1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (C) 1991, 1992 Linus Torvalds
8 #include <linux/syscalls.h>
10 #include <linux/capability.h>
11 #include <linux/compat.h>
12 #include <linux/file.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>
24 #include <asm/ioctls.h>
26 /* So that the fiemap access checks can't overflow on 32 bit machines. */
27 #define FIEMAP_MAX_EXTENTS (UINT_MAX / sizeof(struct fiemap_extent))
30 * vfs_ioctl - call filesystem specific ioctl methods
31 * @filp: open file to invoke ioctl method on
32 * @cmd: ioctl command to execute
33 * @arg: command-specific argument for ioctl
35 * Invokes filesystem specific ->unlocked_ioctl, if one exists; otherwise
38 * Returns 0 on success, -errno on error.
40 long vfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
44 if (!filp->f_op->unlocked_ioctl)
47 error = filp->f_op->unlocked_ioctl(filp, cmd, arg);
48 if (error == -ENOIOCTLCMD)
53 EXPORT_SYMBOL(vfs_ioctl);
55 static int ioctl_fibmap(struct file *filp, int __user *p)
57 struct address_space *mapping = filp->f_mapping;
60 /* do we support this mess? */
61 if (!mapping->a_ops->bmap)
63 if (!capable(CAP_SYS_RAWIO))
65 res = get_user(block, p);
68 res = mapping->a_ops->bmap(mapping, block);
69 return put_user(res, p);
73 * fiemap_fill_next_extent - Fiemap helper function
74 * @fieinfo: Fiemap context passed into ->fiemap
75 * @logical: Extent logical start offset, in bytes
76 * @phys: Extent physical start offset, in bytes
77 * @len: Extent length, in bytes
78 * @flags: FIEMAP_EXTENT flags that describe this extent
80 * Called from file system ->fiemap callback. Will populate extent
81 * info as passed in via arguments and copy to user memory. On
82 * success, extent count on fieinfo is incremented.
84 * Returns 0 on success, -errno on error, 1 if this was the last
85 * extent that will fit in user array.
87 #define SET_UNKNOWN_FLAGS (FIEMAP_EXTENT_DELALLOC)
88 #define SET_NO_UNMOUNTED_IO_FLAGS (FIEMAP_EXTENT_DATA_ENCRYPTED)
89 #define SET_NOT_ALIGNED_FLAGS (FIEMAP_EXTENT_DATA_TAIL|FIEMAP_EXTENT_DATA_INLINE)
90 int fiemap_fill_next_extent(struct fiemap_extent_info *fieinfo, u64 logical,
91 u64 phys, u64 len, u32 flags)
93 struct fiemap_extent extent;
94 struct fiemap_extent __user *dest = fieinfo->fi_extents_start;
96 /* only count the extents */
97 if (fieinfo->fi_extents_max == 0) {
98 fieinfo->fi_extents_mapped++;
99 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
102 if (fieinfo->fi_extents_mapped >= fieinfo->fi_extents_max)
105 if (flags & SET_UNKNOWN_FLAGS)
106 flags |= FIEMAP_EXTENT_UNKNOWN;
107 if (flags & SET_NO_UNMOUNTED_IO_FLAGS)
108 flags |= FIEMAP_EXTENT_ENCODED;
109 if (flags & SET_NOT_ALIGNED_FLAGS)
110 flags |= FIEMAP_EXTENT_NOT_ALIGNED;
112 memset(&extent, 0, sizeof(extent));
113 extent.fe_logical = logical;
114 extent.fe_physical = phys;
115 extent.fe_length = len;
116 extent.fe_flags = flags;
118 dest += fieinfo->fi_extents_mapped;
119 if (copy_to_user(dest, &extent, sizeof(extent)))
122 fieinfo->fi_extents_mapped++;
123 if (fieinfo->fi_extents_mapped == fieinfo->fi_extents_max)
125 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
127 EXPORT_SYMBOL(fiemap_fill_next_extent);
130 * fiemap_check_flags - check validity of requested flags for fiemap
131 * @fieinfo: Fiemap context passed into ->fiemap
132 * @fs_flags: Set of fiemap flags that the file system understands
134 * Called from file system ->fiemap callback. This will compute the
135 * intersection of valid fiemap flags and those that the fs supports. That
136 * value is then compared against the user supplied flags. In case of bad user
137 * flags, the invalid values will be written into the fieinfo structure, and
138 * -EBADR is returned, which tells ioctl_fiemap() to return those values to
139 * userspace. For this reason, a return code of -EBADR should be preserved.
