ioctl.c 16 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624
  1. /*
  2. * linux/fs/ioctl.c
  3. *
  4. * Copyright (C) 1991, 1992 Linus Torvalds
  5. */
  6. #include <linux/syscalls.h>
  7. #include <linux/mm.h>
  8. #include <linux/capability.h>
  9. #include <linux/file.h>
  10. #include <linux/fs.h>
  11. #include <linux/security.h>
  12. #include <linux/export.h>
  13. #include <linux/uaccess.h>
  14. #include <linux/writeback.h>
  15. #include <linux/buffer_head.h>
  16. #include <linux/falloc.h>
  17. #include <asm/ioctls.h>
  18. /* So that the fiemap access checks can't overflow on 32 bit machines. */
  19. #define FIEMAP_MAX_EXTENTS (UINT_MAX / sizeof(struct fiemap_extent))
  20. /**
  21. * vfs_ioctl - call filesystem specific ioctl methods
  22. * @filp: open file to invoke ioctl method on
  23. * @cmd: ioctl command to execute
  24. * @arg: command-specific argument for ioctl
  25. *
  26. * Invokes filesystem specific ->unlocked_ioctl, if one exists; otherwise
  27. * returns -ENOTTY.
  28. *
  29. * Returns 0 on success, -errno on error.
  30. */
  31. static long vfs_ioctl(struct file *filp, unsigned int cmd,
  32. unsigned long arg)
  33. {
  34. int error = -ENOTTY;
  35. if (!filp->f_op || !filp->f_op->unlocked_ioctl)
  36. goto out;
  37. error = filp->f_op->unlocked_ioctl(filp, cmd, arg);
  38. if (error == -ENOIOCTLCMD)
  39. error = -ENOTTY;
  40. out:
  41. return error;
  42. }
  43. static int ioctl_fibmap(struct file *filp, int __user *p)
  44. {
  45. struct address_space *mapping = filp->f_mapping;
  46. int res, block;
  47. /* do we support this mess? */
  48. if (!mapping->a_ops->bmap)
  49. return -EINVAL;
  50. if (!capable(CAP_SYS_RAWIO))
  51. return -EPERM;
  52. res = get_user(block, p);
  53. if (res)
  54. return res;
  55. res = mapping->a_ops->bmap(mapping, block);
  56. return put_user(res, p);
  57. }
  58. /**
  59. * fiemap_fill_next_extent - Fiemap helper function
  60. * @fieinfo: Fiemap context passed into ->fiemap
  61. * @logical: Extent logical start offset, in bytes
  62. * @phys: Extent physical start offset, in bytes
  63. * @len: Extent length, in bytes
  64. * @flags: FIEMAP_EXTENT flags that describe this extent
  65. *
  66. * Called from file system ->fiemap callback. Will populate extent
  67. * info as passed in via arguments and copy to user memory. On
  68. * success, extent count on fieinfo is incremented.
  69. *
  70. * Returns 0 on success, -errno on error, 1 if this was the last
  71. * extent that will fit in user array.
