xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision 7211ec63)
1 /*
2  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3  * All Rights Reserved.
4  *
5  * This program is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU General Public License as
7  * published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope that it would be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12  * GNU General Public License for more details.
13  *
14  * You should have received a copy of the GNU General Public License
15  * along with this program; if not, write the Free Software Foundation,
16  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17  */
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
24 #include "xfs_mount.h"
25 #include "xfs_inode.h"
26 #include "xfs_ioctl.h"
27 #include "xfs_alloc.h"
28 #include "xfs_rtalloc.h"
29 #include "xfs_itable.h"
30 #include "xfs_error.h"
31 #include "xfs_attr.h"
32 #include "xfs_bmap.h"
33 #include "xfs_bmap_util.h"
34 #include "xfs_fsops.h"
35 #include "xfs_discard.h"
36 #include "xfs_quota.h"
37 #include "xfs_export.h"
38 #include "xfs_trace.h"
39 #include "xfs_icache.h"
40 #include "xfs_symlink.h"
41 #include "xfs_trans.h"
42 #include "xfs_pnfs.h"
43 #include "xfs_acl.h"
44 #include "xfs_btree.h"
45 #include <linux/fsmap.h>
46 #include "xfs_fsmap.h"
47 
48 #include <linux/capability.h>
49 #include <linux/cred.h>
50 #include <linux/dcache.h>
51 #include <linux/mount.h>
52 #include <linux/namei.h>
53 #include <linux/pagemap.h>
54 #include <linux/slab.h>
55 #include <linux/exportfs.h>
56 
57 /*
58  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
59  * a file or fs handle.
60  *
61  * XFS_IOC_PATH_TO_FSHANDLE
62  *    returns fs handle for a mount point or path within that mount point
63  * XFS_IOC_FD_TO_HANDLE
64  *    returns full handle for a FD opened in user space
65  * XFS_IOC_PATH_TO_HANDLE
66  *    returns full handle for a path
67  */
68 int
69 xfs_find_handle(
70 	unsigned int		cmd,
71 	xfs_fsop_handlereq_t	*hreq)
72 {
73 	int			hsize;
74 	xfs_handle_t		handle;
75 	struct inode		*inode;
76 	struct fd		f = {NULL};
77 	struct path		path;
78 	int			error;
79 	struct xfs_inode	*ip;
80 
81 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
82 		f = fdget(hreq->fd);
83 		if (!f.file)
84 			return -EBADF;
85 		inode = file_inode(f.file);
86 	} else {
87 		error = user_lpath((const char __user *)hreq->path, &path);
88 		if (error)
89 			return error;
90 		inode = d_inode(path.dentry);
91 	}
92 	ip = XFS_I(inode);
93 
94 	/*
95 	 * We can only generate handles for inodes residing on a XFS filesystem,
96 	 * and only for regular files, directories or symbolic links.
97 	 */
98 	error = -EINVAL;
99 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
100 		goto out_put;
101 
102 	error = -EBADF;
103 	if (!S_ISREG(inode->i_mode) &&
104 	    !S_ISDIR(inode->i_mode) &&
105 	    !S_ISLNK(inode->i_mode))
106 		goto out_put;
107 
108 
109 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
110 
111 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
112 		/*
113 		 * This handle only contains an fsid, zero the rest.
114 		 */
115 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
116 		hsize = sizeof(xfs_fsid_t);
117 	} else {
118 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
119 					sizeof(handle.ha_fid.fid_len);
120 		handle.ha_fid.fid_pad = 0;
121 		handle.ha_fid.fid_gen = inode->i_generation;
122 		handle.ha_fid.fid_ino = ip->i_ino;
123 		hsize = sizeof(xfs_handle_t);
124 	}
125 
126 	error = -EFAULT;
127 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
128 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
129 		goto out_put;
130 
131 	error = 0;
132 
133  out_put:
134 	if (cmd == XFS_IOC_FD_TO_HANDLE)
135 		fdput(f);
136 	else
137 		path_put(&path);
138 	return error;
139 }
140 
141 /*
142  * No need to do permission checks on the various pathname components
143  * as the handle operations are privileged.
144  */
145 STATIC int
146 xfs_handle_acceptable(
147 	void			*context,
148 	struct dentry		*dentry)
149 {
150 	return 1;
151 }
152 
153 /*
154  * Convert userspace handle data into a dentry.
155  */
156 struct dentry *
157 xfs_handle_to_dentry(
158 	struct file		*parfilp,
159 	void __user		*uhandle,
160 	u32			hlen)
161 {
162 	xfs_handle_t		handle;
163 	struct xfs_fid64	fid;
164 
165 	/*
166 	 * Only allow handle opens under a directory.
167 	 */
168 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
169 		return ERR_PTR(-ENOTDIR);
170 
171 	if (hlen != sizeof(xfs_handle_t))
172 		return ERR_PTR(-EINVAL);
173 	if (copy_from_user(&handle, uhandle, hlen))
174 		return ERR_PTR(-EFAULT);
175 	if (handle.ha_fid.fid_len !=
176 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
177 		return ERR_PTR(-EINVAL);
178 
179 	memset(&fid, 0, sizeof(struct fid));
180 	fid.ino = handle.ha_fid.fid_ino;
181 	fid.gen = handle.ha_fid.fid_gen;
182 
183 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
184 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
185 			xfs_handle_acceptable, NULL);
186 }
187 
188 STATIC struct dentry *
189 xfs_handlereq_to_dentry(
190 	struct file		*parfilp,
191 	xfs_fsop_handlereq_t	*hreq)
192 {
193 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
194 }
195 
196 int
197 xfs_open_by_handle(
198 	struct file		*parfilp,
199 	xfs_fsop_handlereq_t	*hreq)
200 {
201 	const struct cred	*cred = current_cred();
202 	int			error;
203 	int			fd;
204 	int			permflag;
205 	struct file		*filp;
206 	struct inode		*inode;
207 	struct dentry		*dentry;
208 	fmode_t			fmode;
209 	struct path		path;
210 
211 	if (!capable(CAP_SYS_ADMIN))
212 		return -EPERM;
213 
214 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
215 	if (IS_ERR(dentry))
216 		return PTR_ERR(dentry);
217 	inode = d_inode(dentry);
218 
219 	/* Restrict xfs_open_by_handle to directories & regular files. */
220 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
221 		error = -EPERM;
222 		goto out_dput;
223 	}
224 
225 #if BITS_PER_LONG != 32
226 	hreq->oflags |= O_LARGEFILE;
227 #endif
228 
229 	permflag = hreq->oflags;
230 	fmode = OPEN_FMODE(permflag);
231 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
232 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
233 		error = -EPERM;
234 		goto out_dput;
235 	}
236 
237 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
238 		error = -EPERM;
239 		goto out_dput;
240 	}
241 
242 	/* Can't write directories. */
243 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
244 		error = -EISDIR;
245 		goto out_dput;
246 	}
247 
248 	fd = get_unused_fd_flags(0);
249 	if (fd < 0) {
250 		error = fd;
251 		goto out_dput;
252 	}
253 
254 	path.mnt = parfilp->f_path.mnt;
255 	path.dentry = dentry;
256 	filp = dentry_open(&path, hreq->oflags, cred);
257 	dput(dentry);
258 	if (IS_ERR(filp)) {
259 		put_unused_fd(fd);
260 		return PTR_ERR(filp);
261 	}
262 
263 	if (S_ISREG(inode->i_mode)) {
264 		filp->f_flags |= O_NOATIME;
265 		filp->f_mode |= FMODE_NOCMTIME;
266 	}
267 
268 	fd_install(fd, filp);
269 	return fd;
270 
271  out_dput:
272 	dput(dentry);
273 	return error;
274 }
275 
276 int
277 xfs_readlink_by_handle(
278 	struct file		*parfilp,
279 	xfs_fsop_handlereq_t	*hreq)
280 {
281 	struct dentry		*dentry;
282 	__u32			olen;
283 	int			error;
284 
285 	if (!capable(CAP_SYS_ADMIN))
286 		return -EPERM;
287 
288 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
289 	if (IS_ERR(dentry))
290 		return PTR_ERR(dentry);
291 
292 	/* Restrict this handle operation to symlinks only. */
293 	if (!d_is_symlink(dentry)) {
294 		error = -EINVAL;
295 		goto out_dput;
296 	}
297 
298 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
299 		error = -EFAULT;
300 		goto out_dput;
301 	}
302 
303 	error = vfs_readlink(dentry, hreq->ohandle, olen);
