4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/syscalls.h>
9 #include <linux/smp_lock.h>
10 #include <linux/capability.h>
11 #include <linux/file.h>
13 #include <linux/security.h>
14 #include <linux/module.h>
15 #include <linux/uaccess.h>
16 #include <linux/writeback.h>
17 #include <linux/buffer_head.h>
18 #include <linux/falloc.h>
20 #include <asm/ioctls.h>
22 /* So that the fiemap access checks can't overflow on 32 bit machines. */
23 #define FIEMAP_MAX_EXTENTS (UINT_MAX / sizeof(struct fiemap_extent))
26 * vfs_ioctl - call filesystem specific ioctl methods
27 * @filp: open file to invoke ioctl method on
28 * @cmd: ioctl command to execute
29 * @arg: command-specific argument for ioctl
31 * Invokes filesystem specific ->unlocked_ioctl, if one exists; otherwise
32 * invokes filesystem specific ->ioctl method. If neither method exists,
35 * Returns 0 on success, -errno on error.
37 static long vfs_ioctl(struct file *filp, unsigned int cmd,
45 if (filp->f_op->unlocked_ioctl) {
46 error = filp->f_op->unlocked_ioctl(filp, cmd, arg);
47 if (error == -ENOIOCTLCMD)
50 } else if (filp->f_op->ioctl) {
52 error = filp->f_op->ioctl(filp->f_path.dentry->d_inode,
61 static int ioctl_fibmap(struct file *filp, int __user *p)
63 struct address_space *mapping = filp->f_mapping;
66 /* do we support this mess? */
67 if (!mapping->a_ops->bmap)
69 if (!capable(CAP_SYS_RAWIO))
71 res = get_user(block, p);
74 res = mapping->a_ops->bmap(mapping, block);
75 return put_user(res, p);
79 * fiemap_fill_next_extent - Fiemap helper function
80 * @fieinfo: Fiemap context passed into ->fiemap
81 * @logical: Extent logical start offset, in bytes
82 * @phys: Extent physical start offset, in bytes
83 * @len: Extent length, in bytes
84 * @flags: FIEMAP_EXTENT flags that describe this extent
86 * Called from file system ->fiemap callback. Will populate extent
87 * info as passed in via arguments and copy to user memory. On
88 * success, extent count on fieinfo is incremented.
90 * Returns 0 on success, -errno on error, 1 if this was the last
91 * extent that will fit in user array.
93 #define SET_UNKNOWN_FLAGS (FIEMAP_EXTENT_DELALLOC)
94 #define SET_NO_UNMOUNTED_IO_FLAGS (FIEMAP_EXTENT_DATA_ENCRYPTED)
95 #define SET_NOT_ALIGNED_FLAGS (FIEMAP_EXTENT_DATA_TAIL|FIEMAP_EXTENT_DATA_INLINE)
96 int fiemap_fill_next_extent(struct fiemap_extent_info *fieinfo, u64 logical,
97 u64 phys, u64 len, u32 flags)
99 struct fiemap_extent extent;
100 struct fiemap_extent *dest = fieinfo->fi_extents_start;
102 /* only count the extents */
103 if (fieinfo->fi_extents_max == 0) {
104 fieinfo->fi_extents_mapped++;
105 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
108 if (fieinfo->fi_extents_mapped >= fieinfo->fi_extents_max)
111 if (flags & SET_UNKNOWN_FLAGS)
112 flags |= FIEMAP_EXTENT_UNKNOWN;
113 if (flags & SET_NO_UNMOUNTED_IO_FLAGS)
114 flags |= FIEMAP_EXTENT_ENCODED;
115 if (flags & SET_NOT_ALIGNED_FLAGS)
116 flags |= FIEMAP_EXTENT_NOT_ALIGNED;
118 memset(&extent, 0, sizeof(extent));
119 extent.fe_logical = logical;
120 extent.fe_physical = phys;
121 extent.fe_length = len;
122 extent.fe_flags = flags;
124 dest += fieinfo->fi_extents_mapped;
125 if (copy_to_user(dest, &extent, sizeof(extent)))
128 fieinfo->fi_extents_mapped++;
129 if (fieinfo->fi_extents_mapped == fieinfo->fi_extents_max)
131 return (flags & FIEMAP_EXTENT_LAST) ? 1 : 0;
133 EXPORT_SYMBOL(fiemap_fill_next_extent);
136 * fiemap_check_flags - check validity of requested flags for fiemap
137 * @fieinfo: Fiemap context passed into ->fiemap
138 * @fs_flags: Set of fiemap flags that the file system understands
140 * Called from file system ->fiemap callback. This will compute the
141 * intersection of valid fiemap flags and those that the fs supports. That
142 * value is then compared against the user supplied flags. In case of bad user
143 * flags, the invalid values will be written into the fieinfo structure, and
144 * -EBADR is returned, which tells ioctl_fiemap() to return those values to
145 * userspace. For this reason, a return code of -EBADR should be preserved.
