xref: /openbmc/linux/fs/ioctl.c (revision baa7eb025ab14f3cba2e35c0a8648f9c9f01d24f)
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  */
35 static 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 
50 static 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)
85 int 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 *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 }
122 EXPORT_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  */
138 int 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 }
149 EXPORT_SYMBOL(fiemap_check_flags);
150 
151 static 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 
173 static int ioctl_fiemap(struct file *filp, unsigned long arg)
174 {
175 	struct fiemap fiemap;
176 	struct fiemap_extent_info fieinfo = { 0, };
177 	struct inode *inode = filp->f_path.dentry->d_inode;
178 	struct super_block *sb = inode->i_sb;
179 	u64 len;
180 	int error;
181 
182 	if (!inode->i_op->fiemap)
183 		return -EOPNOTSUPP;
184 
185 	if (copy_from_user(&fiemap, (struct fiemap __user *)arg,
186 			   sizeof(struct 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 = (struct fiemap_extent *)(arg + sizeof(fiemap));
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((char *)arg, &fiemap, sizeof(fiemap)))
213 		error = -EFAULT;
214 
215 	return error;
216 }
217 
218 #ifdef CONFIG_BLOCK
219 
220 static inline sector_t logical_to_blk(struct inode *inode, loff_t offset)
221 {
222 	return (offset >> inode->i_blkbits);
223 }
224 
225 static 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 
250 int __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 	start_blk = logical_to_blk(inode, start);
277 	last_blk = logical_to_blk(inode, start + len - 1);
278 
279 	do {
280 		/*
281 		 * we set b_size to the total size we want so it will map as
282 		 * many contiguous blocks as possible at once
283 		 */
284 		memset(&map_bh, 0, sizeof(struct buffer_head));
285 		map_bh.b_size = len;
286 
287 		ret = get_block(inode, start_blk, &map_bh, 0);
288 		if (ret)
289 			break;
290 
291 		/* HOLE */
292 		if (!buffer_mapped(&map_bh)) {
293 			start_blk++;
294 
295 			/*
296 			 * We want to handle the case where there is an
297 			 * allocated block at the front of the file, and then
298 			 * nothing but holes up to the end of the file properly,
299 			 * to make sure that extent at the front gets properly
300 			 * marked with FIEMAP_EXTENT_LAST
301 			 */
302 			if (!past_eof &&
303 			    blk_to_logical(inode, start_blk) >= isize)
304 				past_eof = 1;
305 
306 			/*
307 			 * First hole after going past the EOF, this is our
308 			 * last extent
309 			 */
310 			if (past_eof && size) {
311 				flags = FIEMAP_EXTENT_MERGED|FIEMAP_EXTENT_LAST;
312 				ret = fiemap_fill_next_extent(fieinfo, logical,
313 							      phys, size,
314 							      flags);
315 			} else if (size) {
316 				ret = fiemap_fill_next_extent(fieinfo, logical,
317 							      phys, size, flags);
318 				size = 0;
319 			}
320 
321 			/* if we have holes up to/past EOF then we're done */
322 			if (start_blk > last_blk || past_eof || ret)
323 				break;
324 		} else {
325 			/*
326 			 * We have gone over the length of what we wanted to
327 			 * map, and it wasn't the entire file, so add the extent
328 			 * we got last time and exit.
329 			 *
330 			 * This is for the case where say we want to map all the
331 			 * way up to the second to the last block in a file, but
332 			 * the last block is a hole, making the second to last
333 			 * block FIEMAP_EXTENT_LAST.  In this case we want to
334 			 * see if there is a hole after the second to last block
335 			 * so we can mark it properly.  If we found data after
336 			 * we exceeded the length we were requesting, then we
337 			 * are good to go, just add the extent to the fieinfo
338 			 * and break
339 			 */
340 			if (start_blk > last_blk && !whole_file) {
341 				ret = fiemap_fill_next_extent(fieinfo, logical,
342 							      phys, size,
343 							      flags);
344 				break;
345 			}
346 
347 			/*
348 			 * if size != 0 then we know we already have an extent
349 			 * to add, so add it.