304 
305  out_dput:
306 	dput(dentry);
307 	return error;
308 }
309 
310 int
311 xfs_set_dmattrs(
312 	xfs_inode_t     *ip,
313 	u_int		evmask,
314 	u_int16_t	state)
315 {
316 	xfs_mount_t	*mp = ip->i_mount;
317 	xfs_trans_t	*tp;
318 	int		error;
319 
320 	if (!capable(CAP_SYS_ADMIN))
321 		return -EPERM;
322 
323 	if (XFS_FORCED_SHUTDOWN(mp))
324 		return -EIO;
325 
326 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
327 	if (error)
328 		return error;
329 
330 	xfs_ilock(ip, XFS_ILOCK_EXCL);
331 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
332 
333 	ip->i_d.di_dmevmask = evmask;
334 	ip->i_d.di_dmstate  = state;
335 
336 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
337 	error = xfs_trans_commit(tp);
338 
339 	return error;
340 }
341 
342 STATIC int
343 xfs_fssetdm_by_handle(
344 	struct file		*parfilp,
345 	void			__user *arg)
346 {
347 	int			error;
348 	struct fsdmidata	fsd;
349 	xfs_fsop_setdm_handlereq_t dmhreq;
350 	struct dentry		*dentry;
351 
352 	if (!capable(CAP_MKNOD))
353 		return -EPERM;
354 	if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
355 		return -EFAULT;
356 
357 	error = mnt_want_write_file(parfilp);
358 	if (error)
359 		return error;
360 
361 	dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
362 	if (IS_ERR(dentry)) {
363 		mnt_drop_write_file(parfilp);
364 		return PTR_ERR(dentry);
365 	}
366 
367 	if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
368 		error = -EPERM;
369 		goto out;
370 	}
371 
372 	if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
373 		error = -EFAULT;
374 		goto out;
375 	}
376 
377 	error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
378 				 fsd.fsd_dmstate);
379 
380  out:
381 	mnt_drop_write_file(parfilp);
382 	dput(dentry);
383 	return error;
384 }
385 
386 STATIC int
387 xfs_attrlist_by_handle(
388 	struct file		*parfilp,
389 	void			__user *arg)
390 {
391 	int			error = -ENOMEM;
392 	attrlist_cursor_kern_t	*cursor;
393 	struct xfs_fsop_attrlist_handlereq __user	*p = arg;
394 	xfs_fsop_attrlist_handlereq_t al_hreq;
395 	struct dentry		*dentry;
396 	char			*kbuf;
397 
398 	if (!capable(CAP_SYS_ADMIN))
399 		return -EPERM;
400 	if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
401 		return -EFAULT;
402 	if (al_hreq.buflen < sizeof(struct attrlist) ||
403 	    al_hreq.buflen > XFS_XATTR_LIST_MAX)
404 		return -EINVAL;
405 
406 	/*
407 	 * Reject flags, only allow namespaces.
408 	 */
409 	if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
410 		return -EINVAL;
411 
412 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
413 	if (IS_ERR(dentry))
414 		return PTR_ERR(dentry);
415 
416 	kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
417 	if (!kbuf)
418 		goto out_dput;
419 
420 	cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
421 	error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
422 					al_hreq.flags, cursor);
423 	if (error)
424 		goto out_kfree;
425 
426 	if (copy_to_user(&p->pos, cursor, sizeof(attrlist_cursor_kern_t))) {
427 		error = -EFAULT;
428 		goto out_kfree;
429 	}
430 
431 	if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
432 		error = -EFAULT;
433 
434 out_kfree:
435 	kmem_free(kbuf);
436 out_dput:
437 	dput(dentry);
438 	return error;
439 }
440 
441 int
442 xfs_attrmulti_attr_get(
443 	struct inode		*inode,
444 	unsigned char		*name,
445 	unsigned char		__user *ubuf,
446 	uint32_t		*len,
447 	uint32_t		flags)
448 {
449 	unsigned char		*kbuf;
450 	int			error = -EFAULT;
451 
452 	if (*len > XFS_XATTR_SIZE_MAX)
453 		return -EINVAL;
454 	kbuf = kmem_zalloc_large(*len, KM_SLEEP);
455 	if (!kbuf)
456 		return -ENOMEM;
457 
458 	error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
459 	if (error)
460 		goto out_kfree;
461 
462 	if (copy_to_user(ubuf, kbuf, *len))
463 		error = -EFAULT;
464 
465 out_kfree:
466 	kmem_free(kbuf);
467 	return error;
468 }
469 
470 int
471 xfs_attrmulti_attr_set(
472 	struct inode		*inode,
473 	unsigned char		*name,
474 	const unsigned char	__user *ubuf,
475 	uint32_t		len,
476 	uint32_t		flags)
477 {
478 	unsigned char		*kbuf;
479 	int			error;
480 
481 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
482 		return -EPERM;
483 	if (len > XFS_XATTR_SIZE_MAX)
484 		return -EINVAL;
485 
486 	kbuf = memdup_user(ubuf, len);
487 	if (IS_ERR(kbuf))
488 		return PTR_ERR(kbuf);
489 
490 	error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
491 	if (!error)
492 		xfs_forget_acl(inode, name, flags);
493 	kfree(kbuf);
494 	return error;
495 }
496 
497 int
498 xfs_attrmulti_attr_remove(
499 	struct inode		*inode,
500 	unsigned char		*name,
501 	uint32_t		flags)
502 {
503 	int			error;
504 
505 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
506 		return -EPERM;
507 	error = xfs_attr_remove(XFS_I(inode), name, flags);
508 	if (!error)
509 		xfs_forget_acl(inode, name, flags);
510 	return error;
511 }
512 
513 STATIC int
514 xfs_attrmulti_by_handle(
515 	struct file		*parfilp,
516 	void			__user *arg)
517 {
518 	int			error;
519 	xfs_attr_multiop_t	*ops;
520 	xfs_fsop_attrmulti_handlereq_t am_hreq;
521 	struct dentry		*dentry;
522 	unsigned int		i, size;
523 	unsigned char		*attr_name;
524 
525 	if (!capable(CAP_SYS_ADMIN))
526 		return -EPERM;
527 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
528 		return -EFAULT;
529 
530 	/* overflow check */
531 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
532 		return -E2BIG;
533 
534 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
535 	if (IS_ERR(dentry))
536 		return PTR_ERR(dentry);
537 
538 	error = -E2BIG;
539 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
540 	if (!size || size > 16 * PAGE_SIZE)
541 		goto out_dput;
542 
543 	ops = memdup_user(am_hreq.ops, size);
544 	if (IS_ERR(ops)) {
545 		error = PTR_ERR(ops);
546 		goto out_dput;
547 	}
548 
549 	error = -ENOMEM;
550 	attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
551 	if (!attr_name)
552 		goto out_kfree_ops;
553 
554 	error = 0;
555 	for (i = 0; i < am_hreq.opcount; i++) {
556 		ops[i].am_error = strncpy_from_user((char *)attr_name,
557 				ops[i].am_attrname, MAXNAMELEN);
558 		if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
559 			error = -ERANGE;
560 		if (ops[i].am_error < 0)
561 			break;
562 
563 		switch (ops[i].am_opcode) {
564 		case ATTR_OP_GET:
565 			ops[i].am_error = xfs_attrmulti_attr_get(
566 					d_inode(dentry), attr_name,
567 					ops[i].am_attrvalue, &ops[i].am_length,
568 					ops[i].am_flags);
569 			break;
570 		case ATTR_OP_SET:
571 			ops[i].am_error = mnt_want_write_file(parfilp);
572 			if (ops[i].am_error)
573 				break;
574 			ops[i].am_error = xfs_attrmulti_attr_set(
575 					d_inode(dentry), attr_name,
576 					ops[i].am_attrvalue, ops[i].am_length,
577 					ops[i].am_flags);
578 			mnt_drop_write_file(parfilp);
579 			break;
580 		case ATTR_OP_REMOVE:
581 			ops[i].am_error = mnt_want_write_file(parfilp);
582 			if (ops[i].am_error)
583 				break;
584 			ops[i].am_error = xfs_attrmulti_attr_remove(
585 					d_inode(dentry), attr_name,
586 					ops[i].am_flags);
587 			mnt_drop_write_file(parfilp);
588 			break;
589 		default:
590 			ops[i].am_error = -EINVAL;
591 		}
592 	}
593 
594 	if (copy_to_user(am_hreq.ops, ops, size))
595 		error = -EFAULT;
596 
597 	kfree(attr_name);
598  out_kfree_ops:
599 	kfree(ops);
600  out_dput:
601 	dput(dentry);
602 	return error;
603 }
604 
605 int
606 xfs_ioc_space(
607 	struct file		*filp,
608 	unsigned int		cmd,
609 	xfs_flock64_t		*bf)
610 {
611 	struct inode		*inode = file_inode(filp);
612 	struct xfs_inode	*ip = XFS_I(inode);
613 	struct iattr		iattr;
614 	enum xfs_prealloc_flags	flags = 0;
615 	uint			iolock = XFS_IOLOCK_EXCL;
616 	int			error;
617 
618 	/*
619 	 * Only allow the sys admin to reserve space unless
620 	 * unwritten extents are enabled.
