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>
21 #include <linux/fiemap.h>
25 #include <asm/ioctls.h>
27 /* So that the fiemap access checks can't overflow on 32 bit machines. */
28 #define FIEMAP_MAX_EXTENTS (UINT_MAX / sizeof(struct fiemap_extent))
31 * vfs_ioctl - call filesystem specific ioctl methods
32 * @filp: open file to invoke ioctl method on
33 * @cmd: ioctl command to execute
34 * @arg: command-specific argument for ioctl
36 * Invokes filesystem specific ->unlocked_ioctl, if one exists; otherwise
39 * Returns 0 on success, -errno on error.
41 long vfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
45 if (!filp->f_op->unlocked_ioctl)
48 error = filp->f_op->unlocked_ioctl(filp, cmd, arg);
49 if (error == -ENOIOCTLCMD)
54 EXPORT_SYMBOL(vfs_ioctl);
56 static int ioctl_fibmap(struct file *filp, int __user *p)
58 struct inode *inode = file_inode(filp);
59 struct super_block *sb = inode->i_sb;
63 if (!capable(CAP_SYS_RAWIO))
66 error = get_user(ur_block, p);
74 error = bmap(inode, &block);
76 if (block > INT_MAX) {
78 pr_warn_ratelimited("[%s/%d] FS: %s File: %pD4 would truncate fibmap result\n",
79 current->comm, task_pid_nr(current),
88 if (put_user(ur_block, p))
95 * fiemap_fill_next_extent - Fiemap helper function
96 * @fieinfo: Fiemap context passed into ->fiemap
97 * @logical: Extent logical start offset, in bytes
98 * @phys: Extent physical start offset, in bytes
99 * @len: Extent length, in bytes
100 * @flags: FIEMAP_EXTENT flags that describe this extent
102 * Called from file system ->fiemap callback. Will populate extent
103 * info as passed in via arguments and copy to user memory. On
104 * success, extent count on fieinfo is incremented.
106 * Returns 0 on success, -errno on error, 1 if this was the last
107 * extent that will fit in user array.
109 #define SET_UNKNOWN_FLAGS (FIEMAP_EXTENT_DELALLOC)
110 #define SET_NO_UNMOUNTED_IO_FLAGS (FIEMAP_EXTENT_DATA_ENCRYPTED)
111 #define SET_NOT_ALIGNED_FLAGS (FIEMAP_EXTENT_DATA_TAIL|FIEMAP_EXTENT_DATA_INLINE)
112 int fiemap_fill_next_extent(struct fiemap_extent_info *fieinfo, u64 logical,
113 u64 phys, u64 len, u32 flags)
115 struct fiemap_extent extent;
116 struct fiemap_extent __user *dest = fieinfo->fi_extents_start;
118 /* only count the extents */
119 if (fieinfo->fi_extents_max == 0) {
120 fieinfo->fi_extents_mapped++;
121 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
124 if (fieinfo->fi_extents_mapped >= fieinfo->fi_extents_max)
127 if (flags & SET_UNKNOWN_FLAGS)
128 flags |= FIEMAP_EXTENT_UNKNOWN;
129 if (flags & SET_NO_UNMOUNTED_IO_FLAGS)
130 flags |= FIEMAP_EXTENT_ENCODED;
131 if (flags & SET_NOT_ALIGNED_FLAGS)
132 flags |= FIEMAP_EXTENT_NOT_ALIGNED;
134 memset(&extent, 0, sizeof(extent));
135 extent.fe_logical = logical;
136 extent.fe_physical = phys;
137 extent.fe_length = len;
138 extent.fe_flags = flags;
140 dest += fieinfo->fi_extents_mapped;
141 if (copy_to_user(dest, &extent, sizeof(extent)))
144 fieinfo->fi_extents_mapped++;
145 if (fieinfo->fi_extents_mapped == fieinfo->fi_extents_max)
147 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
149 EXPORT_SYMBOL(fiemap_fill_next_extent);
152 * fiemap_prep - check validity of requested flags for fiemap
153 * @inode: Inode to operate on
154 * @fieinfo: Fiemap context passed into ->fiemap
155 * @start: Start of the mapped range
156 * @len: Length of the mapped range, can be truncated by this function.
