc(inode, &iloc);
 741         if (err)
 742                 return err;
 743
 744         raw_inode = ext4_raw_inode(&iloc);
 745         if (!EXT4_FITS_IN_INODE(raw_inode, ei, i_projid)) {
 746                 err = ext4_expand_extra_isize(inode,
 747                                               EXT4_SB(sb)->s_want_extra_isize,
 748                                               &iloc);
 749                 if (err)
 750                         return err;
 751         } else {
 752                 brelse(iloc.bh);
 753         }
 754
 755         handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
 756                 EXT4_QUOTA_INIT_BLOCKS(sb) +
 757                 EXT4_QUOTA_DEL_BLOCKS(sb) + 3);
 758         if (IS_ERR(handle))
 759                 return PTR_ERR(handle);
 760
 761         err = ext4_reserve_inode_write(handle, inode, &iloc);
 762         if (err)
 763                 goto out_stop;
 764
 765         transfer_to[PRJQUOTA] = dqget(sb, make_kqid_projid(kprojid));
 766         if (!IS_ERR(transfer_to[PRJQUOTA])) {
 767
 768                 /* __dquot_transfer() calls back ext4_get_inode_usage() which
 769                  * counts xattr inode references.
 770                  */
 771                 down_read(&EXT4_I(inode)->xattr_sem);
 772                 err = __dquot_transfer(inode, transfer_to);
 773                 up_read(&EXT4_I(inode)->xattr_sem);
 774                 dqput(transfer_to[PRJQUOTA]);
 775                 if (err)
 776                         goto out_dirty;
 777         }
 778
 779         EXT4_I(inode)->i_projid = kprojid;
 780         inode->i_ctime = current_time(inode);
 781         inode_inc_iversion(inode);
 782 out_dirty:
 783         rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
 784         if (!err)
 785                 err = rc;
 786 out_stop:
 787         ext4_journal_stop(handle);
 788         return err;
 789 }
 790 #else
 791 static int ext4_ioctl_setproject(struct inode *inode, __u32 projid)
 792 {
 793         if (projid != EXT4_DEF_PROJID)
 794                 return -EOPNOTSUPP;
 795         return 0;
 796 }
 797 #endif
 798
 799 static int ext4_shutdown(struct super_block *sb, unsigned long arg)
 800 {
 801         struct ext4_sb_info *sbi = EXT4_SB(sb);
 802         __u32 flags;
 803         int ret;
 804
 805         if (!capable(CAP_SYS_ADMIN))
 806                 return -EPERM;
 807
 808         if (get_user(flags, (__u32 __user *)arg))
 809                 return -EFAULT;
 810
 811         if (flags > EXT4_GOING_FLAGS_NOLOGFLUSH)
 812                 return -EINVAL;
 813
 814         if (ext4_forced_shutdown(sbi))
 815                 return 0;
 816
 817         ext4_msg(sb, KERN_ALERT, "shut down requested (%d)", flags);
 818         trace_ext4_shutdown(sb, flags);
 819
 820         switch (flags) {
 821         case EXT4_GOING_FLAGS_DEFAULT:
 822                 ret = freeze_bdev(sb->s_bdev);
 823                 if (ret)
 824                         return ret;
 825                 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
 826                 thaw_bdev(sb->s_bdev);
 827                 break;
 828         case EXT4_GOING_FLAGS_LOGFLUSH:
 829                 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
 830                 if (sbi->s_journal && !is_journal_aborted(sbi->s_journal)) {
 831                         (void) ext4_force_commit(sb);
 832                         jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
 833                 }
 834                 break;
 835         case EXT4_GOING_FLAGS_NOLOGFLUSH:
 836                 set_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
 837                 if (sbi->s_journal && !is_journal_aborted(sbi->s_journal))
 838                         jbd2_journal_abort(sbi->s_journal, -ESHUTDOWN);
 839                 break;
 840         default:
 841                 return -EINVAL;
 842         }
 843         clear_opt(sb, DISCARD);
 844         return 0;
 845 }
 846
 847 struct getfsmap_info {
 848         struct super_block      *gi_sb;
 849         struct fsmap_head __user *gi_data;
 850         unsigned int            gi_idx;
 851         __u32                   gi_last_flags;
 852 };
 853
 854 static int ext4_getfsmap_format(struct ext4_fsmap *xfm, void *priv)
 855 {
 856         struct getfsmap_info *info = priv;
 857         struct fsmap fm;
 858
 859         trace_ext4_getfsmap_mapping(info->gi_sb, xfm);
 860
 861         info->gi_last_flags = xfm->fmr_flags;
 862         ext4_fsmap_from_internal(info->gi_sb, &fm, xfm);
 863         if (copy_to_user(&info->gi_data->fmh_recs[info->gi_idx++], &fm,
 864                         sizeof(struct fsmap)))
 865                 return -EFAULT;
 866
 867         return 0;
 868 }
 869
 870 static int ext4_ioc_getfsmap(struct super_block *sb,
 871                              struct fsmap_head __user *arg)
 872 {
 873         struct getfsmap_info info = { NULL };
 874         struct ext4_fsmap_head xhead = {0};
 875         struct fsmap_head head;
 876         bool aborted = false;
 877         int error;
 878
 879         if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
 880                 return -EFAULT;
 881         if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
 882             memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
 883                        sizeof(head.fmh_keys[0].fmr_reserved)) ||
 884             memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
 885                        sizeof(head.fmh_keys[1].fmr_reserved)))
 886                 return -EINVAL;
 887         /*
 888          * ext4 doesn't report file extents at all, so the only valid
 889          * file offsets are the magic ones (all zeroes or all ones).
