fs/ioctl.c

   1 /*
   2  *  linux/fs/ioctl.c
   3  *
   4  *  Copyright (C) 1991, 1992  Linus Torvalds
   5  */
   6
   7 #include <linux/syscalls.h>
   8 #include <linux/mm.h>
   9 #include <linux/smp_lock.h>
  10 #include <linux/capability.h>
  11 #include <linux/file.h>
  12 #include <linux/fs.h>
  13 #include <linux/security.h>
  14 #include <linux/module.h>
  15 #include <linux/uaccess.h>
  16 #include <linux/writeback.h>
  17 #include <linux/buffer_head.h>
  18 #include <linux/falloc.h>
  19
  20 #include <asm/ioctls.h>
  21
  22 /* So that the fiemap access checks can't overflow on 32 bit machines. */
  23 #define FIEMAP_MAX_EXTENTS      (UINT_MAX / sizeof(struct fiemap_extent))
  24
  25 /**
  26  * vfs_ioctl - call filesystem specific ioctl methods
  27  * @filp:       open file to invoke ioctl method on
  28  * @cmd:        ioctl command to execute
  29  * @arg:        command-specific argument for ioctl
  30  *
  31  * Invokes filesystem specific ->unlocked_ioctl, if one exists; otherwise
  32  * returns -ENOTTY.
  33  *
  34  * Returns 0 on success, -errno on error.
  35  */
  36 static long vfs_ioctl(struct file *filp, unsigned int cmd,
  37                       unsigned long arg)
  38 {
  39         int error = -ENOTTY;
  40
  41         if (!filp->f_op || !filp->f_op->unlocked_ioctl)
  42                 goto out;
  43
  44         error = filp->f_op->unlocked_ioctl(filp, cmd, arg);
  45         if (error == -ENOIOCTLCMD)
  46                 error = -EINVAL;
  47  out:
  48         return error;
  49 }
  50
  51 static int ioctl_fibmap(struct file *filp, int __user *p)
  52 {
  53         struct address_space *mapping = filp->f_mapping;
  54         int res, block;
  55
  56         /* do we support this mess? */
  57         if (!mapping->a_ops->bmap)
  58                 return -EINVAL;
  59         if (!capable(CAP_SYS_RAWIO))
  60                 return -EPERM;
  61         res = get_user(block, p);
  62         if (res)
  63                 return res;
  64         res = mapping->a_ops->bmap(mapping, block);
  65         return put_user(res, p);
  66 }
  67
  68 /**
  69  * fiemap_fill_next_extent - Fiemap helper function
  70  * @fieinfo:    Fiemap context passed into ->fiemap
  71  * @logical:    Extent logical start offset, in bytes
  72  * @phys:       Extent physical start offset, in bytes
  73  * @len:        Extent length, in bytes
  74  * @flags:      FIEMAP_EXTENT flags that describe this extent
  75  *
  76  * Called from file system ->fiemap callback. Will populate extent
  77  * info as passed in via arguments and copy to user memory. On
  78  * success, extent count on fieinfo is incremented.
  79  *
  80  * Returns 0 on success, -errno on error, 1 if this was the last
  81  * extent that will fit in user array.