141 * Returns 0 on success, -EBADR on bad flags.
143 int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags)
147 incompat_flags = fieinfo->fi_flags & ~(FIEMAP_FLAGS_COMPAT & fs_flags);
148 if (incompat_flags) {
149 fieinfo->fi_flags = incompat_flags;
154 EXPORT_SYMBOL(fiemap_check_flags);
156 static int fiemap_check_ranges(struct super_block *sb,
157 u64 start, u64 len, u64 *new_len)
159 u64 maxbytes = (u64) sb->s_maxbytes;
166 if (start > maxbytes)
170 * Shrink request scope to what the fs can actually handle.
172 if (len > maxbytes || (maxbytes - len) < start)
173 *new_len = maxbytes - start;
178 static int ioctl_fiemap(struct file *filp, struct fiemap __user *ufiemap)
180 struct fiemap fiemap;
181 struct fiemap_extent_info fieinfo = { 0, };
182 struct inode *inode = file_inode(filp);
183 struct super_block *sb = inode->i_sb;
187 if (!inode->i_op->fiemap)
190 if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
193 if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
196 error = fiemap_check_ranges(sb, fiemap.fm_start, fiemap.fm_length,
201 fieinfo.fi_flags = fiemap.fm_flags;
202 fieinfo.fi_extents_max = fiemap.fm_extent_count;
203 fieinfo.fi_extents_start = ufiemap->fm_extents;
205 if (fiemap.fm_extent_count != 0 &&
206 !access_ok(fieinfo.fi_extents_start,
207 fieinfo.fi_extents_max * sizeof(struct fiemap_extent)))
210 if (fieinfo.fi_flags & FIEMAP_FLAG_SYNC)
211 filemap_write_and_wait(inode->i_mapping);
213 error = inode->i_op->fiemap(inode, &fieinfo, fiemap.fm_start, len);
214 fiemap.fm_flags = fieinfo.fi_flags;
215 fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
216 if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
222 static long ioctl_file_clone(struct file *dst_file, unsigned long srcfd,
223 u64 off, u64 olen, u64 destoff)
225 struct fd src_file = fdget(srcfd);
232 if (src_file.file->f_path.mnt != dst_file->f_path.mnt)
234 cloned = vfs_clone_file_range(src_file.file, off, dst_file, destoff,
238 else if (olen && cloned != olen)
247 static long ioctl_file_clone_range(struct file *file,
248 struct file_clone_range __user *argp)
250 struct file_clone_range args;
252 if (copy_from_user(&args, argp, sizeof(args)))
254 return ioctl_file_clone(file, args.src_fd, args.src_offset,
255 args.src_length, args.dest_offset);
260 static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
262 return (offset >> inode->i_blkbits);
265 static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
267 return (blk << inode->i_blkbits);
271 * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
272 * @inode: the inode to map
273 * @fieinfo: the fiemap info struct that will be passed back to userspace
274 * @start: where to start mapping in the inode
275 * @len: how much space to map
276 * @get_block: the fs's get_block function
278 * This does FIEMAP for block based inodes. Basically it will just loop
279 * through get_block until we hit the number of extents we want to map, or we
280 * go past the end of the file and hit a hole.
282 * If it is possible to have data blocks beyond a hole past @inode->i_size, then
283 * please do not use this function, it will stop at the first unmapped block
286 * If you use this function directly, you need to do your own locking. Use
287 * generic_block_fiemap if you want the locking done for you.
