xref: /openbmc/linux/fs/ioctl.c (revision 59b4412f)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/fs/ioctl.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  */
7 
8 #include <linux/syscalls.h>
9 #include <linux/mm.h>
10 #include <linux/capability.h>
11 #include <linux/compat.h>
12 #include <linux/file.h>
13 #include <linux/fs.h>
14 #include <linux/security.h>
15 #include <linux/export.h>
16 #include <linux/uaccess.h>
17 #include <linux/writeback.h>
18 #include <linux/buffer_head.h>
19 #include <linux/falloc.h>
20 #include <linux/sched/signal.h>
21 #include <linux/fiemap.h>
22 
23 #include "internal.h"
24 
25 #include <asm/ioctls.h>
26 
27 /* So that the fiemap access checks can't overflow on 32 bit machines. */
28 #define FIEMAP_MAX_EXTENTS	(UINT_MAX / sizeof(struct fiemap_extent))
29 
30 /**
31  * vfs_ioctl - call filesystem specific ioctl methods
32  * @filp:	open file to invoke ioctl method on
33  * @cmd:	ioctl command to execute
34  * @arg:	command-specific argument for ioctl
35  *
36  * Invokes filesystem specific ->unlocked_ioctl, if one exists; otherwise
37  * returns -ENOTTY.
38  *
39  * Returns 0 on success, -errno on error.
40  */
41 long vfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
42 {
43 	int error = -ENOTTY;
44 
45 	if (!filp->f_op->unlocked_ioctl)
46 		goto out;
47 
48 	error = filp->f_op->unlocked_ioctl(filp, cmd, arg);
49 	if (error == -ENOIOCTLCMD)
50 		error = -ENOTTY;
51  out:
52 	return error;
53 }
54 EXPORT_SYMBOL(vfs_ioctl);
55 
56 static int ioctl_fibmap(struct file *filp, int __user *p)
57 {
58 	struct inode *inode = file_inode(filp);
59 	struct super_block *sb = inode->i_sb;
60 	int error, ur_block;
61 	sector_t block;
62 
63 	if (!capable(CAP_SYS_RAWIO))
64 		return -EPERM;
65 
66 	error = get_user(ur_block, p);
67 	if (error)
68 		return error;
69 
70 	if (ur_block < 0)
71 		return -EINVAL;
72 
73 	block = ur_block;
74 	error = bmap(inode, &block);
75 
76 	if (block > INT_MAX) {
77 		error = -ERANGE;
78 		pr_warn_ratelimited("[%s/%d] FS: %s File: %pD4 would truncate fibmap result\n",
79 				    current->comm, task_pid_nr(current),
80 				    sb->s_id, filp);
81 	}
82 
83 	if (error)
84 		ur_block = 0;
85 	else
86 		ur_block = block;
87 
88 	if (put_user(ur_block, p))
89 		error = -EFAULT;
90 
91 	return error;
92 }
93 
94 /**
95  * fiemap_fill_next_extent - Fiemap helper function
96  * @fieinfo:	Fiemap context passed into ->fiemap
97  * @logical:	Extent logical start offset, in bytes
98  * @phys:	Extent physical start offset, in bytes
99  * @len:	Extent length, in bytes
100  * @flags:	FIEMAP_EXTENT flags that describe this extent
101  *
102  * Called from file system ->fiemap callback. Will populate extent
103  * info as passed in via arguments and copy to user memory. On
104  * success, extent count on fieinfo is incremented.
105  *
106  * Returns 0 on success, -errno on error, 1 if this was the last
107  * extent that will fit in user array.