621 	 */
622 	if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
623 	    !capable(CAP_SYS_ADMIN))
624 		return -EPERM;
625 
626 	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
627 		return -EPERM;
628 
629 	if (!(filp->f_mode & FMODE_WRITE))
630 		return -EBADF;
631 
632 	if (!S_ISREG(inode->i_mode))
633 		return -EINVAL;
634 
635 	if (filp->f_flags & O_DSYNC)
636 		flags |= XFS_PREALLOC_SYNC;
637 	if (filp->f_mode & FMODE_NOCMTIME)
638 		flags |= XFS_PREALLOC_INVISIBLE;
639 
640 	error = mnt_want_write_file(filp);
641 	if (error)
642 		return error;
643 
644 	xfs_ilock(ip, iolock);
645 	error = xfs_break_layouts(inode, &iolock);
646 	if (error)
647 		goto out_unlock;
648 
649 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
650 	iolock |= XFS_MMAPLOCK_EXCL;
651 
652 	switch (bf->l_whence) {
653 	case 0: /*SEEK_SET*/
654 		break;
655 	case 1: /*SEEK_CUR*/
656 		bf->l_start += filp->f_pos;
657 		break;
658 	case 2: /*SEEK_END*/
659 		bf->l_start += XFS_ISIZE(ip);
660 		break;
661 	default:
662 		error = -EINVAL;
663 		goto out_unlock;
664 	}
665 
666 	/*
667 	 * length of <= 0 for resv/unresv/zero is invalid.  length for
668 	 * alloc/free is ignored completely and we have no idea what userspace
669 	 * might have set it to, so set it to zero to allow range
670 	 * checks to pass.
671 	 */
672 	switch (cmd) {
673 	case XFS_IOC_ZERO_RANGE:
674 	case XFS_IOC_RESVSP:
675 	case XFS_IOC_RESVSP64:
676 	case XFS_IOC_UNRESVSP:
677 	case XFS_IOC_UNRESVSP64:
678 		if (bf->l_len <= 0) {
679 			error = -EINVAL;
680 			goto out_unlock;
681 		}
682 		break;
683 	default:
684 		bf->l_len = 0;
685 		break;
686 	}
687 
688 	if (bf->l_start < 0 ||
689 	    bf->l_start > inode->i_sb->s_maxbytes ||
690 	    bf->l_start + bf->l_len < 0 ||
691 	    bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
692 		error = -EINVAL;
693 		goto out_unlock;
694 	}
695 
696 	switch (cmd) {
697 	case XFS_IOC_ZERO_RANGE:
698 		flags |= XFS_PREALLOC_SET;
699 		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
700 		break;
701 	case XFS_IOC_RESVSP:
702 	case XFS_IOC_RESVSP64:
703 		flags |= XFS_PREALLOC_SET;
704 		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
705 						XFS_BMAPI_PREALLOC);
706 		break;
707 	case XFS_IOC_UNRESVSP:
708 	case XFS_IOC_UNRESVSP64:
709 		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
710 		break;
711 	case XFS_IOC_ALLOCSP:
712 	case XFS_IOC_ALLOCSP64:
713 	case XFS_IOC_FREESP:
714 	case XFS_IOC_FREESP64:
715 		flags |= XFS_PREALLOC_CLEAR;
716 		if (bf->l_start > XFS_ISIZE(ip)) {
717 			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
718 					bf->l_start - XFS_ISIZE(ip), 0);
719 			if (error)
720 				goto out_unlock;
721 		}
722 
723 		iattr.ia_valid = ATTR_SIZE;
724 		iattr.ia_size = bf->l_start;
725 
726 		error = xfs_vn_setattr_size(file_dentry(filp), &iattr);
727 		break;
728 	default:
729 		ASSERT(0);
730 		error = -EINVAL;
731 	}
732 
733 	if (error)
734 		goto out_unlock;
735 
736 	error = xfs_update_prealloc_flags(ip, flags);
737 
738 out_unlock:
739 	xfs_iunlock(ip, iolock);
740 	mnt_drop_write_file(filp);
741 	return error;
742 }
743 
744 STATIC int
745 xfs_ioc_bulkstat(
746 	xfs_mount_t		*mp,
747 	unsigned int		cmd,
748 	void			__user *arg)
749 {
750 	xfs_fsop_bulkreq_t	bulkreq;
751 	int			count;	/* # of records returned */
752 	xfs_ino_t		inlast;	/* last inode number */
753 	int			done;
754 	int			error;
755 
756 	/* done = 1 if there are more stats to get and if bulkstat */
757 	/* should be called again (unused here, but used in dmapi) */
758 
759 	if (!capable(CAP_SYS_ADMIN))
760 		return -EPERM;
761 
762 	if (XFS_FORCED_SHUTDOWN(mp))
763 		return -EIO;
764 
765 	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
766 		return -EFAULT;
767 
768 	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
769 		return -EFAULT;
770 
771 	if ((count = bulkreq.icount) <= 0)
772 		return -EINVAL;
773 
774 	if (bulkreq.ubuffer == NULL)
775 		return -EINVAL;
776 
777 	if (cmd == XFS_IOC_FSINUMBERS)
778 		error = xfs_inumbers(mp, &inlast, &count,
779 					bulkreq.ubuffer, xfs_inumbers_fmt);
780 	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
781 		error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
782 					sizeof(xfs_bstat_t), NULL, &done);
783 	else	/* XFS_IOC_FSBULKSTAT */
784 		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
785 				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
786 				     &done);
787 
788 	if (error)
789 		return error;
790 
791 	if (bulkreq.ocount != NULL) {
792 		if (copy_to_user(bulkreq.lastip, &inlast,
793 						sizeof(xfs_ino_t)))
794 			return -EFAULT;
795 
796 		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
797 			return -EFAULT;
798 	}
799 
800 	return 0;
801 }
802 
803 STATIC int
804 xfs_ioc_fsgeometry_v1(
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, &fsgeo, 3);
812 	if (error)
813 		return error;
814 
815 	/*
816 	 * Caller should have passed an argument of type
817 	 * xfs_fsop_geom_v1_t.  This is a proper subset of the
818 	 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
819 	 */
820 	if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
821 		return -EFAULT;
822 	return 0;
823 }
824 
825 STATIC int
826 xfs_ioc_fsgeometry(
827 	xfs_mount_t		*mp,
828 	void			__user *arg)
829 {
830 	xfs_fsop_geom_t		fsgeo;
831 	int			error;
832 
833 	error = xfs_fs_geometry(mp, &fsgeo, 4);
834 	if (error)
835 		return error;
836 
837 	if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
838 		return -EFAULT;
839 	return 0;
840 }
841 
842 /*
843  * Linux extended inode flags interface.