350 			 */
351 			if (size) {
352 				ret = fiemap_fill_next_extent(fieinfo, logical,
353 							      phys, size,
354 							      flags);
355 				if (ret)
356 					break;
357 			}
358 
359 			logical = blk_to_logical(inode, start_blk);
360 			phys = blk_to_logical(inode, map_bh.b_blocknr);
361 			size = map_bh.b_size;
362 			flags = FIEMAP_EXTENT_MERGED;
363 
364 			start_blk += logical_to_blk(inode, size);
365 
366 			/*
367 			 * If we are past the EOF, then we need to make sure as
368 			 * soon as we find a hole that the last extent we found
369 			 * is marked with FIEMAP_EXTENT_LAST
370 			 */
371 			if (!past_eof && logical + size >= isize)
372 				past_eof = true;
373 		}
374 		cond_resched();
375 	} while (1);
376 
377 	/* If ret is 1 then we just hit the end of the extent array */
378 	if (ret == 1)
379 		ret = 0;
380 
381 	return ret;
382 }
383 EXPORT_SYMBOL(__generic_block_fiemap);
384 
385 /**
386  * generic_block_fiemap - FIEMAP for block based inodes
387  * @inode: The inode to map
388  * @fieinfo: The mapping information
389  * @start: The initial block to map
390  * @len: The length of the extect to attempt to map
391  * @get_block: The block mapping function for the fs
392  *
393  * Calls __generic_block_fiemap to map the inode, after taking
394  * the inode's mutex lock.
395  */
396 
397 int generic_block_fiemap(struct inode *inode,
398 			 struct fiemap_extent_info *fieinfo, u64 start,
399 			 u64 len, get_block_t *get_block)
400 {
401 	int ret;
402 	mutex_lock(&inode->i_mutex);
403 	ret = __generic_block_fiemap(inode, fieinfo, start, len, get_block);
404 	mutex_unlock(&inode->i_mutex);
405 	return ret;
406 }
407 EXPORT_SYMBOL(generic_block_fiemap);
408 
409 #endif  /*  CONFIG_BLOCK  */
410 
411 /*
412  * This provides compatibility with legacy XFS pre-allocation ioctls
413  * which predate the fallocate syscall.
414  *
415  * Only the l_start, l_len and l_whence fields of the 'struct space_resv'
416  * are used here, rest are ignored.
417  */
418 int ioctl_preallocate(struct file *filp, void __user *argp)
419 {
420 	struct inode *inode = filp->f_path.dentry->d_inode;
421 	struct space_resv sr;
422 
423 	if (copy_from_user(&sr, argp, sizeof(sr)))
424 		return -EFAULT;
425 
426 	switch (sr.l_whence) {
427 	case SEEK_SET:
428 		break;
429 	case SEEK_CUR:
430 		sr.l_start += filp->f_pos;
431 		break;
432 	case SEEK_END:
433 		sr.l_start += i_size_read(inode);
434 		break;
435 	default:
436 		return -EINVAL;
437 	}
438 
439 	return do_fallocate(filp, FALLOC_FL_KEEP_SIZE, sr.l_start, sr.l_len);
440 }
441 
442 static int file_ioctl(struct file *filp, unsigned int cmd,
443 		unsigned long arg)
444 {
445 	struct inode *inode = filp->f_path.dentry->d_inode;
446 	int __user *p = (int __user *)arg;
447 
448 	switch (cmd) {
449 	case FIBMAP:
450 		return ioctl_fibmap(filp, p);
451 	case FIONREAD:
452 		return put_user(i_size_read(inode) - filp->f_pos, p);
453 	case FS_IOC_RESVSP:
454 	case FS_IOC_RESVSP64:
455 		return ioctl_preallocate(filp, p);
456 	}
457 
458 	return vfs_ioctl(filp, cmd, arg);
459 }
460 
461 static int ioctl_fionbio(struct file *filp, int __user *argp)
462 {
463 	unsigned int flag;
464 	int on, error;
465 
466 	error = get_user(on, argp);
467 	if (error)
468 		return error;
469 	flag = O_NONBLOCK;
470 #ifdef __sparc__
471 	/* SunOS compatibility item. */
472 	if (O_NONBLOCK != O_NDELAY)
473 		flag |= O_NDELAY;
474 #endif
475 	spin_lock(&filp->f_lock);
476 	if (on)
477 		filp->f_flags |= flag;
478 	else
479 		filp->f_flags &= ~flag;