  82  */
  83 #define SET_UNKNOWN_FLAGS       (FIEMAP_EXTENT_DELALLOC)
  84 #define SET_NO_UNMOUNTED_IO_FLAGS       (FIEMAP_EXTENT_DATA_ENCRYPTED)
  85 #define SET_NOT_ALIGNED_FLAGS   (FIEMAP_EXTENT_DATA_TAIL|FIEMAP_EXTENT_DATA_INLINE)
  86 int fiemap_fill_next_extent(struct fiemap_extent_info *fieinfo, u64 logical,
  87                             u64 phys, u64 len, u32 flags)
  88 {
  89         struct fiemap_extent extent;
  90         struct fiemap_extent *dest = fieinfo->fi_extents_start;
  91
  92         /* only count the extents */
  93         if (fieinfo->fi_extents_max == 0) {
  94                 fieinfo->fi_extents_mapped++;
  95                 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
  96         }
  97
  98         if (fieinfo->fi_extents_mapped >= fieinfo->fi_extents_max)
  99                 return 1;
 100
 101         if (flags & SET_UNKNOWN_FLAGS)
 102                 flags |= FIEMAP_EXTENT_UNKNOWN;
 103         if (flags & SET_NO_UNMOUNTED_IO_FLAGS)
 104                 flags |= FIEMAP_EXTENT_ENCODED;
 105         if (flags & SET_NOT_ALIGNED_FLAGS)
 106                 flags |= FIEMAP_EXTENT_NOT_ALIGNED;
 107
 108         memset(&extent, 0, sizeof(extent));
 109         extent.fe_logical = logical;
 110         extent.fe_physical = phys;
 111         extent.fe_length = len;
 112         extent.fe_flags = flags;
 113
 114         dest += fieinfo->fi_extents_mapped;
 115         if (copy_to_user(dest, &extent, sizeof(extent)))
 116                 return -EFAULT;
 117
 118         fieinfo->fi_extents_mapped++;
 119         if (fieinfo->fi_extents_mapped == fieinfo->fi_extents_max)
 120                 return 1;
 121         return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
 122 }
 123 EXPORT_SYMBOL(fiemap_fill_next_extent);
 124
 125 /**
 126  * fiemap_check_flags - check validity of requested flags for fiemap
 127  * @fieinfo:    Fiemap context passed into ->fiemap
 128  * @fs_flags:   Set of fiemap flags that the file system understands
 129  *
 130  * Called from file system ->fiemap callback. This will compute the
 131  * intersection of valid fiemap flags and those that the fs supports. That
 132  * value is then compared against the user supplied flags. In case of bad user
 133  * flags, the invalid values will be written into the fieinfo structure, and
 134  * -EBADR is returned, which tells ioctl_fiemap() to return those values to
 135  * userspace. For this reason, a return code of -EBADR should be preserved.
 136  *
 137  * Returns 0 on success, -EBADR on bad flags.
 138  */
 139 int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags)
 140 {
 141         u32 incompat_flags;
 142
 143         incompat_flags = fieinfo->fi_flags & ~(FIEMAP_FLAGS_COMPAT & fs_flags);
 144         if (incompat_flags) {
 145                 fieinfo->fi_flags = incompat_flags;
 146                 return -EBADR;
 147         }
 148         return 0;
 149 }
 150 EXPORT_SYMBOL(fiemap_check_flags);
 151
 152 static int fiemap_check_ranges(struct super_block *sb,
 153                                u64 start, u64 len, u64 *new_len)
 154 {
 155         u64 maxbytes = (u64) sb->s_maxbytes;
 156
 157         *new_len = len;
 158
 159         if (len == 0)
 160                 return -EINVAL;
 161
 162         if (start > maxbytes)
 163                 return -EFBIG;
 164
 165         /*
 166          * Shrink request scope to what the fs can actually handle.
