net/sunrpc/xprt_sock: fix regression in connection error reporting.
[linux-stable.git] / fs / ioctl.c
blob569db68d02b317bbc7f1b4fd040268c3c53afa8a
1 /*
2 * linux/fs/ioctl.c
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
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 <linux/sched/signal.h>
20 #include "internal.h"
22 #include <asm/ioctls.h>
24 /* So that the fiemap access checks can't overflow on 32 bit machines. */
25 #define FIEMAP_MAX_EXTENTS (UINT_MAX / sizeof(struct fiemap_extent))
27 /**
28 * vfs_ioctl - call filesystem specific ioctl methods
29 * @filp: open file to invoke ioctl method on
30 * @cmd: ioctl command to execute
31 * @arg: command-specific argument for ioctl
33 * Invokes filesystem specific ->unlocked_ioctl, if one exists; otherwise
34 * returns -ENOTTY.
36 * Returns 0 on success, -errno on error.
38 long vfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
40 int error = -ENOTTY;
42 if (!filp->f_op->unlocked_ioctl)
43 goto out;
45 error = filp->f_op->unlocked_ioctl(filp, cmd, arg);
46 if (error == -ENOIOCTLCMD)
47 error = -ENOTTY;
48 out:
49 return error;
52 static int ioctl_fibmap(struct file *filp, int __user *p)
54 struct address_space *mapping = filp->f_mapping;
55 int res, block;
57 /* do we support this mess? */
58 if (!mapping->a_ops->bmap)
59 return -EINVAL;
60 if (!capable(CAP_SYS_RAWIO))
61 return -EPERM;
62 res = get_user(block, p);
63 if (res)
64 return res;
65 res = mapping->a_ops->bmap(mapping, block);
66 return put_user(res, p);
69 /**
70 * fiemap_fill_next_extent - Fiemap helper function
71 * @fieinfo: Fiemap context passed into ->fiemap
72 * @logical: Extent logical start offset, in bytes
73 * @phys: Extent physical start offset, in bytes
74 * @len: Extent length, in bytes
75 * @flags: FIEMAP_EXTENT flags that describe this extent
77 * Called from file system ->fiemap callback. Will populate extent
78 * info as passed in via arguments and copy to user memory. On
79 * success, extent count on fieinfo is incremented.
81 * Returns 0 on success, -errno on error, 1 if this was the last
82 * extent that will fit in user array.
84 #define SET_UNKNOWN_FLAGS (FIEMAP_EXTENT_DELALLOC)
85 #define SET_NO_UNMOUNTED_IO_FLAGS (FIEMAP_EXTENT_DATA_ENCRYPTED)
86 #define SET_NOT_ALIGNED_FLAGS (FIEMAP_EXTENT_DATA_TAIL|FIEMAP_EXTENT_DATA_INLINE)
87 int fiemap_fill_next_extent(struct fiemap_extent_info *fieinfo, u64 logical,
88 u64 phys, u64 len, u32 flags)
90 struct fiemap_extent extent;
91 struct fiemap_extent __user *dest = fieinfo->fi_extents_start;
93 /* only count the extents */
94 if (fieinfo->fi_extents_max == 0) {
95 fieinfo->fi_extents_mapped++;
96 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
99 if (fieinfo->fi_extents_mapped >= fieinfo->fi_extents_max)
100 return 1;
102 if (flags & SET_UNKNOWN_FLAGS)
103 flags |= FIEMAP_EXTENT_UNKNOWN;
104 if (flags & SET_NO_UNMOUNTED_IO_FLAGS)
105 flags |= FIEMAP_EXTENT_ENCODED;
106 if (flags & SET_NOT_ALIGNED_FLAGS)
107 flags |= FIEMAP_EXTENT_NOT_ALIGNED;
109 memset(&extent, 0, sizeof(extent));
110 extent.fe_logical = logical;
111 extent.fe_physical = phys;
112 extent.fe_length = len;
113 extent.fe_flags = flags;
115 dest += fieinfo->fi_extents_mapped;
116 if (copy_to_user(dest, &extent, sizeof(extent)))
117 return -EFAULT;
119 fieinfo->fi_extents_mapped++;
120 if (fieinfo->fi_extents_mapped == fieinfo->fi_extents_max)
121 return 1;
122 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
124 EXPORT_SYMBOL(fiemap_fill_next_extent);
127 * fiemap_check_flags - check validity of requested flags for fiemap
128 * @fieinfo: Fiemap context passed into ->fiemap
129 * @fs_flags: Set of fiemap flags that the file system understands
131 * Called from file system ->fiemap callback. This will compute the
132 * intersection of valid fiemap flags and those that the fs supports. That
133 * value is then compared against the user supplied flags. In case of bad user
134 * flags, the invalid values will be written into the fieinfo structure, and
135 * -EBADR is returned, which tells ioctl_fiemap() to return those values to
136 * userspace. For this reason, a return code of -EBADR should be preserved.