 890          */
 891         if (head.fmh_keys[0].fmr_offset ||
 892             (head.fmh_keys[1].fmr_offset != 0 &&
 893              head.fmh_keys[1].fmr_offset != -1ULL))
 894                 return -EINVAL;
 895
 896         xhead.fmh_iflags = head.fmh_iflags;
 897         xhead.fmh_count = head.fmh_count;
 898         ext4_fsmap_to_internal(sb, &xhead.fmh_keys[0], &head.fmh_keys[0]);
 899         ext4_fsmap_to_internal(sb, &xhead.fmh_keys[1], &head.fmh_keys[1]);
 900
 901         trace_ext4_getfsmap_low_key(sb, &xhead.fmh_keys[0]);
 902         trace_ext4_getfsmap_high_key(sb, &xhead.fmh_keys[1]);
 903
 904         info.gi_sb = sb;
 905         info.gi_data = arg;
 906         error = ext4_getfsmap(sb, &xhead, ext4_getfsmap_format, &info);
 907         if (error == EXT4_QUERY_RANGE_ABORT)
 908                 aborted = true;
 909         else if (error)
 910                 return error;
 911
 912         /* If we didn't abort, set the "last" flag in the last fmx */
 913         if (!aborted && info.gi_idx) {
 914                 info.gi_last_flags |= FMR_OF_LAST;
 915                 if (copy_to_user(&info.gi_data->fmh_recs[info.gi_idx - 1].fmr_flags,
 916                                  &info.gi_last_flags,
 917                                  sizeof(info.gi_last_flags)))
 918                         return -EFAULT;
 919         }
 920
 921         /* copy back header */
 922         head.fmh_entries = xhead.fmh_entries;
 923         head.fmh_oflags = xhead.fmh_oflags;
 924         if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
 925                 return -EFAULT;
 926
 927         return 0;
 928 }
 929
 930 static long ext4_ioctl_group_add(struct file *file,
 931                                  struct ext4_new_group_data *input)
 932 {
 933         struct super_block *sb = file_inode(file)->i_sb;
 934         int err, err2=0;
 935
 936         err = ext4_resize_begin(sb);
 937         if (err)
 938                 return err;
 939
 940         if (ext4_has_feature_bigalloc(sb)) {
 941                 ext4_msg(sb, KERN_ERR,
 942                          "Online resizing not supported with bigalloc");
 943                 err = -EOPNOTSUPP;
 944                 goto group_add_out;
 945         }
 946
 947         err = mnt_want_write_file(file);
 948         if (err)
 949                 goto group_add_out;
 950
 951         err = ext4_group_add(sb, input);
 952         if (EXT4_SB(sb)->s_journal) {
 953                 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
 954                 err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
 955                 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
 956         }
 957         if (err == 0)
 958                 err = err2;
 959         mnt_drop_write_file(file);
 960         if (!err && ext4_has_group_desc_csum(sb) &&
 961             test_opt(sb, INIT_INODE_TABLE))
 962                 err = ext4_register_li_request(sb, input->group);
 963 group_add_out:
 964         err2 = ext4_resize_end(sb, false);
 965         if (err == 0)
 966                 err = err2;
 967         return err;
 968 }
 969
 970 int ext4_fileattr_get(struct dentry *dentry, struct fileattr *fa)
 971 {
 972         struct inode *inode = d_inode(dentry);
 973         struct ext4_inode_info *ei = EXT4_I(inode);
 974         u32 flags = ei->i_flags & EXT4_FL_USER_VISIBLE;
 975
 976         if (S_ISREG(inode->i_mode))
 977                 flags &= ~FS_PROJINHERIT_FL;
 978
 979         fileattr_fill_flags(fa, flags);
 980         if (ext4_has_feature_project(inode->i_sb))
 981                 fa->fsx_projid = from_kprojid(&init_user_ns, ei->i_projid);
 982
 983         return 0;
 984 }
 985
 986 int ext4_fileattr_set(struct user_namespace *mnt_userns,
 987                       struct dentry *dentry, struct fileattr *fa)
 988 {
 989         struct inode *inode = d_inode(dentry);
 990         u32 flags = fa->flags;
 991         int err = -EOPNOTSUPP;
 992
 993         if (flags & ~EXT4_FL_USER_VISIBLE)
 994                 goto out;
 995
 996         /*
 997          * chattr(1) grabs flags via GETFLAGS, modifies the result and
 998          * passes that to SETFLAGS. So we cannot easily make SETFLAGS
 999          * more restrictive than just silently masking off visible but
1000          * not settable flags as we always did.