 167          */
 168         if (len > maxbytes || (maxbytes - len) < start)
 169                 *new_len = maxbytes - start;
 170
 171         return 0;
 172 }
 173
 174 static int ioctl_fiemap(struct file *filp, unsigned long arg)
 175 {
 176         struct fiemap fiemap;
 177         struct fiemap_extent_info fieinfo = { 0, };
 178         struct inode *inode = filp->f_path.dentry->d_inode;
 179         struct super_block *sb = inode->i_sb;
 180         u64 len;
 181         int error;
 182
 183         if (!inode->i_op->fiemap)
 184                 return -EOPNOTSUPP;
 185
 186         if (copy_from_user(&fiemap, (struct fiemap __user *)arg,
 187                            sizeof(struct fiemap)))
 188                 return -EFAULT;
 189
 190         if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
 191                 return -EINVAL;
 192
 193         error = fiemap_check_ranges(sb, fiemap.fm_start, fiemap.fm_length,
 194                                     &len);
 195         if (error)
 196                 return error;
 197
 198         fieinfo.fi_flags = fiemap.fm_flags;
 199         fieinfo.fi_extents_max = fiemap.fm_extent_count;
 200         fieinfo.fi_extents_start = (struct fiemap_extent *)(arg + sizeof(fiemap));
 201
 202         if (fiemap.fm_extent_count != 0 &&
 203             !access_ok(VERIFY_WRITE, fieinfo.fi_extents_start,
 204                        fieinfo.fi_extents_max * sizeof(struct fiemap_extent)))
 205                 return -EFAULT;
 206
 207         if (fieinfo.fi_flags & FIEMAP_FLAG_SYNC)
 208                 filemap_write_and_wait(inode->i_mapping);
 209
 210         error = inode->i_op->fiemap(inode, &fieinfo, fiemap.fm_start, len);
 211         fiemap.fm_flags = fieinfo.fi_flags;
 212         fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
 213         if (copy_to_user((char *)arg, &fiemap, sizeof(fiemap)))
 214                 error = -EFAULT;
 215
 216         return error;
 217 }
 218
 219 #ifdef CONFIG_BLOCK
 220
 221 static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
 222 {
 223         return (offset >> inode->i_blkbits);
 224 }
 225
 226 static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
 227 {
 228         return (blk << inode->i_blkbits);
 229 }
 230
 231 /**
 232  * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
 233  * @inode: the inode to map
 234  * @fieinfo: the fiemap info struct that will be passed back to userspace
 235  * @start: where to start mapping in the inode
 236  * @len: how much space to map
 237  * @get_block: the fs's get_block function
 238  *
 239  * This does FIEMAP for block based inodes.  Basically it will just loop
 240  * through get_block until we hit the number of extents we want to map, or we
 241  * go past the end of the file and hit a hole.
 242  *
 243  * If it is possible to have data blocks beyond a hole past @inode->i_size, then
 244  * please do not use this function, it will stop at the first unmapped block
 245  * beyond i_size.
 246  *
 247  * If you use this function directly, you need to do your own locking. Use
 248  * generic_block_fiemap if you want the locking done for you.
 249  */
 250
 251 int __generic_block_fiemap(struct inode *inode,
 252                            struct fiemap_extent_info *fieinfo, loff_t start,
 253                            loff_t len, get_block_t *get_block)
 254 {
 255         struct buffer_head map_bh;
 256         sector_t start_blk, last_blk;
 257         loff_t isize = i_size_read(inode);
 258         u64 logical = 0, phys = 0, size = 0;
 259         u32 flags = FIEMAP_EXTENT_MERGED;
 260         bool past_eof = false, whole_file = false;
 261         int ret = 0;
 262
 263         ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC);
 264         if (ret)
 265                 return ret;
 266
 267         /*
 268          * Either the i_mutex or other appropriate locking needs to be held
 269          * since we expect isize to not change at all through the duration of
 270          * this call.
 271          */
 272         if (len >= isize) {
 273                 whole_file = true;
 274                 len = isize;
 275         }
 276
 277         start_blk = logical_to_blk(inode, start);
 278         last_blk = logical_to_blk(inode, start + len - 1);
 279
 280         do {
 281                 /*
 282                  * we set b_size to the total size we want so it will map as
 283                  * many contiguous blocks as possible at once
 284                  */
 285                 memset(&map_bh, 0, sizeof(struct buffer_head));
 286                 map_bh.b_size = len;
 287
 288                 ret = get_block(inode, start_blk, &map_bh, 0);
 289                 if (ret)
 290                         break;
 291
 292                 /* HOLE */
 293                 if (!buffer_mapped(&map_bh)) {
 294                         start_blk++;
 295
 296                         /*
 297                          * We want to handle the case where there is an
 298                          * allocated block at the front of the file, and then
 299                          * nothing but holes up to the end of the file properly,
 300                          * to make sure that extent at the front gets properly
 301                          * marked with FIEMAP_EXTENT_LAST
 302                          */
 303                         if (!past_eof &&
 304                             blk_to_logical(inode, start_blk) >= isize)
 305                                 past_eof = 1;
 306
 307                         /*
 308                          * First hole after going past the EOF, this is our
 309                          * last extent
 310                          */
 311                         if (past_eof && size) {
 312                                 flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
 313                                 ret = fiemap_fill_next_extent(fieinfo, logical,
 314                                                               phys, size,
 315                                                               flags);
 316                         } else if (size) {
 317                                 ret = fiemap_fill_next_extent(fieinfo, logical,
 318                                                               phys, size, flags);
 319                                 size = 0;
 320                         }
 321
 322                         /* if we have holes up to/past EOF then we're done */
 323                         if (start_blk > last_blk || past_eof || ret)
 324                                 break;
 325                 } else {
 326                         /*
 327                          * We have gone over the length of what we wanted to
 328                          * map, and it wasn't the entire file, so add the extent
 329                          * we got last time and exit.
