xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision b60a5b8d)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
4  * All Rights Reserved.
5  */
6 #include "xfs.h"
7 #include "xfs_fs.h"
8 #include "xfs_shared.h"
9 #include "xfs_format.h"
10 #include "xfs_log_format.h"
11 #include "xfs_trans_resv.h"
12 #include "xfs_mount.h"
13 #include "xfs_inode.h"
14 #include "xfs_ioctl.h"
15 #include "xfs_alloc.h"
16 #include "xfs_rtalloc.h"
17 #include "xfs_itable.h"
18 #include "xfs_error.h"
19 #include "xfs_attr.h"
20 #include "xfs_bmap.h"
21 #include "xfs_bmap_util.h"
22 #include "xfs_fsops.h"
23 #include "xfs_discard.h"
24 #include "xfs_quota.h"
25 #include "xfs_export.h"
26 #include "xfs_trace.h"
27 #include "xfs_icache.h"
28 #include "xfs_symlink.h"
29 #include "xfs_trans.h"
30 #include "xfs_acl.h"
31 #include "xfs_btree.h"
32 #include <linux/fsmap.h>
33 #include "xfs_fsmap.h"
34 #include "scrub/xfs_scrub.h"
35 #include "xfs_sb.h"
36 
37 #include <linux/capability.h>
38 #include <linux/cred.h>
39 #include <linux/dcache.h>
40 #include <linux/mount.h>
41 #include <linux/namei.h>
42 #include <linux/pagemap.h>
43 #include <linux/slab.h>
44 #include <linux/exportfs.h>
45 
46 /*
47  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
48  * a file or fs handle.
49  *
50  * XFS_IOC_PATH_TO_FSHANDLE
51  *    returns fs handle for a mount point or path within that mount point
52  * XFS_IOC_FD_TO_HANDLE
53  *    returns full handle for a FD opened in user space
54  * XFS_IOC_PATH_TO_HANDLE
55  *    returns full handle for a path
56  */
57 int
58 xfs_find_handle(
59 	unsigned int		cmd,
60 	xfs_fsop_handlereq_t	*hreq)
61 {
62 	int			hsize;
63 	xfs_handle_t		handle;
64 	struct inode		*inode;
65 	struct fd		f = {NULL};
66 	struct path		path;
67 	int			error;
68 	struct xfs_inode	*ip;
69 
70 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
71 		f = fdget(hreq->fd);
72 		if (!f.file)
73 			return -EBADF;
74 		inode = file_inode(f.file);
75 	} else {
76 		error = user_lpath((const char __user *)hreq->path, &path);
77 		if (error)
78 			return error;
79 		inode = d_inode(path.dentry);
80 	}
81 	ip = XFS_I(inode);
82 
83 	/*
84 	 * We can only generate handles for inodes residing on a XFS filesystem,
85 	 * and only for regular files, directories or symbolic links.
86 	 */
87 	error = -EINVAL;
88 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
89 		goto out_put;
90 
91 	error = -EBADF;
92 	if (!S_ISREG(inode->i_mode) &&
93 	    !S_ISDIR(inode->i_mode) &&
94 	    !S_ISLNK(inode->i_mode))
95 		goto out_put;
96 
97 
98 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
99 
100 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
101 		/*
102 		 * This handle only contains an fsid, zero the rest.
103 		 */
104 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
105 		hsize = sizeof(xfs_fsid_t);
106 	} else {
107 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
108 					sizeof(handle.ha_fid.fid_len);
109 		handle.ha_fid.fid_pad = 0;
110 		handle.ha_fid.fid_gen = inode->i_generation;
111 		handle.ha_fid.fid_ino = ip->i_ino;
112 		hsize = sizeof(xfs_handle_t);
113 	}
114 
115 	error = -EFAULT;
116 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
117 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
118 		goto out_put;
119 
120 	error = 0;
121 
122  out_put:
123 	if (cmd == XFS_IOC_FD_TO_HANDLE)
124 		fdput(f);
125 	else
126 		path_put(&path);
127 	return error;
128 }
129 
130 /*
131  * No need to do permission checks on the various pathname components
132  * as the handle operations are privileged.
133  */
134 STATIC int
135 xfs_handle_acceptable(
136 	void			*context,
137 	struct dentry		*dentry)
138 {
139 	return 1;
140 }
141 
142 /*
143  * Convert userspace handle data into a dentry.
144  */
145 struct dentry *
146 xfs_handle_to_dentry(
147 	struct file		*parfilp,
148 	void __user		*uhandle,
149 	u32			hlen)
150 {
151 	xfs_handle_t		handle;
152 	struct xfs_fid64	fid;
153 
154 	/*
155 	 * Only allow handle opens under a directory.
156 	 */
157 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
158 		return ERR_PTR(-ENOTDIR);
159 
160 	if (hlen != sizeof(xfs_handle_t))
161 		return ERR_PTR(-EINVAL);
162 	if (copy_from_user(&handle, uhandle, hlen))
163 		return ERR_PTR(-EFAULT);
164 	if (handle.ha_fid.fid_len !=
165 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
166 		return ERR_PTR(-EINVAL);
167 
168 	memset(&fid, 0, sizeof(struct fid));
169 	fid.ino = handle.ha_fid.fid_ino;
170 	fid.gen = handle.ha_fid.fid_gen;
171 
172 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
173 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
174 			xfs_handle_acceptable, NULL);
175 }
176 
177 STATIC struct dentry *
178 xfs_handlereq_to_dentry(
179 	struct file		*parfilp,
180 	xfs_fsop_handlereq_t	*hreq)
181 {
182 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
183 }
184 
185 int
186 xfs_open_by_handle(
187 	struct file		*parfilp,
188 	xfs_fsop_handlereq_t	*hreq)
189 {
190 	const struct cred	*cred = current_cred();
191 	int			error;
192 	int			fd;
193 	int			permflag;
194 	struct file		*filp;
195 	struct inode		*inode;
196 	struct dentry		*dentry;
197 	fmode_t			fmode;
198 	struct path		path;
199 
200 	if (!capable(CAP_SYS_ADMIN))
201 		return -EPERM;
202 
203 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
204 	if (IS_ERR(dentry))
205 		return PTR_ERR(dentry);
206 	inode = d_inode(dentry);
207 
208 	/* Restrict xfs_open_by_handle to directories & regular files. */
209 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
210 		error = -EPERM;
211 		goto out_dput;
212 	}
213 
214 #if BITS_PER_LONG != 32
215 	hreq->oflags |= O_LARGEFILE;
216 #endif
217 
218 	permflag = hreq->oflags;
219 	fmode = OPEN_FMODE(permflag);
220 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
221 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
222 		error = -EPERM;
223 		goto out_dput;
224 	}
225 
226 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
227 		error = -EPERM;
228 		goto out_dput;
229 	}
230 
231 	/* Can't write directories. */
232 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
233 		error = -EISDIR;
234 		goto out_dput;
235 	}
236 
237 	fd = get_unused_fd_flags(0);
238 	if (fd < 0) {
239 		error = fd;
240 		goto out_dput;
241 	}
242 
243 	path.mnt = parfilp->f_path.mnt;
244 	path.dentry = dentry;
245 	filp = dentry_open(&path, hreq->oflags, cred);
246 	dput(dentry);
247 	if (IS_ERR(filp)) {
248 		put_unused_fd(fd);
249 		return PTR_ERR(filp);
250 	}
251 
252 	if (S_ISREG(inode->i_mode)) {
253 		filp->f_flags |= O_NOATIME;
254 		filp->f_mode |= FMODE_NOCMTIME;
255 	}
256 
257 	fd_install(fd, filp);
258 	return fd;
259 
260  out_dput:
261 	dput(dentry);
262 	return error;
263 }
264 
265 int
266 xfs_readlink_by_handle(
267 	struct file		*parfilp,
268 	xfs_fsop_handlereq_t	*hreq)
269 {
270 	struct dentry		*dentry;
271 	__u32			olen;
272 	int			error;
273 
274 	if (!capable(CAP_SYS_ADMIN))
275 		return -EPERM;
276 
277 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
278 	if (IS_ERR(dentry))
279 		return PTR_ERR(dentry);
280 
281 	/* Restrict this handle operation to symlinks only. */
282 	if (!d_is_symlink(dentry)) {
283 		error = -EINVAL;
284 		goto out_dput;
285 	}
286 
287 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
288 		error = -EFAULT;
289 		goto out_dput;
290 	}
291 
292 	error = vfs_readlink(dentry, hreq->ohandle, olen);
293 
294  out_dput:
295 	dput(dentry);
296 	return error;
297 }
298 
299 int
300 xfs_set_dmattrs(
301 	xfs_inode_t     *ip,
302 	uint		evmask,
303 	uint16_t	state)
304 {
305 	xfs_mount_t	*mp = ip->i_mount;
306 	xfs_trans_t	*tp;
307 	int		error;
308 
309 	if (!capable(CAP_SYS_ADMIN))
310 		return -EPERM;
311 
312 	if (XFS_FORCED_SHUTDOWN(mp))
313 		return -EIO;
314 
315 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
316 	if (error)
317 		return error;
318 
319 	xfs_ilock(ip, XFS_ILOCK_EXCL);
320 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
321 
322 	ip->i_d.di_dmevmask = evmask;
323 	ip->i_d.di_dmstate  = state;
324 
325 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
326 	error = xfs_trans_commit(tp);
327 
328 	return error;
329 }
330 
331 STATIC int
332 xfs_fssetdm_by_handle(
333 	struct file		*parfilp,
334 	void			__user *arg)
335 {
336 	int			error;
337 	struct fsdmidata	fsd;
338 	xfs_fsop_setdm_handlereq_t dmhreq;
339 	struct dentry		*dentry;
340 
341 	if (!capable(CAP_MKNOD))
342 		return -EPERM;
343 	if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
344 		return -EFAULT;
345 
346 	error = mnt_want_write_file(parfilp);
347 	if (error)
348 		return error;
349 
350 	dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
351 	if (IS_ERR(dentry)) {
352 		mnt_drop_write_file(parfilp);
353 		return PTR_ERR(dentry);
354 	}
355 
356 	if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
357 		error = -EPERM;
358 		goto out;
359 	}
360 
361 	if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
362 		error = -EFAULT;
363 		goto out;
364 	}
365 
366 	error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
367 				 fsd.fsd_dmstate);
368 
369  out:
370 	mnt_drop_write_file(parfilp);
371 	dput(dentry);
372 	return error;
373 }
374 
375 STATIC int
376 xfs_attrlist_by_handle(
377 	struct file		*parfilp,
378 	void			__user *arg)
379 {
380 	int			error = -ENOMEM;
381 	attrlist_cursor_kern_t	*cursor;
382 	struct xfs_fsop_attrlist_handlereq __user	*p = arg;
383 	xfs_fsop_attrlist_handlereq_t al_hreq;
384 	struct dentry		*dentry;
385 	char			*kbuf;
386 
387 	if (!capable(CAP_SYS_ADMIN))
388 		return -EPERM;
389 	if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
390 		return -EFAULT;
391 	if (al_hreq.buflen < sizeof(struct attrlist) ||
392 	    al_hreq.buflen > XFS_XATTR_LIST_MAX)
393 		return -EINVAL;
394 
395 	/*
396 	 * Reject flags, only allow namespaces.