844  */
845 
846 STATIC unsigned int
847 xfs_merge_ioc_xflags(
848 	unsigned int	flags,
849 	unsigned int	start)
850 {
851 	unsigned int	xflags = start;
852 
853 	if (flags & FS_IMMUTABLE_FL)
854 		xflags |= FS_XFLAG_IMMUTABLE;
855 	else
856 		xflags &= ~FS_XFLAG_IMMUTABLE;
857 	if (flags & FS_APPEND_FL)
858 		xflags |= FS_XFLAG_APPEND;
859 	else
860 		xflags &= ~FS_XFLAG_APPEND;
861 	if (flags & FS_SYNC_FL)
862 		xflags |= FS_XFLAG_SYNC;
863 	else
864 		xflags &= ~FS_XFLAG_SYNC;
865 	if (flags & FS_NOATIME_FL)
866 		xflags |= FS_XFLAG_NOATIME;
867 	else
868 		xflags &= ~FS_XFLAG_NOATIME;
869 	if (flags & FS_NODUMP_FL)
870 		xflags |= FS_XFLAG_NODUMP;
871 	else
872 		xflags &= ~FS_XFLAG_NODUMP;
873 
874 	return xflags;
875 }
876 
877 STATIC unsigned int
878 xfs_di2lxflags(
879 	uint16_t	di_flags)
880 {
881 	unsigned int	flags = 0;
882 
883 	if (di_flags & XFS_DIFLAG_IMMUTABLE)
884 		flags |= FS_IMMUTABLE_FL;
885 	if (di_flags & XFS_DIFLAG_APPEND)
886 		flags |= FS_APPEND_FL;
887 	if (di_flags & XFS_DIFLAG_SYNC)
888 		flags |= FS_SYNC_FL;
889 	if (di_flags & XFS_DIFLAG_NOATIME)
890 		flags |= FS_NOATIME_FL;
891 	if (di_flags & XFS_DIFLAG_NODUMP)
892 		flags |= FS_NODUMP_FL;
893 	return flags;
894 }
895 
896 STATIC int
897 xfs_ioc_fsgetxattr(
898 	xfs_inode_t		*ip,
899 	int			attr,
900 	void			__user *arg)
901 {
902 	struct fsxattr		fa;
903 
904 	memset(&fa, 0, sizeof(struct fsxattr));
905 
906 	xfs_ilock(ip, XFS_ILOCK_SHARED);
907 	fa.fsx_xflags = xfs_ip2xflags(ip);
908 	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
909 	fa.fsx_cowextsize = ip->i_d.di_cowextsize <<
910 			ip->i_mount->m_sb.sb_blocklog;
911 	fa.fsx_projid = xfs_get_projid(ip);
912 
913 	if (attr) {
914 		if (ip->i_afp) {
915 			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
916 				fa.fsx_nextents = xfs_iext_count(ip->i_afp);
917 			else
918 				fa.fsx_nextents = ip->i_d.di_anextents;
919 		} else
920 			fa.fsx_nextents = 0;
921 	} else {
922 		if (ip->i_df.if_flags & XFS_IFEXTENTS)
923 			fa.fsx_nextents = xfs_iext_count(&ip->i_df);
924 		else
925 			fa.fsx_nextents = ip->i_d.di_nextents;
926 	}
927 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
928 
929 	if (copy_to_user(arg, &fa, sizeof(fa)))
930 		return -EFAULT;
931 	return 0;
932 }
933 
934 STATIC uint16_t
935 xfs_flags2diflags(
936 	struct xfs_inode	*ip,
937 	unsigned int		xflags)
938 {
939 	/* can't set PREALLOC this way, just preserve it */
940 	uint16_t		di_flags =
941 		(ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
942 
943 	if (xflags & FS_XFLAG_IMMUTABLE)
944 		di_flags |= XFS_DIFLAG_IMMUTABLE;
945 	if (xflags & FS_XFLAG_APPEND)
946 		di_flags |= XFS_DIFLAG_APPEND;
947 	if (xflags & FS_XFLAG_SYNC)
948 		di_flags |= XFS_DIFLAG_SYNC;
949 	if (xflags & FS_XFLAG_NOATIME)
950 		di_flags |= XFS_DIFLAG_NOATIME;
951 	if (xflags & FS_XFLAG_NODUMP)
952 		di_flags |= XFS_DIFLAG_NODUMP;
953 	if (xflags & FS_XFLAG_NODEFRAG)
954 		di_flags |= XFS_DIFLAG_NODEFRAG;
955 	if (xflags & FS_XFLAG_FILESTREAM)
956 		di_flags |= XFS_DIFLAG_FILESTREAM;
957 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
958 		if (xflags & FS_XFLAG_RTINHERIT)
959 			di_flags |= XFS_DIFLAG_RTINHERIT;
960 		if (xflags & FS_XFLAG_NOSYMLINKS)
961 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
962 		if (xflags & FS_XFLAG_EXTSZINHERIT)
963 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
964 		if (xflags & FS_XFLAG_PROJINHERIT)
965 			di_flags |= XFS_DIFLAG_PROJINHERIT;
966 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
967 		if (xflags & FS_XFLAG_REALTIME)
968 			di_flags |= XFS_DIFLAG_REALTIME;
969 		if (xflags & FS_XFLAG_EXTSIZE)
970 			di_flags |= XFS_DIFLAG_EXTSIZE;
971 	}
972 
973 	return di_flags;
974 }
975 
976 STATIC uint64_t
977 xfs_flags2diflags2(
978 	struct xfs_inode	*ip,
979 	unsigned int		xflags)
980 {
981 	uint64_t		di_flags2 =
982 		(ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK);
983 
984 	if (xflags & FS_XFLAG_DAX)
985 		di_flags2 |= XFS_DIFLAG2_DAX;
986 	if (xflags & FS_XFLAG_COWEXTSIZE)
987 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
988 
989 	return di_flags2;
990 }
991 
992 STATIC void
993 xfs_diflags_to_linux(
994 	struct xfs_inode	*ip)
995 {
996 	struct inode		*inode = VFS_I(ip);
997 	unsigned int		xflags = xfs_ip2xflags(ip);
998 
999 	if (xflags & FS_XFLAG_IMMUTABLE)
1000 		inode->i_flags |= S_IMMUTABLE;
1001 	else
1002 		inode->i_flags &= ~S_IMMUTABLE;
1003 	if (xflags & FS_XFLAG_APPEND)
1004 		inode->i_flags |= S_APPEND;
1005 	else
1006 		inode->i_flags &= ~S_APPEND;
1007 	if (xflags & FS_XFLAG_SYNC)
1008 		inode->i_flags |= S_SYNC;
1009 	else
1010 		inode->i_flags &= ~S_SYNC;
1011 	if (xflags & FS_XFLAG_NOATIME)
1012 		inode->i_flags |= S_NOATIME;
1013 	else
1014 		inode->i_flags &= ~S_NOATIME;
1015 #if 0	/* disabled until the flag switching races are sorted out */
1016 	if (xflags & FS_XFLAG_DAX)
1017 		inode->i_flags |= S_DAX;
1018 	else
1019 		inode->i_flags &= ~S_DAX;
1020 #endif
1021 }
1022 
1023 static int
1024 xfs_ioctl_setattr_xflags(
1025 	struct xfs_trans	*tp,
1026 	struct xfs_inode	*ip,
1027 	struct fsxattr		*fa)
1028 {
1029 	struct xfs_mount	*mp = ip->i_mount;
1030 	uint64_t		di_flags2;
1031 
1032 	/* Can't change realtime flag if any extents are allocated. */
1033 	if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1034 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1035 		return -EINVAL;
1036 
1037 	/* If realtime flag is set then must have realtime device */
1038 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1039 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1040 		    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1041 			return -EINVAL;
1042 	}
1043 
1044 	/* Clear reflink if we are actually able to set the rt flag. */
1045 	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1046 		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1047 
1048 	/* Don't allow us to set DAX mode for a reflinked file for now. */
1049 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1050 		return -EINVAL;
1051 
1052 	/*
1053 	 * Can't modify an immutable/append-only file unless
1054 	 * we have appropriate permission.
1055 	 */
1056 	if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1057 	     (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1058 	    !capable(CAP_LINUX_IMMUTABLE))
1059 		return -EPERM;
1060 
1061 	/* diflags2 only valid for v3 inodes. */
1062 	di_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1063 	if (di_flags2 && ip->i_d.di_version < 3)
1064 		return -EINVAL;
1065 
1066 	ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
1067 	ip->i_d.di_flags2 = di_flags2;
1068 
1069 	xfs_diflags_to_linux(ip);
1070 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1071 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1072 	XFS_STATS_INC(mp, xs_ig_attrchg);
1073 	return 0;
1074 }
1075 
1076 /*
1077  * If we are changing DAX flags, we have to ensure the file is clean and any
1078  * cached objects in the address space are invalidated and removed. This
1079  * requires us to lock out other IO and page faults similar to a truncate
1080  * operation. The locks need to be held until the transaction has been committed
1081  * so that the cache invalidation is atomic with respect to the DAX flag
1082  * manipulation.