 330                          *
 331                          * This is for the case where say we want to map all the
 332                          * way up to the second to the last block in a file, but
 333                          * the last block is a hole, making the second to last
 334                          * block FIEMAP_EXTENT_LAST.  In this case we want to
 335                          * see if there is a hole after the second to last block
 336                          * so we can mark it properly.  If we found data after
 337                          * we exceeded the length we were requesting, then we
 338                          * are good to go, just add the extent to the fieinfo
 339                          * and break
 340                          */
 341                         if (start_blk > last_blk && !whole_file) {
 342                                 ret = fiemap_fill_next_extent(fieinfo, logical,
 343                                                               phys, size,
 344                                                               flags);
 345                                 break;
 346                         }
 347
 348                         /*
 349                          * if size != 0 then we know we already have an extent
 350                          * to add, so add it.
 351                          */
 352                         if (size) {
 353                                 ret = fiemap_fill_next_extent(fieinfo, logical,
 354                                                               phys, size,
 355                                                               flags);
 356                                 if (ret)
 357                                         break;
 358                         }
 359
 360                         logical = blk_to_logical(inode, start_blk);
 361                         phys = blk_to_logical(inode, map_bh.b_blocknr);
 362                         size = map_bh.b_size;
 363                         flags = FIEMAP_EXTENT_MERGED;
 364
 365                         start_blk += logical_to_blk(inode, size);
 366
 367                         /*
 368                          * If we are past the EOF, then we need to make sure as
 369                          * soon as we find a hole that the last extent we found
 370                          * is marked with FIEMAP_EXTENT_LAST
 371                          */
 372                         if (!past_eof && logical + size >= isize)
 373                                 past_eof = true;
 374                 }
 375                 cond_resched();
 376         } while (1);
 377
 378         /* If ret is 1 then we just hit the end of the extent array */
 379         if (ret == 1)
 380                 ret = 0;
 381
 382         return ret;
 383 }
 384 EXPORT_SYMBOL(__generic_block_fiemap);
 385
 386 /**
 387  * generic_block_fiemap - FIEMAP for block based inodes
 388  * @inode: The inode to map
 389  * @fieinfo: The mapping information
 390  * @start: The initial block to map
 391  * @len: The length of the extect to attempt to map
 392  * @get_block: The block mapping function for the fs
 393  *
 394  * Calls __generic_block_fiemap to map the inode, after taking
 395  * the inode's mutex lock.
 396  */
 397
 398 int generic_block_fiemap(struct inode *inode,
 399                          struct fiemap_extent_info *fieinfo, u64 start,
 400                          u64 len, get_block_t *get_block)
 401 {
 402         int ret;
 403         mutex_lock(&inode->i_mutex);
 404         ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
 405         mutex_unlock(&inode->i_mutex);
 406         return ret;
 407 }
 408 EXPORT_SYMBOL(generic_block_fiemap);
 409
 410 #endif  /*  CONFIG_BLOCK  */
 411
 412 /*
 413  * This provides compatibility with legacy XFS pre-allocation ioctls
 414  * which predate the fallocate syscall.