397 	 */
398 	if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
399 		return -EINVAL;
400 
401 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
402 	if (IS_ERR(dentry))
403 		return PTR_ERR(dentry);
404 
405 	kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
406 	if (!kbuf)
407 		goto out_dput;
408 
409 	cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
410 	error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
411 					al_hreq.flags, cursor);
412 	if (error)
413 		goto out_kfree;
414 
415 	if (copy_to_user(&p->pos, cursor, sizeof(attrlist_cursor_kern_t))) {
416 		error = -EFAULT;
417 		goto out_kfree;
418 	}
419 
420 	if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
421 		error = -EFAULT;
422 
423 out_kfree:
424 	kmem_free(kbuf);
425 out_dput:
426 	dput(dentry);
427 	return error;
428 }
429 
430 int
431 xfs_attrmulti_attr_get(
432 	struct inode		*inode,
433 	unsigned char		*name,
434 	unsigned char		__user *ubuf,
435 	uint32_t		*len,
436 	uint32_t		flags)
437 {
438 	unsigned char		*kbuf;
439 	int			error = -EFAULT;
440 
441 	if (*len > XFS_XATTR_SIZE_MAX)
442 		return -EINVAL;
443 	kbuf = kmem_zalloc_large(*len, KM_SLEEP);
444 	if (!kbuf)
445 		return -ENOMEM;
446 
447 	error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
448 	if (error)
449 		goto out_kfree;
450 
451 	if (copy_to_user(ubuf, kbuf, *len))
452 		error = -EFAULT;
453 
454 out_kfree:
455 	kmem_free(kbuf);
456 	return error;
457 }
458 
459 int
460 xfs_attrmulti_attr_set(
461 	struct inode		*inode,
462 	unsigned char		*name,
463 	const unsigned char	__user *ubuf,
464 	uint32_t		len,
465 	uint32_t		flags)
466 {
467 	unsigned char		*kbuf;
468 	int			error;
469 
470 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
471 		return -EPERM;
472 	if (len > XFS_XATTR_SIZE_MAX)
473 		return -EINVAL;
474 
475 	kbuf = memdup_user(ubuf, len);
476 	if (IS_ERR(kbuf))
477 		return PTR_ERR(kbuf);
478 
479 	error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
480 	if (!error)
481 		xfs_forget_acl(inode, name, flags);
482 	kfree(kbuf);
483 	return error;
484 }
485 
486 int
487 xfs_attrmulti_attr_remove(
488 	struct inode		*inode,
489 	unsigned char		*name,
490 	uint32_t		flags)
491 {
492 	int			error;
493 
494 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
495 		return -EPERM;
496 	error = xfs_attr_remove(XFS_I(inode), name, flags);
497 	if (!error)
498 		xfs_forget_acl(inode, name, flags);
499 	return error;
500 }
501 
502 STATIC int
503 xfs_attrmulti_by_handle(
504 	struct file		*parfilp,
505 	void			__user *arg)
506 {
507 	int			error;
508 	xfs_attr_multiop_t	*ops;
509 	xfs_fsop_attrmulti_handlereq_t am_hreq;
510 	struct dentry		*dentry;
511 	unsigned int		i, size;
512 	unsigned char		*attr_name;
513 
514 	if (!capable(CAP_SYS_ADMIN))
515 		return -EPERM;
516 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
517 		return -EFAULT;
518 
519 	/* overflow check */
520 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
521 		return -E2BIG;
522 
523 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
524 	if (IS_ERR(dentry))
525 		return PTR_ERR(dentry);
526 
527 	error = -E2BIG;
528 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
529 	if (!size || size > 16 * PAGE_SIZE)
530 		goto out_dput;
531 
532 	ops = memdup_user(am_hreq.ops, size);
533 	if (IS_ERR(ops)) {
534 		error = PTR_ERR(ops);
535 		goto out_dput;
536 	}
537 
538 	error = -ENOMEM;
539 	attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
540 	if (!attr_name)
541 		goto out_kfree_ops;
542 
543 	error = 0;
544 	for (i = 0; i < am_hreq.opcount; i++) {
545 		ops[i].am_error = strncpy_from_user((char *)attr_name,
546 				ops[i].am_attrname, MAXNAMELEN);
547 		if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
548 			error = -ERANGE;
549 		if (ops[i].am_error < 0)
550 			break;
551 
552 		switch (ops[i].am_opcode) {
553 		case ATTR_OP_GET:
554 			ops[i].am_error = xfs_attrmulti_attr_get(
555 					d_inode(dentry), attr_name,
556 					ops[i].am_attrvalue, &ops[i].am_length,
557 					ops[i].am_flags);
558 			break;
559 		case ATTR_OP_SET:
560 			ops[i].am_error = mnt_want_write_file(parfilp);
561 			if (ops[i].am_error)
562 				break;
563 			ops[i].am_error = xfs_attrmulti_attr_set(
564 					d_inode(dentry), attr_name,
565 					ops[i].am_attrvalue, ops[i].am_length,
566 					ops[i].am_flags);
567 			mnt_drop_write_file(parfilp);
568 			break;
569 		case ATTR_OP_REMOVE:
570 			ops[i].am_error = mnt_want_write_file(parfilp);
571 			if (ops[i].am_error)
572 				break;
573 			ops[i].am_error = xfs_attrmulti_attr_remove(
574 					d_inode(dentry), attr_name,
575 					ops[i].am_flags);
576 			mnt_drop_write_file(parfilp);
577 			break;
578 		default:
579 			ops[i].am_error = -EINVAL;
580 		}
581 	}
582 
583 	if (copy_to_user(am_hreq.ops, ops, size))
584 		error = -EFAULT;
585 
586 	kfree(attr_name);
587  out_kfree_ops:
588 	kfree(ops);
589  out_dput:
590 	dput(dentry);
591 	return error;
592 }
593 
594 int
595 xfs_ioc_space(
596 	struct file		*filp,
597 	unsigned int		cmd,
598 	xfs_flock64_t		*bf)
599 {
600 	struct inode		*inode = file_inode(filp);
601 	struct xfs_inode	*ip = XFS_I(inode);
602 	struct iattr		iattr;
603 	enum xfs_prealloc_flags	flags = 0;
604 	uint			iolock = XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL;
605 	int			error;
606 
607 	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
608 		return -EPERM;
609 
610 	if (!(filp->f_mode & FMODE_WRITE))
611 		return -EBADF;
612 
613 	if (!S_ISREG(inode->i_mode))
614 		return -EINVAL;
615 
616 	if (filp->f_flags & O_DSYNC)
617 		flags |= XFS_PREALLOC_SYNC;
618 	if (filp->f_mode & FMODE_NOCMTIME)
619 		flags |= XFS_PREALLOC_INVISIBLE;
620 
621 	error = mnt_want_write_file(filp);
622 	if (error)
623 		return error;
624 
625 	xfs_ilock(ip, iolock);
626 	error = xfs_break_layouts(inode, &iolock, BREAK_UNMAP);
627 	if (error)
628 		goto out_unlock;
629 
630 	switch (bf->l_whence) {
631 	case 0: /*SEEK_SET*/
632 		break;
633 	case 1: /*SEEK_CUR*/
634 		bf->l_start += filp->f_pos;
635 		break;
636 	case 2: /*SEEK_END*/
637 		bf->l_start += XFS_ISIZE(ip);
638 		break;
639 	default:
640 		error = -EINVAL;
641 		goto out_unlock;
642 	}
643 
644 	/*
645 	 * length of <= 0 for resv/unresv/zero is invalid.  length for
646 	 * alloc/free is ignored completely and we have no idea what userspace
647 	 * might have set it to, so set it to zero to allow range
648 	 * checks to pass.