108  */
109 #define SET_UNKNOWN_FLAGS	(FIEMAP_EXTENT_DELALLOC)
110 #define SET_NO_UNMOUNTED_IO_FLAGS	(FIEMAP_EXTENT_DATA_ENCRYPTED)
111 #define SET_NOT_ALIGNED_FLAGS	(FIEMAP_EXTENT_DATA_TAIL|FIEMAP_EXTENT_DATA_INLINE)
112 int fiemap_fill_next_extent(struct fiemap_extent_info *fieinfo, u64 logical,
113 			    u64 phys, u64 len, u32 flags)
114 {
115 	struct fiemap_extent extent;
116 	struct fiemap_extent __user *dest = fieinfo->fi_extents_start;
117 
118 	/* only count the extents */
119 	if (fieinfo->fi_extents_max == 0) {
120 		fieinfo->fi_extents_mapped++;
121 		return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
122 	}
123 
124 	if (fieinfo->fi_extents_mapped >= fieinfo->fi_extents_max)
125 		return 1;
126 
127 	if (flags & SET_UNKNOWN_FLAGS)
128 		flags |= FIEMAP_EXTENT_UNKNOWN;
129 	if (flags & SET_NO_UNMOUNTED_IO_FLAGS)
130 		flags |= FIEMAP_EXTENT_ENCODED;
131 	if (flags & SET_NOT_ALIGNED_FLAGS)
132 		flags |= FIEMAP_EXTENT_NOT_ALIGNED;
133 
134 	memset(&extent, 0, sizeof(extent));
135 	extent.fe_logical = logical;
136 	extent.fe_physical = phys;
137 	extent.fe_length = len;
138 	extent.fe_flags = flags;
139 
140 	dest += fieinfo->fi_extents_mapped;
141 	if (copy_to_user(dest, &extent, sizeof(extent)))
142 		return -EFAULT;
143 
144 	fieinfo->fi_extents_mapped++;
145 	if (fieinfo->fi_extents_mapped == fieinfo->fi_extents_max)
146 		return 1;
147 	return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
148 }
149 EXPORT_SYMBOL(fiemap_fill_next_extent);
150 
151 /**
152  * fiemap_prep - check validity of requested flags for fiemap
153  * @inode:	Inode to operate on
154  * @fieinfo:	Fiemap context passed into ->fiemap
155  * @start:	Start of the mapped range
156  * @len:	Length of the mapped range, can be truncated by this function.
157  * @supported_flags:	Set of fiemap flags that the file system understands
158  *
159  * This function must be called from each ->fiemap instance to validate the
160  * fiemap request against the file system parameters.
161  *
162  * Returns 0 on success, or a negative error on failure.
163  */
164 int fiemap_prep(struct inode *inode, struct fiemap_extent_info *fieinfo,
165 		u64 start, u64 *len, u32 supported_flags)
166 {
167 	u64 maxbytes = inode->i_sb->s_maxbytes;
168 	u32 incompat_flags;
169 	int ret = 0;
170 
171 	if (*len == 0)
172 		return -EINVAL;
173 	if (start > maxbytes)
174 		return -EFBIG;
175 
176 	/*
177 	 * Shrink request scope to what the fs can actually handle.
178 	 */
179 	if (*len > maxbytes || (maxbytes - *len) < start)
180 		*len = maxbytes - start;
181 
182 	supported_flags |= FIEMAP_FLAG_SYNC;
183 	supported_flags &= FIEMAP_FLAGS_COMPAT;
184 	incompat_flags = fieinfo->fi_flags & ~supported_flags;
185 	if (incompat_flags) {
186 		fieinfo->fi_flags = incompat_flags;
187 		return -EBADR;
188 	}
189 
190 	if (fieinfo->fi_flags & FIEMAP_FLAG_SYNC)
191 		ret = filemap_write_and_wait(inode->i_mapping);
192 	return ret;
193 }
194 EXPORT_SYMBOL(fiemap_prep);
195 
196 static int ioctl_fiemap(struct file *filp, struct fiemap __user *ufiemap)
197 {
198 	struct fiemap fiemap;
199 	struct fiemap_extent_info fieinfo = { 0, };
200 	struct inode *inode = file_inode(filp);
201 	int error;
202 
203 	if (!inode->i_op->fiemap)
204 		return -EOPNOTSUPP;
205 
206 	if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
207 		return -EFAULT;
208 
209 	if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
210 		return -EINVAL;
211 
212 	fieinfo.fi_flags = fiemap.fm_flags;
