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