138 * Returns 0 on success, -EBADR on bad flags.
140 int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags)
142 u32 incompat_flags;
144 incompat_flags = fieinfo->fi_flags & ~(FIEMAP_FLAGS_COMPAT & fs_flags);
145 if (incompat_flags) {
146 fieinfo->fi_flags = incompat_flags;
147 return -EBADR;
149 return 0;
151 EXPORT_SYMBOL(fiemap_check_flags);
153 static int fiemap_check_ranges(struct super_block *sb,
154 u64 start, u64 len, u64 *new_len)
156 u64 maxbytes = (u64) sb->s_maxbytes;
158 *new_len = len;
160 if (len == 0)
161 return -EINVAL;
163 if (start > maxbytes)
164 return -EFBIG;
167 * Shrink request scope to what the fs can actually handle.
169 if (len > maxbytes || (maxbytes - len) < start)
170 *new_len = maxbytes - start;
172 return 0;
175 static int ioctl_fiemap(struct file *filp, unsigned long arg)
177 struct fiemap fiemap;
178 struct fiemap __user *ufiemap = (struct fiemap __user *) arg;
179 struct fiemap_extent_info fieinfo = { 0, };
180 struct inode *inode = file_inode(filp);
181 struct super_block *sb = inode->i_sb;
182 u64 len;
183 int error;
185 if (!inode->i_op->fiemap)
186 return -EOPNOTSUPP;
188 if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
189 return -EFAULT;
191 if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
192 return -EINVAL;
194 error = fiemap_check_ranges(sb, fiemap.fm_start, fiemap.fm_length,
195 &len);
196 if (error)
197 return error;
199 fieinfo.fi_flags = fiemap.fm_flags;
200 fieinfo.fi_extents_max = fiemap.fm_extent_count;
201 fieinfo.fi_extents_start = ufiemap->fm_extents;
203 if (fiemap.fm_extent_count != 0 &&
204 !access_ok(VERIFY_WRITE, fieinfo.fi_extents_start,
205 fieinfo.fi_extents_max * sizeof(struct fiemap_extent)))
206 return -EFAULT;
208 if (fieinfo.fi_flags & FIEMAP_FLAG_SYNC)
209 filemap_write_and_wait(inode->i_mapping);
211 error = inode->i_op->fiemap(inode, &fieinfo, fiemap.fm_start, len);
212 fiemap.fm_flags = fieinfo.fi_flags;
213 fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
214 if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
215 error = -EFAULT;
217 return error;
220 static long ioctl_file_clone(struct file *dst_file, unsigned long srcfd,
221 u64 off, u64 olen, u64 destoff)
223 struct fd src_file = fdget(srcfd);
224 int ret;
226 if (!src_file.file)
227 return -EBADF;
228 ret = -EXDEV;
229 if (src_file.file->f_path.mnt != dst_file->f_path.mnt)
230 goto fdput;
231 ret = do_clone_file_range(src_file.file, off, dst_file, destoff, olen);
232 fdput:
233 fdput(src_file);
234 return ret;
237 static long ioctl_file_clone_range(struct file *file, void __user *argp)
239 struct file_clone_range args;
241 if (copy_from_user(&args, argp, sizeof(args)))
242 return -EFAULT;
243 return ioctl_file_clone(file, args.src_fd, args.src_offset,
244 args.src_length, args.dest_offset);
247 #ifdef CONFIG_BLOCK
249 static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
251 return (offset >> inode->i_blkbits);
254 static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
256 return (blk << inode->i_blkbits);
260 * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
261 * @inode: the inode to map
262 * @fieinfo: the fiemap info struct that will be passed back to userspace
263 * @start: where to start mapping in the inode
264 * @len: how much space to map
265 * @get_block: the fs's get_block function
267 * This does FIEMAP for block based inodes. Basically it will just loop
268 * through get_block until we hit the number of extents we want to map, or we
269 * go past the end of the file and hit a hole.