147 * Returns 0 on success, -EBADR on bad flags.
149 int fiemap_check_flags(struct fiemap_extent_info *fieinfo, u32 fs_flags)
153 incompat_flags = fieinfo->fi_flags & ~(FIEMAP_FLAGS_COMPAT & fs_flags);
154 if (incompat_flags) {
155 fieinfo->fi_flags = incompat_flags;
160 EXPORT_SYMBOL(fiemap_check_flags);
162 static int fiemap_check_ranges(struct super_block *sb,
163 u64 start, u64 len, u64 *new_len)
165 u64 maxbytes = (u64) sb->s_maxbytes;
172 if (start > maxbytes)
176 * Shrink request scope to what the fs can actually handle.
178 if (len > maxbytes || (maxbytes - len) < start)
179 *new_len = maxbytes - start;
184 static int ioctl_fiemap(struct file *filp, unsigned long arg)
186 struct fiemap fiemap;
187 struct fiemap_extent_info fieinfo = { 0, };
188 struct inode *inode = filp->f_path.dentry->d_inode;
189 struct super_block *sb = inode->i_sb;
193 if (!inode->i_op->fiemap)
196 if (copy_from_user(&fiemap, (struct fiemap __user *)arg,
197 sizeof(struct fiemap)))
200 if (fiemap.fm_extent_count > FIEMAP_MAX_EXTENTS)
203 error = fiemap_check_ranges(sb, fiemap.fm_start, fiemap.fm_length,
208 fieinfo.fi_flags = fiemap.fm_flags;
209 fieinfo.fi_extents_max = fiemap.fm_extent_count;
210 fieinfo.fi_extents_start = (struct fiemap_extent *)(arg + sizeof(fiemap));
212 if (fiemap.fm_extent_count != 0 &&
213 !access_ok(VERIFY_WRITE, fieinfo.fi_extents_start,
214 fieinfo.fi_extents_max * sizeof(struct fiemap_extent)))
217 if (fieinfo.fi_flags & FIEMAP_FLAG_SYNC)
218 filemap_write_and_wait(inode->i_mapping);
220 error = inode->i_op->fiemap(inode, &fieinfo, fiemap.fm_start, len);
221 fiemap.fm_flags = fieinfo.fi_flags;
222 fiemap.fm_mapped_extents = fieinfo.fi_extents_mapped;
223 if (copy_to_user((char *)arg, &fiemap, sizeof(fiemap)))
231 #define blk_to_logical(inode, blk) (blk << (inode)->i_blkbits)
232 #define logical_to_blk(inode, offset) (offset >> (inode)->i_blkbits);
235 * __generic_block_fiemap - FIEMAP for block based inodes (no locking)
236 * @inode - the inode to map
237 * @arg - the pointer to userspace where we copy everything to
238 * @get_block - the fs's get_block function
240 * This does FIEMAP for block based inodes. Basically it will just loop
241 * through get_block until we hit the number of extents we want to map, or we
242 * go past the end of the file and hit a hole.
244 * If it is possible to have data blocks beyond a hole past @inode->i_size, then
245 * please do not use this function, it will stop at the first unmapped block
248 * If you use this function directly, you need to do your own locking. Use
249 * generic_block_fiemap if you want the locking done for you.