649 	 */
650 	switch (cmd) {
651 	case XFS_IOC_ZERO_RANGE:
652 	case XFS_IOC_RESVSP:
653 	case XFS_IOC_RESVSP64:
654 	case XFS_IOC_UNRESVSP:
655 	case XFS_IOC_UNRESVSP64:
656 		if (bf->l_len <= 0) {
657 			error = -EINVAL;
658 			goto out_unlock;
659 		}
660 		break;
661 	default:
662 		bf->l_len = 0;
663 		break;
664 	}
665 
666 	if (bf->l_start < 0 ||
667 	    bf->l_start > inode->i_sb->s_maxbytes ||
668 	    bf->l_start + bf->l_len < 0 ||
669 	    bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
670 		error = -EINVAL;
671 		goto out_unlock;
672 	}
673 
674 	switch (cmd) {
675 	case XFS_IOC_ZERO_RANGE:
676 		flags |= XFS_PREALLOC_SET;
677 		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
678 		break;
679 	case XFS_IOC_RESVSP:
680 	case XFS_IOC_RESVSP64:
681 		flags |= XFS_PREALLOC_SET;
682 		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
683 						XFS_BMAPI_PREALLOC);
684 		break;
685 	case XFS_IOC_UNRESVSP:
686 	case XFS_IOC_UNRESVSP64:
687 		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
688 		break;
689 	case XFS_IOC_ALLOCSP:
690 	case XFS_IOC_ALLOCSP64:
691 	case XFS_IOC_FREESP:
692 	case XFS_IOC_FREESP64:
693 		flags |= XFS_PREALLOC_CLEAR;
694 		if (bf->l_start > XFS_ISIZE(ip)) {
695 			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
696 					bf->l_start - XFS_ISIZE(ip), 0);
697 			if (error)
698 				goto out_unlock;
699 		}
700 
701 		iattr.ia_valid = ATTR_SIZE;
702 		iattr.ia_size = bf->l_start;
703 
704 		error = xfs_vn_setattr_size(file_dentry(filp), &iattr);
705 		break;
706 	default:
707 		ASSERT(0);
708 		error = -EINVAL;
709 	}
710 
711 	if (error)
712 		goto out_unlock;
713 
714 	error = xfs_update_prealloc_flags(ip, flags);
715 
716 out_unlock:
717 	xfs_iunlock(ip, iolock);
718 	mnt_drop_write_file(filp);
719 	return error;
720 }
721 
722 STATIC int
723 xfs_ioc_bulkstat(
724 	xfs_mount_t		*mp,
725 	unsigned int		cmd,
726 	void			__user *arg)
727 {
728 	xfs_fsop_bulkreq_t	bulkreq;
729 	int			count;	/* # of records returned */
730 	xfs_ino_t		inlast;	/* last inode number */
731 	int			done;
732 	int			error;
733 
734 	/* done = 1 if there are more stats to get and if bulkstat */
735 	/* should be called again (unused here, but used in dmapi) */
736 
737 	if (!capable(CAP_SYS_ADMIN))
738 		return -EPERM;
739 
740 	if (XFS_FORCED_SHUTDOWN(mp))
741 		return -EIO;
742 
743 	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
744 		return -EFAULT;
745 
746 	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
747 		return -EFAULT;
748 
749 	if ((count = bulkreq.icount) <= 0)
750 		return -EINVAL;
751 
752 	if (bulkreq.ubuffer == NULL)
753 		return -EINVAL;
754 
755 	if (cmd == XFS_IOC_FSINUMBERS)
756 		error = xfs_inumbers(mp, &inlast, &count,
757 					bulkreq.ubuffer, xfs_inumbers_fmt);
758 	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
759 		error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
760 					sizeof(xfs_bstat_t), NULL, &done);
761 	else	/* XFS_IOC_FSBULKSTAT */
762 		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
763 				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
764 				     &done);
765 
766 	if (error)
767 		return error;
768 
769 	if (bulkreq.ocount != NULL) {
770 		if (copy_to_user(bulkreq.lastip, &inlast,
771 						sizeof(xfs_ino_t)))
772 			return -EFAULT;
773 
774 		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
775 			return -EFAULT;
776 	}
777 
778 	return 0;
779 }
780 
781 STATIC int
782 xfs_ioc_fsgeometry_v1(
783 	xfs_mount_t		*mp,
784 	void			__user *arg)
785 {
786 	xfs_fsop_geom_t         fsgeo;
787 	int			error;
788 
789 	error = xfs_fs_geometry(&mp->m_sb, &fsgeo, 3);
790 	if (error)
791 		return error;
792 
793 	/*
794 	 * Caller should have passed an argument of type
795 	 * xfs_fsop_geom_v1_t.  This is a proper subset of the
796 	 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
797 	 */
798 	if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
799 		return -EFAULT;
800 	return 0;
801 }
802 
803 STATIC int
804 xfs_ioc_fsgeometry(
805 	xfs_mount_t		*mp,
806 	void			__user *arg)
807 {
808 	xfs_fsop_geom_t		fsgeo;
809 	int			error;
810 
811 	error = xfs_fs_geometry(&mp->m_sb, &fsgeo, 4);
812 	if (error)
813 		return error;
814 
815 	if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
816 		return -EFAULT;
817 	return 0;
818 }
819 
820 /*
821  * Linux extended inode flags interface.