  72. */
  73. #define SET_UNKNOWN_FLAGS (FIEMAP_EXTENT_DELALLOC)
  74. #define SET_NO_UNMOUNTED_IO_FLAGS (FIEMAP_EXTENT_DATA_ENCRYPTED)
  75. #define SET_NOT_ALIGNED_FLAGS (FIEMAP_EXTENT_DATA_TAIL|FIEMAP_EXTENT_DATA_INLINE)
  76. int fiemap_fill_next_extent(struct fiemap_extent_info *fieinfo, u64 logical,
  77. u64 phys, u64 len, u32 flags)
  78. {
  79. struct fiemap_extent extent;
  80. struct fiemap_extent __user *dest = fieinfo->fi_extents_start;
  81. /* only count the extents */
  82. if (fieinfo->fi_extents_max == 0) {
  83. fieinfo->fi_extents_mapped++;
  84. return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
  85. }
  86. if (fieinfo->fi_extents_mapped >= fieinfo->fi_extents_max)
  87. return 1;
  88. if (flags & SET_UNKNOWN_FLAGS)
  89. flags |= FIEMAP_EXTENT_UNKNOWN;
  90. if (flags & SET_NO_UNMOUNTED_IO_FLAGS)
  91. flags |= FIEMAP_EXTENT_ENCODED;
  92. if (flags & SET_NOT_ALIGNED_FLAGS)
  93. flags |= FIEMAP_EXTENT_NOT_ALIGNED;
  94. memset(&extent, 0, sizeof(extent));
  95. extent.fe_logical = logical;
  96. extent.fe_physical = phys;
  97. extent.fe_length = len;
  98. extent.fe_flags = flags;
  99. dest += fieinfo->fi_extents_mapped;
  100. if (copy_to_user(dest, &extent, sizeof(extent)))
  101. return -EFAULT;
  102. fieinfo->fi_extents_mapped++;
  103. if (fieinfo->fi_extents_mapped == fieinfo->fi_extents_max)
  104. return 1;
  105. return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
  106. }
  107. EXPORT_SYMBOL(fiemap_fill_next_extent);
  108. /**
  109. * fiemap_check_flags - check validity of requested flags for fiemap
  110. * @fieinfo: Fiemap context passed into ->fiemap
  111. * @fs_flags: Set of fiemap flags that the file system understands
  112. *
  113. * Called from file system ->fiemap callback. This will compute the
  114. * intersection of valid fiemap flags and those that the fs supports. That
  115. * value is then compared against the user supplied flags. In case of bad user
  116. * flags, the invalid values will be written into the fieinfo structure, and
  117. * -EBADR is returned, which tells ioctl_fiemap() to return those values to
  118. * userspace. For this reason, a return code of -EBADR should be preserved.
  119. *
  120. * Returns 0 on success, -EBADR on bad flags.
  121. */
  122. int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags)
  123. {
  124. u32 incompat_flags;
  125. incompat_flags = fieinfo->fi_flags & ~(FIEMAP_FLAGS_COMPAT & fs_flags);
  126. if (incompat_flags) {
  127. fieinfo->fi_flags = incompat_flags;
  128. return -EBADR;
  129. }
  130. return 0;
  131. }
  132. EXPORT_SYMBOL(fiemap_check_flags);
  133. static int fiemap_check_ranges(struct super_block *sb,
  134. u64 start, u64 len, u64 *new_len)
  135. {
  136. u64 maxbytes = (u64) sb->s_maxbytes;
  137. *new_len = len;
  138. if (len == 0)
  139. return -EINVAL;
  140. if (start > maxbytes)
  141. return -EFBIG;
  142. /*
  143. * Shrink request scope to what the fs can actually handle.
  144. */
  145. if (len > maxbytes || (maxbytes - len) < start)
  146. *new_len = maxbytes - start;
  147. return 0;
  148. }
  149. static int ioctl_fiemap(struct file *filp, unsigned long arg)
  150. {
  151. struct fiemap fiemap;
  152. struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
  153. struct fiemap_extent_info fieinfo = { 0, };
  154. struct inode *inode = filp->f_path.dentry->d_inode;
  155. struct super_block *sb = inode->i_sb;
  156. u64 len;
  157. int error;
  158. if (!inode->i_op->fiemap)
  159. return -EOPNOTSUPP;
  160. if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
  161. return -EFAULT;
  162. if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
  163. return -EINVAL;
  164. error = fiemap_check_ranges(sb, fiemap.fm_start, fiemap.fm_length,
  165. &len);
  166. if (error)
  167. return error;
  168. fieinfo.fi_flags = fiemap.fm_flags;
  169. fieinfo.fi_extents_max = fiemap.fm_extent_count;
  170. fieinfo.fi_extents_start = ufiemap->fm_extents;
  171. if (fiemap.fm_extent_count != 0 &&
  172. !access_ok(VERIFY_WRITE, fieinfo.fi_extents_start,
  173. fieinfo.fi_extents_max * sizeof(struct fiemap_extent)))
  174. return -EFAULT;
  175. if (fieinfo.fi_flags & FIEMAP_FLAG_SYNC)
  176. filemap_write_and_wait(inode->i_mapping);
  177. error = inode->i_op->fiemap(inode, &fieinfo, fiemap.fm_start, len);
  178. fiemap.fm_flags = fieinfo.fi_flags;
  179. fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
  180. if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
  181. error = -EFAULT;
  182. return error;
  183. }
  184. #ifdef CONFIG_BLOCK
  185. static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
  186. {
  187. return (offset >> inode->i_blkbits);
  188. }
  189. static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
  190. {
  191. return (blk << inode->i_blkbits);
  192. }
  193. /**
  194. * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
  195. * @inode: the inode to map
  196. * @fieinfo: the fiemap info struct that will be passed back to userspace
  197. * @start: where to start mapping in the inode
  198. * @len: how much space to map
  199. * @get_block: the fs's get_block function
  200. *
  201. * This does FIEMAP for block based inodes. Basically it will just loop
  202. * through get_block until we hit the number of extents we want to map, or we
  203. * go past the end of the file and hit a hole.