1083  */
1084 static int
1085 xfs_ioctl_setattr_dax_invalidate(
1086 	struct xfs_inode	*ip,
1087 	struct fsxattr		*fa,
1088 	int			*join_flags)
1089 {
1090 	struct inode		*inode = VFS_I(ip);
1091 	struct super_block	*sb = inode->i_sb;
1092 	int			error;
1093 
1094 	*join_flags = 0;
1095 
1096 	/*
1097 	 * It is only valid to set the DAX flag on regular files and
1098 	 * directories on filesystems where the block size is equal to the page
1099 	 * size. On directories it serves as an inherit hint.
1100 	 */
1101 	if (fa->fsx_xflags & FS_XFLAG_DAX) {
1102 		if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1103 			return -EINVAL;
1104 		if (bdev_dax_supported(sb, sb->s_blocksize) < 0)
1105 			return -EINVAL;
1106 	}
1107 
1108 	/* If the DAX state is not changing, we have nothing to do here. */
1109 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1110 		return 0;
1111 	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1112 		return 0;
1113 
1114 	/* lock, flush and invalidate mapping in preparation for flag change */
1115 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1116 	error = filemap_write_and_wait(inode->i_mapping);
1117 	if (error)
1118 		goto out_unlock;
1119 	error = invalidate_inode_pages2(inode->i_mapping);
1120 	if (error)
1121 		goto out_unlock;
1122 
1123 	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1124 	return 0;
1125 
1126 out_unlock:
1127 	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1128 	return error;
1129 
1130 }
1131 
1132 /*
1133  * Set up the transaction structure for the setattr operation, checking that we
1134  * have permission to do so. On success, return a clean transaction and the
1135  * inode locked exclusively ready for further operation specific checks. On
1136  * failure, return an error without modifying or locking the inode.
1137  *
1138  * The inode might already be IO locked on call. If this is the case, it is
1139  * indicated in @join_flags and we take full responsibility for ensuring they
1140  * are unlocked from now on. Hence if we have an error here, we still have to
1141  * unlock them. Otherwise, once they are joined to the transaction, they will
1142  * be unlocked on commit/cancel.
1143  */
1144 static struct xfs_trans *
1145 xfs_ioctl_setattr_get_trans(
1146 	struct xfs_inode	*ip,
1147 	int			join_flags)
1148 {
1149 	struct xfs_mount	*mp = ip->i_mount;
1150 	struct xfs_trans	*tp;
1151 	int			error = -EROFS;
1152 
1153 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1154 		goto out_unlock;
1155 	error = -EIO;
1156 	if (XFS_FORCED_SHUTDOWN(mp))
1157 		goto out_unlock;
1158 
1159 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1160 	if (error)
1161 		return ERR_PTR(error);
1162 
1163 	xfs_ilock(ip, XFS_ILOCK_EXCL);
1164 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1165 	join_flags = 0;
1166 
1167 	/*
1168 	 * CAP_FOWNER overrides the following restrictions:
1169 	 *
1170 	 * The user ID of the calling process must be equal to the file owner
1171 	 * ID, except in cases where the CAP_FSETID capability is applicable.
1172 	 */
1173 	if (!inode_owner_or_capable(VFS_I(ip))) {
1174 		error = -EPERM;
1175 		goto out_cancel;
1176 	}
1177 
1178 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1179 		xfs_trans_set_sync(tp);
1180 
1181 	return tp;
1182 
1183 out_cancel:
1184 	xfs_trans_cancel(tp);
1185 out_unlock:
1186 	if (join_flags)
1187 		xfs_iunlock(ip, join_flags);
1188 	return ERR_PTR(error);
1189 }
1190 
1191 /*
1192  * extent size hint validation is somewhat cumbersome. Rules are:
1193  *
1194  * 1. extent size hint is only valid for directories and regular files
1195  * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1196  * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1197  * 4. can only be changed on regular files if no extents are allocated
1198  * 5. can be changed on directories at any time
1199  * 6. extsize hint of 0 turns off hints, clears inode flags.
1200  * 7. Extent size must be a multiple of the appropriate block size.
1201  * 8. for non-realtime files, the extent size hint must be limited
1202  *    to half the AG size to avoid alignment extending the extent beyond the
1203  *    limits of the AG.
1204  */
1205 static int
1206 xfs_ioctl_setattr_check_extsize(
1207 	struct xfs_inode	*ip,
1208 	struct fsxattr		*fa)
1209 {
1210 	struct xfs_mount	*mp = ip->i_mount;
1211 
1212 	if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1213 		return -EINVAL;
1214 
1215 	if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1216 	    !S_ISDIR(VFS_I(ip)->i_mode))
1217 		return -EINVAL;
1218 
1219 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1220 	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1221 		return -EINVAL;
1222 
1223 	if (fa->fsx_extsize != 0) {
1224 		xfs_extlen_t    size;
1225 		xfs_fsblock_t   extsize_fsb;
1226 
1227 		extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1228 		if (extsize_fsb > MAXEXTLEN)
1229 			return -EINVAL;
1230 
1231 		if (XFS_IS_REALTIME_INODE(ip) ||
1232 		    (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1233 			size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1234 		} else {
1235 			size = mp->m_sb.sb_blocksize;
1236 			if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1237 				return -EINVAL;
1238 		}
1239 
1240 		if (fa->fsx_extsize % size)
1241 			return -EINVAL;
1242 	} else
1243 		fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1244 
1245 	return 0;
1246 }
1247 
1248 /*
1249  * CoW extent size hint validation rules are:
1250  *
1251  * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1252  *    The inode does not have to have any shared blocks, but it must be a v3.
1253  * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1254  *    for a directory, the hint is propagated to new files.
1255  * 3. Can be changed on files & directories at any time.
1256  * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1257  * 5. Extent size must be a multiple of the appropriate block size.
1258  * 6. The extent size hint must be limited to half the AG size to avoid
1259  *    alignment extending the extent beyond the limits of the AG.
1260  */
1261 static int
1262 xfs_ioctl_setattr_check_cowextsize(
1263 	struct xfs_inode	*ip,
1264 	struct fsxattr		*fa)
1265 {
1266 	struct xfs_mount	*mp = ip->i_mount;
1267 
1268 	if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1269 		return 0;
1270 
1271 	if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb) ||
1272 	    ip->i_d.di_version != 3)
1273 		return -EINVAL;
1274 
1275 	if (!S_ISREG(VFS_I(ip)->i_mode) && !S_ISDIR(VFS_I(ip)->i_mode))
1276 		return -EINVAL;
1277 
1278 	if (fa->fsx_cowextsize != 0) {
1279 		xfs_extlen_t    size;
1280 		xfs_fsblock_t   cowextsize_fsb;
1281 
1282 		cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1283 		if (cowextsize_fsb > MAXEXTLEN)
1284 			return -EINVAL;
1285 
1286 		size = mp->m_sb.sb_blocksize;
1287 		if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1288 			return -EINVAL;
1289 
1290 		if (fa->fsx_cowextsize % size)
1291 			return -EINVAL;
1292 	} else
1293 		fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
1294 
1295 	return 0;
1296 }
1297 
1298 static int
1299 xfs_ioctl_setattr_check_projid(
1300 	struct xfs_inode	*ip,
1301 	struct fsxattr		*fa)
1302 {
1303 	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1304 	if (fa->fsx_projid > (uint16_t)-1 &&
1305 	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1306 		return -EINVAL;
1307 
1308 	/*
1309 	 * Project Quota ID state is only allowed to change from within the init
1310 	 * namespace. Enforce that restriction only if we are trying to change
1311 	 * the quota ID state. Everything else is allowed in user namespaces.
1312 	 */
1313 	if (current_user_ns() == &init_user_ns)
1314 		return 0;
1315 
1316 	if (xfs_get_projid(ip) != fa->fsx_projid)
1317 		return -EINVAL;
1318 	if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1319 	    (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1320 		return -EINVAL;
1321 
1322 	return 0;
1323 }
1324 
1325 STATIC int
1326 xfs_ioctl_setattr(
1327 	xfs_inode_t		*ip,
1328 	struct fsxattr		*fa)
1329 {
1330 	struct xfs_mount	*mp = ip->i_mount;
1331 	struct xfs_trans	*tp;
1332 	struct xfs_dquot	*udqp = NULL;
1333 	struct xfs_dquot	*pdqp = NULL;
1334 	struct xfs_dquot	*olddquot = NULL;
1335 	int			code;
1336 	int			join_flags = 0;
1337 
1338 	trace_xfs_ioctl_setattr(ip);
1339 
1340 	code = xfs_ioctl_setattr_check_projid(ip, fa);
1341 	if (code)
1342 		return code;
1343 
1344 	/*
1345 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1346 	 * before we start any other transactions. Trying to do this later
1347 	 * is messy. We don't care to take a readlock to look at the ids
1348 	 * in inode here, because we can't hold it across the trans_reserve.
