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