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 #define blk_to_logical(inode, blk) (blk << (inode)->i_blkbits)
 232 #define logical_to_blk(inode, offset) (offset >> (inode)->i_blkbits);
 233
 234 /**
 235  * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
 236  * @inode - the inode to map
 237  * @arg - the pointer to userspace where we copy everything to
 238  * @get_block - the fs's get_block function
 239  *
 240  * This does FIEMAP for block based inodes.  Basically it will just loop
 241  * through get_block until we hit the number of extents we want to map, or we
 242  * go past the end of the file and hit a hole.
 243  *
 244  * If it is possible to have data blocks beyond a hole past @inode->i_size, then
 245  * please do not use this function, it will stop at the first unmapped block
 246  * beyond i_size.
 247  *
 248  * If you use this function directly, you need to do your own locking. Use
 249  * generic_block_fiemap if you want the locking done for you.
 250  */
 251
 252 int __generic_block_fiemap(struct inode *inode,
 253                            struct fiemap_extent_info *fieinfo, u64 start,
 254                            u64 len, get_block_t *get_block)
 255 {
 256         struct buffer_head tmp;
 257         unsigned int start_blk;
 258         long long length = 0, map_len = 0;
 259         u64 logical = 0, phys = 0, size = 0;
 260         u32 flags = FIEMAP_EXTENT_MERGED;
 261         int ret = 0, past_eof = 0, whole_file = 0;
 262
 263         if ((ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC)))
 264                 return ret;
 265
 266         start_blk = logical_to_blk(inode, start);
 267
 268         length = (long long)min_t(u64, len, i_size_read(inode));
 269         if (length < len)
 270                 whole_file = 1;
 271
 272         map_len = length;
 273
 274         do {
 275                 /*
 276                  * we set b_size to the total size we want so it will map as
 277                  * many contiguous blocks as possible at once
 278                  */
 279                 memset(&tmp, 0, sizeof(struct buffer_head));
 280                 tmp.b_size = map_len;
 281
 282                 ret = get_block(inode, start_blk, &tmp, 0);
 283                 if (ret)
 284                         break;
 285
 286                 /* HOLE */
 287                 if (!buffer_mapped(&tmp)) {
 288                         length -= blk_to_logical(inode, 1);
 289                         start_blk++;
 290
 291                         /*
 292                          * we want to handle the case where there is an
 293                          * allocated block at the front of the file, and then
 294                          * nothing but holes up to the end of the file properly,
 295                          * to make sure that extent at the front gets properly
 296                          * marked with FIEMAP_EXTENT_LAST
 297                          */
 298                         if (!past_eof &&
 299                             blk_to_logical(inode, start_blk) >=
 300                             blk_to_logical(inode, 0)+i_size_read(inode))
 301                                 past_eof = 1;
 302
 303                         /*
 304                          * first hole after going past the EOF, this is our
 305                          * last extent
 306                          */
 307                         if (past_eof && size) {
 308                                 flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
 309                                 ret = fiemap_fill_next_extent(fieinfo, logical,
 310                                                               phys, size,
 311                                                               flags);
 312                                 break;
 313                         }
 314
 315                         /* if we have holes up to/past EOF then we're done */
 316                         if (length <= 0 || past_eof)
 317                                 break;
 318                 } else {
 319                         /*
 320                          * we have gone over the length of what we wanted to
 321                          * map, and it wasn't the entire file, so add the extent
 322                          * we got last time and exit.
 323                          *
 324                          * This is for the case where say we want to map all the
 325                          * way up to the second to the last block in a file, but
 326                          * the last block is a hole, making the second to last
 327                          * block FIEMAP_EXTENT_LAST.  In this case we want to
 328                          * see if there is a hole after the second to last block
 329                          * so we can mark it properly.  If we found data after
 330                          * we exceeded the length we were requesting, then we
 331                          * are good to go, just add the extent to the fieinfo
 332                          * and break
 333                          */
 334                         if (length <= 0 && !whole_file) {
 335                                 ret = fiemap_fill_next_extent(fieinfo, logical,
 336                                                               phys, size,
 337                                                               flags);
 338                                 break;
 339                         }
 340
 341                         /*
 342                          * if size != 0 then we know we already have an extent
 343                          * to add, so add it.
