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/module.h>
14#include <linux/uaccess.h>
15#include <linux/writeback.h>
16#include <linux/buffer_head.h>
17#include <linux/falloc.h>
18
19#include <asm/ioctls.h>
20
21/* So that the fiemap access checks can't overflow on 32 bit machines. */
22#define FIEMAP_MAX_EXTENTS (UINT_MAX / sizeof(struct fiemap_extent))
23
24/**
25 * vfs_ioctl - call filesystem specific ioctl methods
26 * @filp: open file to invoke ioctl method on
27 * @cmd: ioctl command to execute
28 * @arg: command-specific argument for ioctl
29 *
30 * Invokes filesystem specific ->unlocked_ioctl, if one exists; otherwise
31 * returns -ENOTTY.
32 *
33 * Returns 0 on success, -errno on error.
34 */
35static long vfs_ioctl(struct file *filp, unsigned int cmd,
36 unsigned long arg)
37{
38 int error = -ENOTTY;
39
40 if (!filp->f_op || !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 = -EINVAL;
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 = filp->f_path.dentry->d_inode;
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
218#ifdef CONFIG_BLOCK
219
220static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
221{
222 return (offset >> inode->i_blkbits);
223}
224
225static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
226{
227 return (blk << inode->i_blkbits);
228}
229
230/**
231 * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
232 * @inode: the inode to map
233 * @fieinfo: the fiemap info struct that will be passed back to userspace
234 * @start: where to start mapping in the inode
235 * @len: how much space to map
236 * @get_block: the fs's get_block function
237 *
238 * This does FIEMAP for block based inodes. Basically it will just loop
239 * through get_block until we hit the number of extents we want to map, or we
240 * go past the end of the file and hit a hole.
241 *
242 * If it is possible to have data blocks beyond a hole past @inode->i_size, then
243 * please do not use this function, it will stop at the first unmapped block
244 * beyond i_size.
245 *
246 * If you use this function directly, you need to do your own locking. Use
247 * generic_block_fiemap if you want the locking done for you.
248 */
249
250int __generic_block_fiemap(struct inode *inode,
251 struct fiemap_extent_info *fieinfo, loff_t start,
252 loff_t len, get_block_t *get_block)
253{
254 struct buffer_head map_bh;
255 sector_t start_blk, last_blk;
256 loff_t isize = i_size_read(inode);
257 u64 logical = 0, phys = 0, size = 0;
258 u32 flags = FIEMAP_EXTENT_MERGED;
259 bool past_eof = false, whole_file = false;
260 int ret = 0;
261
262 ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC);
263 if (ret)
264 return ret;
265
266 /*
267 * Either the i_mutex or other appropriate locking needs to be held
268 * since we expect isize to not change at all through the duration of
269 * this call.
270 */
271 if (len >= isize) {
272 whole_file = true;
273 len = isize;
274 }
275
276 /*
277 * Some filesystems can't deal with being asked to map less than
278 * blocksize, so make sure our len is at least block length.
279 */
280 if (logical_to_blk(inode, len) == 0)
281 len = blk_to_logical(inode, 1);
282
283 start_blk = logical_to_blk(inode, start);
284 last_blk = logical_to_blk(inode, start + len - 1);
285
286 do {
287 /*
288 * we set b_size to the total size we want so it will map as
289 * many contiguous blocks as possible at once
290 */
291 memset(&map_bh, 0, sizeof(struct buffer_head));
292 map_bh.b_size = len;
293
294 ret = get_block(inode, start_blk, &map_bh, 0);
295 if (ret)
296 break;
297
298 /* HOLE */
299 if (!buffer_mapped(&map_bh)) {
300 start_blk++;
301
302 /*
303 * We want to handle the case where there is an
304 * allocated block at the front of the file, and then
305 * nothing but holes up to the end of the file properly,
306 * to make sure that extent at the front gets properly
307 * marked with FIEMAP_EXTENT_LAST
308 */
309 if (!past_eof &&
310 blk_to_logical(inode, start_blk) >= isize)
311 past_eof = 1;
312
313 /*
314 * First hole after going past the EOF, this is our
315 * last extent
316 */
317 if (past_eof && size) {
318 flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
319 ret = fiemap_fill_next_extent(fieinfo, logical,
320 phys, size,
321 flags);
322 } else if (size) {
323 ret = fiemap_fill_next_extent(fieinfo, logical,
324 phys, size, flags);
325 size = 0;
326 }
327
328 /* if we have holes up to/past EOF then we're done */
329 if (start_blk > last_blk || past_eof || ret)
330 break;
331 } else {
332 /*
333 * We have gone over the length of what we wanted to
334 * map, and it wasn't the entire file, so add the extent
335 * we got last time and exit.
336 *
337 * This is for the case where say we want to map all the
338 * way up to the second to the last block in a file, but
339 * the last block is a hole, making the second to last
340 * block FIEMAP_EXTENT_LAST. In this case we want to
341 * see if there is a hole after the second to last block
342 * so we can mark it properly. If we found data after
343 * we exceeded the length we were requesting, then we
344 * are good to go, just add the extent to the fieinfo
345 * and break
346 */
347 if (start_blk > last_blk && !whole_file) {
348 ret = fiemap_fill_next_extent(fieinfo, logical,
349 phys, size,
350 flags);
351 break;
352 }
353
354 /*
355 * if size != 0 then we know we already have an extent
356 * to add, so add it.
357 */
358 if (size) {
359 ret = fiemap_fill_next_extent(fieinfo, logical,
360 phys, size,
361 flags);
362 if (ret)
363 break;
364 }
365
366 logical = blk_to_logical(inode, start_blk);
367 phys = blk_to_logical(inode, map_bh.b_blocknr);
368 size = map_bh.b_size;
369 flags = FIEMAP_EXTENT_MERGED;
370
371 start_blk += logical_to_blk(inode, size);
372
373 /*
374 * If we are past the EOF, then we need to make sure as
375 * soon as we find a hole that the last extent we found
376 * is marked with FIEMAP_EXTENT_LAST
377 */
378 if (!past_eof && logical + size >= isize)
379 past_eof = true;
380 }
381 cond_resched();
382 } while (1);
383
384 /* If ret is 1 then we just hit the end of the extent array */
385 if (ret == 1)
386 ret = 0;
387
388 return ret;
389}
390EXPORT_SYMBOL(__generic_block_fiemap);
391
392/**
393 * generic_block_fiemap - FIEMAP for block based inodes
394 * @inode: The inode to map
395 * @fieinfo: The mapping information
396 * @start: The initial block to map
397 * @len: The length of the extect to attempt to map
398 * @get_block: The block mapping function for the fs
399 *
400 * Calls __generic_block_fiemap to map the inode, after taking
401 * the inode's mutex lock.