1001          */
1002         flags &= EXT4_FL_USER_MODIFIABLE;
1003         if (ext4_mask_flags(inode->i_mode, flags) != flags)
1004                 goto out;
1005         err = ext4_ioctl_check_immutable(inode, fa->fsx_projid, flags);
1006         if (err)
1007                 goto out;
1008         err = ext4_ioctl_setflags(inode, flags);
1009         if (err)
1010                 goto out;
1011         err = ext4_ioctl_setproject(inode, fa->fsx_projid);
1012 out:
1013         return err;
1014 }
1015
1016 /* So that the fiemap access checks can't overflow on 32 bit machines. */
1017 #define FIEMAP_MAX_EXTENTS      (UINT_MAX / sizeof(struct fiemap_extent))
1018
1019 static int ext4_ioctl_get_es_cache(struct file *filp, unsigned long arg)
1020 {
1021         struct fiemap fiemap;
1022         struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
1023         struct fiemap_extent_info fieinfo = { 0, };
1024         struct inode *inode = file_inode(filp);
1025         int error;
1026
1027         if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
1028                 return -EFAULT;
1029
1030         if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
1031                 return -EINVAL;
1032
1033         fieinfo.fi_flags = fiemap.fm_flags;
1034         fieinfo.fi_extents_max = fiemap.fm_extent_count;
1035         fieinfo.fi_extents_start = ufiemap->fm_extents;
1036
1037         error = ext4_get_es_cache(inode, &fieinfo, fiemap.fm_start,
1038                         fiemap.fm_length);
1039         fiemap.fm_flags = fieinfo.fi_flags;
1040         fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
1041         if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
1042                 error = -EFAULT;
1043
1044         return error;
1045 }
1046
1047 static int ext4_ioctl_checkpoint(struct file *filp, unsigned long arg)
1048 {
1049         int err = 0;
1050         __u32 flags = 0;
1051         unsigned int flush_flags = 0;
1052         struct super_block *sb = file_inode(filp)->i_sb;
1053
1054         if (copy_from_user(&flags, (__u32 __user *)arg,
1055                                 sizeof(__u32)))
1056                 return -EFAULT;
1057
1058         if (!capable(CAP_SYS_ADMIN))
1059                 return -EPERM;
1060
1061         /* check for invalid bits set */
1062         if ((flags & ~EXT4_IOC_CHECKPOINT_FLAG_VALID) ||
1063                                 ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1064                                 (flags & JBD2_JOURNAL_FLUSH_ZEROOUT)))
1065                 return -EINVAL;
1066
1067         if (!EXT4_SB(sb)->s_journal)
1068                 return -ENODEV;
1069
1070         if ((flags & JBD2_JOURNAL_FLUSH_DISCARD) &&
1071             !bdev_max_discard_sectors(EXT4_SB(sb)->s_journal->j_dev))
1072                 return -EOPNOTSUPP;
1073
1074         if (flags & EXT4_IOC_CHECKPOINT_FLAG_DRY_RUN)
1075                 return 0;
1076
1077         if (flags & EXT4_IOC_CHECKPOINT_FLAG_DISCARD)
1078                 flush_flags |= JBD2_JOURNAL_FLUSH_DISCARD;
1079
1080         if (flags & EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT) {
1081                 flush_flags |= JBD2_JOURNAL_FLUSH_ZEROOUT;
1082                 pr_info_ratelimited("warning: checkpointing journal with EXT4_IOC_CHECKPOINT_FLAG_ZEROOUT can be slow");
1083         }
1084
1085         jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1086         err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, flush_flags);
1087         jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1088
1089         return err;
1090 }
1091
1092 static int ext4_ioctl_setlabel(struct file *filp, const char __user *user_label)
1093 {
1094         size_t len;
1095         int ret = 0;
1096         char new_label[EXT4_LABEL_MAX + 1];
1097         struct super_block *sb = file_inode(filp)->i_sb;
1098
1099         if (!capable(CAP_SYS_ADMIN))
1100                 return -EPERM;
1101
1102         /*
1103          * Copy the maximum length allowed for ext4 label with one more to
1104          * find the required terminating null byte in order to test the
1105          * label length. The on disk label doesn't need to be null terminated.