  204. *
  205. * If it is possible to have data blocks beyond a hole past @inode->i_size, then
  206. * please do not use this function, it will stop at the first unmapped block
  207. * beyond i_size.
  208. *
  209. * If you use this function directly, you need to do your own locking. Use
  210. * generic_block_fiemap if you want the locking done for you.
  211. */
  212. int __generic_block_fiemap(struct inode *inode,
  213. struct fiemap_extent_info *fieinfo, loff_t start,
  214. loff_t len, get_block_t *get_block)
  215. {
  216. struct buffer_head map_bh;
  217. sector_t start_blk, last_blk;
  218. loff_t isize = i_size_read(inode);
  219. u64 logical = 0, phys = 0, size = 0;
  220. u32 flags = FIEMAP_EXTENT_MERGED;
  221. bool past_eof = false, whole_file = false;
  222. int ret = 0;
  223. ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC);
  224. if (ret)
  225. return ret;
  226. /*
  227. * Either the i_mutex or other appropriate locking needs to be held
  228. * since we expect isize to not change at all through the duration of
  229. * this call.
  230. */
  231. if (len >= isize) {
  232. whole_file = true;
  233. len = isize;
  234. }
  235. /*
  236. * Some filesystems can't deal with being asked to map less than
  237. * blocksize, so make sure our len is at least block length.
  238. */
  239. if (logical_to_blk(inode, len) == 0)
  240. len = blk_to_logical(inode, 1);
  241. start_blk = logical_to_blk(inode, start);
  242. last_blk = logical_to_blk(inode, start + len - 1);
  243. do {
  244. /*
  245. * we set b_size to the total size we want so it will map as
  246. * many contiguous blocks as possible at once
  247. */
  248. memset(&map_bh, 0, sizeof(struct buffer_head));
  249. map_bh.b_size = len;
  250. ret = get_block(inode, start_blk, &map_bh, 0);
  251. if (ret)
  252. break;
  253. /* HOLE */
  254. if (!buffer_mapped(&map_bh)) {
  255. start_blk++;
  256. /*
  257. * We want to handle the case where there is an
  258. * allocated block at the front of the file, and then
  259. * nothing but holes up to the end of the file properly,
  260. * to make sure that extent at the front gets properly
  261. * marked with FIEMAP_EXTENT_LAST
  262. */
  263. if (!past_eof &&
  264. blk_to_logical(inode, start_blk) >= isize)
  265. past_eof = 1;
  266. /*
  267. * First hole after going past the EOF, this is our
  268. * last extent
  269. */
  270. if (past_eof && size) {
  271. flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
  272. ret = fiemap_fill_next_extent(fieinfo, logical,
  273. phys, size,
  274. flags);
  275. } else if (size) {
  276. ret = fiemap_fill_next_extent(fieinfo, logical,
  277. phys, size, flags);
  278. size = 0;
  279. }
  280. /* if we have holes up to/past EOF then we're done */
  281. if (start_blk > last_blk || past_eof || ret)
  282. break;
  283. } else {
  284. /*
  285. * We have gone over the length of what we wanted to
  286. * map, and it wasn't the entire file, so add the extent
  287. * we got last time and exit.