402 */
403
404int generic_block_fiemap(struct inode *inode,
405 struct fiemap_extent_info *fieinfo, u64 start,
406 u64 len, get_block_t *get_block)
407{
408 int ret;
409 mutex_lock(&inode->i_mutex);
410 ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
411 mutex_unlock(&inode->i_mutex);
412 return ret;
413}
414EXPORT_SYMBOL(generic_block_fiemap);
415
416#endif /* CONFIG_BLOCK */
417
418/*
419 * This provides compatibility with legacy XFS pre-allocation ioctls
420 * which predate the fallocate syscall.
421 *
422 * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
423 * are used here, rest are ignored.
424 */
425int ioctl_preallocate(struct file *filp, void __user *argp)
426{
427 struct inode *inode = filp->f_path.dentry->d_inode;
428 struct space_resv sr;
429
430 if (copy_from_user(&sr, argp, sizeof(sr)))
431 return -EFAULT;
432
433 switch (sr.l_whence) {
434 case SEEK_SET:
435 break;
436 case SEEK_CUR:
437 sr.l_start += filp->f_pos;
438 break;
439 case SEEK_END:
440 sr.l_start += i_size_read(inode);
441 break;
442 default:
443 return -EINVAL;
444 }
445
446 return do_fallocate(filp, FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
447}
448
449static int file_ioctl(struct file *filp, unsigned int cmd,
450 unsigned long arg)
451{
452 struct inode *inode = filp->f_path.dentry->d_inode;
453 int __user *p = (int __user *)arg;
454
455 switch (cmd) {
456 case FIBMAP:
457 return ioctl_fibmap(filp, p);
458 case FIONREAD:
459 return put_user(i_size_read(inode) - filp->f_pos, p);
460 case FS_IOC_RESVSP:
461 case FS_IOC_RESVSP64:
462 return ioctl_preallocate(filp, p);
463 }
464
465 return vfs_ioctl(filp, cmd, arg);
466}
467
468static int ioctl_fionbio(struct file *filp, int __user *argp)
469{
470 unsigned int flag;
471 int on, error;
472
473 error = get_user(on, argp);
474 if (error)
475 return error;
476 flag = O_NONBLOCK;
477#ifdef __sparc__
478 /* SunOS compatibility item. */
479 if (O_NONBLOCK != O_NDELAY)
480 flag |= O_NDELAY;
481#endif
482 spin_lock(&filp->f_lock);
483 if (on)
484 filp->f_flags |= flag;
485 else
486 filp->f_flags &= ~flag;
487 spin_unlock(&filp->f_lock);
488 return error;
489}
490
491static int ioctl_fioasync(unsigned int fd, struct file *filp,
492 int __user *argp)
493{
494 unsigned int flag;
495 int on, error;
496
497 error = get_user(on, argp);
498 if (error)
499 return error;
500 flag = on ? FASYNC : 0;
501
502 /* Did FASYNC state change ? */
503 if ((flag ^ filp->f_flags) & FASYNC) {
504 if (filp->f_op && filp->f_op->fasync)
505 /* fasync() adjusts filp->f_flags */
506 error = filp->f_op->fasync(fd, filp, on);
507 else
508 error = -ENOTTY;
509 }
510 return error < 0 ? error : 0;
511}
512
513static int ioctl_fsfreeze(struct file *filp)
514{
515 struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
516
517 if (!capable(CAP_SYS_ADMIN))
518 return -EPERM;
519
520 /* If filesystem doesn't support freeze feature, return. */
521 if (sb->s_op->freeze_fs == NULL)
522 return -EOPNOTSUPP;
523
524 /* Freeze */
525 return freeze_super(sb);
526}
527
528static int ioctl_fsthaw(struct file *filp)
529{
530 struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
531
532 if (!capable(CAP_SYS_ADMIN))
533 return -EPERM;
534
535 /* Thaw */
536 return thaw_super(sb);
537}
538
539/*
540 * When you add any new common ioctls to the switches above and below
541 * please update compat_sys_ioctl() too.
542 *
543 * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
544 * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
545 */
546int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
547 unsigned long arg)
548{
549 int error = 0;
550 int __user *argp = (int __user *)arg;
551 struct inode *inode = filp->f_path.dentry->d_inode;
552
553 switch (cmd) {
554 case FIOCLEX:
555 set_close_on_exec(fd, 1);
556 break;
557
558 case FIONCLEX:
559 set_close_on_exec(fd, 0);
560 break;
561
562 case FIONBIO:
563 error = ioctl_fionbio(filp, argp);
564 break;
565
566 case FIOASYNC:
567 error = ioctl_fioasync(fd, filp, argp);
568 break;
569
570 case FIOQSIZE:
571 if (S_ISDIR(inode->i_mode) || S_ISREG(inode->i_mode) ||
572 S_ISLNK(inode->i_mode)) {
573 loff_t res = inode_get_bytes(inode);
574 error = copy_to_user(argp, &res, sizeof(res)) ?
575 -EFAULT : 0;
576 } else
577 error = -ENOTTY;
578 break;
579
580 case FIFREEZE:
581 error = ioctl_fsfreeze(filp);
582 break;
583
584 case FITHAW:
585 error = ioctl_fsthaw(filp);
586 break;
587
588 case FS_IOC_FIEMAP:
589 return ioctl_fiemap(filp, arg);
590
591 case FIGETBSZ:
592 return put_user(inode->i_sb->s_blocksize, argp);
593
594 default:
595 if (S_ISREG(inode->i_mode))
596 error = file_ioctl(filp, cmd, arg);
597 else
598 error = vfs_ioctl(filp, cmd, arg);
599 break;
600 }
601 return error;
602}
603
604SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
605{
606 struct file *filp;
607 int error = -EBADF;
608 int fput_needed;
609
610 filp = fget_light(fd, &fput_needed);
611 if (!filp)
612 goto out;
613
614 error = security_file_ioctl(filp, cmd, arg);
615 if (error)
616 goto out_fput;
617
618 error = do_vfs_ioctl(filp, fd, cmd, arg);
619 out_fput:
620 fput_light(filp, fput_needed);
621 out:
622 return error;
623}
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#include <linux/mount.h>
23#include <linux/fscrypt.h>
24#include <linux/fileattr.h>
25
26#include "internal.h"
27
28#include <asm/ioctls.h>
29
30/* So that the fiemap access checks can't overflow on 32 bit machines. */
31#define FIEMAP_MAX_EXTENTS (UINT_MAX / sizeof(struct fiemap_extent))
32
33/**
34 * vfs_ioctl - call filesystem specific ioctl methods
35 * @filp: open file to invoke ioctl method on
36 * @cmd: ioctl command to execute
37 * @arg: command-specific argument for ioctl
38 *
39 * Invokes filesystem specific ->unlocked_ioctl, if one exists; otherwise
40 * returns -ENOTTY.