 344                          */
 345                         if (size) {
 346                                 ret = fiemap_fill_next_extent(fieinfo, logical,
 347                                                               phys, size,
 348                                                               flags);
 349                                 if (ret)
 350                                         break;
 351                         }
 352
 353                         logical = blk_to_logical(inode, start_blk);
 354                         phys = blk_to_logical(inode, tmp.b_blocknr);
 355                         size = tmp.b_size;
 356                         flags = FIEMAP_EXTENT_MERGED;
 357
 358                         length -= tmp.b_size;
 359                         start_blk += logical_to_blk(inode, size);
 360
 361                         /*
 362                          * If we are past the EOF, then we need to make sure as
 363                          * soon as we find a hole that the last extent we found
 364                          * is marked with FIEMAP_EXTENT_LAST
 365                          */
 366                         if (!past_eof &&
 367                             logical+size >=
 368                             blk_to_logical(inode, 0)+i_size_read(inode))
 369                                 past_eof = 1;
 370                 }
 371                 cond_resched();
 372         } while (1);
 373
 374         /* if ret is 1 then we just hit the end of the extent array */
 375         if (ret == 1)
 376                 ret = 0;
 377
 378         return ret;
 379 }
 380 EXPORT_SYMBOL(__generic_block_fiemap);
 381
 382 /**
 383  * generic_block_fiemap - FIEMAP for block based inodes
 384  * @inode: The inode to map
 385  * @fieinfo: The mapping information
 386  * @start: The initial block to map
 387  * @len: The length of the extect to attempt to map
 388  * @get_block: The block mapping function for the fs
 389  *
 390  * Calls __generic_block_fiemap to map the inode, after taking
 391  * the inode's mutex lock.
 392  */
 393
 394 int generic_block_fiemap(struct inode *inode,
 395                          struct fiemap_extent_info *fieinfo, u64 start,
 396                          u64 len, get_block_t *get_block)
 397 {
 398         int ret;
 399         mutex_lock(&inode->i_mutex);
 400         ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
 401         mutex_unlock(&inode->i_mutex);
 402         return ret;
 403 }
 404 EXPORT_SYMBOL(generic_block_fiemap);
 405
 406 #endif  /*  CONFIG_BLOCK  */
 407
 408 /*
 409  * This provides compatibility with legacy XFS pre-allocation ioctls
 410  * which predate the fallocate syscall.
 411  *
 412  * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
 413  * are used here, rest are ignored.
 414  */
 415 int ioctl_preallocate(struct file *filp, void __user *argp)
 416 {
 417         struct inode *inode = filp->f_path.dentry->d_inode;
 418         struct space_resv sr;
 419
 420         if (copy_from_user(&sr, argp, sizeof(sr)))
 421                 return -EFAULT;
 422
 423         switch (sr.l_whence) {
 424         case SEEK_SET:
 425                 break;
 426         case SEEK_CUR:
 427                 sr.l_start += filp->f_pos;
 428                 break;
 429         case SEEK_END:
 430                 sr.l_start += i_size_read(inode);
 431                 break;
 432         default:
 433                 return -EINVAL;
 434         }
 435
 436         return do_fallocate(filp, FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
 437 }
 438
 439 static int file_ioctl(struct file *filp, unsigned int cmd,
 440                 unsigned long arg)
 441 {
 442         struct inode *inode = filp->f_path.dentry->d_inode;
 443         int __user *p = (int __user *)arg;
 444
 445         switch (cmd) {
 446         case FIBMAP:
 447                 return ioctl_fibmap(filp, p);
 448         case FIONREAD:
 449                 return put_user(i_size_read(inode) - filp->f_pos, p);
 450         case FS_IOC_RESVSP:
 451         case FS_IOC_RESVSP64:
 452                 return ioctl_preallocate(filp, p);
 453         }
 454
 455         return vfs_ioctl(filp, cmd, arg);
 456 }
 457
 458 static int ioctl_fionbio(struct file *filp, int __user *argp)
 459 {
 460         unsigned int flag;
 461         int on, error;
 462
 463         error = get_user(on, argp);
 464         if (error)
 465                 return error;
 466         flag = O_NONBLOCK;
 467 #ifdef __sparc__
 468         /* SunOS compatibility item. */
 469         if (O_NONBLOCK != O_NDELAY)
 470                 flag |= O_NDELAY;
 471 #endif
 472         spin_lock(&filp->f_lock);
 473         if (on)
 474                 filp->f_flags |= flag;
 475         else
 476                 filp->f_flags &= ~flag;
 477         spin_unlock(&filp->f_lock);
 478         return error;
 479 }
 480