213 	fieinfo.fi_extents_max = fiemap.fm_extent_count;
214 	fieinfo.fi_extents_start = ufiemap->fm_extents;
215 
216 	error = inode->i_op->fiemap(inode, &fieinfo, fiemap.fm_start,
217 			fiemap.fm_length);
218 
219 	fiemap.fm_flags = fieinfo.fi_flags;
220 	fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
221 	if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
222 		error = -EFAULT;
223 
224 	return error;
225 }
226 
227 static long ioctl_file_clone(struct file *dst_file, unsigned long srcfd,
228 			     u64 off, u64 olen, u64 destoff)
229 {
230 	struct fd src_file = fdget(srcfd);
231 	loff_t cloned;
232 	int ret;
233 
234 	if (!src_file.file)
235 		return -EBADF;
236 	ret = -EXDEV;
237 	if (src_file.file->f_path.mnt != dst_file->f_path.mnt)
238 		goto fdput;
239 	cloned = vfs_clone_file_range(src_file.file, off, dst_file, destoff,
240 				      olen, 0);
241 	if (cloned < 0)
242 		ret = cloned;
243 	else if (olen && cloned != olen)
244 		ret = -EINVAL;
245 	else
246 		ret = 0;
247 fdput:
248 	fdput(src_file);
249 	return ret;
250 }
251 
252 static long ioctl_file_clone_range(struct file *file,
253 				   struct file_clone_range __user *argp)
254 {
255 	struct file_clone_range args;
256 
257 	if (copy_from_user(&args, argp, sizeof(args)))
258 		return -EFAULT;
259 	return ioctl_file_clone(file, args.src_fd, args.src_offset,
260 				args.src_length, args.dest_offset);
261 }
262 
263 #ifdef CONFIG_BLOCK
264 
265 static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
266 {
267 	return (offset >> inode->i_blkbits);
268 }
269 
270 static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
271 {
272 	return (blk << inode->i_blkbits);
273 }
274 
275 /**
276  * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
277  * @inode: the inode to map
278  * @fieinfo: the fiemap info struct that will be passed back to userspace
279  * @start: where to start mapping in the inode
280  * @len: how much space to map
281  * @get_block: the fs's get_block function
282  *
283  * This does FIEMAP for block based inodes.  Basically it will just loop
284  * through get_block until we hit the number of extents we want to map, or we
285  * go past the end of the file and hit a hole.
286  *
287  * If it is possible to have data blocks beyond a hole past @inode->i_size, then
288  * please do not use this function, it will stop at the first unmapped block
289  * beyond i_size.
290  *
291  * If you use this function directly, you need to do your own locking. Use
292  * generic_block_fiemap if you want the locking done for you.
293  */
294 static int __generic_block_fiemap(struct inode *inode,
295 			   struct fiemap_extent_info *fieinfo, loff_t start,
296 			   loff_t len, get_block_t *get_block)
297 {
298 	struct buffer_head map_bh;
299 	sector_t start_blk, last_blk;
300 	loff_t isize = i_size_read(inode);
301 	u64 logical = 0, phys = 0, size = 0;
302 	u32 flags = FIEMAP_EXTENT_MERGED;
303 	bool past_eof = false, whole_file = false;
304 	int ret = 0;
305 
306 	ret = fiemap_prep(inode, fieinfo, start, &len, FIEMAP_FLAG_SYNC);
307 	if (ret)
308 		return ret;
309 
310 	/*
311 	 * Either the i_mutex or other appropriate locking needs to be held
312 	 * since we expect isize to not change at all through the duration of
313 	 * this call.
314 	 */
315 	if (len >= isize) {
316 		whole_file = true;
317 		len = isize;
318 	}
319 
320 	/*
321 	 * Some filesystems can't deal with being asked to map less than
322 	 * blocksize, so make sure our len is at least block length.