41 *
42 * Returns 0 on success, -errno on error.
43 */
44long vfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
45{
46 int error = -ENOTTY;
47
48 if (!filp->f_op->unlocked_ioctl)
49 goto out;
50
51 error = filp->f_op->unlocked_ioctl(filp, cmd, arg);
52 if (error == -ENOIOCTLCMD)
53 error = -ENOTTY;
54 out:
55 return error;
56}
57EXPORT_SYMBOL(vfs_ioctl);
58
59static int ioctl_fibmap(struct file *filp, int __user *p)
60{
61 struct inode *inode = file_inode(filp);
62 struct super_block *sb = inode->i_sb;
63 int error, ur_block;
64 sector_t block;
65
66 if (!capable(CAP_SYS_RAWIO))
67 return -EPERM;
68
69 error = get_user(ur_block, p);
70 if (error)
71 return error;
72
73 if (ur_block < 0)
74 return -EINVAL;
75
76 block = ur_block;
77 error = bmap(inode, &block);
78
79 if (block > INT_MAX) {
80 error = -ERANGE;
81 pr_warn_ratelimited("[%s/%d] FS: %s File: %pD4 would truncate fibmap result\n",
82 current->comm, task_pid_nr(current),
83 sb->s_id, filp);
84 }
85
86 if (error)
87 ur_block = 0;
88 else
89 ur_block = block;
90
91 if (put_user(ur_block, p))
92 error = -EFAULT;
93
94 return error;
95}
96
97/**
98 * fiemap_fill_next_extent - Fiemap helper function
99 * @fieinfo: Fiemap context passed into ->fiemap
100 * @logical: Extent logical start offset, in bytes
101 * @phys: Extent physical start offset, in bytes
102 * @len: Extent length, in bytes
103 * @flags: FIEMAP_EXTENT flags that describe this extent
104 *
105 * Called from file system ->fiemap callback. Will populate extent
106 * info as passed in via arguments and copy to user memory. On
107 * success, extent count on fieinfo is incremented.
108 *
109 * Returns 0 on success, -errno on error, 1 if this was the last
110 * extent that will fit in user array.
111 */
112#define SET_UNKNOWN_FLAGS (FIEMAP_EXTENT_DELALLOC)
113#define SET_NO_UNMOUNTED_IO_FLAGS (FIEMAP_EXTENT_DATA_ENCRYPTED)
114#define SET_NOT_ALIGNED_FLAGS (FIEMAP_EXTENT_DATA_TAIL|FIEMAP_EXTENT_DATA_INLINE)
115int fiemap_fill_next_extent(struct fiemap_extent_info *fieinfo, u64 logical,
116 u64 phys, u64 len, u32 flags)
117{
118 struct fiemap_extent extent;
119 struct fiemap_extent __user *dest = fieinfo->fi_extents_start;
120
121 /* only count the extents */
122 if (fieinfo->fi_extents_max == 0) {
123 fieinfo->fi_extents_mapped++;
124 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
125 }
126
127 if (fieinfo->fi_extents_mapped >= fieinfo->fi_extents_max)
128 return 1;
129
130 if (flags & SET_UNKNOWN_FLAGS)
131 flags |= FIEMAP_EXTENT_UNKNOWN;
132 if (flags & SET_NO_UNMOUNTED_IO_FLAGS)
133 flags |= FIEMAP_EXTENT_ENCODED;
134 if (flags & SET_NOT_ALIGNED_FLAGS)
135 flags |= FIEMAP_EXTENT_NOT_ALIGNED;
136
137 memset(&extent, 0, sizeof(extent));
138 extent.fe_logical = logical;
139 extent.fe_physical = phys;
140 extent.fe_length = len;
141 extent.fe_flags = flags;
142
143 dest += fieinfo->fi_extents_mapped;
144 if (copy_to_user(dest, &extent, sizeof(extent)))
145 return -EFAULT;
146
147 fieinfo->fi_extents_mapped++;
148 if (fieinfo->fi_extents_mapped == fieinfo->fi_extents_max)
149 return 1;
150 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
151}
152EXPORT_SYMBOL(fiemap_fill_next_extent);
153
154/**
155 * fiemap_prep - check validity of requested flags for fiemap
156 * @inode: Inode to operate on
157 * @fieinfo: Fiemap context passed into ->fiemap
158 * @start: Start of the mapped range
159 * @len: Length of the mapped range, can be truncated by this function.
160 * @supported_flags: Set of fiemap flags that the file system understands
161 *
162 * This function must be called from each ->fiemap instance to validate the
163 * fiemap request against the file system parameters.
164 *
165 * Returns 0 on success, or a negative error on failure.
166 */
167int fiemap_prep(struct inode *inode, struct fiemap_extent_info *fieinfo,
168 u64 start, u64 *len, u32 supported_flags)
169{
170 u64 maxbytes = inode->i_sb->s_maxbytes;
171 u32 incompat_flags;
172 int ret = 0;
173
174 if (*len == 0)
175 return -EINVAL;
176 if (start > maxbytes)
177 return -EFBIG;
178
179 /*
180 * Shrink request scope to what the fs can actually handle.