157 * @supported_flags: Set of fiemap flags that the file system understands
159 * This function must be called from each ->fiemap instance to validate the
160 * fiemap request against the file system parameters.
162 * Returns 0 on success, or a negative error on failure.
164 int fiemap_prep(struct inode *inode, struct fiemap_extent_info *fieinfo,
165 u64 start, u64 *len, u32 supported_flags)
167 u64 maxbytes = inode->i_sb->s_maxbytes;
173 if (start > maxbytes)
177 * Shrink request scope to what the fs can actually handle.
179 if (*len > maxbytes || (maxbytes - *len) < start)
180 *len = maxbytes - start;
182 supported_flags |= FIEMAP_FLAG_SYNC;
183 supported_flags &= FIEMAP_FLAGS_COMPAT;
184 incompat_flags = fieinfo->fi_flags & ~supported_flags;
185 if (incompat_flags) {
186 fieinfo->fi_flags = incompat_flags;
190 if (fieinfo->fi_flags & FIEMAP_FLAG_SYNC)
191 ret = filemap_write_and_wait(inode->i_mapping);
194 EXPORT_SYMBOL(fiemap_prep);
196 static int ioctl_fiemap(struct file *filp, struct fiemap __user *ufiemap)
198 struct fiemap fiemap;
199 struct fiemap_extent_info fieinfo = { 0, };
200 struct inode *inode = file_inode(filp);
203 if (!inode->i_op->fiemap)
206 if (copy_from_user(&fiemap, ufiemap, sizeof(fiemap)))
209 if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
212 fieinfo.fi_flags = fiemap.fm_flags;
213 fieinfo.fi_extents_max = fiemap.fm_extent_count;
214 fieinfo.fi_extents_start = ufiemap->fm_extents;
216 error = inode->i_op->fiemap(inode, &fieinfo, fiemap.fm_start,
219 fiemap.fm_flags = fieinfo.fi_flags;
220 fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
221 if (copy_to_user(ufiemap, &fiemap, sizeof(fiemap)))
227 static long ioctl_file_clone(struct file *dst_file, unsigned long srcfd,
228 u64 off, u64 olen, u64 destoff)
230 struct fd src_file = fdget(srcfd);
237 if (src_file.file->f_path.mnt != dst_file->f_path.mnt)
239 cloned = vfs_clone_file_range(src_file.file, off, dst_file, destoff,
243 else if (olen && cloned != olen)
252 static long ioctl_file_clone_range(struct file *file,
253 struct file_clone_range __user *argp)
255 struct file_clone_range args;
257 if (copy_from_user(&args, argp, sizeof(args)))
259 return ioctl_file_clone(file, args.src_fd, args.src_offset,
260 args.src_length, args.dest_offset);
265 static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
267 return (offset >> inode->i_blkbits);
270 static inline loff_t blk_to_logical(struct inode *inode, sector_t blk)
272 return (blk << inode->i_blkbits);
276 * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
277 * @inode: the inode to map
278 * @fieinfo: the fiemap info struct that will be passed back to userspace
279 * @start: where to start mapping in the inode
280 * @len: how much space to map
281 * @get_block: the fs's get_block function
283 * This does FIEMAP for block based inodes. Basically it will just loop
284 * through get_block until we hit the number of extents we want to map, or we
285 * go past the end of the file and hit a hole.
287 * If it is possible to have data blocks beyond a hole past @inode->i_size, then
288 * please do not use this function, it will stop at the first unmapped block
291 * If you use this function directly, you need to do your own locking. Use
292 * generic_block_fiemap if you want the locking done for you.