323 	 */
324 	if (logical_to_blk(inode, len) == 0)
325 		len = blk_to_logical(inode, 1);
326 
327 	start_blk = logical_to_blk(inode, start);
328 	last_blk = logical_to_blk(inode, start + len - 1);
329 
330 	do {
331 		/*
332 		 * we set b_size to the total size we want so it will map as
333 		 * many contiguous blocks as possible at once
334 		 */
335 		memset(&map_bh, 0, sizeof(struct buffer_head));
336 		map_bh.b_size = len;
337 
338 		ret = get_block(inode, start_blk, &map_bh, 0);
339 		if (ret)
340 			break;
341 
342 		/* HOLE */
343 		if (!buffer_mapped(&map_bh)) {
344 			start_blk++;
345 
346 			/*
347 			 * We want to handle the case where there is an
348 			 * allocated block at the front of the file, and then
349 			 * nothing but holes up to the end of the file properly,
350 			 * to make sure that extent at the front gets properly
351 			 * marked with FIEMAP_EXTENT_LAST
352 			 */
353 			if (!past_eof &&
354 			    blk_to_logical(inode, start_blk) >= isize)
355 				past_eof = 1;
356 
357 			/*
358 			 * First hole after going past the EOF, this is our
359 			 * last extent
360 			 */
361 			if (past_eof && size) {
362 				flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
363 				ret = fiemap_fill_next_extent(fieinfo, logical,
364 							      phys, size,
365 							      flags);
366 			} else if (size) {
367 				ret = fiemap_fill_next_extent(fieinfo, logical,
368 							      phys, size, flags);
369 				size = 0;
370 			}
371 
372 			/* if we have holes up to/past EOF then we're done */
373 			if (start_blk > last_blk || past_eof || ret)
374 				break;
375 		} else {
376 			/*
377 			 * We have gone over the length of what we wanted to
378 			 * map, and it wasn't the entire file, so add the extent
379 			 * we got last time and exit.
380 			 *
381 			 * This is for the case where say we want to map all the
382 			 * way up to the second to the last block in a file, but
383 			 * the last block is a hole, making the second to last
384 			 * block FIEMAP_EXTENT_LAST.  In this case we want to
385 			 * see if there is a hole after the second to last block
386 			 * so we can mark it properly.  If we found data after
387 			 * we exceeded the length we were requesting, then we
388 			 * are good to go, just add the extent to the fieinfo
389 			 * and break
390 			 */
391 			if (start_blk > last_blk && !whole_file) {
392 				ret = fiemap_fill_next_extent(fieinfo, logical,
393 							      phys, size,
394 							      flags);
395 				break;
396 			}
397 
398 			/*
399 			 * if size != 0 then we know we already have an extent
400 			 * to add, so add it.
401 			 */
402 			if (size) {
403 				ret = fiemap_fill_next_extent(fieinfo, logical,
404 							      phys, size,
405 							      flags);
406 				if (ret)
407 					break;
408 			}
409 
410 			logical = blk_to_logical(inode, start_blk);
411 			phys = blk_to_logical(inode, map_bh.b_blocknr);
412 			size = map_bh.b_size;
413 			flags = FIEMAP_EXTENT_MERGED;
414 
415 			start_blk += logical_to_blk(inode, size);
416 
417 			/*
418 			 * If we are past the EOF, then we need to make sure as
419 			 * soon as we find a hole that the last extent we found
420 			 * is marked with FIEMAP_EXTENT_LAST
421 			 */
422 			if (!past_eof && logical + size >= isize)
423 				past_eof = true;
424 		}
425 		cond_resched();
426 		if (fatal_signal_pending(current)) {
427 			ret = -EINTR;
428 			break;
429 		}
430 
431 	} while (1);
432 
433 	/* If ret is 1 then we just hit the end of the extent array */
434 	if (ret == 1)
435 		ret = 0;
436 
437 	return ret;
438 }
439 
440 /**
441  * generic_block_fiemap - FIEMAP for block based inodes
442  * @inode: The inode to map
443  * @fieinfo: The mapping information
444  * @start: The initial block to map
445  * @len: The length of the extect to attempt to map
446  * @get_block: The block mapping function for the fs
447  *
448  * Calls __generic_block_fiemap to map the inode, after taking
449  * the inode's mutex lock.
450  */
451 
452 int generic_block_fiemap(struct inode *inode,
453 			 struct fiemap_extent_info *fieinfo, u64 start,
454 			 u64 len, get_block_t *get_block)
455 {
456 	int ret;
457 	inode_lock(inode);
458 	ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
459 	inode_unlock(inode);
460 	return ret;
461 }
462 EXPORT_SYMBOL(generic_block_fiemap);
463 
464 #endif  /*  CONFIG_BLOCK  */
465 
466 /*
467  * This provides compatibility with legacy XFS pre-allocation ioctls
468  * which predate the fallocate syscall.
469  *
470  * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
471  * are used here, rest are ignored.
