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