290 int __generic_block_fiemap(struct inode *inode,
291 struct fiemap_extent_info *fieinfo, loff_t start,
292 loff_t len, get_block_t *get_block)
294 struct buffer_head map_bh;
295 sector_t start_blk, last_blk;
296 loff_t isize = i_size_read(inode);
297 u64 logical = 0, phys = 0, size = 0;
298 u32 flags = FIEMAP_EXTENT_MERGED;
299 bool past_eof = false, whole_file = false;
302 ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC);
307 * Either the i_mutex or other appropriate locking needs to be held
308 * since we expect isize to not change at all through the duration of
317 * Some filesystems can't deal with being asked to map less than
318 * blocksize, so make sure our len is at least block length.
320 if (logical_to_blk(inode, len) == 0)
321 len = blk_to_logical(inode, 1);
323 start_blk = logical_to_blk(inode, start);
324 last_blk = logical_to_blk(inode, start + len - 1);
328 * we set b_size to the total size we want so it will map as
329 * many contiguous blocks as possible at once
331 memset(&map_bh, 0, sizeof(struct buffer_head));
334 ret = get_block(inode, start_blk, &map_bh, 0);
339 if (!buffer_mapped(&map_bh)) {
343 * We want to handle the case where there is an
344 * allocated block at the front of the file, and then
345 * nothing but holes up to the end of the file properly,
346 * to make sure that extent at the front gets properly
347 * marked with FIEMAP_EXTENT_LAST
350 blk_to_logical(inode, start_blk) >= isize)
354 * First hole after going past the EOF, this is our
357 if (past_eof && size) {
358 flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
359 ret = fiemap_fill_next_extent(fieinfo, logical,
363 ret = fiemap_fill_next_extent(fieinfo, logical,
368 /* if we have holes up to/past EOF then we're done */
369 if (start_blk > last_blk || past_eof || ret)
373 * We have gone over the length of what we wanted to
374 * map, and it wasn't the entire file, so add the extent
375 * we got last time and exit.
377 * This is for the case where say we want to map all the
378 * way up to the second to the last block in a file, but
379 * the last block is a hole, making the second to last
380 * block FIEMAP_EXTENT_LAST. In this case we want to
381 * see if there is a hole after the second to last block
382 * so we can mark it properly. If we found data after
383 * we exceeded the length we were requesting, then we
384 * are good to go, just add the extent to the fieinfo
387 if (start_blk > last_blk && !whole_file) {
388 ret = fiemap_fill_next_extent(fieinfo, logical,
395 * if size != 0 then we know we already have an extent
399 ret = fiemap_fill_next_extent(fieinfo, logical,
406 logical = blk_to_logical(inode, start_blk);
407 phys = blk_to_logical(inode, map_bh.b_blocknr);
408 size = map_bh.b_size;
409 flags = FIEMAP_EXTENT_MERGED;
411 start_blk += logical_to_blk(inode, size);
414 * If we are past the EOF, then we need to make sure as
415 * soon as we find a hole that the last extent we found
416 * is marked with FIEMAP_EXTENT_LAST
418 if (!past_eof && logical + size >= isize)
422 if (fatal_signal_pending(current)) {
429 /* If ret is 1 then we just hit the end of the extent array */
435 EXPORT_SYMBOL(__generic_block_fiemap);
438 * generic_block_fiemap - FIEMAP for block based inodes
439 * @inode: The inode to map
440 * @fieinfo: The mapping information
441 * @start: The initial block to map
442 * @len: The length of the extect to attempt to map
443 * @get_block: The block mapping function for the fs
445 * Calls __generic_block_fiemap to map the inode, after taking
446 * the inode's mutex lock.
449 int generic_block_fiemap(struct inode *inode,
450 struct fiemap_extent_info *fieinfo, u64 start,
451 u64 len, get_block_t *get_block)
455 ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
459 EXPORT_SYMBOL(generic_block_fiemap);
461 #endif /* CONFIG_BLOCK */
464 * This provides compatibility with legacy XFS pre-allocation ioctls
465 * which predate the fallocate syscall.
467 * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
468 * are used here, rest are ignored.