271 * If it is possible to have data blocks beyond a hole past @inode->i_size, then
272 * please do not use this function, it will stop at the first unmapped block
273 * beyond i_size.
275 * If you use this function directly, you need to do your own locking. Use
276 * generic_block_fiemap if you want the locking done for you.
279 int __generic_block_fiemap(struct inode *inode,
280 struct fiemap_extent_info *fieinfo, loff_t start,
281 loff_t len, get_block_t *get_block)
283 struct buffer_head map_bh;
284 sector_t start_blk, last_blk;
285 loff_t isize = i_size_read(inode);
286 u64 logical = 0, phys = 0, size = 0;
287 u32 flags = FIEMAP_EXTENT_MERGED;
288 bool past_eof = false, whole_file = false;
289 int ret = 0;
291 ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC);
292 if (ret)
293 return ret;
296 * Either the i_mutex or other appropriate locking needs to be held
297 * since we expect isize to not change at all through the duration of
298 * this call.
300 if (len >= isize) {
301 whole_file = true;
302 len = isize;
306 * Some filesystems can't deal with being asked to map less than
307 * blocksize, so make sure our len is at least block length.
309 if (logical_to_blk(inode, len) == 0)
310 len = blk_to_logical(inode, 1);
312 start_blk = logical_to_blk(inode, start);
313 last_blk = logical_to_blk(inode, start + len - 1);
315 do {
317 * we set b_size to the total size we want so it will map as
318 * many contiguous blocks as possible at once
320 memset(&map_bh, 0, sizeof(struct buffer_head));
321 map_bh.b_size = len;
323 ret = get_block(inode, start_blk, &map_bh, 0);
324 if (ret)
325 break;
327 /* HOLE */
328 if (!buffer_mapped(&map_bh)) {
329 start_blk++;
332 * We want to handle the case where there is an
333 * allocated block at the front of the file, and then
334 * nothing but holes up to the end of the file properly,
335 * to make sure that extent at the front gets properly
336 * marked with FIEMAP_EXTENT_LAST
338 if (!past_eof &&
339 blk_to_logical(inode, start_blk) >= isize)
340 past_eof = 1;
343 * First hole after going past the EOF, this is our
344 * last extent
346 if (past_eof && size) {
347 flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
348 ret = fiemap_fill_next_extent(fieinfo, logical,
349 phys, size,
350 flags);
351 } else if (size) {
352 ret = fiemap_fill_next_extent(fieinfo, logical,
353 phys, size, flags);
354 size = 0;
357 /* if we have holes up to/past EOF then we're done */
358 if (start_blk > last_blk || past_eof || ret)
359 break;
360 } else {
362 * We have gone over the length of what we wanted to
363 * map, and it wasn't the entire file, so add the extent
364 * we got last time and exit.
366 * This is for the case where say we want to map all the
367 * way up to the second to the last block in a file, but
368 * the last block is a hole, making the second to last
369 * block FIEMAP_EXTENT_LAST. In this case we want to
370 * see if there is a hole after the second to last block
371 * so we can mark it properly. If we found data after
372 * we exceeded the length we were requesting, then we
373 * are good to go, just add the extent to the fieinfo
374 * and break
376 if (start_blk > last_blk && !whole_file) {
377 ret = fiemap_fill_next_extent(fieinfo, logical,
378 phys, size,
379 flags);
380 break;
384 * if size != 0 then we know we already have an extent
385 * to add, so add it.