472  */
473 static int ioctl_preallocate(struct file *filp, int mode, void __user *argp)
474 {
475 	struct inode *inode = file_inode(filp);
476 	struct space_resv sr;
477 
478 	if (copy_from_user(&sr, argp, sizeof(sr)))
479 		return -EFAULT;
480 
481 	switch (sr.l_whence) {
482 	case SEEK_SET:
483 		break;
484 	case SEEK_CUR:
485 		sr.l_start += filp->f_pos;
486 		break;
487 	case SEEK_END:
488 		sr.l_start += i_size_read(inode);
489 		break;
490 	default:
491 		return -EINVAL;
492 	}
493 
494 	return vfs_fallocate(filp, mode | FALLOC_FL_KEEP_SIZE, sr.l_start,
495 			sr.l_len);
496 }
497 
498 /* on ia32 l_start is on a 32-bit boundary */
499 #if defined CONFIG_COMPAT && defined(CONFIG_X86_64)
500 /* just account for different alignment */
501 static int compat_ioctl_preallocate(struct file *file, int mode,
502 				    struct space_resv_32 __user *argp)
503 {
504 	struct inode *inode = file_inode(file);
505 	struct space_resv_32 sr;
506 
507 	if (copy_from_user(&sr, argp, sizeof(sr)))
508 		return -EFAULT;
509 
510 	switch (sr.l_whence) {
511 	case SEEK_SET:
512 		break;
513 	case SEEK_CUR:
514 		sr.l_start += file->f_pos;
515 		break;
516 	case SEEK_END:
517 		sr.l_start += i_size_read(inode);
518 		break;
519 	default:
520 		return -EINVAL;
521 	}
522 
523 	return vfs_fallocate(file, mode | FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
524 }
525 #endif
526 
527 static int file_ioctl(struct file *filp, unsigned int cmd, int __user *p)
528 {
529 	switch (cmd) {
530 	case FIBMAP:
531 		return ioctl_fibmap(filp, p);
532 	case FS_IOC_RESVSP:
533 	case FS_IOC_RESVSP64:
534 		return ioctl_preallocate(filp, 0, p);
535 	case FS_IOC_UNRESVSP:
536 	case FS_IOC_UNRESVSP64:
537 		return ioctl_preallocate(filp, FALLOC_FL_PUNCH_HOLE, p);
538 	case FS_IOC_ZERO_RANGE:
539 		return ioctl_preallocate(filp, FALLOC_FL_ZERO_RANGE, p);
540 	}
541 
542 	return -ENOIOCTLCMD;
543 }
544 
545 static int ioctl_fionbio(struct file *filp, int __user *argp)
546 {
547 	unsigned int flag;
548 	int on, error;
549 
550 	error = get_user(on, argp);
551 	if (error)
552 		return error;
553 	flag = O_NONBLOCK;
554 #ifdef __sparc__
555 	/* SunOS compatibility item. */
556 	if (O_NONBLOCK != O_NDELAY)
557 		flag |= O_NDELAY;
558 #endif
559 	spin_lock(&filp->f_lock);
560 	if (on)
561 		filp->f_flags |= flag;
562 	else
563 		filp->f_flags &= ~flag;
564 	spin_unlock(&filp->f_lock);
565 	return error;
566 }
567 
568 static int ioctl_fioasync(unsigned int fd, struct file *filp,
569 			  int __user *argp)
570 {
571 	unsigned int flag;
572 	int on, error;
573 
574 	error = get_user(on, argp);
575 	if (error)
576 		return error;
577 	flag = on ? FASYNC : 0;
578 
579 	/* Did FASYNC state change ? */
580 	if ((flag ^ filp->f_flags) & FASYNC) {
581 		if (filp->f_op->fasync)
582 			/* fasync() adjusts filp->f_flags */
583 			error = filp->f_op->fasync(fd, filp, on);
584 		else
585 			error = -ENOTTY;
586 	}
587 	return error < 0 ? error : 0;
588 }
589 
590 static int ioctl_fsfreeze(struct file *filp)
591 {
592 	struct super_block *sb = file_inode(filp)->i_sb;
593 
594 	if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN))
595 		return -EPERM;
596 
597 	/* If filesystem doesn't support freeze feature, return. */
598 	if (sb->s_op->freeze_fs == NULL && sb->s_op->freeze_super == NULL)
599 		return -EOPNOTSUPP;
600 
601 	/* Freeze */
602 	if (sb->s_op->freeze_super)
603 		return sb->s_op->freeze_super(sb);
604 	return freeze_super(sb);
605 }
606 
607 static int ioctl_fsthaw(struct file *filp)
608 {
609 	struct super_block *sb = file_inode(filp)->i_sb;
610 
611 	if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN))
612 		return -EPERM;
613 
614 	/* Thaw */
615 	if (sb->s_op->thaw_super)
616 		return sb->s_op->thaw_super(sb);
617 	return thaw_super(sb);
618 }
619 
620 static int ioctl_file_dedupe_range(struct file *file,
621 				   struct file_dedupe_range __user *argp)
622 {
623 	struct file_dedupe_range *same = NULL;