470 static int ioctl_preallocate(struct file *filp, int mode, void __user *argp)
472 struct inode *inode = file_inode(filp);
473 struct space_resv sr;
475 if (copy_from_user(&sr, argp, sizeof(sr)))
478 switch (sr.l_whence) {
482 sr.l_start += filp->f_pos;
485 sr.l_start += i_size_read(inode);
491 return vfs_fallocate(filp, mode | FALLOC_FL_KEEP_SIZE, sr.l_start,
495 /* on ia32 l_start is on a 32-bit boundary */
496 #if defined CONFIG_COMPAT && defined(CONFIG_X86_64)
497 /* just account for different alignment */
498 static int compat_ioctl_preallocate(struct file *file, int mode,
499 struct space_resv_32 __user *argp)
501 struct inode *inode = file_inode(file);
502 struct space_resv_32 sr;
504 if (copy_from_user(&sr, argp, sizeof(sr)))
507 switch (sr.l_whence) {
511 sr.l_start += file->f_pos;
514 sr.l_start += i_size_read(inode);
520 return vfs_fallocate(file, mode | FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
524 static int file_ioctl(struct file *filp, unsigned int cmd, int __user *p)
526 struct inode *inode = file_inode(filp);
530 return ioctl_fibmap(filp, p);
532 return put_user(i_size_read(inode) - filp->f_pos, p);
534 case FS_IOC_RESVSP64:
535 return ioctl_preallocate(filp, 0, p);
536 case FS_IOC_UNRESVSP:
537 case FS_IOC_UNRESVSP64:
538 return ioctl_preallocate(filp, FALLOC_FL_PUNCH_HOLE, p);
539 case FS_IOC_ZERO_RANGE:
540 return ioctl_preallocate(filp, FALLOC_FL_ZERO_RANGE, p);
546 static int ioctl_fionbio(struct file *filp, int __user *argp)
551 error = get_user(on, argp);
556 /* SunOS compatibility item. */
557 if (O_NONBLOCK != O_NDELAY)
560 spin_lock(&filp->f_lock);
562 filp->f_flags |= flag;
564 filp->f_flags &= ~flag;
565 spin_unlock(&filp->f_lock);
569 static int ioctl_fioasync(unsigned int fd, struct file *filp,
575 error = get_user(on, argp);
578 flag = on ? FASYNC : 0;
580 /* Did FASYNC state change ? */
581 if ((flag ^ filp->f_flags) & FASYNC) {
582 if (filp->f_op->fasync)
583 /* fasync() adjusts filp->f_flags */
584 error = filp->f_op->fasync(fd, filp, on);
588 return error < 0 ? error : 0;
591 static int ioctl_fsfreeze(struct file *filp)
593 struct super_block *sb = file_inode(filp)->i_sb;
595 if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN))
598 /* If filesystem doesn't support freeze feature, return. */
599 if (sb->s_op->freeze_fs == NULL && sb->s_op->freeze_super == NULL)
603 if (sb->s_op->freeze_super)
604 return sb->s_op->freeze_super(sb);
605 return freeze_super(sb);
608 static int ioctl_fsthaw(struct file *filp)
610 struct super_block *sb = file_inode(filp)->i_sb;
612 if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN))
616 if (sb->s_op->thaw_super)
617 return sb->s_op->thaw_super(sb);
618 return thaw_super(sb);
621 static int ioctl_file_dedupe_range(struct file *file,
622 struct file_dedupe_range __user *argp)
624 struct file_dedupe_range *same = NULL;
629 if (get_user(count, &argp->dest_count)) {
634 size = offsetof(struct file_dedupe_range __user, info[count]);
635 if (size > PAGE_SIZE) {
640 same = memdup_user(argp, size);
647 same->dest_count = count;
648 ret = vfs_dedupe_file_range(file, same);
652 ret = copy_to_user(argp, same, size);
662 * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
663 * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
665 * When you add any new common ioctls to the switches above and below,
666 * please ensure they have compatible arguments in compat mode.