387 if (size) {
388 ret = fiemap_fill_next_extent(fieinfo, logical,
389 phys, size,
390 flags);
391 if (ret)
392 break;
395 logical = blk_to_logical(inode, start_blk);
396 phys = blk_to_logical(inode, map_bh.b_blocknr);
397 size = map_bh.b_size;
398 flags = FIEMAP_EXTENT_MERGED;
400 start_blk += logical_to_blk(inode, size);
403 * If we are past the EOF, then we need to make sure as
404 * soon as we find a hole that the last extent we found
405 * is marked with FIEMAP_EXTENT_LAST
407 if (!past_eof && logical + size >= isize)
408 past_eof = true;
410 cond_resched();
411 if (fatal_signal_pending(current)) {
412 ret = -EINTR;
413 break;
416 } while (1);
418 /* If ret is 1 then we just hit the end of the extent array */
419 if (ret == 1)
420 ret = 0;
422 return ret;
424 EXPORT_SYMBOL(__generic_block_fiemap);
427 * generic_block_fiemap - FIEMAP for block based inodes
428 * @inode: The inode to map
429 * @fieinfo: The mapping information
430 * @start: The initial block to map
431 * @len: The length of the extect to attempt to map
432 * @get_block: The block mapping function for the fs
434 * Calls __generic_block_fiemap to map the inode, after taking
435 * the inode's mutex lock.
438 int generic_block_fiemap(struct inode *inode,
439 struct fiemap_extent_info *fieinfo, u64 start,
440 u64 len, get_block_t *get_block)
442 int ret;
443 inode_lock(inode);
444 ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
445 inode_unlock(inode);
446 return ret;
448 EXPORT_SYMBOL(generic_block_fiemap);
450 #endif /* CONFIG_BLOCK */
453 * This provides compatibility with legacy XFS pre-allocation ioctls
454 * which predate the fallocate syscall.
456 * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
457 * are used here, rest are ignored.
459 int ioctl_preallocate(struct file *filp, void __user *argp)
461 struct inode *inode = file_inode(filp);
462 struct space_resv sr;
464 if (copy_from_user(&sr, argp, sizeof(sr)))
465 return -EFAULT;
467 switch (sr.l_whence) {
468 case SEEK_SET:
469 break;
470 case SEEK_CUR:
471 sr.l_start += filp->f_pos;
472 break;
473 case SEEK_END:
474 sr.l_start += i_size_read(inode);
475 break;
476 default:
477 return -EINVAL;
480 return vfs_fallocate(filp, FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
483 static int file_ioctl(struct file *filp, unsigned int cmd,
484 unsigned long arg)
486 struct inode *inode = file_inode(filp);
487 int __user *p = (int __user *)arg;
489 switch (cmd) {
490 case FIBMAP:
491 return ioctl_fibmap(filp, p);
492 case FIONREAD:
493 return put_user(i_size_read(inode) - filp->f_pos, p);
494 case FS_IOC_RESVSP:
495 case FS_IOC_RESVSP64:
496 return ioctl_preallocate(filp, p);
499 return vfs_ioctl(filp, cmd, arg);
502 static int ioctl_fionbio(struct file *filp, int __user *argp)
504 unsigned int flag;
505 int on, error;
507 error = get_user(on, argp);
508 if (error)
509 return error;
510 flag = O_NONBLOCK;
511 #ifdef __sparc__
512 /* SunOS compatibility item. */
513 if (O_NONBLOCK != O_NDELAY)
514 flag |= O_NDELAY;
515 #endif
516 spin_lock(&filp->f_lock);
517 if (on)
518 filp->f_flags |= flag;
519 else
520 filp->f_flags &= ~flag;
521 spin_unlock(&filp->f_lock);
522 return error;
525 static int ioctl_fioasync(unsigned int fd, struct file *filp,
526 int __user *argp)
528 unsigned int flag;
529 int on, error;
531 error = get_user(on, argp);
532 if (error)
533 return error;
534 flag = on ? FASYNC : 0;
536 /* Did FASYNC state change ? */
537 if ((flag ^ filp->f_flags) & FASYNC) {
538 if (filp->f_op->fasync)