1106          */
1107         if (copy_from_user(new_label, user_label, EXT4_LABEL_MAX + 1))
1108                 return -EFAULT;
1109
1110         len = strnlen(new_label, EXT4_LABEL_MAX + 1);
1111         if (len > EXT4_LABEL_MAX)
1112                 return -EINVAL;
1113
1114         /*
1115          * Clear the buffer after the new label
1116          */
1117         memset(new_label + len, 0, EXT4_LABEL_MAX - len);
1118
1119         ret = mnt_want_write_file(filp);
1120         if (ret)
1121                 return ret;
1122
1123         ret = ext4_update_superblocks_fn(sb, ext4_sb_setlabel, new_label);
1124
1125         mnt_drop_write_file(filp);
1126         return ret;
1127 }
1128
1129 static int ext4_ioctl_getlabel(struct ext4_sb_info *sbi, char __user *user_label)
1130 {
1131         char label[EXT4_LABEL_MAX + 1];
1132
1133         /*
1134          * EXT4_LABEL_MAX must always be smaller than FSLABEL_MAX because
1135          * FSLABEL_MAX must include terminating null byte, while s_volume_name
1136          * does not have to.
1137          */
1138         BUILD_BUG_ON(EXT4_LABEL_MAX >= FSLABEL_MAX);
1139
1140         memset(label, 0, sizeof(label));
1141         lock_buffer(sbi->s_sbh);
1142         strncpy(label, sbi->s_es->s_volume_name, EXT4_LABEL_MAX);
1143         unlock_buffer(sbi->s_sbh);
1144
1145         if (copy_to_user(user_label, label, sizeof(label)))
1146                 return -EFAULT;
1147         return 0;
1148 }
1149
1150 static int ext4_ioctl_getuuid(struct ext4_sb_info *sbi,
1151                         struct fsuuid __user *ufsuuid)
1152 {
1153         struct fsuuid fsuuid;
1154         __u8 uuid[UUID_SIZE];
1155
1156         if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1157                 return -EFAULT;
1158
1159         if (fsuuid.fsu_len == 0) {
1160                 fsuuid.fsu_len = UUID_SIZE;
1161                 if (copy_to_user(&ufsuuid->fsu_len, &fsuuid.fsu_len,
1162                                         sizeof(fsuuid.fsu_len)))
1163                         return -EFAULT;
1164                 return 0;
1165         }
1166
1167         if (fsuuid.fsu_len < UUID_SIZE || fsuuid.fsu_flags != 0)
1168                 return -EINVAL;
1169
1170         lock_buffer(sbi->s_sbh);
1171         memcpy(uuid, sbi->s_es->s_uuid, UUID_SIZE);
1172         unlock_buffer(sbi->s_sbh);
1173
1174         fsuuid.fsu_len = UUID_SIZE;
1175         if (copy_to_user(ufsuuid, &fsuuid, sizeof(fsuuid)) ||
1176             copy_to_user(&ufsuuid->fsu_uuid[0], uuid, UUID_SIZE))
1177                 return -EFAULT;
1178         return 0;
1179 }
1180
1181 static int ext4_ioctl_setuuid(struct file *filp,
1182                         const struct fsuuid __user *ufsuuid)
1183 {
1184         int ret = 0;
1185         struct super_block *sb = file_inode(filp)->i_sb;
1186         struct fsuuid fsuuid;
1187         __u8 uuid[UUID_SIZE];
1188
1189         if (!capable(CAP_SYS_ADMIN))
1190                 return -EPERM;
1191
1192         /*
1193          * If any checksums (group descriptors or metadata) are being used
1194          * then the checksum seed feature is required to change the UUID.
1195          */
1196         if (((ext4_has_feature_gdt_csum(sb) || ext4_has_metadata_csum(sb))
1197                         && !ext4_has_feature_csum_seed(sb))
1198                 || ext4_has_feature_stable_inodes(sb))
1199                 return -EOPNOTSUPP;
1200
1201         if (copy_from_user(&fsuuid, ufsuuid, sizeof(fsuuid)))
1202                 return -EFAULT;
1203
1204         if (fsuuid.fsu_len != UUID_SIZE || fsuuid.fsu_flags != 0)
1205                 return -EINVAL;
1206
1207         if (copy_from_user(uuid, &ufsuuid->fsu_uuid[0], UUID_SIZE))
1208                 return -EFAULT;
1209
1210         ret = mnt_want_write_file(filp);
1211         if (ret)
1212                 return ret;
1213
1214         ret = ext4_update_superblocks_fn(sb, ext4_sb_setuuid, &uuid);
1215         mnt_drop_write_file(filp);
1216
1217         return ret;
1218 }
1219
1220 static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1221 {
1222         struct inode *inode = file_inode(filp);
1223         struct super_block *sb = inode->i_sb;
1224         struct user_namespace *mnt_userns = file_mnt_user_ns(filp);
1225
1226         ext4_debug("cmd = %u, arg = %lu\n", cmd, arg);
1227
1228         switch (cmd) {
1229         case FS_IOC_GETFSMAP:
1230                 return ext4_ioc_getfsmap(sb, (void __user *)arg);
1231         case EXT4_IOC_GETVERSION:
1232         case EXT4_IOC_GETVERSION_OLD:
1233                 return put_user(inode->i_generation, (int __user *) arg);
1234         case EXT4_IOC_SETVERSION:
1235         case EXT4_IOC_SETVERSION_OLD: {
1236                 handle_t *handle;
1237                 struct ext4_iloc iloc;
1238                 __u32 generation;
1239                 int err;
1240
1241                 if (!inode_owner_or_capable(mnt_userns, inode))
1242                         return -EPERM;
1243
1244                 if (ext4_has_metadata_csum(inode->i_sb)) {
1245                         ext4_warning(sb, "Setting inode version is not "
1246                                      "supported with metadata_csum enabled.");
1247                         return -ENOTTY;
1248                 }
1249
1250                 err = mnt_want_write_file(filp);
1251                 if (err)
1252                         return err;
1253                 if (get_user(generation, (int __user *) arg)) {
1254                         err = -EFAULT;