1349 	 * If the IDs do change before we take the ilock, we're covered
1350 	 * because the i_*dquot fields will get updated anyway.
1351 	 */
1352 	if (XFS_IS_QUOTA_ON(mp)) {
1353 		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1354 					 ip->i_d.di_gid, fa->fsx_projid,
1355 					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1356 		if (code)
1357 			return code;
1358 	}
1359 
1360 	/*
1361 	 * Changing DAX config may require inode locking for mapping
1362 	 * invalidation. These need to be held all the way to transaction commit
1363 	 * or cancel time, so need to be passed through to
1364 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1365 	 * appropriately.
1366 	 */
1367 	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1368 	if (code)
1369 		goto error_free_dquots;
1370 
1371 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1372 	if (IS_ERR(tp)) {
1373 		code = PTR_ERR(tp);
1374 		goto error_free_dquots;
1375 	}
1376 
1377 
1378 	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1379 	    xfs_get_projid(ip) != fa->fsx_projid) {
1380 		code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1381 				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1382 		if (code)	/* out of quota */
1383 			goto error_trans_cancel;
1384 	}
1385 
1386 	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1387 	if (code)
1388 		goto error_trans_cancel;
1389 
1390 	code = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1391 	if (code)
1392 		goto error_trans_cancel;
1393 
1394 	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1395 	if (code)
1396 		goto error_trans_cancel;
1397 
1398 	/*
1399 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1400 	 * overrides the following restrictions:
1401 	 *
1402 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1403 	 * successful return from chown()
1404 	 */
1405 
1406 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1407 	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1408 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1409 
1410 	/* Change the ownerships and register project quota modifications */
1411 	if (xfs_get_projid(ip) != fa->fsx_projid) {
1412 		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1413 			olddquot = xfs_qm_vop_chown(tp, ip,
1414 						&ip->i_pdquot, pdqp);
1415 		}
1416 		ASSERT(ip->i_d.di_version > 1);
1417 		xfs_set_projid(ip, fa->fsx_projid);
1418 	}
1419 
1420 	/*
1421 	 * Only set the extent size hint if we've already determined that the
1422 	 * extent size hint should be set on the inode. If no extent size flags
1423 	 * are set on the inode then unconditionally clear the extent size hint.
1424 	 */
1425 	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1426 		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1427 	else
1428 		ip->i_d.di_extsize = 0;
1429 	if (ip->i_d.di_version == 3 &&
1430 	    (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1431 		ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1432 				mp->m_sb.sb_blocklog;
1433 	else
1434 		ip->i_d.di_cowextsize = 0;
1435 
1436 	code = xfs_trans_commit(tp);
1437 
1438 	/*
1439 	 * Release any dquot(s) the inode had kept before chown.
1440 	 */
1441 	xfs_qm_dqrele(olddquot);
1442 	xfs_qm_dqrele(udqp);
1443 	xfs_qm_dqrele(pdqp);
1444 
1445 	return code;
1446 
1447 error_trans_cancel:
1448 	xfs_trans_cancel(tp);
1449 error_free_dquots:
1450 	xfs_qm_dqrele(udqp);
1451 	xfs_qm_dqrele(pdqp);
1452 	return code;
1453 }
1454 
1455 STATIC int
1456 xfs_ioc_fssetxattr(
1457 	xfs_inode_t		*ip,
1458 	struct file		*filp,
1459 	void			__user *arg)
1460 {
1461 	struct fsxattr		fa;
1462 	int error;
1463 
1464 	if (copy_from_user(&fa, arg, sizeof(fa)))
1465 		return -EFAULT;
1466 
1467 	error = mnt_want_write_file(filp);
1468 	if (error)
1469 		return error;
1470 	error = xfs_ioctl_setattr(ip, &fa);
1471 	mnt_drop_write_file(filp);
1472 	return error;
1473 }
1474 
1475 STATIC int
1476 xfs_ioc_getxflags(
1477 	xfs_inode_t		*ip,
1478 	void			__user *arg)
1479 {
1480 	unsigned int		flags;
1481 
1482 	flags = xfs_di2lxflags(ip->i_d.di_flags);
1483 	if (copy_to_user(arg, &flags, sizeof(flags)))
1484 		return -EFAULT;
1485 	return 0;
1486 }
1487 
1488 STATIC int
1489 xfs_ioc_setxflags(
1490 	struct xfs_inode	*ip,
1491 	struct file		*filp,
1492 	void			__user *arg)
1493 {
1494 	struct xfs_trans	*tp;
1495 	struct fsxattr		fa;
1496 	unsigned int		flags;
1497 	int			join_flags = 0;
1498 	int			error;
1499 
1500 	if (copy_from_user(&flags, arg, sizeof(flags)))
1501 		return -EFAULT;
1502 
1503 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1504 		      FS_NOATIME_FL | FS_NODUMP_FL | \
1505 		      FS_SYNC_FL))
1506 		return -EOPNOTSUPP;
1507 
1508 	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1509 
1510 	error = mnt_want_write_file(filp);
1511 	if (error)
1512 		return error;
1513 
1514 	/*
1515 	 * Changing DAX config may require inode locking for mapping
1516 	 * invalidation. These need to be held all the way to transaction commit
1517 	 * or cancel time, so need to be passed through to
1518 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1519 	 * appropriately.
1520 	 */
1521 	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1522 	if (error)
1523 		goto out_drop_write;
1524 
1525 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1526 	if (IS_ERR(tp)) {
1527 		error = PTR_ERR(tp);
1528 		goto out_drop_write;
1529 	}
1530 
1531 	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1532 	if (error) {
1533 		xfs_trans_cancel(tp);
1534 		goto out_drop_write;
1535 	}
1536 
1537 	error = xfs_trans_commit(tp);
1538 out_drop_write:
1539 	mnt_drop_write_file(filp);
1540 	return error;
1541 }
1542 
1543 STATIC int
1544 xfs_getbmap_format(void **ap, struct getbmapx *bmv)
1545 {
1546 	struct getbmap __user	*base = (struct getbmap __user *)*ap;
1547 
1548 	/* copy only getbmap portion (not getbmapx) */
1549 	if (copy_to_user(base, bmv, sizeof(struct getbmap)))
1550 		return -EFAULT;
1551 
1552 	*ap += sizeof(struct getbmap);
1553 	return 0;
1554 }
1555 
1556 STATIC int
1557 xfs_ioc_getbmap(
1558 	struct file		*file,
1559 	unsigned int		cmd,
1560 	void			__user *arg)
1561 {
1562 	struct getbmapx		bmx = { 0 };
1563 	int			error;
1564 
1565 	/* struct getbmap is a strict subset of struct getbmapx. */
1566 	if (copy_from_user(&bmx, arg, offsetof(struct getbmapx, bmv_iflags)))
1567 		return -EFAULT;
1568 
1569 	if (bmx.bmv_count < 2)
1570 		return -EINVAL;
1571 
1572 	bmx.bmv_iflags = (cmd == XFS_IOC_GETBMAPA ? BMV_IF_ATTRFORK : 0);
1573 	if (file->f_mode & FMODE_NOCMTIME)
1574 		bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1575 
1576 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, xfs_getbmap_format,
1577 			    (__force struct getbmap *)arg+1);