822  */
823 
824 STATIC unsigned int
825 xfs_merge_ioc_xflags(
826 	unsigned int	flags,
827 	unsigned int	start)
828 {
829 	unsigned int	xflags = start;
830 
831 	if (flags & FS_IMMUTABLE_FL)
832 		xflags |= FS_XFLAG_IMMUTABLE;
833 	else
834 		xflags &= ~FS_XFLAG_IMMUTABLE;
835 	if (flags & FS_APPEND_FL)
836 		xflags |= FS_XFLAG_APPEND;
837 	else
838 		xflags &= ~FS_XFLAG_APPEND;
839 	if (flags & FS_SYNC_FL)
840 		xflags |= FS_XFLAG_SYNC;
841 	else
842 		xflags &= ~FS_XFLAG_SYNC;
843 	if (flags & FS_NOATIME_FL)
844 		xflags |= FS_XFLAG_NOATIME;
845 	else
846 		xflags &= ~FS_XFLAG_NOATIME;
847 	if (flags & FS_NODUMP_FL)
848 		xflags |= FS_XFLAG_NODUMP;
849 	else
850 		xflags &= ~FS_XFLAG_NODUMP;
851 
852 	return xflags;
853 }
854 
855 STATIC unsigned int
856 xfs_di2lxflags(
857 	uint16_t	di_flags)
858 {
859 	unsigned int	flags = 0;
860 
861 	if (di_flags & XFS_DIFLAG_IMMUTABLE)
862 		flags |= FS_IMMUTABLE_FL;
863 	if (di_flags & XFS_DIFLAG_APPEND)
864 		flags |= FS_APPEND_FL;
865 	if (di_flags & XFS_DIFLAG_SYNC)
866 		flags |= FS_SYNC_FL;
867 	if (di_flags & XFS_DIFLAG_NOATIME)
868 		flags |= FS_NOATIME_FL;
869 	if (di_flags & XFS_DIFLAG_NODUMP)
870 		flags |= FS_NODUMP_FL;
871 	return flags;
872 }
873 
874 STATIC int
875 xfs_ioc_fsgetxattr(
876 	xfs_inode_t		*ip,
877 	int			attr,
878 	void			__user *arg)
879 {
880 	struct fsxattr		fa;
881 
882 	memset(&fa, 0, sizeof(struct fsxattr));
883 
884 	xfs_ilock(ip, XFS_ILOCK_SHARED);
885 	fa.fsx_xflags = xfs_ip2xflags(ip);
886 	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
887 	fa.fsx_cowextsize = ip->i_d.di_cowextsize <<
888 			ip->i_mount->m_sb.sb_blocklog;
889 	fa.fsx_projid = xfs_get_projid(ip);
890 
891 	if (attr) {
892 		if (ip->i_afp) {
893 			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
894 				fa.fsx_nextents = xfs_iext_count(ip->i_afp);
895 			else
896 				fa.fsx_nextents = ip->i_d.di_anextents;
897 		} else
898 			fa.fsx_nextents = 0;
899 	} else {
900 		if (ip->i_df.if_flags & XFS_IFEXTENTS)
901 			fa.fsx_nextents = xfs_iext_count(&ip->i_df);
902 		else
903 			fa.fsx_nextents = ip->i_d.di_nextents;
904 	}
905 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
906 
907 	if (copy_to_user(arg, &fa, sizeof(fa)))
908 		return -EFAULT;
909 	return 0;
910 }
911 
912 STATIC uint16_t
913 xfs_flags2diflags(
914 	struct xfs_inode	*ip,
915 	unsigned int		xflags)
916 {
917 	/* can't set PREALLOC this way, just preserve it */
918 	uint16_t		di_flags =
919 		(ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
920 
921 	if (xflags & FS_XFLAG_IMMUTABLE)
922 		di_flags |= XFS_DIFLAG_IMMUTABLE;
923 	if (xflags & FS_XFLAG_APPEND)
924 		di_flags |= XFS_DIFLAG_APPEND;
925 	if (xflags & FS_XFLAG_SYNC)
926 		di_flags |= XFS_DIFLAG_SYNC;
927 	if (xflags & FS_XFLAG_NOATIME)
928 		di_flags |= XFS_DIFLAG_NOATIME;
929 	if (xflags & FS_XFLAG_NODUMP)
930 		di_flags |= XFS_DIFLAG_NODUMP;
931 	if (xflags & FS_XFLAG_NODEFRAG)
932 		di_flags |= XFS_DIFLAG_NODEFRAG;
933 	if (xflags & FS_XFLAG_FILESTREAM)
934 		di_flags |= XFS_DIFLAG_FILESTREAM;
935 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
936 		if (xflags & FS_XFLAG_RTINHERIT)
937 			di_flags |= XFS_DIFLAG_RTINHERIT;
938 		if (xflags & FS_XFLAG_NOSYMLINKS)
939 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
940 		if (xflags & FS_XFLAG_EXTSZINHERIT)
941 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
942 		if (xflags & FS_XFLAG_PROJINHERIT)
943 			di_flags |= XFS_DIFLAG_PROJINHERIT;
944 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
945 		if (xflags & FS_XFLAG_REALTIME)
946 			di_flags |= XFS_DIFLAG_REALTIME;
947 		if (xflags & FS_XFLAG_EXTSIZE)
948 			di_flags |= XFS_DIFLAG_EXTSIZE;
949 	}
950 
951 	return di_flags;
952 }
953 
954 STATIC uint64_t
955 xfs_flags2diflags2(
956 	struct xfs_inode	*ip,
957 	unsigned int		xflags)
958 {
959 	uint64_t		di_flags2 =
960 		(ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK);
961 
962 	if (xflags & FS_XFLAG_DAX)
963 		di_flags2 |= XFS_DIFLAG2_DAX;
964 	if (xflags & FS_XFLAG_COWEXTSIZE)
965 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
966 
967 	return di_flags2;
968 }
969 
970 STATIC void
971 xfs_diflags_to_linux(
972 	struct xfs_inode	*ip)
973 {
974 	struct inode		*inode = VFS_I(ip);
975 	unsigned int		xflags = xfs_ip2xflags(ip);
976 
977 	if (xflags & FS_XFLAG_IMMUTABLE)
978 		inode->i_flags |= S_IMMUTABLE;
979 	else
980 		inode->i_flags &= ~S_IMMUTABLE;
981 	if (xflags & FS_XFLAG_APPEND)
982 		inode->i_flags |= S_APPEND;
983 	else
984 		inode->i_flags &= ~S_APPEND;
985 	if (xflags & FS_XFLAG_SYNC)
986 		inode->i_flags |= S_SYNC;
987 	else
988 		inode->i_flags &= ~S_SYNC;
989 	if (xflags & FS_XFLAG_NOATIME)
990 		inode->i_flags |= S_NOATIME;
991 	else
992 		inode->i_flags &= ~S_NOATIME;
993 #if 0	/* disabled until the flag switching races are sorted out */
994 	if (xflags & FS_XFLAG_DAX)
995 		inode->i_flags |= S_DAX;
996 	else
997 		inode->i_flags &= ~S_DAX;
998 #endif
999 }
1000 
1001 static int
1002 xfs_ioctl_setattr_xflags(
1003 	struct xfs_trans	*tp,
1004 	struct xfs_inode	*ip,
1005 	struct fsxattr		*fa)
1006 {
1007 	struct xfs_mount	*mp = ip->i_mount;
1008 	uint64_t		di_flags2;
1009 
1010 	/* Can't change realtime flag if any extents are allocated. */
1011 	if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1012 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1013 		return -EINVAL;
1014 
1015 	/* If realtime flag is set then must have realtime device */
1016 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1017 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1018 		    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1019 			return -EINVAL;
1020 	}
1021 
1022 	/* Clear reflink if we are actually able to set the rt flag. */
1023 	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1024 		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1025 
1026 	/* Don't allow us to set DAX mode for a reflinked file for now. */
1027 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1028 		return -EINVAL;
1029 
1030 	/*
1031 	 * Can't modify an immutable/append-only file unless
1032 	 * we have appropriate permission.
1033 	 */
1034 	if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1035 	     (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1036 	    !capable(CAP_LINUX_IMMUTABLE))
1037 		return -EPERM;
1038 
1039 	/* diflags2 only valid for v3 inodes. */
1040 	di_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1041 	if (di_flags2 && ip->i_d.di_version < 3)
1042 		return -EINVAL;
1043 
1044 	ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
1045 	ip->i_d.di_flags2 = di_flags2;
1046 
1047 	xfs_diflags_to_linux(ip);
1048 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1049 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1050 	XFS_STATS_INC(mp, xs_ig_attrchg);
1051 	return 0;
1052 }
1053 
1054 /*
1055  * If we are changing DAX flags, we have to ensure the file is clean and any
1056  * cached objects in the address space are invalidated and removed. This
1057  * requires us to lock out other IO and page faults similar to a truncate
1058  * operation. The locks need to be held until the transaction has been committed
1059  * so that the cache invalidation is atomic with respect to the DAX flag
1060  * manipulation.
1061  */
1062 static int
1063 xfs_ioctl_setattr_dax_invalidate(
1064 	struct xfs_inode	*ip,
1065 	struct fsxattr		*fa,
1066 	int			*join_flags)
1067 {
1068 	struct inode		*inode = VFS_I(ip);
1069 	struct super_block	*sb = inode->i_sb;
1070 	int			error;
1071 
1072 	*join_flags = 0;
1073 
1074 	/*
1075 	 * It is only valid to set the DAX flag on regular files and
1076 	 * directories on filesystems where the block size is equal to the page
1077 	 * size. On directories it serves as an inherited hint so we don't
1078 	 * have to check the device for dax support or flush pagecache.
1079 	 */
1080 	if (fa->fsx_xflags & FS_XFLAG_DAX) {
1081 		if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1082 			return -EINVAL;
1083 		if (S_ISREG(inode->i_mode) &&
1084 		    !bdev_dax_supported(xfs_find_bdev_for_inode(VFS_I(ip)),
1085 				sb->s_blocksize))
1086 			return -EINVAL;
1087 	}
1088 
1089 	/* If the DAX state is not changing, we have nothing to do here. */
1090 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1091 		return 0;
1092 	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1093 		return 0;
1094 
1095 	if (S_ISDIR(inode->i_mode))
1096 		return 0;
1097 
1098 	/* lock, flush and invalidate mapping in preparation for flag change */
1099 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1100 	error = filemap_write_and_wait(inode->i_mapping);
1101 	if (error)
1102 		goto out_unlock;
1103 	error = invalidate_inode_pages2(inode->i_mapping);
1104 	if (error)
1105 		goto out_unlock;
1106 
1107 	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1108 	return 0;
1109 
1110 out_unlock:
1111 	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1112 	return error;
1113 
1114 }
1115 
1116 /*
1117  * Set up the transaction structure for the setattr operation, checking that we
1118  * have permission to do so. On success, return a clean transaction and the
1119  * inode locked exclusively ready for further operation specific checks. On
1120  * failure, return an error without modifying or locking the inode.
1121  *
1122  * The inode might already be IO locked on call. If this is the case, it is
1123  * indicated in @join_flags and we take full responsibility for ensuring they
1124  * are unlocked from now on. Hence if we have an error here, we still have to
1125  * unlock them. Otherwise, once they are joined to the transaction, they will
1126  * be unlocked on commit/cancel.