1255                         goto setversion_out;
1256                 }
1257
1258                 inode_lock(inode);
1259                 handle = ext4_journal_start(inode, EXT4_HT_INODE, 1);
1260                 if (IS_ERR(handle)) {
1261                         err = PTR_ERR(handle);
1262                         goto unlock_out;
1263                 }
1264                 err = ext4_reserve_inode_write(handle, inode, &iloc);
1265                 if (err == 0) {
1266                         inode->i_ctime = current_time(inode);
1267                         inode_inc_iversion(inode);
1268                         inode->i_generation = generation;
1269                         err = ext4_mark_iloc_dirty(handle, inode, &iloc);
1270                 }
1271                 ext4_journal_stop(handle);
1272
1273 unlock_out:
1274                 inode_unlock(inode);
1275 setversion_out:
1276                 mnt_drop_write_file(filp);
1277                 return err;
1278         }
1279         case EXT4_IOC_GROUP_EXTEND: {
1280                 ext4_fsblk_t n_blocks_count;
1281                 int err, err2=0;
1282
1283                 err = ext4_resize_begin(sb);
1284                 if (err)
1285                         return err;
1286
1287                 if (get_user(n_blocks_count, (__u32 __user *)arg)) {
1288                         err = -EFAULT;
1289                         goto group_extend_out;
1290                 }
1291
1292                 if (ext4_has_feature_bigalloc(sb)) {
1293                         ext4_msg(sb, KERN_ERR,
1294                                  "Online resizing not supported with bigalloc");
1295                         err = -EOPNOTSUPP;
1296                         goto group_extend_out;
1297                 }
1298
1299                 err = mnt_want_write_file(filp);
1300                 if (err)
1301                         goto group_extend_out;
1302
1303                 err = ext4_group_extend(sb, EXT4_SB(sb)->s_es, n_blocks_count);
1304                 if (EXT4_SB(sb)->s_journal) {
1305                         jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1306                         err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1307                         jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1308                 }
1309                 if (err == 0)
1310                         err = err2;
1311                 mnt_drop_write_file(filp);
1312 group_extend_out:
1313                 err2 = ext4_resize_end(sb, false);
1314                 if (err == 0)
1315                         err = err2;
1316                 return err;
1317         }
1318
1319         case EXT4_IOC_MOVE_EXT: {
1320                 struct move_extent me;
1321                 struct fd donor;
1322                 int err;
1323
1324                 if (!(filp->f_mode & FMODE_READ) ||
1325                     !(filp->f_mode & FMODE_WRITE))
1326                         return -EBADF;
1327
1328                 if (copy_from_user(&me,
1329                         (struct move_extent __user *)arg, sizeof(me)))
1330                         return -EFAULT;
1331                 me.moved_len = 0;
1332
1333                 donor = fdget(me.donor_fd);
1334                 if (!donor.file)
1335                         return -EBADF;
1336
1337                 if (!(donor.file->f_mode & FMODE_WRITE)) {
1338                         err = -EBADF;
1339                         goto mext_out;
1340                 }
1341
1342                 if (ext4_has_feature_bigalloc(sb)) {
1343                         ext4_msg(sb, KERN_ERR,
1344                                  "Online defrag not supported with bigalloc");
1345                         err = -EOPNOTSUPP;
1346                         goto mext_out;
1347                 } else if (IS_DAX(inode)) {
1348                         ext4_msg(sb, KERN_ERR,
1349                                  "Online defrag not supported with DAX");
1350                         err = -EOPNOTSUPP;
1351                         goto mext_out;
1352                 }
1353
1354                 err = mnt_want_write_file(filp);
1355                 if (err)
1356                         goto mext_out;
1357
1358                 err = ext4_move_extents(filp, donor.file, me.orig_start,
1359                                         me.donor_start, me.len, &me.moved_len);
1360                 mnt_drop_write_file(filp);
1361
1362                 if (copy_to_user((struct move_extent __user *)arg,
1363                                  &me, sizeof(me)))
1364                         err = -EFAULT;
1365 mext_out:
1366                 fdput(donor);
1367                 return err;
1368         }
1369
1370         case EXT4_IOC_GROUP_ADD: {
1371                 struct ext4_new_group_data input;
1372
1373                 if (copy_from_user(&input, (struct ext4_new_group_input __user *)arg,
1374                                 sizeof(input)))
1375                         return -EFAULT;
1376
1377                 return ext4_ioctl_group_add(filp, &input);
1378         }
1379
1380         case EXT4_IOC_MIGRATE:
1381         {
1382                 int err;
1383                 if (!inode_owner_or_capable(mnt_userns, inode))
1384                         return -EACCES;
1385
1386                 err = mnt_want_write_file(filp);
1387                 if (err)
1388                         return err;
1389                 /*
1390                  * inode_mutex prevent write and truncate on the file.
