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