181 */
182 if (*len > maxbytes || (maxbytes - *len) < start)
183 *len = maxbytes - start;
184
185 supported_flags |= FIEMAP_FLAG_SYNC;
186 supported_flags &= FIEMAP_FLAGS_COMPAT;
187 incompat_flags = fieinfo->fi_flags & ~supported_flags;
188 if (incompat_flags) {
189 fieinfo->fi_flags = incompat_flags;
190 return -EBADR;
191 }
192
193 if (fieinfo->fi_flags & FIEMAP_FLAG_SYNC)
194 ret = filemap_write_and_wait(inode->i_mapping);
195 return ret;
196}
197EXPORT_SYMBOL(fiemap_prep);
198
199static int ioctl_fiemap(struct file *filp, struct fiemap __user *ufiemap)
200{
201 struct fiemap fiemap;
202 struct fiemap_extent_info fieinfo = { 0, };
203 struct inode *inode = file_inode(filp);
204 int error;
205
206 if (!inode->i_op->fiemap)
207 return -EOPNOTSUPP;
208
209 if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
210 return -EFAULT;
211
212 if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
213 return -EINVAL;
214
215 fieinfo.fi_flags = fiemap.fm_flags;
216 fieinfo.fi_extents_max = fiemap.fm_extent_count;
217 fieinfo.fi_extents_start = ufiemap->fm_extents;
218
219 error = inode->i_op->fiemap(inode, &fieinfo, fiemap.fm_start,
220 fiemap.fm_length);
221
222 fiemap.fm_flags = fieinfo.fi_flags;
223 fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
224 if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
225 error = -EFAULT;
226
227 return error;
228}
229
230static long ioctl_file_clone(struct file *dst_file, unsigned long srcfd,
231 u64 off, u64 olen, u64 destoff)
232{
233 struct fd src_file = fdget(srcfd);
234 loff_t cloned;
235 int ret;
236
237 if (!src_file.file)
238 return -EBADF;
239 ret = -EXDEV;
240 if (src_file.file->f_path.mnt != dst_file->f_path.mnt)
241 goto fdput;
242 cloned = vfs_clone_file_range(src_file.file, off, dst_file, destoff,
243 olen, 0);
244 if (cloned < 0)
245 ret = cloned;
246 else if (olen && cloned != olen)
247 ret = -EINVAL;
248 else
249 ret = 0;
250fdput:
251 fdput(src_file);
252 return ret;
253}
254
255static long ioctl_file_clone_range(struct file *file,
256 struct file_clone_range __user *argp)
257{
258 struct file_clone_range args;
259
260 if (copy_from_user(&args, argp, sizeof(args)))
261 return -EFAULT;
262 return ioctl_file_clone(file, args.src_fd, args.src_offset,
263 args.src_length, args.dest_offset);
264}
265
266#ifdef CONFIG_BLOCK
267
268static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
269{
270 return (offset >> inode->i_blkbits);
271}
272
273static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
274{
275 return (blk << inode->i_blkbits);
276}
277
278/**
279 * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
280 * @inode: the inode to map
281 * @fieinfo: the fiemap info struct that will be passed back to userspace
282 * @start: where to start mapping in the inode
283 * @len: how much space to map
284 * @get_block: the fs's get_block function
285 *
286 * This does FIEMAP for block based inodes. Basically it will just loop
287 * through get_block until we hit the number of extents we want to map, or we
288 * go past the end of the file and hit a hole.
289 *
290 * If it is possible to have data blocks beyond a hole past @inode->i_size, then
291 * please do not use this function, it will stop at the first unmapped block
292 * beyond i_size.
293 *
294 * If you use this function directly, you need to do your own locking. Use
295 * generic_block_fiemap if you want the locking done for you.
296 */
297static int __generic_block_fiemap(struct inode *inode,
298 struct fiemap_extent_info *fieinfo, loff_t start,
299 loff_t len, get_block_t *get_block)
300{
301 struct buffer_head map_bh;
302 sector_t start_blk, last_blk;
303 loff_t isize = i_size_read(inode);
304 u64 logical = 0, phys = 0, size = 0;
305 u32 flags = FIEMAP_EXTENT_MERGED;
306 bool past_eof = false, whole_file = false;
307 int ret = 0;
308
309 ret = fiemap_prep(inode, fieinfo, start, &len, FIEMAP_FLAG_SYNC);
310 if (ret)
311 return ret;
312
313 /*
314 * Either the i_mutex or other appropriate locking needs to be held
315 * since we expect isize to not change at all through the duration of
316 * this call.
317 */
318 if (len >= isize) {
319 whole_file = true;
320 len = isize;
321 }
322
323 /*
324 * Some filesystems can't deal with being asked to map less than
325 * blocksize, so make sure our len is at least block length.
326 */
327 if (logical_to_blk(inode, len) == 0)
328 len = blk_to_logical(inode, 1);
329
330 start_blk = logical_to_blk(inode, start);
331 last_blk = logical_to_blk(inode, start + len - 1);
332
333 do {
334 /*
335 * we set b_size to the total size we want so it will map as
336 * many contiguous blocks as possible at once
337 */
338 memset(&map_bh, 0, sizeof(struct buffer_head));
339 map_bh.b_size = len;
340
341 ret = get_block(inode, start_blk, &map_bh, 0);
342 if (ret)
343 break;
344
345 /* HOLE */
346 if (!buffer_mapped(&map_bh)) {
347 start_blk++;
348
349 /*
350 * We want to handle the case where there is an
351 * allocated block at the front of the file, and then
352 * nothing but holes up to the end of the file properly,
353 * to make sure that extent at the front gets properly
354 * marked with FIEMAP_EXTENT_LAST
355 */
356 if (!past_eof &&
357 blk_to_logical(inode, start_blk) >= isize)
358 past_eof = 1;
359
360 /*
361 * First hole after going past the EOF, this is our
362 * last extent
363 */
364 if (past_eof && size) {
365 flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
366 ret = fiemap_fill_next_extent(fieinfo, logical,
367 phys, size,
368 flags);
369 } else if (size) {
370 ret = fiemap_fill_next_extent(fieinfo, logical,
371 phys, size, flags);
372 size = 0;
373 }
374
375 /* if we have holes up to/past EOF then we're done */
376 if (start_blk > last_blk || past_eof || ret)
377 break;
378 } else {
379 /*
380 * We have gone over the length of what we wanted to
381 * map, and it wasn't the entire file, so add the extent
382 * we got last time and exit.
383 *
384 * This is for the case where say we want to map all the
385 * way up to the second to the last block in a file, but
386 * the last block is a hole, making the second to last
387 * block FIEMAP_EXTENT_LAST. In this case we want to
388 * see if there is a hole after the second to last block
389 * so we can mark it properly. If we found data after
390 * we exceeded the length we were requesting, then we
391 * are good to go, just add the extent to the fieinfo
392 * and break
393 */
394 if (start_blk > last_blk && !whole_file) {
395 ret = fiemap_fill_next_extent(fieinfo, logical,
396 phys, size,
397 flags);
398 break;
399 }
400
401 /*
402 * if size != 0 then we know we already have an extent
403 * to add, so add it.