539 /* fasync() adjusts filp->f_flags */
540 error = filp->f_op->fasync(fd, filp, on);
541 else
542 error = -ENOTTY;
544 return error < 0 ? error : 0;
547 static int ioctl_fsfreeze(struct file *filp)
549 struct super_block *sb = file_inode(filp)->i_sb;
551 if (!capable(CAP_SYS_ADMIN))
552 return -EPERM;
554 /* If filesystem doesn't support freeze feature, return. */
555 if (sb->s_op->freeze_fs == NULL && sb->s_op->freeze_super == NULL)
556 return -EOPNOTSUPP;
558 /* Freeze */
559 if (sb->s_op->freeze_super)
560 return sb->s_op->freeze_super(sb);
561 return freeze_super(sb);
564 static int ioctl_fsthaw(struct file *filp)
566 struct super_block *sb = file_inode(filp)->i_sb;
568 if (!capable(CAP_SYS_ADMIN))
569 return -EPERM;
571 /* Thaw */
572 if (sb->s_op->thaw_super)
573 return sb->s_op->thaw_super(sb);
574 return thaw_super(sb);
577 static int ioctl_file_dedupe_range(struct file *file, void __user *arg)
579 struct file_dedupe_range __user *argp = arg;
580 struct file_dedupe_range *same = NULL;
581 int ret;
582 unsigned long size;
583 u16 count;
585 if (get_user(count, &argp->dest_count)) {
586 ret = -EFAULT;
587 goto out;
590 size = offsetof(struct file_dedupe_range __user, info[count]);
591 if (size > PAGE_SIZE) {
592 ret = -ENOMEM;
593 goto out;
596 same = memdup_user(argp, size);
597 if (IS_ERR(same)) {
598 ret = PTR_ERR(same);
599 same = NULL;
600 goto out;
603 same->dest_count = count;
604 ret = vfs_dedupe_file_range(file, same);
605 if (ret)
606 goto out;
608 ret = copy_to_user(argp, same, size);
609 if (ret)
610 ret = -EFAULT;
612 out:
613 kfree(same);
614 return ret;
618 * When you add any new common ioctls to the switches above and below
619 * please update compat_sys_ioctl() too.
621 * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
622 * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
624 int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
625 unsigned long arg)
627 int error = 0;
628 int __user *argp = (int __user *)arg;
629 struct inode *inode = file_inode(filp);
631 switch (cmd) {
632 case FIOCLEX:
633 set_close_on_exec(fd, 1);
634 break;
636 case FIONCLEX:
637 set_close_on_exec(fd, 0);
638 break;
640 case FIONBIO:
641 error = ioctl_fionbio(filp, argp);
642 break;
644 case FIOASYNC:
645 error = ioctl_fioasync(fd, filp, argp);
646 break;
648 case FIOQSIZE:
649 if (S_ISDIR(inode->i_mode) || S_ISREG(inode->i_mode) ||
650 S_ISLNK(inode->i_mode)) {
651 loff_t res = inode_get_bytes(inode);
652 error = copy_to_user(argp, &res, sizeof(res)) ?
653 -EFAULT : 0;
654 } else
655 error = -ENOTTY;
656 break;
658 case FIFREEZE:
659 error = ioctl_fsfreeze(filp);
660 break;
662 case FITHAW:
663 error = ioctl_fsthaw(filp);
664 break;
666 case FS_IOC_FIEMAP:
667 return ioctl_fiemap(filp, arg);
669 case FIGETBSZ:
670 return put_user(inode->i_sb->s_blocksize, argp);
672 case FICLONE:
673 return ioctl_file_clone(filp, arg, 0, 0, 0);
675 case FICLONERANGE:
676 return ioctl_file_clone_range(filp, argp);
678 case FIDEDUPERANGE:
679 return ioctl_file_dedupe_range(filp, argp);
681 default:
682 if (S_ISREG(inode->i_mode))
683 error = file_ioctl(filp, cmd, arg);
684 else
685 error = vfs_ioctl(filp, cmd, arg);
686 break;
688 return error;
691 SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
693 int error;
694 struct fd f = fdget(fd);
696 if (!f.file)
697 return -EBADF;
698 error = security_file_ioctl(f.file, cmd, arg);
699 if (!error)
700 error = do_vfs_ioctl(f.file, fd, cmd, arg);
701 fdput(f);
702 return error;