1578 	if (error)
1579 		return error;
1580 
1581 	/* copy back header - only size of getbmap */
1582 	if (copy_to_user(arg, &bmx, sizeof(struct getbmap)))
1583 		return -EFAULT;
1584 	return 0;
1585 }
1586 
1587 STATIC int
1588 xfs_getbmapx_format(void **ap, struct getbmapx *bmv)
1589 {
1590 	struct getbmapx __user	*base = (struct getbmapx __user *)*ap;
1591 
1592 	if (copy_to_user(base, bmv, sizeof(struct getbmapx)))
1593 		return -EFAULT;
1594 
1595 	*ap += sizeof(struct getbmapx);
1596 	return 0;
1597 }
1598 
1599 STATIC int
1600 xfs_ioc_getbmapx(
1601 	struct xfs_inode	*ip,
1602 	void			__user *arg)
1603 {
1604 	struct getbmapx		bmx;
1605 	int			error;
1606 
1607 	if (copy_from_user(&bmx, arg, sizeof(bmx)))
1608 		return -EFAULT;
1609 
1610 	if (bmx.bmv_count < 2)
1611 		return -EINVAL;
1612 
1613 	if (bmx.bmv_iflags & (~BMV_IF_VALID))
1614 		return -EINVAL;
1615 
1616 	error = xfs_getbmap(ip, &bmx, xfs_getbmapx_format,
1617 			    (__force struct getbmapx *)arg+1);
1618 	if (error)
1619 		return error;
1620 
1621 	/* copy back header */
1622 	if (copy_to_user(arg, &bmx, sizeof(struct getbmapx)))
1623 		return -EFAULT;
1624 
1625 	return 0;
1626 }
1627 
1628 struct getfsmap_info {
1629 	struct xfs_mount	*mp;
1630 	struct fsmap_head __user *data;
1631 	unsigned int		idx;
1632 	__u32			last_flags;
1633 };
1634 
1635 STATIC int
1636 xfs_getfsmap_format(struct xfs_fsmap *xfm, void *priv)
1637 {
1638 	struct getfsmap_info	*info = priv;
1639 	struct fsmap		fm;
1640 
1641 	trace_xfs_getfsmap_mapping(info->mp, xfm);
1642 
1643 	info->last_flags = xfm->fmr_flags;
1644 	xfs_fsmap_from_internal(&fm, xfm);
1645 	if (copy_to_user(&info->data->fmh_recs[info->idx++], &fm,
1646 			sizeof(struct fsmap)))
1647 		return -EFAULT;
1648 
1649 	return 0;
1650 }
1651 
1652 STATIC int
1653 xfs_ioc_getfsmap(
1654 	struct xfs_inode	*ip,
1655 	struct fsmap_head	__user *arg)
1656 {
1657 	struct getfsmap_info	info = { NULL };
1658 	struct xfs_fsmap_head	xhead = {0};
1659 	struct fsmap_head	head;
1660 	bool			aborted = false;
1661 	int			error;
1662 
1663 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1664 		return -EFAULT;
1665 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1666 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1667 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1668 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1669 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1670 		return -EINVAL;
1671 
1672 	xhead.fmh_iflags = head.fmh_iflags;
1673 	xhead.fmh_count = head.fmh_count;
1674 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1675 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1676 
1677 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1678 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1679 
1680 	info.mp = ip->i_mount;
1681 	info.data = arg;
1682 	error = xfs_getfsmap(ip->i_mount, &xhead, xfs_getfsmap_format, &info);
1683 	if (error == XFS_BTREE_QUERY_RANGE_ABORT) {
1684 		error = 0;
1685 		aborted = true;
1686 	} else if (error)
1687 		return error;
1688 
1689 	/* If we didn't abort, set the "last" flag in the last fmx */
1690 	if (!aborted && info.idx) {
1691 		info.last_flags |= FMR_OF_LAST;
1692 		if (copy_to_user(&info.data->fmh_recs[info.idx - 1].fmr_flags,
1693 				&info.last_flags, sizeof(info.last_flags)))
1694 			return -EFAULT;
1695 	}
1696 
1697 	/* copy back header */
1698 	head.fmh_entries = xhead.fmh_entries;
1699 	head.fmh_oflags = xhead.fmh_oflags;
1700 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
1701 		return -EFAULT;
1702 
1703 	return 0;
1704 }
1705 
1706 int
1707 xfs_ioc_swapext(
1708 	xfs_swapext_t	*sxp)
1709 {
1710 	xfs_inode_t     *ip, *tip;
1711 	struct fd	f, tmp;
1712 	int		error = 0;
1713 
1714 	/* Pull information for the target fd */
1715 	f = fdget((int)sxp->sx_fdtarget);
1716 	if (!f.file) {
1717 		error = -EINVAL;
1718 		goto out;
1719 	}
1720 
1721 	if (!(f.file->f_mode & FMODE_WRITE) ||
1722 	    !(f.file->f_mode & FMODE_READ) ||
1723 	    (f.file->f_flags & O_APPEND)) {
1724 		error = -EBADF;
1725 		goto out_put_file;
1726 	}
1727 
1728 	tmp = fdget((int)sxp->sx_fdtmp);
1729 	if (!tmp.file) {
1730 		error = -EINVAL;
1731 		goto out_put_file;
1732 	}
1733 
1734 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1735 	    !(tmp.file->f_mode & FMODE_READ) ||
1736 	    (tmp.file->f_flags & O_APPEND)) {
1737 		error = -EBADF;
1738 		goto out_put_tmp_file;
1739 	}
1740 
1741 	if (IS_SWAPFILE(file_inode(f.file)) ||
1742 	    IS_SWAPFILE(file_inode(tmp.file))) {
1743 		error = -EINVAL;
1744 		goto out_put_tmp_file;
1745 	}
1746 
1747 	/*
1748 	 * We need to ensure that the fds passed in point to XFS inodes
1749 	 * before we cast and access them as XFS structures as we have no
1750 	 * control over what the user passes us here.
1751 	 */
1752 	if (f.file->f_op != &xfs_file_operations ||
1753 	    tmp.file->f_op != &xfs_file_operations) {
1754 		error = -EINVAL;
1755 		goto out_put_tmp_file;
1756 	}
1757 
1758 	ip = XFS_I(file_inode(f.file));
1759 	tip = XFS_I(file_inode(tmp.file));
1760 
1761 	if (ip->i_mount != tip->i_mount) {
1762 		error = -EINVAL;
1763 		goto out_put_tmp_file;
1764 	}
1765 
1766 	if (ip->i_ino == tip->i_ino) {
1767 		error = -EINVAL;
1768 		goto out_put_tmp_file;
1769 	}
1770 
1771 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1772 		error = -EIO;
1773 		goto out_put_tmp_file;
1774 	}
1775 
1776 	error = xfs_swap_extents(ip, tip, sxp);
1777 
1778  out_put_tmp_file:
1779 	fdput(tmp);
1780  out_put_file:
1781 	fdput(f);
1782  out:
1783 	return error;
1784 }
1785 
1786 /*
1787  * Note: some of the ioctl's return positive numbers as a
1788  * byte count indicating success, such as readlink_by_handle.
1789  * So we don't "sign flip" like most other routines.  This means
1790  * true errors need to be returned as a negative value.
1791  */
1792 long
1793 xfs_file_ioctl(
1794 	struct file		*filp,
1795 	unsigned int		cmd,
1796 	unsigned long		p)
1797 {
1798 	struct inode		*inode = file_inode(filp);
1799 	struct xfs_inode	*ip = XFS_I(inode);
1800 	struct xfs_mount	*mp = ip->i_mount;
1801 	void			__user *arg = (void __user *)p;
1802 	int			error;
1803 
1804 	trace_xfs_file_ioctl(ip);
1805 
1806 	switch (cmd) {
1807 	case FITRIM:
1808 		return xfs_ioc_trim(mp, arg);
1809 	case XFS_IOC_ALLOCSP:
1810 	case XFS_IOC_FREESP:
1811 	case XFS_IOC_RESVSP:
1812 	case XFS_IOC_UNRESVSP:
1813 	case XFS_IOC_ALLOCSP64:
1814 	case XFS_IOC_FREESP64:
1815 	case XFS_IOC_RESVSP64:
1816 	case XFS_IOC_UNRESVSP64:
1817 	case XFS_IOC_ZERO_RANGE: {
1818 		xfs_flock64_t		bf;
1819 
1820 		if (copy_from_user(&bf, arg, sizeof(bf)))
1821 			return -EFAULT;
1822 		return xfs_ioc_space(filp, cmd, &bf);
1823 	}
1824 	case XFS_IOC_DIOINFO: {
1825 		struct dioattr	da;