404 */
405 if (size) {
406 ret = fiemap_fill_next_extent(fieinfo, logical,
407 phys, size,
408 flags);
409 if (ret)
410 break;
411 }
412
413 logical = blk_to_logical(inode, start_blk);
414 phys = blk_to_logical(inode, map_bh.b_blocknr);
415 size = map_bh.b_size;
416 flags = FIEMAP_EXTENT_MERGED;
417
418 start_blk += logical_to_blk(inode, size);
419
420 /*
421 * If we are past the EOF, then we need to make sure as
422 * soon as we find a hole that the last extent we found
423 * is marked with FIEMAP_EXTENT_LAST
424 */
425 if (!past_eof && logical + size >= isize)
426 past_eof = true;
427 }
428 cond_resched();
429 if (fatal_signal_pending(current)) {
430 ret = -EINTR;
431 break;
432 }
433
434 } while (1);
435
436 /* If ret is 1 then we just hit the end of the extent array */
437 if (ret == 1)
438 ret = 0;
439
440 return ret;
441}
442
443/**
444 * generic_block_fiemap - FIEMAP for block based inodes
445 * @inode: The inode to map
446 * @fieinfo: The mapping information
447 * @start: The initial block to map
448 * @len: The length of the extect to attempt to map
449 * @get_block: The block mapping function for the fs
450 *
451 * Calls __generic_block_fiemap to map the inode, after taking
452 * the inode's mutex lock.
453 */
454
455int generic_block_fiemap(struct inode *inode,
456 struct fiemap_extent_info *fieinfo, u64 start,
457 u64 len, get_block_t *get_block)
458{
459 int ret;
460 inode_lock(inode);
461 ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
462 inode_unlock(inode);
463 return ret;
464}
465EXPORT_SYMBOL(generic_block_fiemap);
466
467#endif /* CONFIG_BLOCK */
468
469/*
470 * This provides compatibility with legacy XFS pre-allocation ioctls
471 * which predate the fallocate syscall.
472 *
473 * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
474 * are used here, rest are ignored.
475 */
476static int ioctl_preallocate(struct file *filp, int mode, void __user *argp)
477{
478 struct inode *inode = file_inode(filp);
479 struct space_resv sr;
480
481 if (copy_from_user(&sr, argp, sizeof(sr)))
482 return -EFAULT;
483
484 switch (sr.l_whence) {
485 case SEEK_SET:
486 break;
487 case SEEK_CUR:
488 sr.l_start += filp->f_pos;
489 break;
490 case SEEK_END:
491 sr.l_start += i_size_read(inode);
492 break;
493 default:
494 return -EINVAL;
495 }
496
497 return vfs_fallocate(filp, mode | FALLOC_FL_KEEP_SIZE, sr.l_start,
498 sr.l_len);
499}
500
501/* on ia32 l_start is on a 32-bit boundary */
502#if defined CONFIG_COMPAT && defined(CONFIG_X86_64)
503/* just account for different alignment */
504static int compat_ioctl_preallocate(struct file *file, int mode,
505 struct space_resv_32 __user *argp)
506{
507 struct inode *inode = file_inode(file);
508 struct space_resv_32 sr;
509
510 if (copy_from_user(&sr, argp, sizeof(sr)))
511 return -EFAULT;
512
513 switch (sr.l_whence) {
514 case SEEK_SET:
515 break;
516 case SEEK_CUR:
517 sr.l_start += file->f_pos;
518 break;
519 case SEEK_END:
520 sr.l_start += i_size_read(inode);
521 break;
522 default:
523 return -EINVAL;
524 }
525
526 return vfs_fallocate(file, mode | FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
527}
528#endif
529
530static int file_ioctl(struct file *filp, unsigned int cmd, int __user *p)
531{
532 switch (cmd) {
533 case FIBMAP:
534 return ioctl_fibmap(filp, p);
535 case FS_IOC_RESVSP:
536 case FS_IOC_RESVSP64:
537 return ioctl_preallocate(filp, 0, p);
538 case FS_IOC_UNRESVSP:
539 case FS_IOC_UNRESVSP64:
540 return ioctl_preallocate(filp, FALLOC_FL_PUNCH_HOLE, p);
541 case FS_IOC_ZERO_RANGE:
542 return ioctl_preallocate(filp, FALLOC_FL_ZERO_RANGE, p);
543 }
544
545 return -ENOIOCTLCMD;
546}
547
548static int ioctl_fionbio(struct file *filp, int __user *argp)
549{
550 unsigned int flag;
551 int on, error;
552
553 error = get_user(on, argp);
554 if (error)
555 return error;
556 flag = O_NONBLOCK;
557#ifdef __sparc__
558 /* SunOS compatibility item. */
559 if (O_NONBLOCK != O_NDELAY)
560 flag |= O_NDELAY;
561#endif
562 spin_lock(&filp->f_lock);
563 if (on)
564 filp->f_flags |= flag;
565 else
566 filp->f_flags &= ~flag;
567 spin_unlock(&filp->f_lock);
568 return error;
569}
570
571static int ioctl_fioasync(unsigned int fd, struct file *filp,
572 int __user *argp)
573{
574 unsigned int flag;
575 int on, error;
576
577 error = get_user(on, argp);
578 if (error)
579 return error;
580 flag = on ? FASYNC : 0;
581
582 /* Did FASYNC state change ? */
583 if ((flag ^ filp->f_flags) & FASYNC) {
584 if (filp->f_op->fasync)
585 /* fasync() adjusts filp->f_flags */
586 error = filp->f_op->fasync(fd, filp, on);
587 else
588 error = -ENOTTY;
589 }
590 return error < 0 ? error : 0;
591}
592
593static int ioctl_fsfreeze(struct file *filp)
594{
595 struct super_block *sb = file_inode(filp)->i_sb;
596
597 if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN))
598 return -EPERM;
599
600 /* If filesystem doesn't support freeze feature, return. */
601 if (sb->s_op->freeze_fs == NULL && sb->s_op->freeze_super == NULL)
602 return -EOPNOTSUPP;
603
604 /* Freeze */
605 if (sb->s_op->freeze_super)
606 return sb->s_op->freeze_super(sb);
607 return freeze_super(sb);
608}
609
610static int ioctl_fsthaw(struct file *filp)
611{
612 struct super_block *sb = file_inode(filp)->i_sb;
613
614 if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN))
615 return -EPERM;
616
617 /* Thaw */
618 if (sb->s_op->thaw_super)
619 return sb->s_op->thaw_super(sb);
620 return thaw_super(sb);
621}
622
623static int ioctl_file_dedupe_range(struct file *file,
624 struct file_dedupe_range __user *argp)
625{
626 struct file_dedupe_range *same = NULL;
627 int ret;
628 unsigned long size;
629 u16 count;
630
631 if (get_user(count, &argp->dest_count)) {
632 ret = -EFAULT;
633 goto out;
634 }
635
636 size = offsetof(struct file_dedupe_range __user, info[count]);
637 if (size > PAGE_SIZE) {
638 ret = -ENOMEM;
639 goto out;
640 }
641
642 same = memdup_user(argp, size);
643 if (IS_ERR(same)) {
644 ret = PTR_ERR(same);
645 same = NULL;
646 goto out;
647 }
648
649 same->dest_count = count;
650 ret = vfs_dedupe_file_range(file, same);
651 if (ret)
652 goto out;
653
654 ret = copy_to_user(argp, same, size);
655 if (ret)
656 ret = -EFAULT;
657
658out:
659 kfree(same);
660 return ret;
661}
662
663/**
664 * fileattr_fill_xflags - initialize fileattr with xflags
665 * @fa: fileattr pointer
666 * @xflags: FS_XFLAG_* flags
667 *
668 * Set ->fsx_xflags, ->fsx_valid and ->flags (translated xflags). All
669 * other fields are zeroed.