1127  */
1128 static struct xfs_trans *
1129 xfs_ioctl_setattr_get_trans(
1130 	struct xfs_inode	*ip,
1131 	int			join_flags)
1132 {
1133 	struct xfs_mount	*mp = ip->i_mount;
1134 	struct xfs_trans	*tp;
1135 	int			error = -EROFS;
1136 
1137 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1138 		goto out_unlock;
1139 	error = -EIO;
1140 	if (XFS_FORCED_SHUTDOWN(mp))
1141 		goto out_unlock;
1142 
1143 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1144 	if (error)
1145 		return ERR_PTR(error);
1146 
1147 	xfs_ilock(ip, XFS_ILOCK_EXCL);
1148 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1149 	join_flags = 0;
1150 
1151 	/*
1152 	 * CAP_FOWNER overrides the following restrictions:
1153 	 *
1154 	 * The user ID of the calling process must be equal to the file owner
1155 	 * ID, except in cases where the CAP_FSETID capability is applicable.
1156 	 */
1157 	if (!inode_owner_or_capable(VFS_I(ip))) {
1158 		error = -EPERM;
1159 		goto out_cancel;
1160 	}
1161 
1162 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1163 		xfs_trans_set_sync(tp);
1164 
1165 	return tp;
1166 
1167 out_cancel:
1168 	xfs_trans_cancel(tp);
1169 out_unlock:
1170 	if (join_flags)
1171 		xfs_iunlock(ip, join_flags);
1172 	return ERR_PTR(error);
1173 }
1174 
1175 /*
1176  * extent size hint validation is somewhat cumbersome. Rules are:
1177  *
1178  * 1. extent size hint is only valid for directories and regular files
1179  * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1180  * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1181  * 4. can only be changed on regular files if no extents are allocated
1182  * 5. can be changed on directories at any time
1183  * 6. extsize hint of 0 turns off hints, clears inode flags.
1184  * 7. Extent size must be a multiple of the appropriate block size.
1185  * 8. for non-realtime files, the extent size hint must be limited
1186  *    to half the AG size to avoid alignment extending the extent beyond the
1187  *    limits of the AG.
1188  *
1189  * Please keep this function in sync with xfs_scrub_inode_extsize.
1190  */
1191 static int
1192 xfs_ioctl_setattr_check_extsize(
1193 	struct xfs_inode	*ip,
1194 	struct fsxattr		*fa)
1195 {
1196 	struct xfs_mount	*mp = ip->i_mount;
1197 
1198 	if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1199 		return -EINVAL;
1200 
1201 	if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1202 	    !S_ISDIR(VFS_I(ip)->i_mode))
1203 		return -EINVAL;
1204 
1205 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1206 	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1207 		return -EINVAL;
1208 
1209 	if (fa->fsx_extsize != 0) {
1210 		xfs_extlen_t    size;
1211 		xfs_fsblock_t   extsize_fsb;
1212 
1213 		extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1214 		if (extsize_fsb > MAXEXTLEN)
1215 			return -EINVAL;
1216 
1217 		if (XFS_IS_REALTIME_INODE(ip) ||
1218 		    (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1219 			size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1220 		} else {
1221 			size = mp->m_sb.sb_blocksize;
1222 			if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1223 				return -EINVAL;
1224 		}
1225 
1226 		if (fa->fsx_extsize % size)
1227 			return -EINVAL;
1228 	} else
1229 		fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1230 
1231 	return 0;
1232 }
1233 
1234 /*
1235  * CoW extent size hint validation rules are:
1236  *
1237  * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1238  *    The inode does not have to have any shared blocks, but it must be a v3.
1239  * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1240  *    for a directory, the hint is propagated to new files.
1241  * 3. Can be changed on files & directories at any time.
1242  * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1243  * 5. Extent size must be a multiple of the appropriate block size.
1244  * 6. The extent size hint must be limited to half the AG size to avoid
1245  *    alignment extending the extent beyond the limits of the AG.
1246  *
1247  * Please keep this function in sync with xfs_scrub_inode_cowextsize.
1248  */
1249 static int
1250 xfs_ioctl_setattr_check_cowextsize(
1251 	struct xfs_inode	*ip,
1252 	struct fsxattr		*fa)
1253 {
1254 	struct xfs_mount	*mp = ip->i_mount;
1255 
1256 	if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1257 		return 0;
1258 
1259 	if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb) ||
1260 	    ip->i_d.di_version != 3)
1261 		return -EINVAL;
1262 
1263 	if (!S_ISREG(VFS_I(ip)->i_mode) && !S_ISDIR(VFS_I(ip)->i_mode))
1264 		return -EINVAL;
1265 
1266 	if (fa->fsx_cowextsize != 0) {
1267 		xfs_extlen_t    size;
1268 		xfs_fsblock_t   cowextsize_fsb;
1269 
1270 		cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1271 		if (cowextsize_fsb > MAXEXTLEN)
1272 			return -EINVAL;
1273 
1274 		size = mp->m_sb.sb_blocksize;
1275 		if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1276 			return -EINVAL;
1277 
1278 		if (fa->fsx_cowextsize % size)
1279 			return -EINVAL;
1280 	} else
1281 		fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
1282 
1283 	return 0;
1284 }
1285 
1286 static int
1287 xfs_ioctl_setattr_check_projid(
1288 	struct xfs_inode	*ip,
1289 	struct fsxattr		*fa)
1290 {
1291 	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1292 	if (fa->fsx_projid > (uint16_t)-1 &&
1293 	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1294 		return -EINVAL;
1295 
1296 	/*
1297 	 * Project Quota ID state is only allowed to change from within the init
1298 	 * namespace. Enforce that restriction only if we are trying to change
1299 	 * the quota ID state. Everything else is allowed in user namespaces.
1300 	 */
1301 	if (current_user_ns() == &init_user_ns)
1302 		return 0;
1303 
1304 	if (xfs_get_projid(ip) != fa->fsx_projid)
1305 		return -EINVAL;
1306 	if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1307 	    (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1308 		return -EINVAL;
1309 
1310 	return 0;
1311 }
1312 
1313 STATIC int
1314 xfs_ioctl_setattr(
1315 	xfs_inode_t		*ip,
1316 	struct fsxattr		*fa)
1317 {
1318 	struct xfs_mount	*mp = ip->i_mount;
1319 	struct xfs_trans	*tp;
1320 	struct xfs_dquot	*udqp = NULL;
1321 	struct xfs_dquot	*pdqp = NULL;
1322 	struct xfs_dquot	*olddquot = NULL;
1323 	int			code;
1324 	int			join_flags = 0;
1325 
1326 	trace_xfs_ioctl_setattr(ip);
1327 
1328 	code = xfs_ioctl_setattr_check_projid(ip, fa);
1329 	if (code)
1330 		return code;
1331 
1332 	/*
1333 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1334 	 * before we start any other transactions. Trying to do this later
1335 	 * is messy. We don't care to take a readlock to look at the ids
1336 	 * in inode here, because we can't hold it across the trans_reserve.
1337 	 * If the IDs do change before we take the ilock, we're covered
1338 	 * because the i_*dquot fields will get updated anyway.
1339 	 */
1340 	if (XFS_IS_QUOTA_ON(mp)) {
1341 		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1342 					 ip->i_d.di_gid, fa->fsx_projid,
1343 					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1344 		if (code)
1345 			return code;
1346 	}
1347 
1348 	/*
1349 	 * Changing DAX config may require inode locking for mapping
1350 	 * invalidation. These need to be held all the way to transaction commit
1351 	 * or cancel time, so need to be passed through to
1352 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1353 	 * appropriately.
1354 	 */
1355 	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1356 	if (code)
1357 		goto error_free_dquots;
1358 
1359 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1360 	if (IS_ERR(tp)) {
1361 		code = PTR_ERR(tp);
1362 		goto error_free_dquots;
1363 	}
1364 
1365 
1366 	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1367 	    xfs_get_projid(ip) != fa->fsx_projid) {
1368 		code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1369 				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1370 		if (code)	/* out of quota */
1371 			goto error_trans_cancel;
1372 	}
1373 
1374 	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1375 	if (code)
1376 		goto error_trans_cancel;
1377 
1378 	code = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1379 	if (code)
1380 		goto error_trans_cancel;
1381 
1382 	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1383 	if (code)
1384 		goto error_trans_cancel;
1385 
1386 	/*
1387 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1388 	 * overrides the following restrictions:
1389 	 *
1390 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1391 	 * successful return from chown()
1392 	 */
1393 
1394 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1395 	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1396 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1397 
1398 	/* Change the ownerships and register project quota modifications */
1399 	if (xfs_get_projid(ip) != fa->fsx_projid) {
1400 		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1401 			olddquot = xfs_qm_vop_chown(tp, ip,
1402 						&ip->i_pdquot, pdqp);
1403 		}
1404 		ASSERT(ip->i_d.di_version > 1);
1405 		xfs_set_projid(ip, fa->fsx_projid);
1406 	}
1407 
1408 	/*
1409 	 * Only set the extent size hint if we've already determined that the
1410 	 * extent size hint should be set on the inode. If no extent size flags
1411 	 * are set on the inode then unconditionally clear the extent size hint.