1826 		xfs_buftarg_t	*target =
1827 			XFS_IS_REALTIME_INODE(ip) ?
1828 			mp->m_rtdev_targp : mp->m_ddev_targp;
1829 
1830 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1831 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1832 
1833 		if (copy_to_user(arg, &da, sizeof(da)))
1834 			return -EFAULT;
1835 		return 0;
1836 	}
1837 
1838 	case XFS_IOC_FSBULKSTAT_SINGLE:
1839 	case XFS_IOC_FSBULKSTAT:
1840 	case XFS_IOC_FSINUMBERS:
1841 		return xfs_ioc_bulkstat(mp, cmd, arg);
1842 
1843 	case XFS_IOC_FSGEOMETRY_V1:
1844 		return xfs_ioc_fsgeometry_v1(mp, arg);
1845 
1846 	case XFS_IOC_FSGEOMETRY:
1847 		return xfs_ioc_fsgeometry(mp, arg);
1848 
1849 	case XFS_IOC_GETVERSION:
1850 		return put_user(inode->i_generation, (int __user *)arg);
1851 
1852 	case XFS_IOC_FSGETXATTR:
1853 		return xfs_ioc_fsgetxattr(ip, 0, arg);
1854 	case XFS_IOC_FSGETXATTRA:
1855 		return xfs_ioc_fsgetxattr(ip, 1, arg);
1856 	case XFS_IOC_FSSETXATTR:
1857 		return xfs_ioc_fssetxattr(ip, filp, arg);
1858 	case XFS_IOC_GETXFLAGS:
1859 		return xfs_ioc_getxflags(ip, arg);
1860 	case XFS_IOC_SETXFLAGS:
1861 		return xfs_ioc_setxflags(ip, filp, arg);
1862 
1863 	case XFS_IOC_FSSETDM: {
1864 		struct fsdmidata	dmi;
1865 
1866 		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1867 			return -EFAULT;
1868 
1869 		error = mnt_want_write_file(filp);
1870 		if (error)
1871 			return error;
1872 
1873 		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1874 				dmi.fsd_dmstate);
1875 		mnt_drop_write_file(filp);
1876 		return error;
1877 	}
1878 
1879 	case XFS_IOC_GETBMAP:
1880 	case XFS_IOC_GETBMAPA:
1881 		return xfs_ioc_getbmap(filp, cmd, arg);
1882 
1883 	case XFS_IOC_GETBMAPX:
1884 		return xfs_ioc_getbmapx(ip, arg);
1885 
1886 	case FS_IOC_GETFSMAP:
1887 		return xfs_ioc_getfsmap(ip, arg);
1888 
1889 	case XFS_IOC_FD_TO_HANDLE:
1890 	case XFS_IOC_PATH_TO_HANDLE:
1891 	case XFS_IOC_PATH_TO_FSHANDLE: {
1892 		xfs_fsop_handlereq_t	hreq;
1893 
1894 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1895 			return -EFAULT;
1896 		return xfs_find_handle(cmd, &hreq);
1897 	}
1898 	case XFS_IOC_OPEN_BY_HANDLE: {
1899 		xfs_fsop_handlereq_t	hreq;
1900 
1901 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1902 			return -EFAULT;
1903 		return xfs_open_by_handle(filp, &hreq);
1904 	}
1905 	case XFS_IOC_FSSETDM_BY_HANDLE:
1906 		return xfs_fssetdm_by_handle(filp, arg);
1907 
1908 	case XFS_IOC_READLINK_BY_HANDLE: {
1909 		xfs_fsop_handlereq_t	hreq;
1910 
1911 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1912 			return -EFAULT;
1913 		return xfs_readlink_by_handle(filp, &hreq);
1914 	}
1915 	case XFS_IOC_ATTRLIST_BY_HANDLE:
1916 		return xfs_attrlist_by_handle(filp, arg);
1917 
1918 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1919 		return xfs_attrmulti_by_handle(filp, arg);
1920 
1921 	case XFS_IOC_SWAPEXT: {
1922 		struct xfs_swapext	sxp;
1923 
1924 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1925 			return -EFAULT;
1926 		error = mnt_want_write_file(filp);
1927 		if (error)
1928 			return error;
1929 		error = xfs_ioc_swapext(&sxp);
1930 		mnt_drop_write_file(filp);
1931 		return error;
1932 	}
1933 
1934 	case XFS_IOC_FSCOUNTS: {
1935 		xfs_fsop_counts_t out;
1936 
1937 		error = xfs_fs_counts(mp, &out);
1938 		if (error)
1939 			return error;
1940 
1941 		if (copy_to_user(arg, &out, sizeof(out)))
1942 			return -EFAULT;
1943 		return 0;
1944 	}
1945 
1946 	case XFS_IOC_SET_RESBLKS: {
1947 		xfs_fsop_resblks_t inout;
1948 		uint64_t	   in;
1949 
1950 		if (!capable(CAP_SYS_ADMIN))
1951 			return -EPERM;
1952 
1953 		if (mp->m_flags & XFS_MOUNT_RDONLY)
1954 			return -EROFS;
1955 
1956 		if (copy_from_user(&inout, arg, sizeof(inout)))
1957 			return -EFAULT;
1958 
1959 		error = mnt_want_write_file(filp);
1960 		if (error)
1961 			return error;
1962 
1963 		/* input parameter is passed in resblks field of structure */
1964 		in = inout.resblks;
1965 		error = xfs_reserve_blocks(mp, &in, &inout);
1966 		mnt_drop_write_file(filp);
1967 		if (error)
1968 			return error;
1969 
1970 		if (copy_to_user(arg, &inout, sizeof(inout)))
1971 			return -EFAULT;
1972 		return 0;
1973 	}
1974 
1975 	case XFS_IOC_GET_RESBLKS: {
1976 		xfs_fsop_resblks_t out;
1977 
1978 		if (!capable(CAP_SYS_ADMIN))
1979 			return -EPERM;
1980 
1981 		error = xfs_reserve_blocks(mp, NULL, &out);
1982 		if (error)
1983 			return error;
1984 
1985 		if (copy_to_user(arg, &out, sizeof(out)))
1986 			return -EFAULT;
1987 
1988 		return 0;
1989 	}
1990 
1991 	case XFS_IOC_FSGROWFSDATA: {
1992 		xfs_growfs_data_t in;
1993 
1994 		if (copy_from_user(&in, arg, sizeof(in)))
1995 			return -EFAULT;
1996 
1997 		error = mnt_want_write_file(filp);
1998 		if (error)
1999 			return error;
2000 		error = xfs_growfs_data(mp, &in);
2001 		mnt_drop_write_file(filp);
2002 		return error;
2003 	}
2004 
2005 	case XFS_IOC_FSGROWFSLOG: {
2006 		xfs_growfs_log_t in;
2007 
2008 		if (copy_from_user(&in, arg, sizeof(in)))
2009 			return -EFAULT;
2010 
2011 		error = mnt_want_write_file(filp);
2012 		if (error)
2013 			return error;
2014 		error = xfs_growfs_log(mp, &in);
2015 		mnt_drop_write_file(filp);
2016 		return error;
2017 	}
2018 
2019 	case XFS_IOC_FSGROWFSRT: {
2020 		xfs_growfs_rt_t in;
2021 
2022 		if (copy_from_user(&in, arg, sizeof(in)))
2023 			return -EFAULT;
2024 
2025 		error = mnt_want_write_file(filp);
2026 		if (error)
2027 			return error;
2028 		error = xfs_growfs_rt(mp, &in);
2029 		mnt_drop_write_file(filp);
2030 		return error;
2031 	}
2032 
2033 	case XFS_IOC_GOINGDOWN: {
2034 		uint32_t in;
2035 
2036 		if (!capable(CAP_SYS_ADMIN))
2037 			return -EPERM;
2038 
2039 		if (get_user(in, (uint32_t __user *)arg))
2040 			return -EFAULT;
2041 
2042 		return xfs_fs_goingdown(mp, in);
2043 	}
2044 
2045 	case XFS_IOC_ERROR_INJECTION: {
2046 		xfs_error_injection_t in;
2047 
2048 		if (!capable(CAP_SYS_ADMIN))
2049 			return -EPERM;
2050 
2051 		if (copy_from_user(&in, arg, sizeof(in)))
2052 			return -EFAULT;
2053 
2054 		return xfs_errortag_add(mp, in.errtag);
2055 	}
2056 
2057 	case XFS_IOC_ERROR_CLEARALL:
2058 		if (!capable(CAP_SYS_ADMIN))
2059 			return -EPERM;
2060 
2061 		return xfs_errortag_clearall(mp);
2062 
2063 	case XFS_IOC_FREE_EOFBLOCKS: {
2064 		struct xfs_fs_eofblocks eofb;
2065 		struct xfs_eofblocks keofb;
2066 
2067 		if (!capable(CAP_SYS_ADMIN))
2068 			return -EPERM;
2069 
2070 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2071 			return -EROFS;
2072 
2073 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2074 			return -EFAULT;
2075 
2076 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2077 		if (error)
2078 			return error;
2079 
2080 		return xfs_icache_free_eofblocks(mp, &keofb);
2081 	}
2082 
2083 	default:
2084 		return -ENOTTY;
2085 	}
2086 }
2087