  288. *
  289. * This is for the case where say we want to map all the
  290. * way up to the second to the last block in a file, but
  291. * the last block is a hole, making the second to last
  292. * block FIEMAP_EXTENT_LAST. In this case we want to
  293. * see if there is a hole after the second to last block
  294. * so we can mark it properly. If we found data after
  295. * we exceeded the length we were requesting, then we
  296. * are good to go, just add the extent to the fieinfo
  297. * and break
  298. */
  299. if (start_blk > last_blk && !whole_file) {
  300. ret = fiemap_fill_next_extent(fieinfo, logical,
  301. phys, size,
  302. flags);
  303. break;
  304. }
  305. /*
  306. * if size != 0 then we know we already have an extent
  307. * to add, so add it.
  308. */
  309. if (size) {
  310. ret = fiemap_fill_next_extent(fieinfo, logical,
  311. phys, size,
  312. flags);
  313. if (ret)
  314. break;
  315. }
  316. logical = blk_to_logical(inode, start_blk);
  317. phys = blk_to_logical(inode, map_bh.b_blocknr);
  318. size = map_bh.b_size;
  319. flags = FIEMAP_EXTENT_MERGED;
  320. start_blk += logical_to_blk(inode, size);
  321. /*
  322. * If we are past the EOF, then we need to make sure as
  323. * soon as we find a hole that the last extent we found
  324. * is marked with FIEMAP_EXTENT_LAST
  325. */
  326. if (!past_eof && logical + size >= isize)
  327. past_eof = true;
  328. }
  329. cond_resched();
  330. } while (1);
  331. /* If ret is 1 then we just hit the end of the extent array */
  332. if (ret == 1)
  333. ret = 0;
  334. return ret;
  335. }
  336. EXPORT_SYMBOL(__generic_block_fiemap);
  337. /**
  338. * generic_block_fiemap - FIEMAP for block based inodes
  339. * @inode: The inode to map
  340. * @fieinfo: The mapping information
  341. * @start: The initial block to map
  342. * @len: The length of the extect to attempt to map
  343. * @get_block: The block mapping function for the fs
  344. *
  345. * Calls __generic_block_fiemap to map the inode, after taking
  346. * the inode's mutex lock.
  347. */
  348. int generic_block_fiemap(struct inode *inode,
  349. struct fiemap_extent_info *fieinfo, u64 start,
  350. u64 len, get_block_t *get_block)
  351. {
  352. int ret;
  353. mutex_lock(&inode->i_mutex);
  354. ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
  355. mutex_unlock(&inode->i_mutex);
  356. return ret;
  357. }
  358. EXPORT_SYMBOL(generic_block_fiemap);
  359. #endif /* CONFIG_BLOCK */
  360. /*
  361. * This provides compatibility with legacy XFS pre-allocation ioctls
  362. * which predate the fallocate syscall.
  363. *
  364. * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
  365. * are used here, rest are ignored.
  366. */
  367. int ioctl_preallocate(struct file *filp, void __user *argp)
  368. {
  369. struct inode *inode = filp->f_path.dentry->d_inode;
  370. struct space_resv sr;
  371. if (copy_from_user(&sr, argp, sizeof(sr)))
  372. return -EFAULT;
  373. switch (sr.l_whence) {
  374. case SEEK_SET:
  375. break;
  376. case SEEK_CUR:
  377. sr.l_start += filp->f_pos;
  378. break;
  379. case SEEK_END:
  380. sr.l_start += i_size_read(inode);
  381. break;
  382. default:
  383. return -EINVAL;
  384. }
  385. return do_fallocate(filp, FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
  386. }
  387. static int file_ioctl(struct file *filp, unsigned int cmd,
  388. unsigned long arg)
  389. {
  390. struct inode *inode = filp->f_path.dentry->d_inode;
  391. int __user *p = (int __user *)arg;
  392. switch (cmd) {
  393. case FIBMAP:
  394. return ioctl_fibmap(filp, p);
  395. case FIONREAD:
  396. return put_user(i_size_read(inode) - filp->f_pos, p);
  397. case FS_IOC_RESVSP:
  398. case FS_IOC_RESVSP64:
  399. return ioctl_preallocate(filp, p);
  400. }