1412 	 */
1413 	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1414 		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1415 	else
1416 		ip->i_d.di_extsize = 0;
1417 	if (ip->i_d.di_version == 3 &&
1418 	    (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1419 		ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1420 				mp->m_sb.sb_blocklog;
1421 	else
1422 		ip->i_d.di_cowextsize = 0;
1423 
1424 	code = xfs_trans_commit(tp);
1425 
1426 	/*
1427 	 * Release any dquot(s) the inode had kept before chown.
1428 	 */
1429 	xfs_qm_dqrele(olddquot);
1430 	xfs_qm_dqrele(udqp);
1431 	xfs_qm_dqrele(pdqp);
1432 
1433 	return code;
1434 
1435 error_trans_cancel:
1436 	xfs_trans_cancel(tp);
1437 error_free_dquots:
1438 	xfs_qm_dqrele(udqp);
1439 	xfs_qm_dqrele(pdqp);
1440 	return code;
1441 }
1442 
1443 STATIC int
1444 xfs_ioc_fssetxattr(
1445 	xfs_inode_t		*ip,
1446 	struct file		*filp,
1447 	void			__user *arg)
1448 {
1449 	struct fsxattr		fa;
1450 	int error;
1451 
1452 	if (copy_from_user(&fa, arg, sizeof(fa)))
1453 		return -EFAULT;
1454 
1455 	error = mnt_want_write_file(filp);
1456 	if (error)
1457 		return error;
1458 	error = xfs_ioctl_setattr(ip, &fa);
1459 	mnt_drop_write_file(filp);
1460 	return error;
1461 }
1462 
1463 STATIC int
1464 xfs_ioc_getxflags(
1465 	xfs_inode_t		*ip,
1466 	void			__user *arg)
1467 {
1468 	unsigned int		flags;
1469 
1470 	flags = xfs_di2lxflags(ip->i_d.di_flags);
1471 	if (copy_to_user(arg, &flags, sizeof(flags)))
1472 		return -EFAULT;
1473 	return 0;
1474 }
1475 
1476 STATIC int
1477 xfs_ioc_setxflags(
1478 	struct xfs_inode	*ip,
1479 	struct file		*filp,
1480 	void			__user *arg)
1481 {
1482 	struct xfs_trans	*tp;
1483 	struct fsxattr		fa;
1484 	unsigned int		flags;
1485 	int			join_flags = 0;
1486 	int			error;
1487 
1488 	if (copy_from_user(&flags, arg, sizeof(flags)))
1489 		return -EFAULT;
1490 
1491 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1492 		      FS_NOATIME_FL | FS_NODUMP_FL | \
1493 		      FS_SYNC_FL))
1494 		return -EOPNOTSUPP;
1495 
1496 	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1497 
1498 	error = mnt_want_write_file(filp);
1499 	if (error)
1500 		return error;
1501 
1502 	/*
1503 	 * Changing DAX config may require inode locking for mapping
1504 	 * invalidation. These need to be held all the way to transaction commit
1505 	 * or cancel time, so need to be passed through to
1506 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1507 	 * appropriately.
1508 	 */
1509 	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1510 	if (error)
1511 		goto out_drop_write;
1512 
1513 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1514 	if (IS_ERR(tp)) {
1515 		error = PTR_ERR(tp);
1516 		goto out_drop_write;
1517 	}
1518 
1519 	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1520 	if (error) {
1521 		xfs_trans_cancel(tp);
1522 		goto out_drop_write;
1523 	}
1524 
1525 	error = xfs_trans_commit(tp);
1526 out_drop_write:
1527 	mnt_drop_write_file(filp);
1528 	return error;
1529 }
1530 
1531 static bool
1532 xfs_getbmap_format(
1533 	struct kgetbmap		*p,
1534 	struct getbmapx __user	*u,
1535 	size_t			recsize)
1536 {
1537 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
1538 	    put_user(p->bmv_block, &u->bmv_block) ||
1539 	    put_user(p->bmv_length, &u->bmv_length) ||
1540 	    put_user(0, &u->bmv_count) ||
1541 	    put_user(0, &u->bmv_entries))
1542 		return false;
1543 	if (recsize < sizeof(struct getbmapx))
1544 		return true;
1545 	if (put_user(0, &u->bmv_iflags) ||
1546 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
1547 	    put_user(0, &u->bmv_unused1) ||
1548 	    put_user(0, &u->bmv_unused2))
1549 		return false;
1550 	return true;
1551 }
1552 
1553 STATIC int
1554 xfs_ioc_getbmap(
1555 	struct file		*file,
1556 	unsigned int		cmd,
1557 	void			__user *arg)
1558 {
1559 	struct getbmapx		bmx = { 0 };
1560 	struct kgetbmap		*buf;
1561 	size_t			recsize;
1562 	int			error, i;
1563 
1564 	switch (cmd) {
1565 	case XFS_IOC_GETBMAPA:
1566 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
1567 		/*FALLTHRU*/
1568 	case XFS_IOC_GETBMAP:
1569 		if (file->f_mode & FMODE_NOCMTIME)
1570 			bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1571 		/* struct getbmap is a strict subset of struct getbmapx. */
1572 		recsize = sizeof(struct getbmap);
1573 		break;
1574 	case XFS_IOC_GETBMAPX:
1575 		recsize = sizeof(struct getbmapx);
1576 		break;
1577 	default:
1578 		return -EINVAL;
1579 	}
1580 
1581 	if (copy_from_user(&bmx, arg, recsize))
1582 		return -EFAULT;
1583 
1584 	if (bmx.bmv_count < 2)
1585 		return -EINVAL;
1586 	if (bmx.bmv_count > ULONG_MAX / recsize)
1587 		return -ENOMEM;
1588 
1589 	buf = kmem_zalloc_large(bmx.bmv_count * sizeof(*buf), 0);
1590 	if (!buf)
1591 		return -ENOMEM;
1592 
1593 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1594 	if (error)
1595 		goto out_free_buf;
1596 
1597 	error = -EFAULT;
1598 	if (copy_to_user(arg, &bmx, recsize))
1599 		goto out_free_buf;
1600 	arg += recsize;
1601 
1602 	for (i = 0; i < bmx.bmv_entries; i++) {
1603 		if (!xfs_getbmap_format(buf + i, arg, recsize))
1604 			goto out_free_buf;
1605 		arg += recsize;
1606 	}
1607 
1608 	error = 0;
1609 out_free_buf:
1610 	kmem_free(buf);
1611 	return error;
1612 }
1613 
1614 struct getfsmap_info {
1615 	struct xfs_mount	*mp;
1616 	struct fsmap_head __user *data;
1617 	unsigned int		idx;
1618 	__u32			last_flags;
1619 };
1620 
1621 STATIC int
1622 xfs_getfsmap_format(struct xfs_fsmap *xfm, void *priv)
1623 {
1624 	struct getfsmap_info	*info = priv;
1625 	struct fsmap		fm;
1626 
1627 	trace_xfs_getfsmap_mapping(info->mp, xfm);
1628 
1629 	info->last_flags = xfm->fmr_flags;
1630 	xfs_fsmap_from_internal(&fm, xfm);
1631 	if (copy_to_user(&info->data->fmh_recs[info->idx++], &fm,
1632 			sizeof(struct fsmap)))
1633 		return -EFAULT;
1634 
1635 	return 0;
1636 }
1637 
1638 STATIC int
1639 xfs_ioc_getfsmap(
1640 	struct xfs_inode	*ip,
1641 	struct fsmap_head	__user *arg)
1642 {
1643 	struct getfsmap_info	info = { NULL };
1644 	struct xfs_fsmap_head	xhead = {0};
1645 	struct fsmap_head	head;
1646 	bool			aborted = false;
1647 	int			error;
1648 
1649 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1650 		return -EFAULT;
1651 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1652 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1653 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1654 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1655 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1656 		return -EINVAL;
1657 
1658 	xhead.fmh_iflags = head.fmh_iflags;
1659 	xhead.fmh_count = head.fmh_count;
1660 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1661 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1662 
1663 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1664 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1665 
1666 	info.mp = ip->i_mount;
1667 	info.data = arg;
1668 	error = xfs_getfsmap(ip->i_mount, &xhead, xfs_getfsmap_format, &info);
1669 	if (error == XFS_BTREE_QUERY_RANGE_ABORT) {
