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