670 */
671void fileattr_fill_xflags(struct fileattr *fa, u32 xflags)
672{
673 memset(fa, 0, sizeof(*fa));
674 fa->fsx_valid = true;
675 fa->fsx_xflags = xflags;
676 if (fa->fsx_xflags & FS_XFLAG_IMMUTABLE)
677 fa->flags |= FS_IMMUTABLE_FL;
678 if (fa->fsx_xflags & FS_XFLAG_APPEND)
679 fa->flags |= FS_APPEND_FL;
680 if (fa->fsx_xflags & FS_XFLAG_SYNC)
681 fa->flags |= FS_SYNC_FL;
682 if (fa->fsx_xflags & FS_XFLAG_NOATIME)
683 fa->flags |= FS_NOATIME_FL;
684 if (fa->fsx_xflags & FS_XFLAG_NODUMP)
685 fa->flags |= FS_NODUMP_FL;
686 if (fa->fsx_xflags & FS_XFLAG_DAX)
687 fa->flags |= FS_DAX_FL;
688 if (fa->fsx_xflags & FS_XFLAG_PROJINHERIT)
689 fa->flags |= FS_PROJINHERIT_FL;
690}
691EXPORT_SYMBOL(fileattr_fill_xflags);
692
693/**
694 * fileattr_fill_flags - initialize fileattr with flags
695 * @fa: fileattr pointer
696 * @flags: FS_*_FL flags
697 *
698 * Set ->flags, ->flags_valid and ->fsx_xflags (translated flags).
699 * All other fields are zeroed.
700 */
701void fileattr_fill_flags(struct fileattr *fa, u32 flags)
702{
703 memset(fa, 0, sizeof(*fa));
704 fa->flags_valid = true;
705 fa->flags = flags;
706 if (fa->flags & FS_SYNC_FL)
707 fa->fsx_xflags |= FS_XFLAG_SYNC;
708 if (fa->flags & FS_IMMUTABLE_FL)
709 fa->fsx_xflags |= FS_XFLAG_IMMUTABLE;
710 if (fa->flags & FS_APPEND_FL)
711 fa->fsx_xflags |= FS_XFLAG_APPEND;
712 if (fa->flags & FS_NODUMP_FL)
713 fa->fsx_xflags |= FS_XFLAG_NODUMP;
714 if (fa->flags & FS_NOATIME_FL)
715 fa->fsx_xflags |= FS_XFLAG_NOATIME;
716 if (fa->flags & FS_DAX_FL)
717 fa->fsx_xflags |= FS_XFLAG_DAX;
718 if (fa->flags & FS_PROJINHERIT_FL)
719 fa->fsx_xflags |= FS_XFLAG_PROJINHERIT;
720}
721EXPORT_SYMBOL(fileattr_fill_flags);
722
723/**
724 * vfs_fileattr_get - retrieve miscellaneous file attributes
725 * @dentry: the object to retrieve from
726 * @fa: fileattr pointer
727 *
728 * Call i_op->fileattr_get() callback, if exists.
729 *
730 * Return: 0 on success, or a negative error on failure.
731 */
732int vfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
733{
734 struct inode *inode = d_inode(dentry);
735
736 if (!inode->i_op->fileattr_get)
737 return -ENOIOCTLCMD;
738
739 return inode->i_op->fileattr_get(dentry, fa);
740}
741EXPORT_SYMBOL(vfs_fileattr_get);
742
743/**
744 * copy_fsxattr_to_user - copy fsxattr to userspace.
745 * @fa: fileattr pointer
746 * @ufa: fsxattr user pointer
747 *
748 * Return: 0 on success, or -EFAULT on failure.
749 */
750int copy_fsxattr_to_user(const struct fileattr *fa, struct fsxattr __user *ufa)
751{
752 struct fsxattr xfa;
753
754 memset(&xfa, 0, sizeof(xfa));
755 xfa.fsx_xflags = fa->fsx_xflags;
756 xfa.fsx_extsize = fa->fsx_extsize;
757 xfa.fsx_nextents = fa->fsx_nextents;
758 xfa.fsx_projid = fa->fsx_projid;
759 xfa.fsx_cowextsize = fa->fsx_cowextsize;
760
761 if (copy_to_user(ufa, &xfa, sizeof(xfa)))
762 return -EFAULT;
763
764 return 0;
765}
766EXPORT_SYMBOL(copy_fsxattr_to_user);
767
768static int copy_fsxattr_from_user(struct fileattr *fa,
769 struct fsxattr __user *ufa)
770{
771 struct fsxattr xfa;
772
773 if (copy_from_user(&xfa, ufa, sizeof(xfa)))
774 return -EFAULT;
775
776 fileattr_fill_xflags(fa, xfa.fsx_xflags);
777 fa->fsx_extsize = xfa.fsx_extsize;
778 fa->fsx_nextents = xfa.fsx_nextents;
779 fa->fsx_projid = xfa.fsx_projid;
780 fa->fsx_cowextsize = xfa.fsx_cowextsize;
781
782 return 0;
783}
784
785/*
786 * Generic function to check FS_IOC_FSSETXATTR/FS_IOC_SETFLAGS values and reject
787 * any invalid configurations.
788 *
789 * Note: must be called with inode lock held.
790 */
791static int fileattr_set_prepare(struct inode *inode,
792 const struct fileattr *old_ma,
793 struct fileattr *fa)
794{
795 int err;
796
797 /*
798 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
799 * the relevant capability.