1670 		error = 0;
1671 		aborted = true;
1672 	} else if (error)
1673 		return error;
1674 
1675 	/* If we didn't abort, set the "last" flag in the last fmx */
1676 	if (!aborted && info.idx) {
1677 		info.last_flags |= FMR_OF_LAST;
1678 		if (copy_to_user(&info.data->fmh_recs[info.idx - 1].fmr_flags,
1679 				&info.last_flags, sizeof(info.last_flags)))
1680 			return -EFAULT;
1681 	}
1682 
1683 	/* copy back header */
1684 	head.fmh_entries = xhead.fmh_entries;
1685 	head.fmh_oflags = xhead.fmh_oflags;
1686 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
1687 		return -EFAULT;
1688 
1689 	return 0;
1690 }
1691 
1692 STATIC int
1693 xfs_ioc_scrub_metadata(
1694 	struct xfs_inode		*ip,
1695 	void				__user *arg)
1696 {
1697 	struct xfs_scrub_metadata	scrub;
1698 	int				error;
1699 
1700 	if (!capable(CAP_SYS_ADMIN))
1701 		return -EPERM;
1702 
1703 	if (copy_from_user(&scrub, arg, sizeof(scrub)))
1704 		return -EFAULT;
1705 
1706 	error = xfs_scrub_metadata(ip, &scrub);
1707 	if (error)
1708 		return error;
1709 
1710 	if (copy_to_user(arg, &scrub, sizeof(scrub)))
1711 		return -EFAULT;
1712 
1713 	return 0;
1714 }
1715 
1716 int
1717 xfs_ioc_swapext(
1718 	xfs_swapext_t	*sxp)
1719 {
1720 	xfs_inode_t     *ip, *tip;
1721 	struct fd	f, tmp;
1722 	int		error = 0;
1723 
1724 	/* Pull information for the target fd */
1725 	f = fdget((int)sxp->sx_fdtarget);
1726 	if (!f.file) {
1727 		error = -EINVAL;
1728 		goto out;
1729 	}
1730 
1731 	if (!(f.file->f_mode & FMODE_WRITE) ||
1732 	    !(f.file->f_mode & FMODE_READ) ||
1733 	    (f.file->f_flags & O_APPEND)) {
1734 		error = -EBADF;
1735 		goto out_put_file;
1736 	}
1737 
1738 	tmp = fdget((int)sxp->sx_fdtmp);
1739 	if (!tmp.file) {
1740 		error = -EINVAL;
1741 		goto out_put_file;
1742 	}
1743 
1744 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1745 	    !(tmp.file->f_mode & FMODE_READ) ||
1746 	    (tmp.file->f_flags & O_APPEND)) {
1747 		error = -EBADF;
1748 		goto out_put_tmp_file;
1749 	}
1750 
1751 	if (IS_SWAPFILE(file_inode(f.file)) ||
1752 	    IS_SWAPFILE(file_inode(tmp.file))) {
1753 		error = -EINVAL;
1754 		goto out_put_tmp_file;
1755 	}
1756 
1757 	/*
1758 	 * We need to ensure that the fds passed in point to XFS inodes
1759 	 * before we cast and access them as XFS structures as we have no
1760 	 * control over what the user passes us here.
1761 	 */
1762 	if (f.file->f_op != &xfs_file_operations ||
1763 	    tmp.file->f_op != &xfs_file_operations) {
1764 		error = -EINVAL;
1765 		goto out_put_tmp_file;
1766 	}
1767 
1768 	ip = XFS_I(file_inode(f.file));
1769 	tip = XFS_I(file_inode(tmp.file));
1770 
1771 	if (ip->i_mount != tip->i_mount) {
1772 		error = -EINVAL;
1773 		goto out_put_tmp_file;
1774 	}
1775 
1776 	if (ip->i_ino == tip->i_ino) {
1777 		error = -EINVAL;
1778 		goto out_put_tmp_file;
1779 	}
1780 
1781 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1782 		error = -EIO;
1783 		goto out_put_tmp_file;
1784 	}
1785 
1786 	error = xfs_swap_extents(ip, tip, sxp);
1787 
1788  out_put_tmp_file:
1789 	fdput(tmp);
1790  out_put_file:
1791 	fdput(f);
1792  out:
1793 	return error;
1794 }
1795 
1796 static int
1797 xfs_ioc_getlabel(
1798 	struct xfs_mount	*mp,
1799 	char			__user *user_label)
1800 {
1801 	struct xfs_sb		*sbp = &mp->m_sb;
1802 	char			label[XFSLABEL_MAX + 1];
1803 
1804 	/* Paranoia */
1805 	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1806 
1807 	/* 1 larger than sb_fname, so this ensures a trailing NUL char */
1808 	memset(label, 0, sizeof(label));
1809 	spin_lock(&mp->m_sb_lock);
1810 	strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1811 	spin_unlock(&mp->m_sb_lock);
1812 
1813 	if (copy_to_user(user_label, label, sizeof(label)))
1814 		return -EFAULT;
1815 	return 0;
1816 }
1817 
1818 static int
1819 xfs_ioc_setlabel(
1820 	struct file		*filp,
1821 	struct xfs_mount	*mp,
1822 	char			__user *newlabel)
1823 {
1824 	struct xfs_sb		*sbp = &mp->m_sb;
1825 	char			label[XFSLABEL_MAX + 1];
1826 	size_t			len;
1827 	int			error;
1828 
1829 	if (!capable(CAP_SYS_ADMIN))
1830 		return -EPERM;
1831 	/*
1832 	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1833 	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1834 	 * (required) NULL character to test the incoming label length.
1835 	 * NB: The on disk label doesn't need to be null terminated.
1836 	 */
1837 	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1838 		return -EFAULT;
1839 	len = strnlen(label, XFSLABEL_MAX + 1);
1840 	if (len > sizeof(sbp->sb_fname))
1841 		return -EINVAL;
1842 
1843 	error = mnt_want_write_file(filp);
1844 	if (error)
1845 		return error;
1846 
1847 	spin_lock(&mp->m_sb_lock);
1848 	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1849 	memcpy(sbp->sb_fname, label, len);
1850 	spin_unlock(&mp->m_sb_lock);
1851 
1852 	/*
1853 	 * Now we do several things to satisfy userspace.
1854 	 * In addition to normal logging of the primary superblock, we also
1855 	 * immediately write these changes to sector zero for the primary, then
1856 	 * update all backup supers (as xfs_db does for a label change), then
1857 	 * invalidate the block device page cache.  This is so that any prior
1858 	 * buffered reads from userspace (i.e. from blkid) are invalidated,
1859 	 * and userspace will see the newly-written label.
1860 	 */
1861 	error = xfs_sync_sb_buf(mp);
1862 	if (error)
1863 		goto out;
1864 	/*
1865 	 * growfs also updates backup supers so lock against that.
1866 	 */
1867 	mutex_lock(&mp->m_growlock);
1868 	error = xfs_update_secondary_sbs(mp);
1869 	mutex_unlock(&mp->m_growlock);
1870 
1871 	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1872 
1873 out:
1874 	mnt_drop_write_file(filp);
1875 	return error;
1876 }
1877 
1878 /*
1879  * Note: some of the ioctl's return positive numbers as a
1880  * byte count indicating success, such as readlink_by_handle.
1881  * So we don't "sign flip" like most other routines.  This means
1882  * true errors need to be returned as a negative value.
1883  */
1884 long
1885 xfs_file_ioctl(
1886 	struct file		*filp,
1887 	unsigned int		cmd,
1888 	unsigned long		p)
1889 {
1890 	struct inode		*inode = file_inode(filp);
1891 	struct xfs_inode	*ip = XFS_I(inode);
1892 	struct xfs_mount	*mp = ip->i_mount;
1893 	void			__user *arg = (void __user *)p;
1894 	int			error;
1895 
1896 	trace_xfs_file_ioctl(ip);
1897 
1898 	switch (cmd) {
1899 	case FITRIM:
1900 		return xfs_ioc_trim(mp, arg);
1901 	case FS_IOC_GETFSLABEL:
1902 		return xfs_ioc_getlabel(mp, arg);
1903 	case FS_IOC_SETFSLABEL:
1904 		return xfs_ioc_setlabel(filp, mp, arg);
1905 	case XFS_IOC_ALLOCSP:
1906 	case XFS_IOC_FREESP:
1907 	case XFS_IOC_RESVSP:
1908 	case XFS_IOC_UNRESVSP:
1909 	case XFS_IOC_ALLOCSP64:
1910 	case XFS_IOC_FREESP64:
1911 	case XFS_IOC_RESVSP64:
1912 	case XFS_IOC_UNRESVSP64:
1913 	case XFS_IOC_ZERO_RANGE: {
1914 		xfs_flock64_t		bf;
1915 
1916 		if (copy_from_user(&bf, arg, sizeof(bf)))
1917 			return -EFAULT;
1918 		return xfs_ioc_space(filp, cmd, &bf);
1919 	}
1920 	case XFS_IOC_DIOINFO: {
1921 		struct dioattr	da;