1391                  * Read still goes through. We take i_data_sem in
1392                  * ext4_ext_swap_inode_data before we switch the
1393                  * inode format to prevent read.
1394                  */
1395                 inode_lock((inode));
1396                 err = ext4_ext_migrate(inode);
1397                 inode_unlock((inode));
1398                 mnt_drop_write_file(filp);
1399                 return err;
1400         }
1401
1402         case EXT4_IOC_ALLOC_DA_BLKS:
1403         {
1404                 int err;
1405                 if (!inode_owner_or_capable(mnt_userns, inode))
1406                         return -EACCES;
1407
1408                 err = mnt_want_write_file(filp);
1409                 if (err)
1410                         return err;
1411                 err = ext4_alloc_da_blocks(inode);
1412                 mnt_drop_write_file(filp);
1413                 return err;
1414         }
1415
1416         case EXT4_IOC_SWAP_BOOT:
1417         {
1418                 int err;
1419                 if (!(filp->f_mode & FMODE_WRITE))
1420                         return -EBADF;
1421                 err = mnt_want_write_file(filp);
1422                 if (err)
1423                         return err;
1424                 err = swap_inode_boot_loader(sb, mnt_userns, inode);
1425                 mnt_drop_write_file(filp);
1426                 return err;
1427         }
1428
1429         case EXT4_IOC_RESIZE_FS: {
1430                 ext4_fsblk_t n_blocks_count;
1431                 int err = 0, err2 = 0;
1432                 ext4_group_t o_group = EXT4_SB(sb)->s_groups_count;
1433
1434                 if (copy_from_user(&n_blocks_count, (__u64 __user *)arg,
1435                                    sizeof(__u64))) {
1436                         return -EFAULT;
1437                 }
1438
1439                 err = ext4_resize_begin(sb);
1440                 if (err)
1441                         return err;
1442
1443                 err = mnt_want_write_file(filp);
1444                 if (err)
1445                         goto resizefs_out;
1446
1447                 err = ext4_resize_fs(sb, n_blocks_count);
1448                 if (EXT4_SB(sb)->s_journal) {
1449                         ext4_fc_mark_ineligible(sb, EXT4_FC_REASON_RESIZE, NULL);
1450                         jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
1451                         err2 = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
1452                         jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
1453                 }
1454                 if (err == 0)
1455                         err = err2;
1456                 mnt_drop_write_file(filp);
1457                 if (!err && (o_group < EXT4_SB(sb)->s_groups_count) &&
1458                     ext4_has_group_desc_csum(sb) &&
1459                     test_opt(sb, INIT_INODE_TABLE))
1460                         err = ext4_register_li_request(sb, o_group);
1461
1462 resizefs_out:
1463                 err2 = ext4_resize_end(sb, true);
1464                 if (err == 0)
1465                         err = err2;
1466                 return err;
1467         }
1468
1469         case FITRIM:
1470         {
1471                 struct fstrim_range range;
1472                 int ret = 0;
1473
1474                 if (!capable(CAP_SYS_ADMIN))
1475                         return -EPERM;
1476
1477                 if (!bdev_max_discard_sectors(sb->s_bdev))
1478                         return -EOPNOTSUPP;
1479
1480                 /*
1481                  * We haven't replayed the journal, so we cannot use our
1482                  * block-bitmap-guided storage zapping commands.