252 int __generic_block_fiemap(struct inode *inode,
253 struct fiemap_extent_info *fieinfo, u64 start,
254 u64 len, get_block_t *get_block)
256 struct buffer_head tmp;
257 unsigned long long start_blk;
258 long long length = 0, map_len = 0;
259 u64 logical = 0, phys = 0, size = 0;
260 u32 flags = FIEMAP_EXTENT_MERGED;
261 int ret = 0, past_eof = 0, whole_file = 0;
263 if ((ret = fiemap_check_flags(fieinfo, FIEMAP_FLAG_SYNC)))
266 start_blk = logical_to_blk(inode, start);
268 length = (long long)min_t(u64, len, i_size_read(inode));
276 * we set b_size to the total size we want so it will map as
277 * many contiguous blocks as possible at once
279 memset(&tmp, 0, sizeof(struct buffer_head));
280 tmp.b_size = map_len;
282 ret = get_block(inode, start_blk, &tmp, 0);
287 if (!buffer_mapped(&tmp)) {
288 length -= blk_to_logical(inode, 1);
292 * we want to handle the case where there is an
293 * allocated block at the front of the file, and then
294 * nothing but holes up to the end of the file properly,
295 * to make sure that extent at the front gets properly
296 * marked with FIEMAP_EXTENT_LAST
299 blk_to_logical(inode, start_blk) >=
300 blk_to_logical(inode, 0)+i_size_read(inode))
304 * first hole after going past the EOF, this is our
307 if (past_eof && size) {
308 flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
309 ret = fiemap_fill_next_extent(fieinfo, logical,
315 /* if we have holes up to/past EOF then we're done */
316 if (length <= 0 || past_eof)
320 * we have gone over the length of what we wanted to
321 * map, and it wasn't the entire file, so add the extent
322 * we got last time and exit.
324 * This is for the case where say we want to map all the
325 * way up to the second to the last block in a file, but
326 * the last block is a hole, making the second to last
327 * block FIEMAP_EXTENT_LAST. In this case we want to
328 * see if there is a hole after the second to last block
329 * so we can mark it properly. If we found data after
330 * we exceeded the length we were requesting, then we
331 * are good to go, just add the extent to the fieinfo
334 if (length <= 0 && !whole_file) {
335 ret = fiemap_fill_next_extent(fieinfo, logical,
342 * if size != 0 then we know we already have an extent
346 ret = fiemap_fill_next_extent(fieinfo, logical,
353 logical = blk_to_logical(inode, start_blk);
354 phys = blk_to_logical(inode, tmp.b_blocknr);
356 flags = FIEMAP_EXTENT_MERGED;
358 length -= tmp.b_size;
359 start_blk += logical_to_blk(inode, size);
362 * If we are past the EOF, then we need to make sure as
363 * soon as we find a hole that the last extent we found
364 * is marked with FIEMAP_EXTENT_LAST
368 blk_to_logical(inode, 0)+i_size_read(inode))
374 /* if ret is 1 then we just hit the end of the extent array */
380 EXPORT_SYMBOL(__generic_block_fiemap);
383 * generic_block_fiemap - FIEMAP for block based inodes
384 * @inode: The inode to map
385 * @fieinfo: The mapping information
386 * @start: The initial block to map
387 * @len: The length of the extect to attempt to map
388 * @get_block: The block mapping function for the fs
390 * Calls __generic_block_fiemap to map the inode, after taking
391 * the inode's mutex lock.
394 int generic_block_fiemap(struct inode *inode,
395 struct fiemap_extent_info *fieinfo, u64 start,
396 u64 len, get_block_t *get_block)
399 mutex_lock(&inode->i_mutex);
400 ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
401 mutex_unlock(&inode->i_mutex);
404 EXPORT_SYMBOL(generic_block_fiemap);
406 #endif /* CONFIG_BLOCK */
409 * This provides compatibility with legacy XFS pre-allocation ioctls
410 * which predate the fallocate syscall.
412 * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
413 * are used here, rest are ignored.