 481 static int ioctl_fioasync(unsigned int fd, struct file *filp,
 482                           int __user *argp)
 483 {
 484         unsigned int flag;
 485         int on, error;
 486
 487         error = get_user(on, argp);
 488         if (error)
 489                 return error;
 490         flag = on ? FASYNC : 0;
 491
 492         /* Did FASYNC state change ? */
 493         if ((flag ^ filp->f_flags) & FASYNC) {
 494                 if (filp->f_op && filp->f_op->fasync)
 495                         /* fasync() adjusts filp->f_flags */
 496                         error = filp->f_op->fasync(fd, filp, on);
 497                 else
 498                         error = -ENOTTY;
 499         }
 500         return error < 0 ? error : 0;
 501 }
 502
 503 static int ioctl_fsfreeze(struct file *filp)
 504 {
 505         struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
 506
 507         if (!capable(CAP_SYS_ADMIN))
 508                 return -EPERM;
 509
 510         /* If filesystem doesn't support freeze feature, return. */
 511         if (sb->s_op->freeze_fs == NULL)
 512                 return -EOPNOTSUPP;
 513
 514         /* If a blockdevice-backed filesystem isn't specified, return. */
 515         if (sb->s_bdev == NULL)
 516                 return -EINVAL;
 517
 518         /* Freeze */
 519         sb = freeze_bdev(sb->s_bdev);
 520         if (IS_ERR(sb))
 521                 return PTR_ERR(sb);
 522         return 0;
 523 }
 524
 525 static int ioctl_fsthaw(struct file *filp)
 526 {
 527         struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
 528
 529         if (!capable(CAP_SYS_ADMIN))
 530                 return -EPERM;
 531
 532         /* If a blockdevice-backed filesystem isn't specified, return EINVAL. */
 533         if (sb->s_bdev == NULL)
 534                 return -EINVAL;
 535
 536         /* Thaw */
 537         return thaw_bdev(sb->s_bdev, sb);
 538 }
 539
 540 /*
 541  * When you add any new common ioctls to the switches above and below
 542  * please update compat_sys_ioctl() too.
 543  *
 544  * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
 545  * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
 546  */
 547 int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
 548              unsigned long arg)
 549 {
 550         int error = 0;
 551         int __user *argp = (int __user *)arg;
 552
 553         switch (cmd) {
 554         case FIOCLEX:
 555                 set_close_on_exec(fd, 1);
 556                 break;
 557
 558         case FIONCLEX:
 559                 set_close_on_exec(fd, 0);
 560                 break;
 561
 562         case FIONBIO:
 563                 error = ioctl_fionbio(filp, argp);
 564                 break;
 565
 566         case FIOASYNC:
 567                 error = ioctl_fioasync(fd, filp, argp);
 568                 break;
 569
 570         case FIOQSIZE:
 571                 if (S_ISDIR(filp->f_path.dentry->d_inode->i_mode) ||
 572                     S_ISREG(filp->f_path.dentry->d_inode->i_mode) ||
 573                     S_ISLNK(filp->f_path.dentry->d_inode->i_mode)) {
 574                         loff_t res =
 575                                 inode_get_bytes(filp->f_path.dentry->d_inode);
 576                         error = copy_to_user((loff_t __user *)arg, &res,
 577                                              sizeof(res)) ? -EFAULT : 0;
 578                 } else
 579                         error = -ENOTTY;
 580                 break;
 581
 582         case FIFREEZE:
 583                 error = ioctl_fsfreeze(filp);
 584                 break;
 585
 586         case FITHAW:
 587                 error = ioctl_fsthaw(filp);
 588                 break;
 589
 590         case FS_IOC_FIEMAP:
 591                 return ioctl_fiemap(filp, arg);
 592
 593         case FIGETBSZ:
 594         {
 595                 struct inode *inode = filp->f_path.dentry->d_inode;
 596                 int __user *p = (int __user *)arg;
 597                 return put_user(inode->i_sb->s_blocksize, p);
 598         }
 599
 600         default:
 601                 if (S_ISREG(filp->f_path.dentry->d_inode->i_mode))
 602                         error = file_ioctl(filp, cmd, arg);
 603                 else
 604                         error = vfs_ioctl(filp, cmd, arg);
 605                 break;
 606         }
 607         return error;
 608 }
 609
 610 SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
 611 {
 612         struct file *filp;
 613         int error = -EBADF;
 614         int fput_needed;
 615
 616         filp = fget_light(fd, &fput_needed);
 617         if (!filp)
 618                 goto out;
 619
 620         error = security_file_ioctl(filp, cmd, arg);
 621         if (error)
 622                 goto out_fput;
 623
 624         error = do_vfs_ioctl(filp, fd, cmd, arg);
 625  out_fput:
 626         fput_light(filp, fput_needed);
 627  out:
 628         return error;
 629 }