1483                  */
1484                 if (test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb))
1485                         return -EROFS;
1486
1487                 if (copy_from_user(&range, (struct fstrim_range __user *)arg,
1488                     sizeof(range)))
1489                         return -EFAULT;
1490
1491                 ret = ext4_trim_fs(sb, &range);
1492                 if (ret < 0)
1493                         return ret;
1494
1495                 if (copy_to_user((struct fstrim_range __user *)arg, &range,
1496                     sizeof(range)))
1497                         return -EFAULT;
1498
1499                 return 0;
1500         }
1501         case EXT4_IOC_PRECACHE_EXTENTS:
1502                 return ext4_ext_precache(inode);
1503
1504         case FS_IOC_SET_ENCRYPTION_POLICY:
1505                 if (!ext4_has_feature_encrypt(sb))
1506                         return -EOPNOTSUPP;
1507                 return fscrypt_ioctl_set_policy(filp, (const void __user *)arg);
1508
1509         case FS_IOC_GET_ENCRYPTION_PWSALT:
1510                 return ext4_ioctl_get_encryption_pwsalt(filp, (void __user *)arg);
1511
1512         case FS_IOC_GET_ENCRYPTION_POLICY:
1513                 if (!ext4_has_feature_encrypt(sb))
1514                         return -EOPNOTSUPP;
1515                 return fscrypt_ioctl_get_policy(filp, (void __user *)arg);
1516
1517         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1518                 if (!ext4_has_feature_encrypt(sb))
1519                         return -EOPNOTSUPP;
1520                 return fscrypt_ioctl_get_policy_ex(filp, (void __user *)arg);
1521
1522         case FS_IOC_ADD_ENCRYPTION_KEY:
1523                 if (!ext4_has_feature_encrypt(sb))
1524                         return -EOPNOTSUPP;
1525                 return fscrypt_ioctl_add_key(filp, (void __user *)arg);
1526
1527         case FS_IOC_REMOVE_ENCRYPTION_KEY:
1528                 if (!ext4_has_feature_encrypt(sb))
1529                         return -EOPNOTSUPP;
1530                 return fscrypt_ioctl_remove_key(filp, (void __user *)arg);
1531
1532         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1533                 if (!ext4_has_feature_encrypt(sb))
1534                         return -EOPNOTSUPP;
1535                 return fscrypt_ioctl_remove_key_all_users(filp,
1536                                                           (void __user *)arg);
1537         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1538                 if (!ext4_has_feature_encrypt(sb))
1539                         return -EOPNOTSUPP;
1540                 return fscrypt_ioctl_get_key_status(filp, (void __user *)arg);
1541
1542         case FS_IOC_GET_ENCRYPTION_NONCE:
1543                 if (!ext4_has_feature_encrypt(sb))
1544                         return -EOPNOTSUPP;
1545                 return fscrypt_ioctl_get_nonce(filp, (void __user *)arg);
1546
1547         case EXT4_IOC_CLEAR_ES_CACHE:
1548         {
1549                 if (!inode_owner_or_capable(mnt_userns, inode))
1550                         return -EACCES;
1551                 ext4_clear_inode_es(inode);
1552                 return 0;
1553         }
1554
1555         case EXT4_IOC_GETSTATE:
1556         {
1557                 __u32   state = 0;
1558
1559                 if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED))
1560                         state |= EXT4_STATE_FLAG_EXT_PRECACHED;
1561                 if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
1562                         state |= EXT4_STATE_FLAG_NEW;
1563                 if (ext4_test_inode_state(inode, EXT4_STATE_NEWENTRY))
1564                         state |= EXT4_STATE_FLAG_NEWENTRY;
1565                 if (ext4_test_inode_state(inode, EXT4_STATE_DA_ALLOC_CLOSE))
1566                         state |= EXT4_STATE_FLAG_DA_ALLOC_CLOSE;
1567
1568                 return put_user(state, (__u32 __user *) arg);
1569         }
1570
1571         case EXT4_IOC_GET_ES_CACHE:
1572                 return ext4_ioctl_get_es_cache(filp, arg);
1573
1574         case EXT4_IOC_SHUTDOWN:
1575                 return ext4_shutdown(sb, arg);
1576
1577         case FS_IOC_ENABLE_VERITY:
1578                 if (!ext4_has_feature_verity(sb))
1579                         return -EOPNOTSUPP;
1580                 return fsverity_ioctl_enable(filp, (const void __user *)arg);
1581
1582         case FS_IOC_MEASURE_VERITY:
1583                 if (!ext4_has_feature_verity(sb))
1584                         return -EOPNOTSUPP;
1585                 return fsverity_ioctl_measure(filp, (void __user *)arg);
1586
1587         case FS_IOC_READ_VERITY_METADATA:
1588                 if (!ext4_has_feature_verity(sb))
1589                         return -EOPNOTSUPP;
1590                 return fsverity_ioctl_read_metadata(filp,