624 	int ret;
625 	unsigned long size;
626 	u16 count;
627 
628 	if (get_user(count, &argp->dest_count)) {
629 		ret = -EFAULT;
630 		goto out;
631 	}
632 
633 	size = offsetof(struct file_dedupe_range __user, info[count]);
634 	if (size > PAGE_SIZE) {
635 		ret = -ENOMEM;
636 		goto out;
637 	}
638 
639 	same = memdup_user(argp, size);
640 	if (IS_ERR(same)) {
641 		ret = PTR_ERR(same);
642 		same = NULL;
643 		goto out;
644 	}
645 
646 	same->dest_count = count;
647 	ret = vfs_dedupe_file_range(file, same);
648 	if (ret)
649 		goto out;
650 
651 	ret = copy_to_user(argp, same, size);
652 	if (ret)
653 		ret = -EFAULT;
654 
655 out:
656 	kfree(same);
657 	return ret;
658 }
659 
660 /*
661  * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
662  * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
663  *
664  * When you add any new common ioctls to the switches above and below,
665  * please ensure they have compatible arguments in compat mode.
666  */
667 static int do_vfs_ioctl(struct file *filp, unsigned int fd,
668 			unsigned int cmd, unsigned long arg)
669 {
670 	void __user *argp = (void __user *)arg;
671 	struct inode *inode = file_inode(filp);
672 
673 	switch (cmd) {
674 	case FIOCLEX:
675 		set_close_on_exec(fd, 1);
676 		return 0;
677 
678 	case FIONCLEX:
679 		set_close_on_exec(fd, 0);
680 		return 0;
681 
682 	case FIONBIO:
683 		return ioctl_fionbio(filp, argp);
684 
685 	case FIOASYNC:
686 		return ioctl_fioasync(fd, filp, argp);
687 
688 	case FIOQSIZE:
689 		if (S_ISDIR(inode->i_mode) || S_ISREG(inode->i_mode) ||
690 		    S_ISLNK(inode->i_mode)) {
691 			loff_t res = inode_get_bytes(inode);
692 			return copy_to_user(argp, &res, sizeof(res)) ?
693 					    -EFAULT : 0;
694 		}
695 
696 		return -ENOTTY;
697 
698 	case FIFREEZE:
699 		return ioctl_fsfreeze(filp);
700 
701 	case FITHAW:
702 		return ioctl_fsthaw(filp);
703 
704 	case FS_IOC_FIEMAP:
705 		return ioctl_fiemap(filp, argp);
706 
707 	case FIGETBSZ:
708 		/* anon_bdev filesystems may not have a block size */
709 		if (!inode->i_sb->s_blocksize)
710 			return -EINVAL;
711 
712 		return put_user(inode->i_sb->s_blocksize, (int __user *)argp);
713 
714 	case FICLONE:
715 		return ioctl_file_clone(filp, arg, 0, 0, 0);
716 
717 	case FICLONERANGE:
718 		return ioctl_file_clone_range(filp, argp);
719 
720 	case FIDEDUPERANGE:
721 		return ioctl_file_dedupe_range(filp, argp);
722 
723 	case FIONREAD:
724 		if (!S_ISREG(inode->i_mode))
725 			return vfs_ioctl(filp, cmd, arg);
726 
727 		return put_user(i_size_read(inode) - filp->f_pos,
728 				(int __user *)argp);
729 
730 	default:
731 		if (S_ISREG(inode->i_mode))
732 			return file_ioctl(filp, cmd, argp);
733 		break;
734 	}
735 
736 	return -ENOIOCTLCMD;
737 }
738 
739 int ksys_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
740 {
741 	struct fd f = fdget(fd);
742 	int error;
743 
744 	if (!f.file)
745 		return -EBADF;
746 
747 	error = security_file_ioctl(f.file, cmd, arg);
748 	if (error)
749 		goto out;
750 
751 	error = do_vfs_ioctl(f.file, fd, cmd, arg);
752 	if (error == -ENOIOCTLCMD)
753 		error = vfs_ioctl(f.file, cmd, arg);
754 
755 out:
756 	fdput(f);
757 	return error;
758 }
759 
760 SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
761 {
762 	return ksys_ioctl(fd, cmd, arg);
763 }
764 
765 #ifdef CONFIG_COMPAT
766 /**
767  * compat_ptr_ioctl - generic implementation of .compat_ioctl file operation
768  *
769  * This is not normally called as a function, but instead set in struct
770  * file_operations as
771  *
772  *     .compat_ioctl = compat_ptr_ioctl,
773  *
774  * On most architectures, the compat_ptr_ioctl() just passes all arguments
775  * to the corresponding ->ioctl handler. The exception is arch/s390, where
776  * compat_ptr() clears the top bit of a 32-bit pointer value, so user space
777  * pointers to the second 2GB alias the first 2GB, as is the case for
778  * native 32-bit s390 user space.