 415  *
 416  * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
 417  * are used here, rest are ignored.
 418  */
 419 int ioctl_preallocate(struct file *filp, void __user *argp)
 420 {
 421         struct inode *inode = filp->f_path.dentry->d_inode;
 422         struct space_resv sr;
 423
 424         if (copy_from_user(&sr, argp, sizeof(sr)))
 425                 return -EFAULT;
 426
 427         switch (sr.l_whence) {
 428         case SEEK_SET:
 429                 break;
 430         case SEEK_CUR:
 431                 sr.l_start += filp->f_pos;
 432                 break;
 433         case SEEK_END:
 434                 sr.l_start += i_size_read(inode);
 435                 break;
 436         default:
 437                 return -EINVAL;
 438         }
 439
 440         return do_fallocate(filp, FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
 441 }
 442
 443 static int file_ioctl(struct file *filp, unsigned int cmd,
 444                 unsigned long arg)
 445 {
 446         struct inode *inode = filp->f_path.dentry->d_inode;
 447         int __user *p = (int __user *)arg;
 448
 449         switch (cmd) {
 450         case FIBMAP:
 451                 return ioctl_fibmap(filp, p);
 452         case FIONREAD:
 453                 return put_user(i_size_read(inode) - filp->f_pos, p);
 454         case FS_IOC_RESVSP:
 455         case FS_IOC_RESVSP64:
 456                 return ioctl_preallocate(filp, p);
 457         }
 458
 459         return vfs_ioctl(filp, cmd, arg);
 460 }
 461
 462 static int ioctl_fionbio(struct file *filp, int __user *argp)
 463 {
 464         unsigned int flag;
 465         int on, error;
 466
 467         error = get_user(on, argp);
 468         if (error)
 469                 return error;
 470         flag = O_NONBLOCK;
 471 #ifdef __sparc__
 472         /* SunOS compatibility item. */
 473         if (O_NONBLOCK != O_NDELAY)
 474                 flag |= O_NDELAY;
 475 #endif
 476         spin_lock(&filp->f_lock);
 477         if (on)
 478                 filp->f_flags |= flag;
 479         else
 480                 filp->f_flags &= ~flag;
 481         spin_unlock(&filp->f_lock);
 482         return error;
 483 }
 484
 485 static int ioctl_fioasync(unsigned int fd, struct file *filp,
 486                           int __user *argp)
 487 {
 488         unsigned int flag;
 489         int on, error;
 490
 491         error = get_user(on, argp);
 492         if (error)
 493                 return error;
 494         flag = on ? FASYNC : 0;
 495
 496         /* Did FASYNC state change ? */
 497         if ((flag ^ filp->f_flags) & FASYNC) {
 498                 if (filp->f_op && filp->f_op->fasync)
 499                         /* fasync() adjusts filp->f_flags */
 500                         error = filp->f_op->fasync(fd, filp, on);
 501                 else
 502                         error = -ENOTTY;
 503         }
 504         return error < 0 ? error : 0;
 505 }
 506
 507 static int ioctl_fsfreeze(struct file *filp)
 508 {
 509         struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
 510
 511         if (!capable(CAP_SYS_ADMIN))
 512                 return -EPERM;
 513
 514         /* If filesystem doesn't support freeze feature, return. */
 515         if (sb->s_op->freeze_fs == NULL)
 516                 return -EOPNOTSUPP;
 517
 518         /* Freeze */
 519         return freeze_super(sb);
 520 }
 521
 522 static int ioctl_fsthaw(struct file *filp)
 523 {
 524         struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
 525
 526         if (!capable(CAP_SYS_ADMIN))
 527                 return -EPERM;
 528
 529         /* Thaw */
 530         return thaw_super(sb);
 531 }
 532
 533 static int ioctl_fstrim(struct file *filp, void __user *argp)
 534 {
 535         struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
 536         struct fstrim_range range;
 537         int ret = 0;
 538
 539         if (!capable(CAP_SYS_ADMIN))
 540                 return -EPERM;
 541
 542         /* If filesystem doesn't support trim feature, return. */
 543         if (sb->s_op->trim_fs == NULL)
 544                 return -EOPNOTSUPP;
 545
 546         /* If a blockdevice-backed filesystem isn't specified, return EINVAL. */
 547         if (sb->s_bdev == NULL)
 548                 return -EINVAL;
 549
 550         if (argp == NULL) {
 551                 range.start = 0;
 552                 range.len = ULLONG_MAX;
 553                 range.minlen = 0;
 554         } else if (copy_from_user(&range, argp, sizeof(range)))
 555                 return -EFAULT;
 556
 557         ret = sb->s_op->trim_fs(sb, &range);
 558         if (ret < 0)
 559                 return ret;
 560
 561         if ((argp != NULL) &&
 562             (copy_to_user(argp, &range, sizeof(range))))
 563                 return -EFAULT;
 564
 565         return 0;
 566 }
 567
 568 /*
 569  * When you add any new common ioctls to the switches above and below
 570  * please update compat_sys_ioctl() too.