800 */
801 if ((fa->flags ^ old_ma->flags) & (FS_APPEND_FL | FS_IMMUTABLE_FL) &&
802 !capable(CAP_LINUX_IMMUTABLE))
803 return -EPERM;
804
805 err = fscrypt_prepare_setflags(inode, old_ma->flags, fa->flags);
806 if (err)
807 return err;
808
809 /*
810 * Project Quota ID state is only allowed to change from within the init
811 * namespace. Enforce that restriction only if we are trying to change
812 * the quota ID state. Everything else is allowed in user namespaces.
813 */
814 if (current_user_ns() != &init_user_ns) {
815 if (old_ma->fsx_projid != fa->fsx_projid)
816 return -EINVAL;
817 if ((old_ma->fsx_xflags ^ fa->fsx_xflags) &
818 FS_XFLAG_PROJINHERIT)
819 return -EINVAL;
820 }
821
822 /* Check extent size hints. */
823 if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(inode->i_mode))
824 return -EINVAL;
825
826 if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
827 !S_ISDIR(inode->i_mode))
828 return -EINVAL;
829
830 if ((fa->fsx_xflags & FS_XFLAG_COWEXTSIZE) &&
831 !S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
832 return -EINVAL;
833
834 /*
835 * It is only valid to set the DAX flag on regular files and
836 * directories on filesystems.
837 */
838 if ((fa->fsx_xflags & FS_XFLAG_DAX) &&
839 !(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
840 return -EINVAL;
841
842 /* Extent size hints of zero turn off the flags. */
843 if (fa->fsx_extsize == 0)
844 fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
845 if (fa->fsx_cowextsize == 0)
846 fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
847
848 return 0;
849}
850
851/**
852 * vfs_fileattr_set - change miscellaneous file attributes
853 * @mnt_userns: user namespace of the mount
854 * @dentry: the object to change
855 * @fa: fileattr pointer
856 *
857 * After verifying permissions, call i_op->fileattr_set() callback, if
858 * exists.
859 *
860 * Verifying attributes involves retrieving current attributes with
861 * i_op->fileattr_get(), this also allows initializing attributes that have
862 * not been set by the caller to current values. Inode lock is held
863 * thoughout to prevent racing with another instance.
864 *
865 * Return: 0 on success, or a negative error on failure.
866 */
867int vfs_fileattr_set(struct user_namespace *mnt_userns, struct dentry *dentry,
868 struct fileattr *fa)
869{
870 struct inode *inode = d_inode(dentry);
871 struct fileattr old_ma = {};
872 int err;
873
874 if (!inode->i_op->fileattr_set)
875 return -ENOIOCTLCMD;
876
877 if (!inode_owner_or_capable(mnt_userns, inode))
878 return -EPERM;
879
880 inode_lock(inode);
881 err = vfs_fileattr_get(dentry, &old_ma);
882 if (!err) {
883 /* initialize missing bits from old_ma */
884 if (fa->flags_valid) {
885 fa->fsx_xflags |= old_ma.fsx_xflags & ~FS_XFLAG_COMMON;
886 fa->fsx_extsize = old_ma.fsx_extsize;
887 fa->fsx_nextents = old_ma.fsx_nextents;
888 fa->fsx_projid = old_ma.fsx_projid;
889 fa->fsx_cowextsize = old_ma.fsx_cowextsize;
890 } else {
891 fa->flags |= old_ma.flags & ~FS_COMMON_FL;
892 }
893 err = fileattr_set_prepare(inode, &old_ma, fa);
894 if (!err)
895 err = inode->i_op->fileattr_set(mnt_userns, dentry, fa);
896 }
897 inode_unlock(inode);
898
899 return err;
900}
901EXPORT_SYMBOL(vfs_fileattr_set);
902
903static int ioctl_getflags(struct file *file, unsigned int __user *argp)
904{
905 struct fileattr fa = { .flags_valid = true }; /* hint only */
906 int err;
907
908 err = vfs_fileattr_get(file->f_path.dentry, &fa);
909 if (!err)
910 err = put_user(fa.flags, argp);
911 return err;
912}
913
914static int ioctl_setflags(struct file *file, unsigned int __user *argp)
915{
916 struct user_namespace *mnt_userns = file_mnt_user_ns(file);
917 struct dentry *dentry = file->f_path.dentry;
918 struct fileattr fa;
919 unsigned int flags;
920 int err;
921
922 err = get_user(flags, argp);
923 if (!err) {
924 err = mnt_want_write_file(file);
925 if (!err) {
926 fileattr_fill_flags(&fa, flags);
927 err = vfs_fileattr_set(mnt_userns, dentry, &fa);
928 mnt_drop_write_file(file);
929 }
930 }
931 return err;
932}
933
934static int ioctl_fsgetxattr(struct file *file, void __user *argp)
935{
936 struct fileattr fa = { .fsx_valid = true }; /* hint only */
937 int err;
938
939 err = vfs_fileattr_get(file->f_path.dentry, &fa);
940 if (!err)
941 err = copy_fsxattr_to_user(&fa, argp);
942
943 return err;
944}
945
946static int ioctl_fssetxattr(struct file *file, void __user *argp)
947{
948 struct user_namespace *mnt_userns = file_mnt_user_ns(file);
949 struct dentry *dentry = file->f_path.dentry;
950 struct fileattr fa;
951 int err;
952
953 err = copy_fsxattr_from_user(&fa, argp);
954 if (!err) {
955 err = mnt_want_write_file(file);
956 if (!err) {
957 err = vfs_fileattr_set(mnt_userns, dentry, &fa);
958 mnt_drop_write_file(file);
959 }
960 }
961 return err;
962}
963
964/*
965 * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
966 * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
967 *
968 * When you add any new common ioctls to the switches above and below,
969 * please ensure they have compatible arguments in compat mode.
970 */
971static int do_vfs_ioctl(struct file *filp, unsigned int fd,
972 unsigned int cmd, unsigned long arg)
973{
974 void __user *argp = (void __user *)arg;
975 struct inode *inode = file_inode(filp);
976
977 switch (cmd) {
978 case FIOCLEX:
979 set_close_on_exec(fd, 1);
980 return 0;
981
982 case FIONCLEX:
983 set_close_on_exec(fd, 0);
984 return 0;
985
986 case FIONBIO:
987 return ioctl_fionbio(filp, argp);
988
989 case FIOASYNC:
990 return ioctl_fioasync(fd, filp, argp);
991
992 case FIOQSIZE:
993 if (S_ISDIR(inode->i_mode) || S_ISREG(inode->i_mode) ||
994 S_ISLNK(inode->i_mode)) {
995 loff_t res = inode_get_bytes(inode);
996 return copy_to_user(argp, &res, sizeof(res)) ?