294 static int __generic_block_fiemap(struct inode *inode,
295 struct fiemap_extent_info *fieinfo, loff_t start,
296 loff_t len, get_block_t *get_block)
298 struct buffer_head map_bh;
299 sector_t start_blk, last_blk;
300 loff_t isize = i_size_read(inode);
301 u64 logical = 0, phys = 0, size = 0;
302 u32 flags = FIEMAP_EXTENT_MERGED;
303 bool past_eof = false, whole_file = false;
306 ret = fiemap_prep(inode, fieinfo, start, &len, FIEMAP_FLAG_SYNC);
311 * Either the i_mutex or other appropriate locking needs to be held
312 * since we expect isize to not change at all through the duration of
321 * Some filesystems can't deal with being asked to map less than
322 * blocksize, so make sure our len is at least block length.
324 if (logical_to_blk(inode, len) == 0)
325 len = blk_to_logical(inode, 1);
327 start_blk = logical_to_blk(inode, start);
328 last_blk = logical_to_blk(inode, start + len - 1);
332 * we set b_size to the total size we want so it will map as
333 * many contiguous blocks as possible at once
335 memset(&map_bh, 0, sizeof(struct buffer_head));
338 ret = get_block(inode, start_blk, &map_bh, 0);
343 if (!buffer_mapped(&map_bh)) {
347 * We want to handle the case where there is an
348 * allocated block at the front of the file, and then
349 * nothing but holes up to the end of the file properly,
350 * to make sure that extent at the front gets properly
351 * marked with FIEMAP_EXTENT_LAST
354 blk_to_logical(inode, start_blk) >= isize)
358 * First hole after going past the EOF, this is our
361 if (past_eof && size) {
362 flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
363 ret = fiemap_fill_next_extent(fieinfo, logical,
367 ret = fiemap_fill_next_extent(fieinfo, logical,
372 /* if we have holes up to/past EOF then we're done */
373 if (start_blk > last_blk || past_eof || ret)
377 * We have gone over the length of what we wanted to
378 * map, and it wasn't the entire file, so add the extent
379 * we got last time and exit.
381 * This is for the case where say we want to map all the
382 * way up to the second to the last block in a file, but
383 * the last block is a hole, making the second to last
384 * block FIEMAP_EXTENT_LAST. In this case we want to
385 * see if there is a hole after the second to last block
386 * so we can mark it properly. If we found data after
387 * we exceeded the length we were requesting, then we
388 * are good to go, just add the extent to the fieinfo
391 if (start_blk > last_blk && !whole_file) {
392 ret = fiemap_fill_next_extent(fieinfo, logical,
399 * if size != 0 then we know we already have an extent
403 ret = fiemap_fill_next_extent(fieinfo, logical,
410 logical = blk_to_logical(inode, start_blk);
411 phys = blk_to_logical(inode, map_bh.b_blocknr);
412 size = map_bh.b_size;
413 flags = FIEMAP_EXTENT_MERGED;
415 start_blk += logical_to_blk(inode, size);
418 * If we are past the EOF, then we need to make sure as
419 * soon as we find a hole that the last extent we found
420 * is marked with FIEMAP_EXTENT_LAST
422 if (!past_eof && logical + size >= isize)
426 if (fatal_signal_pending(current)) {
433 /* If ret is 1 then we just hit the end of the extent array */
441 * generic_block_fiemap - FIEMAP for block based inodes
442 * @inode: The inode to map
443 * @fieinfo: The mapping information
444 * @start: The initial block to map
445 * @len: The length of the extect to attempt to map
446 * @get_block: The block mapping function for the fs
448 * Calls __generic_block_fiemap to map the inode, after taking
449 * the inode's mutex lock.
452 int generic_block_fiemap(struct inode *inode,
453 struct fiemap_extent_info *fieinfo, u64 start,
454 u64 len, get_block_t *get_block)
458 ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
462 EXPORT_SYMBOL(generic_block_fiemap);
464 #endif /* CONFIG_BLOCK */
467 * This provides compatibility with legacy XFS pre-allocation ioctls
468 * which predate the fallocate syscall.
470 * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
471 * are used here, rest are ignored.