 571  *
 572  * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
 573  * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
 574  */
 575 int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
 576              unsigned long arg)
 577 {
 578         int error = 0;
 579         int __user *argp = (int __user *)arg;
 580
 581         switch (cmd) {
 582         case FIOCLEX:
 583                 set_close_on_exec(fd, 1);
 584                 break;
 585
 586         case FIONCLEX:
 587                 set_close_on_exec(fd, 0);
 588                 break;
 589
 590         case FIONBIO:
 591                 error = ioctl_fionbio(filp, argp);
 592                 break;
 593
 594         case FIOASYNC:
 595                 error = ioctl_fioasync(fd, filp, argp);
 596                 break;
 597
 598         case FIOQSIZE:
 599                 if (S_ISDIR(filp->f_path.dentry->d_inode->i_mode) ||
 600                     S_ISREG(filp->f_path.dentry->d_inode->i_mode) ||
 601                     S_ISLNK(filp->f_path.dentry->d_inode->i_mode)) {
 602                         loff_t res =
 603                                 inode_get_bytes(filp->f_path.dentry->d_inode);
 604                         error = copy_to_user((loff_t __user *)arg, &res,
 605                                              sizeof(res)) ? -EFAULT : 0;
 606                 } else
 607                         error = -ENOTTY;
 608                 break;
 609
 610         case FIFREEZE:
 611                 error = ioctl_fsfreeze(filp);
 612                 break;
 613
 614         case FITHAW:
 615                 error = ioctl_fsthaw(filp);
 616                 break;
 617
 618         case FITRIM:
 619                 error = ioctl_fstrim(filp, argp);
 620                 break;
 621
 622         case FS_IOC_FIEMAP:
 623                 return ioctl_fiemap(filp, arg);
 624
 625         case FIGETBSZ:
 626         {
 627                 struct inode *inode = filp->f_path.dentry->d_inode;
 628                 int __user *p = (int __user *)arg;
 629                 return put_user(inode->i_sb->s_blocksize, p);
 630         }
 631
 632         default:
 633                 if (S_ISREG(filp->f_path.dentry->d_inode->i_mode))
 634                         error = file_ioctl(filp, cmd, arg);
 635                 else
 636                         error = vfs_ioctl(filp, cmd, arg);
 637                 break;
 638         }
 639         return error;
 640 }
 641
 642 SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
 643 {
 644         struct file *filp;
 645         int error = -EBADF;
 646         int fput_needed;
 647
 648         filp = fget_light(fd, &fput_needed);
 649         if (!filp)
 650                 goto out;
 651
 652         error = security_file_ioctl(filp, cmd, arg);
 653         if (error)
 654                 goto out_fput;
 655
 656         error = do_vfs_ioctl(filp, fd, cmd, arg);
 657  out_fput:
 658         fput_light(filp, fput_needed);
 659  out:
 660         return error;
 661 }