  401. return vfs_ioctl(filp, cmd, arg);
  402. }
  403. static int ioctl_fionbio(struct file *filp, int __user *argp)
  404. {
  405. unsigned int flag;
  406. int on, error;
  407. error = get_user(on, argp);
  408. if (error)
  409. return error;
  410. flag = O_NONBLOCK;
  411. #ifdef __sparc__
  412. /* SunOS compatibility item. */
  413. if (O_NONBLOCK != O_NDELAY)
  414. flag |= O_NDELAY;
  415. #endif
  416. spin_lock(&filp->f_lock);
  417. if (on)
  418. filp->f_flags |= flag;
  419. else
  420. filp->f_flags &= ~flag;
  421. spin_unlock(&filp->f_lock);
  422. return error;
  423. }
  424. static int ioctl_fioasync(unsigned int fd, struct file *filp,
  425. int __user *argp)
  426. {
  427. unsigned int flag;
  428. int on, error;
  429. error = get_user(on, argp);
  430. if (error)
  431. return error;
  432. flag = on ? FASYNC : 0;
  433. /* Did FASYNC state change ? */
  434. if ((flag ^ filp->f_flags) & FASYNC) {
  435. if (filp->f_op && filp->f_op->fasync)
  436. /* fasync() adjusts filp->f_flags */
  437. error = filp->f_op->fasync(fd, filp, on);
  438. else
  439. error = -ENOTTY;
  440. }
  441. return error < 0 ? error : 0;
  442. }
  443. static int ioctl_fsfreeze(struct file *filp)
  444. {
  445. struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
  446. if (!capable(CAP_SYS_ADMIN))
  447. return -EPERM;
  448. /* If filesystem doesn't support freeze feature, return. */
  449. if (sb->s_op->freeze_fs == NULL)
  450. return -EOPNOTSUPP;
  451. /* Freeze */
  452. return freeze_super(sb);
  453. }
  454. static int ioctl_fsthaw(struct file *filp)
  455. {
  456. struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
  457. if (!capable(CAP_SYS_ADMIN))
  458. return -EPERM;
  459. /* Thaw */
  460. return thaw_super(sb);
  461. }
  462. /*
  463. * When you add any new common ioctls to the switches above and below
  464. * please update compat_sys_ioctl() too.
  465. *
  466. * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
  467. * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
  468. */
  469. int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
  470. unsigned long arg)
  471. {
  472. int error = 0;
  473. int __user *argp = (int __user *)arg;
  474. struct inode *inode = filp->f_path.dentry->d_inode;
  475. switch (cmd) {
  476. case FIOCLEX:
  477. set_close_on_exec(fd, 1);
  478. break;
  479. case FIONCLEX:
  480. set_close_on_exec(fd, 0);
  481. break;
  482. case FIONBIO:
  483. error = ioctl_fionbio(filp, argp);
  484. break;
  485. case FIOASYNC:
  486. error = ioctl_fioasync(fd, filp, argp);
  487. break;
  488. case FIOQSIZE:
  489. if (S_ISDIR(inode->i_mode) || S_ISREG(inode->i_mode) ||
  490. S_ISLNK(inode->i_mode)) {
  491. loff_t res = inode_get_bytes(inode);
  492. error = copy_to_user(argp, &res, sizeof(res)) ?
  493. -EFAULT : 0;
  494. } else
  495. error = -ENOTTY;
  496. break;
  497. case FIFREEZE:
  498. error = ioctl_fsfreeze(filp);
  499. break;
  500. case FITHAW:
  501. error = ioctl_fsthaw(filp);
  502. break;
  503. case FS_IOC_FIEMAP:
  504. return ioctl_fiemap(filp, arg);
  505. case FIGETBSZ:
  506. return put_user(inode->i_sb->s_blocksize, argp);
  507. default:
  508. if (S_ISREG(inode->i_mode))
  509. error = file_ioctl(filp, cmd, arg);
  510. else
  511. error = vfs_ioctl(filp, cmd, arg);
  512. break;
  513. }
  514. return error;
  515. }
  516. SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
  517. {
  518. struct file *filp;
  519. int error = -EBADF;
  520. int fput_needed;
  521. filp = fget_light(fd, &fput_needed);
  522. if (!filp)
  523. goto out;
  524. error = security_file_ioctl(filp, cmd, arg);
  525. if (error)
  526. goto out_fput;
  527. error = do_vfs_ioctl(filp, fd, cmd, arg);
  528. out_fput:
  529. fput_light(filp, fput_needed);
  530. out:
  531. return error;
  532. }