473 static int ioctl_preallocate(struct file *filp, int mode, void __user *argp)
475 struct inode *inode = file_inode(filp);
476 struct space_resv sr;
478 if (copy_from_user(&sr, argp, sizeof(sr)))
481 switch (sr.l_whence) {
485 sr.l_start += filp->f_pos;
488 sr.l_start += i_size_read(inode);
494 return vfs_fallocate(filp, mode | FALLOC_FL_KEEP_SIZE, sr.l_start,
498 /* on ia32 l_start is on a 32-bit boundary */
499 #if defined CONFIG_COMPAT && defined(CONFIG_X86_64)
500 /* just account for different alignment */
501 static int compat_ioctl_preallocate(struct file *file, int mode,
502 struct space_resv_32 __user *argp)
504 struct inode *inode = file_inode(file);
505 struct space_resv_32 sr;
507 if (copy_from_user(&sr, argp, sizeof(sr)))
510 switch (sr.l_whence) {
514 sr.l_start += file->f_pos;
517 sr.l_start += i_size_read(inode);
523 return vfs_fallocate(file, mode | FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
527 static int file_ioctl(struct file *filp, unsigned int cmd, int __user *p)
531 return ioctl_fibmap(filp, p);
533 case FS_IOC_RESVSP64:
534 return ioctl_preallocate(filp, 0, p);
535 case FS_IOC_UNRESVSP:
536 case FS_IOC_UNRESVSP64:
537 return ioctl_preallocate(filp, FALLOC_FL_PUNCH_HOLE, p);
538 case FS_IOC_ZERO_RANGE:
539 return ioctl_preallocate(filp, FALLOC_FL_ZERO_RANGE, p);
545 static int ioctl_fionbio(struct file *filp, int __user *argp)
550 error = get_user(on, argp);
555 /* SunOS compatibility item. */
556 if (O_NONBLOCK != O_NDELAY)
559 spin_lock(&filp->f_lock);
561 filp->f_flags |= flag;
563 filp->f_flags &= ~flag;
564 spin_unlock(&filp->f_lock);
568 static int ioctl_fioasync(unsigned int fd, struct file *filp,
574 error = get_user(on, argp);
577 flag = on ? FASYNC : 0;
579 /* Did FASYNC state change ? */
580 if ((flag ^ filp->f_flags) & FASYNC) {
581 if (filp->f_op->fasync)
582 /* fasync() adjusts filp->f_flags */
583 error = filp->f_op->fasync(fd, filp, on);
587 return error < 0 ? error : 0;
590 static int ioctl_fsfreeze(struct file *filp)
592 struct super_block *sb = file_inode(filp)->i_sb;
594 if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN))
597 /* If filesystem doesn't support freeze feature, return. */
598 if (sb->s_op->freeze_fs == NULL && sb->s_op->freeze_super == NULL)
602 if (sb->s_op->freeze_super)
603 return sb->s_op->freeze_super(sb);
604 return freeze_super(sb);
607 static int ioctl_fsthaw(struct file *filp)
609 struct super_block *sb = file_inode(filp)->i_sb;
611 if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN))
615 if (sb->s_op->thaw_super)
616 return sb->s_op->thaw_super(sb);
617 return thaw_super(sb);
620 static int ioctl_file_dedupe_range(struct file *file,
621 struct file_dedupe_range __user *argp)
623 struct file_dedupe_range *same = NULL;
628 if (get_user(count, &argp->dest_count)) {
633 size = offsetof(struct file_dedupe_range __user, info[count]);
634 if (size > PAGE_SIZE) {
639 same = memdup_user(argp, size);
646 same->dest_count = count;
647 ret = vfs_dedupe_file_range(file, same);
651 ret = copy_to_user(argp, same, size);
661 * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
662 * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
664 * When you add any new common ioctls to the switches above and below,
665 * please ensure they have compatible arguments in compat mode.
667 static int do_vfs_ioctl(struct file *filp, unsigned int fd,
668 unsigned int cmd, unsigned long arg)
670 void __user *argp = (void __user *)arg;
671 struct inode *inode = file_inode(filp);
675 set_close_on_exec(fd, 1);
679 set_close_on_exec(fd, 0);
683 return ioctl_fionbio(filp, argp);
686 return ioctl_fioasync(fd, filp, argp);
689 if (S_ISDIR(inode->i_mode) || S_ISREG(inode->i_mode) ||
690 S_ISLNK(inode->i_mode)) {
691 loff_t res = inode_get_bytes(inode);
692 return copy_to_user(argp, &res, sizeof(res)) ?