779  *
780  * The compat_ptr_ioctl() function must therefore be used only with ioctl
781  * functions that either ignore the argument or pass a pointer to a
782  * compatible data type.
783  *
784  * If any ioctl command handled by fops->unlocked_ioctl passes a plain
785  * integer instead of a pointer, or any of the passed data types
786  * is incompatible between 32-bit and 64-bit architectures, a proper
787  * handler is required instead of compat_ptr_ioctl.
788  */
789 long compat_ptr_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
790 {
791 	if (!file->f_op->unlocked_ioctl)
792 		return -ENOIOCTLCMD;
793 
794 	return file->f_op->unlocked_ioctl(file, cmd, (unsigned long)compat_ptr(arg));
795 }
796 EXPORT_SYMBOL(compat_ptr_ioctl);
797 
798 COMPAT_SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd,
799 		       compat_ulong_t, arg)
800 {
801 	struct fd f = fdget(fd);
802 	int error;
803 
804 	if (!f.file)
805 		return -EBADF;
806 
807 	/* RED-PEN how should LSM module know it's handling 32bit? */
808 	error = security_file_ioctl(f.file, cmd, arg);
809 	if (error)
810 		goto out;
811 
812 	switch (cmd) {
813 	/* FICLONE takes an int argument, so don't use compat_ptr() */
814 	case FICLONE:
815 		error = ioctl_file_clone(f.file, arg, 0, 0, 0);
816 		break;
817 
818 #if defined(CONFIG_X86_64)
819 	/* these get messy on amd64 due to alignment differences */
820 	case FS_IOC_RESVSP_32:
821 	case FS_IOC_RESVSP64_32:
822 		error = compat_ioctl_preallocate(f.file, 0, compat_ptr(arg));
823 		break;
824 	case FS_IOC_UNRESVSP_32:
825 	case FS_IOC_UNRESVSP64_32:
826 		error = compat_ioctl_preallocate(f.file, FALLOC_FL_PUNCH_HOLE,
827 				compat_ptr(arg));
828 		break;
829 	case FS_IOC_ZERO_RANGE_32:
830 		error = compat_ioctl_preallocate(f.file, FALLOC_FL_ZERO_RANGE,
831 				compat_ptr(arg));
832 		break;
833 #endif
834 
835 	/*
836 	 * everything else in do_vfs_ioctl() takes either a compatible
837 	 * pointer argument or no argument -- call it with a modified
838 	 * argument.
839 	 */
840 	default:
841 		error = do_vfs_ioctl(f.file, fd, cmd,
842 				     (unsigned long)compat_ptr(arg));
843 		if (error != -ENOIOCTLCMD)
844 			break;
845 
846 		if (f.file->f_op->compat_ioctl)
847 			error = f.file->f_op->compat_ioctl(f.file, cmd, arg);
848 		if (error == -ENOIOCTLCMD)
849 			error = -ENOTTY;
850 		break;
851 	}
852 
853  out:
854 	fdput(f);
855 
856 	return error;
857 }
858 #endif
859