1922 		xfs_buftarg_t	*target =
1923 			XFS_IS_REALTIME_INODE(ip) ?
1924 			mp->m_rtdev_targp : mp->m_ddev_targp;
1925 
1926 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1927 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1928 
1929 		if (copy_to_user(arg, &da, sizeof(da)))
1930 			return -EFAULT;
1931 		return 0;
1932 	}
1933 
1934 	case XFS_IOC_FSBULKSTAT_SINGLE:
1935 	case XFS_IOC_FSBULKSTAT:
1936 	case XFS_IOC_FSINUMBERS:
1937 		return xfs_ioc_bulkstat(mp, cmd, arg);
1938 
1939 	case XFS_IOC_FSGEOMETRY_V1:
1940 		return xfs_ioc_fsgeometry_v1(mp, arg);
1941 
1942 	case XFS_IOC_FSGEOMETRY:
1943 		return xfs_ioc_fsgeometry(mp, arg);
1944 
1945 	case XFS_IOC_GETVERSION:
1946 		return put_user(inode->i_generation, (int __user *)arg);
1947 
1948 	case XFS_IOC_FSGETXATTR:
1949 		return xfs_ioc_fsgetxattr(ip, 0, arg);
1950 	case XFS_IOC_FSGETXATTRA:
1951 		return xfs_ioc_fsgetxattr(ip, 1, arg);
1952 	case XFS_IOC_FSSETXATTR:
1953 		return xfs_ioc_fssetxattr(ip, filp, arg);
1954 	case XFS_IOC_GETXFLAGS:
1955 		return xfs_ioc_getxflags(ip, arg);
1956 	case XFS_IOC_SETXFLAGS:
1957 		return xfs_ioc_setxflags(ip, filp, arg);
1958 
1959 	case XFS_IOC_FSSETDM: {
1960 		struct fsdmidata	dmi;
1961 
1962 		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1963 			return -EFAULT;
1964 
1965 		error = mnt_want_write_file(filp);
1966 		if (error)
1967 			return error;
1968 
1969 		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1970 				dmi.fsd_dmstate);
1971 		mnt_drop_write_file(filp);
1972 		return error;
1973 	}
1974 
1975 	case XFS_IOC_GETBMAP:
1976 	case XFS_IOC_GETBMAPA:
1977 	case XFS_IOC_GETBMAPX:
1978 		return xfs_ioc_getbmap(filp, cmd, arg);
1979 
1980 	case FS_IOC_GETFSMAP:
1981 		return xfs_ioc_getfsmap(ip, arg);
1982 
1983 	case XFS_IOC_SCRUB_METADATA:
1984 		return xfs_ioc_scrub_metadata(ip, arg);
1985 
1986 	case XFS_IOC_FD_TO_HANDLE:
1987 	case XFS_IOC_PATH_TO_HANDLE:
1988 	case XFS_IOC_PATH_TO_FSHANDLE: {
1989 		xfs_fsop_handlereq_t	hreq;
1990 
1991 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1992 			return -EFAULT;
1993 		return xfs_find_handle(cmd, &hreq);
1994 	}
1995 	case XFS_IOC_OPEN_BY_HANDLE: {
1996 		xfs_fsop_handlereq_t	hreq;
1997 
1998 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1999 			return -EFAULT;
2000 		return xfs_open_by_handle(filp, &hreq);
2001 	}
2002 	case XFS_IOC_FSSETDM_BY_HANDLE:
2003 		return xfs_fssetdm_by_handle(filp, arg);
2004 
2005 	case XFS_IOC_READLINK_BY_HANDLE: {
2006 		xfs_fsop_handlereq_t	hreq;
2007 
2008 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2009 			return -EFAULT;
2010 		return xfs_readlink_by_handle(filp, &hreq);
2011 	}
2012 	case XFS_IOC_ATTRLIST_BY_HANDLE:
2013 		return xfs_attrlist_by_handle(filp, arg);
2014 
2015 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
2016 		return xfs_attrmulti_by_handle(filp, arg);
2017 
2018 	case XFS_IOC_SWAPEXT: {
2019 		struct xfs_swapext	sxp;
2020 
2021 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2022 			return -EFAULT;
2023 		error = mnt_want_write_file(filp);
2024 		if (error)
2025 			return error;
2026 		error = xfs_ioc_swapext(&sxp);
2027 		mnt_drop_write_file(filp);
2028 		return error;
2029 	}
2030 
2031 	case XFS_IOC_FSCOUNTS: {
2032 		xfs_fsop_counts_t out;
2033 
2034 		error = xfs_fs_counts(mp, &out);
2035 		if (error)
2036 			return error;
2037 
2038 		if (copy_to_user(arg, &out, sizeof(out)))
2039 			return -EFAULT;
2040 		return 0;
2041 	}
2042 
2043 	case XFS_IOC_SET_RESBLKS: {
2044 		xfs_fsop_resblks_t inout;
2045 		uint64_t	   in;
2046 
2047 		if (!capable(CAP_SYS_ADMIN))
2048 			return -EPERM;
2049 
2050 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2051 			return -EROFS;
2052 
2053 		if (copy_from_user(&inout, arg, sizeof(inout)))
2054 			return -EFAULT;
2055 
2056 		error = mnt_want_write_file(filp);
2057 		if (error)
2058 			return error;
2059 
2060 		/* input parameter is passed in resblks field of structure */
2061 		in = inout.resblks;
2062 		error = xfs_reserve_blocks(mp, &in, &inout);
2063 		mnt_drop_write_file(filp);
2064 		if (error)
2065 			return error;
2066 
2067 		if (copy_to_user(arg, &inout, sizeof(inout)))
2068 			return -EFAULT;
2069 		return 0;
2070 	}
2071 
2072 	case XFS_IOC_GET_RESBLKS: {
2073 		xfs_fsop_resblks_t out;
2074 
2075 		if (!capable(CAP_SYS_ADMIN))
2076 			return -EPERM;
2077 
2078 		error = xfs_reserve_blocks(mp, NULL, &out);
2079 		if (error)
2080 			return error;
2081 
2082 		if (copy_to_user(arg, &out, sizeof(out)))
2083 			return -EFAULT;
2084 
2085 		return 0;
2086 	}
2087 
2088 	case XFS_IOC_FSGROWFSDATA: {
2089 		xfs_growfs_data_t in;
2090 
2091 		if (copy_from_user(&in, arg, sizeof(in)))
2092 			return -EFAULT;
2093 
2094 		error = mnt_want_write_file(filp);
2095 		if (error)
2096 			return error;
2097 		error = xfs_growfs_data(mp, &in);
2098 		mnt_drop_write_file(filp);
2099 		return error;
2100 	}
2101 
2102 	case XFS_IOC_FSGROWFSLOG: {
2103 		xfs_growfs_log_t in;
2104 
2105 		if (copy_from_user(&in, arg, sizeof(in)))
2106 			return -EFAULT;
2107 
2108 		error = mnt_want_write_file(filp);
2109 		if (error)
2110 			return error;
2111 		error = xfs_growfs_log(mp, &in);
2112 		mnt_drop_write_file(filp);
2113 		return error;
2114 	}
2115 
2116 	case XFS_IOC_FSGROWFSRT: {
2117 		xfs_growfs_rt_t in;
2118 
2119 		if (copy_from_user(&in, arg, sizeof(in)))
2120 			return -EFAULT;
2121 
2122 		error = mnt_want_write_file(filp);
2123 		if (error)
2124 			return error;
2125 		error = xfs_growfs_rt(mp, &in);
2126 		mnt_drop_write_file(filp);
2127 		return error;
2128 	}
2129 
2130 	case XFS_IOC_GOINGDOWN: {
2131 		uint32_t in;
2132 
2133 		if (!capable(CAP_SYS_ADMIN))
2134 			return -EPERM;
2135 
2136 		if (get_user(in, (uint32_t __user *)arg))
2137 			return -EFAULT;
2138 
2139 		return xfs_fs_goingdown(mp, in);
2140 	}
2141 
2142 	case XFS_IOC_ERROR_INJECTION: {
2143 		xfs_error_injection_t in;
2144 
2145 		if (!capable(CAP_SYS_ADMIN))
2146 			return -EPERM;
2147 
2148 		if (copy_from_user(&in, arg, sizeof(in)))
2149 			return -EFAULT;
2150 
2151 		return xfs_errortag_add(mp, in.errtag);
2152 	}
2153 
2154 	case XFS_IOC_ERROR_CLEARALL:
2155 		if (!capable(CAP_SYS_ADMIN))
2156 			return -EPERM;
2157 
2158 		return xfs_errortag_clearall(mp);
2159 
2160 	case XFS_IOC_FREE_EOFBLOCKS: {
2161 		struct xfs_fs_eofblocks eofb;
2162 		struct xfs_eofblocks keofb;
2163 
2164 		if (!capable(CAP_SYS_ADMIN))
2165 			return -EPERM;
2166 
2167 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2168 			return -EROFS;
2169 
2170 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2171 			return -EFAULT;
2172 
2173 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2174 		if (error)
2175 			return error;
2176 
2177 		return xfs_icache_free_eofblocks(mp, &keofb);
2178 	}
2179 
2180 	default:
2181 		return -ENOTTY;
2182 	}
2183 }
2184