699 return ioctl_fsfreeze(filp);
702 return ioctl_fsthaw(filp);
705 return ioctl_fiemap(filp, argp);
708 /* anon_bdev filesystems may not have a block size */
709 if (!inode->i_sb->s_blocksize)
712 return put_user(inode->i_sb->s_blocksize, (int __user *)argp);
715 return ioctl_file_clone(filp, arg, 0, 0, 0);
718 return ioctl_file_clone_range(filp, argp);
721 return ioctl_file_dedupe_range(filp, argp);
724 if (!S_ISREG(inode->i_mode))
725 return vfs_ioctl(filp, cmd, arg);
727 return put_user(i_size_read(inode) - filp->f_pos,
731 if (S_ISREG(inode->i_mode))
732 return file_ioctl(filp, cmd, argp);
739 int ksys_ioctl(unsigned int fd, unsigned int cmd, unsigned long arg)
741 struct fd f = fdget(fd);
747 error = security_file_ioctl(f.file, cmd, arg);
751 error = do_vfs_ioctl(f.file, fd, cmd, arg);
752 if (error == -ENOIOCTLCMD)
753 error = vfs_ioctl(f.file, cmd, arg);
760 SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
762 return ksys_ioctl(fd, cmd, arg);
767 * compat_ptr_ioctl - generic implementation of .compat_ioctl file operation
769 * This is not normally called as a function, but instead set in struct
772 * .compat_ioctl = compat_ptr_ioctl,
774 * On most architectures, the compat_ptr_ioctl() just passes all arguments
775 * to the corresponding ->ioctl handler. The exception is arch/s390, where
776 * compat_ptr() clears the top bit of a 32-bit pointer value, so user space
777 * pointers to the second 2GB alias the first 2GB, as is the case for
778 * native 32-bit s390 user space.
780 * The compat_ptr_ioctl() function must therefore be used only with ioctl
781 * functions that either ignore the argument or pass a pointer to a
782 * compatible data type.
784 * If any ioctl command handled by fops->unlocked_ioctl passes a plain
785 * integer instead of a pointer, or any of the passed data types
786 * is incompatible between 32-bit and 64-bit architectures, a proper
787 * handler is required instead of compat_ptr_ioctl.
789 long compat_ptr_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
791 if (!file->f_op->unlocked_ioctl)
794 return file->f_op->unlocked_ioctl(file, cmd, (unsigned long)compat_ptr(arg));
796 EXPORT_SYMBOL(compat_ptr_ioctl);
798 COMPAT_SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd,
801 struct fd f = fdget(fd);
807 /* RED-PEN how should LSM module know it's handling 32bit? */
808 error = security_file_ioctl(f.file, cmd, arg);
813 /* FICLONE takes an int argument, so don't use compat_ptr() */
815 error = ioctl_file_clone(f.file, arg, 0, 0, 0);
818 #if defined(CONFIG_X86_64)
819 /* these get messy on amd64 due to alignment differences */
820 case FS_IOC_RESVSP_32:
821 case FS_IOC_RESVSP64_32:
822 error = compat_ioctl_preallocate(f.file, 0, compat_ptr(arg));
824 case FS_IOC_UNRESVSP_32:
825 case FS_IOC_UNRESVSP64_32:
826 error = compat_ioctl_preallocate(f.file, FALLOC_FL_PUNCH_HOLE,
829 case FS_IOC_ZERO_RANGE_32:
830 error = compat_ioctl_preallocate(f.file, FALLOC_FL_ZERO_RANGE,
836 * everything else in do_vfs_ioctl() takes either a compatible
837 * pointer argument or no argument -- call it with a modified
841 error = do_vfs_ioctl(f.file, fd, cmd,
842 (unsigned long)compat_ptr(arg));
843 if (error != -ENOIOCTLCMD)
846 if (f.file->f_op->compat_ioctl)
847 error = f.file->f_op->compat_ioctl(f.file, cmd, arg);
848 if (error == -ENOIOCTLCMD)