668 static int do_vfs_ioctl(struct file *filp, unsigned int fd,
669 unsigned int cmd, unsigned long arg)
671 void __user *argp = (void __user *)arg;
672 struct inode *inode = file_inode(filp);
676 set_close_on_exec(fd, 1);
680 set_close_on_exec(fd, 0);
684 return ioctl_fionbio(filp, argp);
687 return ioctl_fioasync(fd, filp, argp);
690 if (S_ISDIR(inode->i_mode) || S_ISREG(inode->i_mode) ||
691 S_ISLNK(inode->i_mode)) {
692 loff_t res = inode_get_bytes(inode);
693 return copy_to_user(argp, &res, sizeof(res)) ?
700 return ioctl_fsfreeze(filp);
703 return ioctl_fsthaw(filp);
706 return ioctl_fiemap(filp, argp);
709 /* anon_bdev filesystems may not have a block size */
710 if (!inode->i_sb->s_blocksize)
713 return put_user(inode->i_sb->s_blocksize, (int __user *)argp);
716 return ioctl_file_clone(filp, arg, 0, 0, 0);
719 return ioctl_file_clone_range(filp, argp);
722 return ioctl_file_dedupe_range(filp, argp);
725 if (S_ISREG(inode->i_mode))
726 return file_ioctl(filp, cmd, argp);
733 int ksys_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
735 struct fd f = fdget(fd);
741 error = security_file_ioctl(f.file, cmd, arg);
745 error = do_vfs_ioctl(f.file, fd, cmd, arg);
746 if (error == -ENOIOCTLCMD)
747 error = vfs_ioctl(f.file, cmd, arg);
754 SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
756 return ksys_ioctl(fd, cmd, arg);
761 * compat_ptr_ioctl - generic implementation of .compat_ioctl file operation
763 * This is not normally called as a function, but instead set in struct
766 * .compat_ioctl = compat_ptr_ioctl,
768 * On most architectures, the compat_ptr_ioctl() just passes all arguments
769 * to the corresponding ->ioctl handler. The exception is arch/s390, where
770 * compat_ptr() clears the top bit of a 32-bit pointer value, so user space
771 * pointers to the second 2GB alias the first 2GB, as is the case for
772 * native 32-bit s390 user space.
774 * The compat_ptr_ioctl() function must therefore be used only with ioctl
775 * functions that either ignore the argument or pass a pointer to a
776 * compatible data type.
778 * If any ioctl command handled by fops->unlocked_ioctl passes a plain
779 * integer instead of a pointer, or any of the passed data types
780 * is incompatible between 32-bit and 64-bit architectures, a proper
781 * handler is required instead of compat_ptr_ioctl.
783 long compat_ptr_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
785 if (!file->f_op->unlocked_ioctl)
788 return file->f_op->unlocked_ioctl(file, cmd, (unsigned long)compat_ptr(arg));
790 EXPORT_SYMBOL(compat_ptr_ioctl);
792 COMPAT_SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd,
795 struct fd f = fdget(fd);
801 /* RED-PEN how should LSM module know it's handling 32bit? */
802 error = security_file_ioctl(f.file, cmd, arg);
807 /* FICLONE takes an int argument, so don't use compat_ptr() */
809 error = ioctl_file_clone(f.file, arg, 0, 0, 0);
812 #if defined(CONFIG_X86_64)
813 /* these get messy on amd64 due to alignment differences */
814 case FS_IOC_RESVSP_32:
815 case FS_IOC_RESVSP64_32:
816 error = compat_ioctl_preallocate(f.file, 0, compat_ptr(arg));
818 case FS_IOC_UNRESVSP_32:
819 case FS_IOC_UNRESVSP64_32:
820 error = compat_ioctl_preallocate(f.file, FALLOC_FL_PUNCH_HOLE,
823 case FS_IOC_ZERO_RANGE_32:
824 error = compat_ioctl_preallocate(f.file, FALLOC_FL_ZERO_RANGE,
830 * everything else in do_vfs_ioctl() takes either a compatible
831 * pointer argument or no argument -- call it with a modified
835 error = do_vfs_ioctl(f.file, fd, cmd,
836 (unsigned long)compat_ptr(arg));
837 if (error != -ENOIOCTLCMD)
840 if (f.file->f_op->compat_ioctl)
841 error = f.file->f_op->compat_ioctl(f.file, cmd, arg);
842 if (error == -ENOIOCTLCMD)