480 	spin_unlock(&filp->f_lock);
481 	return error;
482 }
483 
484 static int ioctl_fioasync(unsigned int fd, struct file *filp,
485 			  int __user *argp)
486 {
487 	unsigned int flag;
488 	int on, error;
489 
490 	error = get_user(on, argp);
491 	if (error)
492 		return error;
493 	flag = on ? FASYNC : 0;
494 
495 	/* Did FASYNC state change ? */
496 	if ((flag ^ filp->f_flags) & FASYNC) {
497 		if (filp->f_op && filp->f_op->fasync)
498 			/* fasync() adjusts filp->f_flags */
499 			error = filp->f_op->fasync(fd, filp, on);
500 		else
501 			error = -ENOTTY;
502 	}
503 	return error < 0 ? error : 0;
504 }
505 
506 static int ioctl_fsfreeze(struct file *filp)
507 {
508 	struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
509 
510 	if (!capable(CAP_SYS_ADMIN))
511 		return -EPERM;
512 
513 	/* If filesystem doesn't support freeze feature, return. */
514 	if (sb->s_op->freeze_fs == NULL)
515 		return -EOPNOTSUPP;
516 
517 	/* Freeze */
518 	return freeze_super(sb);
519 }
520 
521 static int ioctl_fsthaw(struct file *filp)
522 {
523 	struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
524 
525 	if (!capable(CAP_SYS_ADMIN))
526 		return -EPERM;
527 
528 	/* Thaw */
529 	return thaw_super(sb);
530 }
531 
532 /*
533  * When you add any new common ioctls to the switches above and below
534  * please update compat_sys_ioctl() too.
535  *
536  * do_vfs_ioctl() is not for drivers and not intended to be EXPORT_SYMBOL()'d.
537  * It's just a simple helper for sys_ioctl and compat_sys_ioctl.
538  */
539 int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
540 	     unsigned long arg)
541 {
542 	int error = 0;
543 	int __user *argp = (int __user *)arg;
544 
545 	switch (cmd) {
546 	case FIOCLEX:
547 		set_close_on_exec(fd, 1);
548 		break;
549 
550 	case FIONCLEX:
551 		set_close_on_exec(fd, 0);
552 		break;
553 
554 	case FIONBIO:
555 		error = ioctl_fionbio(filp, argp);
556 		break;
557 
558 	case FIOASYNC:
559 		error = ioctl_fioasync(fd, filp, argp);
560 		break;
561 
562 	case FIOQSIZE:
563 		if (S_ISDIR(filp->f_path.dentry->d_inode->i_mode) ||
564 		    S_ISREG(filp->f_path.dentry->d_inode->i_mode) ||
565 		    S_ISLNK(filp->f_path.dentry->d_inode->i_mode)) {
566 			loff_t res =
567 				inode_get_bytes(filp->f_path.dentry->d_inode);
568 			error = copy_to_user((loff_t __user *)arg, &res,
569 					     sizeof(res)) ? -EFAULT : 0;
570 		} else
571 			error = -ENOTTY;
572 		break;
573 
574 	case FIFREEZE:
575 		error = ioctl_fsfreeze(filp);
576 		break;
577 
578 	case FITHAW:
579 		error = ioctl_fsthaw(filp);
580 		break;
581 
582 	case FS_IOC_FIEMAP:
583 		return ioctl_fiemap(filp, arg);
584 
585 	case FIGETBSZ:
586 	{
587 		struct inode *inode = filp->f_path.dentry->d_inode;
588 		int __user *p = (int __user *)arg;
589 		return put_user(inode->i_sb->s_blocksize, p);
590 	}
591 
592 	default:
593 		if (S_ISREG(filp->f_path.dentry->d_inode->i_mode))
594 			error = file_ioctl(filp, cmd, arg);
595 		else
596 			error = vfs_ioctl(filp, cmd, arg);
597 		break;
598 	}
599 	return error;
600 }
601 
602 SYSCALL_DEFINE3(ioctl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
603 {
604 	struct file *filp;
605 	int error = -EBADF;
606 	int fput_needed;
607 
608 	filp = fget_light(fd, &fput_needed);
609 	if (!filp)
610 		goto out;
611 
612 	error = security_file_ioctl(filp, cmd, arg);
613 	if (error)
614 		goto out_fput;
615 
616 	error = do_vfs_ioctl(filp, fd, cmd, arg);
617  out_fput:
618 	fput_light(filp, fput_needed);
619  out:
620 	return error;
621 }
622