997 -EFAULT : 0;
998 }
999
1000 return -ENOTTY;
1001
1002 case FIFREEZE:
1003 return ioctl_fsfreeze(filp);
1004
1005 case FITHAW:
1006 return ioctl_fsthaw(filp);
1007
1008 case FS_IOC_FIEMAP:
1009 return ioctl_fiemap(filp, argp);
1010
1011 case FIGETBSZ:
1012 /* anon_bdev filesystems may not have a block size */
1013 if (!inode->i_sb->s_blocksize)
1014 return -EINVAL;
1015
1016 return put_user(inode->i_sb->s_blocksize, (int __user *)argp);
1017
1018 case FICLONE:
1019 return ioctl_file_clone(filp, arg, 0, 0, 0);
1020
1021 case FICLONERANGE:
1022 return ioctl_file_clone_range(filp, argp);
1023
1024 case FIDEDUPERANGE:
1025 return ioctl_file_dedupe_range(filp, argp);
1026
1027 case FIONREAD:
1028 if (!S_ISREG(inode->i_mode))
1029 return vfs_ioctl(filp, cmd, arg);
1030
1031 return put_user(i_size_read(inode) - filp->f_pos,
1032 (int __user *)argp);
1033
1034 case FS_IOC_GETFLAGS:
1035 return ioctl_getflags(filp, argp);
1036
1037 case FS_IOC_SETFLAGS:
1038 return ioctl_setflags(filp, argp);
1039
1040 case FS_IOC_FSGETXATTR:
1041 return ioctl_fsgetxattr(filp, argp);
1042
1043 case FS_IOC_FSSETXATTR:
1044 return ioctl_fssetxattr(filp, argp);
1045
1046 default:
1047 if (S_ISREG(inode->i_mode))
1048 return file_ioctl(filp, cmd, argp);
1049 break;
1050 }
1051
1052 return -ENOIOCTLCMD;
1053}
1054
1055SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
1056{
1057 struct fd f = fdget(fd);
1058 int error;
1059
1060 if (!f.file)
1061 return -EBADF;
1062
1063 error = security_file_ioctl(f.file, cmd, arg);
1064 if (error)
1065 goto out;
1066
1067 error = do_vfs_ioctl(f.file, fd, cmd, arg);
1068 if (error == -ENOIOCTLCMD)
1069 error = vfs_ioctl(f.file, cmd, arg);
1070
1071out:
1072 fdput(f);
1073 return error;
1074}
1075
1076#ifdef CONFIG_COMPAT
1077/**
1078 * compat_ptr_ioctl - generic implementation of .compat_ioctl file operation
1079 *
1080 * This is not normally called as a function, but instead set in struct
1081 * file_operations as
1082 *
1083 * .compat_ioctl = compat_ptr_ioctl,
1084 *
1085 * On most architectures, the compat_ptr_ioctl() just passes all arguments
1086 * to the corresponding ->ioctl handler. The exception is arch/s390, where
1087 * compat_ptr() clears the top bit of a 32-bit pointer value, so user space
1088 * pointers to the second 2GB alias the first 2GB, as is the case for
1089 * native 32-bit s390 user space.
1090 *
1091 * The compat_ptr_ioctl() function must therefore be used only with ioctl
1092 * functions that either ignore the argument or pass a pointer to a
1093 * compatible data type.
1094 *
1095 * If any ioctl command handled by fops->unlocked_ioctl passes a plain
1096 * integer instead of a pointer, or any of the passed data types
1097 * is incompatible between 32-bit and 64-bit architectures, a proper
1098 * handler is required instead of compat_ptr_ioctl.
1099 */
1100long compat_ptr_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1101{
1102 if (!file->f_op->unlocked_ioctl)
1103 return -ENOIOCTLCMD;
1104
1105 return file->f_op->unlocked_ioctl(file, cmd, (unsigned long)compat_ptr(arg));
1106}
1107EXPORT_SYMBOL(compat_ptr_ioctl);
1108
1109COMPAT_SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd,
1110 compat_ulong_t, arg)
1111{
1112 struct fd f = fdget(fd);
1113 int error;
1114
1115 if (!f.file)
1116 return -EBADF;
1117
1118 /* RED-PEN how should LSM module know it's handling 32bit? */
1119 error = security_file_ioctl(f.file, cmd, arg);
1120 if (error)
1121 goto out;
1122
1123 switch (cmd) {
1124 /* FICLONE takes an int argument, so don't use compat_ptr() */
1125 case FICLONE:
1126 error = ioctl_file_clone(f.file, arg, 0, 0, 0);
1127 break;
1128
1129#if defined(CONFIG_X86_64)
1130 /* these get messy on amd64 due to alignment differences */
1131 case FS_IOC_RESVSP_32:
1132 case FS_IOC_RESVSP64_32:
1133 error = compat_ioctl_preallocate(f.file, 0, compat_ptr(arg));
1134 break;
1135 case FS_IOC_UNRESVSP_32:
1136 case FS_IOC_UNRESVSP64_32:
1137 error = compat_ioctl_preallocate(f.file, FALLOC_FL_PUNCH_HOLE,
1138 compat_ptr(arg));
1139 break;
1140 case FS_IOC_ZERO_RANGE_32:
1141 error = compat_ioctl_preallocate(f.file, FALLOC_FL_ZERO_RANGE,
1142 compat_ptr(arg));
1143 break;
1144#endif
1145
1146 /*
1147 * These access 32-bit values anyway so no further handling is
1148 * necessary.
1149 */
1150 case FS_IOC32_GETFLAGS:
1151 case FS_IOC32_SETFLAGS:
1152 cmd = (cmd == FS_IOC32_GETFLAGS) ?
1153 FS_IOC_GETFLAGS : FS_IOC_SETFLAGS;
1154 fallthrough;
1155 /*
1156 * everything else in do_vfs_ioctl() takes either a compatible
1157 * pointer argument or no argument -- call it with a modified
1158 * argument.
1159 */
1160 default:
1161 error = do_vfs_ioctl(f.file, fd, cmd,
1162 (unsigned long)compat_ptr(arg));
1163 if (error != -ENOIOCTLCMD)
1164 break;
1165
1166 if (f.file->f_op->compat_ioctl)
1167 error = f.file->f_op->compat_ioctl(f.file, cmd, arg);
1168 if (error == -ENOIOCTLCMD)
1169 error = -ENOTTY;
1170 break;
1171 }
1172
1173 out:
1174 fdput(f);
1175
1176 return error;
1177}
1178#endif