1591                                                     (const void __user *)arg);
1592
1593         case EXT4_IOC_CHECKPOINT:
1594                 return ext4_ioctl_checkpoint(filp, arg);
1595
1596         case FS_IOC_GETFSLABEL:
1597                 return ext4_ioctl_getlabel(EXT4_SB(sb), (void __user *)arg);
1598
1599         case FS_IOC_SETFSLABEL:
1600                 return ext4_ioctl_setlabel(filp,
1601                                            (const void __user *)arg);
1602
1603         case EXT4_IOC_GETFSUUID:
1604                 return ext4_ioctl_getuuid(EXT4_SB(sb), (void __user *)arg);
1605         case EXT4_IOC_SETFSUUID:
1606                 return ext4_ioctl_setuuid(filp, (const void __user *)arg);
1607         default:
1608                 return -ENOTTY;
1609         }
1610 }
1611
1612 long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1613 {
1614         return __ext4_ioctl(filp, cmd, arg);
1615 }
1616
1617 #ifdef CONFIG_COMPAT
1618 long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1619 {
1620         /* These are just misnamed, they actually get/put from/to user an int */
1621         switch (cmd) {
1622         case EXT4_IOC32_GETVERSION:
1623                 cmd = EXT4_IOC_GETVERSION;
1624                 break;
1625         case EXT4_IOC32_SETVERSION:
1626                 cmd = EXT4_IOC_SETVERSION;
1627                 break;
1628         case EXT4_IOC32_GROUP_EXTEND:
1629                 cmd = EXT4_IOC_GROUP_EXTEND;
1630                 break;
1631         case EXT4_IOC32_GETVERSION_OLD:
1632                 cmd = EXT4_IOC_GETVERSION_OLD;
1633                 break;
1634         case EXT4_IOC32_SETVERSION_OLD:
1635                 cmd = EXT4_IOC_SETVERSION_OLD;
1636                 break;
1637         case EXT4_IOC32_GETRSVSZ:
1638                 cmd = EXT4_IOC_GETRSVSZ;
1639                 break;
1640         case EXT4_IOC32_SETRSVSZ:
1641                 cmd = EXT4_IOC_SETRSVSZ;
1642                 break;
1643         case EXT4_IOC32_GROUP_ADD: {
1644                 struct compat_ext4_new_group_input __user *uinput;
1645                 struct ext4_new_group_data input;
1646                 int err;
1647
1648                 uinput = compat_ptr(arg);
1649                 err = get_user(input.group, &uinput->group);
1650                 err |= get_user(input.block_bitmap, &uinput->block_bitmap);
1651                 err |= get_user(input.inode_bitmap, &uinput->inode_bitmap);
1652                 err |= get_user(input.inode_table, &uinput->inode_table);
1653                 err |= get_user(input.blocks_count, &uinput->blocks_count);
1654                 err |= get_user(input.reserved_blocks,
1655                                 &uinput->reserved_blocks);
1656                 if (err)
1657                         return -EFAULT;
1658                 return ext4_ioctl_group_add(file, &input);
1659         }
1660         case EXT4_IOC_MOVE_EXT:
1661         case EXT4_IOC_RESIZE_FS:
1662         case FITRIM:
1663         case EXT4_IOC_PRECACHE_EXTENTS:
1664         case FS_IOC_SET_ENCRYPTION_POLICY:
1665         case FS_IOC_GET_ENCRYPTION_PWSALT:
1666         case FS_IOC_GET_ENCRYPTION_POLICY:
1667         case FS_IOC_GET_ENCRYPTION_POLICY_EX:
1668         case FS_IOC_ADD_ENCRYPTION_KEY:
1669         case FS_IOC_REMOVE_ENCRYPTION_KEY:
1670         case FS_IOC_REMOVE_ENCRYPTION_KEY_ALL_USERS:
1671         case FS_IOC_GET_ENCRYPTION_KEY_STATUS:
1672         case FS_IOC_GET_ENCRYPTION_NONCE:
1673         case EXT4_IOC_SHUTDOWN:
1674         case FS_IOC_GETFSMAP:
1675         case FS_IOC_ENABLE_VERITY:
1676         case FS_IOC_MEASURE_VERITY:
1677         case FS_IOC_READ_VERITY_METADATA:
1678         case EXT4_IOC_CLEAR_ES_CACHE:
1679         case EXT4_IOC_GETSTATE:
1680         case EXT4_IOC_GET_ES_CACHE:
1681         case EXT4_IOC_CHECKPOINT:
1682         case FS_IOC_GETFSLABEL:
1683         case FS_IOC_SETFSLABEL:
1684         case EXT4_IOC_GETFSUUID:
1685         case EXT4_IOC_SETFSUUID:
1686                 break;
1687         default:
1688                 return -ENOIOCTLCMD;
1689         }
1690         return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
1691 }
1692 #endif
1693
1694 static void set_overhead(struct ext4_super_block *es, const void *arg)
1695 {
1696         es->s_overhead_clusters = cpu_to_le32(*((unsigned long *) arg));
1697 }
1698
1699 int ext4_update_overhead(struct super_block *sb, bool force)
1700 {
1701         struct ext4_sb_info *sbi = EXT4_SB(sb);
1702
1703         if (sb_rdonly(sb))
1704                 return 0;
1705         if (!force &&
1706             (sbi->s_overhead == 0 ||
1707              sbi->s_overhead == le32_to_cpu(sbi->s_es->s_overhead_clusters)))
1708                 return 0;
1709         return ext4_update_superblocks_fn(sb, set_overhead, &sbi->s_overhead);
1710 }