415 int ioctl_preallocate(struct file *filp, void __user *argp)
417 struct inode *inode = filp->f_path.dentry->d_inode;
418 struct space_resv sr;
420 if (copy_from_user(&sr, argp, sizeof(sr)))
423 switch (sr.l_whence) {
427 sr.l_start += filp->f_pos;
430 sr.l_start += i_size_read(inode);
436 return do_fallocate(filp, FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
439 static int file_ioctl(struct file *filp, unsigned int cmd,
442 struct inode *inode = filp->f_path.dentry->d_inode;
443 int __user *p = (int __user *)arg;
447 return ioctl_fibmap(filp, p);
449 return put_user(i_size_read(inode) - filp->f_pos, p);
451 case FS_IOC_RESVSP64:
452 return ioctl_preallocate(filp, p);
455 return vfs_ioctl(filp, cmd, arg);
458 static int ioctl_fionbio(struct file *filp, int __user *argp)
463 error = get_user(on, argp);
468 /* SunOS compatibility item. */
469 if (O_NONBLOCK != O_NDELAY)
472 spin_lock(&filp->f_lock);
474 filp->f_flags |= flag;
476 filp->f_flags &= ~flag;
477 spin_unlock(&filp->f_lock);
481 static int ioctl_fioasync(unsigned int fd, struct file *filp,
487 error = get_user(on, argp);
490 flag = on ? FASYNC : 0;
492 /* Did FASYNC state change ? */
493 if ((flag ^ filp->f_flags) & FASYNC) {
494 if (filp->f_op && filp->f_op->fasync)
495 /* fasync() adjusts filp->f_flags */
496 error = filp->f_op->fasync(fd, filp, on);
500 return error < 0 ? error : 0;
503 static int ioctl_fsfreeze(struct file *filp)
505 struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
507 if (!capable(CAP_SYS_ADMIN))
510 /* If filesystem doesn't support freeze feature, return. */
511 if (sb->s_op->freeze_fs == NULL)
514 /* If a blockdevice-backed filesystem isn't specified, return. */
515 if (sb->s_bdev == NULL)
519 sb = freeze_bdev(sb->s_bdev);
525 static int ioctl_fsthaw(struct file *filp)
527 struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
529 if (!capable(CAP_SYS_ADMIN))
532 /* If a blockdevice-backed filesystem isn't specified, return EINVAL. */
533 if (sb->s_bdev == NULL)
537 return thaw_bdev(sb->s_bdev, sb);
541 * When you add any new common ioctls to the switches above and below
542 * please update compat_sys_ioctl() too.
544 * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
545 * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
547 int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
551 int __user *argp = (int __user *)arg;
555 set_close_on_exec(fd, 1);
559 set_close_on_exec(fd, 0);
563 error = ioctl_fionbio(filp, argp);
567 error = ioctl_fioasync(fd, filp, argp);
571 if (S_ISDIR(filp->f_path.dentry->d_inode->i_mode) ||
572 S_ISREG(filp->f_path.dentry->d_inode->i_mode) ||
573 S_ISLNK(filp->f_path.dentry->d_inode->i_mode)) {
575 inode_get_bytes(filp->f_path.dentry->d_inode);
576 error = copy_to_user((loff_t __user *)arg, &res,
577 sizeof(res)) ? -EFAULT : 0;
583 error = ioctl_fsfreeze(filp);
587 error = ioctl_fsthaw(filp);
591 return ioctl_fiemap(filp, arg);
595 struct inode *inode = filp->f_path.dentry->d_inode;
596 int __user *p = (int __user *)arg;
597 return put_user(inode->i_sb->s_blocksize, p);
601 if (S_ISREG(filp->f_path.dentry->d_inode->i_mode))
602 error = file_ioctl(filp, cmd, arg);
604 error = vfs_ioctl(filp, cmd, arg);
610 SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
616 filp = fget_light(fd, &fput_needed);
620 error = security_file_ioctl(filp, cmd, arg);
624 error = do_vfs_ioctl(filp, fd, cmd, arg);
626 fput_light(filp, fput_needed);